Вы находитесь на странице: 1из 596

Process Control and Instrumentation

Prof. A. K. Jana
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture - 1
Introduction to Process Control
The name of this course is Process Control and Instrumentation. Initially, we will cover
the process control and then the instrumentation course.
(Refer Slide Time: 01:02)

So, the first topic is introduction to process control, this topic we will cover first. Now, in
chemical engineering, we have a number of chemical units; for example reactors, for
example distillation column, like pump, compressor, etcetera, these are the different units
which are extensively used in chemical engineering. Now, to constitute a chemical plant,
we need to assemble few of these units if we assemble few of these units then we can
constitute a chemical plant.
Now, what is the objective of a chemical plant suppose, we have a plant now, it receives
raw material so input to this plant is, raw material and output is product. So, basically,
the plant receives raw material using different available sources of energy, the plant
produces products in the most economical way, this is a objective of a plant. It receives
raw material, it uses available sources of energy then it produces product in the most
economical way, this is the objective.

Now, to meet this objectives, we need to meet some requirements porter these
requirements, the primary requirement is safety, say for example, a reactor which is
design to operate within 100 psig pressure. We have one reactor, which is designed to
operate within 100 psig pressure, now to maintain this pressure, we need some external
intervention so that, the reactor operates below this limit, this is the first requirement that
is, safety.
Second one is the production specifications, a process must produce desired amounts of
product and desired quality of product. So, first one is quality and second one is quantity,
this the product specifications, we need to maintain the quantity as well as the quality,
that is a product specification. Third important point is environmental requirements so
there are a number of state and federal laws, which enforce to maintain say for example,
the concentration of chemicals.
(Refer Slide Time: 05:43)

Say for example, it is require to maintain the concentration of chemicals in the effluent
stream now, another example is sulphur dioxide concentration in the stream, which is
rejected to the atmosphere. Second environmental requirement is, to maintain the sulphur
dioxide concentration, this is example which is rejected to the atmosphere. And third
example is the waste water which is retained to the river or lake, these are three examples
under the environmental regulations.

Next important requirement is operational constraints now, the plant have certain
constraints inherent to the operation say for example, the distillation column, this is the
first example. The distillation column should not be flooded, this is one operational
constraints second example is, tanks should not overflow or go dry. Third example is, the
temperature in a catalytic reactor should not exceed the upper limit, if we consider a
catalytic reactor.
So, the temperature should not exceed the upper limit because, if the temperature is
higher than the upper limit, the catalyst may be destroyed so this is all about the
operational constraints. Fifth one is the economics so it is require to control the operating
conditions at given optimum level of minimum operating cost, maximum profit, etcetera.
Say for example, we need to judiciously use the raw materials, the energy used in the
process the human level so that, we can minimize the operating cost and maximize the
profit so these are all about the requirements to meet the plant of ((Refer Time: 09:08)).
Now, it is very obvious from this discussion, that to maintain all these requirements,
there is some external intervention is required and that external intervention is nothing
but, the control system so to maintain, to meet all these requirements, we need to devise
a control system.
(Refer Slide Time: 09:42)

Next, we will discuss three important issues which can dealt by the control system, what
are these issues. First issue is, the influence of the external disturbances, this is the first

issue second one is, the stability of a chemical process, this is the second issue and third
one is, the performance of a chemical process. So, these are the three important issues
which should be dealt by the control system.
Now, the basic aim of a controller is, to separates the influence of external disturbances,
second one is to ensure the stability of a chemical process and third one is, to optimize
the performance of a chemical process. So, we need to device a control system, which
separates the influence of external disturbances, it ensure the stability of a chemical
process and it can optimize the performance of a chemical process. We will discuss all
these three issues in brief with the help of control system.
So, first we will discuss the external disturbance so to discuss this external disturbance,
we will consider a simple example that is, a tank heating system. The schematic of the
tank heating system is something like this, this is a tank, a liquid steam which is entering
this tank with a flow rate of F suffix i, temperature of this inlet steam is T suffix i. Now,
this liquid is while stared, the outlet steam flow rate is F and temperature is T now, the
liquid in this tank has the height of h definitely, the temperature is also T because, we
were considering starrer.
Now, to heat up this process, we need to introduce one coil, through which steam is
going with a flow rate of F suffix s t. So, this is a heating tank system, a liquid is entering
the process with flow rate F i and temperature T i, the outlet flow rate has the rate of F
and temperature T, one coil is introduce for heating the liquid and through the coil, steam
is passing with a flow rate of F s t.
Now, what is the objective of the this process, that we need to mention first, first
objective is, to maintain the liquid temperature T at it is desired value, that desired value
is suppose, T suffix d this is the first objective. The temperature of the liquid should be
maintained at it is desired temperature that is, T d. Similarly, the second objective of this
process is, we need to maintain the height of liquid in the tank at it is desired value that
is, h suffix d so these are the two objectives.
Now initially, what are the steps we need to follow for any chemical process, first is, we
have to follow the startup procedure of a process. After starting of the process, it reaches
at steady state, we are presenting that by ss so after starting up the process, it reaches at
steady state. Now, suppose, the process is at steady state and if there is no external

intervention I mean, if there is no change of F I, there is no change of T i, there is no


change of F s t, the process remains at steady state all the times.
So, if there is no change of any input variables then the process remains at steady state so
there is no need of any control system but, this is not the case in practice. Usually, the
input variables may change with time, not regularly but, maybe in some interval say, for
example this heat, inlet flow rate which is coming from one upstream unit. So, we do not
have any control on F i and T i exactly so this is the practical case where, the input
variables may change with time and that is why, we need to devise a control system.
Now question is, how we can devise that control system and how, we can keep this
objectives say, for example, we are considering the first objective I mean, we need to
maintain the temperature at it is desired value, how we can maintain that. First of all, we
have to measure this temperature of this liquid steam, we can measure the temperature
using one thermocouple say, this is a thermocouple employing, which we can measure
the temperature of this liquid.
And suppose, the thermocouple outlet I mean, the temperature which is measured by the
thermocouple, that is exactly the liquid temperature T. Now, this temperature is then
compared with the desired value that is T d, this input T d is given by the person who is
in charge of operation or you can say, value of this T d is specified by the control
engineer. Now, we compare this T d and T, and the outlet of this comparator is e, that is
nothing but, the error and that error is T d minus T.
So, we can do one thing, we can put here positive sign and here negative sign, this error
signal goes to a controller, this error signal is the input to the controller then this
controller produce or calculates control actions and that control action is implemented
through this control valve. So, this is a heating tank system, we are considering only the
first objective that is, we need to maintain the temperature at it is desired value. We are
not considering the second case so initially, we need to measure the liquid temperature
using one temperature sensor that is, thermocouple.
We have assume, that the thermocouple outlet temperature is exactly equal to the liquid
tank temperature then this temperature is compared with the desired temperature that is,
Td. So, we have used one variable that is, error which is represented by e, e equals to T d
minus T then this error signal goes to the controller, controller calculates and that

calculated action is implemented through this valve. Now suppose, error is greater than 0
this is the first case, error is greater than 0 it means, T d is greater than T that means, the
desired temperature is higher than the temperature exist in the liquid system. So, what is
require to do for the controller?
Student: ((Refer Time: 20:41))
Yes, the controller will increase the F st so it is require to increase the steam flow rate
that means, more steam is required to flow through the control valve, this is one case.
Similarly, if error is less than 0 in that case, the situation is just opposite I mean, T d is
less than T that means, the control should reduce the stem flow rate. Now, we have to
introduce this fact in terms of the disturbance, here the disturbance is basically, the feed
flow rate and temperature.
Suppose, feed flow rate has been increased, this F i feed flow rate increases now, there is
no change of the inlet steam flow rate, only change is in F i. If F i increases and F st
remains same then what happens, temperature increases or decreases, temperature
decreases. So, in this situation, what the controller will do, the controller will increase
the steam flow rate that means, the controller will increase this F s t. So, this is
description in terms of external disturbance so it is very clear from this discussion that,
the controller which separates the effect of external disturbance.
(Refer Slide Time: 22:52)

Next, we will discuss the stability of a chemical process so we will first draw a plot, we
will make a plot, this is x and this is time t. This x may be temperature, x may be
concentration now, s i mention that, initially the process is at steady state. Now, at time t
equals to t naught, the x is disturbed, this is the steady state value of x. Now, first x is
disturbed at t equals t naught now as a result, the x show this type of response that
means, x returns automatically to the steady state.
If this is the case, this type of process is called stable or self regulating process, x returns
automatically to the steady state value and this type of process is called stable process or
self regulating process. So, for this case, there is no need of any external intervention I
mean, no need of any controller for this stable system. We will consider another case
which is unstable so this is x versus time that is t, initially the process is at steady state
and x is disturbed at time t equals to t naught.
Now, in this case, the response is like this so this figure indicates that, x does not return
to the steady state, this figure clearly indicates that x does not return to the steady state
and this type of processes are called unstable process. And for this process, there is a
need of external intervention I mean, there is a need of controller. Now, from this
discussion, we cannot control that for this stable process, there is no need of controller,
we cannot conclude that.
Because, we have mention three important issues so this is only one issue, for other two
issues like the separation of external disturbance and third one is the optimum
performance of the process, for those two issues, we may need the controller. Not only
that, even for this stable process to reach at steady state with a short period of time, we
may need a controller. So, we can say that, here also the need of controller we realize,
next issue is the performance of a chemical process.

(Refer Slide Time: 28:29)

Or optimize the performance of a process this is a third issue now, the main operational
objectives, main objectives are, first one is the safety and second objective is production
specifications, this is the second objective. Once these are achieved, once these two
objectives are achieved, the next goal is to make the operation more profitable. Now, in
that direction, we will consider one example that is, continuous stared tank reactor.
Suppose, this is a jacketed reactor, this is a schematic of a reactor reactants enter the
process, this is the reactant and this is the product. Now, in this process, two consecutive
reactions A to B and then B to C occur, this is two consecutive exothermic reaction
occur. A is the reactant, B is the desired product and C is the undesired product so this is
basically products, A is the reactant, B is the desired product and C is the undesired
product.
Now, what is the economic objective, the economic objective for this process is to
maximize the profit. Now, the profit function is phi and this is integration of 0 to t, t is
nothing but, the operational time and here, one function will be there, this function is in
terms of revenue from the sales of the product B then it includes the cost of reactant A
and then the cost of coolant. Basically, if this is the exothermic reaction, we need to
introduce here coolant to take out the exothermic heat and this is coolant in, this is
coolant out.

So, the profit function which includes the revenue from the sales of product B, cost of
reactant A and the cost of coolant then we need to maximize the profit. And then we can
maximize the performance we can say so these are all about the three issues, which can
be dealt by the control system. Next, we will discuss the classification of variables, the
variables which are extensively used in process controlled course, that we will discuss in
the next.
(Refer Slide Time: 33:45)

So, next topic is classification of variables, the variables is usually 2 types, one is input
variable and second one is output variable. Now, input variable is again 2 types, first one
is disturbance or load variable and it is conventionally represented by LV, second input
variable is manipulated or adjustable variable or some time it is called control variable.
So, input variable is 2 types, one is disturbance or load variable and second one is
manipulated variable or control variable, this we can represent by MV.
Similarly, the output variable again 2 types, one is measured output and second one is
unmeasured output, another variable is also used that is, controlled variable. So, do not
confuse with this controlled variable, controlled variable is manipulated variable and
controlled variable we will discuss that. But at this point, I can say, that the controlled
variables are usually the measured output. Sometimes this is also unmeasured output so
this the controlled variable so these are all about the variables. Now, we will take

different examples and we will select the different variables for that specific example so
we will start with a simple liquid tank system.
(Refer Slide Time: 37:20)

We will first start with a liquid tank so this is a liquid tank system, the input to this
process is the input steam, has a flow rate of F i, outlet flow rate is suppose, F naught.
The liquid in this tank has the height of h and the cross sectional area of tank is A now,
this is a simple liquid tank system. So, what is the objective of this process? The
objective of this process is, to maintain the liquid height in the tank at it is desired value
h d ,this is the objective of this tank system now here, the control variable is height.
The controlled variable we will represent by CV, the controlled variable CV here is
liquid height. Now, if we considered this process so can you classify the variables so
which one is the input variable for this process? F i. Input variable is F i, what are the
output variables, one is F and another one is liquid height, the output variables are one is
outlet flow rate F and second one is liquid height. Now, for this example system, can you
make a pair in between controlled variable and manipulated variable, we have decided
that, for this particular system, liquid height in the tank h is the controlled variable.
So, what will be the corresponding manipulated variable I mean, to maintain this liquid
height, which variable we can adjust F i or F o? ((Refer Time: 40:20))

I think both we can do, F i and F naught. I am just making one controlled configuration
suppose, this F naught is the manipulated variable corresponding to liquid height h so
what will be the control configuration, as we have drawn for the heating tank system. So,
first of all, we need to measure this height by level sensor so here a level sensor we can
place, which can measure the liquid height.
Suppose, this is height then it is compared with the desired value h d, this carries positive
sign and this is negative sign. Now, the output of this comparator is error, error is
basically desired height minus height then this error signal goes to controller and this
control action is implemented here, this is a control valve. So, F naught is the
manipulated variable in this control configuration now, can you tell me, F naught is input
variable or output variable? ((Refer Time: 42:13))
You see the classification, manipulated variable is under input variable or output
variable? Input variable. So, if we consider F naught as the manipulate variable then F
naught is input variable. In this example, it is clear that, if F naught is manipulated
variable then that is input variable, not output variable. Because, manipulated variable is
one type of input variable that is why, you write within bracket F naught but, if we
consider F i is the manipulated variable in that case, F naught is output variable. Next we
will take another example, to know about all these variables.
(Refer Slide Time: 43:15)

That example we have consider heating tank system so this is if heating tank system, the
inlet steam has flow rate F i, temperature T i and it has the outlet flow rate of F and T,
steam is introduced through this coil, it has the flow rate of F st and liquid height here h,
temperature T so this is the heating tank system. Now, what are the inputs in this case, F
i, F st, T i these are the inputs and what about F, may be input and what about the outputs
F, T and h.
We have mentioned the objectives of this process, first objective is to maintain T at it is
desired value and second is we need to maintain height at it is desired temperature. Can
you classify the, can you make the control pairs I mean, manipulated variable and
corresponding controlled variable pairs. So, one controlled variable is temperature
because, that is our objective, second control variable is height, how you can maintain
this temperature, by adjusting F st and for the case of height, how we can maintain the
height, by adjusting F i and F naught, F i and F.
If F is the manipulated variable, as we discuss for the previous example then F will be
one input variable, if F i is the manipulated variable in that case, F will be output
variable. So, F i and F st, these two are suppose for this case, manipulated variable and
this T and height, these two are controlled variable. So, which one is disturbance variable
or load variable, inputs variables are 2 types, one is manipulated variable and another
one is loaded variable, F i and F st they are manipulated variable so rest is T i so this is
load variable. So, these are all about to the variables, different variables next, we will
discuss the controlled configurations. Although we did not discuss the control
configurations in details but, before that, we want to know the different configurations.

(Refer Slide Time: 47:35)

Control configurations, we will consider the distillation example, which is quite complex
compare to this liquid tank and heating tank system. Now, first we will consider one
control scheme that is, the feedback control scheme. So, distillation column, this is the
tower of a distillation column, feed is introduced here with flow rate F and composition
Z, the overhead vapor which leaves the top tram is condensed in this condenser, this is a
condenser. Then, the condensed liquid is accumulated in a drum, which is the reflux
drum, this is also called reflux accumulator.
A part of this liquid is recycled back to the top section of the column and a part is
withdrawn as distillate with flow rate D and composition x D, composition means here
mold fraction. Similarly at the bottom, the liquid is withdrawn and it is subjected to a
reboiler, the produced vapor is recycle back to the bottom tray and some amount of
liquid is taken out as bottoms with flow rate B and composition x B. So, feed is
introduced with flow rate F, composition Z, this is a feed tray, this is the top tray, this is
the bottom tray.
Now, the overhead vapor goes to a condenser, condensation occurs then the condensed
liquid is accumulated in this reflux drum and part of this liquid is recycle to the top tray
as reflux, this is called reflux rate, reflux flow, reflux steam. And another part, a part of
this accumulated liquid is withdrawn as distillate with flow rate D and composition or
you can say mold fraction x D.

Similarly at the bottom, this liquid which is coming from the bottom tray, it goes to a
reboiler, the produced vapor is recycle back just below the bottom tray, this is bottom
tray. And some amount of liquid is withdrawn as bottom product or bottoms, with flow
rate B and composition x B. First we have to know, what is the control objective of this
process, there are basically 2 product, one is distillate another one is bottoms, we will
consider presently the top product.
So, what is the objective, objective is to maintain the top product composition at it is
desired value, this is the objective if you consider the top section only, we need to
maintain the top product composition x D at it is desired value. So, what will be the
control configuration for this case, if this is our control objective that means, this is the
controlled variable so controlled variable manipulated variable pair we have to make. If
x D is the controlled variable then corresponding manipulated variable is reflux flow rate
R.
Can you make the control configuration now, yes we can make it suppose here, the liquid
has composition of x D so we need one composition analyzer to measure x D. Then, this
analyzer gives the value of x D then that x D is compared in this comparator, this is
negative and this is x D, this is capital D desired this is positive then we get the error
signal. This error signal goes to the composition controller then the control action is
implemented through this valve so this is the control configuration.
So, this is basically the feedback control scheme, in the feedback control scheme, the
controlled variable is measured. Anyway, today we do not have time so in the next class,
we will discuss other two control schemes, they are feed forward control scheme and
inference sale control scheme, along with other topics.

Process Control and Instrumentation


Prof. A. K. Jana
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture - 2
Introduction to Process Control (contd.)
Today, we will continue our discussion with the example of distillation column.
(Refer Slide Time: 00:59)

Again I am drawing the schematic of the distillation column, this is the tower now, feed
is introduced in the feed tray. Feed flow rate is F, composition is z then, the operate
vapor lifts the top tray and then, that is condensed in the operate condenser. The
condense liquid is then accumulated in this reflux drum, this is the reflux drum, this is
condenser then, a part of this liquid is taken out as distillate, this is the top product with
flow rate D and composition x D.
Now, another part of this accumulated liquid is recycle back to the top tray, this is the top
tray so, this is the connection for the reflux rate and the bottom liquid which left the
bottom tray, this is bottom tray is reboiled in the reboiler, this is the reboiler. Then, the
produced vapor is recycled back just below the bottom tray and a part of the liquid is
withdrawn as bottoms, this is the bottom product which has the flow rate of B and
composition of x B.

We have discussed with this example and one control configuration we have drawn in
the last class, that control configuration was the feedback control configuration. That was
the feedback control configuration, in that feedback control configuration, we have
majored x D basically for this particular process, the controlled variable is x D and
corresponding manipulated variable is reflux rate, this is reflux rate. We are considering
only the single loop at the top, we are not presently the considering the bottom loop.
Now, in the feedback control configuration we have seen that, the controlled variable
which is x D was directly measured to manipulate the reflux rate. Today, we will
consider another control scheme that is, feed forward control scheme. First, we will
configure this feed forward control scheme for this distillation column, in this feed
forward control scheme, the disturbance is measured. Here, disturbance is the feed
composition as well as feed flow rate but, we are considering this is the majored variable.
Now, for measuring this feed composition, we need one composition analyzer, this is
composition analyzer, which is measuring basically the feed composition z. We are
assuming that, this composition analyzer output is exactly identical with the feed
composition. Then, this information z goes to the feed forward controller, this is the
block for feed forward control scheme then, these feed forward controller calculates the
control action based on this feed composition and this controller action is implemented
through this control valve.
For the feed forward control scheme, disturbance variable or load variable is measured
using one composition analyzer then, that measured feed composition information goes
to the feed forward controller. Then, the controller calculates the control action and that
control action is implemented through this control valve. So, the major difference
between the feedback control and feed forward control is, for the feedback control
scheme, we measured directly the controlled variable and for feed forward control
scheme, we measured the load variable or disturbance variable. There are many other
differences, which we will discuss in the subsequent classes, another control scheme we
will discuss next with this distillation example that is, inferential control scheme.

(Refer Slide Time: 07:07)

So, we will considers the same distillation column to discuss inferential control scheme,
this is a schematic of the same distillation column. Feed is introduced here then, the
operate vapor is condensed and condense liquid is accumulated in the reflux drum, a part
of that liquid is the recycled back to the top tray as reflux stream and a part of the liquid
is withdrawn as distillate. At the bottom section, we incorporate one reboiler, the
vapourised stream is introduced just below the bottom tray and at the bottom, the bottom
product is withdrawn.
Now, in this inferential control scheme, the x D which is controlled variable is not
directly measured, most composition analysis provide large delays in the response.
Secondly, it provides high investment and maintenance cost that is why, sometimes this
composition I mean, the product composition is not directly measured. In that case, we
search for the secondary measurements and then, using those secondary measurements,
we can infer that product composition, that is the purpose here.
So, to estimate x D ,we need to select the secondary variables for this particular example
suppose, this is first tray, this is second tray, this is third tray. So, the temperature of this
first tray is suppose T 1, this is T 2, this is T 3 it is quite easy to measure the temperature
compare to the composition measurement. So, we can say that, these three tray
temperatures we can consider as secondary measurements so, we will measure these
three temperatures using thermocouples.

Then here, we put one block for estimator, these three temperature after measurement,
the temperature information goes to this estimator, this estimator basically nothing but,
one algorithm which consists of some equations and correlations. These correlations
basically, calculate the x D based on the information of the major temperature. So, here,
some mathematical equations are included within this estimator and using those
equations, we can directly calculate based on this measured temperature information x D
value.
Basically, this composition is inferred based on the measure temperatures then as usual,
this estimated composition x D goes to the controller and the controller calculates the
action and those control actions are implemented through this control valve. Since the
product composition, which is the controlled variable is inferred that is why, the
controller is called inferential controller. So, we can write that, the unmeasured output
basically is the function of secondary measurements.
Here, the unmeasured output is x D and the secondary measurements are three
temperatures T 1, T 2 and T 3. So, I have mentioned earlier that, why we prefer this
control scheme, inferential control scheme because, if we directly one to measure this
product composition in that case, we need composition analyser. And most composition
analysis provide large delays in the response, this is the first drawback, most product
composition analysers provide large delays in the response.
Second drawback is high investment and maintenance cost, this is the second drawback
for the use of composition analyser. So, this is all about the different controlled
configurations with the example of distillation column. In the next, we will discuss the
different hardware elements of a control system.

(Refer Slide Time: 13:43)

We will discuss different hardware of a control system, to discuss this topic it is better to
take one example. We will continue the heating tank system, to discuss this topic it is
better to consider one example and we will continue the heating tank system, which will
discussed in the first class. This is the heating tank system, feed is introduced with a flow
rate F i, temperature T i and product is withdrawn with a flow rate of F and temperature
T.
Now, to heat the liquid in this tank, one steam stream is introduced with a flow rate
suppose, F suffix st and for this particular system, we have this control pair I mean,
controlled variable and manipulated variable pair, we want to maintain the temperature
through the manipulation of steam flow rate. Our control objective is to maintain the
temperature of this tank and that is why, this is our controlled variable, corresponding
manipulated variable is steam flow rate.
So, this is basically a close loop system I mean, this is the open loop system if we want
to make it close, we need to include the controlled scheme, this is open loop now, we
want to close it by introducing controlled scheme. So here, temperature is a controlled
variable so, we need to first measure the temperature by using one thermocouple. This is
a thermocouple then, this measured temperature T goes to the comparator, this is the
comparator.

The measured temperature is compared with the desired temperature, this is positive, this
is negative, output of this comparator is the error signal. Then, this error signal goes to
the controller then, the controller calculates the control action and that control action is
implemented through this control valve. So, this is basically the close loop system now,
we will discuss what hardwares are involved in this close loop system so, first one is the
process.
In this process, physical and chemical operations occur, this heating tank system is a
process, this is the first hardware involved in this close loop system. Second hardware is,
measuring instruments or sensors, this measuring instruments or sensors are used to
measure, these are used to measure first one is load variable. These sensors are used to
measure controlled variable and secondary output, third one is secondary output.
Although we have shown here, only the measurement of controlled variable but, we can
measure also the disturbance variable and secondary output by the use of the measuring
device or sensor. So, we will just discuss in brief what are the different measuring
devices used for different variables. So, we can write them in a table form, variable and
sensor used for measuring that variable like first variable is temperature say.
Temperature we can measure by the use of thermocouple, we can use thermocouple, we
can use resistance thermometer.
We can use thermocouple for measuring temperature, we can use resistance thermometer
but, mercury thermometer is not good because, the measurement cannot be transmitted
readily. Mercury thermometer is not good option because, the measurement signal
cannot be readily transmitted. Next we will consider another variable that is pressure,
pressure we can measure by the use of manometer, by the use of diaphragm element. For
measuring the pressure, we can use manometer, we can use diaphragm element.

(Refer Slide Time: 20:44)

Next, we can consider another variable that is a flow rate, flow rate we can measure by
the use of orifice meter, flow rate we can measure by the use of venturi so, these are two
measuring devices, which are extensively use for measuring the flow rate. Like, another
variable is liquid level, liquid level is measured by the use of differential pressure cell,
DP cell. Liquid level in the tank is measured by the use of differential pressure cell,
another one is composition, composition is measured by using chromatographic
analyzer.
Although we have started all these things in the instrumentation part in details but, just to
know, which devices can be use for different variables, we have just discussed in brief
so, our second hardware was the measuring instruments or sensors. Now, third hardware
is transducer now, measurement signal, measurements cannot be used for control until
they are converted to physical quantities. Physical quantities say for example voltage,
say for example current, say for example pneumatic signal.
So, measurements cannot be used for control until they are converted to physical
quantities, which are readily transmitted, these signals we can readily transmit. And this
is the purpose of the use of transducer basically, the transducer physically convert the
measurement signal to the physical quantities. Next one is transmission lines, this is use
to carry the measurement signal from sensor to the controller and from the controller to

the control valve. The transmission lines are used to carry the measurement signal from
sensor to controller and then, from controller to control valve.
If you see the schematic representation of the close loop system, I think you can clearly
understand this so, this is the usage of the transmission line. Now, measurements signal
is sometimes very weak, sometimes the measurement signal is very weak say for
example, in few milli volts. In that case, this transmission line is equipped with
amplifier, when the measurement signal is very weak in that case, the transmission line is
equipped with amplifier, this is the fourth hardware which is involved in the close loop
system.
(Refer Slide Time: 25:45)

Now, fifth hardware is the controller, this is another hardware which receives
measurement signal from the sensor. The controller basically receives measurement
signal from sensor and then, it decides, what action should be taken based on the major
values. So, the controller basically receives the measurement signal from the measuring
device in the next step, the controller calculates the control action based on the measured
signal comparing with the desired value of that control variable.
Sixth hardware is the final control element now, for the case of controller I told, that the
controller basically calculates the control actions. Now, that control action is
implemented through the final control element so, the control action, which is calculated
and produced by the controller that is, physically implemented through this final control

element. Can you give any example of the final control element, yes first example is say,
control valve, another example is variable speed pump, variable speed compressor.
Why variable speed, if the speed is fixed then, that cannot be changed that is why,
variable speed pump and variable speed compressor. Now, next hardware element is a
recording device, this is used to visualize the plant behavior through the measurement
signals. If we used to visualize the plant behavior at different situations, we need one
recording device and that realization we can achieve through the measured values. And
for this purpose, we can use video display unit, which is usually accommodated in the
control room.
Video display unit can be used for this recording purpose, which is usually
accommodated in the control room. So, these are about the hardware elements for a
particular process so far we have discussed, the first topic that is, introduction to process
control. In the next, we will discuss the second topic that is, the mathematical modeling
and the use of mathematical modeling in process control.
(Refer Slide Time: 30:05)

So, we will start to discuss the second topic that is, mathematical modeling first we will
know in brief, what is the mathematical model. Suppose, we have one experimental
setup so, we will represent the experimental setup, we will use one block to represent the
experimental set up. Now, to this experimental setup, the input is introduced I mean,

input is introduced to the setup and if we run the setup, we get output this is a common
thing in our laboratory which happened.
So, what is basically the model, can we represent this experimental setup by some
mathematical equations or correlations we can? We can represent this mathematical
setup by some mathematical correlations or equation. So, suppose, some mathematics are
involved then, we are giving the same input to this block, we will get some output. Now,
this block is the representation of the model, we are just representing this experimental
setup by mathematical correlation.
Now, if these two outputs are close enough then, we can say that, this model is a good
model so, model is the mathematical representation of a process, intended to promote
understanding of the real system. A real system, we can represent by using some
mathematical equations and correlations so, this is the definition of model. And next, we
will discuss about, what is the use of this model, why we will develop the model, what is
the use of mathematical modeling in process control, that we will cover in the next.
(Refer Slide Time: 33:47)

So, use of the process model, first we can write to understand the process behavior
suppose, we have the model and we have the solution of this model, also we are giving
some input to this model, we can get the output only if, we have the solution otherwise,
we cannot get the output. We have develop the model structure for a particular process,

we are giving some input information to the model basically, the input information is
specified and then, if we solve the model then, we can get some output.
Now, suppose, the model initially is at steady state I mean, after start of we can reach at
steady state. Now, at steady state, we are giving some change in this disturbance variable
then, the process will shift from steady state to another state. Basically, if we change in
load variable, we will get some transient response and that transient response we can get
by performing this simulation of this model without performing experiment, is not it. So,
we can understand the process behavior by some change in load variable, to understand
the transient behavior without performing the experimental setup.
So, this is the first purpose of the model secondly, to train the operating personnel, the
model we can use for training purpose, without running the experimental setup or
without running the plant. Suppose, we have the model structure same thing which we
have drawn earlier, we have some input, we have output. Now, some situations can be
irritated using this model like, the feed is introduced to a particular process here, we are
directly using some value for the feed flow rate.
Now, suppose, the pump is not delivering the feed to the process what will happen,
suppose in the distillation column, we have some minimum reflux rate. If we consider
lower than that minimum reflux rate, what will happen so, these types of emergency
situations, we can irritate by the use of this model, without disturbing the process. Or
even some situations cannot be permitted to irritate in the real process so, to train the
operating personnel, we can also use the simulated model.

(Refer Slide Time: 38:05)

The next purpose is selection of control pairs, we have taken few examples and we have
discussed the controlled variable and manipulated variable pair. But, this controlled
variable and manipulated variable pairs, we can select by knowing the model of that
particular process. So, if we know the model of a particular process, we can determine
the pair between controlled variable and manipulated variable so, for this purpose also,
we need the process model.
Fourth purpose is, to develop the model based controller, you know most of the
advanced controllers are model based controller and the name clearly suggests that, the
controller which consists of or which includes the process model, those controllers are
basically model based controller. So, most of these advanced controllers, use the process
model so, in that sense we can say that, we need the process model for the development
of advanced controllers.
Fifth one is, optimize the process operating conditions, to determine the most profitable
operating condition, we need the process model and economic information. So, we need
process model and some economic information to determine the most profitable
operating condition. So, we need the process model, we need some economic
information to determine the most profitable operating condition. So, these are the issues,
for which we need the process model next, we will discuss the classification of process
model.

(Refer Slide Time: 41:14)

There are different ways to classify the process model but here, we will discuss the
classification of process model based on, how they are obtained. There are three different
models I mean, we can classify in three different ways, three different models are there,
first one is theoretical model. The theoretical model is basically developed based on
principles of conservation, this is developed based on the principle of conservation.
So, first type of model that is, theoretical model and this model is develop principle of
conservation, I think you know the principle of conservation like, mass conservation,
energy conservation, momentum conservation. And another type of model that is
empirical model, this model is obtained by fitting experimental data, the second type of
model that is empirical model, this empirical model we can obtain by fitting
experimental data.
Basically, if we have the experimental setup, we have different sets of input output data
now, if we have the input output data by fitting those experimental data, we can
determine the coefficients. By that way, we can construct a model and that is the
empirical model. Third one is just the combination of these two, theoretical and
empirical model, that is called semi empirical model, sometimes this is also called hybrid
model.

Semi empirical model is the combination of theoretical model and empirical model, semi
empirical model is the combination of these two. Next, we will discuss in brief, what are
the advantages and disadvantages of this models.
(Refer Slide Time: 44:45)

So, first is the theoretical model, which is developed based on the conservation principle
so, what is the advantage of this model. First advantage is, it provides physical insight
into process behavior, this is the first advantage. Second advantage is, it is applicable for
a wide range of conditions, this is the second advantage. Similarly, what are the
disadvantages, it is time consuming to develop so, it leads to be time consuming to
develop because, particularly for the complex systems, the theoretical model is too large.
So, in that sense we can say that, this is time consuming and another drawback is, some
model parameters are not readily available, some model parameters for example, reaction
rate coefficient, overall heat transfer coefficient, these are not readily available. In that
case, we have to follow the empirical technique so, these are basically, the two
drawbacks and two advantages of the theoretical model. Now, what about the empirical
model, I think we can say something for this empirical model, based on the discussion of
this theoretical model.
So, what is the advantage, it is easier to develop, this is the advantage because, if you
have the input output data set then, we can feed some correlations or equations using
those input output data set. So, for complex process, we do not have so many equations

like theoretical model so, this is only the advantage for this case. And what are the
disadvantages, this is applicable for a narrow range of conditions, this is applicable for a
limited range of conditions.
So, this is the advantage and this is the disadvantage for the empirical model similarly, I
think you can find, what are the advantages and disadvantages of the semi empirical
model. Now, for the process control, we have discussed earlier different variables I
mean, input variables, output variables then, input variables are again 2 types
manipulated variable and load variable, and output variables are 2 types majored output
and unmajored output. At this time, we will discuss another variable which is extensively
used in process control that is, state variable that we will discuss now.
(Refer Slide Time: 49:36)

State variable basically describes the natural state of a process now, there are basically 3
fundamental quantities. What are these quantities mass, energy and momentum, these are
three fundamental quantities. Now, these fundamental quantities are usually not directly
measured, these three fundamental quantities are not directly and conveniently measured.
And these three fundamental quantities are usually characterized by say, temperature, by
pressure, by composition, by flow rates, etcetera.
There are 3 fundamental quantities mass, energy and momentum, these three
fundamental quantities are not directly and conveniently measured and they are
characterized by these variables temperature, pressure, composition, flow rate and these

variables are called state variable because, they describe the natural state of the system.
So, these are called state variable and it arises naturally in the accumulation term, we will
discuss in the next, the conservation principle and there we will see that, the state
variables usually present within the accumulation term so, it arises in the accumulation
term.
Within the accumulation term the state variable is present and since we have discussed
the state variable, in the next we will discuss state equations. These equations basically
derived by the application of conservation principle on the fundamental quantities to
relate the state variables with other variables are called state equations. The state
equations are derived by the application of conservation principle on the fundamental
quantities to related the state variables with other variables including other state variables
are called state equations.
These are about the state variable and state equations and how, we can see the state
variable within the state equation. I mean, first we will go for the development of state
equation then, we will see which one is the state variable within that state equation. Now,
for that purpose, we have to know the conservation principle.
(Refer Slide Time: 54:48)

The conservation principle, we can represent in general form like rate of accumulation
equals to rate of input minus rate of output plus rate of generation, this is the general
form of the conservation principle. Now, I told about the three fundamental quantities

mass, energy and momentum so, if we include here rate of mass accumulation equals to
rate of mass input minus rate of mass output, that is the conservation of mass.
For the case of conservation of mass, there is no existence of this rate of generation term
and for the energy conservation similarly, we can write rate of energy accumulation
equals to rate of energy input minus rate of energy output plus rate of energy generation.
So, this is about the conservation principle and the state equations, state variable, the
description on that I mean, by the derivation of the state equation, we will know which
one is the state variable, this thing we discuss in the next class.
Thank you.

Process Control and Instrumentation


Prof. A. K. Jana
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture - 3
Mathematical Modeling (Contd.)
(Refer Slide Time: 00:57)

We will continue the topic Mathematical Modeling. In the last class, we have discussed
about the state variables and state impressions and in this class, we will develop the
mathematical model, first for a CSTR, Continuously Stirred Tank Reactor. So, you will
take CSTR example and will derive the mathematical model, this is if jacketed CSTR,
this is a tank, this is a jacket. Now, one medium is introduced here and that medium is
coming out here, first of all the feed which is entering to the reactor has the flow rate of
F i, concentration of the feed is C A i and temperature is T i.
The suffix i indicates the input, A is a component, I mean C A i is the input
concentration of component i and T i is a input temperature. The product steam which is
coming out from this CSTR has the flow rate of F, concentration in terms of component
A is C A and temperature is T. Now, before deriving the mathematical model for this
CSTR, first we will know the units of different flows like F i and F along with the
coolant medium flow rate kept F c.

So, all these flow rates are basically volumetric flow rate, here temperature is T c i, outlet
flow rate is F c and outlet coolant temperature is T c naught. Concentration C A and inlet
concentration C A i, both are molal concentration I mean, unit is say mole per unit
volume. C A and C A i both are molal concentration for example, they are in mole per
unit volume. Next, you will consider the assumptions, so first assumption is perfect
mixing that means, the temperature of this outlet steam T and compositions C A, they are
same with that of the reaction mixture.
The perfect mixing indicates, everywhere in this tank temperature and concentration,
they are identical and the outlet temperature composition also identical with the
temperature and concentration of reacting mixture. Second assumption is liquid density
rho and the heat capacity C p they are constant, third assumption is we are considering a
simple exothermic first order reaction. And to remove this exothermic heat, the coolant
steam is introduced in this jacket, this is the coolant steam.
Fourth assumption is, the reactor is perfectly insulated that means, there is no heat loss
from the reactor to the surroundings it means, no heat loss from the system to the
surroundings. Fifth one is, coolant is perfectly mixed in the jacket and last assumption is,
we will not consider any energy balance for the jacket, there is no energy balance for the
jacket. These are the assumptions adopted for this CSTR system and based on these
assumptions, we will derive the modeling equations, now first we will go for the overall
mass balance.

(Refer Slide Time: 08:48)

First, we will go for overall mass balance so, to develop the overall mass balance of the
CSTR system, we need the conservation of mass. What is that? Conservation of mass is
rate of mass accumulation equals to rate of mass input minus rate of mass output, this is
the conservation of mass. Now, for this CSTR system, what is this term I mean, how we
can represent the mass of accumulation see in this reactor system, you consider the
volume of the liquid is v.
The volume of the liquid in the reactor is v now, if v is the volume, if we multiply with
density so, this whole terms becomes mass. Now, differentiation of this is the mass flow
rate d d t v rho, that is the rate of mass accumulation. Now, rate of mass input, input to
the system is basically F i but, we know that, F i is the volume at the flow rate so, we
have to multiply with rho. So, F i multiplied by rho that is the input mass rate similarly,
what will be the output, output flow rate I mean, the volumetric flow rate that is, F.
So, if we multiply with rho then, this is the output rate now, we have assume that, rho
and C p both are constant. If rho is constant then, we can write this rho d v d t equals to F
i rho minus F rho that means, d v d t equals to F i minus F. Suppose, this is equation
number 1 so, this is a mass balance equation similarly, we can go for mass balance of
component A or you can say, that component mass balance. So, we will consider in the
next, mass balance or component mass balance and the component is here component A.

So, what is the conservation principle for this rate of accumulation of component A, this
is a accumulation term. Then, rate of input of component A then, rate of generation of
component A minus rate of output of component A, this is a conservation principle for
component mass balance. Now, we have to write all these individual term accumulation,
input, output and generation now, for this CSTR system, what is the accumulation term.
(Refer Slide Time: 14:08)

Volume multiplied by C A now, if we one to represent in terms of rate then, d d t of v C


A this is mole per unit volume C A and v is volume so, mole per unit time overall. So,
what is the input, F i is a flow rate of input stream and concentration is C A i so, F i
multiplied by C A i, that is the rate of input of component A. Now, what is a generation,
minus r A into v see, minus r A is the rate of disappearance of A. If we multiply with
minus sign then, we get the generation, minus r A is the rate of disappearance of A so, if
we multiply with minus then, that becomes generation.
And what is the output, output is F multiplied by concentration of output steam that is, C
A so, from the conservation principle, we got this equation. Now, we can write this
equation in this form C A d v d t plus v d C A d t equals to F i C A i minus F C A minus,
minus of r A into v. Now, we know the d v d t term I mean, if we substitute equation 1
then, this equations becomes C A F i minus F plus v d C A d t equals to F i C A i minus
F C A minus, minus of r A into v.

Now, this F multiplied by C A and this F C A will be canceled out then, we can
rearrange this equation to v d C A d t equals to F i C A i minus C A minus, minus of r A
into v. If we divide again both sides by v, we will finally get d C A d t equals to F i
divided by v, C A i minus C A minus, minus of r A, minus of r A is the rate of
disappearance of A. Now, if we consider Arrhenius equation then accordingly, we can
write minus of r A equals to k naught exponential of minus of E divided by R T into C A.
According to the Arrhenius principle, the reactions rate equals to pre exponential factor
then, exponential of minus E divided by R T, C A is the activation energy and R is the
universal gas constant. If we substitute this reaction rate expression here then finally, we
get d C A d t equals to F i divided by v, C A i minus C A minus k naught C A
exponential of minus of E divided by R T. So, this is the final form of component mass
balance equation so, these two equations we have derived based on the conservation of
mass in the next, we will consider the energy balance equation.
(Refer Slide Time: 19:38)

Energy balance equation for the example CSTR system, what is the conservation
principle of energy, conservation principle we can write in this form, rate of energy
accumulation equals to rate of energy input minus rate of energy output minus rate of
energy removed by the coolant plus rate of energy added by exothermic reaction. So,
here 4 terms are involved accumulation, input, output, energy removable and energy
added by the exothermic reaction.

So, next we have to representation all these terms using term mathematical terms I mean,
variables. So, what is the energy accumulation say, volume in the tank is represented by
v and if we multiply with rho, this becomes m then C p, then d t, energy term we can
write in terms of m C p d t. Now, here we are considering reference temperature equals 0
now, if we write here d d t of v rho C p T then, this becomes rate of accumulation of
energy.
Now, next we will consider the rate of energy input, volumetric flow rate of input steam
that is, F i if we multiply with rho then, that becomes mass flow rate. Similarly, C p then,
d t I mean, T i minus T reference here, reference temperature we are assuming 0. So, T i
and you recall, we have consider the C p that is constant, what will be the output rate of
energy output. The flow rate of outlet steam I mean, the volumetric flow rate that is, F
similarly, if we multiply rho, this becomes mass C p and outlet temperature is T.
Now, rate of energy removed by the coolant this one, this will represent by Q, energy
removed by the coolant is represented by here Q. What is Q, how we can calculate Q, we
know flow rate F c, rho c, C p c and then, temperature difference is outlet temperature of
coolant minus inlet temperature of coolant. What is the last term I mean, how we can
represent the last term, last term we have to consider in this way, minus del H that is,
heat of reaction then, minus r A multiplied by v.
Minus del H here is heat of reaction, it is well known to ask that, heat of reaction is
negative I mean, this negative term is used for the case of exothermic reaction and for
endothermic reaction, we use here positive sign. So, this term represents the energy
added by exothermic reaction next, we need to simplify this equation. Now, dividing
both sides by rho C p, v d capital T d small t plus temperature d v d t equals to F i T I,
since we have we are dividing both sides by rho C p.
Next term is F multiplied by T then, Q divided by rho C p and finally, minus del H minus
r A v divided by rho C p. Now, you will substitute this term d v d t, we have the equation
of d v d t obtained from the total mass balance, if we substitute d v d t then, we can write
it like this way, T multiplied by F i minus F equals to F i T i minus F t minus Q divided
by rho C p plus minus of del H minus of r A into v divided by rho C p now, this F T and
this F T we can cancel.

(Refer Slide Time: 27:37)

And then, we get v d T d t equals to F i T i minus T minus Q divided by rho C p plus


minus of del H minus of r A into v divided by rho C p. And finally, we will divide both
sides by this volume term then, we get d T d t equals to F i divided by v T i minus T
minus Q divided by v rho C p plus minus of del H minus of r A divided by rho C p.
Now, again we will substitute here the Arrhenius law then, d T d t equals to F i divided
by v T i minus T minus Q divided by v rho C p plus minus of del H k naught C A
exponential of minus E divided by R T whole divided by rho C p so, this is a energy
balance equation.
So, for the example CSTR system we got three equations, one is based on total mass
balance then, component mass balance and last one is based on energy balance.

(Refer Slide Time: 29:57)

Now, the modeling equations I am writing here, one we got d v d t that is equals to F i
minus F, second equation d C A d t that we got F i divided by v, C A i minus C A minus
k naught C A exponential of minus E divided by R T. And energy balance equation we
got that is, d T d t equals to F i minus v into T i minus T minus Q divided by v rho C p
plus minus of del H C A k naught exponential of minus E divided by R T whole divided
by rho C p.
Here, I have mentioned that Q, Q equals to our coolant flow rate is F c now, density is
suppose rho c, heat capacity if we will consider C p c multiplied by the temperature
difference. What is the outlet temperature of this, T c o minus T c i so, this is the
expression for Q. Now, you will just classify, we will just see what are the different
variables involved in the modeling equations here, what are the input variables. Input
variables are C A i then, F i then, T i then, Q.
We are not considering F c, T c i we are considering Q and F definitely, F will be input
variable. If this is considered as the manipulated variable for controlling the liquid height
or liquid volume so, these are the input variables. What are the output variables, output
variables are here v, C A and T see, in this three modeling equations v, C A and T, they
are present within the accumulation term so, these three variables are also state variables.
So, we can write here these are also state variables because, they are present within the
accumulation term.

Now, among this input variables, which are the manipulated variables, for the example
CSTR system, if we consider liquid volume is a first I mean, one control variable and
another one is say, temperature. Now, the corresponding manipulated variables are, if we
consider F as the manipulated variable for v and Q as the manipulated variable for
temperature then so, Q and F, these two are basically manipulated variables.
So, these three I mean, the rest input variables are load variables or disturbance variables,
among these five input variables, two are the manipulated variables and other three are
the load variables. So, this is the development of model structure for the sample CSTR
and we have seen the different variables, which are involved in this example CSTR.
Before going to discuss another system, we will study about the degrees of freedom
analysis.
(Refer Slide Time: 35:20)

So, we will next study the degrees of freedom so, so far we have discussed about the
modeling of chemical processes, after deriving the mathematical model of a process, we
need to solve those modeling equations. The solution of a model structure is basically
called simulation, we need to stimulate the modeling equations. Now, for the stimulation
of a modeling equation, we need to describe this degrees of freedom. Degrees of
freedom, which suppose is represented by F then, we can write F equals to v minus E,
degrees of freedom we are representing here by F then, F equals to v minus E.

V is the total number of independent process variables and E is the total number of
independent equations. So, degrees of freedom basically, total number of independent
variables minus total number of equations. Now, in the analysis of degrees of freedom,
we will consider three different cases, in the first case we will consider say, F equals to
0. It means, number of independent variables equals to number of independent equations
that means, the system is exactly specified.
When degrees of freedom equals 0 then, we can write v equals to E that means, number
of independent process variables equals the number of independent equations. In this
situation, we can say the system is exactly specified I mean, there is no problem to find
the solution of the modeling equations. In the second case, we will consider F is greater
than 0 that means, v is greater than E so, in this case, the system is called under
specified. How we can make it exactly specified system by the inclusion of F number of
additional equations.
So, to make it exactly specified, we need F additional equations, to make the under
specified system exactly specified, we need F number of additional equations then, we
can only get the solutions of the modeling equations. Last case is, F is less than 0 that
means, total number of independent process variables is less than total number of
independent equations and in this case, the system is called over specified. So, to make
this over specified system exactly specified, we need to remove F number of equations.
Usually in practice, the case 2 is the common I mean, F greater than 0 is the common
case in practice, F greater than 0 that means, v greater than 0. Now, thing is that, if this is
the situation F greater than 0, how we can make it exactly specified basically, there are 2
options, first option is we can specify more number of disturbance variables. So, by
specifying more number of disturbance variable, if we can specify more number of
disturbance variables then, number of unknown variables is reduced.
So, this is one option and in the second option, by incorporating more number of
controller equations. This is a second option, either we have to reduce the number of
unknown variables or we have to increase the number of equations to make the F equals
0. Anyway, we will discuss this degrees of freedom with taking one simple example, we
will consider the stirred tank heater.

(Refer Slide Time: 42:15)

To describe this degrees of freedom analysis, stirred tank heater example for degrees of
freedom analysis. So, before going to analyze the degrees of freedom, we need the model
so, first we will develop the model for the system and then, we will go for the degrees of
freedom analysis. The schematic of this system, we have to draw we have to develop
first so, this is F i and temperature is T i. Now, steam is introduced here through the coil
for heating purpose, this is steam suppose, flow rate is Q, outlet flow rate is F and
temperature is T.
Now, liquid in the tank has the height of h, temperature here also T, cross sectional area
of this tank is A. Now, before going to develop the model, we need to consider some
assumptions so, what are these assumptions, first assumption is, the tank is perfectly
mixed. Second assumption is rho and C p both are constant, third assumption is the tank
is perfectly insulated that means, there is no heat loss from the tank to the surroundings.
So, first we will develop the total mass balance equation so, what is the accumulation of
mass d d t liquid height multiplied by cross sectional area that is, volume.
Volume multiplied by density that is mass, h is the liquid height multiplied by cross
sectional area, this is volume. Now, volume multiplied by density that is mass so, this is
mass flow rate I mean, this is the rate of accumulation. What is the inflow rate, F i rho
minus F rho, this is a outflow rate so, we can write this equation A d h d t equals to F i
minus F. This is the total mass balance equations since rho is constant so, we can get this

from this equation, you give some equation number for this suppose, this is equation
number 1.
In the next, you will go for energy balance, what is the accumulation term, height
multiplied by area that is volume, volume multiplied by rho, mass. Mass, C p,
temperature difference is T minus suppose, T reference so, d d t of this is the
accumulation term, h multiplied by rho that is volume, multiplied by rho, h multiplied by
A that is volume, multiplied by rho that is mass. So, if this is the mass C p and this is
temperature difference, what is the energy input rate, F i rho this is mass flow rate, C p T
i minus T reference this is the energy input rate.
What will be the energy output rate, F rho C p T minus T reference and another term is
energy supplied by steam per unit time. Energy supplied by steam per unit time that is, Q
basically, the unit of Q is here, energy per unit time say for example, ((Refer Time:
47:47)) thermal unit per minute, the unit of this is energy per unit time. So, this is the
energy added or energy supplied by steam per unit time.
(Refer Slide Time: 48:13)

Now, if we simplified this considering T reference equals to 0 and if we simplify the


energy balance equation, we will get A h d T d t equals to F i T i minus T plus Q divided
by rho C p, this is a energy balance equation. If we consider T reference equals to 0 and
if we simplify finally, we will get this energy balance equation. So, there are basically 2

equations, one is based on total mass balance and another one is based on energy
balance.
So, these two equations are, first one is A d h d t equals to F i minus F and this is A h d T
d t equals to F i T i minus T plus Q divided by rho C p, you give some equation number
to this suppose, this is equation number 2 so, this is the model structure. Now, you will
go for the degrees of freedom analysis, how many variables are involved in this equation
h, F i, F, T, T i and Q so, these are the variables. We can write v equals to 6 agree or not,
agree so, there are 6 unknown variables then, how many equations are involved there,
one is equation 1 and another one is equation 2.
So, we can write E equals to 2 so, what is F, 6 minus 2 that is equals to 4 so, degrees of
freedom for the example system is 4. We have discussed, there are two ways to reduce
the degrees of freedom, first option is we can specifies some load variables, what are the
lowered variables in this system, one is F i, another load variables is T i. So, if we can
specify these two load variables then, the degrees of freedom reduces to 4 minus 2 that
is, 2 initially it was 4.
Now, two lowered variables we are specifying, how we can specify, by the direct
measurement. We can measure this flow rate, we can measure this temperature then, we
can get the information of flow rate and temperature that means, F i and T i are known
that means, degrees of freedom we can write 4 minus 2 that is, 2. Another option, I told
by including some controlled equations, for the example liquid heating tank system, what
are the controlled variable and manipulated variable pairs to be considered, one is height,
another one is temperature.
So, we can manipulate this height, we can control this height by the manipulation of
suppose F, we can control this temperature by the manipulation of Q. So, we can develop
two control equations, although we did not study the control equations, I am just
mentioning the simplest control equations for these two control pairs. If F is the
manipulated variable, the control equation we can write like this, F equals to F s plus k c
F multiplied by h d minus h.
F s is the steady state value of F, k c F is one tuning parameter, which the value of that
tuning parameter we need to determine, that is constant k c F basically, h d is the desired
value and h is the liquid height. So, this is one additional equation similarly, if we

consider another control scheme for temperature, in which Q is the manipulated variable,
we can add another equation Q equals to Q s plus k c Q T d minus T. Here, Q s is the
steady state value of Q, k c Q is the control at tuning parameter, that is a constant term
then, this is desired temperature.
So, additionally, we are getting two equations, we had F equals to 2 now, if we can add
two equations then, the degrees of freedom becomes 0. So, we had basically degrees of
freedom 4 additionally, we have specified two load variables through direct
measurement. Then, we have just paired control variable manipulated variable then, we
got to additional equations. And finally, the degrees of freedom becomes 0 that means,
the system is exactly specified.
Thank you.

Process Control and Instrumentation


Prof. A. K. Jana
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture - 4
Mathematical Modeling (Contd.)
Today, we will continue our discussion on Mathematical Modeling and we will develop
today the mathematical modeling of a distillation column.
(Refer Slide Time: 01:04)

Today, we will develop the model of a distillation column previously, we have


configured three control scheme for a distillation column but, we did not discuss the
modeling. So, first the configuration of a distillation column, this is the tower column
section, this is the top tray, this is the feed tray and this is suppose, the bottom tray. Now,
feed is introduced to this feed tray, this is the feed stage or feed tray, feed has the flow
rate of F.
Suppose, this feed mixture contains only two components they are A and B, feed mixture
contains two components namely A and B. The composition of this feed mixture is z
now, the vapor which is living this top tray, this top tray is denoted by N. A vapor steam
which is living this top tray has the flow rate of V N, the vapor steam which is living this
top stage has the flow rate of V N. Now, this vapor steam is condensed in an over hit
condenser, this is a condenser.

After condensation, the produced liquid, condensed liquid is accumulated in a drum, this
is called reflux drum. The condensed liquid is accumulated in this reflux drum suppose, a
holdup of this condensed liquid in the reflux drum is m D. If part of this accumulated
liquid is withdrawn as top product, this is top product, this is also called distillate, top
product is also called distillate and definitely, this is a liquid steam. Now, we will assume
the flow rate of this distillate is D and composition is x suffix D.
The top product flow rate we are representing by capital D and composition is by x D, a
fraction of this accumulated liquid is recycled back to the top tray, this stream is called
reflux stream. This reflux stream has the flow rate of R and composition is same with the
distillate steam I mean, the composition is x D. Similarly, at the bottom section, the
liquid which left this bottom tray, this is we can say first tray, this is first stage, the liquid
which is living this first stage is accumulated in this column base.
Suppose, the holdup in the column base is m suffix B now, this liquid then goes to a
bottom reboiler and vaporization of liquid occurs in this reboiler. Then, the vaporized
stream is recycle back to the bottom stage suppose, this vapor flow rate is V suffix B, the
flow rate of this recycle vapor is suppose V B and the part of that accumulated liquid in
the column base is withdrawn as bottom product. This is bottom product, this is also a
liquid stream, it has the flow rate of suppose B and composition is x B, the flow rate of
this bottom product is suppose B and composition is x B.
So, we are basically introducing to this column is single feed stream and we are getting
two products, top product and bottom product. Now, this condenser is basically a total
condenser, the operate condenser is actually a total condenser because, the operate vapor
is totally condensed, the operate vapor which is entering the condenser is totally
condensed. We can call this condenser as a partial condenser, there is a operate vapor
distillate in forward.
When there is operate vapor distillate involve then, we can only say this is a partial
condenser but here, we are considering, it is a total condenser so, there is no operate
vapor withdrawn. So, this is all about the description of the distillation operation next,
will develop the model based on some assumptions so, what are these assumptions.

(Refer Slide Time: 10:05)

First assumption is, feed is a saturated liquid, the feed which is entering into the column
that is a saturated liquid that means, the feed is at it is boiling point temperature, this is
the first assumption. Second assumption is, the column is perfectly insulated that means,
there is no heat loss from the distillation column to the surroundings. It means, no heat
loss from the process to the surroundings, the column is insulated so that, there is no heat
loss from the column to the surroundings.
Third assumption is, all the trays are ideal that means, the trays are 100 percent efficient
or we can say, tray efficiency is 100 percent so, in this column, we are considering ideal
trays, 100 percent efficient trays. Next assumption is, there is no vapor holdup, vapor
holdup on each tray is neglected because, the density of vapor is much, much lower than
the density of liquid that is why, we are considering this. For the high presser column,
this assumption is taken into account I mean, the vapor holdup is considered on each
tray.
Fifth assumption is, the molar heats of vaporization of both components A and B are
approximately equal. It implies that, 1 mole of condensing vapor releases sufficient heat
to vaporize 1 mole of liquid that means, for the condensation of the vapor, some heat is
evolved. I mean, the vapor which is condensed, that releases some amount of heat and
for vaporization of liquid, heat is required. Now, the 1 mole of condensing vapor releases

the heat that is sufficient to vaporize exactly 1 mole of liquid, the meaning of this
assumption is that.
Sixth assumption is, the column has total 20 trays excluding total condenser and reboiler,
although this is not the assumption, this is basically the configuration of the process but
anyway, we have included under assumptions. So, we have considered N for to represent
the top stage basically, N is 20 here, seventh one is anyway, we can include another thing
with this, the feed stage is the tenth stage. So, feed stage is the tenth stage, the feed enters
the column on ten stage, seventh assumption is perfect mixing on each stage.
That means, if this is a tray suppose, this is the liquid stream so, if we consider perfect
mixing on each tray then, composition of the liquid everywhere same I mean, if we
represent the composition x N, that is identical everywhere on this stage. Next
assumption is, relative volatility of the two components A and B remains constant
throughout the column.
Next assumption is nineth assumption, liquid holdup varies from tray to tray, our fourth
assumption is negligible vapor holdup but, we are considering liquid holdup I mean,
liquid holdup varies from tray to tray but, there is no variation of vapor holdup. Last
assumption tenth that is, condenser and reboiler dynamics are neglected, we are not
considering the dynamics of condenser and reboiler. So, based on this ten assumptions,
we will develop the model for the binary distillation column now, you just see, you just
visit the assumptions 2, 4 and 5.

(Refer Slide Time: 19:07)

Revisit assumptions 2, 4 and 5, second assumption is, the column is perfectly insulated
that means, there is no heat loss. Now, this fourth assumption is, there is no vapor
holdup, vapor holdup is negligible on each tray and fifth assumption is, if 1 mole of
vapor condenses, at the same time 1 mole of liquid is evaporated. Can we write based on
this three assumptions, all the vapor flow rates are identical I mean, the vapor stream
which is living first tray equals to the vapor living second stage like this way, the vapor
living N stage equals to V B, can we write.
So, based on the assumptions 2, 4 and 5, we can write that all the vapor flow rates
throughout the column, they are identical it means, V 1 equal to V 2 equal to V N finally,
V B. Anyway next, we will go to develop the modeling equations dividing the
distillation column into different envelope. So, first we will develop the modeling
equation for the top section, top section means, reflux drum. So, first we will develop the
modeling equation for reflux drum, the schematic of a reflux drum is like this, this is the
reflux drum.
Since we have considered there is no dynamics of condenser so, we can include here
without developing any modeling equation for the condenser. The vapor stream which is
entering into the condenser that we have represented previously by V N, N is basically
20 because, we have consider total number of stages 20. Now, this is the accumulation of

liquid in the reflux drum suppose, the liquid holdup is m D, there are two outgoing
streams, one is distillate composition is x D, another one is R composition is x D.
See, all the compositions are here basically the mole fraction, you consider all the
compositions as mole fractions and flow rates are here basically molar flow rate say, for
the example, D has the unit of mole per unit time. Now, this is the schematic of the top
section and we will consider this as the first envelope, this is the first envelopes. Now,
we have to develop basically the two equations, one is total mole balance and another
one is component mole balance.
So, what will be the total, we will write here total mass balance originally, that is mole
balance and we will writing here, total mass balance. Now, differentiation of m D, this is
accumulation, what is the input to this envelope, V 20. V 20 is the input to this
envelopes, what are the outputs, one is D and second one is R so, d m D d t equals to V
20 minus D minus R, this is a total mass balance equation for this first envelope.
Next, we will develop the component mass balance equation, what will be the
component mass balance equation d m D multiplied by x D d t, equals to V 20 what is
the composition, this is vapor steam. So, vapor composition will represent by y and
liquid composition will represent by x so, V 20 multiplied by y 20, agree vapor flow rate
is V 20 and it is composition is y 20. Now, D multiplied by composition, what is the
composition of distillate x D similarly, R multiplied by x D, the composition of reflux
stream is x D.
So, this is the component mole balance equation or component mass balance equation
now, you use to simply this equation. So, m D d x D d t plus x D d m D d t equals to V
20 y 20 minus d x D minus R x D, we will substitute here d m D d t tomb. So, m D d x D
d t plus x D, what is d m D d t that is, V 20 minus D minus R equal to V 20 y 20 minus D
x D minus R x D. So, this D x D, this D x D will be canceled similarly this and this, R x
D and R x D so, what will be the final expression.

(Refer Slide Time: 26:53)

Final expression will be m D d x D d t equals to V 20 y 20 minus x D, if we further


simplify we will get, D x D d t equal to V 20 by m D y 20 minus x D. So, this is the final
form of component mass balance equation for first envelope, envelope 1 now, in the
next, we will consider the second envelope that is, the top stage.
(Refer Slide Time: 27:48)

So, next we will consider the top stage, that is twentieth stage, first we have to make the
schematic for this so, this is the top section of the column and this is twentieth stage so,
N equals to 20. Now, the vapor which is living this top stage that is, V 20 what is the

incoming to this stage, one is reflux R and another one is the vapor steam, which is
coming from nineteen stage that means, V 19. Outgoing steams one already we have
drawn, that is V 20, another one will be L 20.
So, incoming sterams are, one is V 19, another one is R, outgoing streams are L 20 and
V 20 so, this is the second envelope. Similarly, we have to develop the total mass
balance equation, we will consider for this twentieth stage, the liquid holdup as N 20. If
the liquid holdup is m 20 then, d d t m 20 which is basically the accumulation term,
equals the input flow rates. What are the input flow rates, one is R, another one is V 19
and outgoing streams are L 20 minus V 20, input minus output.
Now, based on the second, fourth and fifth assumptions we have concluded that, all the
vapor flow rates are identical. If that is the case so, V 19, V 20 they are identical so, V 19
and V 20 they are equal to each other. Now, this equation reduces to R minus L 20, got it
since based on the assumptions 2, 4 and 5 we have concluded that, all this vapor flow
rates are same, if that is the case, we have to cancel out this V 19 and V 20. Then finally,
we get the total mass balance equation d m 20 d t equals to R minus L 20.
What will be the component mass balance equation, component mass balance d m 20
corresponding composition is, x 20 d t equal to reflux flow rate and composition x D plus
V 19 y 19 minus L 20 x 20 minus V 20 y 20. Why you did not multiply the composition
with this equation final form I mean, why we have included again V 19 and V 20 terms
in the component mass balance equation. Because, x is varying from tray to tray and y
basically, is the equilibrium composition of x.
So, y varies to tray to tray that is why, we cannot neglect these two terms from this
component mass balance equation. Can we simplify this equation, as we did for the case
of first envelope so, m 20 d x 20 d t plus x 20 d m 20 d t equals to R x D plus V 19 y 19
minus L 20 x 20 minus V 20 y 20. If we substitute the total mass balance x 20 d t plus x
20 d m 20 d t, we will substitute that is, R minus L 20 equals to R x D plus V 19 y 19
minus L 20 x 20 minus V 20 y 20 so, L 20 x 20 is cancelled out.

(Refer Slide Time: 34:16)

So, finally, we will get d x 20 d t equal to 1 by m 20 R x D minus x 20 plus V B y 19


minus y 20, this is a component mass balance equation. Just we have substitute it, V 19
equals to V 20 equals to V B then, we get this equation. So, this is the second envelope,
the modeling equations of the second envelope next, we will consider the third envelope.
(Refer Slide Time: 35:20)

That is any tray, will represent by any tray by n th tray so, n th stage now, this is the
schematic of n th stage, holdup is m suffix n. Input streams are, one is V n minus 1 and
another one is L n plus 1, outgoing streams are V n and another one is L n, this is a

schematic of n stage. So, this is a third envelope, you quickly derive the modeling
equations, what will be the total mass balance for this. D m n d t equals to incoming
streams are L n plus 1 plus V n minus 1 and outgoing streams are L n minus V that
means, the final expression is like this, d m n d t equals to l n plus 1 minus l n.
What will be the component mass balance, d m n x n d t equal to L n plus 1 x n plus 1
plus V n minus 1 y n minus 1 minus L n x n minus V n y n. It is quite strait forward to
write the equation for component mass balance so, we will not proceed further to
simplify this equation, it is quite easy. So, in the next any way so, what will be the n th
stage basically, n th stage is basically just below the top stage that means, nineteenth to
above the feed stages.
That means, eleventh stage and below the feed stage that means, nineth stage to above
the bottom stage that means, second stage. Basically, we have considered common
nomenclature n for this stages that is why, we have represented that by n stage and those
are just nineteen to eleven and ninth to second stage. So, in the next, we will consider the
feed stage, which is not included within this n stage, also the first stage is not included
here.
Because, that has slightly different nomenclature similarly, for the case of column base,
we have to consider to separately. So now, you will consider the feed stage, we will
derive the total mass balance and component mass balance for the feed stage. So, first we
have to draw the feed stage, this is the feed stage that is basically ten stage. So, the
holdup will be m 10, one input stream that is feed stream, another one is L 11, the liquid
which is coming from just above the feed tray. Another input is the vapor steam which is
coming just below the feed tray that is, V 9 and outgoing steams are V 10 L 10.
So, this is the fourth envelope so, what will be the total mass balance equation, total mass
balance equation will be d m 10 d t equals L 11 plus F plus V 9, these are input steams L
11, F and V 9. And what are the outgoing steams, one is L 10, another one is V 10, these
two are the outgoing steam. So, it is straight forward to write the component mass
balance equation, just multiplying the flow rates with their composition.
In the accumulation term, we have to multiply the holdup with composition so, d m 10 x
10 d t equals L 11 x 11 plus F, feed composition is z plus V 9 y 9 minus L 10 x 10 minus

V 10 y 10, this is a modeling equations for the feed stage. See, within the n th stage as I
have mentioned, the bottom tray is not also included.
(Refer Slide Time: 42:56)

So, we will consider in the next the first stage, the schematic representation of the first
stage is like this, this is the first stage. Now, input liquid flow rate one is L 2, another
input steam is V B, which is coming from the reboiler. Outgoing streams are one is L 1,
another one is V 1 so, this is the fifth envelope. What will be the total mass balance, if
holdup is m 1 for the first stage then, d m 1 d t equal to L 2 plus V B minus L 1 minus V
1.
Similarly, component mass balance d m 1 x 1 d t equal to L 2 x 2 plus V B y B minus L
1 x 1 minus V 1 y 1, this is the component mass balance equation. The last envelope we
have to consider that is the, column base which is also not included within n stage. So,
last one is column base, the schematic of that is somewhat like this, this is the first stage
and this is a column base. Some amount of liquid is accumulated in the column base, we
are considering the holdup is m D and one reboiler is installed here.
This is the boiled up vapor which has the flow rate of V B, sometimes the bottom flow
rate is included in this B. The liquid steam which is living first stage, that has the flow
rate of L 1 and this V B is actually introduced to the first stage. So, what will be the total
mass balance d m B d t equal to L 1, any way this is the sixth envelope so, total mass

balance is d m B d t equals to input is L 1 and there are two output streams, one is B
another is V B.
L 1 is a input stream, B and V B both are the outgoing steams from this sixth envelopes,
what will be the component mass balance, d m B x B d t equal to L 1 x 1 minus V B y B
minus V x B, this is the component mass balance equation. So basically, there are total
six envelopes, one is for the reflux drum, second one is for top stage, third one is for n
stage, fourth one is for feed stage, fifth one is for bottom stage and sixth one is for
column base. Now, what are the variables involved in this modeling equations and how,
we can calculate those variables.
(Refer Slide Time: 48:16)

In the modeling equations, we got total mass balance, total mass balance equations are
basically the ordinary deferential equations, by solving those equations what we can
calculate, holdup. By solving the total mass balance equations which are basically the
ordinary deferential equations, we can calculate holdup. Another type of modeling
equations are based on component mass balance, this component mass balance equations
are also ordinary differential equations.
By solving the component mass equation, we can get liquid phase composition and this
is liquid holdup. So, m or total holdup liquid holdup and liquid phase composition, we
can calculate from the developed modeling equations. Based on three assumptions
second, fourth and fifth, we know all these vapor flow rates are identical, equal to V B.

So, this information is also known to us, all vapor flow rates we can calculate, what are
the rest, one is y, another one is liquid flow rate.
How we can calculate these two variables, that we will discuss next, how we can
calculate y. We assumed one thing that, the relative volatility of components A and B
remains identical throughout the column, relative volatility is represented by alpha. So,
relative volatility of any component i with respect to another component j is represented
by alpha i j. Relative volatility is represented by alpha and the suffix is used in this way
alpha i j means, relative volatility of component i with respect to component j. Alpha i j
is related with k by this way I mean, alpha i j equals to k i divided by k j.
K is the vapor liquid equilibrium coefficient, it is also called distribution coefficient
which is basically, represented by y by x I mean, k equals to y by x. So, can we write
here, y i divided by x i whole divided by, this is for j not i, k i by k j. So, y i by x i whole
divided by k j means, y j by x j, can we write this because, k equals to y by x so, we have
just included here the suffix. Now, see our example column is a binary distillation
column that means, there are two components, one is i and another one is j.
So, we can write x i plus x j equals to 1, this is for liquid phase similarly for vapor phase
we can write, y i plus y j equals to 1, agree. Because, our mixture is a binary mixture
now, we will write here, y i divided by x i whole divided by 1 minus y i divided by 1
minus x i. Our intension is to calculate, the vapor phase composition of component i that
is why, we are trying to avoid y j and x j terms. So, y j equals to 1 minus y i and x j
equals to 1 minus x I, if we rearrange this equation, we will get y i equals to alpha i j x i
divided by 1 plus alpha i j minus 1 into x i.
If we rearrange this, finally we will get, y i equals to alpha i j x i divided by 1 plus alpha
i j minus 1 into x i you see, in this equation, alpha i j is defined that is constant, we have
assume that. X value we can calculate from the component mass balance equation so, all
the terms in this equation are known except y i so, we can calculate y i. In the next class,
we will discuss, how we can calculate the liquid flow rate L.

Process Control and Instrumentation


Prof. A. K. Jana
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture - 5
Mathematical Modeling (Contd.)
(Refer Slide Time: 00:59)

Today, we will continue our discussion on modeling of distillation column. So, in the last
class, we have derived mass balance equations, I mean first, we have derived total mass
balance equation along with the component mass balance equation fine. Now, we have
mentioned that the total mass balance equation is use to calculate liquid holdup
represented by m, unit is suppose mole and component mass balance equation, we can
use to calculate liquid composition, which we have represented by x and x is the mole
fraction. And based on some assumptions, we have concluded that all vapor flow rater
are identical that means, V 1 equals to V 2 equals to V N equals to V B.
Now, other variables, I mean apart from these variables include the vapor phase
composition fine. Another one is liquid flow rate, we have discussed the calculation
procedure of vapor flow rate of any component i, that is y i equals to alpha i j x i divided
by 1 plus alpha i j minus 1 into x i. This equation, we have derived and this is called as
equilibrium relationship our assumption is that alpha i j remains constant throughout the

column. So, alpha i j is known, x we can get from the component mole balance equation
or component mass balance equation, so we can calculate the vapor composition.
(Refer Slide Time: 04:08)

Today, we will discuss the calculation of this liquid flow rate that means, we will
discuss liquid hydraulics. The liquid hydraulics is calculated by the use of well known
Francis weir equation, liquid flow rate is calculated by the use of well known, Francis
weir equation. So, for the example distillation column, we will consider the simplified
form of this Francis weir equation and that is the linearised form.
So, we will consider the linearised form of Francis weir equation. So, if L n is the flow
rate of a liquid, leaving n-th stage then the Francis weir equation correlates, L n equals to
L n naught plus m n minus m n naught divided by beta. This is the Francis weir form
Hoyer L n is the flow rate of a liquid steam, flow rate of a liquid steam, leaving n-th
stage, L n naught is the reference value of L n. Similarly m n is the reference value of
sorry, m n naught is the reference value of m n, another term is included in the Francis
weir formula that is beta, beta is hydraulic time constant.
The value of beta is typically, taken in between 3 to 6 minutes, the beta is taken in
between 3 to 6 naught minute seconds or each tray. So, beta is the hydraulic constant it is
typically, 3 to 6 seconds for each tray that means, you see the difference between m n
and m n naught, that is divided by the hydraulic time constant then that amount is

included with L n naught to calculate the L n. This is the simplified version of Francis
weir formula, other forms include the non-linear versions basically.
So, we will consider this amplified form in the example distillation column modeling,
now this is all about the distillation model. So, the model includes basically, the mass
balance, I mean total mass balance, component mass balance then the vapor flow rates,
which are identical for all trays equilibrium relations have to calculate the vapor phase
composition and last one is the tray hydraulics for the calculation of liquid flow rates
fine. Now, we will use this model to analyze the degrees of freedom.
(Refer Slide Time: 08:50)

So, the next topic is the degrees of freedom, for the example distillation degrees of
freedom. Now here, we will consider one example although, we have consider for the in
the development of the distillation model that is N, N is basically the number of trays
excluding condenser and reboiler, will use this nomenclature we will use this notation in
the degrees of freedom analyses. So, you can recall the degrees of freedom represented
by f, which is correlated with V and E by this form, f is the degrees of freedom, V is the
number of independent process variables and E is the number of independent equations.
So, to calculate the degrees of freedom, we need to find both and E for the example
distillation column. So, first we will go for the number of equations calculations, number
of equations and the origin of there those equations will first discuss this part. So, first
one is a equilibrium relationship what is the equilibrium relationship, we have consider

for the example column, that is y i equals to alpha i j x i divided by 1 plus alpha i j minus
1 into x i.
So, how many y, we need to calculate, how many y are involved in the distillation model
y 1 y 2 y n another y is involved, that is y b, the composition of boiled a vapor. So, what
is the number N plus 1. So, total number of equilibrium equations involved in the
distillation model is N plus 1.
Next is hydraulic relationship that is L n naught equals to sorry, L n equals to L n naught
plus m n minus m n naught divided by beta. How many liquid flow rates are involved in
the distillation model L 1 L 2 up to L n, n number of liquid flow rates, we have to
calculate using this Francis weir formula, we cannot consider the distillate flow rate and
bottom flow rate although those are liquid flow rates, because those are not calculated by
the employment of this Francis weir equation.
So, total number of equations N, next is component and total mass ablation equations,
total and component mass balance equation, how many component and total mass
balance equations are involved. First we will consider for N number of trays for N
number of trays, how many equations are there, 2 N one total mass balance equation, one
component mass balance equation for each tray that means, 2 equations for each tray. So,
for N number of equations, N number of stages, we have 2 N equations.
Next, we will consider for the reflux drum number of equations to similarly for the
column base, number of equations 2, one component and another one total mass balance
equations. So, total number of equations how many.

(Refer Slide Time: 14:35)

Total number of equations E that is equals to 4 N plus 5. So, next we have to find that
total number of variables V.
(Refer Slide Time: 15:04)

Next, we will calculate number of variables and type of variables. So, first will consider
the liquid composition, how many liquid compositions are involved x 1 x 2 x n, one for
top section and one for bottom section, N number of liquid compositions N trays, one for
reflux drum and one for column base. So, N plus 2 next vapor composition, how many

vapor compositions are involved y 1 y 2 y n and one for column base, there is no vapor
distillate, we have consider only the liquid distillate. So, 3 is no y D.
Accordingly, we have total N plus 1 equation, N plus 1 variable, those are vapor
compositions. Next one is liquid holdups, one for each tray that means, m 1 m 2 up to m
n, one for reflux drum and one for column base. So, N number of holdup for N trays and
2 for reflux drum and column base. So, total number is N plus 2.
Next one is liquid flow rates, how many liquid flow rates are involved L 1 L 2 L n, we
are not including the flow rates of top and bottom products will consider those
separately. So, total number of liquid flow rates N, now apart from this variables what
are the other variables involved in the distillation modeling. First we will consider in the
feed, what are the variables use for feed F and Z, what about top, I mean top section in
the top section, we have another 2 variables. One is D, another one is reflux flow rate
and what about the bottom section, one is bottom flow rate, another one is boil up.
So, this 2 are representing the feed, this 2 in the top section and this 2, we are
considering for the bottom section, so how many then 6. So, V equals equal to how much
4 N plus 11, we got V equal to 4 N plus 5 and we have just calculated V equals to V
equal to 4 N plus 11.
(Refer Slide Time: 19:40)

So, degrees of freedom f equal to 4 N plus 11 minus 4 n plus 5 that is 6 that means this is
under specified process, since V is greater than E. So, this is underspecified process, we
need to make it exactly specified. So, what we need to do, we have to get f equal to 0,
how we can get that there are 2 O S, we have discussed earlier, one is we need to specify
more number of disturbance variables to reduce if to 0, we can specify more number of
disturbance variables.
So, what are the disturbance variables involved in this distillation process, one is F,
another one is Z. F is feed flow rate and Z is feed composition, we can maser this feed
flow rate using a flow meter then, we have the feed flow rate information, we can maser
this feed composition by any chromatography canalize. Then we have the confirmation
information, if that is the case then F reduces to 4, previously F is 6, we are trying to
specify this 2 disturbance variables that means, minus 2.
So, f becomes 4 what is the another option to make the process exactly specified, we can
incorporate more number of control equations. So, second option is we can incorporate
more number of controller equations. So, before writing the controller equations, we
need to couple controlled variable, manipulated variable pears, what are the control
variable and manipulated variable pears to be consider, for the example column, that we
need to decide. So, control variable, manipulated variable see this is a distillation
example feed is introduced on a particular tray, we are getting 2 products, one is the top
product, another one is the bottom product.
So, what is the control object see for this process, control objective is to mention the
composition products purity that means, we can consider one control variable as top
product purity, which is represented by top product composition. Another controlled
variable, we can consider that is the bottom product composition x B. Now what will be
the manipulated variable for x D, we can consider the manipulated variable that is reflux
flow rate and for bottom purity, we can consider the manipulated variable as vapor boil
operate.
These are the standard control variable, manipulated variable pears along with this 2
pears, we can include another 2 control variable, manipulated variable pear. One for top
section and one for bottom section, you see the reflux drum holdup is represented by m
D, basically in the reflux drums some liquid is accumulated and that we represent by m

D. So, we can control that liquid height and that liquid height, we can control by the
adjustment of distillate flow rate some liquid is accumulated after condensation in the
reflux drum.
So, there is a scope to maintain the liquid holdup or we can say there is a need to
maintain the liquid holdup in the reflux drum. So, if that is our objective, we can
consider m D as the control variable and that can be adjusted by the outlet flow rate D by
the similar fashion, we can consider the holdup in the in the column base m B as a
another control variable.
Some liquid is accumulated in the column base, which is represented by m B. If that is
our objective I mean, if we one to maintain the liquid holdup in the column base then
that is the control variable and this m B can be manipulated sorry, this m B can be
controlled by the manipulation of outlet liquid flow rate that is B.
So, these 4 controls manipulated control variable pears, we can consider for the example
column. So, one manipulated variable is R and the controller equation although did not
discuss that particular discuss later, I am just adding, I am just writing the equations
controller equations. If R is the manipulated variable then one controller equation is R S
plus K C R x D S P minus x D. Previously, we have used suffix D for the desired value
here, we are using S P, S P is a set point. S P means, S P denotes set point.
Set point is nothing, but the desired value and this R S is a steadiest value of R in process
control, we call R S as bias signal, that is basically the steadiest value of reflux rate R. K
C R is the controller parameter, which we need to determine or we calling tuning
parameter that is basically, constant term x D S P is the x D set point that means, that is
the desired value of x D and this is a actual x D.
So, this is one equation similarly, we can write for D, D S plus K C D m D set point
minus m D. For V B, we can write V B equals to V B S plus K C V x B set point minus x
B and last equation is B equals to B S plus K C B m D set point minus m B. These are
the control equations total 4 equations.

(Refer Slide Time: 28:30)

So, if we include 4 equations then degrees of freedom becomes 0, previously we had


degrees of freedom equal to 4 after defining 4 control equations, our degrees of freedom
becomes 0. Now, you will discuss how we can select the manipulated control variable
pear for the example system, how we have selected this 4 control variable manipulated
variable pears.
That will discuss in brief, this is the distillation tower feed flow rate with composition,
one condenser is installed that the top, this is the reflux drum, liquid is accumulated here,
the holdup is m D, this is the top stage. Suppose, this is a control wall, which is use to
manipulate the reflux rate, this is another control wall, which use to manipulate distillate
flow rate, this is reflux flow rate. At the bottom section, this is say bottom tray, this is
feed tray some amount of liquid is accumulated here.
Now, this is the reboiler, this is one control wall, which is use to manipulate vapor boil
operate, another outlet section is included for the out flow of bottom flow rate. Now will
consider first the top section, I mean this section, this is the top section, in this top
section, you see we have selected 2 control variable based of our control objective, one
control variable is x D, another control variable is m D. This 2, we have selected based
on our control objective.
So, how many equations are available for manipulation in the top section, 2 equations, 2
variables are available for manipulation, what are those variables, one is R another one is

D, this 2 variables are available for manipulation. Now, we have to pear the control
variable and manipulated variable, we can pear x D D and m D R another pear may be x
D R and m B D.
So, these are 2 possible pears first we will consider the first pear, we want to manipulate
suppose m D, we want to control m D by the manipulation of R that is no problem, we
can do that. If we consider the control of m D by the manipulation of R, we can easily do
that there is no problem, because this is also outflow, I mean this is also outlet flow, this
is also outlet. So, we can choose any one for controlling the m D.
Another option is x D verses D, suppose we have the distillate composition at present
time 90 percent, I mean the distillate purity presently is 95 mole percent. Now I we want
to change it to 98 mole percent purity, it is perhaps not possible to get 98 mole percent
purity by the manipulation of distillate. But, it may be possible to get 98 mole percent
purity by the manipulation of reflux flow rate.
Because, if we want to get 98 mole percent purity, there is a need to increase the reflux
flow rate, which will affect the separation in the process and they are may be
improvement of composition in the top section. If we want to increase x D, there is a
need to infuse reflux flow rate, that will affect the separation in the process and then
there may be some improvement in composition.
It is usual for binary distillation column, if we want to infuse x D, there is a need to
infuse reflux flow rate, because we are not allowing to live the process, we are again
treating that steam in the process. So, that there is some improvement can be achieved.
So, x D verses D is not a good option and the second option gets preference, I mean x D
verses R and this is not m B, this is m D, m D verses D, this is best option among this 2.
But, the similar O F or the bottom section is this one, this one is the bottom section by
the similar o a for the bottom section, we chose x B verses V B and m B verses B, you
see m B can affect the can affect the process, but B is leaving the process.
So, by the similar way as we discuss for the top section, this is the best option for the
bottom section x D verses V B and m D verses B, this is all about the mathematical
modeling degrees of freedom analyzes of a distillation column. Next, we will start I
mean, we will just try to know in brief the Laplace transform, I hope you have started

this in your basic mathematics course. So, we will not discuss in detail, but will recall the
Laplace transform by some standard formulas.
(Refer Slide Time: 37:28)

So, next we will start the Laplace transform and brief, we will know the standard
formulas basic equation of the Laplace transform of a function f t, we represent by this, f
is a function and t is a time independent variable. This is equal to f bar s the function,
which is written in Laplace domain equal to 0 to infinity, function of t e to the power
minus s t d t, this is the standard formula.
Now, we will just write the formula in terms of time function Hoyer time t is greater than
or equals to 0 and the Laplace transform of those functions. If this function equal to 1,
the Laplace transform is 1 by s function equal to t 1 by s squire, if this is equal to t squire
factorial 2 by s cube.
If this function is t to the power n then this becomes factorial n divided by s to the power
n plus 1, if this is e to the power minus a t Laplace transform of this 1 by s plus a this is
10 to the power n e to the power minus a t Laplace transform is factorial n divided by s
plus a hole to the power n plus 1. If this is sin omega t omega divided by s squire plus
omega squire.

(Refer Slide Time: 40:49)

Then this is equals to suppose, cos omega t s divided by s squire plus omega squre, this
is sin hyperbolic omega t. Laplace transform is omega divided by s squire minus omega
squire, if this is cos hyperbolic omega t, Laplace transform is s divided by s squire minus
omega squire. Next one is exponential of minus a t sin omega t Laplace transform is
omega divided by s plus a hole squire plus omega squire and e to the power minus a t cos
omega t.
It has s plus a divided by s plus a hole squire plus omega squire, there are others
functions are also define in different text books, I mean Laplace transform of different
other functions more functions are available in text book. So, we will not discuss the
Laplace transforms within a more functions.

(Refer Slide Time: 42:45)

So, we will next just consider the Laplace transform of derivatives, for the first order
equation Laplace transform of d f t d t equal to s f bar s minus f naught. This is for first
order differential equation, similarly if we consider second order equation d 2 f t d t
squire, we get s squire f bar s minus s f naught minus first derivative of f naught. This is
for the second order equation.
Similarly, for n third equation f t d t n equal to s to the power n f bar s minus s to the
power n minus 1, f naught minus s to the power n minus 2, first derivative by this way
last term will be n minus 1 derivative at 0. Now, if we consider f as the deviation
variable, suppose f is a deviation variable means, f at any time t, the value of f at any
time t minus the value of f at steady state, then we can call this variable f here, it is
function as a deviation variable.
That means, if we consider this is a function. So, it is a function of deviation variables, if
we consider f as a function f as a deviation function that means, it is a function of
deviation variables. Now the definition of deviation variable is this, the variable at any
time t minus the variable at steadies state. If that is the case I mean, if we consider, if as
the function of deviation variables, what will be this first derivative, I mean the Laplace
transform of first derivative, this will become 0 is not it.
That means, in this t equals to 0. So, f at t equals to 0 minus f steady state, we usually
consider time t equals to 0 is steady stage. So, this becomes 0, similarly this 2 term

become 0, similarly all this terms equal to 0, it is a usual practices in process control the
variables in the model are considered in the form of deviation variable.
(Refer Slide Time: 47:35)

What is the Laplace transform of integrants, Laplace transform of integrals of function of


t d t equal to 1 by s f bar s, if we consider f t as a function in time domain in the
integration of that function is this one. And Laplace transform that function is this one by
is f bar s, other 2 formulas are also used are also use in process control, those are final
value theorem and initial value theorem. Final value theorem is limit t tens to infinity
function of t equal to limit s tens to 0 s f bar s. This is the final value theorem, if our
modeling equations are in time domain then, we have to consider this.
If our modeling equations are converted to Laplace domain in that case, we have to
consider this. Another one is initial value theorem is limit t tens to 0 f t equals to limit s
tens to infinity is f bar s. This is the initial value theorem, similarly if the modeling
equations are in time domain, we have to consider this, if they are in Laplace domain, we
can consider the right hand formula. So, these are about the in brief about the Laplace
transforms. Next, we will go to discuss the different forcing functions, forcing function
are usually the input variable.

(Refer Slide Time: 50:42)

I mean, suppose we have a process, this is a chemical process, this is a input to the
process and this is the output. Now, suppose initially the process is at steady state, we
went to investigate the transient behavior of this process, we want to observe the
dynamics of this process, how we can observe that by the use of simulator.
We have to introduce some change in input variable, if we give some change in input
variable then, we will get some change in output variable that is basically, the transient
behavior of a process or the dynamic characteristics of a process. So, what type of input
variables, we will consider in our process control course. So, that corresponding output
behavior, we can observe from the simulate that, we will discuss next.
Now, there are different input variables, which we will consider, those are like step
input. Another one is pales input, it is better to write step input function, pales input
function then ramp input function. So, these functions are basically called forcing
function, I mean, we will use this name forcing function. Now, we will give change in
input variable say by this fashion will give change in input variable as a step input then
we will observe the output dynamics. Similarly, we can consider the pales chain in input
variable then the effect in terms of output variable, change of input variable as ramp
input, I mean ramp change then the transient behavior in terms of output variable that,
we will discuss in the next class, fine.
Thank you.

Process Control and Instrumentation


Prof. A. K. Jana
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture - 6
Dynamic Behavior of Chemical Processes
(Refer Slide Time: 00:56)

We will discuss the Dynamic Behavior of Chemical Processes. I have mentioned in the
last class about this in brief, I mean what is transient behavior, how we can get the
transient behavior. So, we will start from that only, we have a chemical process, this is
input this is output, if we give a change in input variable, we will get some change in
output variable. And that is the dynamic behavior and that we will observe in terms of
the change in the output variable.

So, we will consider the different forcing functions, first we will start with step function.
So, first I am drawing the step function, what it how it looks the function, we will
represent by f. So, this is f t and this is time t, initially, if the process is at steady state
condition, I mean there is no change. Now, we are changing the input variable in this
way. So, initially it was 0 definitely, we are considering here f as A deviation variable,
that is why that is 0. F we are considering as a deviation variable here, f is not function
here, f is variable note this point

So, since f is the deviation variable, that is why at steady state, f equal to 0. Now, we are
changing this, f value at suppose time t equals to 0 and the magnitude of this change is
suppose A, this is a step function, which we have represented graphically here. Now,
how we can represented mathematically, f t equals to A that is a step function. So, if we
write the Laplace transform of this function f t, what would be the Laplace transform of
this A by S.
So, the Laplace transform of a step function is A by S where, A is the magnitude of that
step function, what is the example of this step function. Suppose, we have a steam valve,
we have one steam valve. Now open the steam valve at t equals to 0, such that Q changes
by A units, we have a steam valve open the steam valve at time t equals to 0.
So, that Q changes by A units, here Q is the steam flow rate and that is basically, f fine
this Q is basically f. So, if we change the Q by A units at t time equals to 0, that is the
example of a step input. So, we can change the load variable, disturbance variable by this
fashion to observe, the transient behavior of the process, this is the first type of forcing
functions, which is commonly used.
(Refer Slide Time: 06:27)

Second forcing function is rectangular pulse function, first we will draw the figure of this
rectangular pulse function so, this is f t and this is time t, initially if f t is 0. Now it
changes at time t equals to 0 by unit of suppose A, then it remains there up to time b then
it returns to the old value, this is the graphical representation of a rectangular pulse

function. Initially f t is 0 that means, the process is at steady state then at time t equals to
0, the input function changes by magnitude A, then it remains, there for a time duration
of b and finely it returns to the old value.
Now, how you can represent it mathematically, f t equals to 0 and time t less than 0, f t
equals to A, when time is in between 0 and b and again this is 0, when time is greater
than b. So, this is the mathematical representation of this pulse function, this pulse
function is basically the combination of 2 input functions, this is the combination of 2
input functions. So, what are these 2 input functions, if we considers one function is f 1
and another is f 2 then the graphical representation of f 1 is like this initially it is at 0
level, this is 0 and this quantity is A.
So, f 1 t equal to 0, when t is less than 0 and this is equal to A, when t is greater 0,
another function is f 2 and this is suppose time t initially, it is at the value of A then it is
coming back to this value and this time is b. So, f 2 t equals to 0, when t is less than b, f 2
t equals to A, when time is greater than b can, we write this for the function second
function f 2 can, we write this, we are considering this is 0.
So, this is a initial value, we are considering 0 and this we are considering A that means,
this is f 1 t minus b, here we are considering increasing, I mean it is increasing the step,
here we are considering decreasing, initial value is 0 and this quantity is a by that way,
we can write f 2 equals to f 1 t minus b.
(Refer Slide Time: 11:30)

Now, f t is a combination of these 2, f 1 t f 2 t minus b see minus sign is included here,


because for f 1 function, we have considered that increase of step. But, for the case of
second function, we have consider the decrease of step, that is why negative sign for f 2
and positive sign for m 2 sorry, this is f 1.
Now, we will take Laplace transform of this, Laplace transform of f t. So, Laplace
transform of f 1 t minus f 1 t minus b, this is equal to f 1 bar s and for this, we can write e
to the power minus b S f 1 bar S. So, f one bar S is 1 by A, S no f 1 bar is A by S, that is
the step function only with the magnitude of A. So, A by S 1 minus e to the power minus
b S, this is the Laplace transform of a rectangular pulse function, this is a final form for
the pulse function.
Now, what is the example of the pulse function, if we consider the steam valve, open the
steam valve at t equals to 0. So, that Q changes by A units, hold at the new valve for the
duration of b then return to the old valve, you just try to remember that configuration, I
mean the graphical representation of pulse function again, I am telling open the steam
valve at time t equals to 0. So, that Q changes by A units hold at the new valve for the
time duration of b and return to the old valve, this is a example of a pulse function. Now,
we will just modify that figure of pulse function, previously we have consider the time
duration of b.
(Refer Slide Time: 14:35)

Now, we will consider the time duration of A, this is f t 0, initially then this is 0 and this
time step is suppose A and we are considering here 1 by A, this is time t. So, what is the
area is 1, A multiplied by 1 by A that equals to 1 and this is called unit pulse function, it
is conventionally represented by del A t. So, this is unit pulse and it is represented by del
At
Now, we use to consider limit A tens to 0 A multiplied by 1 by A, how much is this 1,
we are just approaching 0, we are approaching the time 0. So, how much it will be then
1, how we can represent it graphically, how we can represent this situation, graphically.
The graphical representation of this is somewhat like this, this is time t equals to 0, this is
0, this function is called unit pulse function sorry, unit impulse function.
It is also called dirac function and it is conventionally, represented by del t, this is called
dirac function or unit impulse function and it is represented by del t. So, what is the
definition of this unit impulse function, it is defined as equal to 0 for all times accept for
time t equals to 0. So, we can represent this again by minus infinity to infinity del t d t
equals to 1, if we consider minus infinity to infinity then del t equal to 1 within this range
within this time range. Now, what will be the Laplace transform of this unit impulse
function.
(Refer Slide Time: 18:56)

Laplace transform of this del t equal to Laplace transform of the unit pulse function in
addition, I mean, if we consider limit a tends to then we can write here, unit pulse

function. So, this is unit pulse function that means, integration 0 to infinity limit A tends
to 0, del A t e to the power minus S t d t, we are just considering the definition of
Laplace transform. Integration of 0 to infinity limit A tends to 0 and this is the Laplace
transform of unit pulse function.
We will just take out the limit A tends to 0, integration 0 to infinity del A t e to the power
minus S t d t. This is equal to limit a tends to 0, this part, we have already calculated how
much is that 1 minus e to the power minus S A divided by A S, we have basically,
consideration in the previous example the time duration b, but in this example, we are
considering A.
So, you just replace b by A and in the unit pulse in the discussion of pulse function, we
have consider the magnitude that is A, here we are considering 1 by A. So, we have
derived the Laplace transform of pulse function, that is A by S 1 minus e to the power
minus b S just, we need to replace b by A and A by 1 by A then, we will get this
equation, so how much it will be 1, how by the use of el hospital's rule.
So, this is the Laplace transform of the unit impulse function can you give any example
of this unit impulse function, I mean we can continue the same example say, we have 1
stream valve can you give the example of this impulse function. It is impossible to
realize perfectly, but we can give the example like this open the stream valve and
instantaneously return to the old value open the stream valve at t equal to 0 and as soon
as physically possible. There after return to old value, this is the example of impulse
function.

(Refer Slide Time: 23:06)

Another input function is also used in process control that is ramp function that is ramp
function, we can represent this function in this way time t initially, it is 0 then it changes
like this, suppose the slope is A. Initially, it is 0 and at time t equals to 0, it changes
linearly with slope A. So, if f t is ramp function then, it is 0, when t less than 0, it is A t
when t greater than 0 and the Laplace transform of this is A by S square. So, what will be
the example gradually open the stream valve.
So, that Q increases linearly, gradually increase the stream valve. So, that Q increases or
Q changes linearly and ramp ends, when the valve is fully open. So, these are the forcing
functions, which are used to observe the transient behavior, definitely in presence of
controlled scheme, but presently, we are not discussing the control part, we are just
discussing the simulation racers, I mean the model output.
Next, we will discuss the different how to solve the different model structures,
practically most of the chemical processes are non linear in nature, almost all chemical
processes, so non linear dynamics. Now analytical solution of non linear equations are
not available.

(Refer Slide Time: 26:11)

So, how we can proceed, one option is we can solve the modeling equations numerically,
if we have the non linear equations, non linear model, we can solve it numerically. So,
keeping the numerical simulation of the modeling equations, we can proceed for the
development of non linear model equations. Another option is we can linearize the non
linear model. So, we can linearize the non linear model. So, linearization of non linear
equations or non linear model and we can go for analytical solution, if we can linearize
the non linear equations, we can go for analytical solution.
Third option is we can develop the linear model for the non linear processes and we can
solve the linear equations again analytically. So, there are 3 options, one is development
of non linear model and go for solution by the use of numerically techniques, another
one is linearize, the non linear equations and follow the analytical solutions and third one
is develop the linear model structure and simulate analytically. But, this third option was
used in the past. Now a days it is not used, due to the advancement of different
mathematical techniques.
So, this option is quite old. So, we will not discuss any more this in the first option, we
have option to model the process by the development of non linear equations and the
solution of those equations numerically, I hope you have studied the numerical
simulations in your mathematics courses. So, we will not discuss about that in more
details, I mean, we do not want discuss in this course the numerical simulation of non-

linear equations. Second option is linearization of non linear model structure, so that we
will discuss in brief.
(Refer Slide Time: 29:10)

So, first we will start the linearization of single variable system, we will consider a
simple ordinary differential equation represented by d x d t equal to function of x and t,
suppose this is equation number 1, here x is the only variable, I mean dependent variable
and t is the independent variable that means, x is also varies with time. So, we can omit
the time t within this function, we can write in this form also, this is suppose equation
number 2.
Now, we will linearize this function f according to Taylor series expansion. So, we will
use the Taylor series to linearize f x accordingly, if f x equal to f x naught here, f naught
is the reference x, I mean reference value of x plus d f d x at x equals to x naught x minus
x naught plus d 2, f d x square at x equals to x naught x minus x naught whole square
divided by factorial 2. D n f d x n x equals to x naught x minus x naught to the power n
by factorial n like this.

(Refer Slide Time: 32:05)

This is the Taylor series expansion of f x neglecting the term of order 2 and higher, we
get f x is approximately equal to f x naught plus d f d x at x equals to x naught multiplied
by x naught, you give this Taylor series equation, as equation number 3, the Taylor series
expansion of f x, you put equation 3 for that and this approximate linearization, we are
representing by equation 4.
Now, this linear approximation is satisfactory, when x is very close to x naught, this is f
x, this is x. Suppose this is the function f x, this point is x naught and corresponding f x is
f x naught, now if, we linearize with respect to this point. So, this is basically, if this
linear function is represented by plus d f d x at x naught x minus x naught.
So, from this graphical representation, it is very clear that the linear approximation is
satisfactory, when this x is very close to x naught, if we move away from this x naught
this part is not linear at all fine. So, the deviation increases as we move far from x equals
to x naught.
Now, we can write equation 4 like this way also d x d t equals to f x naught plus d f d x
at x equals to x naught x minus x naught. Because, f x equals to d x d t that is our
equation number 2, we have just replaced f x equals to d x d t in equation 4 and then we
get equation 5, this is a linearization of f x according to Taylor series expansion.

(Refer Slide Time: 35:56)

Now, we will take 1 example to linearize the model of that particular example. So, we
will consider the example of liquid tank system, this is the schematic of the liquid tank
system, liquid height and cross sectional of the tank input flow rate is f i and outlet flow
rate is f naught. We have derived earlier the modeling equation for this system. The
modeling equation is A d h d t equal to F i minus F naught, this equation, we can derive
based on total mass balance.
Now, this outlet flow rate is basically the function of liquid height the outlet flow rate F
naught depends on the liquid height in the tank, we will consider 2 cases. In first case,
we will consider F naught is proportional to h anyway, you give this as equation number
6. In the first case, we will consider F naught is proportional to h that means, F naught
equals to beta h, if we substitute this in equation 6, I mean if we substitute F naught
equals to beta h in equation 6 then, we get A d h d t plus beta h equal to F i. If we
substitute F naught equals to beta h in equation 6, we get this form and this is a linear
ordinary differential equation. So, there is no need of any linearization, we can solved it
analytically.
In another case, we will consider this F naught is proportional to root h, F naught is the
square root of liquid height in the tank that means, F naught equals to say alpha root h.
Now, if we substitute again F naught equals to alpha root h in equation 6, we have a d h
d t plus alpha root h equal to F i, if we substitute F naught equal to alpha root h in

equation 6, we get equation 7, this is not a linear ordinary differential equation. So,
which non linear term is present in this equation, the non linear term is alpha root h. So,
we will linearize this alpha root h term using the Taylor series expansion, we will
linearize this alpha root h term, using the Taylor series expansion and then we substitute
that in equation 7.
(Refer Slide Time: 40:08)

So, according to Taylor series alpha root h equal to alpha root h naught plus
differentiation of alpha root h at point h naught, here h naught is the reference height, h
minus h naught. Similarly, if the second term, the second derivative will be like this at h
equals to h naught h minus h naught whole square divided by factorial 2. We neglect the
terms of order 2 and higher, neglecting terms of order 2 and higher, we get alpha root h
equals to alpha root h naught plus alpha divided by 2 root h naught multiplied by h minus
h naught. This is the final form of alpha root h after linearization.
Now, we will substitute this alpha root h term in equation 7 substitute it, if we substitute
in equation 7, then we get A d h d t plus alpha divided by 2 root h naught multiplied by h
equal to F i minus alpha by 2 root over of h naught. This is suppose equation number 8.
So, this is the linearized form of the non linear model or non linear modeling equation.
Now, if we substitute h equals to h naught, if we substitute h equals to h naught equation
7 and 8, they are identical, you see if we substitute h equals to h naught equation 7 and
equation 8. They are same, equation 7 represents the non linear form and equation 8

represents the linear form exactly at h naught, they are identical. So, that is why, we have
mention earlier that the linear approximation is satisfactory, when x is very close to x
naught. So, this is the linearization. Now, we will try to represent the model in terms of
deviation variables.
(Refer Slide Time: 44:27)

We will represent this model in terms of deviation variables, how we can do that at
steady state x is basically represented by x s, x at steady state, we will represent by x s, if
x equals to x s at steady state then we can write d x s d t equal to 0, equal to function of x
s see the steady state value of x is x s x s is basically constant. That is why the derivative
of that constant term is 0 and this is the function of x, now we are replacing x by x s that
is why that is f x s, this is say equation number 9.
Now, according to Taylor series expansion, we can write f x just by replacing x naught
by x s d x d t equals to f x s plus d f d x at x equals to x s x minus x s the linearization of
f x, we got earlier here, we have just substituted x naught equals to x s and we get
equation 10. Now, we will just subtract, we will subtract equation 9 from equation 10
then we can write differentiation of x minus x s d t equal to d f d x at x equals to x s
multiplied by x minus x s, this is equation number 11.
In the next, we will consider one deviation variable that is represented by x prime equal
to x minus x s x naught is the deviation variable and that is equal to x minus x s then
equation 11 becomes d x prime d t equal to d f d x at x equals to x s x dot. Simply, we

have substituted in place x minus x prime sorry, we have substituted in place of x minus
x s x prime then, we get equation 12. Now, we will continue that example again that
liquid tank system.
(Refer Slide Time: 48:30)

So, previously we got the linearized form for the liquid tank system, we got the
linearized form in equation 8. So, the linearized form is A d h d t plus alpha divided by 2
root h s multiplied by h equals to F i minus alpha by 2 root over of h s say this is
equation number 13, we get equation 13 from equation 8 directly.
So, for the liquid tank system, this is the linearized form at steady state, this equation 13
yields A d h s d t plus alpha root h s equal to F i S to represent the steady state, we have
included suffix S with all variables. Then we get A d h s d t plus alpha root h s equals to
F i S now, we will just subtract equation 14 from equation 13, if we subtract equation 14
from equation 13, we get a differentiation of h minus h s plus alpha divided by 2 root h s
h minus h s equals to F i minus F i s. Now, we will define 2 deviation variables, one is F
i prime that is F i minus F i S, another deviation variable, we will define that is h prime
equal to h minus h s, these are the 2 deviation variables.
Now, if we write the above equation in terms of deviation variables, we finely get A d h
prime d t plus alpha divided by 2 root h s, h prime equals to F i prime, this is suppose
equation number 15. So, for the liquid tank system after linearization, we got one
equation and in terms of deviation variables, we got equation 15. So, this is basically, the

linearization of non linear equations for a single variable system, I mean only one non
linear equation is involved in this example. In the next, we will discuss the linearization
of multivariable systems.
Thank you.

Process Control and Instrumentation


Prof. A. K. Jana
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture -7
Dynamic Behavior of Chemical Processes (Contd.)
In the last call, we discussed linearization of single variable systems, first we have
linearize the non linear equation having single variable using Taylor series expansion
and in the next, we represented that linearized model in terms of deviation variables. So,
today, we will extend our discussion to linearization of multi variables systems.
(Refer Slide Time: 01:30)

So, todays topic is linearization of multi variable system, the approach is quite similar
what we did in the last class. So, today we will consider a multivariable system
represented by these 2 equations, d x 1 d t equal to function of x 1 and x 2 and another
equation is d x 2 d t, second function is f 2 and that is also the function of 2 state
variables one is x 1, another one is x 2 and we will continue the equation numbers.
I mean yesterday, we have considered up to equation number 15. So, today we will start
from equation number 16 and this is equation number 17. So, this is the representation of
a multi variables system. So, by multivariable system, we mean a system having 2 state
variables by multi variable system, we mean a system having 2 state variables, what are
these variables, one is x 1, another one is x 2.

And so, these are 2 state variables one is x 1, another one is x 2, 2 functions are also
involved one is f 1 and f 2, both these functions are non linear functions, these are non
linear functions. So, what we need to. So, we need to linearize these 2 non-linear
equations by the use of Taylor series.
Now, according to Taylor series what will be the linearized form of the first function that
is f 1 x 1 and x 2, what will be the first term will be f 1 x 1 naught x 2 naught. What will
be the next term will be the differentiation f 1 with respect to x 1 and we need to
represent this at a particular operating point, that is suppose x 1 naught x 2 naught, it will
be multiplied with x 1 minus x 1 naught.
In the next step, we need to differentiate it with respect to second state variable that is x
2. So, with respect to x 1 naught x 2 naught multiplied with x 2 minus x 2 naught, next
we have to write the second derivative dou 2 f 1 dou x 1 square at the operating point x 1
naught x 2 naught then whole square of these divider by factorial 2. Similarly the second
derivative with respect to dou x 2 x 1 naught x 2 naught multiplied by x 2 minus x 2
naught whole square divided by factorial 2.
(Refer Slide Time: 06:30)

What will be the next term dou 2 f 1 dou x 1 dou x 2 with respect to I mean at the point x
1 naught x 2 naught multiplied by x 1 minus x 1 naught x 2 minus x 2 naught. So, this is
the linearized form according to that Taylor series expansion of the non linear function f
1.

Similar expression, we will get for the case second case of f 2, so I am not writing that
similar expression, we will get. Now neglecting the order, neglecting the terms of order
two and higher what will get.
(Refer Slide Time: 07:30)

Neglecting the terms of order 2 and higher what will get, if we write from d x 1 d t d x 1
d t. So, it will be f 1 x 1 x 2. Since we are approximating this, so it will be f 1 x 1 naught
x 2 naught plus this x 1 naught x 2 naught multiplied by this plus dou f 1 dou x 2 at x 1
naught x 2 naught multiplied by x 2 minus x 2 naught. This is the linearized form and by
neglecting the terms of order true and higher.
So, we will give equation number 18 for this, we will use equation 18 for this equation
similarly what will be the equation for the second equation. I mean second modeling
equation is write, the second equation also d x 2 d t that will be f 2 x 1 x 2 x 1 naught x 2
naught plus dou f 2 dou x 1 at x 1 naught x 2 naught multiplied by x 1 minus x 1 naught
plus dou f 2 dou x 2 x 1 naught x 2 naught multiplied by x 2 minus x 2 naught.
So, this is suppose equation number 19, what is the next step, we represent this linearized
model in terms of deviation variables. So, what will be the next step, we need to write
the equations considering steady state.

(Refer Slide Time: 10:51)

At steady state, we will assume that x 1 naught equals to x 1 s, basically we are replacing
the suffix 0 by s. Similarly, we will write x 2 naught equals to x 2 s at steady state, we
will consider x 1 s and s 2 s. So, how we can representing the modeling equation at
steady state, first equation will be d x 1 s d t x 1 is s is if constant term. So, we can write
it as 0, this equals to first function and that is the function of x 1 s and x 2 s. So, this is
suppose equation number 20.
Similarly, second equation, we can write d x 2 s d t equal to 0 equal to f 2 x 1 s x 2 s,
suppose this is equation number 21. So, what we need to do in the next step to represent
in terms of deviation variables, we have to subtract the steady state models from
linearized models.
So, next we will just subtract the equation 20 from 18, similarly we have to subtract
equation 21 from 19. Basically we are subtracting the steady state modeling equations
from linearized modeling equations. So, what equation, we will get for the first state
equation x 1 minus x 1 s d t equal to with respect to I mean at x 1 s x 2 s x 1 minus x 1 s
plus dou f 1 dou x 2 x 1 s x 2 s x 2 minus x 2 s. So, we will give some equation number
say equation 22.

(Refer Slide Time: 14:00)

So, similarly we will get for the second state variable that will be d differentiation of x 2
minus x 2 s d t equal to dou x 1 at x 1 s x 2 s multiplied by x 1 minus x 1 s then partial
differentiation with respect to second state variable x 2 minus x 2 s. Suppose, this is
equation number 23.
Now, we assume the deviation variables one is x 1 prime another deviation variable, we
assume x 2 prime, what is x 1 prime x 1 minus x 1 s, similarly x 2 prime is x 2 minus x 2
s, these are 2 deviation variables. Now, if we substitute these 2 deviation variable
expressions in equation 22 and 23, what will be the expressions, I mean equations in
terms of deviation variables, that we need to find in the next. We will just represent the
equations 22 and 23 in terms of x 1 prime and x 2 prime. So, d x 1 prime d t equal to
suppose, we using coefficients a 1 1 a 1 2 and for the next expression, we will use a 2 1 x
1 prime plus a 2 2 x 2 prime, if we write in this way.

(Refer Slide Time: 16:21)

So, what would be the coefficient terms a 1 1 is x 1 s x 2 s a 1 2 is this 1, similarly a 2 1


will be this 1 and forth coefficient is a 2 2, that will be x 1 s x 2 s. So, we got the
equations, now we can represent those linearized equations in terms of deviation
variables in matrix form, we can represent the final expressions like this.
The dot symbol represents the time derivative, so coefficient matrix a 2 1 a 2 2
multiplied by x 1 prime x 2 prime is this equation. So, this is the linearized form for a
multi variable system, in this multivariable system, we have considered only 2 state
variables. So, similarly, we can derive the expression for more than 2 variables. Now, we
will just take 1 example, which has 2 state variables. So, we have considered previously
1 C S T l example, that will continue to linearize using the Taylor series expansion.

(Refer Slide Time: 18:54)

So, next we will consider one example, which we have discussed earlier, this is a
jacketed C S T l, the jacket fluid is entering here and it is coming out here. Feed is
introduced with a flow rate of f, I concentration of C A i and temperature of T i, suffix i
indicates input and product is coming out with a flow rate of F concentration of
component A that C A and temperature T. So, this composition of C A temperature is T
volume of the reacting mixture, volume of the mixture is suppose B.
Previously, we have considered 1 A reversible first order exothermic reaction. So,
irreversible first order exothermic reaction, that is represented by this A is the reactant
and B is the desired product. So, remove the exothermic heat from reactor, we need to
introduce coolant steam in the jacket. So, this is basically A coolant, which has the flow
rate of suppose F c and temperature is T c i, the coolant is coming out at the same flow
rate F c and temperature T c naught.
And we are assuming the temperature is T C naught. So, based on the standard
assumptions, which we have considered earlier, we got 3 modeling equations. So, we
recall these modeling equations one is d v d t equals to F i minus F, F is basically the
volumetric flow rate, this equation we got based on total mass balance.
Similarly we got another equation for component mass balance considering component
mass balance, we got d C A d t equal F i divided by V C A i minus C A minus K naught

exponential of minus e divided by R T into C A. This equation we got considering


component mass balance. Third equation, we got considering energy balance.
So, that equation is d t, d t equal F i divided by V T i minus T minus Q divided by V rho
C P plus minus of dell h K naught C A exponential of minus e divided by R T, whole
divided by rho C P, this equation we got performing energy balance.
You see all these 3 equations are state equations, because these equations have been
derived by the application of the conservation principle on fundamental quantities. Now,
we have considered the formulation of linearization for 2 variable systems. So, in this
particular example, we will consider constant volume. So, that we get 2 variable system.
So, in the we will discuss basically the linearization technique step wise.
(Refer Slide Time: 24:53)

So, in the first step in the first step say this suppose step 1 develop the dynamic model in
the first step, we need to develop the dynamic model, we will continue this example c s t
r, further assuming constant volume system. So, we are basically assuming here, V is a
constant quantity, if that is the case then there will not be in existence of total mass
balance. So, the modeling structure will reduce to a structure having 2 equations.
So, we will write the final model structure d C A d t equal to 1 by tau, C A i minus C A
minus K naught exponential of minus e divided by R T multiplied by C A, here tau is V
divided by F i , tau is V divided by F i. Similarly, we will write another equation that is d

T, d t equal to 1 divided by tau T i minus T minus Q divided by V rho C P plus S K


naught exponential of minus E R T C A. So, what is S is minus of dell H divide by rho C
P, this is S and you see this S is a constant quantity, similarly tau is also constant. So, this
is the final model structure, which we need to develop in step 1.
Next is step 2 next 1 is step 2 in step 2, we need to identify the non linear terms present
in the modeling equations. You see in the first equation I mean, which we got based on
component balance, the first right hand term, this is linear term k naught is pre
exponential factors. So, that is constant
So, only non-linear term is here, this one C A multiplied by exponential of minus E
divided by R T, that is only non linear term present in the first equation. Similarly, in the
second equation first right hand term that is linear term, second right hand term V rho C
P, they are all constant only variable is Q. So, this is also linear term.
In the third right hand term S is constant K naught constant. So, here also the non linear
term is exponential of minus E divided by R T multiplied by C A. So, in this model
structure, there is only a single non linear term that is exponential of minus E divided by
R T multiplied by C A. So, we will basically, linearized this term using the Taylor series
expansion. So, that is the third step.
(Refer Slide Time: 29:53)

So, step 3 linearize the non-linear term, using Taylor series expansion you linearize it the
non-linear term. So, that is exponential of minus E divided by R T multiplied by C A,
this is equals to exponential of minus E divided by R T naught multiplied by C A naught
then, we will derive this E divide by R T C A with respect to temperature at T naught C
A naught multiplied by T minus T naught. Similarly, we will derive we will differentiate
with respect to C A, at T naught C A naught and we will multiplied with C A minus C A
naught
Here also, we are neglecting the terms of order true and higher. So, what will get finally,
we will get exponential minus E divide by R T naught C A naught plus E divided by R T
naught square exponential of E divide by R T naught into C A naught T minus T naught.
Third term will be exponential E divided by R T naught minus E divided by R T naught
into C A minus C A naught. So, in this third step, we have just linearize the non linear
tau.
In the next step, I mean step 4, we will substitute this linearized expression in the non
linear modeling equations, we will substitute the linearized term of E to the power minus
E divided R T C A, in the non linear modeling equation substitute this. In place of this
substitute this term, I mean this part.
(Refer Slide Time: 33:28)

So, in step 4, we need to construct the linearized model. So, for the case of concentration,
we will get this equation d C A d t first right hand term is 1 by tau C A i minus C A.

Next term is K naught multiplied by the non linear term. So, K naught exponential of
minus E divided by R T naught C A naught plus E divided by R T naught square E to the
power, I mean exponential of minus E divided by R T naught C A naught T minus T
naught and the last term is exponential of minus E divided by R T naught multiplied by
C A minus C A naught, am I correct.
So, for the temperature, I mean d T d t equation will be the first right hand term is 1 by
tau in terms of temperature, 1 by tau multiplied by T i minus T and Q divided by V rho C
P is there that is also linear term then the third term was S K naught multiplied by the
non linear term. So, you will just substitute here, exponential minus E divided by R T
naught C A naught then E divided by R T naught square exponential minus E divided by
R T naught multiplied by C A naught multiplied by T minus T naught.
Last term is exponential minus E divided by R T naught C A minus C A naught. So, after
substituting the linearized expression in the non linear model, we get these 2 equations
what is the next step, in the we need to write the modeling equations at steady state,
because ultimately, we have to write the equations linearized equations in terms of
deviation variables.
So, step 5 is we have to write the modeling equations at steady state. So, represent the
model at steady state condition. So, if we represent at steady state then we get this
equations, the it was d C A d t. So, we are just adding the suffix 0 to represent the steady
state, we are adding suffix 0 to represent the steady state 1 by tau C A i naught minus
naught minus K naught exponential minus E divided by R T C A naught, then this is the
steady state representation of the component mass balance equation.
Similarly, d t naught d t equal to 0 equal to 1 by tau T i naught minus T naught minus Q
naught divided by V rho C P plus S K naught exponential minus E divided by RT naught
C A naught, this is the steady state representation of the second modeling equation. So,
sometimes, we are using a suffix S, sometimes we are using suffix 0 to represent the
steady state.

(Refer Slide Time: 39:25)

In the final step that is step 6, we have to derive, the linearized model in terms of
deviation variables. So, we will write the final expression by subtracting steady state
model, steady state modeling equation from linearize modeling equations. So, step 6 is
derive the linearized model in terms of deviation variables, if we subtract steady state
equations from linearized modeling equations, we get these 2 equations, I am writing
those 2 equations.
So, 1 is C A minus C A naught d t equal to 1 by tau C A i minus C A i naught, next 1 is
C A minus C A naught. Then K naught E divided by R T naught square E to the power
minus exponential of minus E divided by R T naught multiplied by C A naught
multiplied by T minus T naught the last term is K naught E to the power minus E divided
exponential of minus E by R T naught C A minus C A naught.
If, we subtract the steady state modeling equations from linearized equation, we get this
component mass balance equation. So, we will represent this in terms of deviation
variables. So, what will be that d C A prime d t equal to 1 by tau C A i prime minus C A
prime minus K naught E divide by R T naught square exponential of minus E divided by
R T naught C A naught T prime. Last term will be K naught E divide by R T naught
square exponential minus E divided by R T naught multiplied by C A prime.
Here, we are considering last term this will not be there no here. So, this will be K
naught exponential of minus E divided by R T naught C A prime. So, here we are

considering basically C A i prime, that is C A i minus C A i naught, this is 1 deviation


variables, we have consider in this equation, another one is C A prime that is C A minus
C A naught, another deviation variable is here T prime, that is T minus T naught, so this
one and this one. So, 3 deviation variables, we have basically considered. So, similarly
we need to write the equation in terms of temperature.
(Refer Slide Time: 44:08)

That will be T minus T naught d t equal to 1 by tau T i minus T i naught minus T minus
T naught then Q minus Q naught divided by V rho C P plus S K naught multiplied by E
divided by R T naught square exponential of minus E divided by R T naught C A naught
T minus T naught plus exponential of minus E divided by R T naught into C A minus C
A naught.
After subtracting the steady state equation, we get to this. So, if we again substitute the
expression of deviation variables, we get d T prime d t equal to 1 by tau T i prime minus
T prime minus Q prime divided by V rho C P then S K naught E divided by R T naught
square exponential of minus E divided by R T naught C naught C A naught T prime plus
exponential of minus E divided by R T naught C A prime.
So, here one deviation variable is T i prime, that is T i minus T i naught, this already we
have defined T prime, another one is Q prime that is Q minus Q naught this is define,
this is define. So, these are the 2 modeling equations for the example C S T S system in

linearized form and in terms of deviation variables. So, next we will discuss the
derivation of transfer functions, which is used in process control.
(Refer Slide Time: 47:11)

So, next topic is transfer functions, usually the process is I mean the modeling equations
are represented in time domain, but the transfer function is represented in S domain
Laplace domain. So, that job we need to do to represent the model in terms of deviation
variables in terms of in Laplace domain.
So, first we will consider one process having inputs and outputs in time domain, suppose
this is the block of a process then, this is the input, which will represent by f and time
will use as independent variable. This is the output of the process, which will represent
by y t, now here by process, we mean the model structure of the process, we will use the
modeling equations of a process to derive the transfer functions.
So, we will consider one example first, which is a single input, single output system first,
we will consider, we will first, we derive the transfer function for a single input single
output system that is S I S O. So, a single input single output process, we can represent
by this form, which is definitely linear or linearized version, a single input single output
system, we are representing by nth order linear or linearized equation n minus 1 y d t n
minus 1, we are trying to formulate in general. That is why, we are we have consider nth
order equation, differential equation a 1 d y d t plus a naught y equal to b f t.

So, y is a output, f is a input a and b, they are the coefficients and what else, t is a
independent variable. So, as I mention that the transfer function should be represented in
S domain. So, we have to take Laplace transform of this equation. What is the Laplace
transform I mean, if we consider the Laplace transform of d n y d t n, this is equals to S
to the power n y bas S minus S n minus 1 y naught minus S n minus 2 y prime naught
like this last term is y n minus 1 naught. This is the Laplace transform of nth order
derivative.
Now, if we consider in this equation, in this process model y and f both are the deviation
variables, these 2 are deviation variables, if that is the case, I mean, if y is the deviation
variable then definitely all these terms become 0. So, we can write this as S n y bar S
similarly, we can consider another derivative. So, if we take the Laplace transform of this
linear or linearized nth order differential equation.
(Refer Slide Time: 52:37)

We get a n S to the power n y bar S plus a n minus 1 S to the power n minus 1 y bar S
like this a 1 S y bar S plus a naught y bar S equal to b f bar S can, we write this. If we
take the Laplace transform of the linear or linearized nth order differential equation, we
get this.
Now, we can rearrange this equation like this equal to b divided by a n S to the power n a
n minus 1 S to the power n minus 1 a 1 S plus a naught. This is the output in S domain,
this the input in S domain, this is the right hand side, this right hand side is basically the

transfer function, which is represented by G S, this is the transfer function, I mean the
transfer function is represented by G S and that is b divided by a n S to the power n a n
minus 1 S to the power n minus 1 a 1 S plus a naught.
So, this transfer function basically represents the linear or linearized input output model
in terms of deviation variables, you see this is the input output model. So, transfer
function represents the linear or linearized input output model in S domain Laplace
domain. So, we have represented using the block the processing time domain, now we
need to represent in terms of Laplace domain the whole process. How we can represent
it, it is very simple.
(Refer Slide Time: 55:20)

We will represent the process by G S here, our input is we will write in s domain our
output is in S domain, you recall the representation in time domain. Now, will Laplace
domain, I mean S domain, we will represent in this way. So, this is the block diagram,
this is the block diagram of this process. So, this is the formulation of transfer function
for a single output single input, S I S O system and in the next day, we will discuss multi
input multi output, multi input single output transfer functions.
Thank you.

Process Control and Instrumentation


Prof. A. K. Jana
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture - 8
Dynamic Behavior of Chemical Processes (Contd.)
(Refer Slide Time: 01:00)

We will continue our discuss on transfer functions. Yesterday, we have considered a


single input single output system and we have derived the transfer function for that SISO
system. So, for a SISO system the process involves Single Input and Single Output,
suppose this is a process, this is the input to the process and this is the output for the
process, we have represented the input by f t and output by y t and the process is
modeled by an nth order linear or linearized non linear equation.
So, we considered in the last class the nth order model represented by a n d n y d t n plus
a n minus 1 d n minus y d t n minus 1 like this a 1 d y d t plus a naught y equals to b f t.
A and b these are constant coefficients and both output and input these are deviation
variables. In the next step, we took Laplace transform of the process model and finally,
we got the expression in f S domain that is like this y S divided by f S equal to b divided
by a n S to the power n plus a n minus 1 S to the power n minus 1 a 1 S plus a naught,
which is basically the transfer function represented by G S.

Then in the next step, we made the block diagram is like this G S is representing the
process the input to that process in Laplace domain in f bar S and output is y bar S, this is
the block diagram for the SISO system. Now today, we will consider one multi input,
single output system and we will try to derive the transfer function for that process.
(Refer Slide Time: 04:47)

Today, we will derive the transfer function for a multi input single output system MISO.
So, the process, we can represent by this. Suppose the process this is the process, this is
one input represented by f 1 t, another input we will represent by f 2 t and single output
that is y t. So, this is the representation of a multi input single output system.
So, next we have to consider the model for this process. So, we will represent the process
by this equation a n d n y d t n plus n minus 1 d n minus y d t n minus 1 like this a 1 d y d
t plus a naught y. And, right hand side includes 2 input variables one is b 1 f 1 t and
another one is b 2 f 2 t. The process model is quite similar with that of the SISO system
only new addition is the input variable, in the previous case, we had 2 input variables,
here one input variable.
So, next we have to take the Laplace transform, if we take the Laplace transform and if
we rearrange then, we will get the expression like this b 1 a n S to the power n, a n minus
1 S to the power n minus 1, a n S plus a naught multiplied by f 1 bar S. Another
expression will be b 2 divided by a n S to the power n, a n minus 1 S to the power n

minus 1 a 1 S plus a naught f 2 bar S. In this discussion, you remember that y and f both
are the deviation variables.
So, this is the representation in S domain of the linear model or linearized non linear
model. So, we can write this equation in this form y bar S equal to G 1 S f 1 bar S plus G
2 S f 2 bar S, we can rearrange the equation, I mean the equation in Laplace domain by
this equation, here this part represent G 1 and this 1 is G 2.
In the next step, we need to develop the block diagram, what will be the block diagram
of this system. So, we have basically 2 transfer function one is G 1 another one is G 2, 2
inputs in Laplace domain one is f 1 S another one is f 2 S and single output that is y S.
(Refer Slide Time: 09:40)

So, one block for G 1 input to this G 1 is f 1 bar S another block, we need to use for
transfer function G 2 input to this block is f 2 bar S. Now the output of these 2 blocks,
we need to add suppose, we are using this symbol to add this 2 signals, one is this one
another one is this one. So, this output is basically f 1 G 1 and this is f 2 G 2. So, f 1 G 1
plus f 2 G 2 that is nothing, but y S agree f 1 multiplied by G 1 that is this signal and this
signal is basically multiplied by G 2. So, if we add this 2, we get y S.
So, this is the block diagram, of the multiple input single output example. Now, if we
consider n number of input variables, but, a single output variable what will be the block

diagram for that. If we consider n number of input variables f 1 f 2 up to f n and if we


consider only one output variable y, what will be the block diagram that you draw.
So, this is G 1, input to this G 1 is f 1, next 1 is G 2 input 2 G 2 is f 2 like this way last
term will be suppose, last transfer function is suppose G n S input to this is m f n S by
adding all these output, we get y S. So, this is for several input and single output, this is
the block diagram with block diagram for process with several inputs and one output or
single output.
Now, to find the transfer function of a process of a non linear process, we need to note 2
important points, one is if the model is non linear, we need to linearize that equation.
This is the first point second point is we need to represent the linearized equation in
terms of deviation variables.
So, these are 2 important points, which we need to follow to derive the transfer function,
one is if the process model is non linear, we need to linearize that equation, if that is
linearize that is, second point is we need to represent the linearized equation in terms of
deviation variables. Another question comes that is what is the use of this transfer
function? We will consider one single input, single output system.
(Refer Slide Time: 14:52)

So, for a S I S O system, we consider one process and input to this process is f t output to
this process is y t for this S I S O system, the block diagram is G S input is f bar S output
is y bar S.
Now, G S is known to us I mean, if we have the process model, we can easily derive the
transfer function for that process. So, G S is derived I mean, G S is known to us. So, G S
is suppose given now what we can do, we can give a change in input variable ultimate
answer is to observe the process dynamics, we use the transfer function fine by the use
the transfer function, we know the transient behavior or dynamic behavior of the process,
but how we can know that we have the process initially at steady state.
Now, we are introducing some change in input variable and that is represented by f t, we
are introducing some change in f t then we can easily transform this, f t to f S by taking
Laplace transform. Say for example, if we introduce a step change with magnitude a
suppose this is a step change with magnitude A, we can change it in S domain, how
much that will be A by S. In the next step, we can determined y bar S, how we can
determined y bar S y bar S is G S f bar S.
So, it is straight forward to determine y bar S, now we need to take the inverse of
Laplace transform. So, that we get y t, if we can determined y S then, we can get y t by
taking inverse of Laplace transform. So, we have a particular process, we are introducing
some change in input variable in terms of f t. Now, we can change to that input in S
domain represented by f bar S, in the next step, we can calculate y bar S, because y bar S
is equal to G S f bar S then by taking inverse of Laplace transform, we can know y t.
So, if we give a change in input variable f t, we can know the transient behavior of that
process in terms of y t. So, that is a use of transfer function. Now, you will consider
some chemical engineer examples and we will try to find the transfer functions first we
will consider a simple example that is liquid tank system.

(Refer Slide Time: 19:03)

So, we will derive the transfer function of a liquid tank system. So, this is a schematic of
a liquid tank system, a liquid is entering at a flow rate of F i and the output is coming out
at a flow rate of F naught. The liquid height is h cross sectional area of tank is suppose
A. This is a schematic of a liquid tank system, F i and F naught both are in terms of
volumetric flow rate.
So, F i and F naught are the volumetric flow rate, I hope, we have consider this example
previously. So, will discuss step wise, in the first step, we need to develop the model in
the first step, the model should be linear or linearized non linear model, if the model is
linear it is fine, if the model is non linear, we have to linearize that using Taylor series.
So, what is the model of this system, I mean what is the modeling equation, if d h d t
equal to F i minus F naught by assuming the liquid density constant. So, this is the
modeling equation for the example liquid tank system in step 2, we need to construct the
model in terms of deviation variables.
So, to construct the model in terms of deviation variables, we have to consider the model
at steady state. So, what will be the modeling equation at steady state condition, at steady
state, the modeling equation becomes F i S minus F naught S, we have just added suffix
S to represent the steady state anyway. This model is the linear equation or non linear
equation, this is the linear equation. So, there is no need of linearization.

So, this is the equation modeling equation at steady state condition. So, how we can
represent the modeling equation in terms of deviation variable, we need to subtract this
steady state equation from this dynamic equation. So, the model in terms of deviation
variables has this form A d h prime d h equal to F i prime minus F naught prime, here F i
prime is F i minus F i S, F naught prime is F naught minus F naught S and h prime is h
minus h S. So, this is the mathematical model in terms of deviation variables.
(Refer Slide Time: 24:03)

In step 3, we need to take Laplace transform and we need to derive the transfer function.
So, in step 3, we will determined, the transfer function. So, for that, we have to take the
Laplace transform, if we take the Laplace transform, we get A S h bar prime S equal to F
i bar prime S minus F naught bar prime S. This is the representation of the model in S
domain.
Now, if we rearrange this equation, we get h bar prime S equal to 1 by A S F i bar prime
S minus 1 by A S F naught bar prime S. So, this is the equation, which we got finally, in
Laplace domain or S domain in the next step, we need to make the block diagram based
on this equation. So, in step 4, we need to make the block diagram can you make the
block diagram based on this equation.
So, one transfer function is 1 by A S another transfer function is 1 by A S, you can
consider or you can consider minus 1 by A S. So, the block diagram will be 1 by A S
input to this block is F i bar prime S. The another transfer function is 1 by A S input is f

naught bar prime S. So, we will use here positive sign and here negative sign, output is h
bar prime S, we are just considering 1 by A S F i bar prime S minus 1 by A S F naught
bar prime S equals to h bar S, why negative sign is there, because F naught is the output
and F i is the input.
So, we can say that F naught is the negative input fine or you can do one thing, if you
want to consider minus 1 by A S as the transfer function then we can get this block
diagram. F i bar prime S and another transfer function is minus 1 by A S F naught bar
prime S, we are adding both of these output and we can get y bar S, this block diagram
also we can get. In the next, we will consider another example that is a C S T R.
(Refer Slide Time: 29:02)

So, we will derive the transfer function of an isothermal C S T R, previously we have


consider one C S T R, I mean, we have develop the model structure of a C S T R
previously and the model structure of that C S T R includes 3 equations, can you
remember this 3 equations.
One is based on mass balance, I mean total mass balance, one equation we got based on
total mass balance, second equation we got based on component mass balance and
another equation, I mean third equation we got based on energy balance, now isothermal
means constant temperature. So, for the case of isothermal C S T R the operation, I mean
the reactor operates at constant temperature. So, there is no variation of temperature, if

that is the case then we have considered one C S T R and we derived modeling equations
based on these 3 balances. So, what is that C S T R.
Yes that is a non-isothermal C S T R, because temperature varies in that C S T R with
time. So, the C S T R, which we have modeled earlier that is the non isothermal C S T R,
so for a non-isothermal C S T R the energy balance equation, we usually derive. Now
here will consider the isothermal C S T R total mass balance, there will not be any total
mass balance equations. How we can avoid this, if we assume the reaction mixture
volume constant, I mean, if we consider the volume of the reaction mixture constant in
that case, we can avoid total mass balance equation. So, we are assuming that.
Second is component mass balance, we will consider fine and third one is energy balance
energy balance does not exist for isothermal case. So, presently we will consider the
considering the isothermal C S T R that includes only the component mass balance
equation.
(Refer Slide Time: 32:58)

Now, the schematic representation of an isothermal C S T R is like this inlet flow rate is
F i concentration is C A i, the output is coming out with a flow rate of F and A
concentration of C A. So, the concentration in the vessel is also C A and will assume the
value is V, since there is no variation of temperature. So, there is no need of any external
heating or cooling medium.

So, what will be the component balance equation then F i and F both are volumetric flow
rates, this is a schematic representation of an isothermal reactor, what will be the
component balance equation. Component mass balance equation previously, we got this
equation d C A d t equal to F i divided by V C A i minus C A minus K naught
exponential of minus E R T into C A. This equation, we derived for non isothermal C S
T R, how we can represent this equation for the case of isothermal C S T R.
See in this is basically, K is not it, this whole is basically K and in this term, you see
temperature is there, since the case is non isothermal, if we consider the isothermal C S T
R, the temperature is constant. So, this whole term becomes constant. So, we can rewrite
this equation for the case of isothermal C A i minus C A minus K C A for the case of
isothermal, there is no variation of temperature, that is way this whole terms becomes
constant and that, we are representing by K.
So, this is the modeling equation for the isothermal C S T R, now we need to represent
this equation in terms of deviation variables anyway before that, we can simplify this
equation like this is out modeling equation d C A d t equal to F i divided by V C A i
minus C A into K C A. Now, we can write this equation in this way.
(Refer Slide Time: 36:53)

D C A d t plus F i divided by V plus K multiplied by C A equal to F i divided by V


multiplied by C A i. It gives d C A d t plus 1 by tau plus K multiplied by C A equal to 1
by tau C A i where, tau is V divided by F i tau is V divided by F i. Next, we will

represent this equation in terms of deviation variables to find the transfer function, you
directly write the equation in terms of deviation variables.
(Refer Slide Time: 38:10)

Write the equation in terms of deviation or perturbation variables d C A prime d t plus 1


by tau plus K multiplied by C A prime equal to 1 by tau C A i prime. This will be the
model equation in terms of deviation variables here, C A prime is C A minus C A S and
C A i prime is C A i minus C A i S to get the transfer function, we need to take Laplace
transform, take Laplace transform of this equation. If, we take the Laplace transform of
this equation, we get S C A bar prime S plus 1 by tau plus K C A bar prime S equal to 1
by tau C A i bar prime S.
Now, the transfer function represented by G S equal to output divided by input equal to 1
by tau S plus K plus 1 by tau. This is the final representation of the transfer function for
the example isothermal C S T R, this is the transfer function. Now, the block diagram is
quite straight forward.

(Refer Slide Time: 40:30)

We can directly write G S equal to 1 by tau divided by S plus K plus 1 by tau input is C
A i bar S output is C A bar prime S. So, this block diagram is represented in terms of
concentration not in terms of flow rates, because we have already assume the volume of
the reactor is constant. So, this is the input concentration C A i and this is the output
concentration. So, this is the block diagram for the example C S T R.
Now, So, for we have discussed the transfer function of single output although, we have
consider multiple input, but our discussion is restrict to the single output. So, in the next,
we will discuss with the consideration of multiple outputs. So, we will consider multiple
input and multiple output systems.

(Refer Slide Time: 42:03)

So, transfer function of a multi input multi output system. Now, the process, we will
represent in this way, this is the process suppose, 1 input is f 1, another input is f 2, they
are in time domain and we will consider 2 outputs, one is y 1 another one is y 2. So, in
this M I M O system, I mean multiple input multiple output system, basically we are
considering 2 by 2 system, 2 input 2 outputs.
Now, first we need to write the input output model for this process. So, input output
model is d y 1 d t a 1 1 y 1 plus a 1 2 y 2 plus b 1 1 f 1 plus b 1 2 f 2, this is our linear
equation and this equation is representing the variation of y 1.
Similarly, for the second output y 2, we have the linear model represented by this
equation d y to d t equal to a 2 1 y 1 plus a 2 2 y 2 plus b 2 1 f 1 plus b 2 2 f 2. So, this is
the model of the 2 by 2 system, I mean M I M O system, we can represent this 2
modeling equations in matrix form and that is represented in this way equal to a 1 1 a 1 2
a 2 1 a 2 2 y 1 y 2 and another matrix for input b 1 1 b 1 2 b 2 1 b 2 2 multiplied by f 1 f
2. Just, we have representing this 2 modeling equations in this matrix form.
Now, here y 1 and y 2 both are perturbation variables or deviation variables. So, y 1 and
y 2 both are deviation variables. Now, if we take the Laplace transform of this 2
equations and if we rearrange, we get finally, the equations, which I am writing next, I
am writing the final expressions.

(Refer Slide Time: 46:18)

So, if we take Laplace transform, taking Laplace transform and rearranging, we get y 1
bar S equal to S minus a 2 2 b 1 1 plus a 1 2 b 2 1 divided by P S is a polynomial, we
will write that expression for P S later.
So, this is the first right hand term and another term related to f 2 that is S minus a 2 2 b
1 2 plus a 1 2 b 2 2 divided by P S multiplied by f 2 bar S. This is a final expression for y
1, similar expression for another output y 2, we get that is y 2 S equal to S minus a 1 1 b
2 1 plus a 2 1 b 1 1 divided by P S, f 1 S plus S minus a 1 1 b 2 2 plus a 2 1 b 1 2 divided
by P S f 2 bar S. These are the 2 expressions and a expression for the characteristics
polynomial P S is like this the P S is the characteristics polynomial represented by a 1 1 a
2 2 S minus a 1 2 a 2 1 minus a 1 1 a 2 2, this P S is a characteristic polynomial.
Now, you see the output is output equal to 1 transfer function multiplied by first input
variable another transfer function multiplied by second input variable. Similarly, for the
case of second output one transfer function multiplied by first input variable second
transfer function multiplied by second input variable.
So, we can write y 1 S equal to G 1 1 S f 1 bar S plus G 1 2 S f 2 bar S, similarly if y 2
bar S equal to G 2 1 S f 1 bar S plus G 2 2 S f 2 bar S. So, we can say that this term is
basically G 1 1, this is another transfer function related to first output that is G 1 2, this
transfer function is involved with y 2 that is G 2 1 and this is G 2 2. So, 4 transfer

functions is involved and the final expressions, we have written here and we can
represent them in matrix form. So, if we represent then matrix form.
(Refer Slide Time: 51:40)

We will get y 1 bar S y 2 bar S equal to G 1 1 G 1 2 G 2 1 G 2 2 another matrix is for 2


input variables f 1 and f 2. So, this is about the derivational transfer function for the M I
M O system. In the next step, we need to make the block diagram, what will be the block
diagram for this system.
(Refer Slide Time: 52:26)

See there are 4 transfer functions, one is G 1 1, second one is suppose G 2 1, third one is
G 1 2 and fourth one is G 2 2, basically the transfer function G has 2 suffix one is i
another one is j, i represents output and another suffix j represents input. So, this is the
nomenclature of the transfer function.
So, G 1 1 this first one means output. So, this is the output is y 1 second suffix 1, that
indicates the input and first input agree, similarly for the last transfer function, first suffix
to indicates the output. So, output will be y 2 second suffix 2, that indicates input, second
input is basically f 2 bar S.
So, how we can add the input and output with this G 2 1, you see G 2 1 first suffix is 2.
So, output second output. So, this is the second output, second suffix indicates the input
that is first input that means f 1. So, we will add f 1, similarly for G 1 2 first suffix is 1
that means, y 1 and second suffix is 2 that indicates second input that means, f 2. So, this
is the block diagram of the multi input multi output system. in the next class, we will
discuss about the quantitative analysis of this transfer function.
Thank you.

Process Control and Instrumentation


Prof. A. K. Jana
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture - 9
Dynamic Behavior of Chemical Processes (Contd.)
(Refer Slide Time: 01:03)

We will start the poles and zeros of a system. So, topic is poles and zeros of a system or
you can say transform function. Suppose, this is a blog diagram for if single input, single
output system G s input to this process is f bar s and output of this process is y bar s. So,
we can write this in this form y bar s divided by f bar s equal to G s. G s is the transform
function we know, now the transform function G s may be, the ratio of 2 polynomials,
one polynomial is Q s another polynomial is P s.
Now, what is the zeros of a transform function, zeros the roots of the polynomial Q s are
the zeros of the transform function or the system. If Q s is the polynomial, then the roots
of the Q s are the zeros of the system, roots of the polynomial Q s are called zeros of a
system or zeros of a transform function, got it is a very simple Q s is a polynomial now,
we need to determined the roots of that polynomial and those roots are zeros of that
system or transform function.

In the similar fashion, roots of the transform function P s are called poles of a system or
poles of a transform function. Now, we will take one simple example, to explain the
concept of poles and zeros.
(Refer Slide Time: 04:32)

So, will take next one example will consider the equation in s domain that is s a G 1 s f 1
bar s, this is the representation of the model for a SISO system in Laplace domain, Hoyer
G 1 is given as k divided by s plus a. You can write this as Q s divided by P s, but the
numerator is basically constant. So, that is not the polynomial of s; that means, there is
no zeros. So, for this example system, there is no zeros fine and what about P s, P s is s
plus a.
So, there is pole at s equals to minus a understood there is no zeros for the example
system and there is a pole at s equals to minus a. Now, one point I want to mention that
is, at the poles of a system, the transfer function becomes what, at the poles of a system
transform function becomes infinity. Our pole, is the pole for the system is at s equals to
minus a, if we substitute in this equation s as minus a the transform becomes infinity,
lesson it, if we put s equals to minus a then G s becomes infinity, but the similar way we
can say at the zeros of a system the transform function becomes 0. So, at the zeros of a
system transform function becomes 0. So, these are the concept of poles and zeros, in the
next we will discuss with the general form of transform function.

(Refer Slide Time: 07:54)

We will discuss the, general form of transform function. So, and nth order system, we
will represent by a linear ordinary deferential equation I mean process will represent by
an nth order ordinary differential equation in linear form. So, that has the form a n d n y
d t n plus a n minus 1 d n minus y d t n minus 1, like this a 1 d y d t plus a not y equal to
b m d m f d t m plus b m minus 1 differentiation. So, this is d t d 1 d f d t plus b not f this
is a general representation of a linear process.
Hoyer all the a and b are constant coefficient and y f both are deviation variables, a and b
these are constant coefficient y and f both are deviation variables. Now, if we take
Laplace transform and rearrange, we get the transform function and that has the form of
G s equal to y bar s by f bar s equal to i equal to 0 to m b suffix i and s to the power i.
Similarly, here i equals to 0 to n a suffix i s to the power i and we get b m s to the power
m plus b m minus 1 s to the power m minus 1 last term is b not a n s to the power n a n
minus 1 s to the power n minus 1 last term is a naught.
Now, if we take common b m divided by a n we get s minus z 1 s minus z 2 s minus z m
hole divided by s minus p 1 s minus p 2 s minus p n. This is a general form of transform
function, here all z represent the zeros and p represents the poles.

(Refer Slide Time: 12:37)

Z i denotes the zeros and p i represents the poles. Now, what about the order n and m, I
mean n will be greater than m or equal to n or less than m for all physically meaning full
system n should be greater than or equal to m, for all physically mining full system n
should be greater than or equal to m, physically realizable systems, this term is used in
process control, physically realizable systems.
I mean all physically meaning full system n should be greater than or equals to m you
can test it, you consider n less m and you check what type of response, you are getting
like for example, a step change in input variable, produces and infinites spike in the
output. If you consider suppose n equals to 0 and m equals to 1, then you give a step
change in input variable f you will see that, there is an infinites spike in the output, for
this reason for any physically meaning full system, we have to consider n greater than or
equals to m. So, next we will discuss transform function and it is analysis.

(Refer Slide Time: 14:51)

Transform function and it is analysis. One important point I mean the use the transform
function we have discussed, that is to observe the transient behavior of a process, we can
use the transform function. Another important I mean the use of transform function is
the, stability analyses, to know the stability of a system, we can use the transform
function that we will discuss in the next.
So, we have mentioned earlier that, this Q s divided by P s this is basically, the
representation of a transform function I mean transform function is the ratio of 2
polynomials, 1 is Q s divided by P s. Now, we will take one general form of this
transform function, that is represented by this equation Q s divided by s minus p 1 s
minus p 2 s minus p 3 hole to the power m s minus p 4 s minus p 4 star s minus p 5, this
is the general representation of a transform function.
Now, s is basically if variable, which is defined in the complex plain, in this o a I mean
this is the form of s, s equals a plus j b, s is defined in the complex plain by this form.
Now, considering this general form of a transform function we will discuss, different
types of poles, for example one system which is 2 poles, they both are real, but distinct,
in another case we will consider a system which has multiple poles and all the poles are
real, what about the stability of that system.
One system which has 2 complex conjugate poles, what about the stability of that
system, like this different situations, so we will consider and all these poles are included

in this general form. So, first we will consider distinct real poles, so in first case will
consider, real, distinct poles. Now, to consider this case we will reduce the general form
of the transform function, to this expression G s equal to any way, we can take the partial
fraction of this equation, before considering the different cases.
So, we can write this equation in this way. So, c 1 s minus p 1 plus c 2 s minus p 2 plus c
3 1 s minus P 3 plus c 3 2 s minus p 3 hole squire, like this way c 3 m s minus p 3 hole to
the power m plus c 4 s minus p 4 plus c 4 star s minus p 4 star plus c 5 divided by s
minus p 5. This is the partial fractions of this transform function, general form of the
transform function.
(Refer Slide Time: 20:01)

I think I am writing again that G s equals to c 1 s minus p 1 c 2 s minus p 2 plus c 3 1 s


minus p 3 c 3 2 s minus p 3 hole squire last term is c 3 m s minus p 3 hole to the power
m plus c 4 s minus p 4 c 4 star s minus p 4 star plus c 5 s minus p 5 I have written that
general form in this equation. Next we will consider the first case that is real distinct
poles. So, in case one we consider real, distinct poles. In this case we will consider just
the 2 poles, both are real and both are distinct.
Those 2 poles are one is p 1 another one is p 2. So, the general form of the transform
function is reduces to G s equals to c 1 by s minus p 1 plus c 2 by s minus p 2 can I write
this, we are presently considering a system which has 2 distinct real poles and this poles

are p 1 and p 2. So, we are considering just this part of the transform function, I have
mention another thing that the s is define in the complex plane, by s equals to a plus j b.
(Refer Slide Time: 22:11)

So, we can represent this 2 poles, in the complex plane in this way, suppose this is real
axis this is imaginary axis. So, both the poles are real and distinct. So, one pole we will
consider in this side, I mean the left hand side of this imaginary axis, another pole we
will consider right hand side of this imaginary axis. One pole of in the left of imaginary
axis and another pole P 2 is in right of the imaginary axis.
(Refer Slide Time: 22:57)

And the transform function is represented by this equation G s equals to c 1 by s minus p


1 plus c 2 by s minus p 2. Now, if we take the inverse of Laplace transform what we get,
inverse of Laplace transform we get c 1 exponential of p 1 t plus c 2 exponential of p 2 t,
if we take the inverse of Laplace transform, we get this expression. Now, if p 1 is greater
than 0, p 1 we have consider in this complex plane sorry this p 1 is less than 0 because,
that is in the left of imaginary axis.
So, p 1 is less than 0. If that is the case, than c 1 e to the power p 1 t I mean exponential
of p 1 t tens to 0 as time tens to infinity, can we write this, we have consider p 1 less than
0. So, exponential of p 1 t multiplied by c 1 tens to 0 as time goes to infinity, we can
represent this graphically also, it is like this going to 0 this is t this is c 1 exponential of p
1 t. In another case, we have considered p 2 that is greater than 0 then c 2 exponential of
p 2 t tens to infinity as t tens to infinity agree, if p 2 is greater than zero then c 2
exponential p 2 t goes to I mean grows 2 infinity as time tens to infinity.
So, we can say that if a system has the pole in the left of imaginary axis, then the system
is decaying to the 0; that means, the system is stable and if the pole is greater than, 0 in
that case it is growing exponentially towards infinity; that means, the system is becoming
unstable. In the second we will consider, the multiple real poles.
(Refer Slide Time: 26:18)

In the second case we will considered multiple real poles, that pole is here p 3 and that is
repeated m times. I mean we will consider multiple poles represented by p 3. So, first we

will locate that p 3 in this axis, complex plane. So, suppose 1 p 3 is here, this is 1 p 3
another p 3 is suppose, here like this way p 3 is repeated m times. And for this multiple
real pole system we will consider the transform function represented by c 3 1 s minus p 3
c 3 2 s minus p 3 hole squire, like this last term is c 3 1 s minus p 3 hole to the power m.
(Refer Slide Time: 27:45)

If you see the general expression of the transform function that is this, this is the general
expression of the transform function. Now, to consider the multiple real roots we are
considering this part, as the transform function that I have return there.
(Refer Slide Time: 28:07)

So, G s equal to c 3 1 by s minus p 3 c 3 2 by s minus p 3 hole squire, like this way this is
c 3 m divided by s minus p 3 hole to the power m. In the similar way we will take the
inverse of Laplace transform G s. So, what will get c 3 1 plus c 3 2 divided by factorial 1
multiplied by t plus c 3 3 divided by factorial 2 multiplied by t squire, like this way c 3 m
divided by factorial m minus 1 multiplied by t to the power m minus 1 this will be
multiplied with exponential of p 3 t.
If take the Laplace inverse of Laplace transform of this transform function, we get this.
Suppose, I am writing this as W multiplied by e to the power p 3 t I mean this hole part
this 1, we are representing by W.
(Refer Slide Time: 30:01)

So, next we will consider p 3 greater than 0 then exponential of p 3 t grows to infinity
agree, as t tens to infinity. So, what about W multiplied by p to the power p 3 t tens to
infinity; that means, again in stability is there because, the system response is going to
infinity, in another case we will consider p 3 less than 0, then exponential of p 3 t tens to
0 as t tens to infinity. So, what about this W multiplied by exponential of p 3 t it will d k
to 0.
And in the third case, if we consider p 3 equal to 0 exponential of p 3 t becomes unity,
exponential of p 3 t becomes 1 as t tens to infinity. So, what about this W multiplied by
exponential of p 3 t grows to infinity, due to this W grows to infinity as t tens to infinity.
So, due to this term, we are getting this W multiplied by exponential of p 3 t becomes

infinity. So, these are basically the different real roots, associated with a process and the
stability of that process. Now, in the next case we will consider the complex conjugate
poles, that is the third case.
(Refer Slide Time: 32:56)

In the third case we will consider, complex conjugate poles. So, for the complex
conjugate poles, will consider the third portion of the general form of transform function.
The general form of transform function I am again writing that c 1 s minus p 1 plus c 2 s
minus p 2 this 2 written terms we have considered for the 2 distinct real poles. Next term
we have consider, for the multiple real poles hole squire 3 m s minus p 3 hole to the
power m .
So, now, we will consider this part c 4 divided by s minus p 4 and c 4 star divided by s
minus p 4 star and in the 4th case we will consider c 5 divided by s minus p 5 this we
have already considered for the distinct to real poles, this part we have considered for
multiple real poles now, we will consider this part I mean the transform function, reduces
to c 4 s minus p 4 plus c 4 star s minus p 4 star. Now, the 2 complex conjugate poles are
p 4 and p 4 star.
So, will consider p 4 equals to alpha plus j beta and p 4 star will consider that is alpha
minus j beta. These are the 2 complex poles. Now, we have to represent them in the
complex plane.

(Refer Slide Time: 36:00)

This is the complex plane, this is real axis, this is imaginary axis suppose, one pole that
is p 4 and another pole that is here p 4 star. So, this is alpha plus j beta and another one is
alpha minus j beta. Now, as usual we need to take the inverse of Laplace transform. If we
take the inverse of Laplace transform, the right hand term includes one exponential term
exponential of alpha t and one sin function, that is beta t plus 5. The right hand term
includes one exponential function, exponential of alpha t and another one is sin sign
function. Now, this right term may be multiplied with some value suppose W. So, next
we will consider, the different situations I mean different values of alpha, then what will
be the response of the process having 2 complex conjugate poles.

(Refer Slide Time: 38:03)

So, in the first case we will consider alpha greater than 0, if alpha greater than 0 then
exponential of alpha t tens to infinity as time tens to infinity agree, then what will be
exponential of alpha t multiplied by sign beta t plus phi. It grows to infinity, in
oscillating manner, it grows to infinity in an oscillating manner, it is going to infinity in
oscillating manner because, of this semi schedule function, so how we can represent the
system response graphically.
So, we represent in this form suppose, this is time and this is output function, this is
output initially like this. So, initially the process was at steady state. Now, we have
introduce some input change, then the output response in the manner, it is going to wards
infinity. So, it is obvious that if alpha is greater than 0 the system has instability problem.
Similarly, we will consider alpha less than 0.

(Refer Slide Time: 48:34)

In the next case we will consider alpha less than 0, then e to the exponential of alpha t
tens to 0 as time tens to infinity. So, exponential of alpha t sin beta t plus phi decays to 0
in an oscillating manner. So, it decays to 0 in an oscillating manner, the graphical
representation is like this, this is the time, this is output. So, it is gradually stabilizing I
mean the process is coming back to the original state.
And next we will consider alpha equal to 0, if alpha equal to 0 then e to the exponential
of alpha t equal to 1 as time tens to infinity, so x exponential of alpha t multiplied by sin
beta t plus phi, what will be the response.
Student: ((Refer Time: 42:51))
Sin a shodel with yes sin a shodel I mean in constant amplitude. So, we can write
oscillates continuously with constant amplitude.

(Refer Slide Time: 43:38)

This is sometimes called marginally stable and for this case graphical representation is,
after this is time, this is output like this, this is a case for alpha equal to 0. So, we have
considered 3 different cases now, another case I mean the last case in which we will
consider.
(Refer Slide Time: 44:15)

This transform function. This is a first case, second case, this is a third case complex
conjugate poles and this is a 4th case, in the 4th case we will consider the pole is at the
origin.

(Refer Slide Time: 44:43)

So, in the 4th case, we will consider poles at the origin I mean if we draw the complex
plane this is real and this is imaginary the pole p 5 is present here, this is p 5; that means,
p 5 equal to 0 or you can write in this form 0 then transform function G s reduces to c 5 s
minus p 5 this is p 5 equal to c 5 divided by s. Now, we will usually take the inverse of
the Laplace transform, then we get c 5 1 constant term.
So, these are the 4 different cases which we have considered hear and we have conclude,
we can conclude based on this 4 observations I mean considering the 4 cases that, a
system is set to be stable, if all the poles lie in the left of imaginary axis, this is our
conclusion based on the 4 observations. A system is stable, if all the poles of it is
transform function lie in the left of imaginary axis. And in the subsequent chapters, we
will study more about the stability, considering different techniques. Now, we will take
one simple example, one simple transform function which we have considered earlier.

(Refer Slide Time: 48:00)

One simple transform function we will consider, that we have discussed earlier that is G
s equals to k divided by s plus a. So, this is a stable system or unstable system, stable
system why this is a stable system.
Student: ((Refer Time: 48:37))
Yes because, s is equal to minus a. So, if a is negative, if a is negative.
Student: Not stable.
Not stable. So, here one pole exists at s equals to minus a. So, we can say the system
having the transform function of G s is stable only if a is positive. If a is negative this is
unstable because, in this case the pole lie in the left side of the imaginary axis and in this
case, the pole lie in the right side of the imaginary axis. Another example we can
consider that is the, liquid level system for the case of liquid level system, we got the
transform function that is 1 divided by A s f i bar prime s minus 1 divided by f s f naught
bar prime s.
And this model we got considering, this system f i and f naught. A is the cross section
gradient, h is the height. Now, we can do one thing, we can consider f is pronominal to h.
We are just trying to modify this example. Now, f naught equals to suppose alpha
multiplied by h.

(Refer Slide Time: 51:24)

Now, if we substitute this in the model equation then what will get A d h d t equal t o f i
minus f naught equal to f i minus alpha into h. So, if A d h d t plus alpha h equal to f i. If
we write in terms of deviation variables we will get alpha h prime equals to f i prime. So,
what will be the transform function for this, h bar prime s divided by f i bar prime s equal
to 1 divided by s A plus alpha, can I write this 1 divided by s a plus alpha agree or not
agree.
So, what is the root I mean what is the pole of this equation.
Student: ((Refer Time: 52:41))
Minus alpha divided by A. So, the system is stable or unstable, the system is stable, area
is always positive. So, if we consider alpha is positive then the system is stable.
Thank you.

Process Control and Instrumentation


Prof. A. K. Jana
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture - 10
Dynamic Behavior of Chemical Processes (Contd.)
(Refer Slide Time: 01:00)

Today, we will study the dynamics of first order systems. So, our topic is dynamics of
first order systems. So, before going to discuss the dynamics of first order system, we
will know what is first order system, what is first order system? First order system is one
whose output y t is modeled by first order differential equations. Suppose, we have a
process, this is the input to the process and this is the output to the process. Input is
represented by f and output is say by y. Now, first order system is one whose output y,
whose output y is modeled by first order differential equation.
Now, we will consider the linear case or if non-linear equation exists, we need to
linearize that equation. So, if we consider the linear case, then first order differential
equation we can write, by this form a 1 d y d t plus a naught y equal to b f t definitely y is
also function of time. Now, in this equation a and b both are constant coefficients, a and
b are the constants. F is the forcing function or input f is the input and y is the output.
Now, if we rearrange this equation, then we get a 1 divided by a naught d y d t plus y
equal to b divided by a naught function of t. Now, before this I want to mention one

thing that we will consider two cases, in one case a 1 equal to 0 and in another case a 1
not is equal to 0.
So, in case 1, we will consider a 1 not is equal to 0, first we will discuss a 1 not is equal
to 0 then in the next case we will consider a 1 equal to 0. Now, after rearranging this first
order differential equation, linear equation we get this. Now, we will represent a 1 by a
naught, by the term tau p and another term that is b divided by a naught will be
represented by k p. Then the linearized form of first order differential equation for the
case of a 1 not is equal to 0 becomes, tau p d y d t plus y equal to k p f t.
Now, this tau p is called time constant of the process, definitely we are discussing first
order process. So, this is a time constant of first order process and k p is called steady
state gain or static gain or only gain or I mean gain of the process. Now, what is gain,
gain is basically the change in output, per unit change in input, gain is the change in
output per unit change in input I mean gain equal to change in output divided by change
in input. Since, we have considered this as static gain.
So, this delta output divided by delta input, they should be considered at steady state
only. So, we know the modeling equation a 1 by a naught d y d t plus y equals to b by a
naught f t. So, what we will be the steady state form of this, steady state form will be if
we consider the first equation I mean a one d y d t plus a naught y equal to b f t this is a
modeling equation.
So, what will be the steady state form of this a naught y s equal to b f s agree because,
this d y d t term does not exist. Now, if we rearrange this y s by f s equal to b by a
naught, agree if we rearrange this equation a naught y s equals to b f s y s divided by f s
becomes b by a naught. So, what is y s by f s that is the gain, output by input change, that
is why we have represented b divided by a naught as steady state gain. And that we have
represented by k p.

(Refer Slide Time: 08:54)

Now, we will go back to the first equation. We got tau p is d y d t plus y equal to k p f t
here, y and f both are deviation variables, that is why we have written here this is as delta
output and this is as delta input. So, y and f both these variables are deviation variables;
that means, y naught equal to f naught equal to 0, can we write this, if y and f both are
deviation variables then we can write y naught f naught both are 0. Now, we will take
Laplace transform of this Laplace transform.
Then we get tau p s y bar s plus y bar s equal to k p f bar s, if we take Laplace transform
of this, then we get this form. Now, rearranging the above equation we get y bar s by f
bar s equal to k p divided by tau p s plus 1, y bar s divided by f bar s becomes k p divided
by tau p s plus 1 and this is nothing, but the transfer function of the first order system.
We will represent this by G s, transfer function of the first order system.
In the next and this is also called as first order lag or exponential transfer lag. We will
see later that, this transfer function involves phase lag, that is why this is first order lag
transfer function, we will consider that later. So, in the first case we have considered a
naught not is equal to 0, in the next case we will consider a naught equal to 0.

(Refer Slide Time: 12:31)

So, in case 2 we will consider a naught equal to 0. Our original equation is a 1 d y d t


plus a naught y equal to b f t, this is original equation. Now, if we consider a naught
equal to 0 then this becomes a 1 d y d t equal to b f t; that means, d y d t equal to b by a 1
f t this is equal to k p prime f t. Here, gain is represented by k p prime and that is b by a 1
k p prime is here, b by a 1. Now, if we take Laplace transform again, then s y bar s equal
to k p prime f bar s, rearranging this form we get the transfer function G s equal to y bar
s divided by f bar s equal to k p prime divided by s.
So, this is the transfer function of a first order system, when we consider a naught equals
to 0. Now, this type of processes are called purely capacitive or pure integrator. For the
example first order system with a naught equal to 0, we got the transfer function this and
this type of processes are called pure integrator, we will discuss with physical example
what type of processes are called pure integrator processes. Now, these are about the
transfer functions in general form, in the next we will discuss with some example.

(Refer Slide Time: 15:47)

We will first consider one example that is a liquid tank system. Will consider a example
that is a liquid tank system. This is a liquid tank height of this liquid is h, cross sectional
area of the tank is suppose A, input is entering the tank with a flow rate of f i and output
is f and here, we are considering resistance that is R. Now, this effluent flow rate f is
related to the hydrostatic pressure, of the liquid height h through resistance R, here R is
the resistance to flow.
This f is linearly related, to the hydrostatic pressure of the liquid level h, through
resistance R. Effluent stream, effluent outlet we can represent by this form f equal to
driving force, for flow divided resistance to flow. It is quite common in heat transfer
also, like q equals to delta t divided by r we have started in our heat transfer course, rate
of heat transfer equals to driving force that is temperature difference, divided by the
resistance.
Similarly, we are writing also the same similar equation, similar a relation that is f equal
to driving force is here, h and resistance is R. So, in the modeling equation we will
consider f equal to h divided by R. What is the modeling equation of this liquid level
system, we have derived the modeling equation for this liquid level system earlier, that is
A d h d t equal to f i minus f. F i and f both are volumetric flow rates. Now, if we
rearrange this equation, we get A dh d t plus f equal to f I, if we rearrange the modeling
equation we get this. Now, we will substitute f equals to h by R.

(Refer Slide Time: 20:14)

Then we get A d h d t plus h divided by R equal to f i. Multiplying both sides by R we


get A r d h d t plus h equal to R f i, multiplying both sides by r we get this equation.
Now, we need to represent this equation in terms of deviation variable. So, for that we
need to consider the equation as steady state, what will be the form of this model at
steady state, it will be h s equal to R f is at steady state condition the above modeling
equation I mean this modeling equation becomes h s equal to R f is we are using suffix s
to represent the steady state.
Now, subtracting this steady state model from this modeling equation, we get the
equation in terms of deviation variable, that is A R dh prime d t plus h prime equal to R f
i prime. Subtracting steady state model from this original model, we get this form in
terms of deviation variables, where h prime equal to h minus h s and f i prime is f i minus
f is this 2 deviation variables are involve.
Now, we can replace A R by the time constant, time constant we have represented by tau
p and here, tau p equal to A R. So, what is the unit of this tau p, can you find the unit of
this tau p, unit of this tau p is.
Student: Time.
Time. Similarly, we will consider k p, k p is here R. So, that will be the unit of R
basically, what is the unit of R, R equals to h by R; that means, time per meter square or

time per area we can say. Now, we will just replace this A R term here by tau p and R by
this k p tau, then we get tau p d h prime d t plus h prime equal to k p f i prime. Again we
will take Laplace transform of this, then we get tau p is h bar prime s plus h bar prime s
equal to k p f i bar prime s.
(Refer Slide Time: 24:27)

Now, if we rearrange then finally, we get h bar prime s divided by f i bar prime s equal to
k p divided by tau p s plus 1. So, here k p is basically the resistance tau p is A R S plus 1,
this is the transfer function of this system, liquid level system.
Now, here you see one thing tau p we have considered that is time constant equal to A
multiplied by R. Now, we can write this as, storage capacitance, multiplied by resistance
to flow. Now, that capacity to store mass, is measured by the help of A the capacity to
store, mass is measured by the help of area A and this is; obviously, resistance R.
Similarly, if we consider one example, which includes only the energy balance equation.
In that case, this storage capacitance is basically the capacity to store energy. If we
consider an example, which is modeled by only energy balance equation, in that case
also we can write, this time constant tau p equal to storage capacitance multiplied by
resistance to heat, then in that case this is the capacity to store energy not mass. In the
next we will take another example, that is the pure capacity system.

(Refer Slide Time: 27:16)

We have consider another example, that is the example of pure capacitive system I mean
how we can get the A naught y term equal to 0 physically, we will continue the liquid
tank system, which we considered earlier. So, this is a liquid tank system height is h, area
is A inlet flow rate is f i in the outlet section one pump is installed, which is delivering
outlet with a flow rate of f.
In the outlet section one pump is installed, which delivers the outlet or effluent stream
with a flow rate of f and this is a constant displacement pump, it means the flow rate is
constant I mean f is constant. So, we have already considered the model of this system in
the previous example, that is if A d h d t equal to f i minus f, this is the modeling
equation for the liquid tank system and this is a linear equation. So, there is no need of
linearization.
Now, this is a modeling equation, what will be the equation at steady state. At steady
state we get d h s d t which is nothing, but 0 equal to f is minus f. I told that for the case
of constant displacement pump f is a constant or f remains constant. Now, subtracting
this steady state model from the actual process model, we get the equation in terms of
deviation variables, that is this subtracting the steady state model from the actual process
model we get this modeling equation in terms of deviation variables.
Now, what will be the transfer function of this, if we take Laplace transform again A S h
bar prime s equal to this. So, h bar prime s divided by f i bar prime s equal to 1 divide by

A S; that means, 1 divided by A whole divided by s. If we write in this form k p prime by


s then here k p prime is 1 by A, this is one example of pure capacitive process.
Next we will discuss, the dynamic behavior of first order system. Previously it was told
that the transfer function is used to observe the transient behavior of a process. Now, that
we will discuss by the use of transfer function, how we can know the transient behavior
of a process, that we will discuss in the next. So, first we will consider the dynamics of a
first order system, considering input as ramp input.
(Refer Slide Time: 32:36)

So, next topic is dynamic behavior and we will consider first order process and we will
change the input as ramp function, I mean the input is ramp input. This we will discuss in
the next. So, we have already derived the transfer function of a first order system in
general form, that is G s equal to y bar s divided by input that is k p divided by tau p s
plus 1 this is the transfer function of a first order system.
Now, to observe the transient behavior we need to introduce some change in the input, to
observe the transient behavior of a process we need to introduce some change in input
variable. Now, here we will consider ramp input. So, f is the forcing function, f is a input
to this process, we will consider here f t equal to A t and this is a ramp function. Now, in
Laplace domain in s domain, we can write this as A by s square, this is the form of ramp
function in s domain.

Now, our transfer function is y s divided by f bar s equal to k p divided by tau p s plus 1.
If we consider f s equals to A by s square, then the output in s domain becomes k p A
divided by s square tau p s plus 1, got it considering f s equal to A divided by s square.
Now, we will just write this form in this way c 1 divided by tau p s plus 1 plus c 2
divided by s plus c 3 divided by s square.
(Refer Slide Time: 36:30)

And, finally we will get y bar s equal to k p A tau p square divided by tau p s plus 1 and
next term k p A tau p divided by s third term involving c 3 becomes k p A divided by s
square, after introducing ramp input to the process, we get this as the output in terms of
in Laplace domain. Now, if we inverse, if we consider the inverse of Laplace transform.
We get y t equal to k p A tau p exponential of minus t divided by tau p plus t divided by
tau p minus 1. Taking inverse of Laplace transform, we get this form, in time domain. If
we rearrange this equation, we get k p A tau p exponential of minus t divided by tau p
plus k p A t minus tau p, rearranging this equation we get this form. Now, we will
consider one case, that is t tends to infinity. If we consider t tends to infinity, then y t
becomes k p A t minus tau p; that means, we can say that the ramp input, yields a ramp
output with slope k p A this is the slope.

(Refer Slide Time: 40:03)

And if we represent this graphically then we get this plot, this is time t, this is y t, this is
y t equal to A k p t minus tau p, this is the tau I mean tau p this is 0. So, we can say that
the ramp input as I told yields, a ramp output with the slope of k p A. Next we will
consider the purely capacitive process, dynamic behavior of a pure capacitive process.
This is the dynamic behavior of the first order system considering ramp input, next we
will discuss dynamic behavior of a pure capacitive process considering step input.
(Refer Slide Time: 41:58)

So, dynamic behavior pure capacitive process considering step input. Now, you can
recall the first order system model, a 1 d y d t plus a naught y equals to b f t and
considering a naught equal to 0, we got y bar s divided by f bar s that is equal to k p
prime by s and this is the transfer function of a pure capacitive process, that we have
derived earlier, this is the transfer function of a pure capacitive process.
Now, we will introduce a step change in input variable. So, f t, we will consider equal to
1 I mean this is a unit step change, to observe the transient behavior of a process we need
to introduce some change in the input variable. So, for that we are considering a unit step
change. So, in Laplace domain we can write this as 1 by s then what will be y bar s, y bar
s becomes k p prime divided by s square because, f bar s equals to 1 by s. Now, we will
take the inverse of Laplace transform, inverting we get, the output in time domain equal
to k p prime t, this is a linear equation, with slope k p prime.
(Refer Slide Time: 45:13)

Now, we will represent it graphically, if we consider t tends to infinity what happens y t


tends to infinity. So, this is y t and this is time t this is the output and slope is k p prime.
This type of processes are called non self regulating processes, this type of processes are
called non self regulating systems. Question is why, we will consider one system, that is
liquid tank system, to explain this concept we will consider one liquid tank system.
In fact, we have considered this earlier also, one constant displacement pump is installed
here, h A. Now, we are introducing a unit step change in f I, then what happens the

height increases because, this pump is delivering constant output. So, height gradually
increases with time agree. So, if we consider here, positive change I mean if we increase
f i then the liquid in the tank I mean the height of liquid in the tank increases and this
tank becomes I mean this tank is flooded agree. So, flooding takes place.
Similarly, if we decrease f I, then height decreases because, pump is delivering constant
quantity and the tank becomes empty, that is why we called this process as non self
regulating process. If there is no pump, you see the fact if there is no pump with the
increase of f i the height increases instantly, then hydrostatic pressure increases, which in
turn increases the out flow rate and equilibrium is established, that is basically the new
steady state, in that case we call the we use the self regulation term, if there is no pump
this process is called self regulating process. Anyway next we will consider another
dynamic study, that is the dynamics of first order system considering step change,
previously we have consider considering ramp change.
(Refer Slide Time: 49:21)

So, now, we will discuss dynamic behavior of first order systems considering step
change. So, the transfer function of a first order system is k p divided by tau p s plus 1,
this is the transfer function of a first order system. Now, we will consider step change,
with suppose magnitude A this is step input; that means, f bar s equal to a by s agree,
then y bar s becomes k p A divided by s into tau p s plus 1. Now, we can write this as k p
A multiplied by 1 divided by s minus tau p divided by tau p s plus 1.

Taking inverse of Laplace transform, we get the output y in time domain as k p A 1


minus exponential of minus t divided by tau p. This is a process output of a first order
system subjected to step input, in time domain. Now, here y is the deviation variable.
Now, in the next class we will discuss the, graphical representation of this system I mean
first order system subjected to step input.
Thank you, what you told.

Process Control and Instrumentation


Prof. A. K. Jana
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture - 11
Dynamic Behavior of Chemical Processes (Contd.)
In the last class, we just started the discussion on Dynamic Behavior of first order system
that we will continue.
(Refer Slide Time: 01:02)

So, the topic is dynamic behavior for first order process introducing step change in input
variable. So, you know the first order system can be represented by the first order
differential equation and we have considered this equation, previously a naught y equal
to b f t this is the model first order process. And transfer function we got, that is G s
equal to y bar s divided by f bar s equal to k p divided by tau p s plus 1 and we have
considered step change in input variable with magnitude A I mean f t equal to A and in
Laplace domain this is A by s.
So, y bar s becomes k p A divided by s tau p s plus 1 inverting this, we get y t equal to A
k p 1 minus exponential of minus t divided by tau p up to this we had discussed in the
last class. Now, we will try to plot the y versus time. So, to plot this equation I mean we
will make the plot in terms of dimensionless quantities. So, dimensionless quantities
means, we will consider y t divided by A k p versus t divided by tau p we will make a

plot between y t divided by A k p versus t divided by tau p. This is the dimensionless


quantity, which we will consider along y axis and this is also the dimensionless quantity
which we will consider along x axis.
(Refer Slide Time: 04:33)

So, what will be the plot y t divided by A k p and t divided by tau p say we are
introducing a step change with magnitude A. So, what will be the final steady state, final
steady state we can find considering t tends to infinity. So, if we consider y t divide by A
k p with t tends to infinity how much it is 1. So, final steady state value is 1 and this is
the, starting state I mean t time equals to 0 is a starting point. Now, the dynamics of y t is
represented graphically by this curve.
So, this is the dynamic response of the first order system, if we introduce a step change
in the input variable with magnitude f magnitude A. Now, you will note down few
points, this is basically a self regulating process. Previously we have considered one case
that is, pure capacity process and that is non self regulating process. If we take one
example, then I think it will be clear, why this is self regulating process, we can consider
a liquid tank system, input is f i output is f height is h.
So, if f i increases then what happen, height increases then hydrostatic pressure increases
which in term increases the out flow rate. And after a time period and equilibrium is
established, upper a time period and equilibrium is established and at that equilibrium

state the value of y t by A k p is the new steady state value, that is the reason for which
we are considering this is the self regulating process, this is the first remark.
(Refer Slide Time: 08:48)

Second remark, what is the slop this response at time t equal to 0 can you calculate the
slop at time t equals to 0 I mean d y t divided by A k p d t by tau p at time t equals to 0
how much is this, exponential of minus t by tau p at t equals to 0, so slop is 1.
(Refer Slide Time: 09:51)

If we see this dynamic response then this state line has the slop of 1, what will be the
corresponding t by tau p value, if slop is 1 what is the corresponding t by tau p value 1

because, this value is 1 so, it is 1. So, can we say that, if the initial rate of change of y t
wire to be maintained the response would reach it is final steady state value in 1 time
constant, can we say if this is 1; that means, t by tau p 1; that means, t equal to tau p.
(Refer Slide Time: 11:00)

So, we can conclude that, if the initial rate of change of y t wire to be maintained, then
the response would reach it is final value in one time constant. Because, t divided by tau
p is 1 so; obviously, t equal to tau p this is the second remark. Next we will go for the
third remark.
(Refer Slide Time: 12:10)

But, what happens originally I mean if we see this plot if this is the response, then only
the final steady state value we obtain at t equal to tau p, but that is not the original case,
the dynamic response is this 1. So, at t equal to tau p how much is the y t by A k p value,
can we calculate that.
(Refer Slide Time: 12:50)

See we have the equation y t divided by A k p equal to 1 minus exponential of minus t


divided by tau p. Considering t divided by tau p equals 1, we get y t divided by A k p
equals 0.632, considering t by tau p equals 1 we get y t divided by a k p 0.632; that
means, originally the process reaches 63.2 percent of it is final value. So, this is 0.632
originally in one time constant, the process reaches 63.2 percent of it is final value.
So, write we can write in this tabular form, that when t equals tau p this value is 0.632.
Similarly, considering 2 tau p and using this equation, we get 0.865 I mean in the two
time constants, the process reaches 86.5 percent of the final. Similarly, if we consider 3
tau p we will get 0.95 4 tau p this is 0.98 this is our third remark.
Fourth remark, we discussed one thing that, the gain equals change in output divided by
change in input. Recall this correlation, which we discussed earlier, gain equals change
in output divided by change in input. So, for this system, what is the change in output,
what is the change in y that is k p multiplied by A. Because, we have seen earlier that
considering t tends to infinity y t divided by A k p equals 1; that means, y equals A k p
and how much change we introduced in the input that is A.

So, this gain becomes k p and originally we have considered in the transfer function k p
at the gain. But, remember that this is the steady state gain and this is we determined
considering steady state only. So, these are four comments, we can make on the dynamic
of first order system considering step change in input variable with magnitude A.
(Refer Slide Time: 17:21)

Our next topic is variable time constant tau p and gain k p, so for we have discussed
considering time constant and gain as constant quantities. Now, we will consider one
case, in which we will observe that tau p and k p are not constants. So, for we have
considered tau p and k p both are constants. Now, we consider one example and we will
observe that these two are variables, we will consider one that simple liquid tank system
input flow rate is f i and output flow rate is f, suppose f is proportional to square of liquid
height.
So, f equals to alpha root h, we derived the modeling equation for this system
considering f equals alpha root h I am writing that modeling equation which we derived
earlier f i. Then in this equation this is the non linear term we linearized it considering
Taylor series. And finally, we got A d h d t plus alpha divided by 2 root h s h equals f I
we got this equation in linearized form and considering the deviation variables, we can
write this equation in this form A d h prime d t plus alpha divided by 2 root h s h prime
equals to f i prime.

Now, we can represent this equation introducing tau p and k p by this form tau p d h
prime d t plus h prime equals k p f i prime. Now, here tau p equals 2 A root over of h s
divided by alpha, if we compare these 2 equations we get the expression for tau p equals
2 A root over h s divided by alpha. Similarly, we get the expression for k p that is k p
equals 2 root over h s divided by alpha. So, this is the expression for time constant and
this is the expression for gain are they constant or variable. See for a particular f i value,
suppose that is f i s 1 we get the corresponding height that is suppose, h s 1.
Now, if we change this f i value to f i s 2 we get different height, that is suppose h s 2;
that means, we can conclude that, this h s where is depending on the f i s value the h s is
obtained. So, we can say that this tau p and k p both are not constant in this particular
example, they are the variables. So, this is all about the dynamic of first order systems, in
the next we discuss the dynamics of second order systems.
(Refer Slide Time: 23:16)

So, next topic is the dynamics of second order systems what is the definition of second
order system, second order system is 1 whose output y is modeled by the second order
differential equation. So, the output is y t and this should be modeled by second order
differential equation. Now, we will consider a second order differential equation. So, for
the second order process, the modeling equation we can represent by this form a 2 d 2 y
d t square plus a 1 d y d t plus a naught y equals b f t a b both are constant coefficients. y
is the output of the process and f is the input to the process.

Now, if we divide both sides of this equation by a naught, we get a 2 divided by a naught
d 2 y d t square plus a 1 divided by a naught d y d t plus y equals b by a naught f t. This
equation we get by dividing both sides by a naught. Now, we will represent this
equation, by this form.
(Refer Slide Time: 27:10)

Tau square d 2 y d t square plus 2 zeta tau d y d t plus y equals k p f t. We will represent
this equation, the previous equation in this form tau square d 2 y d t square plus 2 zeta
tau d y d t plus y equals k p f t here, tau square equals a 2 by a naught, 2 zeta tau equals a
1 by a naught and k p is b a naught.
This tau is called natural period of oscillation of the system, tau is the natural period of
oscillation of the system and zeta is here, dumping factor zeta is called dumping factor
and k p is the steady state or static gain, considering y and f both are deviation variables.
We finally, get the transfer function of this second order system, taking Laplace
transform of this equation as G s equals y bar s divided by f bar s equals k p divided by
tau square s square plus 2 zeta tau s plus 1, this is the transfer function of the second
order system. Now, the second order system transfer function includes one term that is
dumping factor.

(Refer Slide Time: 31:19)

Dumping factor provides a measure of the amount of dumping in the system that is the
degree of oscillation in a process response after a perturbation. Dumping factor provides
a measure of the amount of dumping in the system, that is the degree of oscillation in a
process response after a perturbation. Now, small value of zeta what is indicates, small
value of zeta means little dumping small value of zeta implies little dumping, but a large
amount of oscillations.
Suppose, this is the output and the response we are getting in this way, another response
is like this. So, in this case we can say large dumping, in the second case small dumping,
when the dumping factor zeta becomes 0, when zeta equals 0 in that case the oscillation
occurs with constant amplitude. So, zeta becomes 0 belongs undumped system, it means
oscillation with constant amplitude, for the case of undumped system, we observe
oscillations with constant amplitude.
In another case we considered zeta less than 0 what it indicates, oscillations with
increasing amplitude. So, it belongs to unstable system, for this case the output is
somewhat like this. So, for all stable systems zeta should be greater than 0. Now, earlier
we have considered first order systems, presently we are discussing the second order
systems now, what processes are called second order systems.

(Refer Slide Time: 36:40)

So, we will discuss now the systems with second or higher order dynamics, how we can
obtain the second or higher order dynamics, that will discuss now. If 2 or more first order
systems are connected in series, then we obtain second or higher order systems, if 2 first
order systems are connected in series, the overall dynamics is second order dynamics. If
we connect more than 2 first order systems in series, we obtain higher order dynamics.
So, this is the first option I mean by this way we can get the higher order dynamics.
So, under this category the processes consist of two or more capacities and they are first
order systems in series. If a process incorporates a controller, then also we can obtain
higher order dynamics say for example, we have a first order system, if one controller is
employed with this first order system, then the overall response may be second order
response. If a process includes a controller, then we can obtain second or higher order
dynamics and this is the example, if we consider a first order system that includes a
controller then the overall response may by second order dynamics.
Thirdly few processes are inherently higher order, few processes are inherently second
order systems, but this is very rare in chemical engineering and we will not discuss this
third option. We will discuss only first and second options.

(Refer Slide Time: 40:30)

Now, we will discuss the types of second order system response. Previously, we obtained
the transfer function for the second order system in general form, that is G s equals k p
divided by tau square s square 2 zeta tau s plus 1 this transfer function we derived earlier
for the second order system. Now, we want determined the poles of the second order
system, for determining the poles of the transfer function, we write tau square s square
plus 2 zeta tau s plus 1 equals 0 for determining the poles of the transfer function or
poles of the system, we write tau square s square plus 2 zeta tau s plus 1 equals 0.
So, if we represent the pole by p then p equals minus 2 zeta tau plus minus root over 2
zeta tau whole square minus 4 tau square divided by 2 tau square, see this is the
quadratic equation. So, we can easily obtain the poles, we have represent the poles by p
and the expression for the p is this 1. So, we will get basically 2 poles p 1 equals minus
zeta divided by tau plus root over of zeta square minus 1 divided by tau, another pole is p
2 and p 2 equals minus zeta divided by tau minus root over of zeta square minus 1
divided by tau these are the 2 poles. Now, we will categories the system response, based
on the value of zeta, we will now categories the system response based on the value of
zeta.

(Refer Slide Time: 44:22)

So, first we will consider zeta greater than 1, in case 1 we will consider zeta greater than
1. If zeta greater than 1 then we obtain two distinct real poles, you see from the
expression of p 1 and p 2 if we consider zeta greater than 1 we get true distinct real
poles. The corresponding system response is called over dumped response, this is case 1
where we have considered dumping factor greater than 1 and dumping factor greater than
1 implies two distinct real poles.
In the next case we will consider zeta equals 1. So, what about the poles, two equal real
poles I mean you can say 1 pole. So, we can write 2 equal real poles, it means 1 real pole
and that is minus zeta divided by tau, we obtain for the case of zeta equals 1 a single pole
and that is minus zeta divided by tau. Then the corresponding system response is called
critically dumped response.
So, when zeta equals one the response is critically dumped, in third case we will consider
0 less than zeta less than 1 in the third case we consider zeta is in between 0 and 1 in this
case what about the poles, we get two complex conjugate poles and the poles are p 1, 1 is
p 1. And another one is p 2 the expression for p 1 is this one p 1 equals minus zeta by tau
plus i root over 1 minus zeta square divided by tau, another pole is minus zeta by tau
minus i root over 1 minus zeta square divided by tau. So, when zeta is in between 0 and 1
the corresponding response is called under dumped response.

So, this is the under dumped response. So, these are three different cases, depending on
the value of zeta. Now, we need to take the invert of inverse of Laplace transform to
obtain the expression for y in time domain, that will discuss in the next class, the
expression for y in time domain thank you, what you are saying.
Student: ((Refer Time: 49:33))
Yeah, that I will correct.

Process Control and Instrumentation


Prof. A. K. Jana
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture - 12
Dynamic Behavior of Chemical Processes (Contd.)
(Refer Slide Time: 01:00)

In the last class, we just started to discuss the time domain output. So, we will continue
the topic that is step response time domain solutions, fine. Previously we derived the
transfer function for the second order system that is represented by G s equals y bar s
divided by f bar s equals K p divided by tau square s square 2 zeta tau s plus 1. Now, we
will consider a step change in input variable, input variable is f t, if we considers step
change with a magnitude of A that means f t equals A, and in Laplace domain, we can
write this as A divided by s, fine.
Now, introducing this f bar s expression in this general form of transfer function, we
obtain y bar s equals K p A divided by s tau square s square plus 2 zeta tau s plus 1, fine.
Now, we will take the inverse of Laplace transform of this expression to obtain output in
time domain, I mean y t. Now, we already considered three different cases depending on
the value of zeta; one is over damped, second one is critically damped, and third one is
under damped. So, we will get three different expressions of y t for these three different
cases.

So, first we will write the expression for the case of over damped response in the case of
over damped response, we have zeta greater than 1, fine. And the expression is y t equals
K p A multiplied by 1minus exponential minus zeta t divided by tau cos hyperbolic root
over zeta square minus 1 divided by tau into t plus zeta divided by root over of zeta
square minus 1 sin hyperbolic root over zeta square minus 1 divided by tau into t, fine.
This is the expression for the case of over damped response the derivation is not shown
here, but if we take the inverse of Laplace transform of this equation we get this
expression for the case of over damped system, fine. The derivation is left for the
students.
(Refer Slide Time: 05:36)

Similarly, for the in the second case, we consider critically damped response where zeta
equals 1 and the expression of output in time domain we obtain by inverting as y t equals
K p multiplied by A 1 minus 1 plus t by tau exponential minus t divided by tau this is the
expression of output y in time domain for the case of critically damped systems, fine. For
the case of under damped response, where zero less than zeta less than 1. We obtain the
expression for output y in time domain as y t equals k p multiplied by A 1 minus
exponential minus zeta t divided by tau cos root over of 1minus zeta square divided by
tau into t plus zeta divided by root over of 1 minus zeta square sin root over of 1minus
zeta square divided by tau into t.

This is the expression of output y for the case of under damped response, we can write
this equation in this form where y t equals K p A1 minus 1divided by root over of
1minus zeta square exponential minus zeta t divided by tau sin omega t plus phi
rearranging the expression of y in time domain. We get this last equation where the
radian frequency omega radian frequency omega equals root over of 1 minus zeta square
divided by tau this is the expression for radian frequency omega. Similarly, if the
expression for phase angle phi, we obtain as phi equal tan inverse root over 1minus zeta
square divided by zeta this is phase angle this is phase angle, fine. So, these are the three
expressions of y in time domain just by inverting the equation of y in s domain.
(Refer Slide Time: 10:16)

Now, if we graphically represent the outputs of this three y expressions the float looks
like this t by tau and in this direction, we are considering y t divided by A K p suppose
the final steady state value is 1. So, the output is like this type of response, we obtain for
the value of zeta 1 that is suppose 0.5, fine. This type of dynamic response, we obtain for
the value of zeta. Say 0.5 in another case, we get the response like this type of response.
We get suppose for the value of zeta equals 1and another type of response we get like
this for the value of zeta suppose 2.0. This is a zeta 3 and this is zeta 2. So, obviously
here zeta 3 greater than zeta 2 greater than zeta 1, fine. So, this is the case of under
damped, where oscillations are there. And zeta 1equals 0.5 that is the case of under
damped response and oscillations are there. In the second case, we have considered zeta

2 equals 1. There is no oscillation and that is the case of critically damped response. And
in the third case, we have considered zeta 3equals 2 that is over damped response and
there is no oscillation. Usually these two cases zeta 2 and zeta 3 are characterized as
sluggish. The responses are under critically damped and under damped response are
characterized as sluggish, fine. If you reduce the value of zeta 1 then 0.5, you will get
more oscillation, fine. So, these are the responses of y at three different cases. Next you
will discuss the under damped response in details.
(Refer Slide Time: 13:51)

Almost all under damped responses in a chemical plant are caused by the interactions of
controllers and process units. Almost all under damped responses, almost all under
damped responses are caused in a chemical plant due to the interactions of controllers
with the process units. That is why we are more interested to analyze the under damped
responses, fine. Now, we will draw the response suppose this is time t this is y t by
suppose K p, we are considering here A equals 1, fine. Now this is the new steady state
value response is like this the time corresponding to this is say represented by t p.
Sorry, we will use another suffix that is t r, fine; t r is rise time. This is the first maximum
peak the corresponding t is or corresponding time is t p. That is peak time or time to first
peak t p is peak time or time to first peak, fine. This quantity is suppose B. And this
quantity is suppose A. And this 1 is suppose C. This is the period represented by t. So,
here t is the period, fine. And these are indicating plus minus 2 percent deviation. And

corresponding time, we will represent by t s, where t s is the settling time or response


time, fine. Now, we will know of these times t r t p t s along with the quantities a b c 1
by 1. So, first we will know about the rise time, it is very obvious from the figure that it
is the time required for the output y to first reach its steady state value.
(Refer Slide Time: 19:25)

So, the definition of rise time is, it is the time required for the output y to reach, to first
reach its final steady state value. Rise time is the time required for the output to first
reach its final steady state value and it is very obvious from the figure, fine. So, what
about the peak time, it is the time required for the output to reach its first maximum
value. It is the time required for the output to reach its first maximum value. What is the
response or settling time? Response or settling time, it is the time required for the output
to come within some prescribed band of the final steady state value. It is the time
required for the output to come within some prescribed band of the final steady state
value, fine.
Next term is the decay ratio. Decay ratio decay ratio is define by the ratio of c and a.
Decay ratio is defined by the ratio of c and a. We have the expression of y in time
domain for the output y, fine. Now, from that equation, we can easily find the expression
for this decay ratio that is. So, the expression of decay ratio is equal to exponential of
minus 2 pi zeta divided by root over 1 minus zeta square. We have the expression of y in
time domain for the under damped response from that expression, we can easily derive

the expression for decay ratio and that is this 1 decay ratio equals exponential of minus 2
pi zeta divided by root over of 1 minus zeta square another term is there that is over
shoot.
(Refer Slide Time: 23:12)

Another term is there that is called over shoot it is defined by the ratio of A and B over
shoot is A divided by B, fine. And the expression for the over shoot is mathematical
expression for the over shoot is represented by this equation, exponential minus pi zeta
divided by root over of 1 minus zeta square. This is the expression for over shoot this
expression also, we can derive from the time domain expression of y. If you see the
expression of the over shoot and decay ratio is there any correlation between this two
decay ratio is the square of over shoot or we can write over shoot is root over of decay
ratio square root of decay ratio, fine. Now, over shoot increases with decreasing zeta, if
you see this expression it is clear that over shoot increases with decreasing zeta. So, first
remark we can make that is over shoot increases with decreasing zeta, fine. What
happens? ((Refer Time: 25:27)) zeta becomes 1, there is no over shoot fine.
So, over shoot approaches 0. Over shoot approaches 0 and zeta approaches 1, we made 1
float comparing the responses of different zeta values. In 1 case we considered zeta
equals 0.5. In another case, we considered zeta equals 1, and in the third case, we
considered zeta equals 2. And you can see that float where zeta becomes 1 there is no
over shoot, fine. It is very obvious from this mathematical explanation.

(Refer Slide Time: 26:37)

So, next we will discuss the period of oscillation period of oscillation for an under
damped response, We got previously, the expression of radian frequency for the under
damped response, We got the expression for radian frequency previously, where omega
equals root over of 1 minus zeta square divided by tau, fine. This equation we have used
in the expression of y in time domain for under damped case. Now, we will consider if
cyclical frequency, unit is cycle spark unit time. So, if the cyclical frequency is f. Then
the period is 1 by f, if f is the cyclical frequency then; obviously, the period becomes 1
by f fine. So, what is the radian frequency omega. Omega equals 2 pi f again this equals
2 pi divided by t. So, we can write t equals 2 pi divided by omega.
Substituting, the expression of omega, we finally obtain 2 pi tau divided by root over of
2 minus zeta square. f is the cyclical frequency, then we get, we know the period equals
1 by f s. Now, this is the expression for radian frequency omega equals 2 pi f means 2 pi
divided by t finally, we can write the period equals 2 pi by omega, and here we
substituted the expression of omega. And we finally get the expression for the period,
fine. Next we will cover the natural period of oscillation, natural period of oscillation,
fine. And in this case zeta equals 0.

(Refer Slide Time: 29:30)

In this case zeta equals 0; that means, we observe the natural period of oscillation for undamped systems fine. Now we know the transfer function of a second order system, that
is K p divided by tau square s square plus 2 zeta tau s plus 1. This is the transfer function
of second order systems, if we consider zeta equals 1. Then we get K p divided by tau
square s square plus 1, considering zeta equals 1, we get the transfer function K p
divided by tau square s square plus 1, we can write this as K p divided by tau square
divided by s minus i 1 by tau s plus i 1 by tau, fine.
So, 2 poles are involved and they are complex conjugate poles fine. Can you, can you
observe the position of the poles, this is the real axis, this is the imaginary axis there is
no real part. So, both the poles lie on the imaginary axis, 1 is here, another 1 is here, the
two poles are involved here and for both the poles there is no real part. Therefore, both
the poles lie on the imaginary axis. So, what about the natural period? What about the
natural frequency? Omega n, we know the expression of omega that is root over of 1
minus zeta square divided by tau.
We know the expression of omega that is omega equals root over of 1 minus zeta square
divided by tau, substitute zeta equals 1 then will get omega n, so omega n equal to 1 by
tau. Here you just substitute 0 1 minus 0 divided by tau that means, this will 1 by tau
which is written here, fine. Then the period we can write by this way t n equals to 2 pi by
omega n. that means, 2 pi tau, agree. So, this is the expression for natural period of

oscillation t n equals 2 pi tau, fine. Next we will discuss the multi capacity processes as
second order systems.
(Refer Slide Time: 34:12)

Next we will discuss the multi capacity processes as second order systems. We can
obtain overall the second order dynamics if we connect to first order systems in series,
fine. So, we can obtain a second order system by connecting to first order systems in
series. In another way if we include a controller with a first order process then we obtain
the second order dynamics.
So, in the first option if we connect to first order systems in series, we obtain second
order dynamics in another case if we include a controller with a first order system. We
obtain second order dynamics, fine. So, these 2 options we will discuss in details. So,
first we will consider the connection of 2 first order systems in series. We will consider 2
simple liquid tank systems. So, how we can connect the 2 liquid tank systems in series?
This is tank 1. Tank 1 input to this tank is F i cross sectional area is suppose A 1 height
of liquid in the tank is suppose h 1 the output of tank 1 enters tank 2. This is tank 2 and
this flow rate is suppose f 1 here resistance to flow is say R 1. In the tank 2 consider the
cross sectional area A 2 and liquid height h 2. the stream which is coming out from tank
2 that has the flow rate of f 2 and resistance is suppose R 2, fine.
So, in this schematic representation to first order systems are connected in series. Now,
how the tanks interact to each other? You see tank 1 effects tank 2, but tank 2 does not

affect tank 1, because the flow rate to tank 2 inlet stream 2. Tank 2 is the outlet of tank 1.
So, what about the interaction? This is 1 way interacting system tank 1 affects tank 2, but
tank 2 does not affect tank 1. And we will use the term for this type of systems that is
non-interacting system, fine. If the 2 liquid tanks are connected in series in this way, we
will use the term non-interacting tanks. In another way also, we can connect the 2 tanks.
(Refer Slide Time: 39:38)

This is tank 1 cross sectional area is A 2 height is h 1 inlet flow rate is f i. outlet is f 1,
and resistance is R 1. Second tank is placed here. Cross sectional area A 2 height h 2.
Output is f 2, fine and resistance is here R 2. This is the second scheme in which we have
connected the two tanks in series. Now, what about the interaction in this system? This is
the case of both way interaction I mean tank 1 interacts tank 2 and tank 2 also interacts
tank 1. So, for this type of systems we will use the term interacting system. So, for this
type of systems we will use the term interacting system or interacting tanks, fine. Now,
we will try to derive the overall transfer function for the case of non-interacting tanks
and in the next, we will consider for the case of interacting tanks, fine.

(Refer Slide Time: 41:57)

So, now our topic is non-interacting capacities. And we will discuss the general case,
fine. So, 2 tanks are or 2 first order systems are connected in series, those 2 first order
systems have the transfer functions of suppose G 1 and G 2. So, this is one tank G 1 and
this is another tank G 2. input is suppose f 1 par s. Output of the first first order system is
y 1 bar s. and output of the second first order system is y 2 bar s, fine. So, if we consider
both the systems are first order, then what is the modeling equation for first system I
mean for g 1 the modeling equation can be written by this equation, tau p 1 d y 1 d t plus
y 1 equals K p 1 f 1 t. So, we can write this is the modeling equation for say first
capacity, which has the transfer function of G 1.
Similarly for the second capacity which has the transfer function of G 2, we can write the
modeling equation as tau p 2 d y 2 d t plus y 2 equals K p 2 this will be y 1 t, fine. This is
for the second capacity. So, for the first capacity the time constant is tau p 1, gain is K p
1, output is y 1, and input is f 1 for the second capacity tau p 2 is the time constant K p 2
is the gain y 2 is the output and y 1 is the input. So, using this two modeling equations it
is straight forward to derive the transfer functions. So, we can write the expression for
transfer function for the first capacity G 1 is equals y 1 is divided by f 1 s that will be K p
1 divided by tau p 1 s plus 1 fine.
This is for the first capacity the expression of G 1, we can obtain by taking Laplace
transform of this mode. Similarly, for the second capacity, we obtain the transfer

function s G 2 s equals y 2 bar s divided by y 1 bar s equals K p 2 divided by tau p 2 s


plus 1. This is for second capacity, agree. Now, our objective is to find the overall
transfer function, if we consider this is a single system. So, what will be the overall
transfer function? That will be y 2 bar s divided by f 1 bar s. isnt the overall transfer
function will be.
(Refer Slide Time: 47:32)

If we represent by G naught s as G naught is equals y 2 bar s divided by f 1 bar s for the


overall system y 2 is the output and f 1 is the input. So, we can write this as y 2 bar s
divided by y 1 bar s multiplied by y 1 bar s divided by f 1 bar s, agree. So, this is the
expression for overall transfer function. This means G naught is equals G 1 s multiplied
by g 2 s, this is the transfer function G 1 and this is the transfer function G 2. So, if we
substitute the expression for G 1 s and G 2 s, we obtain G naught s as K p 1 divided by
tau p 1 s plus 1 multiplied by K p 2 divided by tau p 2 s plus 1, fine.
So, we can write this expression as K p prime divided by tau prime whole square s
square plus 2 zeta prime tau prime s plus 1, we can write this equation in this form. G
naught is equals K p prime divided by tau prime whole square s square plus 2 zeta prime
tau prime s plus 1, where tau prime whole square equals tau p 1 tau p 2, fine. And 2 zeta
prime tau prime equals tau p 1 plus p 2 and third term that is K p prime equals K p 1 K p
2, fine. So, this is the prime expression, this is the expression for 2 zeta prime tau prime
and this is an expression for K p prime. Now, we will make some conclusions.

(Refer Slide Time: 51:09)

So, what about the overall response is that first order or second order? the overall
response is second order. So, it is proved that if two first order systems are connected in
series the overall response becomes second order response, fine. So, we have considered
2 first order systems and we got second order response. Secondly, what are the poles in
the overall transfer functions? One pole is minus 1 divided by tau p 1. Another pole is
minus 1 divided by tau p 2.
If we see the overall transfer function G naught s we have two poles 1 is minus 1 divided
by tau p 1, another 1 is minus 1 divided by tau p 2. So, what about the overall response
over damped or critically damped or under damped? We have to distinct real poles the
response is over damped response, but in some cases tau p 1 may be equal to tau p 2. In
that situation the response is critically damped, fine. So, if 2 first order systems are
connected in series, we obtain over damped or critically damped response, but never
under damped, fine. this is the second naught third one is we have connected two 2 first
order systems in series if we connect n number of first order systems what will be the
overall transfer function?
Suppose this is G 1 this is G 2 like this way we are connecting n number of systems this
is f 1 and finally, we are getting y n s, fine. So, overall we can write G naught s equals G
1 s G 2 s G small n s equals K p 1 K p 2 K p n divided by tau p 1 s plus 1 tau p 2 s plus 1

like this tau p n s plus 1, fine. So, if n first order systems are connected in series we
obtain the overall transfer function like this.
Thank you.

Process Control and Instrumentation


Prof. A. K. Jana
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture - 13
Dynamic Behavior of Chemical Processes (Contd.)
(Refer Slide Time: 01:12)

Today, we will continue our discussion on dynamics of second order systems. And we
will start today, with a physical example of noninteracting system. Example, of the
noninteracting system it was told that if 2 first order systems or 2 first order tanks are
connected in series the overall response is second order response. So, today we will take
that example I mean, 2 first order systems those are first order tank systems connected in
series.
So, this is tank 1, input to the tank is Fi cross sectional area is A1 and height is h1output
is F1 and resistant to flow is suppose R1. This outlet from Tank 1 enters Tank 2, this is
Tank 2 cross sectional area of this Tank 2 is a 2 and liquid height in the Tank is h2,
output stream from Tank 2 is F2 and suppose this resistance is R2 this is a example of a
noninteracting system.
Now, what are the modeling equations for these two Tanks? If we consider Tank 1 the
modeling equation is A1 dh 1 dt equals input flow rate that is Fi minus output flow rate
that is R1. This is a modeling equation for Tank 1. Now, F1 is again we can write by this

form h1 divided by R1 equals h1 divided by R1 driving force divided by resistants to


flow. If, we substitute this F1 equals h1 divided by R1 in the modeling equation then, we
have A1 dh 1 dt equals Fi minus h1 divided by R1. Substituting the expression of F1 we
finally get this, rearranging this equation we have A1 dh 1 dt plus h1 divided by R1
equals Fi fine, rearranging we get this.
Now, if we write this equation in terms of deviation variables, then we obtain A1 dh 1
prime dt plus h1 prime divided by R1 equals Fi prime. This is modeling equation in
terms of deviation variables. Fine now, if we multiplied both sides by R1 then we get A1
R1 dh 1 prime dt plus h1 prime equals R1 Fi prime, this equation we get by Multiplying
both sides with R1 fine.
(Refer Slide Time: 07:25)

Now, you will assume that A1 R1 equals tau p1 and R1 is equal to Kp 1we assume A1
R1 equals tau p1 which is for the time constant First Tank and R1 equals kp 1 this is the
game of the first tank. Then, the equation modeling equation becomes tau p1 dh 1 prime
dt plus h1 prime equal to this is Kp 1 Fi prime this is the first order differential equation
of the output h1.

(Refer Slide Time: 09:39)

Now, if we take Laplace transform and rearrange then we get the transfer function of the
First tank as h a bar prime is divided by Fi bar prime is equals Kp 1 divided by tau p1 s
plus 1 this is a transfer function of the First tank that is tank 1. Now, you will represent
this as g1 s similarly we will consider tank 2 for tank 2 we can write the modeling
equation as A2 dh 2 dt equals input flow rate that is F1 minus output flow rate that is F2.
Now, F2 is again the driving force divided by resistant to flow. So, FT F2 equal to h2
divided by R2 if you substitute that in this modeling equation we get A2 dh 2 dt equal to
F1 minus h2 divided by R2. Multiplying both sides R2 and rearranging we get A2 R2 dh
2 dt plus h2 equal to R2 multiplied by F1 fine multiplying both sides by resistant R2 and
rearranging we get this equation.
Now, similarly we have to represent this equation in terms of deviation variables then we
get A2 R2 dh 2 prime dt plus h2 prime, equal to R2 F1 prime. This is represented in
terms of deviation variables fine. Now, we assume A2 R2 equals tau p2 and R2 is equal
to Kp 2, we assume A2 R2 is equal to the time constant for tank 2 and R2 as gain of the
tank 2. Then we get tau p2 dh 2 prime dt plus h2 prime equal to Kp 2 F1prime.

(Refer Slide Time: 13:07)

And taking Laplace transform finally, we get the transfer function of tank 2 as G2 s is
equal h2 bar prime is divided by F1 bar prime s equals Kp 2 divided by tau p2 s plus 1.
This is the transfer function of the second tank that is tank 2. Now, if you see the
schematic of the 2 tank system. We can say that the overall transfer function will be h2
bar prime s divided by Fi bar prime s fine. The overall transfer function of this 2 tank
systems can be written G naught s equals h2 bar prime s this is a output of tank 2 divided
by the input of input 2 tank 1 fine. This is a overall transfer function.
Now, we can write again the overall transfer function by this form, h2 bar prime s
divided by F1 bar prime s multiplied by F1 bar prime s divided by Fi bar prime s can we
write this, the overall transfer function we can write in this form.
So, this is equal to h2 bar prime s divided by F1 bar prime s this is nothing, but the
transfer function of tank 2. And what is F1 bar prime s we know F1 bar equals h1 bar
divided by R1, for the tank 1 we have consider this correlation. F1 bar equals h1 bar
divided by R1 that means, h1 bar divided by kp 1. Can we write this?
Now, this relationship will substitute here, I mean F1 bar F1 prime this is not bar prime
f1 prime equals h1 prime divided by kp 1. So, if we substitute here, h1 bar prime s
divided by kp 1 Fi bar prime s. Can we write this? There is no change of this term. Now,
in place of F1 prime we have substitute it h1 prime divided by kp 1 and then, Fi prime
remains there. So, the first term this 1 represents Gp 2 not Gp 2 G2 and multiplied by

this which is G1 divided by kp 1 that means, kp 2 divided by tau p2 s plus 1 multiplied


by 1 divided by kp 1 multiplied by kp 1 divided by tau p1 s plus 1 agree. So, the overall
transform function G naught s we can write as, kp 2 divided by tau p1 s plus 1 multiplied
by tau p2 s plus 1. This is a overall transfer function of the example non-interacting
system.
(Refer Slide Time: 18:47)

So, G naught s is basically h2 bar prime s divided by Fi bar prime s. We can write in this
form now, you will introduce a unit step change in the input variable, considering a unit
step change in Fi and taking inverse of Laplace transform we get the output in time
domain as h2 prime t equals kp 2, 1 plus 1 divided by tau p2 minus tau p1 tau p1
exponential of minus t divided by tau p1 minus tau p2 exponential of minus t divided by
tau p2 considering a unit step change in Fi bar prime s I mean, Fi bar prime s equals 1 by
s and taking inverse of Laplace transform we get the expression of output h to in time
domain like this fine.

(Refer Slide Time: 20:49)

Now, you will conclude based on this discussion. So, First conclusion is that, by
connecting to First-order systems in series. We get overall Second-order dynamics fine,
by connecting to First-order system in series we obtain Second-order dynamics. So, by
connecting to First-order systems in series the overall dynamics becomes Second-order
dynamics this is a First conclusion. Second thing is we used to observe the behavior
making a plot.
So, this is h2 prime t divided by kp 2 verses time. The steady state is represented by the
dotted line and the dynamics is like this, is for First-order system. This is the dynamics
of First-order system, this is the dynamics of Second-order or 2 non-interacting 2 noninteracting tanks system and this is the dynamics of 4 non-interacting tanks fine.
For the First-order system this should be h by kp for 2 non-interacting tanks this should
be h2 divided by kp 2 or h2 prime by kp 2 for 4 non-interacting tanks this should be h 4
prime t divided by kp 4. So, this is the dynamic behavior of 3 systems. Now, what
conclusion we can make initially you see it changes slowly for the 2 non-interacting tank
systems initially it changes slowly then it fix up the speed, the behavior for second and
higher order systems is s shaped, fine the behavior is s shaped. So, initially it changes
slowly then its fix up the speed fine.
So, this is basically the over time response, and 3rd conclusion is as the number of
capacities in series increases the delay in the initial response becomes more pronounced.

You see if you compared these two curves, the delay is more for the case of 4 noninteracting tank system. So, that is the 3rd conclusion. As the number of capacities in
series increases the delay in the initial response becomes more pronounced this is the 3rd
conclusion fine.
(Refer Slide Time: 26:02)

So, this is the example physical example of a non-interacting tank system. In the next we
will consider an interacting tank systems; we will connect 2 tanks in series. So, that they
are interactive to each other. So, next example is, the Example of the Interacting System
fine. For the case of interacting system similarly, you consider 2 tanks which are
connected in series. This is tank 1 cross sectional area of A1 and h1 input is Fi resistant
to flow is R1, this is Tank 2 cross-sectional area A2 height is h2. This flow rate is F1 and
the stream which is coming out from Tank 2 has the flow rate of F2 and resistance is R2.
So, the modeling equation for Tank 1 we can write in this form A1 dh 1 dt equals Fi
minus F1. What is F1 here? What is the relationship of F1? F1 equals h1 minus h2
divided by R1. Since h1 is higher than h2 that is why there is a flow. So, if we substitute
this, then we obtain A1, R1 dh 1 dt plus h1 minus h2 equals R1, Fi. If we substitute the
expression of F1 in the modeling equation. Then we get this equation. Similarly, for
Tank 2 the modeling equation we can write in this form A2 dh 2 dt plus h2 divided by
R2 equals h1 minus h2 divided by R1. Can we write this, for Tank2 A2 dh 2 dt plus F2
equal to F1 fine. Now, these 2 modeling equation we need to write in terms of deviation

variables anyway before that, we can rearrange this equations and we obtained A2 R2 dh
2 dt plus 1 plus R2 divided by R1, h2 minus R2 divided by R1 h1 equal to 0. By
rearranging the modeling equation for Tank 2 finally, we get this equation fine.
(Refer Slide Time: 30:46)

Now, we use to write these 2 equations in terms of deviation variables. So, for Tank 1 we
can write the modeling equation in terms of deviation variables, and we obtained A1 R1
dh 1 prime dt plus h1 prime minus h2 prime, equals R1 Fi prime. This is for Tank 1.
Similarly for Tank 2, we can write the modeling equation in terms of deviation variables
as A2 R2 dh 2 prime dt plus 1 plus R2 divided by R1, h2 prime minus R2 divided by R1
h1 prime equals 0, this is for Tank 2 fine.
Now, we will take Laplace transform of these 2 equations. Taking Laplace transform we
get for, Tank 1 A1 R1 S plus 1, h1 bar prime s minus h2 bar prime s equals R1 Fi bar
prime s. Taking Laplace transform for the case of Tank 1, I mean the modeling equation
of Tank 1 we obtain this form. Similarly if we take the Laplace transform of this, we
obtained minus R2 divided by R1 h1 bar prime s, plus A2 R2 s plus 1 plus R2 divided by
R1 h2 prime s equals 0. If we take the Laplace transform for the modeling equation of
Tank 2 we obtain this expression.
Now, we need to solve these 2 equations, to obtain the transform function with respect to
Tank 1. As well as Tank 2, solving these 2 equations we get the transfer function for
Tank 1 as h1 bar prime s, equals tau p2 R1 s plus R1 plus R2, divided by tau p1 tau p2 s

square, plus tau p1, tau p2, plus A1 R2 multiplied by s plus 1 Fi bar prime s. This is for
Tank 1 fine solving these 2 equations we obtain the transfer function for Tank 1 in this
form.
(Refer Slide Time: 35:22)

Similarly, for Tank 2 we obtain the transfer function as h2 bar prime s equals R2 divided
by tau p1, tau p2 s square plus tau p1 plus tau p2, plus A1 R2 s plus 1, Fi bar prime s.
This is a transfer function with respect to h2 fine. So, can say that this is the
representation of overall transfer function because, overall transfer function we can
represent by h2 bar prime s divided by Fi bar prime s. So, this overall transfer function
we obtain for the case of Interacting System. This overall transfer function we obtain for
the case of Interacting System. What is the overall transfer function of the noninteracting system? For the non-interacting system we obtained, the overall transfer
function in this form kp 2 divided by tau p1 s plus 1 multiplied by tau p2 s plus 1, just we
obtain before this discussion.
So, we can write this as R2 divided by tau p1 tau p2 s square plus tau p1 plus tau p2 s
plus 1 it will be multiplied with Fi bar prime s. So, Fi bar prime s this is the overall
transfer function for non-interacting system. Can we compare these two overall transfer
function, if we compare we see that 1 extra term is there for the case of interacting
system. That is A1 R2 this term this is the additional term which is present in the overall
transfer function of interacting tank system.

So, this term may be thought of as the interaction factor. These term A1 R2 may be
thought of as a interaction factor fine. If we compare the 2 transfer functions we see that
1 additional term A1 R2 is present in the overall transfer function of interacting system,
and these term may be thought of as the interaction factor.
(Refer Slide Time: 39:42)

Now, we will conclude based on this discussion for the case of interaction, Interacting
System. What is the First conclusion? First conclusion is that the overall response is
Second-order response this is the First conclusion. By connecting to First-order liquid
tanks in series we obtain the overall response as Second-order response.
So, the overall response is Second-order response. Second conclusion in the second
conclusion we use to first find the roots. What are the roots? There are 2 roots because, if
you see the denominator that is if Second-order polynomial. That is it quadratic that is
given in quadratic form. So, the roots we can write in this form, tau p1 plus tau p2 plus
A1 R2 plus minus root over, tau p1 plus tau p2 plus A1 R2 whole square minus 4 tau p 1
tau p2, divided by 2 tau p1 tau p2 fine there are 2 roots and they are represented by this
form.
Now, tau p1 plus tau p2 plus A1 R2, whole square minus 4 tau p1 tau p2 is greater than 0
fine. That means, 2 distinct real poles so, what about the overall response? The overall
response is over damped response. So, we obtained zeta greater than 1 and the response
of interacting capacities is always over damped.. So, these are about, the two tank

systems I mean, we observe that if 2 First-order systems are connected in series the
overall response is Second-order response.
Another case we mentioned in the previous class that, Second-order system we can
obtained by employing one controller around process, we mentioned 3 cases. 1 is by
connecting to First-order systems in series, 1 is by employing 1 controller with a process,
and in the 3rd case we consider that few systems are inherently higher order systems.
(Refer Slide Time: 43:20)

So, now we will consider the process having a controller. We use to observe the higher
order dynamics. I mean Second-order dynamics for the case of a first-order process
configured with a controller if a controller is employed for a first-order process what will
be the overall dynamics that we use to observe.
So, we will first consider 1 example that is the liquid tank system. This is a liquid tank
system, inlet flow rate is FI, and outlet flow rate is F naught. This is height crosssectional area is A, control objective is to maintain the height of the liquid in the tank.
So, for that we need to employ 1 level controller. So, our control objective is to maintain
the height of liquid in the tank; that means the control variable is h and the corresponding
manipulated variable is F naught fine, this is the control pair for this case.
Now, what is the modeling equation? Modeling equation in terms of deviation variables,
we can write as A dh prime dt equals Fi prime minus F naught prime fine. Now since,

the controller is manipulating F naught. So, we can correlate F naught with the height, by
a controller equation suppose a controller equation is given as F naught equals F naught
s, plus kc h prime plus kc divided by tau i integration of h prime dt plus kc tau d, dh
prime dt. This is the equation of a controller which we will discuss in the subsequent
classes fine.
This is a controller equation which is correlating the manipulated variable Fnaught with
height; h prime is h minus hs fine. Now, we can write this equation as F naught prime
equals kc h prime plus kc divided by tau i integration h prime dt plus kc tau d, dh prime
dt ,where F naught prime equals F naught minus F naught s. And in this equation kc tau i
and tau d these 3 are controller tuning parameters. In this controller equations kc tau i
and tau d are constant parameters fine. Now, we can substitute the expression of f naught
prime in the modeling equation, by substituting the expression of F naught prime in the
modeling equation.
(Refer Slide Time: 48:45)

We obtain A dh prime dt plus kc h prime plus kc divided by tau i integration h prime dt


plus kc tau d, dh prime dt equals Fi prime. Substituting the expression of F naught prime
in the modeling equation we obtain this. Now, taking Laplace transform we get, A plus
kc tau d s h bar prime s, plus kc h bar prime s, plus kc divided by tau i s, h bar prime s
equals Fi bar prime s fine. Taking Laplace transform we obtain this. So, we can write this

as tau i divided by kc A plus kc tau d s square h bar prime s, plus tau i is s h bar prime s
plus h bar prime s equals tau i s divided by kc Fi bar prime s.
Now, we can represent this as, tau square s square h bar prime s, plus 2 zeta tau s h bar
prime s plus h bar prime s equals kp multiplied by s Fi bar prime s fine. So, the transfer
function can be represented by h bar prime s divided by Fi bar prime s equals kp s
divided by tau square s square, plus 2 zeta tau s plus 1. The transfer function yields this
form, h bar prime s divided by Fi bar prime s equals kp s divided by tau square s square
plus 2 zeta tau s plus 1.
(Refer Slide Time: 52:24)

Now, here kp is equals to tau i divided by kc. Similarly tau square is equal to tau i
divided by kc A plus kc tau d and 2 zeta tau is equal to tau i fine. We have considered in
the derivation of a transfer function these simplified forms. I mean this correlation have
been used in the derivation of transfer function.
So, from these equations we obtain, zeta equals half root over of kc tau I, divided by a
plus kc tau d fine. These expression for zeta we obtain from these 2 forms. Now we want
to conclude. So, our first conclusion is that by employing a controller around a first-order
system, we obtain overall second-order response. This is our first conclusion by
employing a controller; around a first-order system we obtain overall second-order
system. And in the second conclusion we one to know whether, it is over damped or
critically damped or under damped response

If, root over of kc tau i divided by a plus kc tau d less than 2 then zeta is less than 1; That
means, this is under damped response. If this is equal to 2; that means, zeta is equal to 1.
So, it is critically damped fine similarly if this is greater than 2 then zeta is greater than 1
then it is the case of over damped response fine. So, our first conclusion is the overall
response become second-order and it may be over damped may be under damped may be
critically damped depending on the value of this zeta.

Process Control and Instrumentation


Prof. A.K.Jana
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture - 14
Dynamic Behavior of Chemical Processes (Contd.)
(Refer Slide Time: 01:00)

Today, we will start the topic that is dynamics of higher-order systems. Previously, we
discuss the dynamics of first order system then the dynamics of second order system. So,
today we will discuss dynamics of higher than second order system. Now systems higher
than second order dynamics are not uncommon in chemical processes systems with
higher than second order dynamics are not uncommon in chemical process example
include the processes having n first order systems connected in series.
So, we can say that we are presently interested to discuss n order systems dynamics here
n is greater than 2 now, n order system dynamics i mean the examples includes processes
having n first order systems connected in series. Second example is processes with deadtime examples includes processes with dead time examples include processes with
inverse response. So, these are the examples of n order systems and all these three
processes will discuss one by one. So, first we will discuss n first order processes
connected in series first we will discuss n capacities in series fine first we used to discuss
n capacities connected in series. So, it is it has basically we can classified it into two
different systems one is n non interacting capacities in series another one is n interacting

capacities in series. So, it has it is two different I mean we can classify these into two
classes one is n capacities n non interacting capacities and another one is n interacting
capacities.
So, if we consider n non interacting capacities then we need to connect n first order
systems in series. So, we will connect this n number of first order systems the transfer
function of the first system is G1 S input to this transfer function is say F1 bar S output
of these first system is y1 bar S. The second first order system has the transfer function
of G2 S. Output of these Second first order system is y2 bar S like this we place the last
or n first order system having transfer function of n input to this n first order system is y
n minus 1 bar S output is y n bar S. This is n non interacting capacities connecting in
series.
(Refer Slide Time: 06:52)

So, what will be the overall transfer function? Overall transfer function of these n non
interacting capacities is represented by y n bar S divided by F1 bar S which is equal to
G1 S multiplied by G2 S like this Gn S fine multiplication of transfer functions from G1
to Gn.
Now, what is the transfer function of a first order system? I mean G1 equals Kp 1
divided by tow p1 S plus 1 this we form we derived earlier similarly for the second first
order system we can write Kp2 divided by tow p 2 S plus 1 and last transfer function is

represented by Kpn divided by tow pn S plus 1 fine. So, this is the overall transfer
function for n non interacting capacities in series.
Similarly, for the case of n interacting capacities in series we obtain more complex form
fine for the n interacting capacities in series we can obtain the overall transfer function in
more complex form. Now, we will conclude based on the dynamics presenting in a plot.
So, this is the dynamics of different systems here we will consider height divided by Kp
say this is the steady state value.
So, for the first order system we get the dynamics like this. This is for the two noninteracting tanks in series this is for two interacting tank in series and this is for four non
interacting tank in series. This is for this is the dynamics of first order system see this is
the dynamic this is the representation of dynamics of a series of tank system.
So, the first curve represents the dynamics of a single liquid tank system the second
curve represents the dynamic behavior of the two non-interacting tanks system this is the
dynamic behavior of two interacting tanks system and this is the Dynamics of four non
interacting tanks system. So, this representation of Dynamics of liquid tank system.
If we consider a single liquid tank system then we get this type of behavior for two noninteracting two interacting and four non interacting tank system we obtain S staped
behavior. And as I mentioned in the last class this h represents the height of first tank for
single tank system the height h represents the height of second tank and Kp represents
the gain of second tank for two tank system height h represents height Kp represents gain
for fourth tank for four tank systems.

(Refer Slide Time: 12:31)

Now, we used to conclude based on these observation. So, first we will consider the noninteracting case. So, as I mentioned that we got S-Shaped response that means, the
response as that characteristics of over damped system. So, we can say that the response
has the character characteristics of an over damped systems that is the S-Shaped
behavior.
And the response is also sluggish over damped means zeta is greater than 1 the overall
response is sluggish. This is the first remark fine, second one is increasing the number of
capacities increases the sluggishness of the response. If you see if you compare the
dynamic behavior of two non-interacting tanks and four non interacting tanks. If, you see
the behavior of two non-interacting tanks and four non interacting tanks then we can
conclude that increasing the number of capacities increases the sluggishness of the
response agree. So, this is the second comment increasing the number of capacities in
series increases the sluggishness of the response sluggishness. This is the second
conclusion we can make for non-interacting capacities systems.
Similarly, if we can conclude for the interacting case as interacting increases the
sluggishness of the response if you again compare the dynamic behavior of two noninteracting tanks and two interacting tanks which is providing slow response two
interacting tank. So, we can say we can conclude that interaction increases the
sluggishness the response. We can conclude that interaction increases the sluggishness of

the overall response this is the conclusion we can draw for the case of interacting tanks
system.
(Refer Slide Time: 16:35)

So, next we will discuss the processes with dead time which is also an example of higher
order system next we will consider the processes with dead time. To discuss this topic
we will first take an example say this is a Fi through which a liquid is flowing this is a Fi
through which a liquid is flowing with the inflow rate of Fi and say the outflow rate F
naught. Now, we are introducing a stape change at a particular time instant suppose we
are introducing a stape change in Fi at time t equals to 0. Initially if it is at steady state I
mean there was no change of flow rate with time now we are introducing a stape change
in Fi.
So, suddenly we cannot get any change in the output of F naught. If, we give a stape
change at time t equal to 0 in Fi at that time instant we cannot get any change in Fi. So,
we can also make one plot for F naught verses t this is suppose representing time t equals
0 F naught changes after their time duration of td fine. This td is the time delay.
So, if we introduce a stape change in Fi at time t equals 0 at that time instant we do not
get any change in F naught. F naught changes a after time interval represented by t suffix
d that is the time delay.

Now, we will represent the time delay writing the number of blocks like we can consider
one block for a process input to this process is suppose Ft fine output of the process is
say yt now we need to include one block for dead time this time delay we again called
dead time. So, one block we need to include for dead time. What will be the output of
this dead time block? Output will be yt minus td if td is the dead time.
So, td is time delay or dead time this block is representing a process input to the process
is Ft output from the process is yt and another block we have included for the dead time
output for the dead time block is yt minus td now, we will represent this block diagram
considering the variables in Laplace domain. So, this will be the process.
Say this is a first order process. So, we can write the transfer function Kp divided by tow
p S plus 1 this is the transfer function of the process and we can say this is the transfer
function of the first order process. Input to this process in is domain we can represent by
this form Laplace transform of Ft output of this process we can represent by y bar S and,
this is the dead time block output we can write Laplace transform of yt minus td. This is
basically Laplace transform of yt fine and this block represent the Dead time in Laplace
domain.
(Refer Slide Time: 23:22)

So, what will be the transfer function of dead time. What will be the dead time of
Laplace domain? Laplace transform of dead time we can find by this way Laplace
transform of yt minus td divided by Laplace transform of yt. So, this is exponentials of

minus td S y bar S divided by y bar s that means, exponentials of minus tds. So, this is
the transfer function of dead time block. So, we can put here the transfer function as
exponentials of minus tds.
Next we need to determine the overall transfer function. What will be the overall transfer
function? So, if we represent the overall transfer function by G naught S then G naught S
equal Laplace transform of yt minus td divided by Laplace transform of Ft agree. This is
the output of the overall system and this is the input of the overall system in Laplace
domain if we consider first order process then the overall transfer function becomes Kp
exponentials of minus tds divided by tow p S plus 1 if we consider the first order process
then the overall transfer function becomes Kp exponentials of minus tds divided by tow
p S plus 1 this is called first order plus dead time system.
(Refer Slide Time: 26:10)

Similarly, if we consider second order process then the overall transfer function we can
write as G naught S equals Laplace transform of yt minus td divided by Laplace
transform of Ft Kp exponentials of minus tds divided by tow square S square plus 2 zeta
tow S plus 1 this is the second order plus dead time system.
Now, we want to represent it graphically the first order and second order plus dead time
systems this is time and this is suppose y this is the steady state of y. So, this value is
representing yss that means, steady state value of y. Now, if there is no dead time then
the response is like this if we (Refer Time: 28:06) the dead time then the response is like

this. This is the response of first order system and if we consider the dead time then it
becomes like this.
So, this is the first order process response with introducing it a stape change in input
variable this is the First order Dynamics with introducing a stape change in input
variable and this is the dynamic representation coin we this reader the dead time. So, no
dead time I mean no td. So, this difference is the dead time. So, this is the Dynamics of
First order system with dead time and without dead time. Similarly if we consider the
second order system this is time this is y this value is the steady state value of y then for
the second order system we get this type of response definitely this is the second order
underdem response fine zeta less than 1 if there is dead time then the response is like
this. So, this is representing the second order process response fine this is the Dynamics
of second order response and there is no dead time and this is Dynamics of second order
response and dead time is there.
(Refer Slide Time: 31:34)

Now, it was discussed earlier that and we commented that it is convenient to represent
the transfer function as the ratio of two polynomials it is easy to analyze the transfer
function by representing it as the ratio of two polynomials and that is like G naught S
equals Q S divided by P S if we represent the transfer function by this form it is easy to
analyze.

Now, if there is a dead time then how we can represent the overall transfer function by
the ratio of two polynomials that we will discuss next. So are title is polynomial
Approximation to the dead time term I mean exponentials of minus tds. So, we can
approximate this exponentials term by Pade Approximation. By Pade Approximation we
can approximate the exponentials term the forms are like this exponentials of minus tds
is approximately equal to 1 minus td divided by 2 S divided by 1 plus td divided by 2 S
we can write the exponential term by this form and this is called First order Pade
Approximation.
Similarly, if we can write the form Second order Pade Approximation 1 minus td divided
by 2 S plus td square S square divided by 12 1 plus td divided by 2 S plus td square S
square divided by 12 we can write the exponentials term by this form and this is called
Second order Pade approximation fine by another way we can also approximate that is
Taylor series expansion by Taylor method also we can approximate this exponential
term.
(Refer Slide Time: 35:12)

So, that is the second method Tayler series expansion according to this the exponentials
of minus tds is approximately equal to 1 minus tds the exponentials of minus tds is
approximately equal to 1 minus tds this is also First order Approximation this is first
order Tayler Approximation. It as another form the Taylers by using Tayler method we
can write this as 1 by exponential of tds which is equal to 1 divided by 1 plus tds

exponentials of minus tds equals 1 by exponential of tds and that is equal to 1 divided by
1 plus tds this is also first order approximation first order Taylor approximation. In the
next we will discuss in (Refer Time: 37:10) I mean how we can represent the higher
order transfer function to lower order transfer function? That we used to discuss.
(Refer Slide Time: 37:25)

Approximation of Higher order transfer function to lower order transfer function fine
next we used to discuss the approximation of Higher order transfer function to lower
order transfer function, but why will go for these? By doing this we can easily design the
controller. So, the first advantage is by performing the approximation of higher order to
lower order we can easily design the controller.
So, it is easy to design the controller by transforming Higher order transfer function to
lower order transfer function another advantage also it is easy to analyze it is easy to
analyze the lower order transfer function. Now, you just take one example to discuss this
approximation we will take one example to discuss this transformation suppose the
overall transfer function is even for a particular system by this forms equals the process
gain Kp minus 0.2 S plus 1 divided by 6 S plus 1 multiplied by 3 S plus 1 multiplied by
S plus 1.
To discuss the transformation of higher order to lower order transfer function we have
taken this example having the overall transfer function of these now we used to
transform this transfer function to a first order plus dead time system. So, we we used to

transform this transfer function in the form of Gs equals Kp exponentials of minus t ds


divided by tow pS plus 1. So, this is higher order transfer function and our purpose is to
transform these to lower order transfer function.
Now, this we can do by the use of Taylor method. So, first we will use the Taylor
method. Now according to the Taylor method minus 0.2 S plus 1 is approximately equal
to exponentials of minus 0.2 S agree according to the Taylor method we can write minus
0.2 S plus 1 is approximately equal to exponentials of minus 0.2 S. Similarly, we can
write 1 divided by 3 S plus 1 approximately equal to exponentials of minus 3S again we
can write 1 divided by S plus 1 equals exponentials of minus S.
(Refer Slide Time: 42:20)

Now, retaining the dominant time constant what is that? That is equal to 6 we get Gs
equal to Kp exponentials of minus 0.2 another one is 3 and 3rd one is 1 multiplied by S
and the largest time constant remains in the denominator. So, can we write this from the
higher order transfer function. So, it becomes Kp exponentials of minus 4.2 S divided by
6 S plus 1. So, this is the representation of higher order transfer function by the first
order plus dead time system transfer function.

(Refer Slide Time: 44:27)

This is one method another method we want to discuss that is Skogestads half rule
method using which we can also transfer the higher order transfer function to lower order
transfer function. So, next we will discuss another technique that is Skogestads half rule.
The overall transfer function which we have taken is represented again I am writing that
divided by 6S plus 1 3S plus 1 S plus 1. Now we will transform this transfer function to
first order plus dead time system by the use of Skogestads half rule. So, what is the
neglected highest neglected time constant.
The largest neglected time constant considered in the Taylor method is equal to 3 fine. In
the earlier discussion we have neglected this time constant which is largest one now,
according to this Skogestads half rule half of these I mean 1.5 is added to the next
largest time constant try to remember the stape in the Taylor method the largest
neglected time constant considered is 3.
According to this Skogestads half rule 50 percent of this value that is 1.5 is added to the
next largest time constant to the next largest time constant what is the next largest time
constant? 6 next largest time constant is 6 that means, the time constant becomes 6 plus
1.5 that is 7.5.
The time constant becomes 6 plus 1.5 that is 7.5 and the others half provides a new time
delay provides a new time delay of 1.5 that means, another half is considered as time

delay that will be added within the exponentials terms. Then, what will be the reduced
transfer function according to this half rule transfer function finally, becomes Kp
exponentials of minus 0.2 is there half of 3 that is 1.5 is added and 1 remains there the
time constant is already obtained 7.5.
So, the denominator becomes 7.5 S plus 1 that means, Kp exponentials of minus 2.7 S
divided by 7.5 S plus 1 it has been observed that this transfer function obtained by the
use of Skogestads Half Rule provides better performance then the Taylor than the
transfer function obtained by the use of Taylor method.
It has been observed that the performance provided by the reduced transfer function
obtained by the use of Skogestads Half Rule is better than the response provided by the
reduced transfer function obtained by the use of Taylor method.
(Refer Slide Time: 49:55)

So, next we will discuss the Dynamics of processes having Inverse Response. This type
of processes also provide nth order Dynamics. Now, this type of behavior is observed
when there is a positive 0 this is the main point associated with this topic. This type of
behavior I mean the Inverse Response behavior is observed when there exists a positive
0.
So, we will consider a process we will consider a two processes acting in parallel one has
the transfer function of K1 divided by tow p1 S plus 1 another one has the transfer

function of K2 divided by tow p two S plus 1. Now, we want to find the condition at
which this process shows Inverse Response behavior we used to find the condition at
which this process provides Inverse Response behavior. So, input to this process is f bar
S and this process is output is y bar S.
(Refer Slide Time: 52:52)

The two processes are acting in parallel. Now, first we need to find the overall transfer
function the overall transfer function y bar S divided by f bar S which is equal to K1
divided by tow p 1 S plus 1 this is Kp 1 and another transfer function is Kp 2 divided by
tow p two S plus 1. Now, we get from this Kp 1 plus Kp 2 multiplied by Kp 1 tow p 2
plus Kp 2 tow p 1 divided by Kp 1 plus Kp 2 plus 1 whole divided by tow p 1 S plus 1
tow p 2 S plus 1.
We get this transfer function this overall transfer function now, we can represent this
again as Kp tow S plus 1 divided by tow p 1 S plus 1 multiplied with tow p 2 S plus 1.
Here Kp is Kp 1 plus Kp 2 where Kp is Kp 1 plus Kp 2 and tow is Kp 1 tow p 2 plus Kp
2 tow p 1 divided by K. This is the tow. So, for positive 0 tow should be less than 0 agree
that means, we can write Kp 1 tow p 2 plus Kp 2 tow p 1 less than 0 agree.
So, rearranging this again we obtain Kp 2 divided by Kp 1 greater than tow p 2 divided
by tow p 1 rearranging finally, we get this condition. So, at this condition the process
shows Inverse Response behavior. So, Inverse Response is the net result of the two
opposing effects if you see the sign tow p 1 tow p 2 both are positive. So, if one Kp is

positive another Kp should be negative. So, we can conclude that Inverse Response is the
net result of the two opposing effects.

Process Control and Instrumentation


Prof. A. K. Jana
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture - 15
Feedback Control Schemes
(Refer Slide Time: 01:00)

Now, we start the topic introduction to Feedback Control. We will discuss the topic
introduction to Feedback Control. Now, before discussing the feedback controlling
schemes, we will develop the block diagram for a closed loop system. So, development
of block diagram, of a closed loop process we have develop the block diagram for open
loop process. So, now we used to develop the block diagram of closed loop process, for
this purpose we will consider one example that is hitting tank system, for the
development of block diagram, we consider the hitting tank system.
Steam is introduced through this coil with a flow rate of f s t in food flow rate is f i and
temperature is T i outlet stream is going out at the flow rate of f and temperature T. Now,
what is the control of objective the control of objective is to maintain the temperature at
it is desired value. The control of objective of this example process is to maintain the
temperature at it is desired set point value and we assuming that, f i and if their identical
and their constant quantities. We are assuming that a f i if both are identical and their
constant.

So, what is the control variable, what is the manipulated variable and what is the load
variable, that we can detect. Now, control variable is temperature, what is the
manipulated variable, corresponding manipulated variable is steam flow rate and load
variable is T i. Now, this is the schematic of an open loop system, this is the schematic of
the open loop hitting tank system. Now, we have assumed that the inlet and outlet flow
rates are identical and they are also constant, it indicates there is no variation of height, it
indicates height is also constant. Now, we will configure the control around this process,
we have already dictated that control variable, manipulated variable and load variable.
Now, we will configure the feedback control scheme for maintaining the liquid
temperature.
(Refer Slide Time: 06:19)

So, this the hitting tank system we considered as an example, now first we need to
measure the temperature by using a measuring device, we need to measure the
temperature using a measuring device. So, this is the block for temperature sensor which
is employed to measure, the liquid temperature. Now, this measure temperature usually
defers, from the original temperature habit, this measure temperature we can represent by
t suffix n, in the next step this measure temperature is compared with it is set point value
represented by T s p. So, this is positive and this is negative.
Then the output of this comparator is the output of the comparator is represented by it is
epsilon, it is epsilon is equal to set point temperature minus measure temperature, the

comparator output can be represented by it is epsilon or epsilon is equal to T s p minus t


m. Then this information is supplied to the controller, this information is supplied to the
controller, based on the error signal the controller takes action and that action is
physically implemented to these control hall, the controller action is physically
implemented to the final control element that is control hall.
This is the configuration of the controller, employed around this hitting tank system. And
this is the schematic of the closed loop hitting tank system, this is the schematic of the
closed loop hitting tank system or closed loop process, by the introduction of the
controller the loop is closed. Now, in the next week we used to develop the block
diagram for this closed loop process. So, we will consider a block process, this block is
representing the process, input to this process is suppose, m bar is input to the process is
represented by m bar s that is nothing, but the final control output.
And another input is considered, that is the disturbance d bar s. So, there are two inputs
affecting the process, one is the final control element output, another one is the
disturbance. Disturbance is represented by d and final control element is represented by
here m, the output of the process is y in Laplace domain we can write y bar s, this is the
open loop process, this is the block diagram of the open loop process, there is no
controller involve with these open loop process.
Now, what we doing first, we are first measuring the temperature using a measuring
device. So, one block we can draw for the measuring device, this measuring device is
measuring the temperature, basically this y is T only. The control variable y is here the
temperature of the liquid in the tank. So, first we measuring the temperature using this
measuring device. What is the output of this measuring device, output is y suffix m that
is nothing, but measure temperature sensor output is measure temperature T m.
Now, this measure temperature, this measurement signal goes to the comparator and the
sign is negative, you just compare the configuration with this block diagram. Now,
another input to this comparator is y said point, that is y suffix s p and we use positive
sign for this. So, there are two input signals considered for the comparator, what is the
comparator output, epsilon in Laplace domain we can write epsilon bar, this comparator
output is supplied to the controller.

So, one block we can draw for the controller, the error signal epsilon goes to the
controller, then the controller calculates the control action. We can represent the
controller output by c and in Laplace domain c bar, this controller output goes to the final
control element, that is the control half, the controller output c is supplied to the
controller output c goes to the final control element and final control element output is
m.
So, this the block diagram of the closed loop process this the process which as two
inputs, one is the final control element output and another one is the disturbance, these
two inputs effect the process. If we consider the liquid tank system hitting tank system,
the output is the temperature this temperature is first measured by using one measuring
sensor.
Sensor output is T suffix m that is measure temperature, then that measure temperature is
compared with it is said point value, comparator output is represented by here it is
epsilon, that information is goes to the controller. And using that information controller
calculates the control action, then control information is physically implemented to this
final control element. So, this the closed loop block diagram of a process.
(Refer Slide Time: 16:39)

Next we will discuss the different types of controllers. Now, if we redraw the controller
block, then we see that input is m bar s output is sorry input is epsilon bar and output is c
bar s. So, the basically the controller relates c to it is epsilon. So, various types of

controller differ in the o a they relate c to it is epsilon. So, first we will consider three
types of controller, which are classical controllers and which are definitely fee back
controllers.
These three controllers are proportional controller, this is also called I mean this is also
named as p only controller. Second controller is Proportional Integral controller, which
we can call as PI controller, proportional integral. Proportional integral controller
additional includes the integral term. Third controller is Proportional Integral Derivative
controller, which we can call as PID controller, these are the basic feedback controllers.
And we will discuss one by one so, first we will consider proportional controller.
(Refer Slide Time: 19:45)

First we will discuss proportional or p only controller the controller output c which can
be written in terms of divisional variables as c prime t is proportional to the error signal.
If you see the control block, will observe that its epsilon t is the input and c t is the
output. Now, we have written c in terms of divisional variables. So, c prime t is
proportional to the error signal that is the input to the controller and that is why the name
I mean that is why it is called proportional controller. So, you can write this against as c
prime t equals k c it is epsilon t.
And finally, we write c t equals c s plus k c it is epsilon t or c prime t is equal to c t
minus c s. I have mentioned that c prime is the divisional variable. So, c prime t we can
write as c t minus c s. Now, k c is the proportional gain of the controller. It is very

obvious you see k c we can write as c prime t divide by epsilon t can we write this, k c is
equal to c prime t is divided by epsilon t you just see this block c t is the output. If
epsilon t is the input I have already mentioned that gain is change in output per units
change in input, it is obvious in this correlation.
So, we can say k c as the proportional gain. Now, what is c s, c s is the bias signal,
controllers bias signal. We can define it as c s is the controller output and there is no
error, if you see this equation. We can write that c t is equal to c s when the error is 0. So,
bias signal is the controller output and error equals 0. This k c is one tuning parameter
controller tuning parameter, the value of that parameter we need to determine. So, one
equivalent term is also used in different soft wares and industrial factories, that is
proportional band.
One equivalent term of k c called proportional band is also used in different process
simulator and in industrial practice. Proportional band, which is 100 divided by k c
proportional band is 100 divided by k c. Now, we used to determine the trans perform of
the controller.
(Refer Slide Time: 25:20)

You see for the p only controller we got c prime t is equal to k c it is epsilon t. Now, if
we take the Laplace transform for this, then we obtain c prime bar s equals k c epsilon s,
if we take Laplace transform we obtain c prime bar s is equal to k c epsilon s. Now, what
is the output of the controller, c prime bar s what is the input to the controller epsilon s,

and this can be represented as transfer function of the controller G c s and G c s is equal
to k c. So, for PI controller transfer function is sorry for the p only controller transfer
function is equal to k c.
Now, we will conclude the p only controller, with the increase of k c the controller
becomes more sensitive to error, this is the first conclusion which we can draw for the p
only controller. With the increase of k c the controller becomes more sensitive to error
epsilon. If we increase the k c value the controller become more sensitive to it is epsilon.
Now, we want to see the behavior of the p only controller, how the controller behaves
that we want to represent graphically.
So, remember the controller equation that is c t equals c s plus k c epsilon t this is the p
only controller equation in time domain. Now, if we plot c verses epsilon, we obtain the
c s when it is epsilon is equal to 0. So, this pointing is representing c s because, we know
that c t equals to c s when it is epsilon is equal to 0. So, we are considering this or epsilon
is equal to 0, so the corresponding c value is c s and when we increase epsilon, it
increase like this, it is a slope of k c. So, slope is k c.
And the last value of c is c max maximum value of c we cannot consider c as infinity due
to the some physical limitation of final control element, we have to put constants on c.
So, the higher limit is c max and lower limit is c min. Now, this variation is obtained
when it is epsilon is positive similarly, we obtain this line, when it is epsilon is negative
and the minimum value of c is c min. So, this is the second conclusion on p only
controller. Next we will discuss how the controller response, which it is take place in
error signal.

(Refer Slide Time: 31:02)

So, we will use the controller equation in Laplace domain. This is the p only controller,
equation in Laplace domain. Now, we will consider a unit place change, we will consider
a unit step change in error signal, then the controller equation is c bar prime s is equal to
k c by s. If we take in words of transform, then we get c prime t equals k c if we take in
terms of transform we obtain c prime t equals k c; that means, c t is equal to c s plus k c.
Finally we get c t equals c s plus k c.
Now, we will produce the plot to represent this concept, this is time t and this is c t. Now,
initially the process was, initially the controller has the value of c s. Now, we are
introducing a unit step change at time t equals 0, then the controller output provides the
value of c s plus k c. This is the controller behavior against a unit step change in error
signal. Next we will discuss the second controller that is PI controller Proportional
Integral controller.

(Refer Slide Time: 33:50)

Next we will discuss, proportional integral controller c t equals c s plus k c it is epsilon t


this the representation of p only controller. Now, additionally we include the integral
term in poly n controller, to obtain PI controller. That integral term is represented by k c
divide by tau i integration of error t d t. So, this is the integral action and this the
proportional action, combining this two actions we obtain t i controller.
Now, we can represent this terms in divisional variables for finding the trans form
function of PI controller. Accordingly we write, c prime t equals k c it is epsilon t plus k
c divided by tau i integration of error t d t by taking Laplace transform and rearranging
we obtain the transform function, as G c s equals c bar prime s divided by it is epsilon by
s equals k c 1 plus 1 divided by tau i s by taking Laplace transform and rearranging
finally, we obtain the transfer function of PI controller as G c s k c multiplied by k c 1
plus 1 divide by tau i s this is the transfer function of PI controller. Now, we will make
some demarks on these PI controller.

(Refer Slide Time: 37:04)

The integral action eliminates the offset, first conclusion is the integral action eliminates
the offset. Offset is basically the steady state error. If we plot y verses time t we may get
this type of response under a controller, this is the set point value and the value who
which we are getting finally, that is the steady state value. So, this corresponding value,
this value is the steady state value y steady state. Now, the difference between the set
point value and steady value is called offset.
So, offset is basically y set point minus y steady state. Now, this theory controller can
eliminate the offset due to the inclusion of integral action I mean p only controller cannot
eliminate the offset due to the inclusion of integral action. I mean the p only controller
can not eliminate offset, this is the first conclusion. Now, due to the additional of integral
action, the overall response becomes more sluggish, this is the second conclusion due to
the addition of integral action the overall response become more sluggish, due to the
increase of order by one due to the increase of integral action the overall response
becomes more sluggish.
Say for example, if we consider a first order system and if we employ the PI controller,
then the overall system becomes second order system. If we consider a first order system
and if a controller is employed, then the overall system response becomes second order
dynamites, that means order is increased by one. Now, in plot we have compared the
dynamites of first order, second order and fourth order systems and we have observed

that increasing the number of capacities in series, increases the sluggishness. So, by that
line we can say that by inclusion of integral action the order is increased by one, it means
the overall response becomes more sluggish. So, this is the second conclusion. Now, we
will consider a unit step change in error signal and we will observe the response of the PI
controller.
(Refer Slide Time: 41:56)

So, PI controller equation in Laplace domain, we can write as c bar prime s equals it is
epsilon s multiplied by k c 1 plus 1 divided by tau i s can we write this, the PI controller
in Laplace domain. Now, we will consider a unit step change in error signal. We will
consider a unit step change in error signal; that means, epsilon bar s equals 1 by s, then
we obtain c bar prime s equals k c divided by s plus k c divided by tau i 1 by s square.
Now, inverting this, we obtain c prime t equals k c plus k c divided by tau i multiplied by
t. Taking inverse of Laplace transform we obtain finally, c prime t equals k c plus k c
divided by tau i multiplied by t. Now, we will represent this graphically, this is time t this
is c t. Now, at time t equals 0 the controller as the value of c s and we are introducing a
unit step change at time t equal was 0 accordingly the controller output becomes c s plus
k c.
So, c s plus k c we can obtain from this c s plus k c plus k c divided by tau i into t. So, it
is obvious that the time t equals 0 c t equals c s plus k c and when time is greater than 0
the c increases linearly, with a slope of a c divided by tau i slope equals k c divided by

tau i. This is the behavior of PI controller against the unit step change in error signal.
And that error signal the change in error signal introduced at time t equals 0.
(Refer Slide Time: 45:26)

Now, this integral action, we can write as k c divided by tau i epsilon t d t, k c is the
proportional gain and tau i is the integral time constant, one additional tuning parameters
is there, that is tau i and tau i is called as integral time constant, it is also called as reset
time, unit of tau i is time.
Now, for the PI controller we got c prime t equals k c plus k c divided by tau i into t by
introducing a unit step change in error signal. We obtain for PI controller c prime t plus k
c plus k c divided by tau i into t. Now, if we consider t equals tau i. The controller output
is c prime t equals k c plus k c. So, this term is given this yielding k c understood or not,
for the PI controller one additional term is there that is integral action. The integral
action can be represented by this form, integral action equals k c divided by tau i
integration of error t d t.
Now, the tuning parameter k c is present also for p only controller. Now, the additional
term tau i is present with this PI controller that tau i is called integral time constant or
reset time. The unit of these integral time constant is time. Now, we obtain these form c
prime t equals k c plus k c divided by tau i into t considering it a unit step change in error
signal this form we obtain.

Now, we are just considering time t equals tau i then the controller output is c prime t
equals to k c can we define now reset time. What is reset time, reset time is a time
required for the controller, reset time is the time required by the controller to repeat the
initial proportional action change in it is output you see the initially we had one
proportional gain that is k c. Now, due to the addition of integral term and considering
time equals tau i we are getting another proportional action, and therefore we can say
that, the reset time is the time required by the controller to repeat the initial proportional
action change, in it is output.
Thank you.

Process Control and Instrumentation


Prof. A.K. Jana
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture - 16
Feedback Control Schemes (contd.)
So, in the last class we discussed two control schemes. one is P-only controller and
another one is PI controller.
(Refer Slide Time: 01:06)

So, replace we discuss P-only controller, followed by PI controller we made few


conclusions on this P-only controller. Like replace conclusion on P-only controller it
speeds of the response of the close look process, it speeds up the response. Fine and with
the increase of Kc values the controller becomes more sensitive to error signal, with the
increase of Kc values the controller becomes more sensitive to error signal with the
increase of the Kc to error signal Fine. The P-only controller does not eliminate offset,
third conclusion the P-only controller, or proportional action we can say does not
eliminate offset. P-only controller does not eliminate offset. The PI controller consists of
proportional action and integral action. PI controller includes two control actions those
are proportional action and integral action. So, if we consider the integral action then, we
can say that integral action eliminates the offset.

This integral action eliminates the offset, and that is why is this integral tom is included
with P-only controller to make PI controller. Now, today will study the elimination of
offset by PI controller. How the PI controller eliminates the offset that we want to
absorb?
(Refer Slide Time: 04:54)

Now, we will discuss the elimination of offset, What is offset? As mentioned earlier
offset is described as steady state deviations of the controlled variable from set point.
Offset is the steady state devitaions, it is mention as earlier that the offset is defined as
steady state deviations of the controlled variable form set point. The offset is defined as
steady state deviations of the controlled variable form set point. Now, we will consider
one example this is a tank heater Fi is the inlet steam flouride and Ti is the tempertaure
of the inlet steam, the outlet is coming out with flouride F and temperture T, height of
liquid in the tank is h and temperture is T
A heating medium a steam is introduced through this coil with a fluoride of if suffix s T.
Now, we are assuming here, that the inlet and outlet fluoride are identical and both are
constants. We are assuming the inlet fluoride and outlet fluoride as constants and there
both are identical. So, what it indicates there is no variation of height. So, height is
constant if the both the inlet and outlet fluorides are equal and constants then there is no
variation of liquid height. So, for this process what is controlled variable which one is
controlled variable temperature, manipulated variable is steam fluoride which one is load

variable. Fi is constant that is why load variable is Ti Fi is constant so Ti is the only load
variable. Now, we are considering the steady state conditions as Ti equals 25 degrees 60
Celsius T 60 degrees Celsius and at steady state the steam valve is 50 percent open, these
are steady state conditions. Next we will consider state change in the load variable,
another thing I want to mention here, that this valve is ear to open. It means if, the input
pressure is increased the valve opening increases; with the increase of inlet pressure, the
valve opening is increases.
(Refer Slide Time: 10:51)

Next we will consider the step change in load variable, step change in load variable. As
mentioned Ti is the load variable and steady state Ti is 25 degree Celsius. Now, we will
introduce a step change in Ti that is represented in this plot this is Time, this is Ti,
initially Ti is 25 degree Celsius now it is changed to Suppose 20 degree Celsius at time
t equals 0. We are introducing the step change in Ti that is 25 degree Celsius to 20 degree
Celsius. Now we will Note down few points, the larger the values of Kc, the smaller the
offset. values If we introduce this step change I mean 25 to 20 degree Celsius, then the
process response under P-only controller like this, is t this is temperature. What is steady
step values, values of temperature? 60 degree Celsius Now, if we introduced a step
change at say Time t equals 0, the close look process responds like this, this at a
particular Kc values.

So, we can write the title as process response, under P-only controller. If P-only
controller is employed around that tank heater now, we introduced a step change in Ti
then the process I mean close loop process responds in this manner. If we consider
another Kc values, then process responds like this. If we consider a different proportional
gain values then it responds like this. Now here Kc 1 is greater than Kc 2, this is the
observation if we introduced a step change in Ti the tank heater system under P-only
controller responds in this way. That is why, we concluded here that the larger values of
Kc the smaller the offset. This the offset for the case Kc 1 this the offset, and for Kc 2
this is the offset. Now if this is the conclusion I mean larger values of Kc the smaller the
offset the
Why not set maximum gain? Immediately one question arise is why not set a maximum
gain, for most of the process is there is maximum values of Kc there is a limit of Kc for
the most of the chemical process is, there is a maximum values of Kc that we can
represent by Kc max. If, we consider Kc greater than Kc max then, the process goes
unstable for the most of the process is there is maximum values of Kc represented by Kc
max if, we consider Kc greater than Kc max there is high probability of in stability
problem. This Kc max is called ultimate Kc, this Kc max called ultimate gain.
Represented by Kcu. So, we cannot consider arbitrarily large Kc values to reduce the
offset, there certain limitation. Now, with the decrease of tow I what happens. In this
case I mean we have consider decrease of tow I if, we decrease tow I if we decrease of
tow I the temperature of liquid in the tank decreases agreed.
If Ti is in decreased the temperature of the liquid in the tank also decreases then, what is
require to do for the controller? To keep the temperature at 60 degree Celsius, the
controller needs to increase the steam fluoride. Now, we have consider at steady state the
valve is 50 percent open but, if Ti is decreased to 20 degree Celsius then suppose the
controller needs to open the value 60 percent.

(Refer Slide Time: 18:45)

At steady state we have considered 50 percent opening of steam valve if the Ti changes
from 25 to 20 degree Celsius, in that case suppose it is require to open it is require 60
percent open of the valve to maintain temperature at 60 degree Celsius. Now, our
manipulated variable is Fst if, we consider P-only controller then we can write Fst equals
Fsts plus Kc error t for the case of P-only controller, our manipulated variable is Fst. So,
if we consider P-only controller then we can represent we can write this equation Fst
equals Fst s plus Kc (Refer Time: 20:15) t. What is Fst s? Steady state values of Fst. If
we represent Fst in terms of percentage opening then, Fst s is nothing but 50 percent
because ,that is steady state values of Fst. How much the 2nd term will be? How much
will Kc epsilon t? Remaining 10 percent, agree or not a Fst as is the steady state values
of Fst.
We have consider 50 percent opening at steady state. So, Fsts we can write as 50 percent
and rest Kc epsilon t will be reaming 10 percent. So What it indicates? I mean Kc epsilon
t equals 10 percent. So, how much epsilon t? At least epsilon Tis not equals to 0. If Kc
epsilon t equals 10 percent then epsilon t not equals to 0 because the controller gain is
non 0 since Kc is non 0. So, epsilon t not equals to 0, it indicates there will be an offset
because error is error not equals to 0 it indicates there is an offset under P-only
controller. Next we will consider the PI controller, we will continue the same example
and we will consider.

(Refer Slide Time: 22:41)

PI controller as long as the error is present the PI controller keeps changing its output by
integrating the error. If, we see the controller equation Fst equals Fst s, plus proportional
term Kc epsilon t plus integral term; that means, Kc divided by tow y integration error dt.
This the PI controller equation. So, this is the integral term as long as the error is present
the PI controller keeps changing its output by integrating the error.
Now, if we consider Fst 60 percent. How much is Fst s? 50 percent, How much is this 1?
PI controllers eliminate the offset. If that is the case then error should be 0. So, Kc if
silent is 0 and this term, is 10 percent. If we see the plot, we observe that the PI
controller the process under PI controller reaches at steady state values I mean the
process PI controller remains at steady state temperature. But for the case of P-only
controller there is an offset. So, this is the process response under P-only controller and.
So, this is the process response under PI only controller. Now, you see if we introduced
state change in PI from t 25 to 20 degree Celsius then there is need to open the steam
value60s percent opening there is need to maintain 60 percent opening of steam valve.
So, this is 60 percent and steady state values is 50 percent if, PI controller eliminates the
offset then definitely error should be 0 that means this is 0. So, remaining 10 percent is
indicating the integral term this remaining 10 percent is provided by the integral term.
So, it is not 0. It remains constant at, the last values I mean epsilon at a last Times step is
0. But the sum of remaining errors, pervious errors provides 10 percent the last error that
is epsilon t equal to 0 and the sum of pervious errors, provides 10 percent opening. So,

this indicates that the error I mean, steady state reduced to 0 under PI controller. So, we
will discuss this with more rigors proof later.
(Refer Slide Time: 27:51)

Next topic is reset windup of reset windup happens for PI controller, Now, for PI
controller we know that, as long as error is present PI controller keeps changing it is
output by integrating the error. For PI controller we know that as long as error is present
the controller keeps changing it is output integrating the error. As long as error is present,
the PI controller keeps changing it i s output by integrating the error. When a sustained
error is happens the integral term becomes quite large and controller output eventually
saturates an a sustain error
Further buildup of integral term is referred to as reset windup further buildup of integral
term, is referred to as reset windup. So, as long as error is present the controller keeps
changing it is output integrating the error minus sustain error occurs integral term
becomes quite large, and the controller output eventually saturates, further buildup of
integral term is called reset windup. Now, to reduce the reset windup there need to devise
anti reset windup.
So, there are many os to devise this. So, one we will discuss here, I mean one we will
mention here reset windup is reduced by temporarily halting the integral action.
Whenever, the controller output saturates, reset windup is reduced by temporarily
halting, the integral control action whenever the controller output saturates this is the one

way to devise the anti reset windup. So, next we discuss the third controller that is PID
controller proportional integral derivative controller.
(Refer Slide Time: 33:28)

Next we will discuss proportional integral derivative controller. In soft this is PID
controller the name suggest that, the controller includes three actions proportional
actions, integral actions, and derivative actions. And if, we consider manipulated variable
if we represent manipulated variable as c, the PID controller can written as Ct equals C
plus Kc epsilon t plus Kc divided by tow y integration error dt plus Kc tow d, d error dt
this the proportional action this is the integral actions and this one is the derivative
action. and C s is the bios signal I mean the error is 0 the controller output is C s.
So, we can we can rearrange this equation as C prime t, equals Kc error t plus 1 divided
by tow i integration of error dt plus tow d, d error dt we can rearrangePID controller in
this form. Now, we should derive the transverse function of PID controller by taking lap
plus transform if we taking lap plus transform and if, we rearrange then we get is C bar
prime is divided by epsilon bar is which is Gc and Gc equals Kc, 1 plus 1 divided by tow
is plus tow ds this is the transverse function of PID controller.
Now, you observe one thing that, if the error is non 0 What about, the derivative action?
If error is constant and non 0 what about the derivative action? 0 there is no existence of
derivative action. Because derivative action is equals to Kc tow D dr dt if error is
constant then the differentiation of that with respect to Time becomes 0. So, this point we

can note down as replace point. Second if for a noise response with almost 0 error. For a
noise response I mean for noise y, because it is Siloam is y set point minus y.
So, noise response means noise y, with almost 0 error D term leads to large control
action. Although that is not needed for noise response 0 error d term leads to large action
derivative term leads tolarge control action although that is not required. So, you can say
this unnecessary you can say this is the drawback this two are the drawbacks of the PID
controller, and error is constant there is no action from the derivative action, and there is
noisy error the derivative action unnecessarily is large. So, these two are the drawbacks
of the PID controller. Next we will discuss the control actions one is deduct action
another one is reverse action.
(Refer Slide Time: 39:51).

.
We will discuss next Direct and Reverse Action. So, as the input signal to the controller
increases the output signal from the controller should increase for direct acting controller
you note down this point, as the input signal to the controller increases the output signal
from controller must increase.
For direct acting controller, similarly we can say that, as the input signal to the controller
decreases the output signal from the controller must decrease, for direct acting controller.
So, we can say that if, the input signal increases the output signal should increase for
direct acting controller in reverse we can say that if, the input signals to the controller
decreases the output signal from the controller should decrease. Therefore, for deduct

acting controller these two terms are used increase another one is decrease we will take
one example to discuss this action direct action. This is a liquid tank system our
controlled variable is liquid height and manipulated variable is suppose if not, So, to
maintain height we need to employ one level controller, I am not showing the sensor
comparator extra here, direct deduct we put here the block for level controller. Now, if
height increases suppose if height increases. What is required to do for the controller?
The controller should increase if not or decrease if not, the controller should increase if
not. If height increases the controller should increase if not to maintain the height at it is
desired values. You see the input signal to the controller that is height, if increases the
controller is increasing the output signal if not. So, this is direct acting controller this
level controller is direct acting controller. The input signal height if, increases the
controller increases accordingly if not.
So, this direct acting controller. Now, we will write the controller equation. Suppose is
level controller is P-only controller according to the controller equation is written as if
naught equals if naught is, plus Kc height set point minus height suppose this level
controller is a P-only controller. Accordingly we can write the controller equation as, F
naught equals f naught is plus Kc error error means eighth set point minus 8. If height is
greater than 8 set point.
So, what will be F naught it will be greater than F naught s or less than F naught s it
should be greater the F naught s agree. If, height is greater than 8 set point. Then the
controller require to maintain this condition, I mean F naught should be greater than F
naught s. If the height is greater than 8 set point then what about this term positive or
negative? It is negative now, how it is possible to maintain f naught is greater than F not
is Kc if Kc is negative, if height goes above, set point values of 8. Then from this
equation it is clear that the error becomes negative. Now, controller is required to
maintain F naught is greater than F naught s that is only possible and Kc is negative. So,
we need to remember that we should remember that for direct acting controller gain is
negative. This is the most important feature of direct acting controller for direct acting
controller the controller gain should be negative.

(Refer Slide Time: 47:21)

.
Reverse- next we will discuss the reverse Action. It is just reverse of direct action. So,
the definition is as the input signal to the controller increases the output signals from
controller should decrease for reverse acting controller, as the input signal to the
controller increases. The output signals from controller should decreases for reverse
acting controller, or as the input signal to the controller decreases the output signals from
controller should increase for reverse acting controller. If we continue the same example,
What will be the manipulated variable? For reverse acting controller, height is the
controlled variable. So, if I should be the manipulated variable for reverse acting case.
You see if, we increase the height What is require to do for the controller? The controller
should decrease a Fi to maintain height at its desired values. If height increases the
controller should decrease a Fi to maintain height at it is desired values; That means, it is
reverse acting.
Now, we will write similarly the controller equation a Fi equals a Fi is plus Kc, h set
point minus 8. Now if, height is greater than 8 set point a Fi should be less than a Fi if
height is greater than 8 set point the controller should maintain this condition that a Fi is
less than a Fi. Now, if height is greater than its 8 set point then, the error becomes
negative. Now, how the controller can maintain the Fi less than a Fi s? If Kc is positive
agree if height is greater than 8 set point then the error becomes negative. Now, to
maintain a Fi less than Fi is Kc should be positive. So, similarly I mean like direct acting

controller unlike deduct acting controller the controller gain should be positive for
reverse acting controller.
(Refer Slide Time: 51:16)

Next we will discuss another controller that is ON-OFF Controller. ON-OFF controller
this is the simple inexpensive feedback controller. like, an ideal ON-OFF controller. We
can represent as, Ct equals Cmax if, error is greater than or equal to 0 C t is the controller
output. So, error is greater than 0 Ct provides maximum values Cmax; that means, we
considering control value C provides 100 percent opening. So, if we considering control
valve as a Final controlling element then Cmax indicates full opening of valve.
Similarly this Ct becomes Cmim if, error less than 0 the controller output becoming the
minimum value of C and error less than 0 if we consider control value as a Final
controlling element. This indicates full closing I mean this indicates fully closed, there is
no intermediate control action either maximum values or minimum values that is why the
name On-off. Now, we will considering few cases like, for digital computer
implementation Cmax indicates100 percent and Cmim indicates 0 percent. If, we
consider a current based electronic controller this, Cmax indicates 20 million empire and
Cmim 4 million empire, for current base electronic controller the maximum values is 20
million empire and minimum is 4 million empire similarly we consider numeric
controller pneumatic controller maximum values is 15 psig and minimum values is 3
psig.

If we consider pneumatic controller c max is 15 psig pressure and minimum is 3 psi g.


What is the use of this on-off controller? On-off controller is used, in home heating
systems. controller is used in domestic refrigerator, on-off controller is also used in
laboratory furnaces, Another thing we can say that if, Kc values is very large for P-only
controller then controller behaves like on-off controller. If Kc values is very large for Ponly controller then, the controller behaves like on-off controller. Definitely the
proportional controller are more efficient than on-off controller. The proportional
controllers like P, PID more efficient than on-off controller.

Process Control and Instrumentation


Prof. A. K. Jana
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture - 17
Feedback Control Schemes (Contd.)
In the last class we discussed three different control schemes, proportional controller
proportional integral controller and proportional integral derivative controllers.
(Refer Slide Time: 01:07)

We discussed P only controller having the transfer function of G c equals K c, we discussed P I


controller having the transfer function of G c equals K c 1 plus 1 divided by tau i s. We
discussed the P I D controller having the transfer function of G c equals K c 1 plus 1 divided by
tau s plus tau d s. Now, here K c is basically the proportional gain and this gain equals del output
I mean change in output bar unit change in input, so the unit of proportional gain K c depends on
the unit of output as well as the unit of input.
And tau i is the integral time constraint, the unit of tau i is time, I mean it may be minute may be
hour, so the unit of integral time constant is say minute. Another term which is involved in
derivative term that is tau d, tau d is derivative time constant, unit of derivative time constant is
minute or it may be hour fine. So, these three parameter are basically the control parameters, K c
tau i and tau d are the control parameters, tuning of this parameter will be discuss later. How you
can select the values of these three parameters that will be discuss later fine.

(Refer Slide Time: 04:53)

So, today we will start with the dynamic behavior of closed loop process, first we will discussed
in generalized form, then we will considered few specific cases, we will considered few
examples. Now, to discuss this we need to develop the block diagram of the close loop process,
the block diagram of the close loop process we already discuss in the previous class, so I am just
redrawing the block diagram of a typical closed loop process.
So, this block is representing the process having the transfer function of G suffix P, this block is
representing the process having the transfer function of G suffix P, in is the input to the process
m is the input to the process which is basically the output of final control element. Now, the
output of this is added with this output fine, this two outputs are added d is the disturbance and
G d is the transfer function of the process with respect to disturbance, output is y bar process
output is represented by y, and in Laplace doming that is y bar.
Now, what is run first the process output is measured using tape measuring devise, the process
output is measured using a measuring devise, say the measuring devise as the transferring
function of G m. Now, the measure output, usually defers from the actual output, so we will
represent the measured output by y m, then this measured output is compared with the step one
value of process output. The measured output is compared in the comparator with its step one
value, the output of this comparator is epsilon, and this output is supply to the controller, the
controller transfer function we are representing by G c, controller output is say c.

And this c is implemented through the final control element FCE, and this final control element
has the transfer function of suppose G suffix f. So, this is the block diagram of a closed loop
system, now we will discuss different elements of this close loops transfer function, different
elements or process, measuring device, controller final control element. These are the different
elements of this close loop process one is the process, another one is measuring device, then
controller and final control element and we will try to represent this elements by mathematical
forms.
So, if we considered the process we can write for the process y bar s equals G p s multiplied by
m bar s plus G d s d bar s, can we write this, the process can be for the process; we can write this
equation this two signals are added, this and this signal is added now this equation is
representing the process.
(Refer Slide Time: 12:05)

Next one is the measuring device output is y m, so in Laplace domine y m bar is equals G m s y
bar s, for the measuring device we can write this equation output equals transfer function
multiplied by input. Next element is the controller, the input signal to the controller epsilon bar
equals y set point bar s minus y m bar s, input signal to the controller epsilon equals y set point
minus y ,m this is representing the comparator. Similarly, the controller output is c bar s
controller transfer function is G c s and input to the controller is epsilon, so c bar s equals G c s
multiplied by epsilon bar s this is representing the control block.

Now, final control element FCE, for final control element the output is m bar transfer function is
G F and input is c bar s, so for final control element we can write output m equals transfer
function G c multiplied by c fine. Now, this m bar is we can write again as G F multiplied by c s,
c s means what c s means G c s multiplied by epsilon bar s, can we write this, c s equals G c s
multiplied by epsilon bar s.
Now, we will substitute the expression for epsilon bar s now, so G F G c epsilon bar s is y set
point bar minus y m bar fine, epsilon bar s is y set point bar s minus y m bar s. In the next step
we will substitute the expression of y m, so G F G c y set point bar minus y m bar, what is the
expression of y m bar G m s multiplied by y bar s, we are not writing s anymore, so this is the
expression of m bar agree.
(Refer Slide Time: 16:28)

Now, we have written the expression for the process as y bar equals G p m bar plus G d d bar,
we are written equation previously for the process fine, now here will substitute the expression
of m bar, so G p multiplied by m bar; that means, G F G c y set point minus G m y bar, this is m
bar. Now, the second term is G d d bar fine, we are just substituted the expression of m bar in
the equation of process. Now, if we rearrange this equation we finally, get y bar equals G p G F
G c divided by 1 plus G p G F G c G m y set point bar plus second term is G d divided by 1 plus
G p G F G c G m d bar, by rearranging we get this equation.
Now, the first right hand term this provides the effect of y set point on y, similarly the second
term second right hand term provides the effect of disturbance on y, now this equation is the

closed loop transfer function, this equation is called closed loop transfer function CLTF fine,
this is the expression of closed loop transfer function.
(Refer Slide Time: 19:56)

Now, if we considered G p G F G c equals G considering G p G F G c equals G, we get the


closed loop transfer function s y bar equals G divided by 1 plus G G m y set point bar plus G d
divided by 1 plus G multiplied by G m d bar fine. Considering G p G F G c equals G we get this
form of close loop transfer function, now again will consider this as G set point multiplied by y s
p plus G load multiplied by d bar.
Here, G set point equals G divided by 1 plus G G m and G load equals G d divided by 1 plus G
G m, considering this two expression for G set point and G load finally, we get this closed loop
transfer function. Can we corresponding block diagram what will be the block diagram of this
final form of closed loop transfer function? It is very simple we can consider one block for G set
point, what is the input? Input is y set point, now this output is added, which the output of the
this load block I mean this two outputs are added and finally, we get y fine.

(Refer Slide Time: 22:53)

This is the block diagram of the final closed loop transfer function, so ours original transfer
function is G c G p G F divided by 1 plus G c G p G F divided by 1 plus G c G p G F G m y set
point bar plus G d divided by 1 plus G c G p G c G m d bar. We will use this transfer function
mostly in the analysis of different feature fine, you see here the transfer function of measuring
devise does not exist in the numerator. And in the denominator in this term particularly all the
transfer functions are included, all the individual transfer function are included in the
denominator fine.
Now, the control performance is usually investigated by performing two testes, one is servo test
another one is regulatory test, so what is servo test, in the case of servo test there is no change
introduce in the disturbance, in the servo test there is no change introduce in the disturbance.
That means, d bar equals 0 only the set point change is considered, now if d bar becomes 0, what
is the closed loop transfer function for the servo problem.
So, close loop transfer function for the servo problem, become G c G p G s divided by 1 plus G
c G p G F G m y set point, this is the close loop transfer function for the servo problem. Now,
introducing step change may be sinusoidal change in y set point and by taking inverse of
Laplace transform we can know the tangent response y t by that wave we can test the controller.
This is called servo problem and sometimes another term is use that is set point tracking, so set
point tracking performance is absorbed by performing servo test.

Another test is regulatory test, regulatory test is conducting by considering no change in set
point; that means, y set point bar equals 0. Regulatory test is conducting by considering no
change in set point value accordingly, the closed loop transfer function become y bar equals G d
divided by 1 plus G c G p G F G m d bar. So, this is the closed loop transfer function for the
case of regulatory problem, another term is again used for this that is disturbance rejection
performance, so to absorb the disturbance reaction performance of a controller regulatory test is
conducted, set point tracking performance is absorbed by conducting servo test and disturbance
rejection performance is absorb by conducting regulatory test.
(Refer Slide Time: 27:16)

So, next we will considered one example, to discuss the formation of closed loop block diagram,
so formation of formation of closed loop block diagram, for a liquid level system. So, first we
will draw the schismatic of liquid level system, this is the liquid tank, F i is the inlet floret and
out let floret is say F naught, height of liquid in the tank is h. Now, the objective is to maintain
the liquid height in the tank by manipulating the outlet floret the objective is to maintain the
liquid height in the tank at its desired value, accordingly height is the controlled variable.
So, there are two options of manipulated variable I mean we can select either F i or F naught,
suppose F naught is the manipulated variable, then what will be the load variable F i. If we
considered F naught as the manipulated variable then F i will be the load variable, now we will
consider the different elements. So, first we will considered the process I mean, we need to
develop the transfer function of the process in terms of G p and G d.

So, for developing the transfer function of a process we need the model, so model we can
represent by this ordinary differential equation A d h d t equals F i minus F naught, this is the
modeling equation for the liquid level system. Now, here we are not using the prime to represent
the deviation variable, we are not representing the prime superscript to indicate the deviation
variable, I mean a h F i F naught, these are deviation variables.
So, this equation is written in terms of deviation variable, we are not representing prime
superscripts, if we take Laplace transform then we get A S h bar s equals F i s minus F naught s.
Taking Laplace transform we get this form, now rearranging h bar s becomes one by a is
multiplied by F bar s minus 1 bar A s F naught bar s, dividing both sides by A S we get this
equation.
Now, what is the general form, see general form is y bar s equals G p s m bar s plus G d s d bar
s, this is the general form and this is for the liquid tank system. Comparing this last two equation
we can easily get the transfer function G p n G d, so what is G p, see all G p is 1 by s. Yes, G p
is minus 1 by A s, because our disturbance is F i and aim is F naught.
(Refer Slide Time: 33:02)

Here aim is F naught and disturbance d is F I, so we get the transfer function G p s equals minus
1 by A s, and we get G d s equals 1 by A s. So, these two transfer function we obtain, now what
about the next element, I mean what is the next element next element is the measuring device.
So, we will considering the measuring device for finding the transfer function.

(Refer Slide Time: 34:08)

I mean in this liquid level system, first we can configure the control scheme, if we configure the
control scheme then we need the final control element, sorry first we need the measuring device.
So, to measure this liquid height the differential pressure cell is extensively used to measure the
liquid height differential pressure cell d p c can be use as a sensor, so what is the output of this
differential pressure cell h m, the output of this differential pressure cell is h m.
Original height is h and measured height is h m, this output is compared with the set point value
h s p, measured value is compared with set point value. Then this error signal is supplied to the
controller and controllers action is implemented through the final control element, this is the
control configuration of the example liquid level system. So, here measuring device is the
differential pressure cell, we need to find the transfer function of that differential pressure cell.

(Refer Slide Time: 36:05)

What is the input to this differential pressure cell, input is the pressure difference del p and we
can considered del p proportional to liquid height. This del p proportional to liquid height, what
is the output of this differential pressure cell h suffix m input to this differential pressure cell is
differential pressure and output is h m. Now, considering second order dynamics of this
differential pressure cell, we can write the modeling equation as tau square d 2 h m by d t square
plus 2 zeta tau d h m d t plus h m equals k p del p.
Considering second order dynamics of the differential pressure cell, we can write this equation
now the right hand side is k p del p, so we can write again that is equal to k p beta h, del p is
proportional to height, so del p equals beta height beta multiplied by h. Taking Laplace
transform and rearranging we get the transfer function as G m s equals output h m bar s divided
by input h bar s which is equal to K p beta divided by tau square s square plus 2 zeta tau s plus 1.
Taking Laplace transform and rearranging we get finally, this transfer function for the
measuring device, next element is the controller, we know the transform function for P PI PID
controllers.

(Refer Slide Time: 39:20)

Now, if we considers the P I controller then the transfer function we can write as G c equals K c
1 plus 1 divided by tau i s, if we consider p i controller for the example liquid level system then
we can write the transfer function in this form. Another element is the final control element
assuming first order dynamics of the control bulb, we can write the transfer function as G c
equals K v divided by tau v s plus 1. Assuming first order dynamics of the control bulb we can
write the transfer function as G this is f this is G f, so G F equals K v divided by tau v s plus 1, K
v is the gain of the control and tau v is the time constraint.
Can we make the close loop block diagram, first we need to develop the process having the
transfer function of G p and G d, how much is G p. G p is minus 1 by A s, what is G d is 1 by A
s and what is the disturbance, disturbance is F i what is the output, output is height and what is
the input to the process, I mean what is the final control element output that is F naught. So, this
is the process, the transfer function of the process with respect to F naught is minus 1 by A s,
transfer function of the process with respect to F i is 1 by A s output of that process is h.
For the measuring device we have derived the transfer function that is K p beta, K p is the gain
of the measuring device, hole divided by tau square s square plus 2 zeta tau s plus 1. So, this is
the transfer function of measuring device I mean this is equal to G m, output is measured height,
this is the comparator h set point is compared with measured height. Then the error signal goes
through the controller which has the transform function of K c 1 plus 1 divided by tau y s, this is
the transfer function of the controller P I controller output is c bar.

The transfer function for the final control element, we have derived that is K f equals K v
divided by tau v s plus 1, this is the transfer function of the final control element. So, this is the
block diagram of the example liquid level system, this is the closed loop block diagram of the
liquid level system. In the next we will discuss the effect of proportional action, we have we
discussed the three controllers P PI and P I D, basically there are three actions one is
proportional action, another one is integral action and derivative action. So, we use to discus
effects of all these actions individually, so first we will discuss the effect of proportional action.
(Refer Slide Time: 45:40)

So, first we will discuss the effect of proportional control we can write proportional action or
proportional control, because it only includes the proportional action. For this purpose we will
consider one process one open loop process we need to considered. So, first we will considered
an open loop process, now to absorb the effect of proportional controller we need to close the
loop. So, for closing what do we need one controller, that is p only controller because we are
going to discuss the effect of p only controller, then we need one measuring device and also we
need one final control element to implement the control action.
So, for absorbing the effect of proportional controller, we need to consider a process that is open
loop process, now to close the loop we need to include the controller, we need to include the
measuring device and final control element. So, fort open loop process we will consider, we will
consider two processes one is first order process and second one is second order process and we
will discuss one by one.

So, first we will consider first order process in the next we will consider second order process
example, next element is the controller. So, here we will include the p only controller, so for p
only controller what will be the transfer function for p only controller the transfer function G c
equals K c. So, for measuring devise we will consider G m equals 1 and for final control element
we will consider G F equals 1. So, for simplicity we are assuming the transfer function of
measuring device and final control element both are one fine.
So, we will start from the closed loop transfer function we will start from the closed loop
transfer function, the general form of close loop transfer function is y bar equals G c G p G F
divided by 1 plus G c G p G F G m y set point bar plus G d divided by 1 plus G c G p G F G m d
barThis is the general form of close loop transfer function. Now, in this close loop transfer
function we need to substitute the individual transfer functions, like for the process it is mention
that first we will consider first order process; that means, G p equals K p divided by tau p s plus
1.
For a first order process this is the transfer function, now if we substitute the transfer function G
p G c equals K c G f G m equal to 1, then what we will get if we substitute all the transfer
functions then will get the close loop transfer function for this example system. Now, for the
process these is the transfer function with respect to manipulated variable, another transfer
function is involved that is G d with respect to G d, so what that will be.
(Refer Slide Time: 51:38)

So, for that purpose we will first considered the process, we need to ride the modeling equation
for the process, the modeling equation for the process we can write as tau p divide d t plus y
equals K p m plus K d d. The first order process we can represent in time domain by this form,
where y is the output, m is the input to the process m is basically the output of the final control
element and d is the disturbance.
And y m and d all are in terms of deviation variables, now K p is the gain of the process with
respect to m, another gain is involved in this equation that is K d, K d is the gain with respect to
disturbance. Now, if we take Laplace transform and if you rearrange then finally, we get y bar s
equals K p divided by tau p s plus 1 m bar s plus K d divided by tau p s plus 1 d bar s, taking
Laplace transform and rearranging we get y bar s equals K p divided by tau p s m bar s plus k d
divided by tau p s d bar s.
So, G p is equal to k p divided by tau p s plus 1 and G d is equal to K d divided by tau p s plus 1,
from this equation we can get the transfer function of the process with respect to m that is G p
and the transfer function of the process with respect to disturbance that is G d. Now, all
individual transfer functions are known to us G p G d G c G F and G m, so in the next step we
need to substitute all this transfer functions in the close loop transfer function, so that we will
discuss in the next class.
Thank you.

Process Control and Instrumentation


Prof. A. K. Jana
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture - 18
Feedback Control Schemes (Contd.)

(Refer Slide Time: 01:02)

We will continue our topic which we started in the last class. That is the effect of
proportional controller, and we will observe first for the first order system, and then for
the second order system. So, initially will discuss the effect of proportional controller for
the first order process, for the first order process we derived the transfer function G p as
K p divided by tau p s plus 1. The transfer function G d equals K d divided by tau p s
plus 1 the controller is proportional controller, so the transfer function of the proportional
controller is j c equals K c.
The transfer function for P only controller can be written has G c equals K c and for
simplicity we are assuming G f and G m both are unity, for simplicity we are assuming
the transfer function of measuring device and final control element both are unity. Now,
we will try to develop the closed loop transfer function, so for that purpose we will start
from the general form of close loop transfer function which can be written as y bar

equals G c G p G f divided by 1 plus G c G p G f G m y set point bar plus G d divided by


1 plus G c G p G f G m d bar.
Now, we need to substitute all the individual transfer functions in this generalized form,
if we do that than we will get y bar equals K p K c divided by tau p s plus 1 plus K p K c
y set point bar plus K d divided by tau p s plus 1 plus K p K c d bar. Substituting all
individual transfer functions in this generalized closed loop transfer function, we obtain
the close loop transfer function for the first order system in this form.
(Refer Slide Time: 05:50)

So, this equation we can write again as K p prime divided by tau p prime s plus 1 y set
point bar plus K d prime divided by tau p prime s plus 1 d bar, this form we can write
again by this form. However, tau p prime equals tau p divided by 1 plus K p K c, K p
prime equals K p K c divided by 1 plus K p K c and K d prime equals K d divided by 1
plus K p K c. Now, what conclusion we can make based on the final form of the close
loop transfer function, due to the addition of proportional controller is there any change
up order in the our own system.
No, there is no change of order due to the inclusion of proportional controller, so no
change of order of the overall system, due to the inclusion of proportional action with the
first order system there is no change up order. Second conclusion is tau p prime second
conclusion can be made based on this, tau p prime less then tau p and K p K c is always
positive and therefore, tau p prime is less then tau p.

So, what it indicates, due to the inclusion of proportional action the overall response
becomes faster because time constant is decreased, so this is a second conclusion due to
the inclusion of proportional action the overall response becomes faster. Now, as I
mention that that control performance is observed by conducting two tests one is servo
test and other one is regulatory test.
(Refer Slide Time: 08:57)

So, in the next you will conduct these two tests for the first order process and the P only
control, so first we will discuss the servo problem. Our close loop transfer function is
represented as y prime equals y bar equal K p prime divided by tau p prime s plus 1 y set
point bar plus K d prime divided by tau p prime s plus 1 d bar. So, for servo problem d
bar equals 0; that means, the transfer function becomes y bar equals K p prime divided
by tau p prime s plus 1 y set point bar.
This is for the case of servo test, because for servo test there is no change in the
disturbance, no change in the disturbance is considered. Now, we will introduce if state
change in the set point value and you will observe the transient behavior of first order
process under P only control. We will consider a state change in the set point with a
magnitude of A, and we will observe the transient behavior of the close loop process.
So, y bar becomes K p prime divided by tau p prime s plus 1 multiplied by A by s agree
by considering this set point change in by considering the step change in set point value
we get this form. Now, if we take the inverts of Laplace transform, we get y t it is equal

to K p prime A 1 minus exponential of minus t divided by tau p prime, taking inverse of


Laplace transform we get this form of y in time domain.
(Refer Slide Time: 12:22)

Now, you will try to discuss the offset I mean we will try to determine either the process
under P only controller shows any offsets or not. Offset is calculated by representing as
new set point minus ultimate value of the response, offset is calculated using this form
new set point minus ultimate value of the response. Now, ultimate value of the response
means limit t tends to infinity y t ultimate value is calculated by using the final volume
theorem, that is limit tends to infinity y t or we can represent in Laplace domain also by
limit s tends to 0 s multiplied by y bar s.
So, what is the ultimate value of response for this case, set point value is A, because we
have introduced a step change with magnitude A and what is the ultimate value K p
prime A. If we apply the final value theorem we can obtain the ultimate value as K p
prime A, now offset becomes A multiplied by 1 minus K p prime substituting the
expression of K p prime finally, we get a divided by 1 plus K p K c.
Offset is calculated by using this form new set point minus ultimate value, here set point
is a ultimate value is K p prime A, substituting the expression of K p prime finally, we
get the offset equals A divided by 1 plus K p K c. So, if we produce the plot in terms of
transient response and y set point t, now we are introducing step change in y set point, so
initially it was 0, now we are giving a change in y set point, this is A at time t equals 0.

Now, the process under P only control response like this, this is the process response
under P only control y and this is representing y set point. This plot represents the
process behavior under P only control scheme and also by introducing step change in set
point value with a magnitude of A. Now, this deviations steady state deviation of the
controlled variable from set point value is offset, this difference is offset and
mathematically it is A divided by 1 plus K p K c, mathematically, we obtain offset
equals a divided by 1 plus K p K c.
(Refer Slide Time: 17:47)

Now, how K c tends to infinity, what about offset we got offset equals 1 divided by 1
plus K p K c offset equals A divided by 1 plus K p K c. Now, how when K c tends to
infinity offset approaches 0, but we discussed in the last class that there is a maximum
limit of K c beyond which the process goes un stable. So, we cannot arbitrarily select a
large value of K c. So, K c tends to infinity this cannot be considered in practice, due to
some maximum limit of K c value.
Next we will consider the regulatory test, next we will discuss the regulatory problem, so
our close loop transfer function for the first order system under P only control is K p
prime divided by tau p prime s plus 1 y set point bar plus K d prime tau p prime s plus 1
d bar. This is a close loop transfer function for the first order system under P only
control, now for the regulatory problem there is no change of y set point occurred.

So, for the regulatory problem the close loop transfer function, becomes y set point
equals K d prime divided by tau p prime s plus 1 d bar. This is the close loop transfer
function for regulatory test, why are no change in y set point is considered, similarly we
will consider a step change in the disturbance with a magnitude of A, we get the close
loop transfer function as y bar equals K d prime divided by tau p prime s plus 1
multiplied by A by s.
Taking inverse of Laplace transform we get the process output y in time domain has y t
equals K d prime a multiplied by 1 minus exponential of minus t divided by tau p prime,
taking inverse of Laplace transform we obtain the y in time domain by this, now you will
try to determine the offset.
(Refer Slide Time: 21:59)

Offset equals new set point minus ultimate value of the response, the ultimate value of
the response can be determined by the application of final value theorem. Anyway, how
much is the set point, set point is 0 and how much is the ultimate value of the response K
d prime A, ultimate value of the response is K d prime A. Substituting the expression of
K d prime, we obtain offset equals K d A divided by 1 plus K p K c, substituting the
expression of K d prime we finally, get the offset equals minus K d A divided by 1 plus
K p K c.
Now, you will produce a plot to observe the transient behavior we did not consider any
change in y set point, so y set point remains constant throughout the operation, but we

have introduced a step change in the disturbance at time t equals 0. So, the process
output changes from t equals 0, this is the process output under p control and if there is
no control then the process response behaves like this. We did not introduce any change
in the set point therefore, the set point remains constant throughout the operation, but the
disturbance is changed at time t equals 0.
Therefore, the process changes to a different steady state value starting from time t
equals 0, and if there is no controller then the process response like this. Now, this is the
steady state error; that means, this is the offset which is equal to minus K d A divided by
1 plus K p K c, now if you see the expression of offset again we can say that how when
K c approaches infinity, the offset approaches 0.
(Refer Slide Time: 26:47)

If we see the expression of the offset in the regulatory test we can say that if K c
approaches infinity the f offset approaches 0, now you will conclude. So, if we see the
expression of tau p prime we can say that the tau p prime is less than tau p, if we see
from this expression we can say that tau p prime is less than tau p. So, due to the
inclusion of proportional controller the overall response becomes faster, another
observation is with the increase of K c the tau p prime decreases.
So, if you recall this plot which includes the process response under different K c value,
this is K c 2, so here K c 1 is greater than K c 2. Now, if K c 1 is higher than the process

under K c 1 which is the steady state quickly then K c 2 that we need to include also in
this figure.
(Refer Slide Time: 29:22)

Secondly, we can say that the process under P only controller shows offset, that means it
is proved that the P only controller cannot eliminate offset. So, next we will discuss the P
only control of an integrating process, so for this purpose we need to consider an
integrating process this is a liquid tank, two input streams are involved in this example.
One is F 1, another one is F 2, one outlet stream is involved which as the flow rate of F
naught.
And one constant displacement pump is installed in the out let section, this is A constant
displacement pump, it indicates F naught is a constantan, F naught is constant. Now, in
this particular example, we can say that the control variable is liquid height, suppose the
manipulated variable is F 1, so load variable will be F 2. If we consider a F 1 as a
manipulated variable then F 2 is the load variable, now will consider will develop the
transfer functions of the different elements.
So, first is the process performing mass balance, we can write A d h d t equals F 1 plus F
2, all this variables are represented all these variables are here deviation variables. So,
this equation is written in terms of deviation variables, and therefore there is no F naught
term. So, to represent the deviation variable we are not using super script prime, now if

we take Laplace transform and if we rearrange then we get h bar s equals 1 divided by a s
F 1 bar s plus 1 divided by A s F 2 bar s, taking Laplace transform we get this form.
So, from this form we get the transfer function of the process, G p with respect to m s 1
by A s and G d s is also equal to 1 by A s, from this equation we get the transfer function
G p and G d s and both are equal to 1 by A s.
(Refer Slide Time: 34:18)

So, next element is the controller, our controller is the P only controller, so the transfer
function of P only controller is G c equals K c, this is a transfer function of P only
controller. Now, for measuring device and final control element we will consider both
the transfer functions as unity, the transfer function of measuring device G m and final
control element G f are unity.
Now, we will try to develop the close loop transfer function for the example system, the
generalized form is y bar equals G c G p G f divided by 1 plus G c G p G f G m y set
point bar plus G d divided by 1 plus G c G p G f G m d bar. This is of generalized form
of close loop transfer function, now if we substitute the form of individual transfer
function.
Then we get h bar s equals 1 divided by a by K c s plus 1 h set point bar s plus 1 divided
by K c A divided by K c s plus 1 F 2 bar s, substituting the expressions of individual
transfer functions in the closed loop transfer function, we get this form. Now, we will

conduct the servo and regulatory tests to observe the closed loop process behavior, so
first we will consider the servo test.
(Refer Slide Time: 37:18)

So, this is servo problem, so for servo problem we can write that F 2 bar s equals 0,
because there is no change of disturbance, according if the close loop transfer function
becomes h bar s equals 1 divided by a by K c s plus 1 h set point bar s. This is for the
servo test, considering F 2 bar equals 0, we get this closed loop transfer function for
servo problem.
(Refer Slide Time: 38:32)

Now, will consider a unit step change in the set point value of in the set point of h; that
means, h set point bar s equals by s. So, considering a unit step change in h s p, we get
the transfer function s h bar s equals 1 divided by a by K c s plus 1 multiplied by 1 by s.
So, what will be the offset, offset we can calculate by finding the new set point minus
ultimate value of the response, what is new set point 1, what is ultimate value.
Ultimate value we can find by the application of final value theorem that is also 1, so
offset becomes 0, so it is very interesting that for an integrating process the P only
controller provides offset free response. It is very interesting that the P only controller
provides for an integrating process offset free response, definitely by conducting the
servo test we got offset equals 0, but we need to also conduct the regulatory test.
So, in the next we will conduct the regulatory test and we will see what is the offset, our
transfer function close loop transfer function is h bar s equals 1 by K c divided by A by
K c s plus 1 F 2 bar s. This is the close loop transfer function for the integrating system,
how when there is no change is introduced in the set point. So, this is for the regulatory
test, introducing a unit step change in F 2 s, we obtain h bar s equals 1 by K c whole
divided by a by K c s plus 1 multiplied by 1 by s. Introducing a unit step change in F 2
we obtain the close loop function in this form, can you calculate, now the offset.
(Refer Slide Time: 42:34)

The offset will be calculated by finding the new set point minus ultimate value of the
response, so new set point is 0 and ultimate value is minus 1 by K c, so this is equal to

minus 1 by K c. So, for the regulatory problem, we obtain none 0 offset, but it is
important to note that usually we are not interested in maintaining the liquid level system
at the desired value, but within a certain range. We are not interested in maintaining the
liquid level exactly at the desired value, but at a certain range in that sense we can say
that the P only controller is sufficient to control the liquid level system.
Another important thing is that if K c is quite large, then this offset is acceptable if K c is
large then the offset represented by minus 1 by K c is acceptable. So, P only controller
can be used to maintain liquid level system, so this is all about the effect of proportional
control on the first order system, similarly we will consider we will observe the effect of
P only control on the second order system.
(Refer Slide Time: 45:22)

So, next we will discuss second order systems, under P only control next we will discuss
the dynamic sub second order system under P only control and here we will only
consider the servo problem. The first element of the close loop process is the process I
mean the open loop process, the transfer function of a second order process has this form
G p s equals y bar s divided by m bar s equals K p divided by tau square s square plus 2
zeta tau s plus 1, this is the transfer function of a second order process.
Next element is the controller and the controller is P only controller, so the transfer
function can be written has G c equals K c, this is the transfer function of P only
controller. Final control element and measuring device will consider the transfer

functions equal unity, to the transfer function of final control element and measuring
device are unity. Now, you will write the close loop transfer function for the second
order system under P only control, substituting all these individual transfer functions in
the generalized form of close loop transfer function.
We obtain y bar equals G p K c divided by 1 plus G p K c y set point bar plus G d
divided by 1 plus G p K c d bar, substituting all the transfer functions except the process
transfer function, we obtain this form. Now, we will consider only the servo test, we will
only conduct the servo test, so for the servo case the transfer function yields y prime
equals K c K p divided by tau square s square plus 2 zeta tau s plus 1 divided by 1 plus K
c K p divided by tau square s square plus 2 zeta tau s plus 1.
Substituting the transfer function of the open loop process we obtain this expression,
now we can represent this transfer function as y prime equals K p prime divided by tau
prime square s square plus 2 zeta prime tau prime s plus 1 y set point bar, here we need
to multiply y set point.
(Refer Slide Time: 50:07)

Now, K p prime as the form of K p prime equals K p K c divided by 1 plus K p K c,


similarly tau prime as the form of tau prime equals tau divided by root over of 1 plus K p
K c and zeta prime equals zeta divided by root over of 1 plus K p K c. So, due to the
inclusion of P only controller with the second order process, there is no change of order.

So, first observation is that due to the inclusion of P only controller with the second order
process there is no change up order in the overall system.
Another observation we can note that it is tau prime less than tau, if we see the
expression of tau prime we can say that tau prime less than tau. Similarly, zeta prime less
than zeta, similarly if we see the expression of zeta prime we can say that zeta prime less
than zeta; that means, under P only control we may obtain oscillatory response I mean
the over dam system may change to under dam system under P only control.
Thank you.

Process Control and Instrumentation


Prof. A. K. Jana
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture No. # 19
Feedback Control Schemes

In the last class we are discussing in the P control of a second order process, in the last
class we are discussing of proportional control of a second order process we will
continue the topic today.
(Refer Slide Time: 01:09)

So, p control of a second order process and we will consider servo test, basically you
want to see the closed loop response of a second order process under P only control. So,
what are the different element of closed loop process, first element is the process that is
basically open loop process. The open loop process with second order system has the
transfer function G p and that is equal to K p divided by tau square s square plus 2 zeta
tau is plus 1.
This is the transfer function of base a second order process, next element is the controller
we are interested to observe the close loop performance under the P only control.

Accordingly, the controller equation has the form of G c equals K c, this is the transfer
function of P only controller, for simplicity will consider G m and G f both are equal to
unity.
The transfer function of the measuring device that is the G m and the transfer function of
the final control element that is G f or equal to 1, now what is the expression general
expression for closed loop transform function. General expression of close loop transfer
function is represented as y bar equals G c G p G f divided by 1 plus G c G p G f G m y
set point bar plus G d divided by 1 plus G c G p G f G m d bar, this is the general form of
closed loop transfer function.
Now, considering or substituting all the individual transfer function in this generalized
form, we obtain y bar equals K c K p divided by tau square s square plus 2 zeta tau s plus
1 whole divided by 1 plus K c K p by tau square s square plus 2 zeta tau s plus 1
multiplied by y set point bar. And this is the close loop transfer function for servo
problem. Substituting all the individual transfer function G c G p G f G m in the
generalized form of this close loop transfer function, we obtain this form for the servo
vocals. Now, if we simplify it than we get K c K p divided by t square s square plus 2
zeta tau is plus 1 plus K c K p multiplied by y set point bar.
(Refer Slide Time: 06:45)

If we divide numerator and denominator by 1 plus K c K p, dividing numerator and the


denominator by 1 plus K c K p, than we obtain y bar s equals K p prime divided by tau

prime square s square plus 2 zeta prime tau prime s plus 1 multiplied by y set point. But,
dividing numerator and the denominator 1 plus K c K p, we obtain this form higher tau
prime equals tau Divided by root over of 1 plus K p K c. Zeta prime equals zeta divided
by root over of 1 plus K p K c and prime K p prime equals K p K c divided by 1 plus K p
K c.
These are the expression with which happen used in this closed loop form, now you will
conclude, so due to the inclusion of proportion controller with the second order process,
the order does not change the order remains same. So, we will note down few points due
to the inclusion of proportion controller there is no change of the order overall system,
this is the first point.
Secondly, if we see the expression of tau prime we can say that tau prime less than tau,
secondly zeta prime less than zeta it means overall response under P only control
becomes more oscillatory, because damping factor is decreased. As a result the overall
response under the p control becomes more oscillatory in the last class we discuss up to
this.
(Refer Slide Time: 09:49)

So, next will determine the offset, now out tools loop transfer function is y bar equals K
p prime divided by tau prime s square plus 2 zeta prime tau prime s plus 1 y set point bar,
this is the close loop transfer function we which we derived for servo case. Now, we will
consider a step change in set point with magnitude A, we use to determine the offset

introducing a step change for set point with magnitude A. So, considering this y set point
equal A by s that close loop transfer function, becomes K p prime divided by tau prime
square s square plus 2 zeta prime tau prime s plus 1 multiply by A by s.
Now, according to final value theorem the ultimate response we determine as limit s
tends to 0 s y bar s, so limit s tends to 0 K p prime A divided by tau prime s square plus 2
zeta prime tau prime s plus 1, which is equal to K p prime A. The ultimate response for
the case of second order system under P only control with the introduction of set p
change point with the magnitude A.
We obtain K p prime A than what about of offset we know the offset equals new set
point minus ultimate value of the response to new set point is A ultimate value is K p
prime A. , so A multiplied by 1 minus K p prime. Now, we will substitute the expression
of K p prime, so A 1 minus K p prime is K p K c divided by 1 plus K c, what is the
expression of K p prime, K p prime is K p K c divided by 1 plus K p K c. So, the offset
becomes A divided by 1 plus K p K c, this is the offset, now when the offset becomes 0
or when offset approach is to 0 ,when K c is to infinity.
(Refer Slide Time: 14:13)

When K c tends to infinity, the offset approach is 0, but this is not through in practice
due to some limitation of K c value, now you will try to produce one plot to observe the
dynamic behavior of the closed loop system. Now, we obtain the expression of zeta
prime as zeta prime equals zeta divided by root over of 1 plus K p K c, this expression

we obtain. So, with the decrease of K c, zeta prime increases; that means, with the
increase of K c zeta prime decreases.
So, if the increase of K c value the overall response may be more oscillatory, so with the
increase of K c the overall response becomes more oscillatory. This is one point, another
point is we know the offset equals A divided by 1 plus K p K c, the larger the K c value
the smaller the offset. From this expression we can say that the larger the K c value the
smaller the offset, but the considering this two points with the increase of K c, the overall
this becomes more oscillatory and another point is the larger the K c the smaller the
offset.
(Refer Slide Time: 17:23)

We will try to produce the plot and in the close loop response keeping in mind this two
points. So, we will produce y verses time y and y set point will consider and here time,
now we consider a step change in set point with magnitude A, so this is magnitude A,
initially it was say 0 and we are introducing step change a time equals 0.
Now, the close loop look response is like this, for a particular K c that is suppose K c 1
this is for another K c suppose K c 2, so here K c 1 is greater than K c 2 fine. So, for the
last K c that is K c 1 we obtain more oscillatory response, one point an second point is
for larger K c the offset is relatively small and this is for the case of K c 2 offset is large
for small K c.

So, above two points are taken into account in the development of this close loop
response, in the next we will discuss the effect of integral control action. So, for we
discussed the effect of proportion action on the overall response of a process, in the next
we will discuss the effect of integral action on the overall response of a process.
(Refer Slide Time: 20:10)

So, next topic is effect of integral control action and we will consider some servo
problem only, the regulatory problem is left for the students. So, we will consider
different elements and there expressions first, first element is the process and discuss to
element integral control action we consider first order process. So, for the first order
process is transfer function can be written as G p equals K p divided by tau p s plus 1,
this is the transfer function of a first order process.
Next element is the controller, the controller is the integral controller what is the transfer
function of integral controller, the transfer function of the integral controller is written as
G c equal K c divided by tau i s. For G i control is the transfer function is K c multiplied
by 1 plus 1 divided by tau I s; that means, K c plus K c divided by tau i s. So, this is the
transfer function of p and this is the transfer function of i, so if we consider only the
integral controller than the transfer function is G c equal K c divided by tau i s.
And simplicity we will consider the transfer function of measuring device and the
transfer function of final control element, both are unity; that means, G m and G f equal
to 1. Now, we will try to develop the close loop transfer function substituting all this

individual transfer functions, in the next we will develop the closed loop transfer
function.
The general form is G c y bar equals G c G p G f divided by 1 plus G c G p G f G m y set
point bar, now we will substitute the individual transfer function in this close loop
transfer function for the case of servo problem G c is K c divided by tau i s. This is G c G
p is K p divided by tau p s plus 1 and G p is equal to 1 than 1 plus K c divided by tau i s
multiplied by K p divided by tau p s plus 1 and y set point.
(Refer Slide Time: 24:45)

If we simplify this expression than we obtain y bar s equals 1 divided by tau square s
square plus 2 zeta tau s plus 1, simplifying we obtain this expression multiplied by y set
point bar. First we substituted all the individual transfer function in generalized from of
those transfer function, than after simplification we are obtain this expression, where are
tau is root over of tau i tau p divided by K p K c. Here, tau has the this expression I mean
tau has this relationship, similarly zeta is equal to half root over of tau i divided by tau p
K p K c.
So, these are the two expressions which have in consider in the development of close
loop transfer function, you see found the co-relation is clear that the tau and zeta both
depend on the value of K c and tau y. From these two co-relation, it is obvious that the
tau and zeta both depend on the value of K c and tau y. Where, K c and tau y are the

tuning parameters, controller tuning parameters, in the next we will try to determine the
offset.
Now, for the determining offset we will consider the step change in set point with
magnitude A; that means, y set point bar is equal to A bar s. We are consider in step
change with magnitude A, than the close loop transfer function becomes 1 divide y bar is
equals 1 divide by tau square s square plus 2 zeta s plus 1 multiplied by A by s. Then
offset is equal to new set point minus ultimate value of the response, we know this corelation offset equals minus ultimate value of the response.
(Refer Slide Time: 28:51)

So, new set point is A, and ultimate value of the response is how much A, so offset
becomes 0. So, we can say that the integral control action can eliminate the offset, it is
very obvious from this that the integral control action can eliminate the offset. Now, we
will note down some few important points, is any change of order due to the inclusion of
integral action. Yes, so due to the integral action the order changes, the order of the
overall system changes, due to the inclusion of integral action this is the first point.
Second point is we have seen that due to the inclusion of integral action the offset
become 0, so second important point is the integral action can eliminate the offset,
integral control action eliminates the offset. Third important point is as I mentioned zeta
want depends on the value K c and tau i, if we see the expression of zeta we can say that
zeta depends on the value of K c and tau i.

So, close loop system provides over dammed or under damped, critically damped
response depending on the value of K c and tau i, closed loop system provides over
damped or under damped or critically damped response depending on the value of K c
and tau i. So, the selection of K c and tau i is very important.
(Refer Slide Time: 32:30)

Fourth point is if we again see the expression of zeta that is zeta is half root over of tau i
divided by tau p K p K c, we can say that with the increase of K c, zeta decreases. So, the
over damped response may move to under damped response with large over suite and big
ratio. If K c is sufficiently large the over damped response may move to under damped
response with large over suite and big ratio, in the same line we can conclude again that
with the decrease of tau i zeta decreases. You see the expression of zeta it is clear that
with the decrease of tau i zeta decreases.
So, if the tau i is sufficiently small the over damped response may move to under
damped response with large over large suite and big ratio. Now, we is to produce plots
best on the common 4 and 5, in the fourth point we have mention that with the increase
of K c zeta decreases; that means, more oscillatory response.
So, that we will reflect through this plot y verses time, a step change is introduced with
magnitude A, this is suppose 0 and the step change is introduce that time equals 0. Now,
this is the close loop response for a particular K c value, suppose that is K c 1 another
close loop response we obtain for the value of K c 2. So, which one is greater K c 1 or K

c 2, K c 1, because our fourth point is with the increase of K c we get more oscillatory
response. So, the K c 1 is larger than K c 2.
So, that is why we obtain more oscillatory response, but definitely here tau y is kept
constant we only varied the K c value. Similarly, we will produce the close loop response
based on 0.5 that is with the decrease of tau i zeta decreases; that means, with decrease of
tau i the overall response become more oscillatory. So, we will try to produce y verses t,
similarly a step change is introduce in set point with magnitude, suppose A at that time t
equals 0. This is the close loop response for a particular tau value, suppose that a is tau i
1, another sponge will we obtain one tau i 2.
So, which one is larger among this two tau i 2; that means, tau i 1 is less than tau i 2,
with the decrease of tau i zeta decreases; that means, more oscillatory response. So, first
one is the close loop response with the valuation of K c and second one is the close loop
response with the valuation of tau i. So, these two points 4 and 5 clearly say that we need
judicially tune the values of K c and tau i. So, we discuss effect of proportion action and
integral action and the close loop response, the third action is the derivative action.
(Refer Slide Time: 39:13)

We will discuss that I mean discuss the next derivative action on close loop response,
effect of derivative action, derivative control action and we will consider the servo
problem. In the discussion of derivative action we will consider the first order process
and servo problem, in a close loop black diagram the important element is process. The

transfer function of the process of a first order process can be written as G p equals K p
divided by tau p s plus 1, this is the transfer function of a first order process.
Next element is the controller recently we discuss derivative action, so the controller is
only the derivative controller, what is the function of derivative control, K c tau p s. This
is derivative action, because the transfer function of P I D controller, we have written K c
K c plus divided by tau i s plus K c tau p i s. This is the transfer function of proportion
action, this is the transfer of integral action and this is the transfer function derivative
action.
And simplicity for will come consider the transfer function G f and G m both are equal to
1, now we will try to develop the close loop transfer function substituting of all these
individual expressions. In the next will develop close loop transfer function which is
express in generalized form as y bar equals G c G p G f divided by 1 plus G c G p G f G
m y set point bar, this is generalized close loop transfer function for servo base.
Now, if we substitute the individual transfer functions, than we obtain G c as K c tau D s,
next transfer function is G p; that means, K p divided by tau p s plus 1 and G f is 1,
whole divided by 1 plus K c tau D s multiplied by K p divided by tau p s plus 1
multiplied by y set point bar. Now, if we this is the close loop transfer function we
obtain, substituting all individual transfer functions.
(Refer Slide Time: 43:56)

Now, if we rearrange further than we obtain the final close loop transfer function as y bar
equals K p K c tau D s divided by tau p plus K p K c tau D multiplied by s plus 1 y set
point bar, this is the final form of close loop transfer function. Now, this is the time
constant of the close loop transfer function, so this time we can say we can call as
effective time constant denoted by tau p prime. So, tau p prime equals tau p plus K p K c
tau D, this is called as effective time constant we can call this time constant as effective
time constant.
So, which one is larger tau p or tau p prime, tau p prime now we will note down few
important points, due to the inclusion of derivative action is there any change of order.
No, the order of overall system does not change, we have taken first order process and
overall processing is first order. So, there is no change of order of the overall system due
to the inclusion of derivative function there is no change order of the overall system, this
is the first point.
Second point is regarding the time constant, the time constant tau p prime is greater than
tau p, because K p K c is positive and tau D is also positive. K p K c is positive
multiplication of K p K c I mean multiplication of K p and K c that is positive and tau D
is time constant time constant is always positive. So, this is also positive therefore, tau p
prime is greater than tau p, so what it indicates the overall response becomes ((Refer
Time: 47:34)).
Third point is as K c increases, then what happens for tau p prime, if K c increases then
tau p prime also increases with the increase of K c tau p prime increases; that means, the
overall response becomes slower due to the increase of K c. Similarly, what happens for
tau D, if tau D increases the effective time constant increases. So, in the derivative action
there are two tuning parameters involved, one is K c and another one is tau D. So, that is
why we observe the effects of K c as well as tau D with the increase of K c tau p prime
increases and with the increase of tau D also tau D increases; that means the overall
response becomes slower due to the increase of K c and tau D.
Fifth point is related to the offset, can you calculate the offset with introducing step
change point with set point magnitude A. What about offset? Offset is A, so fifth point is
the derivative controller cannot eliminate offset, these are the five important points
related to the close loop response and under derivative controller action. So, we have

discussed about three particular controller proportions one is P only, second one is P I
and third one is P I D.
(Refer Slide Time: 50:17)

Now we conclude these control actions one by one, so first one is P only controller, what
you conclude the P only controller. It speeds up the response, if you see the response of
over loop system and if you see the response of a process, same process under P only
controller, we obtain first response. In case of process under P only controller, if we
include a proportional controller with a process the overall response become faster than
the open look response. Therefore, we conclude that the P only controller speeds of the
response.
Secondly, we can say that with the increase of K c value, the proportional only controller
output becomes more sensitive to error, with the increase of K c value the P only
controller output becomes more sensitive to the actuating error signal. This is the second
conclusion on P only controller, third one is it cannot eliminate offset, it does not
eliminate offset, these are the three important points related to P only controller.
Next is P I controller, most important point related to P I controller is that eliminate the
offset, so which part eliminate the offset integral part integral action eliminates the
offset. Second point is the order for P only controller there is no change of order, but P I
controller order is increased. So, second point is order increases and as the order
increases the overall response will sluggish.

Now, how we can improve this situation, due to the inclusion of a integral action with p
action, the overall response becomes sluggish. Now, we want to speed of the response,
how we can do that either by increasing K c and decreasing tau i is in did. We can speed
of the response either by increasing K c or by decreasing tau i, if you see the controlled
equation, P I it is controller equation cleared that with the increase of K c the overall
system response become faster. But, if K c is larger than the sorter value or if tau i is
sufficiently small, the process those constant I mean there some instability problem.
So, if the K c value is sufficiently large or tau i is sufficiently small the instability
problem may raise, for the case of P I controller. Next is the P I D controller, so first we
can say that it eliminates the offset P I D controller have eliminates the offset and this is
the definitely effect to the integral action.
Now, for the case of P I controller, we observe that with the increase of K c value I mean
when K c is larger the certain limit or tau i is sufficiently small there is some instability
problem, that can be improve by the addition of derivative action. So, P I D control has
batter stability criteria and this is due to derivative action only. Third point is if error is
constant, then there is no derivative action because deep silent d t is there, if error is
constant then there is no derivative action. And last point is for a noisy is almost 0 error,
the derivative term is needs to large control action although that is not require. So, these
are the concluding remarks on three proportion controller actions.
Thank you.

Process Control and Instrumentation


Prof. A. K. Jana
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture No. # 20
Feedback Control Schemes (Contd.)

(Refer Slide Time: 01:00)

Today, we will covered the topic stability analysis, today will discuss stability analysis.
So, first we will try to know the definition of stable system, what is the definition of
stable system? A dynamic system is said to be stable when for all bounded inputs, the
dynamic system produces bounded outputs, and this is the definition of stable system.
So, here to new terms or inward in this definition; one is bounded input and another one
is bounded output, these are quite new. So, next we will know what is bounded input,
bounded input is an input that remains within an upper limit and lower limit.
Similarly, we can defined the bound output, bounded output is an output, that remains
within an upper limit and a lower limit. If we produce a plot between time t versus input
u, then it looks like is this, see u is the input. So, it has upper limit, that we can denote by
suppose u max and it has the lower limit I mean the lower value that is a u min. Now, the
input variable always remains within this upper limit and lower limit, it should not

exceed the upper as well as lower limit that is why it is called bonded input, can we give
one example of bounded input.
Student: Sinusoidal
Sinusoidal input is one bounded input, step input is another bounded input, what about
ramp input, ramp input is not a bounded input. Now, just opposite of this bounded, that is
un bounded, there is no existence of the term un bounded in reality, because of the
physical limitation of all instruments and equipments. There is no existence of the term
un bounded in reality, it only exist in theory, because of the physical limitation of all
instruments and equipments.
So, for we have discussed a little about the stability of open loop system, we have
determined the roots of the denominators of a transfer function. And we discussed about
the position of those roots in the complex plane, best on who which we can comment on
stability of open loop system, that we have discussed, so for. Now, we will discussed the
stability of closed loop system, so before discussing we if net to know few important
points.
(Refer Slide Time: 07:32)

We derive the closed loop transfer function in generalize form, the closed loop transfer
function we can represent by this expirations, y bar is equal to G c G p G f divided by 1
plus G c G p G f G m y set point bar plus G d divided by 1 plus G c G p G f G m d bar,

this the general form of the closed loop transfer function. Now, the denominator of these
transfer function, we can write as 1 plus G c G p G f G m, if we write this equals 0, then
that is called the characteristics equation. This is the characteristics equation of the
closed loop transfer function for stability analysis we will use this equation.
This is the characteristics equations of the closed loop transfer function, these equation
can also be write in this form 1 plus open loop transfer function G suffix O L open loop,
which is equal to 0. The characteristic equation we can also write by this form 1 plus the
transfer of open loop equals 0. Now, we will write the definition of stable closed loop
system, a feed back control system is stable, when all the roots of it is characteristic
equation, lie in the left side of complex plane, a feed back control system is stable, when
all the roots of it is characteristic equation, have negative real part.
A dynamic a feed back control system is set to be stable, if all the roots of the
characteristic equation lie in the left side of the complex plane. So, if this is real axis and
this is imaginary axis all the roots would lie in the left side, this is the left side, then we
can say the feedback control system stable.
(Refer Slide Time: 12:00)

Now, we will just discus different situations, so first we will consider the type of roots,
then what will be the corresponding root location in complex plane. And we will know
the corresponding closed loop response, for a step change in set point SP, SP means Set
Point. Suppose, it characteristic is equation has a number of roots, among them say a

characteristic equation has a number of roots and the roots are lie in the left side of the
imaginary axis; that means, all the roots have negative real parts negative real root not
parts negative real root.
So, this is the complex plane this is real axis and this is imaginary axis, so suppose we
are considering a single negative real root is there. So, the position of that would be here
what it indicates the system is the feedback system is stable are un stable, this is stable.
So, if the characteristic equation has negative real root then the system is stable, what
will be the transient response time t verses y, it will be like this definitely against step
change in set point value the transient response will be like this.
The it was the earlier state I mean the process was add this positions now we are going to
change, we are going to introduce is step change in set point that is why the process is
going towards achieving a new steady state. In the second situation we will considered
among the roots one has positive real part, so among the root one is positive real root,
now we need to locate that root in the complex plane, this will be the location positive
real root. So, this is stable or un stable this is un stable since the root has positive real
part, so the feedback control system is un stable. What will be the corresponding
transient response?
This is time verses y plot, so it will response like this, it is going to un stable zone, in the
third situation we will consider complex conjugate roots, with negative real part. Now, if
will locate the roots in the complex plane, these are the two roots who which are
complex conjugate with negative real parts. Now, what about the stability stable, because
the real parts of negative, so the feedback control system is stable, what will be the
transient response t verses y, it will reach a new steady state with decreasing oscillation,
oscillation with decreasing amplitude.
And in the fourth situation will consider complex conjugate roots with positive real parts,
so if we locate the roots in the complex plane, it looks like this these are two roots
complex conjugate roots, which have positive real parts. What about the stability un
stable, the closed loop system is un stable when it has two complex conjugate roots with
positive real parts. And what will be the corresponding transient behavior, oscillation
with growing amplitude like this, so these are the four deferent situations created in a
closed loop system.

(Refer Slide Time: 19:23)

We will note down 5 important points on stability roots, axis, response real roots near the
imaginary axis, result in slow response. This is the first important point real roots near
the imaginary axis result in slow response, if we compare in the complex plane it looks
like this, one root is this one another root is this one. So, the root which is near to the
imaginary axis that will provide slower response if we consider two roots, so this is near
to the imaginary axis, so it will provide slower response compared to another root,
complex roots near the imaginary axis result in slow response.
Suppose, this is one set this is another set, so this is near to the imaginary axis, so this set
will provide slower response, among these two complex conjugate roots this one is near
to the imaginary axis. So, this have the complex the closed loop system having these
roots will slow with, so slower response, another important point is when the complex
roots are away from the real axis.
The more oscillatory the transient response will be say this is one set and another one is
this one, so this set is away from the real axis, so it this set will provide more oscillatory
response. So, these are three important points on the transient response, speed of the
transient response you can say. Now, in the next we will discuss the one very popular
stability analysis technique, that is Routh Hurwitz test.

(Refer Slide Time: 24:35)

This is purely an algebraic test Routh Hurwitz test is purely an algebraic test, by
perfuming this test we can know how many roots have positive real roots, how many
roots have positive real parts, it is interesting that without determining the values of the
roots. We can know, we can say how many roots have a positive real parts by perfuming
this test, there is known it to determine the roots, just by perfuming this test we can say
how many roots have positive real parts.
And another point is there is known in of any production of plot, there is known it to
production any plot for this particular case and this method this test is applicable when
the characteristic equation is expressed in polynomial form. This roots test is applicable
when the characteristic equation is in polynomial form, you can say this is the limitation
of these roots test, we discussed the characteristics equation that it has this form, I mean
we can express these as 1 plus G c G p G f G m equals 0.
Now, suppose the process is a first order get time process suppose the transfer function
of the process has this form, G p equals K p exponential of minus tau d s divided by tau p
s plus 1 the process is the first order plus date time system. Now, if we substituted G p in
this characteristic equations, we do not get polynomial form, so what we can do in this
case, in this case we can go for approximation may be Pade approximation may be trailer
series expansion. Remember in that case we can do all the approximate stability analysis,

if we approximate this exponential term say by Pade approximation, in that case exact
stability analysis is not possible ally approximate stability analysis is possible.
(Refer Slide Time: 29:16)

Now, we will known the algorithm I mean the test procedure, we will know the test
procedure, what are the stets we need to follow to perform this Routh stability analysis.
First point is we need to develop the closed loop transfer function, so in the first we need
to develop the closed loop transfer function this is the first point. Secondly write the
characteristics equation in standard polynomial form, that is 1 plus G c G p G f G m
equals 0, now we will write this in standard polynomial form, as a naught s to the power
n plus a 1 s to the power n minus 1 a n minus 1 is plus a n equals 0.
We need to the represent the characteristics equations in this standard polynomial form,
which have which is in thought are polynomial. Let a naught be positive, under the
second point few important ((Refer Time: 31:31)) we need to include, first point is let a
naught be positive, if a naught is negative multiplied both sides of this polynomial by
minus 1. Now, inspect the coefficients, in the next inspect the coefficients a 1 a naught a
1 up to a n, inspect the coefficients if any 1 coefficient is negative then the closed loop
system is unstable, there is no need of any further analysis needed.
If all the coefficient are positive, the system may be stable or unstable, if all the
coefficients are positive, it is difficult to say whether the system is stable or unstable. So,
we can say that if all the coefficients are positive, it may be stable may be un stable, so in

this situation who in need to go for further analysis. So, this is the second point a naught
should be positive we need to assuming a naught be positive, if that is not positive we
need to multiply both sides of the characteristics equation by minus 1.
In the next step we need to inspect the coefficients in any one coefficients is negative the
system is unstable, so there is no need of further analysis, if all the coefficients are
positive the system may be stable or may not be stable. So, to conclude weathered the
system is stable or unstable we need to go for further analysis.
(Refer Slide Time: 35:38)

So, that is basically the third point I mean third step of this test, in the third step we need
to form Routh array, the Routh array consist of a number of rows. So, we will write the
row one by one this is first row, second row, third fourth like this in plus 1, so you
remember if the polynomial is in order the number of rows should be n plus 1. So,
number of rows is equal to n plus 1, where in is the order of the polynomial, now we will
write in this array all the coefficients of the standard polynomial. You we just re visit the
characteristic equation the characteristic equation, we have return in this standard
polynomial form which in clouds the coefficients starting from a 1 to a n.
Now, we will write the coefficients like this, in the first row we will write a naught, in
the second row we will write a 1, then again a 2 a 3 a 4 a 5 a 6 a 7 like this, last one will
be a minus 1, a n minus 1 and this one will be a n. So, the elements of the first row
include a 1, a 2 a 4 a 6 a n minus 1, the second row includes a 1 a 3 a 5 a 7 through a n.

So, this is the ((Refer Time: 38:14)) we need to write all the coefficients in first row and
second row, the elements of the other rows we need to calculate.
The elements of the other rows starting from 3 to n plus 1, we need to calculate, that
calculation procedure has been propose by Routh suppose elements are A 1 capital A 1
capital A 2 capital A 3 like this. The elements of the fourth row are assumed B 1 B 2 B 3
and the elements of the last row n plus 1 row or Z 1 Z 2 Z 3, this is the root array. Now, if
how we can I mean what are the expirations for calculating the elements starting from
third to n plus 1 row.
Capital A 1 which is the first element of third row is calculated as a naught they are
calculated as a 1 a 2 minus a naught a 3 divided by a 1, a 1 a 2 divided by minus a 1
minus a naught a 3 divided by a 1. This the expiration for a 1 the expiration of capital A 1
is a 1 a 2 minus a naught a 3 divided by a 1, the expiration for capital A 2 is a 1 a 4 minus
a naught a 5 divided by a 1, I think now we can write the expiration for all other
elements.
Expiration for A 2 you we just left that column within which a 2 is present, so a 1 a 4
minus a naught a 5 divided by a 1, similarly the expiration for a 3 is a 1 a 6 minus a
naught a 7 divided by a 1. a 3 is present in the third column, so you will left the elements
of third column and we will stared from first column always that is why A 3 has the
expiration and that is expect by a 1 a 6 minus a naught a 7 divided by a 1.
What will B the expiration for B 1 then B 1 will be capital A 1 small a 3 minus small a 1
capital A 2 divided by capital A 1, the expiration for B 1 we can write as capital A 1 a 3
minus small a 1 capital A 2 divided by capital A 1. Similarly, we can write B 2 equals a 1
a 5 minus a 1 capital A 3 divided by capital A 1 capital B 2, we can write as capital A 1
small a 5 minus small a 1 capital A 3 divided by capital A 1.

(Refer Slide Time: 43:15)

Another element is B 3, so B 3 we can write as Capital A 1 capita small a 7 minus small


a 1 capital A 4 divided by capital A 1 B 3, so we need to write here the element a 4, so
capital A 1 small a 7 minus small a 1 capital A 4 divided by capital A 1, that is that
expiration for B 3. So, this is the third point in which we have discuss the formation of
root array, next we will analyze the stability in the forth point.
(Refer Slide Time: 44:24)

In the forth point we will discuss the analysis of this root array, so what are the elements
are a naught a 1 capital A 1 capital B 1 last element is z 1, these are the elements in the

fast column. Now, the close loop system is stable if all the elements are positive and non
0, so this is another point is if any one element in the first column is negative then at
least one root has the positive real part.
If any one element in the first column is negative, then at least one root has positive real
part, another point is if one element in the first column is negative. Then at least one root
at least one root has positive real part, I mean that root lie in the right half plane, see we
are mentioning here at least; that means, there may be more than one roots lie in the
write half plane.
Now, question is how we can say more than one root are off ward situation more than
one roots lie in the right half plane, so best on that there is one point number of roots lie
in the write half plane equal to the number of sign changes in the first column. So, in the
next point we will write the number of right half plane roots is equal to number of sign
changes in the first column, suppose this is the first column which includes the elements
a naught a 1 capital A 1 capital B 1 2 z 1, among these element one root among these
elements one is negative.
That mean say this capital A 1 is negative, so how many sign changes are involved to one
sign change from second column to third column, another sign change from third column
to third row to fourth row. So, two sign changes involved; that means, two roots have
positive real parts, suppose among this elements in the first column one element is 0,
then what about the stability of that closed loop system. See, if all the roots are positive
then the system is stable, if an one root is negative system is un stable, so if one element
is 0 that is the critical stability condition.
So, at critical stability condition we get sub stain oscillation, so if one root one element
in the first column is 0 then we get sub stain oscillation, I mean oscillation with constant
amplitude. This is all about the Routh Hurwitz stability test and one point is very clear
that according to this Routh Hurwitz test we can only now whether the closed loop
system is stable or un stable. Using this test we cannot determine the values of tuning
parameters, see we need to conduct the root stability test, to know the stability of a close
loop system.
That means, the process including the controller, we want to know the stability of a
closed loop system, closed loop system means the process which includes the controller.

Now, according to this test we can only say that the power all system in stable or un
stable, using or conducting this test we never determined of exact values of the tuning
parameters. So, this is not basically the tuning test, this is basically a stability test.
Thank you.

Process Control and Instrumentation


Prof. A. K. Jana
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture No. # 21
Feedback Control Schemes (Contd.)

In the last class we discussed Routh-Hurwitz test, today we will solve one problem based
on that test.
(Refer Slide Time: 01:07)

So, today will discuss a problem on Routh-Hurwitz test, now in this problem the
characteristics equation is given, it is 2 s cube plus 3 s square plus 1 plus K c s plus 2,
which is equal to 0, and K c is the proportional gain of the controller. Now, we need to
compare this characteristics equation with the standard polynomial form of
characteristics equation. What is the standard form, that is a naught s cube plus a 1 s
square plus a 2 s plus a 3 which is equal to 0, this the standard form of characteristics
equation.
Now, it is obvious that, a naught equals 2, a 1 equal to 3, a 2 equal to 1 plus K c and a 3
equal to 2. a naught, a 1 and a 3 these are positives coefficients but we cannot say about a

2. But, if a 2 is negative, there is no need of further analysis, we can directly say that, the
close loop process is definitely unstable. Therefore, for the time being, we will assume
that 1 plus K c is positive quantity and we will try to find the stability condition based on
the value of K c.
So, we are assuming all coefficients starting from a 1 to a 3 are positive coefficients then
we cannot say whether system is stable or unstable. If all the coefficients are positive, we
cannot say, whether system is stable or unstable. So, to make any conclusion, we need to
go for further analysis and in the next step, what is require to do, we need constitute
routh array. Routh array consists of n plus 1 number of rows, it includes n plus 1 number
of rows.
Here, n is 3 so, routh array should constitute with 4 rows so, this is a row 1, row 2, row 3,
row 4. Now, you will just write all the coefficients of characteristics equation in the first
and second row only. So, first one is 2 which is a naught, second one is a 1 which is 3
then next one is 1 plus K c, next one is 2 and there is no more coefficient. So, we can put
0, 0. So, what will be the first element of third row, can you determine the first element
of third row, 3 multiplied by 1 plus K c minus 4 divided by 3.
What will be second element, second element will be 0 then what is the first element in
fourth row, 2 so, we got all the elements in the first column based on which we proceed
for stability analysis so, we got all the elements of the first column.
(Refer Slide Time: 07:10)

So, the first Colum we can write as 2, 3 then third element is 3 K c minus 1 divided by 3,
can I write the element of third row and last element is 2. So, this is the first column in
the routh array now, you will try to consider some values of K c arbitrarily. Suppose, if K
c is equal to 1, what about the third element in the first column, the third element in the
first column becomes two third. So, the system is stable or unstable, system is stable
because all the elements in the first column are positive.
So, the close loop system is stable and K c is equal to 1 because all the elements in the
first column are positive. We will consider another value suppose, K c is equal to 1 by 6,
if K c is equal to 1 by 6, the third element in the first column becomes how much, minus
1 by 6. So, the process is, the close loop system is stable or unstable, unstable if K c is
equal to 1 by 6, the close loop system is unstable. Now, if this is unstable, how many
poles lie in the right half plane, 2 because if we consider K c is equal to 1 by 6 then the
first column becomes this.
If we consider K c equals 1 by 6 then first column becomes this now here, two sign
changes are involved one sign changes is from second row to third row, second sign
changes is from second row to fourth row. Since two sign changes are involved so, if K c
is equal to 1 by 6, 2 poles lie in the right half plane. In the next, we will try to find the
critical stability condition no, before critical stability condition, we will find stability
condition.
(Refer Slide Time: 10:30)

I mean, for what value of K c, the close loop system suite always be stable, can you find
that. So, the system will always be positive, always be stable when 3 K c minus 1
divided 3 is greater than 0 then we can say that, the close loop system remains always
stable. So, from these, we can find that K c should be greater than one third so, this is the
stability condition. If K c is greater than one third then the close loop system remains
always stable in the next, we will discuss the critical stability condition.
See, one thing is clear that, if K c is greater than one third then the close loop system is
stable, if K c is less than one third then system is unstable. If K c is equal to one third
then we can say that, this the critical stability condition so, this the critical stability
condition. And K c is equal to one third, this is the stability condition, can you determine
the third element of first column when K c is equal to one third, 0. When K c is equal to
one third, the third element in the first column becomes 0 then what will be the all
elements in first column.
(Refer Slide Time: 13:03)

First element is 2, second element is 3, third element is 0 so, upto third row, we can
determine the elements in the first column. If the elements in the third row vanish then
what about the close loop response, we have discussed in last class, sustain oscillation.
So, if all the elements from the third row vanish then we can conclude that, the close
loop response provides sustain oscillation I mean, oscillatory response with constant
amplitude.

In this situation, we can say that, we get sustain oscillations when K c is equal to one
third. When K c is equal to one third, the close loop response provides oscillatory
response with constant amplitude that means, the poles lies on the imaginary axis, agree
it indicates that the poles lie on the imaginary axis. Can you determine those poles which
are basically complex conjugate poles with real part is 0, we can determine say, this is
first row, this is second row, this is third row, this is fourth row suppose, the elements
vanish from p th row suppose, all the elements banish from p th row.
Now, to determine the complex conjugate poles with real part is equals 0, we need to
take the elements from p minus 1 th row, to determine the complex conjugate poles with
real parts equals 0 we have to take the elements from p minus 1 th row. That means, if
we write other elements then routh array becomes like this agree, this is the routh array
we are saying that, this is p th row. Now, all the elements vanish from this row now, we
will take the elements just from p minus 1 th row so, one element is 3, another element is
2 which is equal to 0.
And you remember, we only take the elements from p minus 1 th row and from first and
second column. All the elements vanish from p th row, we will take elements from p
minus 1 th row and from first and second column. So, we can write 3 s square plus 2
equals 0 by solving this, we can write, s equals plus minus j root over of 2 by 3. So, these
are the two complex conjugate poles, which lie on the imaginary axis.
So, when K c is equal to one third, we get sustain oscillation and the corresponding
complex conjugate roots are this, which lie on the imaginary axis. In the next, we will
discuss another technique which is root locus technique.

(Refer Slide Time: 18:37)

The Routh-Hurwitz taste was purely algebraic but this root locus technique is a
graphical technique, root locus technique is a graphical technique. Now, question arise is,
how we can produce the graph, we can make the graph by plotting all the poles, all the
roots of characteristics equation in the complex plane. So, we can produce the graph by
plotting the roots of the characteristics equation in the complex plane. This is a graphical
technique and we can produce the graph by plotting all the roots of characteristics of
equation in the complex plane.
The location of all the roots is seen at a glance in the plot and we can conclude on
stability, by seeing the overall picture. We have the characteristics equation, we can
determine the roots now, we will consider one complex plane and we locate all the roots
in the plane. That means, at a glance, we can see the position of all the roots in this graph
and we can conclude on stability based on this overall picture, this is the root locus
technique.
Now, we will start to discuss this root locus technique taking one example, we will
discuss this root locus technique taking one problem.

(Refer Slide Time: 21:57)

The close loop block diagram is require to develop first suppose, we are considering the
process is under P only controller, the process is under P only controller. That means,
transfer function of the controller G c equals K c, the transfer function of the final control
element G f equals 2. The transfer function of the process we are considering as G p
equals 0.25 divided by s plus 1 multiplied by 2 s plus 1. So, this is a second order
process, this is the transfer function of the process, output is y bar.
Transfer function of measuring device G m equals 2, the output of these measuring
device is suppose, y and the output of the comparator is epsilon. So, this is a close loop
diagram, this is the closed loop block diagram for servo case or regulatory case or for
both, this for servo case, because we considering no disturbance. Now, you will analyze
the closed loop system based on root locus technique, can we find characteristics
equation, find the characteristics equation.
The general form of characteristics equation is 1 plus G c G p G f G m equals 0, this is
the general form. Now, we substitute all the individual transfer function in this general
form, if we substitute all the individual transfer functions, we get 1 plus K c divided by s
plus 1 multiplied by 2 s plus 1 equals 0. This is a transfer function, close loop transfer
function now, another form also we write that is, 1 plus G suffix O L equals 0 where, G
suffix O L is the transfer function of the open loop system.

(Refer Slide Time: 26:49)

Now, comparing the last two equations we can write, G OL equals K c divided by s plus
1, 2 s plus 1, this the form of the open loop system. Can you find the poles of open loop
system transfer, what are the poles, the open loop transfer function has two poles, one is
at s equals half, another one is at s equals minus 1 so, two poles are there. What about
zeros, no zeros so, there are two poles and no zeros of the open loop transfer function
remember, open loop transfer function G OL. Now, we will try to find the roots of
characteristics equation.
(Refer Slide Time: 28:12)

So, our characteristics equation is 1 plus K c divided by s plus 1 multiplied by 2 s plus 1


equals 0. We can write this equation by simplifying s, 2 s square plus 3 s plus 1 plus K c
equals 0, can we write the characteristics equation in this form so, this is the
characteristics equation which we got after simplifying. Then we can write the root as r
equals minus 3 plus minus root over 9 minus 8, 1 plus K c divided by 4.
The characteristics equation in quadratic form obviously, there are two roots so, we will
write as r 1 and r 2, r 1 is minus 3 by 4 plus root over of 1 minus 8 K c divided by 4 this
is the first root. Second root r 2 is equal to minus 3 by 4 minus root over of 1 minus 8 K
c divided by 4 this is second root. Now, you will generate the roots varying K c value
from 0 to infinity, the root locus diagram is produced varying K c from 0 to infinity.
(Refer Slide Time: 30:44)

So, we will just produce one plot varying K c and we will determine the corresponding r
1 and r 2 values. So, first we will first consider K c equals 0 then what is r 1, how much
is r 1 you keep on determining the values. If K c is equal to 0 how much is r 1, minus
half, how much is r 2, minus 1. If K c is 1 by 8, how much is r 1, minus 3 by 4 and r 2 so,
both roots are equal. If K c is equal to 1 by 4, how much is r 1, minus 3 by 4 plus j 1 by
4, another one will be minus 3 by 4 minus j 1 by 4 basically, they are complex conjugate
roots.
If we consider K c equals half, r 1 becomes minus 3 by 4 plus j root 3 by 4 and second
root will be minus 3 by 4 minus j root 3 by 4. Next, we will consider K c is equal to 1

then r 1 becomes minus 3 by 4 plus j root 7 by 4 and second root is minus 3 by 4 minus j
root 7 by 4 like this, we can go up to infinity. Now, based on these data I mean, we will
produce in next, the root locus diagram putting these two roots.
(Refer Slide Time: 34:02)

So, our complex plane is this one, this is real axis, this is indicating imaginary axis when
K c is equal to 0, r 1 is minus half say, if r 1 is minus half indicates this point, r 1 is
minus half and corresponding K c value is 0. When K c is equal 0, another root r 2 is
minus 1, corresponding K c value 0 so, when K c equal 0, these two roots we have
determine and they are located now in the complex plane. In the next, we will consider
suppose, K c equals 1 by fourth, before that we will consider K c equal to 1 by 8.
If K c equal to 1 by 8 then we have only one root that is, minus 3 by 4 and minus 3 by 4
is in between minus 1 and minus half agree, corresponding K c values I am writing here
that is, 1 by 8. Next, we will locate the roots when K c equals one fourth see, when K c
equals one fourth, we have two complex conjugate roots. Real part is minus 3 by 4 that
means, this point and complex part is 1 by 4 and another one is minus 1 by 4. So, I am
just writing here, this root corresponds to K c equals one fourth, this is also we got when
K c equals one fourth.
So, these two points we obtained when K c equals 0, these two points we obtained when
K c equals one fourth, the single root we obtain when K c equals 1 by 8. Next, we will
consider K c equals half then these two roots we can locate in complex plane, K c equals

half, K c equals half. You see, the real part is same minus 3 by 4, only imaginary part
gradually increasing with the increase of K c value. So, by this way, it will increase upto
infinity as K c approach is infinity now, you will just connect all these points.
So, we get branches like this, the deduction is this, like this so, these are the branches, by
connecting all the points we got this picture. So, this is when K c approaches infinity, this
is when K c approaches infinity. Now, this is the root locus diagram for the example
system, which we produced locating the roots of characteristics equation. Now, we will
make some remarks based on this plot.
(Refer Slide Time: 39:24)

We will note down few important points based on this plot so, how many root loci are
there, how many branches are there. Number of root loci or number of branches is equal
to 2 so, that means, number of poles of G OL. So, we can conclude based on these root
locus diagram that, the number of root loci or branches equals number of poles of G OL
that is, 2. Two branches are there and we obtain also two poles of the open loop transfer
function here second point is, the root loci originate from which point, root loci originate
from the poles of G OL that is I mean, there minus half and minus 1, agree.
The root loci originates from the poles of G OL, the poles are minus half and minus 1
can I say, the root loci originate when K c equals 0, yes. The root loci originate when K c
equals 0 and the root loci terminate when K c is equals infinity recall that, the root locus

diagram is produced varying K c from 0 to infinity. So, when K c equals 0 I mean, the
root loci originate when K c equals 0 and root loci terminate when K c equals infinity.
What about the stability, is there any crossing of root loci from left side to right side no,
if you see the root locus diagram, it is clear that, there is no crossing of root loci from left
side to right side. That means, the close loop system is stable irrespective of K c value I
mean, you can take any large value of K c, there will not any in stability problem. So,
third point is the close look system is stable irrespective of K c value, there is no in
stability problem.
You can observe another thing that, as the value of K c increases, the poles of the closed
loop system move away from the real axis.
(Refer Slide Time: 43:37)

With the increase of K c value, the poles of the closed loop system they move away from
the real axis. See here, this root we got when K c equals one fourth, if we increased to K
c half, the root is moving away from the real axis in both sides what it indicates,
oscillatory response

(Refer Slide Time: 44:11)

So, you write as the value of K c increases, the poles of the closed loop system move
away from the real axis that means, the response becomes more oscillatory so, these are
the conclusions, we can make best on the results we obtained. In the next, we will try to
find some general features, best on the result we obtained in solving this problem.
(Refer Slide Time: 45:20)

We will try to obtain some general characteristics based on the results we obtain in
solving the pervious problem, general characteristics of a root locus plot. Now, root locus
plot is produced by plotting, by locating the roots of characteristics equation. The general

form characteristics equation is written as 1 plus G c G p G f G m equals 1 plus G OL


equals 0, this the general form of the characteristics equation. Now, we will write this
equation as 1 plus K p K c K f K m, N s by D s equals 0.
We are writing the characteristics equation in this form where, K p K c K f K m are the
gains of process, gain of the controller, gain of the final control element, gain of the
measuring device and N s is a polynomial of order n. Similarly, if in the denominator,
another polynomial exist that is D s, D s is the polynomial of order r, these are the two
polynomials. Now, for any physically realizable system, r should be physical
realizability, r should be less or equals to n. For any physically realizable system it will
be just opposite, r should be greater or equals to n.
Now, we will just simplifying characteristics equation, if we simplifying characteristics
equation we get, D s plus K p K c K f K m N s which is equals to 0. Suppose, this the
equation 1 and this is the equation 2 now, what is the order of this equation, r. If r is
greater than n then definitely order of the equation is r. So, the number of roots of this
characteristics equation is number of poles of G OL.
We can write that, number of roots of characteristics equation equal to number of poles
to G OL. And that is equal to r so, this is the first characteristics next, I mean, the second
characteristics that we will discuss.
(Refer Slide Time: 51:17)

So, in the next, we will discuss that, the root loci originate when K c equals 0 and it
terminate when K c approach is infinity that means, it originates at the poles, terminate at
the zeros. Now, we got the characteristics equation D s plus K p K c K f K m N s which
is equal to 0, which we got in the last slide as equation 2. Now, we will consider K c
equal to 0 so, what the characteristics equation gets I mean, the characteristics equation
becomes D s equals 0.
When K c equal to 0, D s becomes 0 so, we should start the root loci diagram, plotting
the roots of D s equals 0 that means, the root loci originate at the poles of G OL. Next,
we will consider K c tends to infinity, if K c approaches infinity what form of the
characteristics equation we get, we have to write the characteristics equation by this way,
D s by K c plus K p K f K m N s equals 0.
Now, when K c approaches 0, we get K p K f K m N s equals 0 so, we should aim the
roots locus diagram plotting the roots of the N s equals 0. So, it proves that, the root
locus diagram terminate when K c tends to infinity.
Thank you.

Process Control and Instrumentation


Prof. A. K. Jana
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture No. # 22
Feedback Control Schemes (Contd.)

(Refer Slide Time: 00:58)

We will start today with the topic tuning of control parameters, tuning of control
parameters. We have disused earlier three different control schemes, p only controller, PI
controller and PID controller, for p only controller one tuning parameter is there, that is k
c propositional gain. In PI controller two tuning parameters are there, one is k c another
one is tau i, tau i is the integral time constant. And for PID controller there are three
tuning parameters, proportional gain k c, integral time constant tau i and derivative time
constant tau D, so these are the tuning parameters.
How will go to apply the control equations, we need to use the values of this tuning
parameters, so that we get best possible close loop performance. If we take simple
example, here we make a plot between closed loop process output y t and time t. Now, in
this process a step change in set point value is introduced, this is the set point value.

We are getting this type of close loop response, for a particular set of tuning parameter
values and if we use another tuning set, we get suppose this type of response. Say this is
one type of response and this A 1 type of response and this response is suppose A 2, here
A 1 and A 2 are two close loop responses provided by two tuning sets of a controller. A
1 and A 2 are two close loop responses provided by two tuning sets of a controller, we
get response A 1 and we considers k c 1 and tau i 1 suppose, this is the close loop
response of a process under PI controller.
This is the closed loop response and the process includes a PI controller. So, PI controller
has two tuning parameters one is k c another one is tau i. Now, we get A 1 response and
we consider k c 1 and tau i 1, we get the response A 2 and we consider another tuning set
namely k c 2 and tau i 2. So, the close loop response varies with the tuning parameter
values. Now, there are different criteria based on which we need to select the tuning
parameter values.
(Refer Slide Time: 06:18)

Suppose our first criteria is to keep the rise time as small as possible our target is to keep
the rise time as small as possible, this is our target. So, which tuning set we will prefer k
c 1 and tau i 1. So, if this is our target I mean, if we need to keep the rise time as small as
possible then we will prefer the tuning set corresponding to A 1 I mean we prefer k c 1
and tau i 1. Now, suppose another criteria is keep the over soot as small as possible.

So, if this is our target which tuning set we will use, second tuning set I means k c 2 and
tau i 2 agree, like this there are different characteristic features, like rise time over soot
there are few more characteristic feature, they are like settling time. Settling time should
be as small as possible, another characteristic feature is decay ratio. Now, one thing is
clear that we are in dilemma in selecting the tuning set, which one will select k c 1 tau i 1
or k c 2 tau i 2.
(Refer Slide Time: 08:37)

Now, you see one thing, if our target is keep to rise time as small as possible, then we
need to consider the close loop response at a particular time instant. And we will proceed
to tune the controller; that means, we can tune the controller, considering the response at
particular time instant, but it will be nice if we consider the response throughout the
operation period, in selecting the tuning parameter values. If we consider the entire time
period if we consider the response throughout this time period and if we tune the control
parameters, we will defiantly get better performance.
So, in that way the time integral performance criteria is consider for tuning the control
parameters.

(Refer Slide Time: 09:59)

So, we will consider the time integral performance criteria, which take in to account the
response throughout the period. Now, there are different time integral performance
criteria, one is Integral of the Square Error integral ISE, it is represented as ISE equals 0
to infinity epsilon square t d t, this is basically not infinity, this is the time period for
which or based on which we select the tuning parameters. Epsilon t is the difference
between set point value of y and process y, this is the integral square error 0 to infinity
epsilon square t delta t epsilon t d t.
Another time integral criteria is integral of the absolute value of the error, Integral of the
Absolute value of the Error IAE, integral of the absolute value of the error. IAE is equal
to 0 to infinity absolute value of epsilon t d t, this is the expression for IAE. Another time
integral performance criteria is, Integral of the Time weighted Absolute Error, which is
denoted by ITAE in short form ITAE. ITAE is written as ITAE equals 0 to infinity t
epsilon t d t this are the three time integral performance criteria which are used in
selecting the control parameters values.
Now, we will discuss the suitability of this three time integral criteria

(Refer Slide Time: 14:27)

Suitability, if we want to strongly suppress large errors, it is better to use ISE


performance criteria. If we want to strongly suppress large errors ISE performance
criteria is used. It has this form, now if the errors, if we square the errors the error
becomes larger therefore, this ISE performance criteria is used, if we use to suppress
small errors, then IAE performance criteria is used.
If small errors are involved ISE performance criteria is not used because if you take
squares of small errors, this term because in smaller therefore, this IAE performs criteria
gets preference, when small errors are there. And when the errors persist for a long time,
errors persist for a long time then ITAE performance criteria is used. Now, we will
considers the response of a process suppose this is indicating the set point now, this is the
close loop response.
Now, it is obvious from this figure that as time progresses, the error decreases. So, as
time increases, the error decreases definitely for it is stable system. Now, in the ITAE
performance criteria, the time is multiplied with error it is written by this form t
multiplied by e t. Now, in this figure it is clear thus at as time increases the error
decreases. Now, if we multiply this time which the error, the enter process response gets
weights in selecting the control parameters, if we multiply this time which the error, the
enter process response gets weights in selecting the control parameter values.

Therefore, this ITAE performance criteria is used in selecting the control parameter
values. Now, how you can tuning the control parameters using this time integral
performance criteria, that we will discuss with an example.
(Refer Slide Time: 19:10)

How we can tune the parameters by using this time integral performance criteria that we
will discuss in the next. So, in this example first we will consider one close loop block
diagram and in this example we consider a PI controller, which has the transfer function
G c equals k c 1 plus 1 divided by tau i s this is the controller block, for simplicity we are
considering G f equals 1 and the transfer function of the process, has this form G p
equals 2 divided by s plus 1 this is the disturbance block and this is equals G d output is
y bar, for simplicity again we are considering g m equals 1.
And error signal is going to the control block. This is the closed loop block diagram.
Now, we will try to derive the close loop transfer function, the general form of close loop
transfer function is written as y bar equals G c G p G f divided by 1 plus G c G p G f G
m y set point bar plus G d divided 1 plus by 1 plus G c G p G f G m d bar, if we
substitute all the individual transfer functions then finally, we get y bar equals tau i s
divided by tau square s square plus 2 zeta tau s plus 1 y set point bar plus tau i divided by
2 k c multiplied by s divided by tau square s square plus 2 zeta tau s plus 1 d bar.
If we substitute the individual transfer function expressions in this general form, we get
this close loop transfer function.

(Refer Slide Time: 23:36)

Here, tau equals tau i divided by 2 k c and zeta equals half root over of tau i divided by 2
k c 1 plus 2 k c you see both tau and zeta are the function of k c and tau i. Now, we will
consider only the servo problem for tuning, then the close loop transfer function becomes
y bar equals tau i s plus 1 divided by tau square s square plus 2 zeta tau s plus 1 y set
point bar, this is for the case of this is plus 1 tau i s plus 1 divided by tau square s square
plus 2 zeta tau s plus 1 y set point bar, this is the close loop transfer function for the
servo case.
In the next we will consider a step change in y set point; in the next step we will consider
a unit step change in set point. Accordingly, we will write y bar equals tau i s divided by
tau square s square plus 2 zeta tau s plus 1 multiplied by 1 by s and we consider, unit step
change in y set point. Next we need to take the inverse of Laplace transform, if we take
the inverse of Laplace transform we get the expression for y in time domain.

(Refer Slide Time: 26:26)

If we take the inverse of lap less transform of this, then we get the expression for y as y t
equals 1 plus 1 divided by root over of 1 minus zeta square, exponential minus zeta t
divided by tau multiplied by tau i divided by tau sin root over of 1 minus zeta square t by
tau minus sin root over 1 minus zeta square t by tau plus tan inverse root over 1 minus
zeta square divided by zeta, if we take the inverse of lap less transform of this equation
we get y t expression. Now, among the three time integral performance criteria suppose,
we is to consider ISE performance criteria.
So, ISE is equal to integration of 0 to infinity y set point t minus y t whole square d t
among the three time integral performance criteria, we are interested to tune the control
parameters using ISE performance criteria. So, ISE is equal to 0 to infinity y set point t
minus y t whole square d t, this y set point t is supplied externally by the person who is in
charge of operation, this is supplied externally y set point t and the expression for y t we
have derived. So, for y t use the derived equation.

(Refer Slide Time: 29:44)

Now, what is our target, what is we required to do, we the ISE should be minimized with
respect to tau and with respect to zeta. If we minimize the ISE we will get the optimum
values of tau and zeta, if we minimize the ISE with respect to tau and zeta we will get the
optimum values of tau and zeta. Suppose, this optimum values are tau star and zeta star.
Now, what are the expression of tau and zeta, the expression of tau and zeta we obtained
previously as this is the expression for tau and this is the expression for zeta.
Now, we will write tau star equal root over of tau i divided by 2 k c and similarly, zeta
star equals half root over of tau i divided by 2 k c multiplied 1 plus 2 k c, this two values
tau square and zeta star we obtained by minimizing ISE. Now, we have the expressions
for tau and zeta those are written in this form now, you see we have 2 equations and 2
unknowns, we have 2 equations and 2 unknowns tau i and k c solving this two equations
we can obtain the optimum values of tau i and k c solving this two equation we get
optimal values of tau i and k c.
So, by this way by using the time integral performance criteria, we can obtain the
optimal values of tuning parameters. Similarly, we can consider PID controller, we can
consider other time integral performance criteria for the approaches like this. Next we
will discuss another tuning approach that is Cohen coon technique. Now, so for we
discuss routh Hurwitz, we have discus root locus technique and time integral
performance criteria.

(Refer Slide Time: 33:06)

First we discussed routh Hurwitz test. In this test we need the process model, in the form
of transfer function. In the routh Hurwitz test, we need the process model in the form of
transfer function. Secondly, we started the root locus technique, which is a graphical
technique and in the root locus technique also we need the process model because for
root locus technique we need to produce the plot by positioning the roots of the
characteristic equation.
So, for root locus technique also we need the process model, but these two techniques
cannot provide the exact values of tuning parameters, this two techniques only can
provide the range of the tuning parameters, this two technique do not provide the exact
values of tuning parameters, they only provide the ranges of tuning parameters therefore,
there not called as tuning techniques, it is better to call them instability technique.
Thirdly we discuss the time integral performance criteria, I mean controller tuning by
time integral performance criteria, this technique requires the process model because in
the time integral performance criteria we need the expression for y t in the time integral
performance criteria we need the expression for y t; that means, process model is
required and this technique provides the exact values of tuning parameters. So, this is a
tuning method. Now, best on this first point we can say that, it can provide better
performance by using more accurate model, this time integral performance criteria
involves the process model.

Now, if we give more accurate model, we will get better performance. So, it provides
better performance, when we use more accurate model. So, therefore, we look for an
efficient tuning method that uses empirical rules. Since, the time integral performance
criteria uses or depend heavily on the process model, we look for an efficient control
technique that uses the empirical rules.
(Refer Slide Time: 38:12)

So, next we will discuss one technique which is based on the empirical rules, that is
Cohen coon method, this method was proposed by Cohen and coon. This is an empirical
tuning method, this method is also called as process reaction curve method. Now, before
discussing this method we need to draw the block diagram, close loop block diagram. So,
this is y set point, this is controller output, this is the block for final control element then
the process, disturbance, process output, this process output is connected with the block
of measuring device and the measuring device output is y m bar.
Now, according to this Cohen coon method, the closed loop block diagram is opened by
disconnecting the final control element from controller, according to the Cohen coon
method this block is this loop is opened by disconnecting the final control element from
the controller. Now, if we disconnect the final control element from the controller, then
we can write the output y m bar divided by input c bar s equals G f G p G m, definitely
this is for the case of servo problem, can we write the output y m bar divided by the input
c bar equals G f G p G m and this is the transfer function of the process reaction curve.

Now, it is obvious that the output y m depends not only on the process, it depends also
on that final control element and measuring devise, the output y m depends not only on
the process, but also on the final control element and measuring device. Now, Cohen and
coon they absorbed that for most of the chemical process they got sigmoidal response.
(Refer Slide Time: 42:55)

Cohen and coon absorbed that for most of the processes, they obtained sigmoidal
response, sigmoidal shape, like this. Suppose, this quantity is B; that means, the set point
value, not set point it may be the y value at new study state, B is the value at new study
set, for most of the chemical process Cohen and coon absorb that, the process look like
sigmoidal shape. Now, this response can be approximated, by the first order plus date
time, this is basically the response of a first order plus date time.
So, we can write the first order plus date time as k exponential minus t d s divided by tau
s plus 1, so most of the chemical processes the response looks like this. Now, this shape
can be approximated by first order plus date time and this is the expression of first order
plus date time response, the first order plus date time response can be represented by this
equation, here k is static gain. So, what will be the expression for gain, for most of the
chemical process Cohen and coon absorb this type of response and a step change is
introduce in c.

(Refer Slide Time: 45:58)

Sigmoidal shape is obtained and a step change with magnitude a is introduced in c.


(Refer Slide Time: 46:20)

Now, can you find the expression for k change in output that is B divided by change in
input that is A agree, gain is the change in output divided by the change in input. Next
one is the t d, t d is date time, date time value you can obtain from this plot, this is the
date time. Date time value we can obtain from the plot. Now, if we exclude the date time
plot, then the response becomes the response of the first order system. If we exclude the

date time part from this response, then the response becomes the response of first order
system.
If we exclude the date time part from this over all response, then the remaining part is the
response of first order system. Now, how much is this quantity for a first order system
we have plotted y verses a k p verses t divided by tau p and we got this equals 1 for the
first order system we plotted y divided by a k p versus t divided by tau p and we obtained
at time t equals 0 this quantity equals 1. So, how much is this quantity, if you plot y
verses t this is tau t divided by tau p equals 1; that means, t equals to tau this quantity is
tau.
If we consider the slope of this is s this slope of this tangent is s, then we can write tau
equals B by s can we write, if s is the slope of this tangent at starting from t equals 0 then
we can write tau equals b by s, so by this why you can determine, all the parameters
involved in this expression. Now, after finding the parameters of the expression Cohen
and coon derived the empirical equations, introducing change in load variable, after
finding the parameters involved in the first order plus date time expression Cohen and
coon derived empirical equations.
(Refer Slide Time: 50:26)

Empirical equations and they have derived the equations introducing change in load
variable, for p only controller they got the expression for k c here, k c equals 1 by k tau
by t d 1 plus t d by 3 tau for PI controller they got 2 equation, one for k c another for tau

i. This expression they got for k c and for tau i they obtain tau i equals t d 30 plus 3 t d by
tau divided by 9 plus 20 t d by tau for PID controller they derive three equations, one for
k c one for tau i and one for tau d.
This expression they got from k c for tau i the expression can be written as tau i equals t
d 32 plus 6 t d divided by tau divided by 13 plus 8 t d divided by tau and for tau D they
got tau D equals t d 4 divided by 11 plus 2 t d divided by tau, this empirical correlations
Cohen and coon derived by introducing change in load variable.
Now, if you compare the three equations for p PI and PID 3 k c expressions, which one
is the largest I mean which k c is largest k c for p PI or PID k c value of pi controller is
the lowest and k c value of PID controller is the highest, reason is very obvious if you
introduce integral action there may be some instability problem therefore, k c is taken
lowest and PID controller as better stability criteria therefore, highest k c value for PID is
consider.
Thank you.

Process Control and Instrumentation


Prof. A. K. Jana
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture No. # 23
Feedback Control Schemes (Contd.)

(Refer Slide Time: 01:00)

In the last class, we discuss Cohen coon technique. We discussed the algorithm on
Cohen coon method. So, today we will solve one problem based Cohen coon technique I
mean we will determine the tunic parameter values of the controller using this technique.
The close loop black diagram is given with the transfer function of controller as G c
equals k c 1 plus 1 divided by tau i s plus tau d s this is the transfer function of PID
controller, then as we discussed the final control eliminate and controller blocker
disconnected.
So, that is 1 here, then the final control eliminate output goes to the process, having the
transfer function G p equals 1 divided by s plus 3 whole square. This is the transfer
function of the process. Output of this block is y bar, then this output is measured, the
measured signal is then compared with set point of value, so this is basically not the
close loop black diagram, we have disconnected here. Suppose, a units step change is

introduced in the controller output signal, input a unit step change introduced in input
signal to the final control eliminate.
So, this is the process, first we need to form the transfer function of the process reaction
curve. So, transfer function Process Reaction Curve of PRC is equal to y m bar divided
by c bar, which is equal to G p G f G m for simplicity we have consider G m equals 1
similarly, we will consider G f equals 1. Now, if we substitute individual the transfer
functions G p G f g m then we get GPRC equals 1 divided by s plus 3 whole square. So,
this is the transfer function of the system example system, as mentioned that unit step
change introduced in the input signal to the final control element.
So, the process this transfer function becomes, not transfer function this output becomes
1 divided by s s plus 3 whole square, when c bar s equals 1 by s we are just substitute c
bar s equals 1 by s in this process reaction curve. Now, we basically need the y m in the
time domain so, for that we need to expand into partial reactions.
(Refer Slide Time: 07:05)

I mean we can write this as A 1 divided by s plus 3 plus A 2 divided by s plus 3 whole
square plus B 1 divided by s in the next we need to determine the values of this co
efficient say A 1 A 2 and B 1. So, finally, we will get this form y m bar s equals minus 1
divided by 9 s plus 3 minus 1 divided by 3 s plus 3 whole square plus 1 divided by 9 s.
That means we got A 1 equals minus 1 by 9 similarly, A 2 equals minus 1 3rd and B 1
equals 1 by 9. So, the determination of that is co efficiency is left for you.

So, this is the ultimate formed of y m bar. Next we take to inverse of lap lace transform,
take the inverse of lap lace of transform. If we take the inverse of lap lace form then we
get y m in time domain as y m t equals minus 1 by 9 exponential minus 3 t minus 13rd t
e to exponent especial of minus 3 t plus 1 by 9. So, this is the expression of y m in time
in domain, in the next step we need to determine the difference derivative of first
derivative and second derivative of y m, find the first and second derivative of y m.
The first derivative of y m represented by y m dot equals t exponential of minus 3 t
check this, first derivative of y m we get as t exponential of minus 3 t, this is the first
derivative. Now, second derivative we obtain as y m here, approximate use 2 dot
exponential of minus 3 t 1 minus 3 t this is the second derivative.
(Refer Slide Time: 11:32)

In the location of the inflection point obtain by setting the second derivative equals 0. So,
if we do that, then we get t equals 1 3 rd, this is the route of interest in this problem t
equals 13rd. So, the point of inflection of occurs at t equals 1 3rd. In the next we will
determine the slope of the tangent line, the slope of the tangent line through this point is
s equals first derivative at t equals 1 3 rd find the value, the expression is t exponential of
minus 3 t, if we substitute t equals 1 3 rd then we obtain 0.123. The slope we obtain
0.123 I think at this stage, it is good through drawer response.

(Refer Slide Time: 14:05)

This is time and this is y m t. So, at this point t equals 1 3rd, this point is correspond to
equal 1 3rd means this is the point of inflection. Now, this point represents the debt
time. Now, the slope we obtain 0.123. So, this is the process reaction curve, can you find
the corresponding y m value at t equals 1 3rd, find the corresponding y m value at t
equals 1 3rd.
(Refer Slide Time: 16:17)

We have the expression for y m that is minus 1 by 9 exponential minus 3 t minus 1 3rd t
exponential minus 3 t plus 1 by 9, if we substitute t equals 1 3rd than, we obtain this,
which is equal to 0.0294. So, the y m at t equals 1 3rd we obtain 0.0294.
(Refer Slide Time: 17:16)

So, this is 0.0294. Now, it is obvious from this flat that 0.0294 minus 0 divided by 1 3rd
minus t d equals what slope, it is obvious from this figure that this equals slope, a slope
is how much 0.123 from this co relation we can determine the value of their time, from
this co relation we can determine the value of debt time, how much is that 0.0943. So, we
obtain the debt time 0.0943 can you determine the ultimate value of y m t.

(Refer Slide Time: 18:50)

We represent the ultimate value of y m by B. So, B equals how much you calculate
ultimate value of y m t that is limit t tense to infinity y m t equals 1 by 9. Now, we
determine they gain k, we determine the gain k, k equals to B divided by A, B this is the
ultimate value of the response and A is the magnitude of input function.
So, B we obtain 1 by 9 and A is 1 because we have introduced in needs to change; that
means, k is 1 by 9. Next we need to determine the time constant tau, which is represented
by B by s. Now, B will obtained 1 by 9 and rates is 0.123. So, it gives 0.9033. So, tau is
0.9033 in Cohen coon method basically response of chemical processes is subjected to
step input can be approximated by the response of first order plus get time system. So,
accordingly we have determine the k tau and t d.
Now, what controller we have used PID controller, they have proposed the expression
for k c tau i and tau b, as the function of this k tau and t d like for PID controller the
expression for k c is represented as k c equals 1 by k tau by t d multiplied by 4 by 3 plus t
d divided by 4 tau. So, we know the values of tau t d and k substituting those values, we
obtain k c equals 117.2 substituting the values of k tau and t d obtain k c equals 117.2.
Similarly, the expression for tau i as the form tau i equals t d 32 plus 6 t d by tau divided
by 13 plus 8 t d by tau which is equals to 0.222 and the derivative time which is equal to
t d 4 divided by 11 plus 2 t d by tau and we can calculate this tau d equals 0.0336 by this

way we can tune the controller, using the Cohen-coon technique. Next we will discuss
the frequency response analysis of linier systems.
(Refer Slide Time: 24:10)

Our next topic is, frequency response analysis of linier processes. So, we need to
consider the linier process and the linier process is substituted to a sinusoidal input, this
it is ultimate response, after a long time is also a substantial sinusoidal wave, it can be
prove that, a linier system a subjected to a sinusoidal input. It is ultimate response after a
long time is also a substantial sinusoidal wave A A naught of amplitude, omega is the
frequency and phi is the phase leg.
So, this is the input which has the amplitude A, this is the output which has the amplitude
A naught and this output is out of phase by phi degrees or radian, this is representing phi,
phase leg, these are 2 o fs. Now, it is obvious that both inputs and outputs are sinusoidal
and the output is out of phase by phi degrees or radians. The ratio of output and input
amplitudes, is known as amplitude ratio enamel a is represented by A R, which is
amplitude ratio and phi is call the phase angle.
Now, this amplitude ratio and phi both vary with omega, how the amplitude ratio an phi
vary with omega is the main concern of frequency response analysis, amplitude ratio and
phi values vary with omega. Now, how the amplitude ratio an phi, vary with omega is
the main concern of frequency response analysis. Now, we discuss the fundamental
frequency response result.

(Refer Slide Time: 29:54)

The fundamental frequency response result. A complex number denoted by z whose real
part is a and imaginary part is b, can be expressed as z equals a plus j b, j b is the
complex number. Now, in the complex plane we will locate the z. So, this is the real part
a and this is the imaginary part b. Now, the magnitude of z, we can find as magnitude of
z equals root over of a square plus b square and argument of z which we can write by this
and it can be determined using this form, I mean argument z equals tan inverse b by a.
This is the argument z. Now, for a particular process we usually have the transfer
function G s. Now, if we substitute s equals j omega, then we get frequency response
transfer function replacing s by j omega, we obtain frequency response transfer function.
Now, amplitude ratio we can determine I mean the amplitude ratio is the magnitude of
this G j omega, amplitude ratio is the magnitude of the frequency response transfer
function. Similarly, the phase angle phi is the argument of this frequency response
transfer function, it can be represented by this notation.
So, this is the fundamental of frequency response result. Next, we discuss the general
procedure for frequency response for finding the response frequency.

(Refer Slide Time: 34:17)

General procedure for finding the frequency response. First we discuss generalizing then
we will take few examples to finding the frequency response are to analyze the
frequency response, say G s is the transfer function of a process. Now, input to this
process we can write as u bar s and output is y bar s in the first step, we need to replace s
by g omega in the first step we should replace s by g omega to obtain the frequency
response transfer function, this is step one, substitute s equals g omega and find the
frequency response transfer function.
In the next step represent this frequency response transfer function in Cartesian form;
that means, G j omega equals real part plus imaginary part multiplied by j. Now, this way
we can clearly identify the real part and imaginary part, this is step 2. In step 3 we can
determine the amplitude ratio and phase angle, we know amplitude ratio equals root over
of square of this real part plus square of this imaginary part. We can determine the
amplitude ratio using this similarly, the phase and here we can determine from this have
phi equals tan inverse imaginary part divided by real part.
So, we just substitute s equals j omega to get the frequency response transfer function,
then represent that frequency response transfer function in this form I mean just you try
to find out the real and imaginary part, after doing that use these 2 expressions to obtain
amplitude ratio and phase angle. Now, the alternative technique in the alternative
technique we can represent the frequency response transfer function in polar form.

(Refer Slide Time: 38:49)

So, we can represent the frequency response transfer function in polar form, amplitude
ratio exponential of j phi. Now, we can write this frequency response transfer function as
j omega equals the numerator divided by the denominator part. Now, we will apply the
polar form I mean we can write the numerator as amplitude and then the exponential
term. Similarly, we can write the denominator as first tau a magnitude of the
denominator multiplied by the exponential term, this is amplitude, this is magnitude of
this numerator multiplied by this exponential term.
Similarly, for the denominator this is the magnitude of the denominator D and this is the
exponential term. Now, this we can write again as numerator magnitude divided by
denominator magnitude, exponential j in j omega minus j d j omega, the polar form of
the frequency response transfer function we have written here amplitude ratio
exponential of j 5. Now, we will compare this 2, then we obtained the expression for
amplitude ratio as, amplitude ratio equals this, the magnitude of N numerator divided by
the magnitude of denominator and we obtained the expression for phi also.
So, we can determine the amplitude ratio phi by this way also, here the steps are
substitute s equals j omega to obtain the frequency response transfer function, then you
represent the frequency response transfer function by this form, this numerator by
denominator, then use this 2 expression to determine the amplitude ratio and phi. So, we

will take one example to find the frequency response I mean to discuss the frequency
response analysis.
(Refer Slide Time: 42:54)

So, we will consider is first order processes, first order system. The transfer function of a
first order system suppose, is given as G s equals k divided by tau s plus 1, k is the study
state gain of the first order system and tau is the time constant. Now, we proceed step by
step, in the first step we need to replace s by j omega. So, you substitute s equals j
omega, then we can write G j omega equals k 1 divided by 1 plus j tau omega.
So, we can write this again as k multiplied by 1 divided by 1 plus j tau omega 1 minus j
tau omega divided by 1 minus j tau omega, can we write this. So, this is equal to k 1
minus j tau omega divided by 1 plus tau square omega square. After substitute s equals j
omega we obtain this form. Now, we need to represent this frequency response transfer
function Cartesian form that is step 2. In Cartesian form we can write the frequencies
transfer response function has G j omega equals k multiplied by 1 plus tau square omega
square minus j tau omega divided by 1 plus tau square omega square, we can write this
from this expression.
Now, we need to find the real part and imaginary part. So, if you compare this with this,
we obtained the real part s, this real part equal k divided by 1 plus tau square omega
square and comparing this 2 again we get the imaginary part also. Imaginary part is equal
to minus k tau omega divided by 1 plus tau square omega square. So, you obtain in step

2 the real part and imaginary part. In step 3 we will determine the amplitude ratio and phi
using the expressions.
How much is this, amplitude ratio is equal to k divided by root over of 1 plus tau square
omega square. Similarly, we can determine the face angle phi.
(Refer Slide Time: 48:27)

Which is argument and this phi equals tan inverse, imaginary part divided by real part
substituting real and imaginary part, we obtain minus tan inverse tau omega. So, for a
first order system having the transfer function G s equals k divided by tau s plus 1 we
obtained the expression for amplitude ratio and phi, representing the frequency response
transfer function in Cartesian form. Similarly, we will represent the transfer function
frequency response transfer function in polar form, for finding the amplitude ratio and
face angle.

(Refer Slide Time: 49:47)

The frequency response transfer function we obtained, as G j omega equals k divided by


1 plus j tau omega, which can be written has N j omega divided by D j omega. Now,
what is N j omega, N j omega is equal to k. So, you can write this k plus j multiplied by 0
and we can write the denominator as 1 plus j tau omega, then the magnitude of N j
omega is how much k and argument is equal to tan inverse 0 by k which is equal to 0.
So, we obtain the magnitude and amplitude for the numerator. Similarly, for the
denominator we obtain the magnitude as 1 divided by tau square omega square agree, the
real square plus imaginary square whole to the power half and the argument is equal to
tan inverse tau omega. Now, we know the amplitude ratio is the ratio of 2 magnitudes,
magnitude of numerator divided by denominator magnitude of denominator, then we
obtain k divided by root over of 1 plus tau square omega square and phi is equal to
argument of N j omega minus argument of D j omega which is equal to minus tan
inverse tau omega.
Now, this 2 expressions we obtained also by representing the frequency response transfer
function in Cartesian form.
Thank you.

Process Control and Instrumentation


Prof. A. K. Jana
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture No. # 24
Feedback Control Schemes (Contd.)

(Refer Slide Time: 01:07)

Today, we will continue our discussion on frequency response analysis, we will start
with the discussion of frequency response of a second order system. In the last class we
discussed the frequency response of a first order system. So, the transfer function of a
second order system in general form can be written as K divided by tau square S square
plus 2 zeta tau S plus 1 this is a transfer function of a second order system in general
form. Now, we will follow the different steps for frequency response analysis. In the first
step we replace S by j omega, in the first step we put S equals by j omega in the transfer
function.
So, the frequency response transfer function represented by j omega becomes K divided
by 1 minus tau square omega square plus j 2 zeta tau omega. If we substitute S equals j
omega, we get this form this is frequency response transfer function, in the next step we
write the frequency response transfer function as K divided by 1 minus tau square omega

square plus j 2 zeta tau omega multiplied by 1 minus tau square omega square minus j 2
zeta tau omega divided by 1 minus tau square omega square minus zeta minus j 2 zeta
tau omega.
Can we write rearranging this we obtain K multiplied by 1 minus tau square omega
square minus j 2 zeta tau omega divided by 1 minus tau square omega square whole
square plus 2zeta tau omega whole square. So, substituting S equals j omega and
rearranging finally, we get this form of frequency response transfer function, in the
second step we need to represent this frequency response transfer function in Cartesian
form.
(Refer Slide Time: 05:30)

So, we can write the frequency response transfer function as K multiplied by 1 minus tau
square omega square divided by 1 minus tau square omega square whole square plus 2
zeta tau omega whole square, this is a real part. And the imaginary part, becomes minus
K two zeta tau omega divided by 1 minus tau square omega square whole square plus 2
zeta tau omega whole square. We have written this frequency response transfer function
has a combination of real and imaginary parts, so it is obvious that this part is the real
part, and this part is the imaginary part real part plus j imaginary part.
Now, in the third step we find the amplitude ratio and the phase angle, amplitude ratio is
determined using this form. So, how much is amplitude ratio, if we substitute real and
imaginary parts the amplitude ratio becomes K divided by root over of 1 minus tau

square omega square whole square plus 2 zeta tau omega whole square, so this is the
amplitude ratio for the second order system.
Similarly, we can find the phase angle phi using this form phase angle phi equals tan
inverse imaginary part by real part, accordingly we obtain minus tan inverse 2 zeta tau
omega divided by 1 minus tau square omega square. So, first we substituted S equals j
omega then we represented the frequency response transfer function in Cartesian form,
and in the step we determine the expressions for amplitude ratio and phase angle. Now,
we will represent the frequency response transfer function in polar form for finding the
expressions for amplitude ratio and phi.
(Refer Slide Time: 10:25)

We will represent the frequency response transfer function in polar form and we will find
the expressions for amplitude ratio and then phase angle, so first we write the frequency
response transfer function as the ratio of numerator and denominator. So, we got this
expression K divided by 1 minus tau square omega square plus j 2 zeta tau omega, this
expression we got after substituting S equals j omega. Now, if this is the ratio of
numerator term divided by denominator term then; obviously, this numerator term is
equal to K., so we can write this as K plus j 0.
Similarly, the denominator term we can write as 1 minus tau square omega square plus j
2 zeta tau omega, this is a numerator this is the denominator of the frequency response
transfer function. Now, what is the amplitude of this numerator term, amplitude is K

what is the amplitude of denominator term that is root over of 1 minus tau square omega
square whole square plus 2 zeta tau omega whole square, this is the magnitude of
denominator.
Now, what is the argument of this numerator 0 what is the argument of this denominator
tan inverse 2 zeta tau omega divided by 1 minus tau square omega square, these are the
argument of numerator and denominator and magnitude of numerator and denominator.
Now, amplitude ratio is written as magnitude of N j omega divided by D j omega, this is
of expression for amplitude ratio, if we substitute the magnitude of a N and D we obtain
K divided by root over of 1 minus tau square omega square whole square plus 2 zeta tau
omega whole square.
After substituting the magnitude of a N and D we obtain this form for amplitude ratio,
similarly the phase angle is obtained as magnitude sorry the argument of N minus the
argument of denominator. If we substitute the individual argument then we obtain minus
tan inverse 2 zeta tau omega divided by 1 minus tau square omega square, so these are
the two expressions for amplitude ratio and phi. The similar expressions we obtained
representing the frequency response transfer function in Cartesian form.
(Refer Slide Time: 15:45)

Next we will discuss the frequency response of several systems connected in series, in
the frequency response of N none interacting systems, say N none interacting systems
are connected in series, and the individual transfer function are G 1 then G 2 like this the

last one is G N. So, input to the process G 1 is a prime the output of G 1 is introduced as
the input 2 G 2 and finally, we obtain y N bar.
So, this is the, this is indicating the connection of N none interacting systems and now
the amplitude ratio of this overall system, can be represented as the multiplication of
individual amplitude ratios. The overall amplitude ratio this is the overall amplitude ratio
is the multiplication of individual amplitude ratio. Similarly, the overall phase angle phi
is the submission of the individual phase angles, this is a overall amplitude ratio, this is
the overall phi.
Now, amplitude ratio A R 1 and phase angle phi 1, these 2 correspond to G 1. Similarly,
amplitude ratio A R N and phase angle phi N this two correspond to G N, now we will
proof this two co-relations, we will try to find the overall amplitude ratio and phi. If N
none interacting systems are connected in series, then we can write the overall transfer
function represented by G equals G 1 G 2 G N, can we write this, they are multiplied.
Now, we will represent the transfer function in polar form, so we can write amplitude
ratio exponential j phi, this we have written for G for the overall system. Similarly, we
can write for G 1 as amplitude ratio 1 exponential j phi 1, this is for G 1, for G 2
similarly we can write amplitude ratio 2 exponential j phi 2 by this way the G N becomes
amplitude ratio N exponential j phi N. We represented the transfer function in polar
form, so generally we write G equals amplitude ratio exponential of j phi in polar form,
we can write the transfer function in this way.
Now, the right hand side we can write has amplitude ratio 1 amplitude ratio 2 amplitude
ratio N and exponential of j then submission of all the phase angles, can we write? We
can write the right hand terms in this way, now if we compare left hand side and right
hand side we obtain amplitude ratio is the multiplication of individual amplitude ratios.

(Refer Slide Time: 21:44)

If we compare the left hand side and right hand side we get the overall amplitude ratio as
the multiplication of individual amplitude ratios, now if we take logarithm in both sides.
Then we obtain this form, similarly for the phase angle we can write phi has the
submission of all individual phase angles, comparing left hand side and right hand side.
We obtain phi as the submission of all individual phase angles, so these are the two
expressions.
Sometimes this amplitude ratio is represented in logarithmic form and this is a
expression for phi, so these are basically the fundamentals of frequency response
analysis. In the next stage you plot different diagrams and then finally, will go for
stability analysis using this frequency response concept.

(Refer Slide Time: 23:46)

So, in the next we will discuss the bode plot, first we will discuss the bode plot in the
next we will discuss nyquist plot, then these two plots will be used for stability analysis.
So, this bode plot is a graphical representation of amplitude ratio and phi, bode plot is a
common graphical representation of amplitude ratio and phase angle, and this bode plot
consists of 2 graphs.
In one graph log amplitude ratio is plotted against log omega, in the bode plot bode plot
consists of two graphs in the first graph the amplitude ratio in logarithmic form is plotted
with log omega. And in the second plot the phi is plotted with log omega, so these are
the two plots which constitute the bode plots. Now, sometimes the amplitude ratio is
replaced by magnitude ratio, so we can write sometimes the bode plot includes log
magnitude ratio verses log omega.
This is magnitude ratio, it is nothing but the scaled amplitude ratio, and it is interesting to
mention also that this omega is also sometimes multiplied by some scaling factor. So,
those things we will discuss in different cases different tau A, now since we are
interested to plot log amplitude verses log omega therefore, log log graph paper is used
for this.
So, for the first plot log log graph paper is used and in the second plot phi is plotted with
log omega, so for this semi log graph paper is used. Now, we will take different
examples and we will try to make the bode plot, so first we will take a pure gain element,

we will make the bode plot taking different examples first we will consider a pure gain
element.
(Refer Slide Time: 27:40)

As mentioned the bode plot is produced by plotting the amplitude ratio and phi therefore,
first we need to determine the expressions for the amplitude ratio and phi. So, to do those
to do that we need to know the transfer function the transfer function of a pure gain
element is represented as K. So, this is the transfer function of a pure gain element, now
there is no need to express in terms of real and imaginary terms, we can directly find the
expressions for amplitude ratio, what is amplitude ratio K.
So, we can write this again as amplitude ratio by K equals 1, so here amplitude ratio by
K is the magnitude ratio; that means, magnitude ratio is equal to 1. Now, we will
determine the expression for phase angle phi, phase angle will be how much tan inverse
0, because imaginary part is 0 and real part is K, so 0 by K that is 0, so this is 0. Now,
bode plot considers the variations of amplitude ratio and phi with omega, but here you
see both magnitude ratio and phi they are independent of omega, so here both magnitude
ratio and phi are independent of omega.
Next we will produce the bode plot, so first plot is magnitude ratio verses omega, radians
per unit time. And this is log log graph paper, now since we obtain magnitude ratio
equals 1, so in the next step we can write as logarithmic of magnitude ratio equals log 1,

can we write this log of magnitude ratio equals log 1. Now, we will consider this, so
what it indicates log magnitude ratio equals log 1, this is a horizontal line.
So, the first plot we obtained like this, this is a horizontal line and it is equals 1, second
plot is made between phi verses log omega and remember this is the semi log graph
paper. This is the semi log graph paper has commented that phi is independent of omega
and that is equal to 0. So, first we will consider the axis this is suppose 0 degree, this is
minus 45 degree, this is minus 90, this is 45 degree and it looks like this. It is obvious
that both magnitude ratio and phi are independent of omega, so this is the bode plot for
the pure gain element.
(Refer Slide Time: 34:20)

Next we will consider another example that is pure capacity system, this is example two,
we will produce bode plot for pure capacity system. What is the transfer function for the
pure capacity system, G S is equal to K by S, this is a transfer function of a pure capacity
system. So, we will follow the ((Refer Time: 35:23)) steps I mean in the first step we will
substitute S equals j omega, then the frequency response transfer function yields G j
omega equals K divided by j omega.
Now, this we can write as K j divided by minus of omega a multiplied j in both
numerator and denominator, then we can write as 0 minus j K by omega. So, what is
amplitude ratio, we can write as the square of 0 square of K by omega and then square
root. So, this is equal to K by omega, how much is phi, phi is tan inverse minus infinity

minus 90, that is always that this phi is independent of omega, but this amplitude ratio is
dependent of omega and phi is independent of omega.
Now, we can write the amplitude ratio as amplitude ratio by K equals 1 by omega, can
we rearrange the expression of amplitude ratio in this form. If we take logarithm, then we
obtain log amplitude ratio by K equals minus log omega. So, this is basically is straight
line with slope minus 1, this is if straight line with slope minus 1, now we will produce
the plot bode plot.
(Refer Slide Time: 38:51)

So, this is the log log graph paper this is 0, starting point is 0.01next one is 0.1 1 10 100
0.01 1 and 100, now if we consider the, this equation this type line log magnitude ratio
equals minus log omega. Then we obtain this straight line with slope minus 1, if we
consider the expression of amplitude ratio in the form of magnitude ratio, we obtain this
straight line with slope minus 1, it has the slope of minus
1. In the next plot we consider phi and omega starting value is minus 100 degree another
value we are taking that is minus 90 0 degree then suppose this is 100 degree. Since, phi
is independent of omega and that is always minus 90 degree, so we get this horizontal
line for phi and this plot is produced in semi log graph paper. So, this is the bode plot of
a pure capacity system, so for we discussed the construction of bode plot with taking two
examples.

(Refer Slide Time: 42:13)

So, in the third example we will consider that time, in example three we will consider
pure time delay system, how we can represent the pure time delay system I mean what is
the transfer function. G S is equal to exponential of minus t d S t d is the date time, so
this is the transfer function of a time delay system. So, what is amplitude ratio, then we
can directly represent this transfer function in polar form, so if we express the frequency
response transfer function, we obtain G j omega equals amplitude ratio exponential of j
phi equals exponential minus j t d omega.
We can write this by this, now if we compare this and this then we obtain amplitude ratio
equals 1 and phi equals minus t d omega. So, the expression of amplitude ratio is
independent of omega, but phi depends on omega, so this is dependent of omega. Next
we will produce the bode plot for this pure time delay system, this is omega and this is
amplitude ratio, here we are considering amplitude ratio because no scaling factor is
there. This is log log plot, so the x axis starts from 10 to the power minus 2, 2 suppose 10
and this 1 starts from 10 to the power minus 1, 2 10.Now, this is a horizontal line
amplitude ratio equals 1, so we get this type of line in the first plot.

(Refer Slide Time: 46:07)

And in the next plot we consider phi verses omega, as usual this is the semi log plot and
for the phi we consider from minus 200 degree to 0 degree, this is suppose minus 50
degree. Now, if we produce some intermediate data, then we obtain this type of plot for
phi, to know the nature of this plot we need to produce some intermediate data varying
the omega value.
Ultimately, we get this type of response for phi, what it indicates, the phase angle
decreases infinitely as frequency decreases, as frequency increases, this is an interesting
finding for this time delay system. Next you discuss the construction of bode plot for a
first order system, which is quite complicated.

(Refer Slide Time: 48:55)

First order system, so to construct the bode plot we need to consider the transfer function
of a first order system, the transfer function of first order system is represented as say G
S equals K divided by tau S plus 1. We are not considering here the suffix p, which we
considered earlier, so this is the transfer function of a first order system. And in the last
class we determine the expressions of amplitude ratio and phi, in the last class we
obtained the expressions for amplitude ratio and phi for this first order system.
There amplitude ratio equals K divided by root over 1 plus tau square omega square, and
phase angle phi equals tan inverse minus tau omega. For a first order system this is the
expression for amplitude ratio and this is the expression for phase angle. Now, if we take
the logarithm of this expression, then we can write log amplitude ratio by K equals
minus half log 1 plus tau square omega square. Can we write this, if we take logarithm of
this expression we obtain this form.
So, in the next class we will discuss the asymptotic considerations, for constructing the
bode diagram of this first order system. So, today we discussed the frequency response
analysis of second order system, for N none interacting systems connected in series, then
we took three examples to discuss the construction of bode plot.
Thank you.

Process Control and Instrumentation


Prof. A. K. Jana
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture - 25
Feedback Control Schemes (Contd.)
(Refer Slide time: 01:08)

So, we will continued discussion on the constructions of prode plot in the last class we
discuss the construction of prode plot with taking three examples, fine. So, today will
discuss the construction of prode plot for a first order system. So, the transfer function of
a first order system is return as K divided by tau s plus 1 this is the transfer function of a
first order system now we untainted the expirations for amplitude ration and phi in the
previous class.
As amplitude ratio equals K divided by root over of 1 plus tau squire omega squire and
we obtained the expiration for phase angle as 5 equals ten inverse minus tau omega now
we will take the logarithm of this amplitude ratio expiration in both said then we can
write as amplitude ratio by K log amplitude ratio by K equals minus half log 1 plus tau
squire omega squire fine
So, this is the magnitude ratio basically next we will considered the asymptotic
considerations before making the plot we need to discuss the asymptotic considerations

first we will considered the amplitude ratio first we will discuss the amplitude ratio then
we will discuss the phase angle fine now if we considered tau omega tends two 0.
If we considered tau omega tens two 0 then amplitude ratio by K tens to 1 here the
amplitude ratio is multiply do it 1 by K that mean this is scaled amplitude ratio and here
wear considering scaled omega I mean tau omega we are considering now if we
considered tau omega tends to 0 then amplitude ratio by k tends to 1. So, we can write
that log amplitude ratio by K tends to log 1 fine. So, at the very low sequence is the log
log plot of magnitude ratio verses tau omega tends to the horizontal line.
At very low sequence is the log log plot of magnitude ratio verses tau omega tends to a
horizontal line represented by magnitude ratio equals amplitude ratio by K in equals 1
fine since this situations arise is at low frequence is this line is cold low frequency
asymptode since this situation arise is act low frequencies this line is cold low frequency
asymptode fine now if we if produce the plot considering this equation I mean this line
then what we obtained.
(Refer Slide time: 06:56)

. So, here wear considering tau omega and here magnitude ratio this is a log log plot. So,
wear using log log graph paper this is ten to the power minus 2 10 to the power minus 1
1 10 and this is suppose 10 to the power plus 2 and here wear considering 10 to the
power minus 2 10 to the power minus 1 than suppose this is 1, fine.

Now, we obtained this form log amplitude ratio by K approximately equal to log 1 when
we considered low frequencies; that means, it is clearly a horizontal line fine. So, this is
and this is the low frequency asymptote this is low frequency asymptote fine now under
asymptotic consideration next we will consider tau omega tends to infinity.
(Refer Slide time: 09:36)

In the next we will considered tau omega tends to infinity as tau omega tends to infinity
then we can write log amplitude ratio by K tends to minus log tau omega when tau
omega is much much better one we can neglect this one term I mean this is the general
form when tau omega tens to infinity we can neglect this one; that means, this right hand
said becomes minus log tau omega agree. So, when tau omega tends to infinity we can
write log amplitude ratio by K tends to minus log tau omega. So, the log log plot of
magnitude ratio verses tau omega tends to the line represented log magnitude ratio
approximately equal to minus log tau omega.
in the log log plot of magnitude ratio verses tau omega we obtain we need to considered
this equations log magnitude ratio approximately equals to minus log tau omega this is
nothing, but a straight line with slope minus one this is a straight line with slope minus
one and this situation arise is act high frequencies there for this straight line is cold as
high frequency asymptote since this situation arise is act high frequencies this straight
line is cold as high frequency asymptote fine. So, using I mean we will next considered
this equation in the prode plot. So, it is basically straight line with slope minus 1. So, this

is representing the high frequency asymptote this is representing high frequency


asymptote.
(Refer Slide time: 13:45)

It has slope of minus 1 fine in the next step we need to determined the point of
intersection in the third step we need to determined the point at which the two asymptote
intersect when we considered tau omega tends to 0, we got log amplitude ratio by K is
nearly equal to 1. This we considered low frequency asymptote when we considered tau
omega tends to infinity we got now this we got log 1. So, when we considered tau omega
tends to infinity we got log amplitude ratio by K which is approximately equal to minus
log tau omega fine now if we equate these two equations equating these two equations
we obtain.
Tau omega equals one agree equating these two equations this 1 and this 1 we obtain tau
omega equals one; that means, they inter set act tau omega equals one, now in this
situation we usually use one suffix for this omega that is in act this situations we use the
suffix in for omega; that means, we can write tau omega n equals 1 the value of omega
act which this two frequency to asymptote inter sets that is cold natural frequency. So,
here omega in is the natural frequency.
And most popularly it is cold as corner frequency it is also cold as break frequency fine
now can we determined the exact values of the magnitude receive add this corner
frequency what is the exact value of magnitude ratio act this corner frequency we have

the expiration of magnitude ratio that is magnitude ratio equals 1 divided by root over of
1 plus tau squire omega squire fine. So, this is the exact value of magnitude ratio act
corner frequency according the we need to construct the prode diagrams I mean this
magnitude ratio verses tau omega plot should lops like this like this and the
corresponding tau omega is 1 here.
(Refer Slide time: 18:38)

This is the corresponding natural frequency in this first plot. So, this is the asymptotic
considerations in terms of amplitude ratio and we obtained this plot similarly we need to
considered the asymptote consideration in term of phase angle. So, next we will
considered the phase angle as tau omega tends to 0 phi tends to what zero fine in the
second as tau omega tends to infinity how much is phi [vocalized- noise] minus 90; that
means, phi tends to minus 90.
So, phi never we lay less then minus 90 fine, it clearly indicates that phi never we less
then minus 90 and in the third situation, we will considered the corner frequency at the
corner frequency; that means, tau omega equals 1 how much is phi phi is minus 45 minus
tan inverse 1.

(Refer Slide time: 20:31)

So, this is minus forty five. So, these are the asymptote considerations in terms of phase
angle now we will drove the of phi verses log tau omega plot this is the semi log graph
paper and this is 1 which correspond to the corner frequency ten to the power minus 2 10
to the power minus 1 this is ten to the power plus 1 and 10 to the power plus 2, this is
suppose minus 100 degree minus 90 degree this is a minus 45 degree this is a 0 degree at
low frequencies at low frequencies the tau omega the five is 0 fine. So, it is somewhat
like this at low frequencies phi is 0 and when tau omega tends to infinity we obtain 5
minus 90.
And when tau omega equals 1 we obtained minus 45 agree at low frequencies this phi is
near to near 2 0 and act as tau omega tends to infinity this is minus 90 and act corner
frequency this is minus 45. So, this is the prode plot for a first order system this is the
prode plot of a first order system in the next we will discuss the construction of prode
plot for a second order system.

(Refer Slide time: 23:31)

In the next we will discuss the prode plot of a second order system. So, the transfer
function a second order system we considered earlier taking this form g s equals K
divided by tau squire a squire plus 2 zeta tau s plus 1 and in the last class we derived the
expirations for amplitude ratio and phi. So, they are amplitude ratio divided by k equals
one divided by root over of one minus tau squire omega squire hole squire plus 2 zeta tau
omega hole squire this expirations this expirations we derived earlier similarly the
expiration of phi is return as tan inverse minus 2 zeta tau omega divided by 1 minus tau
squire omega squire fine this 2 expirations we derived the earlier now we will discuss the
asymptotic considerations in terms of first amplitude ratio forward by phase angle.
Asymptotic consideration first we will consider amplitude ratio as tau omega tends to
zero amplitude ratio by K tends to one and log amplitude ratio by K tends to log 1 fine as
tau omega tends to 0 we obtained amplitude ratio by K tends to 1 and accordingly log
amplitude ratio by K tends to log 1.
Now, it is clear that the low frequency asymptote is a horizontal line it clearly says that it
clearly says that this is a horizontal line next we will consider tau omega tends to
infinity. So, as tau omega tends to infinity log amplitude ratio by K tends to minus 1 by2
log tau to the power 4 omega to the power 4 here we neglect all lower order terms, fine.
So, this is equal to basically minus two log tau omega as we considered tau omega tends
to infinity we obtained log amplitude ratio by K tends to minus half log tau to the power

4 omega to the power four here all low order. Terms have been neglected. So, it is than
becomes minus two log tau omega. So, this is a straight line with slope minus two. So,
this is a straight line with slope minus 2, so the low frequency.
(Refer Slide time: 29:20)

Asymptote is a horizontal line and the high frequency asymptote is a straight line with
slope minus 2 for the first order system we obtained horizontal line as low frequency
asymptote, but high frequency asymptote is a straight line with slope minus 1.
(Refer Slide time: 30:18)

And for second order system the slope is minus two we now we can produce the plot of
magnitude ratio verses tau omega before that we need to determined the point of
intersection fine. So, this is say third step. So, we got basically log amplitude ratio by K
which is approximately equal to log 1 this equation we obtain at low frequencies and for
high frequencies. We obtained minus two log tau omega. So, equating these two
equations we obtained again tau omega equals one again we obtain tau omega equals
one, but here we cannot determine the exact values of magnitude ratio because that is
also function of zeta here the exact value at corner frequency we can obtain ally knowing
the value of zeta. So, here the magnitude ratio
(Refer Slide time: 32:19)

Can be obtain by knowing the value of zeta and this should be deflected in the prode plot
fine next we will construct the prode plot and first we consider magnitude ratio verses
tau omega say this is zero point one than one and this is ten zero point zero one the
starting value of magnitude ratio next one is one and this is a 100. So, act low
frequencies we obtained a horizontal line fine and at high frequencies we obtained a
straight line with slope minus two it was mention that the magnitude ratio at the point of
intersection depends on the value of zeta. So, accordingly we will considered the three
different cases whirring the zeta values.
This type of response we obtained if zeta is say zero point six which is quit lace this one
we obtained when zeta is say zero point nine and this plot we obtained zeta is one fine

considering three different values of zeta we obtained three different plots as magnitude
ratio verses tau omega.
(Refer Slide time: 35:40)

So, these are the three plots this tope one corresponds to zeta equals 0.6 next 1
corresponds to zeta equals 0.9 and this bottom was to 1 corresponds to that equals 1 next
we will discuss the asymptotic considerations in term of phase angle next we will discuss
the asymptotic consideration in term of phase angle when tau omega tends to 0 we can
write phi equals tan inverse minus 2 zeta tau omega can we write this.
This we can write when tau omega becomes much less than unity when tau omega much
less than unity fine now if we consider tau omega tends to 0 now if we consider tau
omega tens to 0 this phi approaches 0 agree if tau omega is much less than 1 we can
write the phi expression like this tan inverse minus 2 zeta tau omega now if we consider
tau omega tends to 0 the phi approaches 0. Secondly, we consider tau omega tends to
infinity. So, in this case we can write the phi expiration as 10 inverse minus 2 zeta tau
omega divided by 1 minus tau squire omega squire this is the original expiration, we can
write this as 10 inverse minus 2 zeta divided by 1 by tau omega minus tau omega agree.
So, this is equal to tan inverse two zeta by tau omega when we considered tau omega
tends to infinity agree. So, this one we can write in this form considering tau omega
tends to infinity. So, than phi tends to 10 inverse 2 zeta by tau omega which is equal to

tan inverse 0 agree this expirations we got now we are considering tau omega tends to
infinity.
So, we can write tan inverse 0 and this is equal to minus 180 degree this is considered as
minus 180 fine in the third case we will considered tau omega equals ,1 I mean the point
of intersection tau omega equals 1 now in this case re gall is the value of zeta phi
becomes tan inverse minus infinity which is minus 90, which is minus 90 tau omega
equals 1 corresponds to the corner frequency and regal less of the value of the zeta we
obtain phi equals tan inverse minus infinity; that means, this is minus 90 now you see
initially we consider tau omega tense to 0 and this is tau omega equals 1. So, this is the
lowest one this is the intermediate 1 and this is the highest 1 fine here the phi value is
maximum this is intermediate and this is minimum. So, there is a train in prode plot fine.
(Refer Slide time: 41:10)

Now, we will try to produce the plot between phi verses tau omega and this is phi 0.1
this is indicating one and this is ten and this is suppose minus 200 degree minus 180
degree minus 90 degree 0 degree this plot also considered the different values of zeta
This is one plot this is the second plot this is the third plot this top one we obtain for zeta
zero point six second one we obtain I mean intermediate 1 we obtain considering zeta 0.9
and this one we obtained considering zeta equals 1 fine. So, this is the prode plot of a
second order system the top one is plotted in using log log graph paper and the bottom
one is obtain considering semi log graph paper, fine.

(Refer Slide time: 44:29)

In the last example we will consider a feedback control system a feedback controller and
we will construct the prode plot. So, next we vest to discuss the construction of prode
plot for a p only controller the transfer function of a p only controller in K c I mean g c
equal K c. So, we can write amplitude ratio equals K c and the phase angle we can write
as phi equals 0. So, both are independent of omega. So, it is easy to develop the prode a
plot this is indicating omega and this is indicating amplitude ratio since K c is
independent of omega this should be a straight line agree in the bottom plot. We consider
phi verses omega and phi remains 0 all the time I mean that is also a horizontal line
having phi 0 degree. So, this is the prode plot of a p only controller.

(Refer Slide time: 46:25)

Next is the P i controller the P i controller has the transfer function and it can be written
as G c equals K c equals one plus one divided by tau i s replacing a s by j omega
replacing a s by j omega. We can write for the frequency response transfer function as K
c multiplied by one plus one divided by j tau i omega we can write this as k c one minus j
tau i omega agree. So, from this cartesian form it is straight forward to determine the
expression for amplitude ratio and phi. So, amplitude ratio we obtain as K c root over of
1 square plus 1 divided by tau square omega square fine
Similarly, we can write the expression for phi phi equals tan inverse minus 1 divided by
tau i omega this is the expression for phi now we can write this expression of amplitude
ratio by taking algorithm as log amplitude ratio by K c equals half log 1 plus one divided
by tau i omega whole square.
If we take the algorithm of this expression we can write log amplitude ratio by K c
equals half log 1 plus one divided by tau i square omega square fine in the next we
considered the asymptotic considerations, first we consider the amplitude ratio as omega
tends to 0 1 by tau i omega whole square because much much higher than 1 agree.
So, we can write as log amplitude ratio by K c tends to half log 1 by tau i square omega
square which is equal to minus log tau i omega agree if we considered omega tense to 0
we can say that one by tau i square omega is much much higher than one than this
expression eldes this. So, the low frequency asymptote is a straight line.

(Refer Slide time: 51:29)

So, this is a straight line with slope minus 1 fine. So, what will be phi than if we consider
omega tends to 0 phi becomes tan inverse minus infinity agree which is equal to minus
90 degree it is very straight forward, because we have this expression we are just
considering omega tense to 0. In the next step we will consider omega tense to infinity
than one by tau i square omega square approach a 0. Accordingly we can write log
amplitude ratio by K c tense to log 1. So, at high the frequency into the horizontal line
high frequency asymptote is a horizontal line and how much is phi if omega tense to
infinity phi becomes 0 degree next one is the determination of corner frequency we
obtained the corner frequency, when tau i omega n equals 1.0 fine accordingly phi
becomes minus tan inverse 1 which is minus 45 degree and amplitude ratio by K c
becomes root 2 which is 1.4 1 4, fine. So, these are the asymptote consideration an next
we will make the plot amplitude ratio by K c and next one is phi verses tau i omega.

(Refer Slide time: 54:03)

So, when omega tense to 0 we obtain straight line with minus 1 somewhat like this and
this is indicating the corner frequency and when omega tends to infinity this is a straight
line like this fine. So, we get this type of prode a plot similarly here, we obtained minus
90 we obtain like this plot for phi verses tau i omega here this this should be like this is
45 I am re join it tau i omega this is suppose 1 this is minus 90 degree
So, this is minus 45 and this is 0 degree. So, it will go like this fine. So, this is indicating
the corner frequency at which phi is minus 45 and at low frequency this is minus 90 and
at high frequency at high frequency this is 0. So, you just do not considered this 1 fine.
So, this is the prode plot of a P i controller prode plot of a P i controller thank you any
question minus 180. You know in the prode plot if there is no trained that cannot be
used.

Process Control and Instrumentation


Prof. A. K. Jana
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture - 26
Feedback Control Schemes (Contd.)
So, presently we are discussing the frequency response analysis. Under frequency
response analysis, today we will cover the construction of polar plot, previously we
started the construction of polar plot.
(Refer Slide Time: 01:12)

Now, we will discuss the construction of polar plot, this plot is also called Nyquist plot,
because this plot is used for the stability analysis using Nyquist criteria. So, the polar plot
is also called Nyquist plot. Now in Nyquist plot is an alternative way of polar plot to
represent the frequency response characteristics of dynamic systems. Nyquist plot is an
alternative way of prode plot to represent the frequency response characteristics to
represent the frequency response characteristics of a dynamic system.
So, this is basically a alternative to prode plot and this can be used for stability analysis
later. This plot is construction using transfer function in the complex plane, fine. Now,
we will take the transfer function in general form I mean you can take the transfer
function as G s now replacing S by G omega you can write as G j omega which is

frequency response transfer function, and this you can write again in Cartesian form as
real G j omega plus j imaginary part, fine.
We can write the transfer function like this now you will produce 1 plot in the complex
plane. So, this is real axis this is imaginary axis, and this is a unit circle which as the
radius of 1 this is a unit circle fine; that means, this point corresponds to 10 this is the
origin, and this point correspond to minus 10 this 1 minus 10 this is 10, and this is the
origin. Now from the frequency response transfer function we can write.
(Refer Slide Time: 06:02)

Expression for amplitude ratio is root over of the real part whole square plus imaginary
part whole square fine amplitude ratio we know is represented by output amplitude
divided by the input amplitude I mean amplitude of the output divided by amplitude of
the input, and this transverve function G.
Which is represented as output divided by input agree here what is the amplitude of f S ,
because we can write this term divided by 1 I can write this term divided by 1 plus G
multiplied by 0; that means, amplitude of output is this amplitude of input is 1 agree. So,
we can say that here amplitude ratio is becoming output amplitude now. What will be th
fails it phi phi we can write as tan inverse imaginary part divided by the real part, fine. If
we consider this as complex number.

(Refer Slide Time: 08:35)

Then suppose this part OB represent the real part BA represent the imaginary part. So,
how much is the OA this is the amplitude ratio agree, if we consider G s is this 1 and
amplitude ratio is this 1 square root of real part square plus imaginary part square then
this OA represent the amplitude ratio.
(Refer Slide Time: 09:26)

So, I can write this amplitude ratio is basically, OA this is basically the magnitude of this
G j omega how much is argument this 1 phi.

(Refer slide Time: 09:47)

That means this angle is phi agree, and this OA is representing the amplitude ratio. So,
this is the polar plot of transverve function G s. So, this is the polar plot of transverve
function G s next we will discuss the cons traction of polar plot for first order system
second order system extra. So, in the next we will consider the first order system.
(Refer Slide Time: 10:51)

I mean the constraction of polar plot and we will consider first order system the
transverve function of first order system we can write as K divided by tau S plus 1, fine.
This the transverve function of a first order system we determine the expression for first

order system we determine the expression for amplitude ratio that is K divided by root
over of tau square omega plus 1 we determine the expression for faze angle phi that is
tan inverse minus tau omega this 2 expressions we have determine earlier.
Now, we will consider both the process gain and time constant are equal to 1, we will
consider both K, and tau are equal to 1, fine. Now bode plots sorry polar plots polar plot
starts at omega equal to 0 fine the polar plot starts at omega equal to 0, if we consider
omega equal to 0 what is amplitude ratio I mean how much is amplitude ratio amplitude
ratio is 1 how much is phi; phi is 0, fine. So, if we consider this point I mean omega
equals 0.
(Refer Slide Time: 13:34)

What will be the polar plot this is real axis this is imaginary axis. So, first we will draw
the unit circle this is the unit circle amplitude ratio 1 means this is basically 1 I mean this
point represent 1 agree when phi equals to phi is equals to 0. So, amplitude ratio 1, and
phi equal to 0 represent this point agree. So, we will write here and omega equal to 0 the
starting point we get this 1.

(Refer Slide Time: 14:49)

Next we will consider the end point the plot terminates at omega equals infinity plot.
Terminates an omega is equal to infinity if omega is equal to infinity how much is
amplitude ratio amplitude ratio is 0 how much is phi; phi is minus 90, fine. Because phi
minus tan inverse tau omega. So, can you locate end point.
End point is this 1 here phi is equal to minus 90 and amplitude ratio is 0. So, this is the
end point next we need to consider the intermediate points to draw this polar plots now
when omega is in between 0 and infinity how much will be amplitude ratio when we
consider omega 0 we got amplitude ratio 1 when we consider infinity we got amplitude
ratio 0. So, we can write like this agree when we consider omega 0 we obtain amplitude
ratio 1 when we consider omega infinity we obtain amplitude ratio 0.
So, we can write this accordingly similarly we can write for phi when omega equals 0
phi is 0 when omega is infinity phi is minus 90 degree. So, we can write this 0 greater
than phi greater than minus 90 if use intermediate points in this complex plane then we
will get the polar plot like this and this is the deduction this is starting point, and this is
the end point and we need take few intermediate points. Then we will get this type of
plot. So, this is the polar plot for the example first order system this is the polar plot for
the example first order system, fine.

(Refer Slide Time: 18:49)

Similarly, we will try to construct the polar plot for second order system in the next polar
plot for second order system we know the transverse function is represented in general
form as K divided by tau square S square plus 2 zeta tau S plus 1 this is the transverse
function of a second order system, fine. Previously we determine the expressions for
amplitude ratio and phi for second order system they are amplitude ratio equals K
divided by root over 1 minus tau square omega square whole square plus 2 zeta tau
omega whole square this expression we determine earlier.
Similarly, if we determine the expression for phi as tan inverse minus 2 zeta tau omega
divided by 1 minus 2 square omega square, fine. Now we will construct the polar plot
and for that we will consider K and tau both are equal to 1 for simplicity we are
considering the K and tau equal to 1 now the polar plot starts at omega equal to 0. So,
plot start at omega equal to 0 what is the corresponding amplitude ratio 1 how much is
phi; phi is 0, when omega equal to is equal to 0 amplitude ratio becomes 1 and phi
becomes 0.

(Refer Slide Time: 21:56)

Now, we will locate the starting point in the complex plane. This is the unit circle when
amplitude ratio is equal to 1 and phi is equal to 0 we get this point. So, when omega
equal to 0 the polar plot should start from this point.
(Refer Slide Time: 22:36)

Secondly if we will consider the we will try to find the end point this polar plot
terminates at omega is equal to infinity now if this the case then how much is the
amplitude ratio 0 how much is phi; phi is minus 180 pi, we can write as tan inverse 2

zeta y tau omega tan inverse 0 this we consider minus. So, what will be the end point in
this complex plane.
(Refer Slide Time: 23:46)

Here, but the polar plot should go through this quadrant next we need to generates some
intermediate data see here another parameter is involved that is zeta, fine. So, when zeta
is greater or equal to 1 amplitude ratio becomes less than or equals to 1 it is easy to test
that when we consider zeta is greater than or equal to 1 the amplitude ratio becomes less
than or equals to 1 what it indicates the polar plot will be within the unit circle amplitude
ratio less than 1 means the polar plot should be within the unit circle.
Similarly, if we consider zeta less than 1 it is easy to investigate the amplitude ratio
becomes greater than 1 when zeta is less than 1 amplitude ratio becomes greater than 1;
that means, this outside of the unit circle the polar plot goes outside of the unit circle. So,
this is an omega equals infinity this point an omega equals 0. So, when zeta greater than
1 we obtain this plot, because it is within the unit circle this we obtain an zeta greater
than 1 similarly when zeta less than 1 we obtain this plot I mean this polar plot is outside
of the unit circle, fine. So, it is clear that the polar plot depends on the value of zeta this
is the polar plot of the example. Second order system, fine. Next we will consider the
third order system.

(Refer Slide Time: 27:24)

Construction of polar plot for a third order system the transverse function of third order
system can be written as 1 divided by tau 1 S plus 1 tau 2 S plus 1 tau three S plus 1,
fine. This is the transverse function of a third order system. Now, it is easy to show that
when omega equal to 0 amplitude ratio is 1 and phi is 0 degree, similarly when omega
tends to infinity amplitude ratio tends to 0 and phi tends to minus 270 degree for the first
order system we obtain minus 90 for second order minus 180 for third order system
minus 270, fine. Similarly we can generate some intermediate data this is the unit circle
in complex plane starting point is this 1 and end point is this 1 and polar plot we obtain
like this. So, this is the polar plot of the example third order system, fine. This is the
polar plot of a third order system we can take another example like if pure delay system
how we can construct.

(Refer slide Time: 30:54)

Polar plot of a pure dead time the transverse function of this element pure dead time
element we can write as exponential minus t d S this is the transverse function of the
pure time delay system how much is amplitude ratio you can compare this with the
standard polar form that is amplitude ratio exponential of j phi compare this with this
how much is amplitude ratio 1 how much is phi minus t d omega fine minus t d omega.
So, it is obvious that amplitude ratio is independent of omega it is obvious that amplitude
ratio is independent of omega that is always 1. So, the polar plot is if circle of radius 1 it
clearly indicates that the polar plot should be a circle of radius 1. So, the polar plot is if
circle of radius 1 this is real axis this is imaginary axis amplitude ratio 1 means the polar
plot this is the polar plot we can say the polar plot is a circle of radius 1 as omega
increases what happens as omega increases this will start rotating it is obvious from this
expression if you keep on increasing the omega holus we will see this phi will changing
like this that means.
As omega increases this will start rotating now next question is what is the deduction of
that rotation? I mean we have to show the deduction of this polar plot also that will be
clock wise. So, deduction of rotation will be clock wise. So, deduction of rotation will be
clock wise clock wise rotation. So, this is the polar plot of pure dead time system. Now
what is the purpose of this frequency response analysis started. So, far what is the
purpose of this.

(Refer Slide Time: 35:15)

Frequency response analysis. So, far we have started under frequency response analysis
the determination of amplitude ratio and phi. So, we have started under frequency
response analysis the determination of amplitude ratio and phi. So, we have started the
cons traction of polar plot, but what is the purpose of doing all this what is the purpose of
doing all this the frequency response analysis is performed for two reasons; oner is for
stability analysis another one for controller tuning frequency response analysis is
performed for two reasons one is for stability analysis of the closed loop system.
Frequency response analysis is performed for tuning of control parameters and it is
interesting to note that for the stability analysis and for control tuning we will use
thisbode plot and polar plot. So, far we have just started construction those plots in the
next we will use those plots for stability analysis, and for control tuning. So, first we will
discuss thebode stability criteria, and we will use for that purpose thebode plot.

(Refer Slide Time: 38:05)

So, first we will discuss the bode stability criteria, fine. Now for this purpose will
consider 1 closed loop system, because it is mentioned that the bode stability criteria is
used for stability analysis of closed loop system. So, we will consider a closed loop
system and presently we are making that close look block diagram having p only
controller then considering G f equals 1 and G p equals exponential of minus 0.123 S
divided by 0.1 S plus 1 and considering G m equals 1, fine.
This is a close loop block diagram of a sharp o problem this close loop block diagram of
a sir loop proble, fine. Now you know that for constricting in the bode plot we need the
expression for amplitude ratio and we need the expression for phi for constricting in the
polar plot we need the expression for amplitude ratio and phi now for deriving the
expression for amplitude ratio and phi what we need we need the open loop transverse
function open loop transverve function we denote by G suffix o l which is nothing, but a
multiplication of all four tansverve function G c G p G f G m now all the tansverve
function are defined in this block diagram accordingly we obtain K c exponential of
minus 0.123 S divided by 0.1 S plus 1.
Now, we can write it like this K c divided by 0.1 S plus 1 and separately this exponential
term. So, this is the transverse function of first order system and this is a dead time part
in the next we will try to find the expression for faze angle phi. So, if we consider first
order system and dead time element then we can write as phi equals phi 1 plus phi 2 this

is representing the first system I mean 1 and this the second. So, how much this will be
for a first order system we know the phi is tan inverse minus tau omega. So, tan inverse
how much is the tau 0.1 omega, fine. And how much is phi for this dead time element
minus 0.123 omega. So, these in degrees this is in radians. So, we will multiply this with
180 degree divided by pi defiantly we write in the right hand side minus 180 degree you
can write minus phi also accordingly you write the left hand side presently we are
representing all in degrees fine can you calculate from this expression the holus of
omega we are considering phi equals minus 180 degree.
(Refer Slide Time: 44:15)

Now, you calculate from this the value of omega that we obtain 17 radian per minute the
value of omega at phi equals minus 180 degree is called cross over frequency the
frequency which is calculated at phi equals minus 180 degree that omega is called cross
over frequency and which is represented by omega suffix c o cross over frequency this
also called as critical frequency the frequency which is calculated at phi equals minus
180 degree is called cross over frequency or critical frequency.
Now phi if phi is always greater minus 180 degree, then the there is no chance of
instability if phi is always greater minus 180 degree then the there is no chance of
instability no chance when phi is less than minus 180 degree, then there is probability of
instability I mean that is in unstable system. Therefore, we can say that phi equals minus
180 degree that is critical value, fine. So, we got cross over frequency equal to 70 radian

per minute here the dead time is given in minute and time constant is also in minute what
how we can calculate the amplitude ratio?.
If we consider the first order system and dead time system; that means, there are 2
system. So, we can write the overall amplitude ratio equal to the multiplication of the 2
individual amplitude ratios for the first order system we know the amplitude ratio is
written as process gain divided by tau square omega square plus 1 this amplitude ratio
for a first order system and for the time dead system the amplitude ratio is 1, now we will
try to determine the value of this several amplitude ratio at cross over frequency tau is
0.1 cross over frequency we determine 17 accordingly we obtain this.
Now, we obtain amplitude ration 1 when K c equal to 2, I mean if we write this
amplitude ratio by K c we obtain 0.5 fine amplitude ratio is this calculating this
denominator we can write amplitude ratio divided by K c equals to 0.5 then we can say
that we obtain amplitude ratio equals to 1 when K c equal to 2, fine in this line we can
say that amplitude ratio is greater than 1 and K c is greater than 2 we can conclude from
this expression amplitude ratio becomes greater than 1 when K c is greater than 2
similarly we can write amplitude ratio less than 1 when K c is less than 2 fine can we
draw the bode plot.
(Refer Slide Time: 50:19)

Now, we will construct the polar plot based on this information this omega phi omega
amplitude ratio this is a log plot and this is a semi log plot this is 0 degree now phi equals

minus 180 degree we obtain the omega which is omega c o this is cross over frequency
the frequency at phi equals minus 180 is cross over frequency, and this we obtain 17
radian per minute correct. Now this is the plot when k c equal to 2 at cross over
frequency how much is amplitude ratio 1 definitely when K c equal to 2 agree see that
definition I mean we have derived cross over frequency as we consider phi equals minus
180 degree we obtain this now we obtain amplitude ratio 1 and K c is 2 at cross over
frequency we obtain amplitude ratio 1 when K c equal to 2 and omega equal to omega c
o.
Therefore, we have drawn this I mean this is the 1 at cross over frequency and K c is
equal to 2 and K c is greater than 2 this is indicating K c is greater than 2 how much will
be amplitude ratio greater than 1 it is clearly indicating that and K c is less than 2 how
much is amplitude ratio less than 1 fine. So, this is phi verses log omega plot at phi
equals minus 180 degree we obtain cross over frequency that is minus that is 17 radian
per minute amplitude ratio is equal to 1 and K c is equal to 2 at omega c o that is shown
here other 2 cases are also shown here they are K c is less than 2 and K c greater than 2 it
is obvious from this plot that and K c less than 2 and amplitude ratio less than 1 and
when K c greater than 2 amplitude ratio become becomes greater than 1.So, this is the
bode plot. So, in the next day we will discuss the stability criteria continuing this portion
thank you I hope presently what we have discussing that is most difficult part of process
control. So, do not sleep sitting in the first bench.

Process Control and Instrumentation


Prof. A. K. Jana
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture - 27
Feedback Control Schemes (Contd.)
So, in the last class we started the topic Bode Stability Criterion we will continue that
topic and we took one example also. So, today we will continue Bode Stability Criterion
with that example.
(Refer Slide Time: 01:11).

Bode Stability Criterion we considered ap only controller having transfer function kc.
We considered the 1st order process dead time system for the process and the transfer
function is exponential of minus 0.123S divided by 0.1S plus 1 for simplicity we
considered Gm Gf both equal to 1. First we need to write the open loop transfer function.
That we can write as kc divided by 0.1S plus 1 exponential of minus 0.123S this is first
order and this is dead time part. So, the expression for phi we can write as tan inverse
minus 0.1 omega minus 0.123 omega multiplied by 180 degree divided by phi this is the
expression for phi. If, we consider phi equals minus 180 degree then we obtain omega
equals 17 radian per minute here time is in minute. Now, this is basically cross over
frequency cross over frequency or critical frequency is the frequency which is calculated
at phi equals minus 180 degree.

In the next step we can write the expression for Amplitude Ratio that we can write as K c
divided by tau square omega square plus 1 here tau is 0.1 omega we are considering
cross over frequency whole square plus 1 then we obtain Amplitude Ratio divided by Kc
equal to 0.5. We obtain Amplitude Ratio divided by Kc equal to 0.5 now, we obtain
Amplitude Ratio 1 when Kc equal to 2. We obtain amplitude ratio equal to 1 when Kc
equal to 2.
(Refer Slide Time: 05:28)

So, we can say that Amplitude Ratio we can obtain greater than 1 considering Kc greater
than 2 similarly, we obtain amplitude ratio less than 1 when Kc less than 2. So, Kc
equals to 2 that is the critical value of Kc 2 is the critical value for Kc. I hope up to this
we discussed in the last class. So, today we will start with the open loop response next
we will find the open loop response of the example system. Now, how we can open the
loop? Previously we considered the closed loop. We want to open that than we need to
disconnect something and that is shown here. This is the block for first order plus dead
time system output is y bar this is the comparator this is set point and this is measuring
device negative this is positive.
Now, we are disconnecting the measurement signal from this comparator then we can
say that this is open loop. We use to know the open loop response by disconnecting the
measurement signal from this comparator in addition we will consider the followings: 1st
will consider Amplitude Ratio equal to 1 phi equal to minus 180 degree if, we consider

phi equal to minus 180 degree we obtain cross over frequency 17 radian per minute
agreed if we consider phi equal to minus 180 degree we obtain cross over frequency 17
radian per minute. Now, at this cross over frequency we obtain Kc equal to 2 at
Amplitude Ratio equals 1. So, this is our first consideration Amplitude Ratio equal to 1
and phi equal to minus 180 degree.
(Refer Slide Time: 09:42)

Secondly, set point changes in a Sinusoidal manner with omega equals 17 radian per
minute and input amplitude equals to 1. This is the second consideration set point
changes in a Sinusoidal manner with omega equals 17t and input amplitude equals 1 that
means, this is sin 17t.

(Refer Slide Time: 11:00)

So, in time domain I can write this y set point equals sin 17 t. Here, we can write y set p
equals sin 17t than how much is the error signal error? Error is also sin 17t because, we
have disconnected this measurement signal from the comparator. So, error signal is also
sin 17t which is equal to y set point and there written in time domain. Now, our first
consideration is Amplitude Ratio equals 1.
(Refer Slide Time: 11:56)

Amplitude ratio equal to 1 Amplitude Ratio is nothing, but output amplitude divided by
input amplitude again we consider in the second consideration input amplitude equals 1.

So, how much is output amplitude? Output amplitude becomes 1. So, if the input is
changed in Sinusoidal manner. What is the output? Output is a out sin omega t plus phi
this is the output. So, amplitude of this output ais1 omega is 17 and phi is minus 180.
We have considered in our first point phi is minus 180 that means, this is minus sin 17t
agree. So, if we introduce set point change in Sinusoidal manner the output is minus sin
17t this is also Sinusoidal. So, we can write Ym equals sin 17t this is the open loop
response find the process is subjected to a change in set point in Sinusoidal manner.
In the next we will discuss the close loop system response. So, we will considered 2
events at a time I mean we will consider at the same time two events they are no change
in set point that means, Y set point equal to 0 at the same time we consider another point
that is close the loop this is the second point.
(Refer Slide Time: 15:30)

So, what would be the corresponding block diagram then? Controller, is this one final
control element is final control element has transfer function of unity process transfer
function is exponential of minus 0.123S is divided by 0 point 1S plus 1 y bar Gm equal
to 1 comparator y set point equal to 0. This, we are simply connecting the measurement
signal with this comparator. So, this is the close loop system. Here, we are considering
no set point change and we are closing this. The output signal was minus sin 17t in the
previous block diagram the output was minus sin 17t and we are suddenly closing it and

at the same time we are not introducing any change in set point. So, this is negative this
is positive. So, how much will be error now? Sin17t.
So, you notice thsat the error remains in the open loop system error vise sin 17t and
when you are closing this loop and we are introducing no change in set point we are
getting same error signal that is sin 17t. So, error is sin 17t and we are getting output
minus sin17t What it indicates? Sustained oscillation I mean if we consider Kcequal to 2
then Amplitude Ratio is equal to 1.
When phi is equal to minus 180 degree if Kc is equal to 2 which is a critical Kc value
then Amplitude Ratio equal to 1 when phi equal to minus 180 degree and in this situation
we obtain sustained oscillation I mean oscillation with constant amplitude.
What about stability? If, we obtain sustained oscillation what about stability marginally
stable we can use this term marginal stability. Now, similarly we can conclude that if Kc
is greater than 2 then Amplitude Ratio greater than 1 when phi is minus 180 degree. So,
What type of response we can obtain in this condition? Oscillations with increasing
amplitude then, what about stability unstable similarly if Kc is less than 2 then
Amplitude Ratio less than 1 and phi is minus 180 degree if Kc is less than 2 then
Amplitude Ratio becomes less than 1.Whenphi is minus 180 degree.
So, in this condition we obtain oscillations with decreasing amplitude and this is stable.
Now, based on this observation we can write the Bode Stability Criteria that is a
feedback control system is unstable if the Amplitude Ratio of the corresponding open
loop transfer function is greater than 1 at cross over frequency.

(Refer Slide Time: 22:19)

Bode Stability Criteria is stated as a feedback control system is unstable if the amplitude
ratio of the corresponding open loop transfer function is larger than 1 at the cross over
frequency this is the Bode Stability Criteria . A feedback control system is unstable if the
Amplitude Ratio of the open loop transfer function is larger than 1 at the cross over
frequency. Next we will discuss the Stability Analysis based on this Bode Stability
Criteria we will take a number of examples and we will apply this Bode Stability
Criteria.
(Refer Slide Time: 24:10)

So, our topic is Stability Analysis based on Bode Criteria. Firstt we will consider a First
order process. For a First order process we can consider the transfer function as Kp
divided by tau pS plus 1 this is the transfer function of a First order system. We consider
p only controller. So, Gc equals Kc this is the transfer function of p only controller for
simplicity we consider Gm and Gf both equal to 1. Now, at the first step we need to find
the open loop transfer function the open loop transfer function for this example system
we can write as Gc Gp Gf Gm and that is equal to Kc Kp divided by tau pS plus 1 and
we can write this as K divided by tau pS plus 1 here K is Kc Kp.
Now, what about the phi if we consider omega equal to 0? How much will be phi? Phi
will be 0 this is a First order system I mean this is the transfer of a First order system
when omega is infinity, How much is phi for a first order system? Minus 90 when omega
is 0 phi is 0 when omega is infinity phi is minus 90. For a first order system phi remains
in between 0 and minus 90. So, what it indicates? Phi is always greater than minus 180
agreed phi is always greater than minus 180 degree that means, there is no existence of
cross over frequency.
So, what about stability stable if there is no cross over frequency then the system always
remains as stable. So, for a First order system the process always remains at stable state
irrespective of the Kc value only thing of you that if you considered large K c than
oscillations will be more, but there is no instability problem.
(Refer Slide Time: 29:04)

So, we can say that first we need to check weather cross over frequency exists or not. If,
the answer is no then the system is stable if the cross over frequency does not exists then
the system is stable system. So, you can use any Kc value at least there will not be any
instability problem, but as I mentioned with the increase of Kc value now oscillation also
the response also becomes more oscillatory now, if the answer is yes cross over
frequency exists then instability problem is there. So, how we can make it stables this
unstable system? We can make it stable by tuning the controller we can make this
unstable systems stable by tuning the controller. So, if there is any instability problem
Tune Kc. So, initially we considered the first order process in the next we will consider a
First order process dead time system. Next we will discuss a First order with dead time
process transfer function of a First order process a dead time system we can write as Kp
exponential minus 0.15S divided by tau pS plus 1 here dead time is 0.15.
So, this is a transfer function of a First order plus dead time system. Here, we are
considering Gm equal to Km controller is p only controller. So, Gc equals Kc and for the
final control element Gf is equal to suppose Kf. So, these are the individual transfer
function for the examples system.
(Refer Slide Time: 32:52)

Now, we need to write the open loop transfer function. What will be the open loop
transfer function? Kc Kp Kf Km exponential minus 0.15S divided by tau pS plus 1 this is
the open loop transfer function. So, this transfer function we can rewrite as K

exponential of minus 0.15S divided by tau pS plus 1 where K is Kc Kp Kf Km. So, this
is overall the transfer function of a first order plus dead time system. Now, what will be
the expression for phi? Ppi will be tan inverse minus tau p omega this is for the dead
time free part and for dead time part the phi will be minus 0.15 omega this is for dead
time free part and this is for dead time part.
Now, we can multiply with 180 degree divided by phi. Now when omega tense to 0 what
about phi if omega tense to 0 then phi tense to 0 degree when omega tense to infinity
What about phi? Phi will be minus infinity agree if omega tense to 0 then we can write
phi tense to 0 degree and when omega tense to infinity phi tense to minus infinity. So,
what about cross over frequency? Cross over frequency exists now at large value of Kc
Amplitude Ratio may become greater than 1 at cross over frequency at large Kc value
Amplitude Ratio may become greater than 1 at cross over frequency it clearly indicates
there is an instability problem.
So, first point is the cross over frequency exists and at large Kc value the Amplitude
Ratio may become greater than 1 at cross over frequency that means, the there is some
instability problem. So, overall we can say that dead time is a principle source of this
stabilizing effect in chemical process control systems we can conclude based on this
example that dead time is the principle source of this, stabilizing effects in chemical
process control systems.
Now, when we started the Coherence technique it was mentioned that the response of
most of the chemical processes can be approximated by the response of First order plus
dead time system when we started the Coherence technique it was mentioned that the
response of most of the chemical process is can be approximated by the response of First
order plus dead time that means, almost all chemical processes have instability problem
due to the inclusion of dead time part. So, we can say that almost all chemical processes
include the dead time part.

(Refer Slide Time: 38:17)

That means almost all chemical processes have instability problem. Now as mentioned
before to make it stabilize we need to tune the controller. So, controller tuning becomes a
crucial task. Next we will consider a third order system and will apply the Bode Stability
Criteria. Higher order systems here before Third order system we will consider second
order system will consider next second order system then third order system. So, the
transfer function for a second order system we can write as Kp divided by tau pS plus 1
now this is the transfer function of a first order system.
Now, we are considering Gf equal to Kf we are considering Gc equal to Kc and Gm
equal to Km divided by tau mS plus 1. These are four individual transfer function now,
overall transfer function we can write as K divided by tau pS plus 1 tau mS plus 1 can we
write this? Overall transfer function becomes K divided by tau pS plus 1 multiplied by
tau mS plus 1 this is the second order system I mean this is the transfer function of a
Second order system and obviously, here K equal to Kp Kc Kf Km.
Now, if we consider omega equal to 0. How much is phi for the Second order system? 0
degree we are not writing the expression for phi because, we discuss this several times
when omega is equal to 0 for the Second order system phi becomes 0 degree when
omega tense to infinity phi tense to minus 180 degree for the second order system the
phi varies from 0 to minus 180 degree.

So, what about the existence of cross over frequency there is no existence of cross over
frequency at finite value of omega that means, there is no instability problem. So, see
First order system without dead time no instability problem Second order system without
dead time no instability problem next we will consider the third order system.
(Refer Slide Time: 42:42)

So, next we will consider the third order system. Suppose the transfer function of the
process is Kp divided by tau pS plus 1 transfer function for the final control element is
considered as Kp divided by tau fS plus 1 controller is p only controller and the transfer
function for the measuring device is considered as Km divided by tau mS plus1. These
are the four individual transfer function so obviously, the open loop transfer function
becomes Third order transfer function. I mean we can write it as K divided by tau pS
plus 1 tau mS plus 1 tau fS plus 1 this is the open loop transfer function for the example
system.
Now, when omega is equal to 0 phi becomes 0 degree and when omega tense to infinity
phi tense to minus 270 degree. So, for a Third order system the phi where is in between 0
and minus 270. So, what about the cross over frequency cross over frequency exists. So,
there is instability problem. So, we can note 1 point that up to Second order system if
there is no dead time the systems do not have any instability problem First order system
without dead time no instability problem Second order system without dead time no
instability problem Third order system without dead time instability problem is there. So,

up to Second order there is no instability problem if there is no dead time, but, if dead
time is there from First order system itself stability instability problem arises.
Next we will discuss another topic under this frequency response analysis that is Gain
margin and Phase margin.
(Refer Slide Time: 46:04)

This gain margin and phase margin of the typical performance specifications associated
with the frequency response analysis. Gain margin and Phase margin are the
performance specifications associated with the frequency response analysis. So, for
discussing this Gain margin and Phase margin it is better to consider an example and we
will continue the example which we considered earlier having the open loop transfer
function equals Kc exponential minus 0.123S divided by 0.1S plus 1.
To discuss the Gain margin and Phase margin concept we considered this example
where Gf and Gm both are equal to 1 and the transfer function for the process we
consider exponential minus 0.123S divided by 0.1S plus 1 and Gc equals Kc recall the
fact that if Kc equal to 1 if Kc equals 2 then Amplitude Ratio becomes 1. When phi
equals minus 180 degree. We concluded these for this example are here if Kc is equal 2
than Amplitude Ratio becomes 1 when phi equals minus 180 degree and this corresponds
to Marginal stability.

(Refer Slide Time: 49:15)

Now, at phi equals minus 180 degree we know that at phi equals minus 180 degree we
obtain the cross over frequency. So, we will try to show it in the bodeplot this is omega
this is phi have the plot is say like this. This is suppose 0 degree and this is suppose
minus 180 degree. So, at minus phi equals minus 180 degree we obtain cross over
frequency agree.
So, the frequency corresponding to this phi equals minus 180 degree is cross over
frequency which is shown in this figure. Now, above this we will draw Amplitude Ratio
verses omega. Suppose this A B curve we obtain for Kc equals 1 this A B curve I mean
this is the Amplitude Ratio verses omega plot we obtain considering K c equals one that
means, the Kc value we have taken lower than the critical Kc value now at this cross
over frequency, How much is this value? At cross over frequency Amplitude Ratio is
suppose m agree and critical value of Kc is 2 suppose this is the Amplitude Ratio verses
omega plot when Kc equal to 2.
So, at cross over frequency how much is this Amplitude Ratio 1. So, when omega equals
omega co we obtain Amplitude Ratio equal to M which is lower than 1 agree. If, we
obtain 1 than marginal stability if this is lower than 1 that means, stable system. So,
when omega equals omega co I mean at cross over frequency if Amplitude Ratio is less
than 1 than that is the indication of stability.

So, we can say that the response is stable response when Kc equal to 1 because, at cross
over frequency Amplitude Ratio is less than 1. I am repeating again the steps this is phi
verses omega plot for this example system this now cross over frequency is the
frequency which is obtained at phi equals minus 180.
So, we consider phi equals minus 180 corresponding omega is cross over frequency now
for the example system we know the critical value of Kc is 2 here, we first consider Kc
equals 1. I mean it indicates stability. Now, A B is the amplitude verses omega plot when
Kc equals 1 at cross over frequency it is clear from this bode plot that Amplitude Ratio is
M then, we can make another amplitude verses omega plot at the critical value of Kc that
is 2 and at cross over frequency.
The corresponding Amplitude Ratio is 1. So, it says that when we consider Kc equals 1
Amplitude Ratio less than 1 at cross over frequency now this difference is the Gain
margin you see this is nothing, but a safety factor this difference is called the Gain
margin and this is nothing, but a safety factor in the next class we will discuss the Phase
margin.

Process Control and Instrumentation


Prof. A.K. Jana
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture - 28
Feedback Control Schemes (Contd.)
In the last class we discussed the topic gain margin and phase margin, we will continue
that topic.
(Refer Slide Time: 00:58)

Gain Margin and phase Margin. And we considered the example having open loop
transfer function, equal to Kc exponential minus 0.123 is divided by 0.1s plus.. Now,
when K c if equal to 2 we obtain marginal stability, this is the marginal stability
condition. We developed in the last class the bode plot like this. This is omega this is phi
and this is a semi log graph paper. Suppose this is the phi verses omega plot in semi log
graph paper. Now, cross over frequency we obtain considering phi equals minus 180
degree, So, this is cross over frequency.
Now, above this we consider amplitude ratio verses omega and this plot is produced in a
log graph paper. Suppose this is amplitude ratio verses omega plot when Kc equal to 1.
So, at cross over frequency the amplitude ratio we obtain suppose that is M. Now, as
mentioned the marginal stability corresponds to Kc equal to 2. Suppose this is a
amplitude ratio verses omega plot, when Kc equal to 2.

That means at cross over frequency, we obtain the amplitude ratio equal to 1.So, this is
for Kcc equals 1 so, we can say that we obtain stable response and Kc equal to 1 because
at cross over frequency amplitude ratio is less than 1. So, we obtain stable response when
Kcc equal to 1. Because, amplitude ratio is less than 1 at cross over frequency. Now, this
difference is called as gain margin. So, it is obvious that it is a safety factor. Now, How
can we represent the gain margin mathematically? Gain margin we can represent as 1
divided by M, difference is log 1 minus log M. So, we can write gain margin as 1 divided
by M..
Now, when amplitude ratio equals 1 suppose this is I am just extending this, like this
when amplitude ratio equals 1. How much is the corresponding phi? See we are
considering Kcc 1. Previously also we considered Kcc 1 and we obtained amplitude ratio
M. Similarly here, we are considering Kcc equals 1. And the corresponding value of phi
is this, quantity suppose this is, phi prime this 1. So, when amplitude ratio equals 1. The
corresponding phi, we are indicating by phi prime agree. And this phi prime is greater
than you see minus 180 degree agree.
So, we can say that we obtain stable response when Kcc equals 1 because of the nonexistence of cross over frequency. We obtain stable response when Kcc equals 1 because
of the non-existence of cross over frequency. See this observation is based on this plot,
and the previous observation was based on this plot. And this difference is called phase
margin. You see this, phi prime is higher than 180. So, similarly we can say that this
phase margin is also a safety factor, and this phase margin is represented as 180 minus,
the absolute value of phi prime. The phase margin we can represent or we can calculate
by this correlation. So, these are 2 safety factors and based on the value of gain margin
and phase margin we can tune the control parameters.

(Refer Slide Time: 09:43)

Now, next we will note down few points. So, Gain Margin is a measure of how for the
close loop process is from the brink of instability. This is also true, for phase margin. So,
they are the measure of how for the closed loop system is from the brink of instability.
Now, the higher the gain margin implies the higher the safety factor. Now, question is.
How we can select the values of gain margin and phase margin? First we will discuss the
selection of gain margin, then we will discuss the selection of phase margin. The
heuristics is that, when the process parameter values are poorly knownFirst we have to
know the process parameters.
Process parameters mean process gain, time constant, and dead time etcetera. These are
the process parameters, so, process gain is 1 parameter. Then time constant then dead
time and so on. These are the process parameters, when these parameters are poorly
known imprecisely known the gain margin is taken in between 1.7 to 3.This is the
heuristics. When there well known, they are known quite accurately in that case the gain
margin is taken in between 1.4 and 1.7. These are the safety limits. And for phase margin
it is usually taken as a greater than 30 degree phase margin is usually taken larger than
30 degree. So, these are the heuristics for selecting the gain margin and phase margin
values.

(Refer Slide Time: 14:08)

In the next you will solve one problem, based on this phase margin and gain margin
concept. The open loop transfer function is given has 0 point Kc exponential of minus
1.74 s, divided by 5 s plus 1, 10 s plus 1, 15 s plus 1. So, this is basically the transfer
function of a third order plus dead time system. And only P controller is employed for
the system. So, this is the open loop transfer function of the example system. Now, we
want to tune the controller p only controller, tune P-only controller. Taking gain margin
equal to 1.7 tune the P-only controller I mean find the Kc value considering gain margin
equal to 1.7. Now, the open loop transfer function we can write as 1 divided by 5 s plus
1, 1 divided by 10 s plus 1, 1 divided by 15 s plus 1.
So, first order systems connected in series next part is the dead time part. We can rewrite
the open loop transfer function in this form. Now, in the first step we need to calculate
the cross over frequency, for finding the cross over frequency we need to write the
expression for phi and we need to consider phi equals minus 180 degree.
So, what will be the expression for phi tan inverse minus tau p omega is the expression
for phi for First order system. So, for this First order system tan inverse minus 5 omega
for the second First order system tan inverse minus 10 omega, and for this First order
system tan inverse minus 15 omega, and for the dead time part minus 1.74 omega
multiplied by 180 divided by phi. And we can write this is equal to, minus 180 degree

agree. See when we write phi equals minus 180 degree then, that omega is crossed over
frequency
So, we can include the suffix co omega co. Here also omega co everywhere omega co.
Now, solving this, we obtain the cross over frequency 0.16 radian per minute. Here, time
is in minute solving this equation we obtain cross over frequency 0.16 radian per minute.
Now, we need to write in the next the expression for amplitude ratio. So, for the First
order system I mean for this system for this transfer function, what would be the
amplitude ratio 5 omega whole square plus 1. Now, one thing we missed in this open
loop transfer function here that is 0.8 Kc we missed this numerator term 0.8 Kc. So, we
can write here that term 0.8 Kc. For the next transfer function 10 omega whole square
plus 1 for Third 1 we can write 1 divided by 15 omega whole square plus 1.
(Refer Slide Time: 20:41)

Now, we will substitute omega equals omega co then, we can write the expression for
amplitude ratio as 0.8 Kc divided by 5 into 0.16 whole square plus 1, 10 multiplied by
0.16 whole square plus 1, 15 multiplied by 0.16 whole square plus 1. This is equal to 0.8
Kc divided by the first term yields 1.28 next term gives 1.89 and third one gives 2.6
which is equal to 0.127 Kc. So, at cross over frequency we obtain amplitude ratio this
much, suppose this is equal to M now gain margin is given 1.7 this is given. In the
problem gain margin of 1.7 is given. Now, how much is gain margin 1 by M this is gain

margin; that means, we can write 1 divided by 0.127 Kc equal to 1.7. Then how much is
Kc, Kc becomes 4.62.
So, for the example system if, we consider k c equal to 4.62 then amplitude ratio at cross
over frequency is lower than 1 . So, this is a answer for the 1st part. Now, in this example
see we calculated this Kc value considering the dead time 1.74 suppose, that value is
wrong the actual dead time value is suppose 2.4 minute.
(Refer Slide Time: 24:01)

True value of dead time is 2.42 minute instead of 1.74 minute. What about stability if,
we use Kc equals 4.69. Now, question is what about, stability if we use Kc equals 4.62
and original td is 2.42 minute. This is a second question we found Kc value 4.62 based
on the wrong dead time value. Now, question is, What about stability if we use Kc equals
4.62 and the original dead time is 2.42 minute? Now, in this situation 1st we will found
the cross over frequency, considering this dead time.
So, again we need to write the phi expression tan inverse minus 5 omega C naught, plus
tan inverse minus 10 omega C naught plus tan inverse minus 15 omega C naught minus
of 2.42 omega C naught multiplied by 180 divided by phi, and this is equal to minus 180
degree. In this phi expression we have changed only this td value. Solving this we obtain
cross over frequency 0.15 radian per minute. Next we will calculate the gain margin our
aim is to check the gain margin value. So, further we need to determine the amplitude
ratio amplitude ratio expression we know 0.8 Kc divided by root over of 5 omega C

naught whole square plus 1, 10 omega C naught whole square plus 1, 15 omega C naught
whole square plus 1 f 0.8 Kc we determined 4.62 and we are still using that, 4.62 value
divided by 5 multiplied by 0.15 square plus 1, 10 multiplied by 0.15 whole square plus 1.
Next is 15, 0.15 whole square plus 1, and we obtain this is equal to 0.666 this is the
amplitude ratio value considering td equals 2.42.
Then how much is gain margin? 1 divided by amplitude ratio I mean, 1 divided by M
0.666 and this is equal to 0.5. What about stability? This is stable, because when M is
less than 1. We say that the system is stable. So, gain margin is basically 1 by M. So, if 1
by M is greater than 1 than, the system is stable when M is less than 1 definitely gain
margin should be greater than 1. So, this is the condition for stability if, gain margin is
greater then 1 than we can say that the system is stable. So, the problem we have solved
based on the gain margin value. We will continue this problem considering the phase
margin of 26 degree.
(Refer Slide Time: 30:35)

So, next question is, tune the P-only controller, when the phase margin is 26 degree find
the Kc value when this phase margin is 26 degree. Now, we know the expression for
phase margin. Phase margin we express as 180 minus phi prime this is the expiration for
phase margin 180 minus phi prime 180 minus phi prime. So, what is the expiration for
phi? That we will write here tan inverse phi omega, absolute term we are writing I mean
this is absolute value no, plus tan inverse 10 omega plus tan inverse 15 omega plus 1.74

omega 180 divided by phi equal to 26. And this we have return based on this transfer
function Gol equals 0.8 Kc exponential minus 1.74 s divided by 5 s plus 1, 10 s plus 1,
15 s plus 1.
We are solving we are solving this problem considering this open loop transfer function I
mean, using this open loop transfer function fine the value of Kc when phase margin is
taken as 26 degree. Now, if we solved this, tan we obtain omega as 0.12 radian per
minute, If, we considered this part equal to 26 degree and if, we solved then we have
obtained omega equals 0.12 radian per minute. Now, amplitude ratio equals 1 basically
phi prim is the phi value at amplitude ratio is equals 1.
So, we will considered this here now what is the expiration for amplitude ratio, 0.8 Kc
divided by root over of 5 omega whole squire plus 1 10 omega whole squire plus 1, 15
omega whole squire plus 1 and this is equal to 1 this is the expiration for amplitude ratio
0 point 8 Kc divided by 5 omega hole squire plus 1 root over of tan omega whole square
plus 1 root over of 15 omega whole squire plus 1 and that is equal to 1.
(Refer Slide Time: 35:16)

Now, if we solve this, we obtain Kc equal to 4.69 here, only if Kc is unknown if, we
substitute omega equal to 0.12. So, finally, we obtain k c equal to 4.69. So, previously
we obtain Kc equal to 4.62, considering gain margin 1.7. Now, here were getting Kc
equal to 4.69 considering phase margin 1.26 degree. Now next question is the dead time
we have taken wrongly. So, the original dead time is the true value of dead time d is

suppose 6 minute, instead of 1.74 minute. Now question is. What about stability? If, we
use Kc equal to 4.69 when the actual dead time is 6 minute. .
First we need to calculate phi prime, expiration is 10 inverse minus phi omega tan
inverse minus 10 omega than 10 inverse minus 15 omega, dead time is minus 6. So,
minus 6 omega 180 degree divided by phi. This is the expiration for phi, and considering
omega equals 0.1 to radian per minute. We obtain this as 183.35 substituting omega
equals 0.12 radian per minute we obtain the phi prime as minus 183.35. What about
phase margin then? 180 degree minus absolute value of phi prime so, 180 minus 183.35
minus 3.35 degree which is less than 0z. So, what about stability? Unstable. So, this is
unstable problem I mean unstable system. If the dead time is 6 minute and if, we use Kc
equals 4.69 than the system becomes unstable fine. So, this is the problem related to this,
gain margin and phase margin. So, first we if constructed the bode plot then, by the use
of the bode plot we discussed the bode stability criteria under which we determined the
control parameter values. Similarly we discussed the construction of Nyquist plot are
polar plot.
(Refer Slide Time: 40:14)

So, now we will use that plot for stability analysis. So, next topic is Nyquist stability
criteria. So, we discussed a number of examples, under this frequency response analysis.
And we observe that, with the increase of omega both amplitude ratio and phi decrease.
So, far we have discussed the frequency response analysis taking a number of examples,

and it is observed that with the increase of omega both amplitude ratio and phi decrease,
this is our observation amplitude ratio and phi decrease monotonically with the increase
of omega. If, this is the situation then, it is recommended to use the bode stability criteria
If, this is the situation the bode stability criteria leads to rigorous conclusions. But, there
are few systems for which amplitude ratio and phi do not decrease continuously with the
increase of omega. There are very few systems in chemical engineering for which the
amplitude ratio and phi do not decrease monotonically with the increase of omega.
(Refer Slide Time: 43:29)

And the plot looks somewhat like this. This is suppose amplitude ratio verses omega
plot, in log graph paper here, suppose the value is 1. So, it goes like this , similarly the
phi verses omega plot not the corresponding plot I, want to say for few systems the phi
verses omega plot is somewhat like this. So, it is very clear that the amplitude ratio and
phi do not decrease monotonically with the increase of omega. And in this situation it is
not recommended to use bode stability criteria. And it is recommended to use Nyquist
stability criteria which to be discussed in the next. In this situation the bode plot bode
criteria leads to erroneous conclusions and it recommended to use the bode Nyquist
stability criterion.

(Refer Slide Time: 46:27)

This criteria was proposed by Nyquist in 1932, and which has been published in journal
namely Bell Systems Technical Journal. The Nyquist stability criteria was proposed in
1932 the title of the work was Regeneration theory, and this work was published in the
journal namely Bell System Technical Journal no theorem, no proof of the theorem will
be discussed, in this course and you are strongly recommended to go through this paper.
If, you are interested to know more about the Nyquist stability criteria.
So, what is that criteria? Nyquist stability criteria we, can write has a closed loop control
system is unstable. If the open loop Nyquist plot of the feedback system, encircles the
point minus 10 as the frequency omega takes any value, in between minus infinity and
infinity.

(Refer Slide Time: 50:01)

So, previously we construct the polar plot in the complex plane. So, this is the complex
plane first we draw 1, unit circle. this, is a unit circle. So, this point represent suppose
this o is the origin, this is point A and this is point B, A is basically 10 and B is minus 10.
Now, if, the Nyquist plot Do not encircle this, then the system is said to be stable it is
somewhat like this, suppose for a particular system we are getting say for a Third order
system we are getting this polar plot. And this plot we obtain, varying suppose omega 2
from 0 to infinity.
Another plot we are getting suppose, this 1.So, this plot we can obtain varying omega
minus infinity to 0. And this we obtain varying omega from 0 to infinity. Now, you see
this Nyquist plot do not encircle this minus 10. So, we can say that the example system is
stabled.

Process Control and Instrumentation


Prof. A. K. Jana
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture - 29
Feedback Control Schemes (Contd.)
So in the last class we discussed Nyquist Stability Criteria. So, today we will continue
that topic with an example.
(Refer Slide Time: 01:02)

So, today we will continue the Nyquist Stability Criteria with an example. Now, the close
look transfer function is given as 0.8Kc divided by 5s plus 1 multiplied by 10s plus 1
multiplied by 15s plus 1. The open look transfer function is given in this form open loop
transfer function here, Kc is the proportional gain controller gain now, we will perform
the stability analysis by developing the polar plot with varying omega from minus
infinity to infinity.
So, in first step we will replace s by j omega. P Accordingly we get the frequency
response transfer function as GOL equals 0.8 Kc divided by 5 j 5 omega plus 1
multiplied by 10j omega plus 1 multiplied by 15j omega plus 1 this is the frequency
response transfer function.
Now, we can write the expression for Amplitude Ratio and phase angle 5 5. So, we can
say that this is the I mean 3 First order systems are connected in series this indicates that

the 3 First order systems are connected in series. So, it is straight forward to write the
expression for Amplitude Ratio 0.8Kc divided by 1 plus 5 5 omega square multiplied by
1 plus 10 omega square multiplied by 1 plus 15 omega square. T his is the expression for
Amplitude Ratio similarly we can write the equation for phase angle 5 that is tan inverse
minus 5 omega plus tan inverse minus 10 omega plus tan inverse minus 15omega these
are the 2 expressions 1 is for Amplitude Ratio and another is for 5.
(Refer Slide Time: 05:48)

Next we will go to the 2nd step. In this vary omega from 0 to infinity I have mentioned
that we will develop the polar plot by varing omega minus infinity to infinity. So, that
range we will divide into 2 ranges 1 is from minus infinity to 0 another 1 is from 0 to
infinity. So, in step 2 we will vary omega from 0 to infinity under this we can say that the
polar plot starts at omega equals to 0.
Now, at omega equals 0. What is the corresponding Amplitude Ratio that we can get
from that expression which we have written when omega equals 0 then Amplitude Ratio
becomes 0.8 Kc. Similarly if we can determine the values of phase angle. How much is
the phase angle when omega equals 0? 0.
So, when omega equals 0 we can obtain Amplitude Ratio 0.8 Kc and 5 0 degree now,
plot polar plot terminates at omega equals infinity when omega approaches infinity the
polar plot ends. So, when omega equals infinity we obtain the corresponding Amplitude

Ratio as 0. Similarly if we obtain 5 minus 270 degree when omega approaches infinity
we obtain Amplitude Ratio is 0 and 5 minus 270 .
So, an Amplitude Ratio is 0.8 Kc and 5 0 the starting point on real axis you see 5 is 0. So
the starting point will be on real axis similarly when Amplitude Ratio is 0 and 5 minus
270 the end point will be on in origin this indicates the origin agreed.
Now, we will draw the polar plot varying omega from 0 to infinity. So, this is the
complex plane real axis imaginary axis this is the unit circle. Now, this point indicates 1,
0 and this point indicates minus 1, 0 and this is the origin now the plot starts at omega
equals to 0 when Amplitude Ratio 0.8 Kc.
So, we will consider Kc values unity 1 that means, this will be 0.8 Amplitude Ratio is 0.8
if Kc is 1 and 5 is 0. So, we can locate the point say here this is the point starting point.
So, this starting point we obtain and omega equals to 0 this is starting point which
indicates Amplitude Ratio is 0.8 and 5 is 0 what about the end point? At end point
Amplitude Ratio is 0 and 5 is minus 270 that means, this. So, if we join then we obtain
the plot like this. So this is the polar plot and deduction is this, this is the deduction of
this plot definitely we obtain polar plot considering Kc equal to 1 this the starting point
at omega equals 0 this is the end point origin when omega equals infinity and this is the
polar plot and Kc is equals 1. So, this is polar plot with varying omega from 0 to infinity.

(Refer Slide Time: 12:49)

.
In next step PI mean in step 3 varying omega from minus infinity to 0 in the third step
we will vary omega from minus infinity to 0. Now plot starts at omega equals to minus
infinity polar plot at omega equal to minus infinity. What will be the corresponding
Amplitude Ratio? Corresponding Amplitude Ratio is 0. What about phi? Phi is 270
degree. So, at omega equal to minus infinity we obtain Amplitude Ratio as 0 and phase
angle phi phi as plus 270 degree. So, what about point location of point starting point
origin. So, location of point starting point is origin.
Similarly if plot terminates the plot terminates at omega equals to 0. What about the
corresponding Amplitude Ratio? Amplitude Ratio is 0.8 Kc k c. What about phi 0
degree. So, when omega equal to 0 we obtain Amplitude Ratio as 0.8 Kc and phi 0
degree what about the location on the real axis the location of the end point should be on
real axis.
Now, we will draw the polar plot for this case I mean from varying from varying omega
from minus infinity to 0 this is real axis in complex plane this is imaginary axis similarly
we need to draw the unit circle this is the unit circle. So, this is 1, 0 and this point is
minus 1, 0 and this is the origin. So, plot starts at origin and Amplitude Ratio 0 and phi is
minus 270 and plot ends when Amplitude Ratio is 0.8 Kc and phi 0.
So, for this case in this case we will consider Kc 1. So this is the starting point and this is
the end point starting point we obtain considering omega minus infinity end point we

obtain considering omega 0. So, the polar plot is obtained as this is the deduction clock
wise. So, this plot we obtain when Kc equal to 1 this is suppose figure 2 polar plot with
varying omega from minus infinity to 0 this is figure 2 which is basically the polar plot
with varying omega from minus infinity to infinity and previously we obtain polar plot
with varying omega from 0 to infinity.
Suppose this is figure 1 in figure 1 obtain polar plot with varying omega from 0 to
infinity and in figure 2 we obtain polar plot with varying omega from minus infinity to 0.
Now, you see figure 2 is the mirror image of figure 1 it can be observed that this figure 2
is mirror image of the figure 1 with respect to the real axis.
(Refer Slide Time: 19:10)

Previously it was mentioned that we will produce the polar plot by varying omega from
minus infinity to infinity, but, we have drawn 2 plots in 1 case we have varied from 0 to
infinity and in another case we have varied from minus infinity to 0. Now, we will
combine this 2 plots. So, in the next step we perform Stability analysis and for that we
will combine 2 figures I mean figure 1 and figure 2. So, that we can show the polar plot
with varying omega from minus infinity to infinity. So, this is complex plane this is real
axis this is imaginary axis this is the unit circle this point indicates 1, 0 and this point
indicates minus 1, 0. So, we will consider Kc equal to 1 then when we varied omega
from 0 to infinity then we obtain this plot you can say this is figure 1.

Similarly if as we varied omega from minus infinity to 0 we obtain this plot agree. So,
this plot we obtained by varying omega from 0 to infinity and this plot we obtain by
varying omega from minus infinity to 0 and this polar plot corresponds Kc equal to 1
now, this point is called cross point the point C is called the cross point C is called the
cross point. Now, we will note down some important points on stability. You see when
Kc equals to 1 the cross point is to the right of minus 1 Kc equals to 1 the cross point is
to the right of minus 1 it indicates stability and when the cross point is to left of the
minus 1 that indicates in stability I mean the system is unstable .
(Refer Slide Time: 24:23)

So, when Kc if equal to 1 the cross point C is to right of minus 1 that means, the system
is stable we can define it in another way you see the plot corresponding to Kc equal to 1
does not encircle minus 1, 0 therefore, the system with Kc equal to 1 is stable the polar
plot corresponding to Kc equal to 1 does not encircle minus 1, 0 that means, the system
is stable. If, we consider Kc equals to 2 what happens?

(Refer Slide Time: 25:49)

Suppose this is the unit circle, unit circle in complex plane. If we consider Kc equals to 2
then this becomes 0.8 I mean 0.8 multiplied by 2 is 1.6 . So, it will be like this the polar
plot will be if suppose this is indicating 1.6 this is the polar plot by varying omega from
0 to infinity and this the polar plot by varying omega from minus infinity to 0. This the
polar plot when varying omega from 0 to infinity and this the polar plot when we vary
omega from minus infinity to 0. Which 1 is the cross point? This point and this is the
minus 1, 0 point you see when Kc is equals to 0 the cross point C is to the left of minus
1.
So, when Kc equals to 2 the cross point C is to the left of minus 1. So, we say that the
system is unstable by this way we can apply the Nyquist plot for stability analysis and
this criteria can be used for tuning the controllers. So, this is all about Nyquist plot next
we will discuss another Tuning Method that is Zn method Ziegler Nicolas technique.

(Refer Slide Time: 28:58)

.
In the next we will discuss another Tuning Method that is Zn tuning method Ziegler
Nicolas Tuning Method. This was proposed by Ziegler and Nicolas in 1942. This
technique is also based on Frequency response technique analysis and this technique is
known as closed loop tuning method. This technique can also used online I mean when
the process is in the operation we can use this technique for tuning purpose therefore,
this technique is also sometimes called online tuning method. So, this technique is called
closed loop as well as online tuning method.
So, in the next we will discuss how we can use this technique for tuning purpose? We
will know the several sequences states for applying this technique for tuning purpose. So,
we will discuss the step wise procedure at first you bring this process to the steady state
condition. In first set we need to bring the system to the steady state condition.
So, bring the system to the steady state condition this is the first step. Secondly, use the P
only controller you cannot use PI PPI or PID . So, first you use P only controller at the
same time you close the loop. So, close the feedback loop apply the Ponly controller and
close the feedback loop. Then, introduce a set point change. So, first we need to use we
need employ Ponly controller not PPI or PPID at the same time you close the feedback
loop P then introduce a set point change then we can tune the controller I mean we will
vary the Kc values. So, that we can obtain sustain oscillation.

So, next keep on varying Kc value. So, I am writing vary Kc value which is proportional
gain until the system oscillates with constant amplitude. So, under this second step use P
only controller at the same time close the feedback loop then introduce a set point change
then vary Kc values. So, that we can obtain sustain oscillations.
(Refer Slide Time: 35:19)

This is step 2 in the procedure next Determine ultimate gain. You see last point under
second step PPI we have mentioned that vary Kc value. So, that we can obtain sustain
oscillations that Kc values is basically ultimate gain. So, ultimate gain means the process
gain that provides the conditions of Amplitude Ratio equal to 1at phi equals to minus 180
degree. We can say that the values of Kc at which we obtain sustain oscillations and this
is represented by Kc suffix u Ku is the values of Kc at which we obtain sustain
oscillations and this corresponds to the condition of Amplitude Ratio equal to 1 at phi
equal to minus 180 degree.
In the next step we need to determine the ultimate period in 4 steps. We need to
determine the ultimate period of sustained cycling and this ultimate period is represented
by suffix u. What is ultimate period? Ultimate period is time between two successive
peaks in continuously oscillating process output continuously it is somewhat like this.
Suppose, this is y this is t it is continuously oscillating like this. So, this is the ultimate
period Pu you see this is the continuous oscillation. So, the time between 2 successive
peaks that is Pu now, we can determine this mathematically. What is period of

oscillations which we represented earlier by t? period of oscillations t equal to1 byF. F is


the cyclical frequency this is equal to 2 phi by omega basically omega is 2 phi A by
omega.
(Refer Slide Time: 39:52)

If, we consider sustain oscillation that means, what omega equal to omega co cross over
frequency therefore, we can write this Pu equal to 2 phi divided by omega co. So, this is
the ultimate period of sustained cycling. So, in step 3 we knew the calculation of ultimate
gain and in the next step we discussed the determination of ultimate period now,
knowing the values of Ku and Pu we need to use some correlations recommended by
Ziegler and Nicolas.

(Refer Slide Time: 41:50)

.
So, we will know the different correlations which have been proposed by Ziegler and
Nicolas controller type proportional gain Kc time constant tow y in minute derivative
tow d in minute. For P only controller Ziegler and Nicolas proposed the expression for
Kc which is equal to Ku divided by 2 for PI controller the correlations is Ku divided by
2.2 for Kc Pu divided by 1.2 for tow I. Next is PID controller for PID controller Kc is
Ku divided by 1.7 tow I equals Pu divided by 2 and tow d is Pu divided by 8 you see all
the correlations depend on Ku and Pu. So, knowing the values of ultimate gain Ku and
ultimate period Pu we can use this correlations recommended by Ziegler and Nicolas for
finding the controller parameter values.
Now, important points we need to note down now. First point is the tuning relations are
determined empirically to provide 1 quarter decay ratio response this correlations are
determined to provide 1 quarter decay ratio response this is the first point this
correlations are have been determined to provide 1 quarter decay ratio response. I mean
decay ration should be 1 by 4. Secondly, for the P only controller the safety margin is 2 P
because, Kc is equal to half of the stability limit. You see here because Kc is equal to half
of the stability limit that is Ku divided by 2 therefore, the safety margin for P only
controller is 2.

(Refer Slide Time: 45:36)

Next if you see the correlations we observe that the proportional gain of P only controller
is greater than proportional gain of PI controller if you see the correlations we observe
that the Kc of the P only controller is higher than the Kc of PI. What is the reason?
Because integral mode introduces additional phase leg with destabilizing effect on the
system.
Another important point is under this lower Kc maintains approximately the same safety
margins, in the last point we will discuss we will compare the Kc values of all 3
controllers you see the Kc values of PID controller is greater than Kc of the P only that
is again greater than Kc of PI.
So, the Kc values of PID controller is the largest 1 because the derivative action
introduces the phase lead with stabilizing effect on the system for PID controller we
consider large Kc values because derivative action introduces additional phase lead you
write phase lead derivative action introduces phase lead with stabilizing effect on the
system. So, these are the steps related to Ziegler and Nicolas technique next we will
solve 1 problem in which will apply this Ziegler and Nicolas technique for finding the
tuning parameters values.

(Refer Slide Time: 49:03)

Example open loop transfer function is given as k c exponential of minus 0.15s divided
y 0.5s

plus 1 open loop transfer function is given as Kc exponential of minus 0.15s

divided y 0.5s plus 1 here, time constant is 0.5 and dead time is 0.15. point. Now, tune
the PID controller using Zn technique. So, first you determine cross over frequency see
in this open loop transfer function we have considered P controller only. So, this
basically Gc GpGf Gm know here we considered Gc equals to Kc.
First you determine cross over frequency for determining cross over frequency we write
phi equals minus 180 degree What is the phi expression for open loop transfer function
given? Tan inverse minus 0.5 omega co 0.15 omega co multiplied by 180 divided by phi
that is equals minus 180 degree. If, you solve we will obtain cross over frequency as 11.6
radian per minute this is the values for cross over frequency. Then, determine ultimate
gain we can obtain ultimate gain if we consider Amplitude Ratio equal to 1 at cross over
frequency Ku is the values of Kc at which we obtain sustained oscillation and that
corresponds to the condition of Amplitude Ratio equal to 1 at cross over frequency. So,
What is the expression for Amplitude Ratio? Kc divided by 0.5 multiplied by 11.6 whole
square plus 1 equal to 1. This is the expression for Amplitude Ratio and that is equal to
1.

(Refer Slide Time: 53:30)

So, if we solve it we obtain Kc values equal 5.89 and this is nothing, but ultimate period
ultimate gain Ku. Next determine ultimate period. What is the expression for Pu? Pu is 2
phi divided by omega co that is equals to 0.542 min per cycle. So, ultimate period is
ultimate gain is 5.89 ultimate period is 0.542 now you can easily use the correlations to
determine the 3 tuning parameter values for PPID . So, for PPID controller Kc is Ku
divided by 1.7 using the value of Ku we obtain 3.465 this is Kc integral time constant
has the form Pu divided by 2 using the value of Pu we obtain 0.271.
Now, derivative time is Pu divided by 8 and this is equal to 0.068. So, by this way we
can tune P only and PPI controller also. So, today we discussed the Nyqumist Stability
Criteria we can example then we discussed the theory involved Ziegler Nicolas
technique then we discussed 1 example to apply the Zn technique.

Process Control and Instrumentation


Prof. A.K. Jana
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture - 30
Advanced Control Schemes
(Refer Slide Time: 01:25)

So, for we discussed a number of interesting topic under process control part and in the
next phase we will discuss advanced control schemes under advanced control schemes
today we will discuss this system with large dead time. So, our topic is systems with
large dead time fine this under advanced control schemes now a closed loop process
consist of a number of dynamic elements like the process the measuring device the
controller the final control element a closed loop process includes a number of elements
one is the controller the final control element the process, and the measuring device.
So, if we connect all these elements we obtain the close loop block diagram this is the
close loop block diagram almost all the dynamic element may include time delays almost
all the dynamic element may include time delays like time delay is involved in
transportation of fluids over long distances see here controller final control element
process and measuring device construct the close loop block diagram all this dynamic
element may include time delays like time delay is involved in transportation of fluids
transportation of fluids over long distances fine time delays like time delay is involved in
transportation of fluids over long distances this is related to which element this is related

to the process time delay is involved in sampling. And measurement time delay is
involved in sampling and measurement it is obvious in this close loop block diagram that
the controller needs output information y that the controller needs output information y
now that information we can get by collecting the sample and that and measuring that
sample fine. So, that takes some time therefore, we can say that the time delay is
involved in sampling and measurement of y.
So, this is related to which element measuring device this is related this is related to
measuring device similarly time delay is involved in devil upping the actuating sig signal
time delay is involved in devil upping the actuating signal and this is related to the final
controlling element this is related to the final controlling element time delay is involved
in human controller time delay is involved in human controller to thing and take proper
control action to thing and take proper control action and; obviously, this related to the
controller now in all of the situations noted above the feedback controller would provide
unsatisfactory performance due to the presence of dead time in all of the above situations
the feedback controller would provide unsatisfactory performance.
(Refer Slide Time: 08:14)

Due to the presence of dead time and we will discuss few other reasons which are related
to dead time like the effect disturbance is detected after a long period of time. So, the
effect disturbance is detected after a significant period of time why the controller provide
unsatisfactory performance, we are discussing the reasons first reason is the effect of

disturbance is detected after a significant period of time second reason is the controller
calculated the control action at present time step based on the last measurement is not
adequate next is the controller or the control action calculated at present time step based
on the last measurement is in adequate see the controller calculates the the control action
at present time step based on the last measurement originally, it is not last measurement
due to the dead time this is actually the measurement some steps back. So, this originally
not last measurement this is some steps back measured signal which is some steps back
therefore, the control action is not adequate for the existing situation third one is the
control action. Also takes time to make its effect felt by the process these are the reasons
for which the feedback controller would provide unsatisfactory response.
(Refer Slide Time: 11:49)

In the next we will take one example to observe the effect of dead time in the next we
will consider one example to investigate the effects of dead time. Consider the open loop
trans verve function given as K c exponential of minus t d s divided by 0.1s plus 1 this is
the open loop trans verve function which we considered earlier fine now t d cross over
frequency and ultimate gain if t d is 0.123, as we considered earlier we can easily
calculate the cross over frequency seventeen radian per minute if t d is 0.123 minute then
the cross over frequency is 17 radian per minute this we can calculate and the
corresponding ultimate gain is 2 similarly if we increase t d 0.236 then the cross over
frequency is reduced to 10 radian per minute and ultimate gain is also reduced to 1.4, if
we increase further t d to suppose 0.535 then the cross over frequency is decreased to 5

and the ultimate gain is also decreased to 1.1, this datum. We obtain based on this open
loop trans verve function fine now it is very obvious that with the increase of t d the
cross over frequency and ultimate gain both decreased. So, with the increase of with the
increase of t d the cross over frequency is decreases it means the close loop response is
sensitive to lower frequency periodic disturbances it means the the close loop response is
sensitive to lower frequency periodic disturbances this is the first observation. Secondly,
with the increase of t d ultimate gain decreases.
(Refer Slide Time: 15:33)

That means, we are force to select K c values low low values we have to select I mean
we have to we need to reduce K c to avoid instabilities fine because the ultimate gain is
the maximum values we can take. So, if K u decreases we have to select lower value. So,
if we select lower K c value then what happen? The response becomes shiggish. So, if
we decrease K c value then the close look response becomes sluggish. So, these are the
effect of dead time. So, due to the presence of dead time we observe that the close loop
response or the controller performance is degraded therefore, there is a need to devise a
dead time compensative, fine.
So, our next topic is the derivation of dead time compensative dead time compensative
for deriving the dead time compensator we will consider the simple close look block
diagram the block diagram looks like this, this is the controller block we are assuming G
f G m both are one this is the process fine this the process this the process.

(Refer Slide Time: 17:19)

So, this is the closed loop process for servo case fine now we assume here that all the
dead time is caused by the process here we assume here that all the dead time is caused
by the process fine usually we consider we we denote g p to represent the process. So,
here G p is equal to G s multiplied by exponential of minus t d s this is the dead time free
part G s is the dead time free part and this is the dead time part fine this is the trans verve
function of the process which we consider in this block diagram.
(Refer Slide Time: 20:52)

Now, the open loop response we write G o l which is equal to y m bar divided by y set
point bar, and this is written as G p G f G m G c multiplication of four individual trans
verve functions this is the generalized form of open loop trans verve function fine we
know all these individual trans verve. We know the expression for all these trans verve
functions substituting those we obtain y bar here y m bar and y bar both are identical
because we are considered G m equals one then we obtain G c G s exponential of minus t
d s y set point bar this is the expression for output response y, if we substitute all these
individual trans verve functions in this equation we obtain this now it is obvious that this.
is delayed by time t d it is obvious from this expression that it is delayed by tow d minute
Now what is our aim basically this information y bar is feed back to the comparator our
target is to gives dead time free response to this comparator. So, that the controller is not
effected fine. Now, if we consider this is the ideal response I mean if we represent
anyways first I am just mentioning the aim aim is to remove this dead time part from the
input signal to the comparator this is our aim aim is to remove this dead time part from
the input signal to the comparator, fine.
So, ideally the response should be G c G y set point bar agree this response includes this
dead time part we have to devise the comparator. So, that the input signal to the
comparator does not include this dead time part. So, if we represent the ideal response by
y star then y star should be G c G y set point. Now how we can obtain this y stat we can
obtain this y star this suppose y star and the information which is coming from process to
deduct measurement that is y bar we are just adding some thing with this y bar.

(Refer Slide Time: 25:50)

So, that we can obtain y star anyway I am just drawing this figure this is G c this is our
process that includes dead time part this y bar this is comparator y set point our target is
to give y star information to the comparator, but originally we getting y bar fine, but we
need to give dead time free response to this comparator. How that is possible suppose we
adding another signal denoted by y bar with this y to obtain y star it is straight forward to
calculate this y prime I mean y star is basically if the summation of y and y prime y star
is known to us that is G c G multiplied by y set point y is G c G exponential of minus t d
s y set point what will be the y prime that we calculate that will be if one minus
exponential minus t d s multiplied by G c G y set point. So, if we add this expression
with the expression of y then only we get dead time free response which can be feedback
to the comparator fine.

(Refer Slide Time: 29:06)

Now we need to include this expression in this block diagram how we can do that you
see, we can write this equation further as y prime divided by y set point which is equal to
G c G. So, this is the output and this is the input then the trans verve function can be
represented by this expression and this is nothing, but the trans verve function of the
dead time compensator this is the trans verve function of the compensator output of the
compensator is y prime input is y set point, fine. Now we will just include this
compensator in this block diagram. So, this is the block diagram this is the expression for
compensator this is the block diagram output is y prime and input is y set point. So, I am
writing here one minus exponential minus t d s can I connect this block like this, output
is y prime which is equal to this multiplied by G this is G multiplied by G c this is G c
multiplied by y set point this is y set point can we do this no we cannot do this because
the actual process includes G and this exponential term in the actual process, we cannot
separate I mean we cannot get information from the actual process excluding this dead
time part.
So, this not correct we can connect like this then you see in the expression output is y
prime then this part which is written here multiplied by G c this is G c multiplied by G c
this G c multiplied by g we can write here G fine. So, this block represent the dead time
compensator this block represent the dead time compensator and this compensator is
proposed by O J M smith this is this was proposed by O J M smith therefore, sometimes
it is called as smith predicator it is also called as smith predicator fine.

(Refer Slide Time: 33:54)

So, we obtain the block diagram for the compensator G compensator output is y prime
and input is y set point which is represented by G c G one minus exponential of minus t d
s fine now in this expression G and t d are involved now G is the process model and t d is
the dead time. So, t d is the dead time. So, the dead time compensative depends on the
accuracy of process model as well as the value of dead time this dead time compensator
depends on the process model represented by G this also depends on the value of t d we
can get perfect compensation only if G d and t d r exactly known we can get perfect
compensation if the process model and t d value are exactly known, but in practice this
never known exactly.
So, they are known approximately fine in practice the both process model and t d are
known approximately if that is the case then we will represent the approximate model as
g prime and the approximate t d as t d prime fine we will represent the G prime for the
approximate model and t d prime for approximate dead time, then what will be if y star
we know y star is y bar which is which is the actual process output plus y prime which is
the output of dead time compensative y bar is G c G exponential minus t d s.

(Refer Slide Time: 36:28)

This y bar and how much will be y prime if G and t d are not exactly from this equation
from this block y prime will be one minus exponential of minus t d prime s multiplied by
G c G prime and this will be multiplied by y set point bar agree just in this expression of
y prime we considered t d prime and G prime in place of t d and G rearranging this we
obtain G c multiplied by G prime plus G exponential of minus t d s minus g prime
exponential of minus t d prime s y set point bar rearranging we obtain this expression for
y star fine.
(Refer Slide Time: 38:55)

So, what conclusion we can make based on these expression when we get perfect
compensation we can get perfect compensation when G equal to and t d equal to t d
prime I mean when the process model and dead time value exactly known then only we
can get perfect dead time compensation the larger the modeling error that is G minus G
prime and t d minus t d prime that less effective is the compensation fine third remark
which one is more crucial
on the dead time compensation I mean error in G or error in t d has sever effect on dead
time compensation t d because it is related to exponential term. So, this is more crucial to
the dead time compensation because of the exponential function fourth remark see In this
dead time compensator expression t d is considered as constant quantity, but some time
the t d is vary with the time some time the dead time t d may vary with time that is not
taken into the account in the dead time compensator expression. So, that is you can say
that is a draw back.
(Refer Slide Time: 41:23)

Next we will discuss the dead time compensation and the effect of error in t d next we
will discuss the dead time compensation and the effect of error in t d we observe that
there may the possibility of the error in both G and t d, but we are considering only error
in t d. So, for this purpose we are consider a process represented by G s exponential of
minus t d s which is again 0.1s plus 1 exponential minus 0.535 s fine to discuss this topic
we are considering this trans verve function to represent the process see this the trans

verve function of the true process I mean this is g. So, this true trans verve function of
the process; obviously, G s is 1 divided by 0.1 s plus one and t d is 0.535 it is obvious
that G s is 1 divided by 0.1 s plus 1 and t d is 0.535 we will consider P only controller.
So, controller trans verve function, we can write as K c this P only controller we will
consider here both G m and G f equal to one for simplicity, we are considering G f G m
and G f both are one.
(Refer Slide Time: 44:06)

Now we will consider three different cases in the first case we will consider there is no
compensator case 1 in case 1 we consider no compensator fine the open loop trans verve
function which we are considering is represented by K c exponential minus 0.535 s
divided by 0.1 s plus 1 fine this is given to us. So, based on this trans verve function we
can calculate the corresponding cross over frequency that is 5 radian per minute.
In fact, this problem we discussed previously and we obtain the cross over frequency is 5
radian per minute corresponding ultimate gain is 1.1 fine. So, what value for K c we can
select we have to select the K c value lower than 1.1 fine this is the ultimate value at
which we get sustain oscillation.

(Refer Slide Time: 45:44)

So, for stability if we K c should be less than 1.1 for stable response, we have select K c
value lower than 1.1 if that is the case what will be the offset maximum holus we can
select for K c that is 1.1 fine at which we obtain sustain oscillation if that is the case what
will be the offset the expression for offset for the example system we can derive as
1divided by 1 plus K p K c this is the expression for the offset. For the example system
fine. So, if we substitute the values then we obtain 1 divided by 1 plus how much is K p
one how much is K c, suppose we are considering 1.1 then it is0.48 this is the large offset
if we consider the maximum holus of K c that is 1.1 then we obtain the offset 0.48 say
this is time this y this is suppose one this is the offset which is equal to 0.48.

(Refer Slide Time: 47:33)

That means, the corresponding values is 0.52 this is closed loop response with K c equal
to 1.1 fine. So, this case 1 if we do not use dead time compensator and if we use
maximum permissible K c value that is 1.1 we obtain offset 0.48.
(Refer Slide Time: 48:47)

Next, we will consider case 2 and that is perfect compensation in case 2, we consider
perfect compensation in case 2, we consider perfect compensator it is true when G is G
prime is equal to G perfect compensation is possible when G prime is equal to G; that
means, process model is perfectly known. Secondly, this t d prime is t d; that means, the

dead time value is exactly known for the case of perfect compensation y star by y set
point is equal to G c G I mean there will be not dead time part. So much is this 1 K c
divided by 0.1 s plus 1 agree there is no dead time part. So, for perfect compensation we
get y star by y set point equal to k c divided by 0.1 s plus 1 now this is first order trans
verve function this is a trans verve function of a first order system K c by 0.1 s plus 1 is
there any cross over frequency no because if we vary omega from 0 to infinity we obtain
the variation 5 from 0 to minus 90. So, there is no existence cross over frequency if there
is no existence of cross over frequency then we can use any large value of K c we can
use any large value of K c there will not be instability problem fine since there is no
existence cross over frequency. So, we can use any large value of K c suppose this value
is hundred since there is no instability problem we are considering large value of K c and
that is 100.
If K c is 100 what will be the offset 1 divided by 1 plus K p K c that is 0.01 fine without
using dead time compensator we got 0.48 we using perfect compensator we are getting
insignificant offset this is insignificant offset fine, but we mentioned that in practice g
and t d they are not known exactly.
(Refer Slide Time: 52:54)

So, that will consider in third case in case three we consider imperfect compensator,
imperfect compensator fine process gain for the example system is one time constant for
the example system is 0.1 suppose these are perfectly known; that means, G prime equal

to G equal to 0.1 s plus 1 this is the K p which is 1 this tow p which is 0.1 the dead time,
that is not perfectly known this is not perfectly known to us true value of this dead time
0.535 this is true value and approximate holus is 0.3 fine original holus is 0.535, but we
have the devil up the compensator based on this holus of 0.3 now question is we are
using K c holus 100, we know that our compensator is perfect question is what about
stability when K c equal to 100 what about stability when K c equal to 100 original holus
is 0.535 we are using 0.3. So, how much is the residual dead time residual dead time is
0.535 minus 0.3 that is 0.235, that is residual dead time it is clear that the compensation
is not perfect.
Now uncompensated dead time corresponds to omega c not 10 radian per minute fine
uncompensated dead time corresponds to K u 1.4 at beginning of today class we got this
values fine for the dead time of 0.235, we obtain cross over frequency 10 radian per
minute and ultimate gain 1.4, but how much we are using what holus of K c we are using
K c equal to 100, but maximum I mean I mean limit is up to 1.4 what about stability
unstable system becomes unstable fine. Now conclusion is be conservative in selecting
the K c holus for the process having dead time be conservative for selecting the K c
holus for the process having dead time.
(Refer Slide Time: 57:13)

So, this is all about the smith predictor and dead time compensator. In the next class, we
will discuss the another compensator that is inverse response compensative.
Thank you

Process Control and Instrumentation


Prof. A. K. Jana
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture - 31
Advanced Control Schemes
(Refer Slide Time: 00:58)

So, today, we will discuss inverse response compensator. In the last class, we discuss the
dead time compensator. Inverse response compensator can be derived based on the same
concept. Now, inverse response behavior we started earlier fine. Inverse response
behavior is the need result of 2 opposing effects. Inverse response behavior is the net
result of 2opposing effects. What is the condition for inverse response behavior? What is
the condition for inverse response? Transform function has a positive zero, transform
function should have positive zero then only the inverse response behavior it is obtained.
Now, this inverse response compensator was proposed by Linoya and Altpeter in 1962
based on the general concept of smith predictor or dead time compensator. They have
proposed this inverse response compensator based on the general concept of smith
predictor or dead time compensator. Next, we will discuss development of inverse
response compensator.

(Refer Slide Time: 04:05)

.
Next, we will discuss the development of inverse response compensator. Consider to first
order processes acting in parallel. Consider to first order processes acting in parallel and
the first order processes are represented by K p 1 divided by tau p 1 s plus 1. And
another process has the transform function of K p 2 divided by tau p 2 s plus 1, this two
first order processes acting in parallel Input to this first order processes is f just is a input.
Now, it is mention that inverse response behavior is the net result of 2 opposing effect.
So, we will consider one output positive another positive negative. Then the output is say
y fine it is obvious, that this 2 process first order processes are acting in parallel and they
have opposing effects. Now, we can write the expression for y as K p 1 divided by tau p
1 s plus 1 minus K p 2 divided by tau p 2 s plus 1 multiplied by f bar. Can we write this
output equal to transform function multiplied by input? Rearranging this we can write as
K p 1 tau p 2 minus K p 2 tau p 1 s plus K p 1 minus K p 2 divided by tau p 1 s plus 1
multiplied by tau p 2 s plus 1 f. This we can get from this expression. Now, can we find
the 0 from this expression? What will be the expression for 0?

(Refer Slide Time: 07:38)

Minus K p 1 minus K p 2 divided by K p 1 tau p 2 minus K p 2 tau p 1 this is a 0 agree.


So, what is the condition for inverse response behavior? This 0 should be positive; that
means, the condition is this should be greater than 0, this is the condition for inverse
response behavior. Next, we will implement the controller i mean we want to develop the
closed loaf block diagram.
(Refer Slide Time: 08:39)

So, our process has the transform function of K p 1 divided by tau p 1 s plus 1 and K p 2
divided by tau p 2 s plus 1 with opposite effects. The transform function for the

controller is G c input to the comparator is set point fine here we consider G m and G f
both are equal to 1 for simplicity. So, this is the closed loaf process, fine we have just
included the transform for the controller with the process sowing inverse response.
(Refer Slide Time: 10:33)

Now, what is the open loop transfer function? Open loop transfer function; we can write
as y by y set point; this is the open loop transfer function. And this can be expressed in
terms of individual transfer function as G c multiplied by K p 1 divided by tau p 1 s plus
1 minus K p 2 divided by tau p 2 s plus 1. Because we assumed G m G f both are 1. So,
G c multiplied by G p this we can write as G c K p 1 tau p 2 minus K p 2 tau p 1 s plus K
p 1 minus K p 2 divided by tau p 1 s plus 1 tau p 2 s plus 1. Now, as we determined f
prime for dead time compensator. Similarly, here also we need to find the expression for
f prime f prime is nothing but the output of the compensator a prime is the output of the
compensator.
And this is written as G c K 1 divided by tau p 2 s plus 1 minus 1 divided by tau p 1 s
plus 1 y set point bar fine basically this is the output y bar. Now, this is not a prime this
is y prime y is the process output y prime is the compensator output y prime is
calculated. Show that ideally, we obtained inverse free response and that is y star, y is the
process output y prime is the compensator output. If we add this 2, we obtain y star that
is the ideal response which is inverse free response. We did the same thing for the case
of dead time compensator and another thing I want to mention here that this K basically

the, determine the condition. Based on this K fine we will determine the condition based
on the K, any way what is the ideal, what should be the ideal response y star?
Now, we can calculate this y star substituting the expressions for y and y prime
substitute. Now, G c multiplied by K p 1 divided by tau p 1 s plus 1 minus K p 2 divided
by tau p 2 s plus 1. Then K divided by tau 2 s plus 1 minus K divided by tau p 1 s plus 1
multiplied by y set point substituting the expressions of y. And y prime we obtain this;
this is y star, we can write this as G c K p 1 tau p 2 minus K p 2 tau p 1 plus K multiplied
by tau p 1 minus tau p 2 s plus K p 1 minus K p 2 divided by tau p 1 s plus 1 tau p 2 s
plus 1 y set point. Rearranging the initial expression previous expression we get this. Can
you find the expression for 0, what will the expression for 0, which we can obtain for this
expression?
(Refer Slide Time: 17:12)

The expression for 0 will be minus K p 1 minus K p 2 divided by K p 1 tau p 2 minus K


p 2 tau p 1 plus K tau p 1 minus tau p 2. Basically 0, we obtain by considering this
enumerator equal to 0. So, this is the expression for 0. Now, to avoid the inverse
response behavior this should be less than or equal to 0 agree. The condition for inverse
response behavior is this 0 should be greater than 0. Now, to avoid the inverse response
behavior it should be less than 0. Now, question is that what condition of K this is true at
what condition of K this is true; this is true when K is greater than or equal to K p 2 tau p

1 minus K p 1 tau p 2 divided by tau p 1 minus tau p 2. This 0 should be less than 0 when
this condition is satisfied.
Now, what is the transfer function for the inverse response compensator? The transfer
function for inverse response compensator, we can write as K 1 divided by tau p 2 s plus
1 minus 1 divided by tau p 1 s plus 1. This is the transfer function for the inverse
response compensator. This is the transfer function for the inverse response
compensator; you just consider the expression which we consider for y prime. This is a
output; output is inverse response compensator transfer function then this is the
controller transfer function multiplied by set point. So, it is obvious, that this is the
transfer function for the inverse response compensator which I have written here. Now,
we vesture incorporate the block of inverse response compensator in the closed loop
block diagram as we deed for the case of dead time compensator.
(Refer Slide Time: 20:49)

Now, our process has the transfer function of K p 1 divided by tau p s plus 1 and K p 2
divided by tau 2 s plus 1. They have the opposing effect output is y this is the controller
G c this is set point y. Now, we will place the compensator here you see the transfer
function, the transfer function of the compensator can be expressed as K multiplied by 1
divided by tau p 2 s plus 1 minus 1 divided by tau 1 s plus 1 agree. This is the transfer
function of the inverse response compensator definitely it is multiplied with G c. So, we

can consider this input output is y prime and y is a output of the process this is y. if this 2
signals are added.
Then we obtain ideal response i mean inverse response inverse feed response. Now, if
you see this expression this expression y prime is the output y prime is the output. That is
multiplied that is equal to G c this is G c multiplied with K into this one which is the
inverse response compensator then this s y set point y set point. So, to obtain this y prime
this block is included in the closed flop block diagram. So, this is basically the inverse
response compensator. Now, how I mean at what situation we can get perfect
compensation at what situation we can get perfect compensation?
If it is represented by K p 2 divided by tau p 2 s plus 1 minus K p 1 divided by tau p 1 s
plus 1. This is the original transfer function of inverse response compensator; we can get
perfect compensation if the inverse response compensator has this transfer function. You
can substitute the y prime related to this expression in the expression of y star to see the
perfect compensation. So, this is first point we can get perfect compensation if the
inverse response compensator has this form. Second point is inaccurate model degrades
the inverse response compensator performance. Inaccurate model degrades the I R C
performance inaccurate model degrades the inverse response compensator performance.
So, this is all about the inverse response compensator. So, we discussed the 2
compensators one is dead time compensator. Another one is inverse response
compensator. In the next, we will discuss another advanced control scheme that is
cascade control scheme.

(Refer Slide Time: 26:58)

Cascade control scheme previously we discussed 3 different controllers P P I and P I D


controllers all are feedback controllers. And we will designate them as simple feedback
control schemes previously we discussed P P I and P I D which we are calling now
simple feedback controller. Now, for those controllers one measurement and one
manipulated variable are involved. If I write a simple equation I mean if I write a
equation for the simple P only controller we can writes that as c equal to c s plus K c y
set point minus y.
This is the equation for P only controller you see in this equation the measurement y and
manipulated variable c are involved. Because c e s c s bias signal which is basically
constant and K c is the tuning parameter and this is pacified by the control engineer. So,
in the simple feedback control engineer one measurement and one manipulated variable
are involved in a single loop. Now, we will discuss cascade control scheme in which
more than 1 measurement more than 1 measurement. And 1 manipulated variable are
involved in cascade control scheme for the simple feedback scheme 1 measurement.
And 1 manipulated variable are involved, but for the case of cascade controller more
than 1 measurement and 1manipulated variable are involved. Secondly, it involves dual
loop controller involves dual loop controller the cascade controller involved dual loop
controller 1 controller is called primary or master controller. Aanother controller is called
secondary or slave controller. The cascade controller also involve say dual loop

controller; first one is called the primary or master controller and second one is called
secondary or slave controller fine Now, to discuss this cascade control scheme, we will
take one example.
(Refer Slide Time: 31:42)

We will consider the example of a jacketed C S T R will consider the example of a


jacketed C S T R. This is a schematic of a jacketed CSTR in lets steam has the flow rate
of f with concentration C A i. And temperature T i outlet steam has a flow rate of F
composition concentration C A and temperature T an exothermic reaction occurs in the
reactor and exothermic a reversible a reaction occurs in the reactor. Now, the exothermic
heat is removed by the coolant exothermic heat is removed by passing a coolant through
the jacket.
So, suppose the coolant flow rate is F C and the temperature of the coolant is T c i the
outlet coolant flow rate is F c and temperature is T c. We assume that both the inlet and
outlet flow rates are identical and they remain constant F i equal to F not equal to F. And
they both are constant for this example system out control objective is to mention the
reactor temperature T at its set point value. The control objective is to mention the
reactor temperature T at its set point value. Now, we will try to find the pears controlled
variable is reactor temperature T corresponding manipulated variable is coolant flow rate
F c and possible disturbances are T i and T c i. Control variable is reactor variable T

corresponding manipulated variable is F c and possible disturbances are T i and T c i.


Now, we will configure the simple feedback control scheme
(Refer Slide Time: 36:18)

Inlet flow rate has the flow rate of F concentration c i temperature T I, outlet flow rate is
F temperature T and concentration C A fine our contain objective is to mention the
temperature. So, first it is required to measure the temperature T by using a sensor
temperature measuring device. Then this measure temperature T m is feedback to the
temperature controller T c measure temperature T m is compared first with set point
value T s p. Then the calculated control action is implemented through the final control
element.
This is the simple feedback control configuration temperature T is measured using the
temperature measuring device represented by T T. Then the measure temperature T m is
compared with set point value T s p. Then the control action is calculated and that action
is implemented through this valve. So, this is the simple feedback control configuration.
Now, in this configuration which one is the measured variable measured variable is T
which one is the manipulated variable manipulated variable is F c. Now, it is obvious,
that T response much faster to changes in T i then to changes in T c. There are 2
disturbances there are two possible disturbances; one is T i another one is T c. Now, this
T response much faster to changes in T i. Then to changes in T c T response much faster

to changes in T i then to changes in T c it is very obvious you see T i can directly affect
T.
Because it is affecting to the reaction mixture directly, but T c is first effect the jacket
then to the process fine T c is not directly going to the reactor. It is going to the jacket
then the effect is then the effect goes to the I mean the reactor is affected. Then therefore,
we can say that the reactor temperature response much faster to changes in T i then to
changes in T c. So, accordingly can we say the simple feedback controller is more
effective to reduce the effect of disturbance in T i, and less effective to reduce the effect
of disturbance in T c. The simple feedback controller is more effective to reduce the
effect of disturbance in T i and the controller is less effective.
To reduce the effect of disturbance in T c this conclusion, we can draw based on this
simple feedback control configuration. The simple feedback controller is more effective
to reduce the effect of disturbance in T i and less effective to reduce the effect of
disturbance in T c. Now, basically we want more effective for both the cases I mean the
controller should be more effective to reduce the effect of T i. And more effective to
reduce the effect of disturbance in T c and therefore, we need the cascade controller.
Now, we will discuss the cascade controls scheme, how it is more effective for both the
cases.
(Refer Slide Time: 43:21)

So, this is the CSTR jacketed CSTR. Now, first we measure the temperature T using one
temperature measuring device which we are representing by T T 1 this is the first
measuring device I mention earlier that cascade controller involves to more than 1
measurement. So, this is a first measuring device T T 1. Then the measure temperature
information which is T suffix m goes to the controller the temperature controller which is
T c 1. Then the output of this first temperature controller goes to another temperature
controller that is T c 2. The second measuring device T T 2 is use to measure the
disturbance T c.
This s negative this is positive and then the control action is implemented through this
final control element.. So, I am repeating the steps again our control objective is to
mention the temperature the T. So, this reactor temperature is measured by the measuring
device T T 1. Then the measure temperature T T m is compared with its set point value
then T c one calculates the control action for the simple feedback controller the control
action of T c 1 is implemented. But here another controller is included that is T c 2 here
the first temperature controller T c 1 calculates basically the set point value of coolant
temperature.
The first temperature controller basically calculates the set point of coolant temperature
that is T c set point. And the second measuring device T T 2 basically measure T c which
is the disturbance involved with the manipulated variable. So, this signal is basically T c
m measured T c. So, T c set point compared with T c m and the second controller T c 2
calculates the control action that action is implemented through this control valve. So, T
c 1 is the primary controller or master controller and T c 2 is the secondary controller or
slave controller. So, here 2 measurements are involved; one is reactor temperature T and
second one is coolant temperature T c. But manipulated variable is one I mean only one
manipulated value is involve that is F c fine
For the feedback controller scheme there is no action with respect to any change in T c,
but this controller takes action if there is any change in T c this is the difference. So,
disturbance in T i disturbance in T i will be taken care of will be taken care of by T c 1.
And this is done by primary controller T c one is primary controller; this primary
controller basically takes action after T is effected by T i this primary controller takes
action after T is effected by T i. I want to say that if there is any change in T i that should
be reflected through T then only the controller takes action. So, this disturbance first

effect the process then the controller takes action. So, disturbance in T i will be taken
care of by this primary controller. And it takes action after T is effected by T i next point
is disturbance in T c.
(Refer Slide Time: 51:12)

Can be efficiently compensated by T c 2; this is the secondary controller. If there in any


disturbance in T c that can be efficiently compensated by T c 2. This is the secondary
controller and it takes action before the effect of T c is felt by the mixture. So, previously
I mentioned that there are 2 possible disturbances; one is T i another one is T c. Now, if
there is any change in T i that will be that will taken care of by T c 1 and if there is any
disturbance in another load variable T c. That can be efficiently compensated by
secondary controller T c 2. But primary controller takes action after t is effected by T i
and secondary controller takes action before the effect of T c is felt by the mixture.
So, here 2 measurements are involved; one is T, another one T c, but manipulated
variable is 1 I mean one manipulated variable is involved that is F c. Now, basically this
jacketed CSTR consists of 2 processes; one is process 1. Process one is basically the
reactor excluding the jacket, the jacket its CSTR consists of 2 processes. One is process
one which is the reactor including the jacket and here measurement variable is T
controller is T c. One another process is process 2 that is the jacket and for process 2 the
measurement is coolant temperature and controller is T c 2. So, the jacket is CSTR
consists of 2 processes; one is process 1, another is process 2. Process 1 is the reactor

excluding the jacket and process 2 is the jacket. So, in the next class, we will discuss will
derive basically the closed loop block diagram for this cascade control scheme.
Thank you

Process Control and Instrumentation


Prof. A. K. Jana
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture - 32
Advanced Control Schemes (Contd.)
(Refer Slide Time: 00:59)

In the last class, we have discussing the cascade control scheme. So that topic will
continue today. Cascade control scheme to discuss this cascade control scheme we
considered a jacketed CSTR example. So, I am drawing that jacketed CSTR configured
with cascade control scheme. First, we measure the reactor temperature using one
measuring device, temperature measuring device T T 1, then this information goes to
temperature controller one which is the primary controller, the measure temperature T m
is compared with it set point value that is T s p. Then the temperature controller output
goes to another controller, that is T C 2; coolant temperature is measured by a measuring
device namely T T 2, and this measured coolant temperature T C m is compared with T
C set point supplied by T C 1.
Then this control action is implemented through the control valve to manipulate the
coolant flow rate F c, fine. So, this is the cascade control scheme, and coolant outlet flow
rate is F c with temperature T C. This is cascade control of a jacketed CSTR. We also
discussed that this process includes two processes, I mean process one and process two.
Process one is the reactor excluding the jacket, process one is the CSTR excluding the

jacket, fine. And for the process one measurement is reactor temperature T, measuring
device is T T 1 and controller is T C 1, this is basically the primary controller or master
controller, fine.
(Refer Slide Time: 05:59)

And process two is the jacket. Process two is the jacket, measurement involved in
process two is coolant temperature T C measuring device used is T T 2 and for process
two, the controller is T C 2, which is the secondary controller fine.
(Refer Slide Time: 07:22)

In the next step, we will develop the block diagram for this complete closed loop
process. So, process one is the reactor excluding jacket. So, we can use one block for
this, and we consider for process one disturbance d 1 fine, what is the output is the
reactor temperature T, fine. So, what we did first we measure this temperature using T T
1. So, this is T T 1 temperature transmitter, which is used as a measuring device. So, first
we measure the reactor temperature T and we denote the measure temperature as Tm
fine. Then this measure temperature is compared with set point value, you see the
schematic based on that, we are only developing this block diagram, this measure
temperature is compared with its set point value, that is T suffix s p then the error signal
goes to first temperature controller.
That is primary controller denoted by T C 1 agree, error signal goes to the primary
controller denoted by T C 1. T C 1 output is the set point of coolant temperature T c s p
fine, it is done in the schematic diagram. I mean we configure the cascade controller in
this way T C 1 output is the coolant temperature set point, and this set point value is
compared with T c m, agree, this T C set point which is supplied by the primary
controller is compared with measured coolant temperature measured coolant temperature
we obtain from T T 2 fine, then the error signal T C set point minus T C m goes to
another controller that is secondary controller T C 2, T C 2 output is implemented
through the final control element this action is implemented through the final control
element, and this control action goes to process two that is the jacket process two.
Secondary controller output enters the process two through the final control element.
Now, we can consider the disturbance for process two denoted by d2. Like for process
one, we consider d1 then this output effect process one that is nothing but T C, and this
TC is measured basically using temperature transmitter two, understood. This is the
block diagram for the closed loop cascade control system of the example jacketed
system. I am repeating this process one is the reactor excluding jacket our control
objective is to maintain the reactor temperature at its desired value. So, process one
output is temperature T that is first measured using T T 1 measured signal we denote by
T suffix m that then T m is compared with its set point value T suffix s p output of this
comparator is the error signal which goes to primary controller T C 1.
The primary controller output is the coolant temperature set point, which we represent by
T C set point. Now, this set point is compared with measured coolant temperature that is

T suffix c m. We measure the coolant temperature by the use of this temperature


transmitted two. Now, if we compare T C set point and T m then we obtain the error
signal i mean the error signal is then goes to the secondary controller T C 2, and T C 2 is
T C 2 output is implemented through the controlled valve and this action goes to the
process two that is a jacket we used here the disturbance d2 for the second process and
the output is coolant temperature which effects process one and coolant temperature here
we measured by the user T T 2. So, this is the block diagram.
This is the close loop block diagram, fine. Now, cascade controller is usually used when
the disturbances are associated with manipulated variable it is very obvious from this
example that that cascade controller takes care of the effect of disturbance associated
with manipulated variable our manipulated variable is F c. So, related disturbance is
coolant temperature T C. So, this is almost common that the cascade controller is useful
when the disturbances are associated with manipulated variable. This is useful
particularly when the disturbances are associated with the manipulated variable another
situation is when the final control element exhibits non-linear behavior, the final control
element shows non-linear behavior, here final control element control valve, fine. So,
these are the two common cases here the cascade control scheme is used. So, we
developed the cascade control configuration for a jacketed CSTR. We will take another
two examples for configuring this cascade control scheme.
(Refer Slide Time: 17:26)

So, now we will consider distillation top section. We will configure the cascade control
scheme for the top section of distillation column. At the top of this tower a vapor steam
namely if overhead vapor is condensed by the use of a condenser the condensed liquid is
then accumulated in a reflux drum, this is a reflux drum a part of this liquid is withdrawn
as top product that is distillate. Another part is recycling back to the top tray of the
column fine. This is the top section of a distillation column the overhead vapor is
condensed then the liquid is accumulated in the reflux drum a part of this accumulated
liquid is withdrawn as the top product namely distillate or distillate product and a
fraction is recycle back to the top section of the distillation column.
Now our objective our control objective is to maintain the top tray temperature, fine. So,
we are considering this is the measured variable, fine. So, this is the top tray. So, we are
first measuring the top tray temperature by the use of temperature measuring device
represented by TT, fine. Then this goes to the primary controller that is T C measured
temperature is then compared with t set point, fine. Primary controller output goes to
another controller that is flow controller F c. We measure this flow rate reflux flow rate
by the use of one flow measuring device.
So, this is the set point of this flow rate, and this is the measured value if we represent
this reflux flow rate by R then this is R measured R m and the primary controller output
is R set point, then this flow controller action is implemented through this control valve,
fine. So, this is the cascade control configuration for the distillation top section. Our
control objective is to maintain the top tray temperature. So, we consider that as the
measured variable this top temperature is measured by the use of one temperature
measuring device represented by TT then this measured temperature T m goes to the
temperature controller in which, the measured temperature is compared with its set point
value.
Then this primary controller calculates the set point value for R represented by R suffix s
p. And this flow transmitter which is measuring this flow rate as R suffix m. Now, these
two are compared in this flow controller F c and that action is implemented trough this
final control element. So, this is the cascade control scheme for the top section of the
distillation column. So, here this is basically the control variable. Top tray temperature is
the control variable what is the corresponding manipulated variable manipulated variable

is reflux rate, which is R disturbance associated with the manipulated variable


disturbance associated with the manipulated variable that is also reflux rate.
(Refer Slide Time: 25:13)

Which one is the primary controller? Primary controller is the temperature controller
denoted by T C, and secondary controller is the flow controller denoted by F c, fine. So,
this is the second example for configuring the cascade control scheme.
(Refer Slide Time: 25:54)

Next, we will consider another example that is a heat exchanger heat exchanger. Next,
we will configure the cascade control scheme for the heat exchanger this is the cell side

of a heat exchanger through which a process stream is flowing we are interested to


maintain the temperature of this process stream T, fine. So, our control objective is to
maintain the temperature of this process stream at its set point value now to maintain the
temperature we need one heating medium, which flows through the cell side possible I
told this is cell side.
Now, this is tube side through tube side the process stream is flowing our control
objective is to maintain the temperature at its set point value for maintaining the
temperature we need a heating medium. So, for this particular case we need a heating
medium for maintaining the temperature of the process stream. So, we will consider the
process stream temperature as the as the measured variable, we can measure this
temperature by using one temperature measuring device then this information is supplied
to the primary controller T C, this is a T suffix m this measured temperature is compared
with its set point value T s p.
This is basically indicating the steam flow rate then this is the primary controller T C
primary controller output goes to the secondary controller denoted by F c that is a flow
controller, and this flow rate is measured the steam flow rate is measured by the use of
one flow meter F t. So, this temperature controller gives the set point value of this flow
rate, and this is the measured flow rate these two are compared I mean this is positive
this is negative then this flow controller flow control action is implemented through this
control valve, fine. So, this is the cascade control scheme for the example heat
exchanger.
So, our control objective is temperature and we considered that as a measured variable T
T is used to measure the temperature then this measurement signal is compared with its
set point value then the temperature controller calculates the set point value for this flow
rate. So, this is the set point value of this flow rate, and this is the measured value of this
flow rate there again compared within this block F c. Then this flow controller calculates
action and that action are implemented through this control valve. So, this is the cascade
control scheme for a heat exchanger here control objective is T.
So, this also the control variable process stream outlet temperature is the controlled
variable what is the corresponding manipulated variable manipulated variable is the
steam flow rate. Disturbance associated with the manipulated variable that is also steam

flow rate then primary controller primary controller is the T C temperature controller
which one is the secondary controller secondary controller is the flow controller, fine.
So, for we discussed three examples for configuring the cascade control scheme one is
jacketed C S T R then distillation top section and finally a heat exchanger.
(Refer Slide Time: 33:17)

In all three cases we observe that the secondary controller is the flow controller fine in
all three cases we observe that the secondary loop is the flow control loop, and it is
almost common for all processes and therefore, we can say that flow control loops are
almost always cascaded with other control loops. Therefore, we can say that flow control
loops are almost always cascaded with other control loops; other control loops means
this is primary loop basically, fine.

(Refer Slide Time: 34:39)

So, next we will discuss the closed loop behavior of cascade controller, fine. To discuss
this close loop behavior we will consider a simplified block diagram. Process one has the
transfer function of say G P 1. We are representing the transfer function of process one
as GP1 previously consider the disturbance for process one is d1 output is say y, G m1 is
the measuring device for process one suppose this is one. So, this is the comparator set
point of y is y set point primary controller transfer function is say G c 1 another
comparator secondary controller transfer function is say G c 2 process two transfer
function is suppose G p 2 disturbance for process two is suppose d 2.
We are considering G m 2 equal to 1. Fine previously we develop the block diagram for
the cascade control system it is same thing only. We have included here the transfer
functions, and in addition to that we are considering here G m 1 G m 2 both are equal to
1, and G f 1 G f 2 both are equal to 1 for simplicity we are considering this, fine. So to
discuss the close loop behavior of cascade controller, we will consider this block
diagram. So, you see this is the primary loop. And this is the secondary loop this is the
primary loop, and this is the secondary loop. We will consider one by one, first we
consider secondary loop.

(Refer Slide Time: 38:53)

So, what is the open loop transfer function of the secondary loop, how we can represent
the open loop transfer function of the secondary loop open loop transfer function is
determined as the multiplication of four individual transfer functions G c G p G f G m;
Here for secondary controller G c is G c 2 for secondary loop process has a transfer
function G p 2 and G f 2 G m 2 both are 1, fine. The general expression for open loop
transfer function is written as G p G f G m G c for the secondary loop you see G m 2 and
G f 2 they are 1, g m two g f two. So, we can write G secondary G c 2 G p 2 agree.
Similarly, can you write the close loop transfer function for the secondary loop what will
be the close loop transfer function of the secondary loop; suppose an output is y prime
and the output to the secondary loop is y set point prime this is the secondary loop output
is y prime and input is y set point prime. So, what will be the expression for the close
loop transfer function output is y prime which is equal to G c 2 G p 2 divided by 1 plus G
c 2 G p 2 y set point prime plus 1 divided by 1 plus G c 2 G p 2 d 2 prime here, G d is
one, fine. So, this is the close loop transfer function for the secondary loop. Can we
include this close loop transfer function in the block diagram? Yes, we can.
Then the block diagram becomes like this G c 2 G p 2 divided by 1 plus G c 2 G p 2 then
the disturbance is added 1 divided by 1 plus G c 2 G p 2 then the process one disturbance
for process one is d1, output is y. So, previously we considered that the primary

controller output is y set point prime and input to the process one is y prime, if you
consider this equation I hope this is incorporated in this block, fine.
Y prime is equal to this; this is one multiplied by y set point prime plus this and these are
added plus this is this one multiplied by d 2, fine. Now for the secondary controller how
we can analyze the stability I mean what is the characteristics equation which can be
used for stability analysis. For secondary controller the characteristic equation we can
write as 1 plus G secondary equal to 1 and this yeah equal to zero, and this can be used
for stability analysis. So, this is the characteristics equation for the secondary controller.
(Refer Slide Time: 45:35)

Next we will consider the primary loop, what is the expression of open loop transfer
function of the primary loop. So, primary which is equal to G c 1 G p 1 G c 2 G p 2
divided by 1 plus G c 2 G p 2, agree. I mention that the open loop transfer function is
calculated by multiplying four individual transfer function G p G f G m G c. Now, if you
considered this block this block diagram and if you multiply the four individual transfer
functions then we obtain the open loop transfer function for the primary loop, fine. Like
the secondary loop for stability analysis, we need the characteristic equation for primary
loop and that we represent as 1 plus G primary equal to 0, this is the characteristic
equation for the primary loop which can be used for stability analysis, fine.

(Refer Slide Time: 47:47)

In the next, we will note down few important notes in the cascade control system there
are two control schemes, one is the primary controller, and another one is the secondary
controller. So, what type of feedback controller we can use? For the primary loop or
sorry and secondary loop usually for primary controller P, PI and PID, these three
feedback controllers are conventionally used, and commonly PI and PID controllers are
used commonly; PI and PID controllers are used as primary controller. For secondary
controller P and PI controllers are used and P only is the most common as secondary
controller P only controller is the most common because if offset exists under P only
controller that is not important because our control objective is not to maintain the output
of the secondary loop.
If offset exists under P only controller that is not so important, because our control
objective is not maintain the output of secondary loop our control objective is to
maintain the output of primary loop I mean our control of objective is the measured
variable T not T C. So, this is the first point. So, if you see the characteristic equation for
primary and secondary loops. We can say that the dynamics secondary loops are much
faster than the dynamics of primary loop, the dynamics of secondary loop is much faster
than the dynamics of primary loop, fine. You see the characteristics equation for primary
and secondary loop, and you observe the order of the polynomial with the increase of the
order of a system the response becomes more sluggish.

Therefore, we can conclude that the dynamics of this secondary loop is much faster than
the dynamics of primary loop. So, based on this can we write the phase lag for secondary
loop is less than phase lag for closed primary loop based on this common can we say that
phase lag of closed secondary loop is less than the phase lag of closed primary loop.
Again can we say the cross over frequency of secondary loop is greater than the cross
over frequency of primary loop, if first comment is true then we can make the second
comment if second comment is true, then we can say that the cross over frequency of
secondary loop is higher than the primary loop.
(Refer Slide Time: 52:54)

If the third comment is true, we can say that we can use higher gains for the secondary
loop higher gains for the secondary controller; this is basically K c controller gain. So,
we can conclude that for secondary controller we can use higher gain compared to the
primary controller, fine. In the third note we to discuss the cascade controller tuning so,
cascade controller basically involve two controllers; one is primary controller, and
another one is secondary controller. Secondary controlled is tuned first, secondary
controller is tuned first, and it is tuned tightly as high a proportional gain value as
possible it is tuned tightly means as high a proportional gain value as possible, and for
tuning, we can use the Cohen cutn technique ziglar nicoles technique, we can use phase
margin gain margin technique.

(Refer Slide Time: 55:09)

We can use time integral performance criteria, fine. After tuning the secondary controller
the primary controller is tuned the primary controller is tuned. First, we tune the
secondary controller then primary controller is tuned, and for this we usually use
frequency response techniques frequency response techniques, like phase margin gain
margin method, zeglar nicoles technique, etcetera. So, by this way we tune the cascade
controller, fine.
Thank you.

Process Control and Instrumentation


Prof. A. K. Jana
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture - 33
Advanced Control Schemes (Contd.)
In the last class we discussed one advanced controlled scheme that is cascade
control scheme. And today we will start another scheme, that is constraint controller or
override controller.
(Refer Slide Time: 01:06)

So, today we will discuss override control, and it is also called as constraint control.
During the normal operation of the plant or during the startup or start down, some
abnormal situations may arise, which lead to the distraction of equipment and operating
personal. During the normal operation of the plant or during the startup or start down,
some abnormal situations may arise, some abnormal situations may arise or some
dangerous situations may arise, which lead to destruction of equipment and operating
personal.
In these situations a special type of switch is used. One switch is like HSS - high selector
switch, this is one type of switch which is used in abnormal situations, and this switch
prevents to exceed the upper limit, it prevents to exceed the upper limit. Fine in every
operation there is a certain limit in terms of upper limit and lower limit that should be

((Refer Time: 04:18)). So, we can say that this upper limit is a constraint fine, therefore
the name constraint controller
Another switch is also used that is LSS - low selector switch, and it prevents to exceed
the lower limit, which is also a constraint, fine. So, these two switches are commonly
used in abnormal situations. Now, we will discuss the application of this special control
scheme, constraint controller with taking one example.
(Refer Slide Time: 05:45)

So, we will take a boiler system to discuss the application of this switch we will take the
example of a boiler system. So, water enters at the bottom feed is water, and product is
the steam this is a steam boiler fine. So, water is introduced here now for producing
steam from water we need a heating medium, which is passing through this coil. This is
heating medium, which is supplying heat and this is indicating the water level this is a
water level this is the discharge line a control valve is installed in the discharge line fine.
So, this is the schematic of the process
Now, we use to include the controlled scheme. So, first step pressure transmitter is
installed employed to measure steam pressure, then a pressure controller. Here we use
one switch that is LSS we will discuss why we have selected LSS. Another controlled
scheme is also employed to maintain the liquid level. So, for that we need to employ one
level measuring device that is LT then one level controller then the controller signal goes
to LSS. LSS output is implemented through the control valve fine. This is the constraint

control for a boiler system fine. So, what is our primary control objective our primary
control objective is to maintain the pressure of the steam.
(Refer Slide Time: 09:49)

So, our primary objective is to use a pressure control loop for maintaining the pressure of
the steam, suppose this is loop one and this is loop two fine. So, the steam pressure is
controlled by the use of a pressure control loop this is the pressure loop I mean loop 1.
So, primary objective is to maintain the pressure by the use of loop 1 at the same time it
is require to maintain the liquid level. So, that this heating coil is immersed in liquid in
water
So, along with the control of the pressure it is require maintaining the liquid level. So,
that this heating coil is immersed in water fine. If it is not immersed in water this coil
burns out. So, another objective is we need to maintain the liquid level. I mean liquid
level should not fall below a lower limit purpose is to keep this heating coil immersed in
water; otherwise the coil burns out fine. So, our purpose is to maintain the lower limit for
the liquid level, which switch we can select LSS because it prevents to exceed the lower
limit therefore, we have selected here LSS.
Now, we will consider one situation how it functions we will consider one situation
suppose, a liquid level falls below the lower limit accordingly the level controller
reduces its output fine. If the liquid level falls below the lower limit the level controller
reduces its output to close the valve. So, the control action taken by level controller that

is level controller reduces its output to close the valve. This is the action taken by level
controller fine. This is the situation the liquid level falls below the lower limit this is the
action taken by the level controller.
Now, can you explain it, why the level controller attains to close the valve. If the valve is
closed what happens the pressure inside the boiler increase. So, what about the boiling
rate boiling rate decreases. If boiling rate decreases the liquid level starts improving if
boiling rate decreases then liquid level improves agree. Therefore, the level controller
reduces its output to close the valve, fine. It closes the valve because it exceeds the lower
limit. So, therefore, the crucial action is required to take from the level controller sides.
(Refer Slide Time: 15:57)

So, this is a situation and corresponding action taken by the level controller. Now I am
just trying to highlight the steps initially the pressure controller is in action initially the
pressure controller or pressure control loop is in action then the water level falls below
the lower limit water level falls below allowable limit. This situation we have
considered, and then level controller reduces its output to close the valve fine.
Now, when the level controllers output drops below the pressure controllers output the
LSS selects the level controllers output signal this is very important point. If level
controller output falls below the pressure controllers output, then LSS selects the lower
value LSS selects the level controllers output. So, you see previously the LSS, this is
LSS previously the LSS selects pressure controller output. Now as the liquid level falls

below the allowable limit the level controllers output reduces, and when the level
controller outputs falls below this pressure controller output the LSS select this lowest
amount I mean level controllers output.
So, we can say that the level controller overrides the pressure controller. We can say that
the level controller overrides the pressure controller fine, therefore this controller is also
called override controller. Previously we discussed the reason to call it constraint
controller and this is the reason to call it override controller fine level controller
overrides the pressure controller. So, in this example we have discussed the application
of LSS, LSS selects basically the lower output signal for case of HSS the fact is opposite
HSS selects the higher output signal fine. So this is about the constraint controller.
(Refer Slide Time: 19:39)

Next we will discuss another advanced controlled scheme that is split-range control. We
discussed the cascade control scheme previously. And in the cascade control scheme
there was more than one measurement, but one manipulated input for the cascade control
scheme more than one measurement and one manipulated variable are involved. But for
this split-range control the fact is opposite. One measurement, but more than one
manipulated variable involved, for split-range control one measurement which is
basically the controlled output or you can say controlled variable. And more than one
manipulated variable are involved fine. This split-range control provides added safety

and operational optimality. And these controllers are not very common in chemical
engineering fine, split-range controller is not very common in chemical engineering.
So, we will discuss the concept of split-range control with taking one example. We will
take one example to discuss this split-range control we will take a non-isothermal batch
reactor to discuss this control non-isothermal batch reactor. In this non-isothermal batch
reactor the temperature varies with time fine. The specified temperature program is given
as the batch reactor starts, I mean at the starting at the beginning of the batch operation it
operates at 15 degree Celsius. At the beginning of the example operator the temperature
is 15 degree Celsius. And at the end of batch operation the temperature is 100 degree
Celsius. So obviously, the temperature changes with respect to time in the example batch
operation and it is common for batch operations.
So, if we want to maintain 15 degree Celsius or near to that temperature we need to use
one cooling medium. If we want to maintain 100 or nearby temperature then we need to
use one heating medium so here basically if cooling medium is cooling water cooling
water can be used as cooling medium, which is available at 5 degree Celsius. Similarly,
if as heating medium steam is steam can be used, which is supposing available at 180
degree celsius? So, these are the two medium, which can be used to maintain the
temperature of the example non-isothermal batch reactor.
(Refer Slide Time: 25:39)

Now, we will try to configure the split-range control for this example system. This is a
schematic of the non-isothermal batch reactor this is jacketed batch reactor fine. One
valve is used for cooling medium and another valve is used for heating medium. Now the
split-range temperature controller can be used to manipulate the cooling water, as well as
the steam SRTC is the split-range temperature controller fine. This is the jacket
temperature T j. So, we need to give the set point value of T j is the jacket temperature
and sp has been used to represent the set point.
Now, the control action directs to manipulate the steam, and this is also used to
manipulate cooling water, then this action in combined form is implemented fine. So,
this is the split-range temperature control for a non-isothermal batch reactor. This is the
jacket temperature, which is measured and this set point of the jacket temperature is also
supplied to this split-range controller. Then this controller manipulates both steam as
well as cooling water for maintain the temperature throughout the batch operation, which
varies from 15 to 100 degree Celsius, fine. So, this is the split-range controller.
(Refer Slide Time: 29:36)

And this controller is used for broad span temperature control of small plant this is used
for broad span temperature control of small plant or pilot scale plant. This is also used
for ear bound heating and cooling of office buildings, this is also used for ear round
heating and cooling of office buildings. This is also used to control PH in waste water
treatment plant control PH in waste water treatment by manipulating both acid and base

fine. So, these are the typical examples of split-range control scheme. In the next we will
discuss another advanced control scheme that is feed forward control scheme, we have
started feedback control scheme like PEPI and PID controllers. So, now, we use to
discuss another control scheme that is feed forward control scheme.
(Refer Slide Time: 31:50)

So, before discussing this feed forward control scheme, we are interested to know the
limitations of feedback control schemes fine. Then we will proceed to discuss this feed
forward control scheme. So, what are the limitations of feed forward control scheme
sorry limitations of feedback control scheme, this point we use to discuss. So, for that we
can take one example like a heating tank system, which we discuss earlier.
So, we will take a heating tank system to absorb the limitation feed is entering at a flow
rate of Fi with temperature Ti. The product is coming out at flow rate F and temperature
T. A heating medium is used to maintain the temperature, suppose that heating medium
is steam, which has the flow rate of Q fine. We use to discuss the limitation with this
example taking this example. So, what is the control objective is to maintain temperature
at its set point value. So, we can say that temperature is the controlled variable and
corresponding manipulated variable is steam flow rate that is Q, fine.
Now, we use to employ one feedback control scheme. So, we can configure the feedback
control scheme. Now first we need to measure the temperature by using a temperature
measuring device denoted by TT then that measures signal is compared with set point

value. Then this signal is used to calculate the control action and then this action is
implemented through the control valve fine. So, this is the feedback control of the
heating tank system feedback control configuration for the heating tank system.
Now, you see here we have used one controller that is TC, suppose this is a PI controller
the temperature controller, which we use to maintain the temperature is suppose a PI
controller that is a feedback controller. Now how can we write the I mean, what will be
the equation for this PI controller manipulated variable is Q. So, Q is equal to plus Kc
multiplied by error plus Kc divided by tau i integration of error dt this is the temperature
controller equation fine.
Now, in this equation you see Q s, K c and tau i are constant parameters. In this PI
controller equation Q s which is the bias signal K c is a proportional gain and tau i is a
time constant, these three are constant parameters fine, last two are the tuning
parameters. Now Q changes with the change of which variable error signal so we can say
that the Q changes only if error. That means, if temperature changes because T set point
is again the constant Q changes only if error, that means, only if the temperature changes
fine. So, what we want to say that this feedback control scheme takes action only the
temperature changes detected otherwise you can it cannot take action, anyway we will
consider one situation to elaborate this.
(Refer Slide Time: 39:10)

Suppose there is a change in T I, T i is a disturbance variable, there is a change in T i.


You see in this example T i is the disturbance variable initially the process is at steady
state now we are considering a situation that is there is a change in T i. Suppose T i has
increased. So, it will affect the temperature in the next step we can say that it affects the
tank temperature agree, what will happen temperature will increase or decrease
temperature will increase. So, what about the value of epsilon changes, fine. Basically
this error signal decreases and it becomes negative, if initially the process is at steady
state fine at steady state error is zero, because t set point and t both are equal.
Now, if temperature increases then this the value of epsilon becomes negative, then the
controller will take action, because error signal has changed, so according to this
equation, if error the value of epsilon changes then Q changes fine. So, we can say that in
the next step controller takes action how Q decreases or increases Q decreases.
So, you see the controller acts after the effect of a disturbance has been felt by the
process the controller acts after the effect of T i has been felt by the process agree. So, it
is not possible to afford the impact of T i by the use feedback controller. It means that the
perfect control can never be achieved by the use of feedback controller agree. So, how
can we get the perfect control, we can get the perfect control. If the controller takes
perfect action before the disturbance effects the process isnt it we can get perfect control
if the controller takes perfect action before the disturbance effects the process.
(Refer Slide Time: 44:08)

So, perfect control we can achieve, if the controller takes perfect action before the
disturbance effects the process agrees. So, what is require to do to take the perfect
control action. So, for perfect control action first it is require to measure the disturbance
which is effecting the process this is the first point. It is required to measure the
disturbance which is effecting the process. So, what is require to do to achieve the
perfect control action first it is require to measure the disturbance variable, fine. First it is
require to measure the disturbance variable in the next step the controller should act
before the effect is felt by the process. So, for perfect control action it is require to
measure the disturbance to know the change in disturbance variable, and then the
controller should act before the effect is felt by the process. And this is the mechanism of
feed forward control scheme and theoretically it can provide perfect control fine.
So, we can say that the feedback controller acts after the fact in compensatory manner
and feed forward controller acts beforehand in anticipatory manner. So, feedback
controller acts after the fact. I mean after effecting the disturbance to the process in a
compensatory manner, where is a feed forward controller acts beforehand in an
anticipatory manner fine. So, theoretically we can achieve the perfect control from the
feed forward controller. Anyway before going to discuss the theory, we use to just
configure the feed forward controller taking one two examples.
(Refer Slide Time: 48:39)

Before going to discuss the or before going to analyze this feed forward controller, we
just use to configure the feed forward controller with taking two examples. So, first we
will take one heat exchanger, this is a heat exchanger this is process stream and this is
the output of the process stream. Here one heating medium is used, that is steam for
maintaining the temperature of the process stream fine. Now for the feedback controller
the controlled variable is usually measured, but for the feed forward controller the
disturbance is measured. For feedback control scheme the controlled variable is
measured, it is very obvious if you see the equation of any feedback controller like P or
PID fine, and feed forward controller the disturbance is measured, fine.
Now, for this heat exchanger our control objective is to maintain the temperature of the
process stream. Our control objective is to maintain the temperature at its set point, what
is the manipulated variable; manipulated variable is steam flow rate. This is the
manipulated variable, what are the disturbances? One is process stream flow rate another
one process stream temperature T i. So, flow rate and temperature T i; these two are the
disturbance variables.
So, we can use one measuring device for the flow rate measurement and another
measuring device for the temperature T i. So, this is for F and this is for T i, then one
feed forward controller block can be included and then this control action is implemented
through this control valve fine. So, two disturbances are involved in this process and they
are measured, then that measure signal go to this feed forward control block and then the
control action is calculated and that is implemented through this final control element.
So, this is the feed forward control scheme for this heat exchanger.

(Refer Slide Time: 54:03)

Another example we can take that is a distillation column. One control valve is employed
for reflux rate and another valve is employed for heating medium fine. So, this is
distillate, which has the flow rate of D and composition of x d that is mole fraction
basically, and in the bottom loop the flow rate is suppose B and composition is X b. Feed
is introduced, here with feed flow rate is F and composition is Z. So, what is the control
objective, the control objective is to maintain X d at X d set point. Similarly X b at X b
set point. Control objective is to maintain the product quality at the top as well as the
bottom.
So, what are disturbances involved here, one is feed flow rate another one is
composition. So, we can use flow measuring device FT and composition measuring
device CT. Then one block is included for feed forward controller fine. For this top loop
X D the manipulated variable is reflux flow rate R, and for the bottom loop the
manipulated flow rate is manipulated variable is Q R, Q R is basically the revolver duty.
Then this is implemented for Q R and another one is for R fine. So, control objective is
to maintain X D at X D set point X B at X B set point disturbances are F and Z and
manipulated variables are R for X D and Q R for X B. So, this is the feed forward control
scheme for the distillation scheme. In the next class, we will discuss the theory involved
with this feed forward controller.
Thank you.

Process Control and Instrumentation


Prof. A. K. Jana
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture - 34
Advanced Control Schemes (Contd.)

In the last class, we discussed about the configuration of feed forward control scheme.
And today, we will discuss the design of feed forward control schemes.
(Refer Slide Time: 01:08)

So, first we will discuss a generalized form of feed forward controller fine. So, for this
we first draw the process. So, there are 2 transfer; functions one is G p, another one is G
d d is the disturbance and n is the input to the process output is y fine. So, this is block
diagram of the open loop process this is the block diagram of the open loop process.
Now, we can represent this process mathematically as output y bar s which is equal to G
p m bar s plus G d d bar s. This is the representation of the open loop process shown here
fine. Now, our target is to keep with the y at its set point value our target is to keep y at it
is set point valued this is our target fine. So, accordingly the feedback controller sorry
feed forward controller is require to take the control action.
Feed forward controller needs to take action to keep y at its desired value that is y set
point. So, if we substitute y equal to y set point in this equation then we obtain y set point

s equal to G p s m bar s plus G d s d bar s. This expression we obtain just substituting y


equal to y set point in this open loop expression Now, if we rearrange this equation, we
obtain m bar s which is equal to one by G d s y set point bar s minus d bar s multiplied
by G d s divided by G p s. If we rearrange this equation, we get this expression in the
form of in which is the input to the process. Now, we can write this expression as.
(Refer Slide Time: 05:36)

We can write this expression as m bar s equal to G set point s y set point bar s minus d
bar s multiplied by G c s. G c is the transfer function of the controller and G s p is the
transfer function of the set point element. G suffix s p is the transfer function of the set
point element and G c is the transfer function of the controller. Now, if we compare the
last 2 equations if we compare the last 2 equations, we obtain G s p equal to 1 by G d s.
If we compare the last 2 equations, we obtain G s p equals to 1 divided by G d s; this is
transfer function of the set point element.
Similarly, we obtain G c s as G d s divided by G p s if we compare the last 2 equations
we obtain similarly, the expression for controller. This is the transfer function of the feed
forward controller feed forward controller fine. So, these 2 transfer functions, we obtain
in the next we will develop the block diagram of these feed forward control scheme. We
will develop the block diagram for the feed forward control scheme.

(Refer Slide Time: 08:14)

First we draw the open loop process; the open loop process includes 2 transfer functions;
one is G p, another one is G d. So, this is a open loop transfer fine next we use to include
the feed forward control scheme. With this open loop transfer function, you see the
expression this is a expression of the feed forward control scheme, this is a comparator
this is d see here minus d is involved. So, we can put here minus sign next G s p y s p.
So, we can put here set point element and input to this set point element is y set point.
This is positive fine the output of this comparator is suppose error and this is a error G
set point y set point minus d. That is the error; this error is this is suppose the error
epsilon this error is multiplied with G c. So, we can put here one block for G c fine
output is m this is output.
So, first this disturbance is measured and this disturbance is subtracted from G s p
multiplied by y s p. So, we can consider G s p multiplied by y s p minus d as error signal.
So, this is basically the error then this error is multiplied with G c therefore, G c is placed
here and this output signal of this controller is m. So, this is the block diagram of the
feed forward control scheme fine Now, in this block diagram you see the final control
element and measuring device are not included the final control and measuring device
are not included if we use to include that then we obtain the close loop block diagram
somewhat like this; this is disturbance. So, first we measure this disturbance using one
measuring device which has transfer function of G m.

Then this measure disturbance is compared in the comparator; this is a set point element
then here controller here final control element fine. So, this is the block diagram for the
feed forward loop block diagram; for the feed forward loop fine we have just included
here one measuring device and final control element. Now, we will note down few
important points on this feed forward control scheme say if we see the equation of the
feed forward control scheme this one. It is obvious, that the feed forward controller is not
like P or PI or PID controller. If you compare this equation with any expression of P or
PI or PID it is obvious that the feed forward controller is not like P PI or PID controller.
(Refer Slide Time: 14:39)

This is the first important point from the equation of feed forward controller it is
obvious, that it is not like P PI or PID controller. It is a special purpose computing
machine, it can be viewed as a special purpose computing machine, it can be viewed as a
special purpose computing machine. Second important point is if you see this equation it
is obvious, that the feed forward controller heavily depends on the process model you see
in this equation G d G p transfer functions are involved. That means the feed forward
controller depends heavily on the process model. It depends on the process model fine it
is impossible to develop the accurate model in practice practically it is not possible to
develop the accurate model. So, we cannot achieve perfect performance from this feed
from this feed forward controller practically it is not possible to develop the accurate
model. And we can consider this is a drawback of feed forward controller fine. Next, we

will take one example to discuss to derive the design equation for this feed forward
controller.
(Refer Slide Time: 17:01)

We will consider an example of this feed forward control scheme. So, we will take if
hitting tank system which we discuss earlier F i is the input flow rate, T i is the
temperature, F is the outlet flow rate. And T is the temperature liquid height in the tank is
h and T is the temperature. Now, the heating medium is passed through this heating coil
suppose steam is the heating medium. And it has the flow rate of Q I mean it carries Q
amount of heat fine here F F i both are volumetric flow rate volumetric. Flow rate h is
the height of liquid in the tank a is the cross-sectional area of the tank a is the crosssectional area of the tank h is the liquid height in the tank fine.
Now, for this process we developed earlier 2 equations one is based on mass balance
another one is based on energy balance. So, we obtain the mass balance equation as a d h
d T equal to F i minus F this is a mass balance equation. Similarly, we obtain the energy
balance equation as a h d T d T equal to F i T i minus T plus Q divided by rho c p. This
is a energy balance equation; this is a mass balance equation to discuss a feed forward
control scheme. We assume that F i and F both are equal and they are constant here. We
are assuming that F i equal to F and that is constant quantity, accordingly from the mass
balance.

(Refer Slide Time: 20:54)

We obtain d h d T equal to 0; that means, h is again a constant quantity. So, what is our
control objective?
(Refer Slide Time: 21:23)

Our control objective is to maintain the temperature. So, temperature is our control
variable what is the corresponding manipulated variable Q we have assume that F i is a
constant quantity. Therefore, T i is the lowered variable therefore, T i is the lowered
variable. So, next we will derive the design equation for the feed forward control scheme
next we will derive the design equation for the feed forward control scheme. So, we will

start from a energy balance equation which is a h d capital T d small T equal to F i T i


minus T plus Q divided by rho c p. This is the energy balance equation, we have written
earlier.
Now, we can write this equations as v divided by F i we are just dividing both sides by F
i. We are just dividing both sides by F i then we obtain v divided by F i d T d T plus T
equal to T i plus Q divided by F i rho c p here we assume v equal to A h fine. So, by
dividing both sides by F i and rearranging, we obtain this equation. In the next step, we
can write this equation in terms of deviation variables as v divided by F i d T prime d T
plus T prime equal to T i prime plus Q prime divided by F i rho c p. This equation is
written in terms of deviation variables fine here 3 deviation variable are involved; one is
T T i and third one is Q.
(Refer Slide Time: 24:48)

Now, if we take the Laplace transform, we obtain T bar prime s T i bar prime s divided
by tau s plus 1 plus 1 divided by F i rho c p 1 divided by tau s plus 1 Q bar prime s where
tau is equal to v divided by F i. This is the retention time of liquid in the tank retention
time of liquid in the tank, if we take the Laplace transform of this equation. And after
rearranging, we obtain this expression where tau is the retention time of liquid in the
tank. As mention earlier the objective is to keep the temperature T at its set point value
this is the objective.

Accordingly the feed forward controller computes the control action the objective is to
maintain the temperature T at its desired value that is T set point and accordingly the
feed forward controller takes action. So, substituting T s p in place of T in this equation
and rearranging finally, we obtain Q bar prime s equal to F i rho c p tau s plus 1 T set
point bar prime s multiplied by T i bar prime s fine. Substituting T equal to T s p in this
expression and rearranging finally, we obtain this expression the expression of Q fine.
So, this is the design equation of the feed forward control scheme; this is the design
equation of the feed forward control scheme. Now, can we configure the feed forward
control scheme, considering this equation for the heating tank system? In the next, we
will try to configure the feed forward control scheme.
(Refer Slide Time: 28:27)

Incorporating this expression I mean applying this expression. So, this is the example
system fine in the feed forward control scheme the disturbance is measured here
disturbance is T i. So, we can measure this T i using one temperature measuring device
fine. Then this T i is compared with the set point value fine you see T s plus 1 this 1
minus T i this is T i. Then the output of this error signal is multiplied with F i rho c p. So,
this is F i rho c p and that is basically Q prime tau s plus 1 multiplied by T s p that is this
1 minus T i minus T i.
So, this part we can consider as error. So, this is error then this error is multiplied with F
i rho c p this is F i rho c p this is F i rho c p and output is Q. And that is written here in

terms of deviation variable agree. So, this is a feed forward control scheme for the
heating tank system this is a feed forward control scheme for the tank heater fine. Now,
if see this feed forward control scheme, we have develop considering the dynamic
equation dynamic energy balance equation. Therefore, this feed forward controller is
sometimes called as dynamic state feed forward controller. Since, we have used the
dynamic energy balance equation for deriving the feed forward controller. Therefore, this
is sometimes called as dynamic state feed forward controller fine.
(Refer Slide Time: 32:25)

Similarly, we can derive the steady-state feed forward controller. Similarly, we can
derive the steady-state feed forward controller. And for steady-state feed forward
controller, we write the energy balance as this we have just considered a h d T d T equal
to 0 left hand side of the energy balance equation. We consider zero then we obtain this
steady-state equation fine. Now, if we rearrange it we obtain Q equal to F i rho c p
multiplied by T minus T i if we rearrange this steady-state form of energy balance
equation we obtain this fine. Now, this is the design equation sorry we just have to put T
equal to T set point. Then finally, we obtain this and this is the design equation for the
steady-state feed forward controller. This is the design equation for the steady-state feed
forward controller.
Now, if you compare the dynamic state feed forward controller and steady-state feed
forward controller. We see that only this term does not exist for steady-state feed forward

controller this is equal to 0 fine. We can get the steady-state feed forward controller just
substituting tau is equal to 0 in the dynamic state feed forward controller. So, we have
configure the dynamic state feed forward controller with the sitting tank system shown
here for steady-state feed forward controller. You will just write inside this block one for
steady-state feed forward controller; you replace tau s plus 1 by 1 fine. So, that is the
steady-state feed forward controller.
(Refer Slide Time: 35:38)

In the next, we will discuss another advanced control scheme that is ratio control. We
will discuss another advance control scheme that is the ratio control scheme. This ratio
control scheme is a special type of feed forward controller. Ratio controller is a special
type of feed forward controller its objective is to maintain the ratio of 2 variables.
Therefore, the n m its objective is to maintain the ratio of 2 process variables the ratio of
2 process variables at a specified value. Its objective is to maintain the ratio of 2 process
variables at its specified value and these 2 process variables are usually flow rates. So,
the objective of the ratio control scheme is to maintain the ratio of 2 flow rates fine both
flow rates are measured.
But only one is controlled both flow rates are measured, but only one can be controlled,
only one can be controlled fine now which is controlled that is called manipulated
variable. So, this is manipulated variable and another flow rate is not controlled that is
basically disturbance variable the variable which is not controlled that is basically

disturbance variable. So, the objective of the ratio controller is to maintain the ratio of 2
flow rates, both the flow rates are measured. But only the manipulated input is controlled
the flow rate which is not under controlled is the disturbance variable. And since this is
not controlled it is called as wild stream fine since the manipulated variable is not under
controlled it is called as wild stream.
(Refer Slide Time: 39:22)

And the ratio is the ratio denoted by R is equal to u divided by d u is the flow rate of the
manipulated input flow rate of a manipulated variable and d is a flow rate of disturbance
variable fine. Since the flow rate of manipulated variable and flow rate of disturbance
variable the both are measured. So, sometimes they are written as u measured divided by
d measured and accordingly the ratio is denoted by R suffix m since the manipulated
input. And the disturbance variable both are measured quantity. Therefore, they are
sometimes denoted by u suffix m and d suffix m. Accordingly the ratio R is denoted by
R suffix m and another important thing is that both u and d are physical variable not
deviation variables u and d. Both are physical variables not deviation variables fine. In
the next, we will try to develop the ratio control configuration, in the next, we will
develop we will discuss the ratio control configurations.

(Refer Slide Time: 41:45)

So, this is one stream say this is the disturbance stream fine and this is another stream
which we can call as manipulated stream u and this is d now it is mention that both the
flow rates are measured. So, we can employ one flow measuring device that is suppose F
T one for measuring the flow rate of disturbance stream. Similarly, we can measure this
manipulated variable using another flow measure in defines that is say F T 2. So, this is
said d suffix m and this is suppose u suffix m it is mention that both the flow rates are
measured accordingly. We can employ one flow measuring device for disturbance
stream another flow measuring device for manipulated input. Now, in the next it is
required to compute the ratio.
So, this is representing the divider this is the divider which is computing the ratio u m by
d m in this divider r m is calculated using this expression u m by d m. So, the output of
this divider is basically r suffix m measured ratio fine this measured ratio is compared
with the set point value of reflux ratio R R s p is the set point value of ratio. So, the
output of this comparator is basically the error signal output of the comparator is error
represented by epsilon. And this output goes to the ratio controller G R c is the transfer
function of the ratio controller. Then the output of this ratio controller is implemented
through the control bulb fine. It is mention that both the flow rates are measured
accordingly F T one can be employed to measured disturbance variable F T 2 can be
employed to measure this manipulated variable.

Then this is the divider element in which the reflux the ratio r m is computed considering
this expression u m by d m. So, output of this divider is r m which is compared by the set
point value of this ratio r s p. Then this error signal goes to the controller G R c is the
transfer function of the ratio controller and the controller action is use to control this
manipulated input. So, this is a ratio controllers this is a ratio controller fine. So, I am
repeating again the steps both the flow rates d and u are measured. Then compute the
measured ratio compute the measured ratio r m which is equal to u m divided by d m
fine. Now, the output of the divider element is sent to the ratio controller.
(Refer Slide Time: 47:58)

Output of the divider element is sent to a ratio controller fine. In this ratio controller first
the error is calculated, first the error is calculated by comparing R m with R s p fine and
then the ratio controller takes action. Now, the ratio controller is takes typically a PI
controller ratio controller is typically a PI controller fine. Now, at the beginning I have
mention that the ratio controller is a special type of feed forward controller. Now, we are
saying that the ratio controller is a PI controller which is feedback controller. So, what is
the justification behind this? Actually in this ratio control scheme the disturbance is
measured. Therefore, the ratio control is in essence is a special type of feed forward
controller fine. In this ratio control scheme the disturbance is measured therefore, the
ratio control is in essence is a special type of feed forward controller.

(Refer Slide Time: 50:23)

Next, we will discuss the application of this ratio controller. Can you give few examples
where we can employ this ratio controller where basically the ratio is involved? Like one
example is reflux ratio in this relation. So, to control the reflux ratio in distillation
columns the ratio controller is used to maintain the reflux ratio in distillation columns.
The ratio controller is used to control the ratio up to blended streams maintain the ratio in
a blending process. The ratio controller can be employed to maintain the stoichiometric
ratio of reactance in the reactor or reactive distillation. Maintain the ratio of
stoichiometric to maintain the stoichiomatric ratio of reactance to a reactor or reactive
distillation.
The ratio controller can be used if any reactant is in access that can come out as the
product in the product and that degrades the product quality. Therefore, it is require to
maintain the ratio stiochiomatric ratio of reactants which are fate to the reactor or
reactive distillation to maintain the fuel air ratio to a furnace. The ratio controller can be
employed to maintain the fuel air ratio to a furnace the ratio controller can be employed.
So, these are the applications of ratio controller and ratio controller is extensively used in
chemical engineering processes. So, today we discussed the derivation of design
equation for the feed forward controller scheme. We discuss the feed forward controller
with taking one example that is tank heater. And finally, we discussed a special type of
feed forward controller that is ratio controller. Thank you.

Process Control and Instrumentation


Prof. D. Sarkar
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture - 35
Instrumentation: General Principles of Measurement Systems
In this module of process control and instrumentation, we will now learned about
instrumentation specifically we will discuss about measuring instruments core sensors
that at used in process industry to measure common process variables such as
temperature pressure flow rate of a fluid or liquid level. Science of measurement is very
old and well advanced now. So, it is for us not possible to teach you every details of call
the instruments that are available and for as it is not required also, but the purpose of this
course or purpose of the module of this course is to give you a basic introduction to the
instruments, that are used in industry for measurements of common process variables for
process monitoring process control and safe profitable operation.
(Refer slide Time: 01:27)

So, I will talk about the instrumentation part of process control and instrumentation we
are now familiar with this block diagram which is a feedback control block diagram for
feedback control system process control has been the topic of discussion so far. So, we
can recognize very well all the blocks of this block diagram.

What we intend to do here is we have a process which is influence by some disturbance


which we cannot manipulate the process is also influenced by flow of energy and our
material which we can manipulate we want to control the process meaning we define a
desired state for the process and the control variable has to be kept at that desired state.
So, how does this feedback control work the first thing that is done is we need to
measure the control variable and we do that using a sensor or measuring element. So, the
sensor or measuring element measures information about the controlled variable and the
information goes to controller where this value is compare with the desired value of the
control variable that is set point and then here error signal generated.
Depending on this error signal the controller decides what action to be taken. So, that this
control variable is the went to the desired value of control variable. So, controller take
thus decision and directs final control element to manipulate this input energy and
material which we can manipulate such that control variable is driven to the desired state.
You can write such block diagram for any control system such that you can think of the
control of temperature in this water bath by manipulating the flow rate of steam. So, you
measure the bath temperature then information goes to the controller and the controller
decides what should be the opening of the control valve.
Similarly, you can also talk about the level control in this tank very use a level indicator
which is nothing, but level measuring instrument which measure the level of the liquid
determines whether these are desired state or not then information goes to controller the
controller decides what is to be done and accordingly changes the opening of this valve.
So, you see that every control every feedback control system will have at least one
measuring sensor or measuring instrument feedback system is widely use for control of
temperature pressures flow rates liquid levels in process industry. So, it is a at most
important that we learned about this sensor or the measuring elements.

(Refer slide Time: 05:20)

So, this is the objective of the module we we intend to learn the fundamentals of
industrial instrumentation as we as I told you in the beginning that it is for as not possible
to learn every details of all the instruments that are available, because there are numerous
instruments for measurement of various process variables. So, what will do is we will
learn general principles of measurement systems this will be done without any reference
to in a specific hardware then youll talk about transducer elements.
Transducer elements are those elements which changes the energy from one from to
another let us say a pressure signal is converted to an electrical signal a displacement
signal is converted to an electrical signal. So, transducer elements are useful in process
instrumentation although by transducer elements are those elements which converts the
energy from one form to another typically by transducer we mean those elements which
converts the energy in one form to an electrical signal, Then specialize measurement
systems that will be the body of this course will learn about various pressure measuring
elements various temperature measuring elements various flow measuring instruments
and various level measuring instruments and finally, the control valve and
instrumentation diagram.

(Refer slide Time: 07:04)

So, towards the end of this course we will be able to answer the following questions what
are sensors and transducers what are the building blocks of an instrument; that means,
can I break down an instrument in to various building blocks and describe the operation
of an instrument what instrument characteristics affect the quality of measurement. What
do we get an how do we an estimate of error in our measurement since every
measurement is in error, we should be able to get an estimate of how much of error is
there in my measurement how do we measure temperature how do we measure pressure
how do we measure very low pressure or high vacuum how do we measure flow rate of a
fluid say in a process flow rate of a process fluid in a pipe, how do we measure liquid
level let us say how do we measure liquid level in the storage tank how do we interpret
an instrumentation diagram, what what are the what are the meaning of the symbols that
are present in an instrumentation diagram.

(Refer slide Time: 08:29)

So, towards the end of this course we should be able to answer these questions here is the
details of model module contents is divided into four parts in part a we talk about general
principles of measurement systems we will start with an introduction to measurement
systems and today will discuss that we will talk about what do you mean by
measurement why should you measure what is the purpose of measurement what is
direct measurement what is indirect measurement what are the various types of
application of measurement then will talk about functional elements of instrument this is
an important concept which which helps us to describe the operation of instrument in a
very generalized manner. So, we will try to break down an instrument into various
functional elements and try to see how the instrument works
Then classification of instruments then instruments can be classified in various ways. So,
we will see some of the some of the ways how we classify the instruments we will talk
about input output configurations of instruments then will go to performance
characteristics of instruments what are the various performance characteristics that affect
the quality of measurement broadly we talk about two different types of characteristics
one is call static characteristics another is dynamic characteristics as the name sagest
static characteristics a those characteristics which we must consider on the instrument is
being use to measure the value of a condition which is not changing with tile and
dynamic characteristics such the name sagest a those characteristics which we must

consider when the instrument is being used to measure a condition which is changing
with time.
Them will aim part a with brief discussion on error analysis since every measurements
are in error we should have an estimate of error in our measurements in third b we will
talk about transducer elements we just discuss the transducers transducer elements are
those elements which converts energy in one form to another and practically we mean
those elements which converts energy in one form to electrical energy
Well talk about four different types of transducer elements linear variable differential
transformer or l v d t which is the displacement displacement transducer it converts
displacement signal one electrical signal then resistance strain gauges resistance strain
gauge if strained it resistance changes. So, it converts the strain to an electrical signal
capacitive type displacement capacity type transducer which is again a displacement type
transducer and piezoelectric transducers which is like if a crystalline material like quartz
is distorted a charge is produced. So, we will talk about piezoelectric transducers in
detail also
(Refer Slide Time: 12:29)

Part c will be the main body of this module here we will talk about various process
variable measuring instruments in detail we start with various pressure measuring
instruments we will see how we classify the various pressure measuring instruments

what are the various pressure measuring instruments available what are the are working
principle what are the ranges of this instruments and so on and so for.
Similarly, we will treat various temperature measuring instruments various flow
measuring instruments and various level measuring instruments mainly we will focus on
liquid level measuring instruments. Then in part d we will talk about control valve and
instrumentation diagram the contains will be the construction and working principle of a
control valve how does a control valve work then we will see what do you mean by
instrumentation diagram what is it what are the symbols used in an instrumentation
diagram how do we interpret an instrumentation diagram.
(Refer Slide Time: 13:57)

This is the list of text books or reference books number one measurement system
application and design by dulblein number two process control instrumentation
technology by johnson number three principles of industrial instrumentation by
patranabis number four industrial instrumentation by eckman, and finally process
industrial instruments and controls handbook by considine.

(Refer Slide Time: 14:57)

So, let us start part a with a definition of measurement measurement is an essential


activity in every branch of science and technology measurement means quantification of
a parameter or quantification of a quantity or condition a measuring instrument is thus a
device the determines the value of a quantity or condition an instrument converts a
physical or chemical plant condition for example, heat pressure or light to a signal which
is often electrical, but not necessarily. So, that can be measured or interpreted
So, simple instrument model may be as follows we have a process or measured medium
let us say you are interested in measuring the temperature of a liquid in a tank. So, this is
the process or measured medium for you then I take a temperature measuring instrument
this temperature measuring instrument interacts with this process, and produces an
output which is in some sense related to this input and this output should be in a form
which can which can be interpreted by the observer.

(Refer Slide Time: 17:00)

So, why should we measure what is the purpose of measurement lord kelvin said when
you can measure, but you are speaking about and express it in numbers you know
something about it, but when you cannot measure it when you cannot express it in
numbers you knowledge is of a meager and unsatisfactory kind within court and un court
that defines the purpose of instrument the purpose of measurement very nicely in a
process industry if you visit even you work for a process industry you for a sell
frequently us or we will want to know what is the temperature in reactor what is the
temperature in furnace or in the reboiler what is the flow rate of a process fluid in a pipe
or what is the level of liquid in the storage tank or what is the consultation of a product in
the product steam.
So, to go to an answer to these questions we definitely need to take help of various
instruments. So, the fundamental purpose of measurements in industrial manufacturing
and processing is to obtain a numerical value because generally we are interested in
quantification corresponding to the variable being measured. So, that we can determine
and improve the quality of a product or the efficiency of production. The fundamental
purpose of measurements in industrial manufacturing and processing is to obtain a
numerical value corresponding to the variable being measured. So, that you can
determine and improve the quality of a product or the efficiency of the production by
efficiency of production you mean the process operations should be profitable and safe
for human environment as well as equipment.

(Refer Slide Time: 19:26).

.
Now, let us come to types of measurement applications the measurements can be done
for various reasons we have various process variables to measure we have various
instruments to measure those variables. So, there can be various types of measurements
there can be various types of measurement applications fortunately the application of any
measurement can usually be put into any of the following three categories a monitoring
of process operations b control of processes and operations and three experimental
engineering analysis.
What do you mean by monitoring of process and operations by monitoring of process
and operations we mean the instruments simply measures and display what it is measure
usually control action is not taken. So, it basically keeps track of some quantity or
condition for example, let us talk about weather monitoring thermometers barometers
radars simply indicate weather condition it does not take any control action.
So, thermometers barometers radars they all indicate the weather condition and weather
condition can be monitor that way a homely example is a water meter or electric meter
which keeps track of how much of water or how much of electricity has been used. So,
monitoring of processes and operations where the instruments simply measures and
display what it has measure it is essentially for keeping track of some quantity or
condition and usually no control action is taken a homely example is water meter or
electric meter which tells us how much of water or electricity is used.

Two control of processes and operations perhaps is the most important applications of
measurements and we basically refer to automatic control we are briefly talked about a
feedback control system in the beginning of todays discussion and also it has been the
topic of this discussion. So far Say you know that every feedback control system has at
least one measuring instrument involve and this is this is definitely one of the most
important application of measurements.
And number three experimental engineering analysis, we can measure variables or some
quantity or condition to test the validity of predictions from theories we can perform
measurements to develop model from raw data that you collect using instruments we can
also perform carryout measurements for characterization of materials devices etcetera.
So, application of any measurement can generally we put into one of these three
categories process monitoring and operations where instruments simply measures and
display no control action taken number two control of processes and operations well
control action is actually taken and the control in a feedback control system control
action has to be taken on the basis of the measurement of the control variable and
number three is for experimental engineering analysis when you use it for testing the
validity of predictions from theories or you build up a model using the data that you
collect using instruments.
(Refer Slide Time: 24:41)

The measurements can broadly classified into two categories direct measurement and
indirect measurement direct measurements what do you mean by direct measurements in
direct measurements we compare directly the parameter that we intend to measure with
an accepted standard for example, measuring the length of a wooden block by a scale if i
have an accepted standard accepted standard I can directly measure the length of wooden
block using that accepted standard for example, you measure the length of this class
room using a tape which can be consider as an accepted standard.
Indirect measurement as the name sagest we measure a parameter by measuring another
parameter which is more convenient to measure. So, essentially we infer here we do not
measure directly the parameter that we intend to measure, but you measure some other
parameter which is more convenient to measure and infer the value of the other
parameter
Let us take an example to clarify this i have task at an find out how many bacteria are
there in a tube. So, we have a tube and there are some bacteria in the tube i want to find
out how many bacteria are present in the tube what can be the direct measurement here a
direct measurement can be you spread the bacteria out on microscope slide and directly
count under the microscope.
So, you take a microscope slide spread out the bacteria put it under microscope and
count it that will that will of course, be a direct measurement, but you understand is
going to be very inconvenient and cumbersome. So, what can be indirect measurement
use a spectrophotometer use an instrument call spectrophotometer a spectrophotometer
works by signing light in one side of a tube and measuring how much of light passes
through a passes through to the other side if more light passes it means there is less
bacteria in the tube.
Say you can bac calcular and can measure how many bacteria are present in the test tube
there can be another way of measuring the bacteria count indirectly we give some
substrate or food to the bacteria to it engrow. So, the bacteria will consume the substrate
and grow unless measure this consumption rate higher the consumption rate higher the
bacteria count. So, this will be another way of measuring the bacteria count indirectly of
course, the direct measurement always preferred, but as I notice here there it is not

always possible to go for direct measurement it may not always be possible or even if it
is possible it may be very inconvenient.
(Refer Slide Time: 29:16)

Well the instrument measures the value of a condition the following functions may be
performed by an instrument transmitting signaling registering indicating and recording
accordingly, you can say transmitting instrument signaling type of instrument registering
instrument indicating type instrument and recording instrument.
What do you mean by transmitting or transmitting type of instrument here the instrument
conveys information concerning the measured quantity over some distance to a remote
point. So, the instrument will measure the quantity and convey the information from one
place to another a homely example is telephone signaling type of instruments here the
instrument indicates the general value or range of values of its measured quantity some
grosses scale indicates the general value of what it has measured.
Registering type of instruments the instruments merely indicates by numbers or some
other symbols of discrete increments the value of some quantity for example, a water
meter which uses discrete numbers to indicate how much of water has been consume
indicating type of instruments the instrument indicates, the value of the measured
quantity using a calibrated pointer and scale you have seeing this type of instruments in
laboratory say commonly use for pressure gauge which uses a pointer n scale to indicate
the value of the measured quantity which is pressure in case of pressure gauge.

And recording type the instrument keeps a written record of measured quantity and
usually it is recorded against time. So, all these functions transmitting signaling
registering indicating recording all the functions may be found in a instrument in any
combination.
(Refer Slide Time: 32:22)

Now, let us come to an important concept called functional elements is a very useful
concept to describe the operation of an instrument in a generalized way an instrument
consist of several elements these elements perform a prescribed functions in converting a
quantity or condition which is measuring medium into a corresponding indication which
we called measured value.

(Refer Slide Time: 33:33)

If you examine various physical instruments with a view to a generalization we may be a


able to identify a set of elements that are similar with regard to their function we call
these elements functional elements. So, what you mean is we have various instruments
let us say we have various temperature measuring instruments we have various pressure
measuring instruments.
Now, if we examine all these instruments and several other instruments very closely with
a view towards generalization we may be able to identify a set of elements that are
similar with regard to their function what you mean is if I take let us say this is a
temperature measuring instrument and this is the pressure measuring instruments perhaps
i may able to see an element here and similar element here which does similar job.
So, in terms of function they are similar we call these elements functional elements. So,
we will define functional elements as if we examine various physical instruments with a
view toward generalization, we may be able to identify a set of elements that are similar
with regard to their function we call these elements as functional elements.
An instrument thus can be broken down to a limited number of functional elements and
this gives us a way to treat the operation of an instrument in a generalized way, but the
question now is how do we break down an instrument into its building blocks or various
functional elements it can be done in a number of ways and looks like there is nothing

like universally accepted generalized scheme. So, there are different schemes propose in
different looks and youll see here at least two of them.
(Refer Slide Time: 35:56)

Let us look at this scheme which describes how you break down an instrument into
various functional elements you have a process or measured medium which i am
interested in measuring having a instrument in contact with this the instrument interact
with this measure medium and presents may what it has measured.
Now, if I examine the instrument how it works or if I if I able to breakdown instruments
into various parts which performs different functions I can identify these elements the
first element is sensing element sensing element the sensing element interacts with the
process first it produces an output which is related to process variable being measured.
So, this is the element which first interacts with the measured medium a receive signal
from the measured medium which we call measuring and produces a signal which is
related to the process variable that is being measured.
And example can be a thermocouple you know which a thermocouple measure
temperature a thermocouple has two junctions one is call measuring junction another is
call cold junction we will talk about thermometer in detail later. So, the hot junction or
the measuring junction is brought in contact to the medium whose temperature i am
going to measure. So, depending on the temperature or the temperature difference that

exist between the measuring medium and the cold junction thermocouple produces an e
m f as output.
So, thermocouple is produces e m f which is in the range of millivolt depends on
temperature. So, the output signals of the sensing element is related to the process
variable that is being measured or thermocouple is temperature here is millivolt here
similarly strain gauge where it resistance depends on the mechanical strain. So, if a
resistance strain gauge is strain differently is resistance will change differently.
The next element in line is signal conditioning element the signal conditioning element
will convert the output signal of the sensing element to a form or to a signal which is
more suitable for further processing. So, it conditions the signal that comes out of
sensing element So, it converts the output signal of the sensing element to another signal
which is more suitable for further processing it may be a d c voltage or current it is
require when output signal of the sensing element is not in the form which is more
suitable which is suitable for further processing say the output of the primary signal is a
displacement signal anyone may be interested in converting into a electrical signal which
can be processed much more easily. So, you need a signal conditioning element which
will converts the displacement signal when electrical signal.
Examples may be amplifier which will converts millivolts to volts and oscillator which
converts an impedance change into a variable frequency voltage the output of the signal
conditioning element goes to signal processing element a signal processing element
converts the output signal from the signal conditioning element to a form which is more
suitable for presentation for presentation to the observer it may be a analog to digital
converter.
An analog to an analog to digital converter will convert voltage signal into a digital form
for input to a computer it can also be a micro computer which calculates measured value
of the of the variable from incoming digital data and finally, we have data presentation
element the data presentation element presents the data to the observer and it has to be in
a form which the observer can recognize. So, the data presentation element presents data
to the observer in a recognizable form examples are pointer and scale a chart recorder or
a visual display unit.

So, in this thing we identify four functional elements firs is sensing element which firs
receives information about the measured medium and produces an output which is
somewhere related to the variable we are going to measure second in the line is signal
conditioning element which will changes the output signal of the sensing element to a
form which is more suitable for processing next is signal processing element which will
convert the output signal of the signal conditioning element to a signal which is more
suitable for the purpose of presentation and the data presentation element will present
this signal to the observer in a recognizable form.
(Refer Slide time: 44:02)

Let us discuss another scheme as proposed by doebelin we have a process or measure


medium as before again consider it is liquid whose temperature I am going to measure.
So, I bring a temperature measurement measuring instrument in contact with this
measure medium and analyze how the instrument works I mean able to identify these
functional elements.
The firs lock is primary sensing element the primary sensing element the primary sensing
element is that which first receives energy from the measured medium and produces an
output depending in some way on the measured quantity. So, this is same as what you
saw in the scheme one we called sensing element there that is called primary sensing
element here because you can have if you have more than if you have more than two

sensing elements in the one interacts with the system first may be called primary sensing
element.
So, this is the element which first receives information from the measured media and and
produces an output depending on some way on the measured quantity if you if you take
the temperature measurement as an example the primary sensing element will first
extract some amount of thermal energy from the measuring medium. So, immediately we
know that we are disturbing the system that we are going to measure because it is it is
extracting some amount of thermal energy from the system and that is necessary for the
act of measurement which makes perfect measurements extremely difficult or
theoretically impossible this is call loading effect the amount of energy that is necessary
for the purpose of measurement is called loading effect. So, will good instrument will
receive very minimum energy from the measuring media for the purpose of
measurement. So, good instrument will have minimum loading effect.
So, the next is variable conversion element the output signal of the primary sensing
element is some physical variable such as displacement or voltage say if you are using
thermocouple as a temperature measuring instrument the output signal of the primary
sensing element will be a millivolt signal for the instrument to perform the desired
function it may be necessary to convert this variable into another more suitable variable
while preserving the information content of the original signal and variable conversion
element will do this function.
So, the variable conversion element will convert the physical nature of the signal that
comes out of the primary sensing element to another signal which is more suitable for the
purpose of measurement. It may be noted here that all instruments may not have variable
conversion element some instrument may have more than one variable conversion
element next in the line is variable manipulation element in performing the measurement
and instrument may require the signal represented by some physical variable the
manipulated some way that is we may be interested in changing the numerical value of
the signal while preserving the physical nature of the variable for example, an electronic
amplifier excepts a small voltage signal as input and produces an output signal that is
also voltage, but it some constant times the input.

So, variable manipulation element will manipulate the signal that comes out of variable
conversion element by manipulation we essentially mean changing the magnitude of the
signal unlike variable conversion element it will not change the physical signal of the
nature for example, an amplifier will change the mini hold signal to voltage signal. It
may also be noted here there is not necessary that variable manipulation element will
always follow the variable conversion element it may be present before this block or it
may present elsewhere in the diagram or in may not be present at all.
Next is data transmission element as the name sagest data transmission element transmits
the data when functional elements are actually physically separated then it becomes
necessary to transmit the data from one place to another and the data transmission
element will transmits the data from one place to another the data transmission element
may be as simple as a shaft and bearing assembling or it may be as complicated as a
telemetry system for transmitting signals from satellites to ground equipment by radio.
And finally, we have data presentation element the data presentation element
communicates information about the measured quantity to the observer for monitoring or
control purpose and as we have said earlier the data must be communicated such that it is
recognizable by human sense solvent the common examples are pointer moving on scale
is a data presentation element recording using a pain moving over a chair is another data
presentation element.
Finally we may also have a data storage or playback element which will stored data
which can be restore later whenever it is require examples may be a computer a magnetic
tape recorder etcetera. So, this block diagram tells us all possible functional elements that
you can present in an instrument.
I repeat again that it is not necessary the functional elements will be present strictly in
thus in this order in any instrument and also it is not necessary the there will be only one
variable conversion element or there will be one variable manipulation element. So, there
may be more than one variable conversion element or there will be there may be more
than one variable manipulation element at that may they may be present anywhere in the
chain, but we can describe the operation of any instrument in terms of these functional
element.

(Refer Slide Time: 55:00)

So, let us summarize what we did today define the module objective detail module
content then is started with general principles of measuring systems where we give a
brief discussion on why we measure purpose of measurement types of application of
measurement we talked about direct indirect measurement we stares that direct
measurement is always preferred, but we may not be possible always.
Then we talked about different functions that instrument perform, and finally we talked
about functional elements where we define what functional elements are and what are
various schemes available which will allow us break down an instrument in to various
functional elements in the next class we will take an example and try to identify the
functional elements that are present in that instrument.

Process Control and Instrumentation


Prof. D. Sarkar
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture - 36
General Principles of Measurement Systems (Contd.)
(Refer Slide Time: 00:36)

In our previous lecture, we studied our discussion on process instrumentation part of the
course, process control and instrumentation. Today, let us start with a brief review of
what we learnt in previous lecture. We started with general principles of measurement
systems. We gave a very brief introduction to measurement systems, we talked about
purpose about measurement, types of measurement applications, direct measurements
verses indirect measurement, functions of an instrument. Then we introduced a concept
call functional elements of an instrument, we defined what are the functional elements?
We also talked about two different schemes, which we can use to break down and
instrument into its various functional elements.

(Refer Slide Time: 01:34)

So, today will took an will take an example now to understood better how to break down
and instrument to its various functional elements. We talked about this scheme in our
previous lecture, this block diagram represents the all possible functional elements that
you can expect to see in an instrument, if you want to analyze it.
This is the first this is the process or measuring medium, we call this measuring, the first
functional element which is call primary sensing element receives inter interacts with the
measuring medium first and receives information about the measured quantity and
produces an output, which is somewhere related to the input to the primary sensing
element that comes from the measuring medium. Then we are variable conversion
element, which when convert the physical nature of the output signal from the primary
sensing element, if it is required for the purpose of measurement. So, this output signal
can be different from this input signal.
Then if it is necessary to magnify the magnitude of this signal we can have a variable
manipulation element next, then if the different parts of the instruments are separated by
distance; we need to transmit data from one point two another, so we have a data
transmission element which will transmit data from 1.2 another. Then finally, will have
an element which will present the data in a recognizable form to the observer so we have
a data presentation element.

Before we taken example and breakdown the instrument into is various functional
elements, let us stressed on the falling points. These represent a possible arrangement of
all the functional elements that is necessary to describe the operation of an instrument.
Remember, we are talking about functional elements not physical elements; in this block
diagram it is shown that this primary sensing element variable conversion element and
all this elements are separated into separate blocks. Instruments are in general not
separated into such blocks; instruments may involve the basic functions in any number
and any combinations.
So, let us take an example of pressure thermometer; so this is an instrument used for
measurement of temperature. You have a bulk here, then a tube is connected to the bulk
and the other end of the tube, goes to a pressure measuring element. What you see from
here is just the pointer in the scale; the bulk and the entire tube is field with a liquid call
filling liquid.
(Refer Slide Time: 05:29)

Now let us see, the part which we cannot see this is the pressure thermometer; and this is
what it will look like if you get to see what is inside this. You have a bulb here the same
bulb the bulb is connected to a tube caplet tube, to other end of the tube. A bourdon tube
is attached it may be a spiral bourdon tube or it may be a seaborne tube. Will see what is
bourdon tube is in few minutes, then the other end of the bourdon tube or the free end of

the bourdon tube is connected to a pointer and scale through some gear and linkage
mechanism.
(Refer Slide Time: 06:50)

So, before we before we analyze, how this is to met works let us see what a bourdon
tube; we take a tube with circular perception. Then pattern the tube to change the cross
section of the tube from circular to electrical. And then seal one another tube and finally,
then the tube in the form of a c, this is call bourdon tube name the Now, if I keep this
end of the bourdon tube reject fixed and apply the fluid pressure inside the bourdon tube;
the tip of the bourdon tube will deflect. This is bourdon tube tip now the tip deflection
depends on various factor such as the material of construction of the bourdon tube use a
metals with good elastic properties chosen. It depends on the angle through which it is
bend it depends on its cross section, but most importantly it depends on the amount of
pressure that is being apply inside the bourdon tube.
So, for a given bourdon tube the deflection of the tip can be directly related to the
pressure that is being applied inside the bourdon tube. So, for a given bourdon tube the
tip deflection proportional to the pressure being applied. So, if through some gear and
linkage mechanism; I can attach a pointer and scale the deflection of the pointer over this
scale can be directly take an as a measure of the bourdon tube bourdon tube pressure.

So, that is scale can be directly calibrated in terms of pressure units. So, the deflection of
the tip of the bourdon tube, which is magnified by this point can be considered as a
measure of the pressure.
Now let us come to a previous example to analyze how we can breakdown this pressure
thermometer into is various functional elements. So, this is the tip of this is the bulb of
the pressure thermometer. Next a capillary tube which is connected to a bourdon tube
which is not bent in the form of a see, but it is a spiral bourdon tube. The spiral bourdon
tube is chosen here because the tip deflection even further increases, if we consider, if we
take a spiral take bourdon tube inside instead of a c type bourdon tube.
So, we essentially increase the sensitivity of the bourdon tube by converting it from a c
type bourdon tube to a spiral type bourdon tube. Now this entire part of the bulb the
capillary tube and the inside of this spiral bourdon tube use field with a liquid called
filling liquid. In certain applications it can out we field with gas. Now this bulb is put
into the medium, which temperature I want to measure. So, bulb receives thermal energy
from the media and the filling liquid inside the bulb expanse. Now due to the strict
thermal expansion of the filling liquid of a series build up; and the pressure is transmitted
to the bourdon tube through this capillary tube.
Then if the same as applying pressure to a bourdon tube; so the tip of the bourdon tube
will deflect, and the deflection will depend on how much of pressure that has been
transmitted through this tube to the bourdon tube. So, this tip deflection will cause a
movement of this pointer again this scale, which is being magnified by this gear and
linkage system. So, let us now try to breakdown the instrument into its various functional
blocks.

(Refer Slide Time: 13:52)

Let us redraw the figure here; you have the bulb of the bourdon tube, bulb of the pressure
thermometer. Then the tube and a bourdon tube disconnected for simplicity, let us draw a
c type bourdon tube. So, this is bulb, this is the tube or capillary tube, this is the bourdon
tube, this is gear and linkage system, this is pointer and scale and this is bulb is put to the
medium which temperature I want to measure. So, you can put to a see breathable liquid
is temperature we want to measure.
So, you have the measuring medium, which is essentially this; the bulb interacts with the
measuring medium and this is information about the temperature. So, the primary
sensing element, this is information about temperature first and then what happens due to
restrict thermal expansion of the filling liquid. A pressure is develop that also happen
inside the bulb. So, the temperature signal is being converted to a pressure signal. So,
there is a variable conversion element and both the functions are performed by bulb
causes bulb, which first receives information from the measuring medium about the
temperature.
And due to restrict thermal expansion of the filling liquid pressure gets developed inside
the bulb show and like this signal is no more a temperature signal, but a pressure signal.
Then this pressure is transmitted through this capillary tube. So, this tube box as a data
transmission element. So, this is tube I am talking about this tube, which is capillary tube
which t will have pressure signal here. Then is pressure goes to bourdon tube, which

convert, which converts this pressure signal to a displacement or motion signal. Because
when the pressure actions are the bourdon tube the tip of the bourdon tube will reflect.
So, the bourdon tube receives pressure as put signal and gives motion or displacement as
output signal. So, this is a variable conversion element. So, bourdon tube is variable
conversion element.
Now, the output of the bourdon tube is no more a pressure signal, but a displacement or
motion signal. This goes to this goes through gear linkage system, which magnifies the
reflection. So, gear linkage system works as a variable manipulation element. So, you
stellar motion signal, but now has been magnified. Then this signal is presented to the
observer by this pointer and scale. So, pointer and scale works as data presentation
element.
So, finally, data is presented to observer, you can see here we have a primary sensing
element, we have a variable conversion element, we have a data transmission element
then again we have a variable conversion element, we have a manipulation element and
finally, we have a data presentation element. Let us take another example and again
breakdown the instrument into various functional elements. This time let us take a very
simple pressure gage.
(Refer Slide Time: 22:50)

Will take a very simple pressure gage, which measures will pressure of a fluid and the
measure pressure is indicated by deflection of the pointer against the scale. Consider a

piston here, a piston rod and a spring here and a pointer in scale is attached and let the
fluid pressure act only this term surface. So, how does this, the fluid pressure acts on the
piston surface. So, force will be developed, that force is transmitted through the to the
spring through this piston load.
So, the spring gives displacement or motion as output signal, which can be used to move
this pointer against this scale. So, the deflection of the pointer against this scale can be
rate as a measure of this fluid pressure. It is a very simple pressure gage; so let us now
try to analyze the instrument by breaking down into its various functional elements. So,
piston rod which first interacts with the need measuring media receives the information
about the fluid pressure, so this is of course the primary sensing element.
So, we have measuring medium, which is fluid pressure here; the piston works as
primary sensing element. But the fourth piston also converts this pressure, which is
forced by unit area to force. So, piston also axes variable conversion element. So, piston
is both a primary sensing element, as well as a variable conversion element. So, the
output from the piston is no more a pressure, but a force signal. Now this force is
converted, the force is transmitted by the piston rod. So, piston rod works as data
transmission element swistle has fourth signal here.
Now this fourth signal is transmitted to the spring, which converse this first signal to a
motion signal. So, spring works as variable conversion element. So, you know have a
motion signal, this this magnified with the linkage, which is variable manipulation
element. You can write it here and put linkage here, so we have a magnified motion;
which is presented to observer by data presentation element, which is pointer or scale
here. So, data is presented to observer. So, you see it is possible to breakdown and
instrument into its various functional elements and the functional elements together will
describe how the instrument works.

(Refer Slide Time: 30:32)

Lets now move on to classification of instruments; the instruments can be classified in


various ways. So, depending on the basis of classification that we chose of instruments
various instruments can be classified variously. For example, we can classify the
instruments on the basis of energy consideration; that means, weather it is self operating
instruments or it requires an auxiliary source of power like, weather it is automatic
instrument or a manual instrument, weather it is necessary to bring the instrument in
intimate contact with the measuring medium or it is not necessary to bring the instrument
in close contact with the measuring medium. So, there are various bases classification
and now will see some of the ways how we can classify various instruments.
So, let us start to the classification of the basis of energy consideration; on the basis of
energy consideration we can classify instruments into two categories passive instruments
and active instruments. Passive instruments are those whose output energy is supplied
entirely almost entirely bits input signal or bit the energy of the input signal. So, you do
not need any auxiliary source of power here. These instruments are self operated
instruments; it may be noted here, that the output signal and the input signal may be of
same form or they are may be an energy conversion.
Some of the examples of passive instruments are ordinary mercury in glass thermometer
or clinical thermometer, a bourdon tube, which is use for measurement of pressure or a
pitot tube. All this instruments are self operated instruments their own need any power

source. The energy required for the act of measurement is supplied entirely by the energy
of the input signal it receives.
On the other hand active instruments are those, which an auxiliary source of power. So,
there not self operated instruments, but there power operated instruments. Again the
output and input signals may be of same form or there may be an energy conversion
involved. Examples of active instruments are electronic amplifier differential transformer
for displacement measurement, etcetera. So, this qualification is on the basis of weather
and instrument requires auxiliary power or not. If the instruments requires auxiliary
power for its operation; then instrument will be called an active instrument, if the
instrument does not required any auxiliary source of power the instrument will be called
a passive instrument.
(Refer Slide Time: 35:08)

We can also classified the instruments on the basis of analog and digital mode of
operation, if the instrument follows analog mode of operation we call the instruments, as
analog types of instruments or if the instruments follows digital mode of operation, we
call those as digital types instruments. Majority of the primary sensing elements are of
analog types. They present the information of about the measure variable in the form of
continuous variation with respect to time.
On the other hand digital types of instruments represent the measure variables by discrete
quantities or digital quantities, which are discrete in nature. So, this classification is on

the basis of whether a particular instrument follows analog mode of operation or digital
mode of operation.
(Refer Slide Time: 36:37)

Still another we have classifying instruments is on the basis of operation on a null or


deflection principle. Deflection-type instruments the physical effect generated by the
measuring quantity produces a similar, but opposite effect in some part of the instrument
and this effect is close the related to some variable like mechanical displacement or
deflection in the instrument; that can be easily observed by human operators. So, the
physical effect in the deflection type of instrument that is generated by the measuring
quantity produces a similar, but opposite effect in some part of the instrument and this
effect is closely related to some variable like mechanical displacement or deflection in
the instrument; that can be easily observed by an human operator. An example can be the
typical spring balance or a bourdon tube we are just in bourdon tube. So, let us see the
typical spring balance.

(Refer Slide Time: 38:41)

So, this is a deflection type weight measuring instrument; what a simple spring balance
weight of the object weight of this object is measured by the deflection of the spring
cause by it. So, this is a very simple example of deflection principle the example of
bourdon tube we have just seen in our discussion on functional elements Now, null types
the null types attempts to mention deflection is zero just provide it with either a manually
operated or automatic balancing device; that generates and equivalent opposing effect to
nullify the physical effect cause by the measuring quantity. These instruments are
provided with either a manually operated or automatic balancing device; that generates
an equivalent opposing effect to nullify the physical effect cause by the measuring
quantity. Examples are a dead with pressure gage equal arm balance electrical resistance
measurement by wheatstone bridge in all this a null point is attend during act of
measurement.
Lets take example of let say an equal arm balance; so we draw the figure of an equal
arm balance. We have unknown weight you have standard it. This is the null point; this
is equal arms, we call this balance equal arm beam balance. They information about the
weight of this unknown object can be measure directly by putting standard weights here
and mentioning a null point here.

(Refer Slide Time: 44:45)

Still another example can be a dead weight pressure gage; lets draw a simple diagram
for dead weight pressure gage you. Have a platform, when the fluid force when the fluid
pressure acts here, this will try to go which can be counter balance by putting standard
weights here. So, by mentioning a null point, we can measure the fluid pressure. It may
be noted here; that accuracy attainable in null method is in general higher compared to
deflection method. In a null type instrument defect produce by the measuring quantity
can be compared directly with a primary standard.
As we have seen in the example of spring balance and equal arm balance in case of equal
arm balance, we are measuring the unknown weight by comparing directly with standard
weights. But in case of deflection type spring balance, we are comparing the weights
again the spring, but spring is not primary standard for weights. So, in general the null
methods are superior in terms of accuracy of measurement. However, for dynamic
measurements deflection methods have distanced have advantage; because they are
faster. However, it is possible to increase the speed of the null method by using
automatic balancing devices. So, the advantage of null method is increase accuracy, but
disadvantages. It is generally slow because of an type it is manual; however, it can be
improved using automatic balancing devices.

(Refer Slide Time: 48:55)

The next classification is based on whether the instrument is contacting type or non
contacting type. As the name suggest the instrument as contacting type, when it is
necessary to bring the instrument in direct contact with the measuring medium. For
example clinical thermometer, the ordinary clinical thermometer you put in your mouth
wait for some time how much time, you wait usually depends on the time constant of the
instrument. And finally, you get the reading directly of the scale let us to the clinical
thermometer.
It is necessary for the purpose of measurement that you bring the thermometer in direct
contract with the measuring medium. So, this is contacting type most of the instruments
are contacting type; other examples can be a thermocouple a pressure gage and so on. So,
fore while non-contacting types you do not have to bring the instrument in direct contact
with the measuring medium. So, the instrument can measure the variable without being
in physical contact with the measuring medium.
For example an optical pyrometer or radiation pyrometer measures the temperature of
the furnace; but you do not have to bring the pyrometer in direct contact with the
furnace. It is inconvenient to that to do that as well, will see later how this instrument to
work we need to talk about temperature measurement in detail. But this is an instrument
which is of non-contacting type. Similarly ultrasonic level measurement, which makes

use of the speed of the ultra sound, to get an indication of the level of ugly quick is also
non-contacting type.
(Refer Slide Time: 51:43)

Let us now move on to input output configuration of instruments; we are asking here can
you develop a generalized configuration, which represent significant input output
relationships present in an instruments. Let us consider this general model of an
instrument an instrument receives input signal from the measuring medium and gives us
and output signal, which is related to this input signal.
So, you can consider that there is a transfer function here, which acts on this input signal
and gives as output signal. Now let us try to see, what are the various types of
instruments that effect this instrument any particular instrument. The input quantities can
be broadly classified into three categories, desired input, interfering inputs and
modifying inputs. Desired inputs measuring and quantities that the instrument is
designed to be measure; so the instrument has been designed to be sensitive two desired
inputs only; interfering inputs are those quantity is that unintentionally affect, the
instrument, as a consequence of the principles use to acquire and process the desired
inputs.
So, the instruments becomes unintentionally sensitive to interfering inputs; so while an
intent to measure desired inputs and instruments has been design to be sensitive to
desired input only. The instruments becomes unintentionally sensitive; to some other

inputs and we call those inputs as interfering inputs, and finally, modifying inputs.
Modifying inputs are those undesired quantities, that affect the by altering the inputoutput relations for desired and interfering inputs, if we consider this as desire input and
there is a terms of function, which works on the desired input and gives gives us output.
Now, modifying inputs will tend to change this transfer functions; in other words
modifying inputs affect the modifying output affect the relationship that exist between
input and output. And in this in this block diagram, we are saying this transfer function
relates this modifying input and output. So, modifying input will tend to change this
transfer function. if we have to speak mathematically.
(Refer Slide Time: 55:30)

So, we can have a generalized input-output configuration of instruments like this, we


have desired inputs, modifying inputs and interfering inputs. We have a transfer function
corresponding to desired inputs. So, this transfer function works on this desired input and
gives as output component due to desired input. We have an interfering input and there is
transfer function corresponding to interfering input with gives us an output component
due to interfering input. But modifying input, will tend to change the relationship that
exist between desired input and the output as well as interfering input and output.
So, we have a transfer function corresponding to modifying input, corresponding input
modifying input verse desire input interactions and the modifying input interfering input
interactions. So finally, we have the component for the output that is due to desire input

or modifying input you also have a component due to interfering input and modifying
input, when these two signals are edit together we have the final output.
(Refer Slide Time: 57:12)

We can also explain this in terms of a simpler block diagram as follow, we have
measuring variables; this goes to the input instrument or sensor. So, this is the desired
input, it also receives interfering input as input signal and also modifying input as in
input signal, and finally gives us the output. As understanding you that modifying input
will tend to modifying will tend to relationship between to this input, and these two
inputs, it can explain graphically like this.
Suppose this is input-output relationship when there is no interfering input, sorry, when
there is no modifying input. In presence of modifying input this presence you can be
shifted like this. So, will stop here for the day and in our next class will see more on
interfering input, desire input, and modifying input. What corrective measures, we can
take, so the our instrument you sensitive to only desired inputs. In other words how can
we cancel out the effect of interfering input and modifying input.

Process Control and Instrumentation


Prof. D. Sarkar
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture - 37
Instrumentation: General Principles of Measurement Systems (Contd.)
(Refer Slide Time: 00:29)

In your previous lecture, we studied our discussion on Input Output Configuration of


Instruments. We were trying to develop a generalized configuration that represent,
significant input output relationship that exist in an instrument. In this context we define
three different types of inputs. We said that input quantities can be broadly classified into
these categories. Desired Inputs we defined desired inputs as the quantities that the
instrument is designed to be measured. So, the instrument has been design to be sensitive
to desired inputs only. Interfering Inputs quantities that unintentionally affect the
instrument as a consequence of the principles used acquire, and process the desired
inputs. And, Modifying Inputs are undesired quantities that affect the output by altering
the input output relations for desired and interfering inputs.

(Refer Slide Time: 01:27)

So, we presented this block diagram in previous class which represents a general Input
Output Configuration of Instruments. You have Desired Input, Modifying input and
Interfering Input. Let, us consider this as a Transfer Function which works on desired
input this one or let, us 1st consider this the Transfer Function works on interfering input.
And since modifying input alters the relationship that exist between desired input and
output and interfering input and the output we have a transfer function TF M,D which
alters, which represents interactions between these two and a transfer function which
alters the which interacts with modifying input and interfering input.
So, you can consider that that this transfer function which works on the desired input
signal gives us and signal which can be consider as component of the Final Output. And
this, component is due to desired input as well as modifying input. Similarly, the
component here is the result of interfering input as well as modifying input when these
two components are summed up we get the Final Output.

(Refer Slide Time: 03:25)

It can also be represented by mode gen by moves simple block diagram like this. Where,
the instrument receives not only the signal corresponding to the Physical variable where
interested in measuring but, also Interfering input as well as modifying input. So, the
instrument receives all these different types of inputs and finally, responds with its
output.
(Refer Slide Time: 03:59)

So, let us now take some Example of Desired input, Interfering input and Modifying
Inputs.

(Refer Slide Time: 04:20)

First let us consider, a manometer we are considering a U Tube manometer. You know U
Tube manometer is used to measure different self ratio. Let us consider the two links of
U tube manometer is connected to two different pressure sources. We consider P1 to be
greater than P2 and there is manometer liquid. So, P1 is greater than P2 this, difference
of liquid level between these two tubes to be called it x, x can be considered as a measure
of P1 minus P2 the differential pressure. So, Desired Input here is P1, P2 for the
differential pressure. Now, let us take the same U tube manometer and put it in a vehicle
which is accelerating. So, we have the same U tube manometer.
(Refer Slide Time: 06:33)

We put the U tube manometer on a vehicle that is Accelerating. Now, even if we apply
same pressure to both the lames of the manometer will see a reading. We are considering
here, P1 equal to P2. If, P1 is equal to P2 and if this is not mounted on a vehicle which is
accelerating x would be equal to 0 but, if I mount the U tube manometer on a vehicle
which is accelerating and even if the limes to both the limes of the U tube manometer is
connected to same pressure source it will show a differential pressure gradient. So,
Acceleration is working as an interfering input. The manometer has become
unintentionally sensitive to acceleration. Similarly, if you tilt the manometer with an
angle then, also you can see that even if both the limes of the manometer connected to
the same pressure source the manometer show a non 0 really.
So, the tilt angle again works as an interfering input. So, what can be the modifying input
here? We know the modifying input changes the relationship that exist between input
and output or, in other words it affects the transfer function between the desired input
and the output and also interfering input and the output. We know temperature affects the
density of the manometer liquid. So, if temperature changes the change in temperature
will affect the manometer really. Since, the density of the manometer liquid will play a
role in the celebration of the manometer and the density changes with time density
changes with temperature. So, the temperature can act as modifying input.
(Refer Slide Time: 10: 13)

Let us take still another examples. Let us take an example of Strain Gauge. A strain
gauge is a device which Measures the strain of a specimen. This is an example of a strain
gauge will talk about strain gauge in more detail later for the time being let us consider
this is a simple resistance wire. Now, this resistant wire is cemented on a specimen. Say
if, the specimen is strain the resistance wire also gets strained.
Now, the resistance wire when strain it will change the resistance of the resistance wire.
So, if I take help of an withsten barge, and put the strain gauge here and then, 1st
maintain null point and then I strain there is specimen. So, the resistance wire gets strain
which will cause a change in the resistance of the strain gauge. So, there will be an
unbalance able circuit and the current can be taken as measure of the strain.
So, the Desired Input here is Strain. The strain gauge has been design to be sensitive to
strain. So, the temperature can act as an interfering input because, temperature can
change the gauge resistance even if there is no strain on the gauge. Because, the
resistance of the wire is a function of temperature and if there is a change in temperature
in the ambient the gauge resistance can change which can show an unbalance current in
the circuit. So, temperature works as an Interfering Input. Temperature can also Change
the gauge resistance due to the strain resulting from the differential expansion of the
gauge and the specimen. Ordinarily the strain gauge on this specimen will be of materials
with different thermal expansion coefficient. So, a change in temperature can calls
differential expansion of the gauge on the specimen
So, this can calls a strain and again temperature this way also works as interfering input.
Temperature also works as a Modifying Input here because, Gauge factor is dependent
on temperature, we can define the gauge factor of this strain gauge as Change in
resistance, Change in gauge resistance is equal to Gauge Factor multiplied by gauge
resistance when unstrained multiplied by Unit strain. Since, Gauge Factor depends on
Temperature. So, temperature will work as modifying input.

(Refer Slide Time: 14:46)

So, next let us talk about what we can do to make our instrument respond to only desired
input. So, we want to filter out the effect of or the spurious effect of modifying input as
well as interfering input. Now, method of Signal Filtering uses filters to block the
undesired inputs. So, we can block the undesired inputs namely Modifying Input as well
as Interfering Input by introducing suitable filters. The filters may be applied to input
signal as well as output signal depending on whether we apply the filter to input signal or
output signal we can have input filtering or output filtering.
So, this is a Corrective Measure for Undesired Inputs. In case of Input Filtering we want
to apply the filter to input signal. So, that it blocks the undesired input. Again, let us
consider the block diagram which represents the relationships between Desired Input,
Modifying input, Interfering Input on one hand and the output on the other hand. I want
this output to be sensitive to desired input only. But, the instrument also response in the
process of being sensitive to desired input only to modifying input as well as interfering
input.
So, if I can introduce a filter here, and if I can design a suitable filter it may block the
modifying input. In other words you can consider that I want Transfer Function here
which is actually 0. So, when I M the input modifying input is being worked upon by a
transfer function which is equal to 0 will give a signal equal to 0.

Similarly, to block the effect of interfering input I can introduce a suitable filter here.
Again, mathematically speaking or in terms of our block diagram notation we should
have a Transfer Function here, which is equal to 0. If, we can do that this component will
be this signal will be essentially 0 and this signal will be void of effect of modifying
input. So, this Final Output will be surely result of desired input. So, by introducing
suitable filters to input signal it is possible theoretically to make my instrument sensitive
to desired input only. Or to make my measurement to be sensitive to desired input only.
(Refer Slide Time: 19:17)

I can also introduce filters to the output signal again consider the same block diagram. In
case of Input Filtering we put filters to the modifying input signal as well as to
interfering input signal. In case of Output Filtering will introduce filters at output signal.
So, if I can split this output here which is a result of desired input modifying input as
well as interfering input if, I can split this output signal into components of output
component coming from only desired input output component coming from only
modifying input and output component coming from only interfering input. And then,
introduce filters to the component that comes from modifying input to the component
that comes from interfering input my Final Output will be sensitive only to desired input.
So, the concepts are very much similar in case of input filtering will introduce filters to
input signal in case of output filtering we introduce filters to the output.

(Refer Slide Time: 21:10)

So, now let us take an example of input filtering.


(Refer Slide Time: 21:26)

Let us consider a Thermocouple. Thermocouple is a temperature measuring instrument.


It is made up two dissimilar metals we will talk about thermocouple in more detail later.
Thermocouple is made of two dissimilar metals we make two junctions one we call
Measuring junction another one is called Reference junction. So, if you take dissimilar
metals let us say metal A and metal B and form two junctions if, you keep this two

junctions at two different temperature them EMF is produced, and this EMF depends on
the difference between this two junction temperature.
So, now if I keep one junction temperature constant and call this as Reference junction
and expose this junction which I call Measuring junction to the measuring medium
whose temperature I want to measure then, this EMF produce can be taken as a direct
measure of the temperature. The EMF produce depends on the difference between this
junction temperature and this junction temperature but, by keeping one junction
temperature constant I can measure the temperature directly by looking at the EMF value
of course, after suitable calibration. So, now, let us come back to this. This is same as
this. So, you have and EMF measuring instrument here a mille voltmeter the EMF is the
order of mille volt.
Now, this reference junction has to be kept at some constant temperature. So, if there is
Varying Ambient Temperature this can work as an interfering input to the thermocouple.
So, one way to put a filter to block the effect of varying ambient temperature will be put
a Thermal Insulation around the reference junction. So, by putting a Thermal Insulation
around reference junction I can block the effect of varying ambient temperature. So, this
is an example of Input Filtering. Again another example of blocking the effect of
interfering input can be in strain gauge.
(Refer Slide Time: 26:11)

We just talked about Strain Gauge which measures strain in a specimen consider this to
be the strain gauge put in one norm of the western bridge. We know the unbalance
current is a measure of the strain in the specimen. Now, this strain gauge can become
sensitive unintentionally to magnetic field. So, to block the effect of interfering magnetic
field we can put a magnetic sealed around the circuit. So, we have a Magnetic Shield
which blocks the effect of interfering magnetic field. So, again this is an example of
Input Filtering. Another way of taking Corrective Measure for Undesired Inputs will be
method of Opposing Inputs. Here, we intentionally introduce an input that cancels the
effect of undesired inputs. So, we introduce intentionally an input to the instrument that
tend to cancel the bad effect of modifying input and a interfering input.
Let us look at this block we have Desired Input corresponding Transfer Function. We
have Interfering Input corresponding Transfer Function. We intentionally introduce this
Interfering Input. We have this Unavoidable Interfering Input. So, we have a desired
input signal we have an interfering input signal which is undesirable which is
unavoidable but, we intentionally introduce another interfering input signal and we
choose this signal such that, the effect of unavoidable interfering input signal and the
intentionally introduced interfering input signal cancels each other out. So, this is
essentially a Zero signal. So, the Final Output becomes this component alone or, in other
words the Final Output becomes result of desired input alone. So let us now, take an
example of method of opposing input.
(Refer Slide Time: (31:29)

We know Milli voltmeter is essentially a current sensitive device. But, the scale of a milli
voltmeter can also be calibrated in volts because volt and current are related let us
consider this as schematic of milli voltmeter. So, you have the Coil Ambient
Temperature works as Modifying input here because, the ambient temperature can cause
a change in the Coil Resistance. So, how do I nullify the effect of ambient temperature?
We can add another resistance into the circuit and it is materially such that it has a
temperature coefficient of resistance opposite to that of coil ambient temperature can
change the resistance of the coil. So, it can work as a modifying input. Now, if I add a
compensatory resistance in the circuit and I choose the material such that it has a
temperature coefficient of resistance opposite to that of coil I can cancel out the effect of
ambient temperature. So, this is an example of Method of Opposing Input. I intentionally
introduce a resistance to cancel out the effect of ambient temperature the changes in
resistance of coil. Let us take another example of method of opposing input we are talked
about Bourdon tube.
(Refer Slide Time: 35:58)

.
Let us consider a schematic diagram of Bourdon tube. We know the deflection of the T
is the measure of the pressure being apply inside the bourdon tube. Now, the tip
deflection depends on various factors but, for given bourdon tube other things being un
change this tip deflection can be consider as a function of Pressure alone. Among the
various factors on which the tip deflection will depend the elasticity of the material is 1.
So, modules of elasticity can change with temperature.

Now, if it changes sufficiently the tip deflection can no longer be considered as a


function of Pressure alone. So, in sit we can get additional deflection for changes in
temperature. One solution can be we put a Bimetal Strip here. Suppose, we are talking
about a situation where the tip of the bourdon tubes is deflecting mode then it should
correspond to the given pressure being applied and that additional deflection is the result
of change in Ambient Temperature.
Now, if I can design a Bimetal Strip which, deflects when subjected to a temperature let
me explain what a bimetal strip is? You take two different metallic strip they are
different thermal expansion coefficient they are bounded rigidly at some temperature
called T0 Bonding Temperature. Now, if I subject this bimetal strip to a different
temperature say New temperature the bimetal strip since they have different thermal
expansion coefficient can reflect like this or can reflect like this.
So, now come back to the bourdon tubes example. If, I can suitably design a bimetal strip
such that the extra reflection that is caused in the bourdon tubes by the change in
Ambient Temperature and the deflection of the bimetal strip due to change in ambient
temperature is same in magnitude but, opposite in direction then this deflection of the
point are against this scale will be surely a function of this Pressure. So, by putting a
suitably design bimetal strip I can cancel out the extra deflection that has cause due to
Change in Ambient Temperature. So, this is again an example of Method of Opposite
Inputs.

(Refer Slide Time: 41:44)

Now, let us talk about Performance Characteristics of Instruments. The question you ask
here is. How do you describe the quality of measurement? Or, How do I evaluate the
performance of an instrument? They are various instruments available if, I want to
compare an instrument with other possible alternatives I must have some quantitative
basis for comparison. So, Performance Characteristics of Instruments are those
quantitative bases for comparing the performance of instrument with possible
alternatives.
Two classes of characteristics are used to evaluate or compare instruments namely Static
Characteristics and Dynamic Characteristics. Static Characteristics are associated with
static measurement these are set of criteria that are used when you measure a quantity or
a condition that is either constant changes very slowly with time. So, Static
Characteristics are those characteristics which we must consider any one to analyze the
performance of an instrument that is measuring a medium which is constant or varying
very slowly with time.
So, you are essentially talking about static medium or steady states situations. On the
other hand Dynamic Characteristics are associated with dynamic measurements. So,
these are set of criteria that are used when we measure a quantity that is rapidly varying
with time. So, Dynamic Characteristics are those characteristics which we must consider
if inputs are varying with time while, static characteristics are those characteristics which

we must consider when you measure a quantity that is not varying with time. We would
like to emphasize here, that Static Characteristics also influence the quality of
measurement under dynamic environment.
So, static characteristics not only influence the quality of measurements under static
environment it can also influence the quality of measurement under dynamic
environment. But, usually we treat them separately the reason is as follows will see later
that the dynamic characteristics can be describe usually in terms of differential equations.
Now, if I one to incorporate that if static characteristics on to this dynamic characteristics
the resulting differential equations will become extremely difficult to handle. So, a
general procedure is stood state static characteristics and the dynamic characteristics
separately. And we can always have a qualitative super in position of both the
characteristics to make a final choice about, the instrument or to have a final say on the
performance of the instruments or quality of measurements.
(Refer Slide Time: 46:31)

These are the set of Static Characteristics we should be familiar with as we define these
are the characteristics which we must consider when the measuring medium is not
changing with time or changing very slowly with time. The Static Characteristics can be
classified into two categories Desirable Static Characteristics and Undesirable Static
Characteristics. Under Desirable Static Characteristics we have Accuracy, Sensitivity
and Precision as various characteristics.

Similarly under Undesirable Static Characteristics we have Static Error, Dead zone and
Drift. Dead zone can be result of Threshold or Hysteresis. It can also be noted here, that
accuracy and static error are related in the same stat if your instrument is accurate
enough or if the desirable static characteristic accuracy is high the undesirable static
characteristic static error will be low. So, corresponding to accuracy which is desirable
static characteristic we have static error which is undesirable static characteristics. So,
corresponding to sensitive which is desirable we have dead zone which is undesirable
corresponding to precision which is desirable we have drift which is undesirable.
(Refer Slide Time: 49:03)

So, let us start our discussion on Static Characteristics of Instruments with a term called
Static Calibration. This is a procedure use to assess the relation between the output and
the input of a instrument if all influencing parameters are controlled within working
range. So, calibration is the act or result of quantitative comparison between the known
standard and the output of the measuring system measuring the same quantity.
Calibration is accomplished by applying known magnitudes of the input and observing
the measurement system output.
So, what do you do is as follows. All inputs desired interfering or modifying inputs
except one are kept at some constant values, then the one input under steady is varied
over some range of constant values which, causes the outputs to vary over some range of
constant values the input output relations developed in this way comprise a static

calibration valid under the stated conditions of all the other inputs. So, all inputs except
one are kept some constant values then one input under steady is vary over some range
of constant values which, causes the outputs to vary over some range of constant values.
The input output relations developed in this way comprise Static Calibration which is
valid under the static condition static constant conditions of all the other inputs.
So, essentially it is determining a relationship between input and output. So, we keep all
the influencing parameters or all input except one constant and vary this constant over
some range each time we know the instruments output and can generate the relationships
that exist between this set of inputs and outputs. Depending on whether the instrument is
Liner or Nonlinear I can have a Linear relationship that exist between input and output
represented by a straight line or, a Nonlinear curve that represents the relationships that
exist between input and output. Once an instrument is selected it must be calibrated
against known standard.
(Refer Slide Time: 52:45)

Let us now, define the Static Characteristics of Instruments. Range the range of an
instrument is the lowest and highest values it can measure within its specifications. Span
the span of an instrument is the high end of the range minus the low end of the range. So,
if I have an instrument where the lowest point of calibration is a units and the highest
point of calibration is b units then the instruments range is a to b the lowest point of
calibration to the highest point of calibration and the instruments span is b minus a, so

highest point of calibration minus lowest point of calibration. Resolution is the smallest
amount of input signal change that the instrument can detect reliably. It should be noted
that we should select instrument such that its range includes all values a process will
normally encounter, including expected disturbances and possible failures.
(Refer Slide Time: 54:25)

Accuracy of a measurement describes how close the measurement approaches the true
value of the process variable. So, it is a measure of how close the measured value is to
the true value. Accuracy can be expressed in many ways accurate within plus minus x
percent means accurate to within plus minus x percent of instrument span at all points of
the scale. Accuracy can also be expressed in terms of percentage True Value which can
be expressed as Measured Value minus True Value divided by True value multiplied by
100. Similarly, accuracy can also be expressed in terms of percentage Full scale
deflection and this can be represented as Measured Value minus True Value divided by
Maximum Scale Value multiplied by 100.

(Refer Slide Time: 55:40)

If, accuracy is measure of how close the measured value is to the true value static error
must be the difference between the measured value and the true value of the quantity and
of course, we are talking about Static Characteristics, so, under static condition. So,
Static error is the difference between the measured value and the true value of the
quantity. So, True Value plus Static Error is actually what instrument gives us as its
reading. So, True Value is Instrument Reading plus some correction factor which we
called Static Correction.
Precision is the Ability of an instrument to reproduce a certain set of readings with a
given accuracy. So, precision is an ability of an instrument to reproduce a certain set of
reading with a given accuracy. This instrument can have precision low accuracy, it can
have poor precision poor accuracy and so on and so forth. In our next class will see mode
of static characteristics and also talk about dynamic characteristics.

Process Control and Instrumentation


Prof. D. Sarkar
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture - 38
Instrumentation: General Principle of Measurement Systems (Contd.)
In our previous discussion we started with Static Characteristics of Instruments. We
discuss that there are set of Desirable Characteristics of Instruments and there are set of
Undesirable Characteristics of the Instruments. The desirable characteristics like
Accuracy, Sensitivity, Precision and the undesirable static characteristics we talked about
our Static Error, Dead zone and Drift.
(Refer Slide Time: 00:40)

Today, let us talk more about the Static Characteristics of Instruments. So, we have set of
desirable static characteristics and we have set of undesirable static characteristics. The
desirable static characteristics accuracy has a corresponding undesirable static
characteristic static error similarly, sensitivity and dead zone precision on the desirable
static characteristic list and corresponding undesirable static characteristic is drift. Dead
zone can come from Threshold or Hysteresis.

(Refer Slide Time: 01:43)

We define these terms in our previous lecture let us do a quick review. The range of an
instrument is the lowest and highest values it can measure within its specification, while
the span of an instrument is the high end of the range minus the low end of the range. For
example, the lowest point of calibration say a units and the highest point of calibration is
b units that the instruments range is specified as a to b while instrument span is b minus
a.
Resolution is defined as the smallest amount of input signal change that the instrument
can detect reliably. It should be noted that we should select instruments such that, its
range includes all values a process will normally encounter and that we should include
expected disturbances as well as possible failures.

(Refer Slide Time: 03:03)

Accuracy of a measurement describes how close the measurement approaches the true
value of the process variable. Accuracy can be expressed in many ways when you say
accurate within plus minus X percent we mean accurate to within plus minus X percent
of instrument span at all points of the scale. Accuracy can also be expressed as
percentage True Value, percentage true value can be defined as Measured Value minus
True Value divided by True Value multiplied by 100.
Accuracy can also be expressed as percent Full-scale deflection. Percent full-scale
deflection can be described as, Measured Value minus True Value divided by Maximum
Scale Value multiplied by 100. For example, if a pressure gauge of range 0 to 10 bar has
a quoted inaccuracy of plus minus 1 percent of full-scale reading then, the maximum
error to be expected in any reading is 0.1 bar. This means, that when the instrument is
reading 1 bar the possible error is 10 percent of this value.

(Refer slide Time: 04:58)

Corresponding to desirable static characteristic accuracy we have undesirable static


characteristics static error. Static error is defined as the difference between the true value
of the quantity under static condition and the measured value that is value indicated by
the instrument. So, it is a difference between the true value of the quantity and the value
indicated by the instrument and, we are talking about a condition which is not changing
with time or changing very slowly. So, that it can be considered static. Static error is
expressed as plus a units or minus a units.
So, True Value plus Static Error is Instruments Reading. If, true value plus static error is
equal to instrument reading and if the static error is positive the instrument will read high
and if the static error is negative the instrument will read low. So, True Value can be
equated to Instrument Reading plus Static Correction. In other words Static Correction is
equated to minus of static error.
A typical static error curve may look like this. While, on x axis is the Measured variable
percent span and, on y variable I have plotted static error again expressed as percent
span. So, corresponding to different value of Measured variable I can have an estimate of
Static error if, I have this Static error curve.

(Refer Slide Time: 07:25)

Next, let us talk about another desirable static characteristic Precision. Precision is
defined as ability of an instrument to reproduce a certain set of readings within a given
accuracy. Say it is the ability of an instruments to reproduce a certain set of readings
within a given accuracy this, can be best explained by an example. Let us consider a
person is doing a shooting practice. So, the person has few bullets in hand. In fact, 8 in
number and the Shooting target is this boy which is intersection of these two lines.
Now, here is a set of outcomes of the persons shooting practice. In this case he miss the
target, but all the bullets are here. So, this is a case of High Precision, but Poor accuracy
Precision is high because, all the bullets are together. So, there is ability to reproduce, but
accuracy is very poor because the person could even hit the target once. So, it is not
necessary that an instrument is highly precise will also be highly accurate.
Let us consider this scenario. Here, the bullets are all scattered. So, this is a case of Poor
Precision Poor Accuracy. In this example, each time the person has been able to hit the
target exactly all the 8 bullets have a this centre. So, this is obviously, a case of High
Precision and High Accuracy.
Now, let us consider the last scenario. Here, the person has not been able to exactly hit
the target even for once, but all the bullets are around here. So, we can say that this is a
case where Good Average Accuracy is there but, there is no precision. So, it is a case of
Poor Precision, but Good Average Accuracy.

Now, if I plot the measurement values verses number of measurements I will get this
frequency plot. This represents the mean. So, High accuracy will signify that the mean of
the measurements is close to that true value. So, high accuracy will mean the mean is
very close to the true value. While, High precision will mean that there is less scattered
in other words the standard deviation is low.
(Refer Slide Time: 12:37)

Let us define another static characteristic call Static Sensitivity. This is defined as Ratio
of change in the input or response of instrument to a change of input or measured
variable. So, this is change in output or response divided by change in input or measured
variable. So, this can be defined as delta output divided by delta input where, delta
output is change in output and delta input is change in input. So, it is easy to recognize
that this is nothing, but slop of calibration curve.
So, let us look at this calibration curve. This is a calibration curve for a Linear instrument
where, the input output follows a linear relationship. So, the Slope of this curve or slope
of this straight line which is given by delta output divided by delta input will be the
Static Sensitivity of the instrument. This is a calibration curve for a Nonlinear instrument
where, the instruments input and instruments output follows a nonlinear relationship.
Here, the static sensitivity can be defined at different points by the slope at that point.
Let us take this plot. Let us say these are all calibration curve for different instruments. It
is easy to recognize that the sensitivity of this instrument is highest next sensitive is this

instrument and last is this one because, the gradient is stiff here compare to these two.
So, delta output by delta input value is maximum here compare to this and then this. So,
What should be the unit of static sensitivity? It is defined as delta output by delta input.
So, its unit will also be unit of output by unit of input.
Let us say The resistance value of a resistance thermometer changes with temperature
changes when the temperature increases. See if, we change the temperature the resistance
of a Resistance Thermometer will change so obviously, the unit of sensitivity for
Resistance Thermometer will be unit of resistance divided by unit of temperature. So, it
may be Ohmper degree Celsius. So, for a mercuryinglass thermometer if, mercury level
moves by 1 centimeter when the temperature changes by 10 degree Celsius its sensitivity
will be 1 centimeter divided by 10 degree Celsius. In other words it is 1 millimeter per
degree Celsius.
(Refer Slide Time: 17:25)

Let us take an Example. The output of a platinum resistance thermometer which is called
RTD in brief is recorded as follows. On the left I have temperature column and in right
we have resistance corresponding resistance column. So, at 0 degree Celsius the
resistance of the RTD is 0. At 100 degree Celsius the resistance of RTD is 200 Ohm, at
200 degree Celsius the resistance of RTD is 400 Ohm, at 300 Celsius the RTD has
resistance of 600 Ohm and at 400 degree Celsius RTD has resistance 800 Ohm.

We need to calculate the sensitivity of the resistance thermometer device or RTD. It can
be easily done if we just Draw an input verses output graph. So, I will make a graph of
temperature on x axis and the resistance in Ohm on y axis. So, the Slope of the graph
will give me the value of static sensitivity. If, you look at the data table it is obvious that
the input which is temperature in degree Celsius and the output which is resistance in
Ohm follows a linear relationship. So obviously, we are going to get a straight line. So, is
easy to calculate the slope from the data given.
So, slope of the graph will be if we consider these pair of points 400 minus 200 divided
by 200 minus100. So, this is 200 divided by 100 which is equal to 2 Ohm per degree
Celsius. Also simply by looking at the table as I just discuss that the data obviously,
produces a linear relationship and for every 100 degree Celsius increase in temperature
there is 200 Ohm increase in the resistance. So, the measurement sensitivity is 200
divided by 100 is 2 Ohm per degree Celsius.
(Refer Slide Time: 21:00)

Next, static characteristic we will discuss is Dead Zone. Dead Zone is the Largest range
of values of a measured variable to which the instrument does not respond. So, dead zone
is the largest range of values of a measured variable to which the instrument does not
respond. The Hysteresis is the characteristic loop we find when the instrument is
calibrated first in one direction and then, in the other direction this is caused by friction
or backlash.

(Refer Slide Time: 22:28)

Let us consider an instrument with a scale as follows. We have 0 at lower end and let us
say 100 at high end. So, is the point. Let us say this is the pressure gradation we go on
increasing the pressure from 0 degree from 0 say atmosphere to 100 atmosphere and
then, slowly realize the pressure from 100 atmosphere to 0 atmosphere. So, now if I
make a plot of input which for this case is Pressure verses output which can be
considered as the Angular rotation of the point at against this scale which is of course,
calibrated in terms of pressure units. So, if while increasing the pressure from 0 units to
100 unit I follow this path while, lowering down the pressure from 100 units to 0 units I
may follow another path as follows. So, Hysteresis is this loop this is the non
coincidence of this Loading and Unloading curve. So, while I am increasing pressure or
where in I am putting more energy to the instrument and while I am reducing the
pressure from 100 units to 0 units, I am recovering that energy. That energy is being
released.
So obviously, I am unable to recover the entire amount of energy. So, this non
coincidence of this loading and this unloading curve is called Hysteresis which is a result
of friction or backlash. If, we have a scale like say 0 at the middle and we have say
minus 100 units at this end and 100 units at the other end, if I do the same experiment.
So, if we start from 0 go on increasing up to 100 units then comeback to minus 100 units
and then finally, return to my initial starting point, if I now make the same plot. The plot

will now look like or may look like start from 0, I reach here, comeback to this unit and
then this. So, again this is the Hysteresis loop.
(Refer Slide Time: 27:33)

Next, let us define a static characteristic called Drift. Drift is the Change in the indicated
reading of an instrument over time when the value of the measuring quantity remains
constant. So, drift is defined as the change in the indicated reading of an instrument over
time when the value of the measuring quantity remains constant. The drift can be
classified as Zero drift or zero error, Span drift or sensitivity drift and finally, Zonal drift
which is a combination of zero and span drift.
In case of Zero drift or zero error the instrument calibration may gradually shift by the
same amount. You can see that in some of the instruments let us say a pressure gauge
when there is no pressure input to the instrument, the instrument still shows of small
positive value which, is a zero error we should take a note of this zero error, so that we
can apply the correction factor to the measured value being indicated by the instrument.
So, Zero drift or zero error The whole instrument calibration gradually shift by the same
amount it may be cause by permanent set or slippage.
In case of Span drift or sensitivity drift A gradual change in which the calibration from
zero upward changes a proportional amount. This may be cause by gradual change in the
spring characteristics of the instruments. Well, Zonal drift can be viewed as a
combination of zero drift as well as span drift or sensitivity drift.

(Refer Slide Time: 30:38)

So, this is an example while the instrument shows zero drift. If, this is the Nominal
characteristic between input for this particular case let us we are talking about a Pressure
gauge. So, pressure is the input and this is the output Scale reading. So, if this is the
nominal characteristic or the calibration curve when there was no zero error this may be
the case when the instrument has develop a zero error or zero drift. So, the whole
calibration has shifted by this amount which is the magnitude of zero error.
This is the case of span drift or sensitivity drift. This is the Nominal Characteristic while
the characteristic with the sensitivity drift or the span drift shows a gradual change from
zero upward. So, this is a case of sensitivity drift or span drift. Finally, this is the case
where we have zero drift as well as sensitivity drift. You can see that if the instrument
had no zero error this would have this point should have coincided with the origin and, if
the instrument had no sensitivity drift, but only zero drift this line would have been
parallel to this Nominal characteristic line.

(Refer Slide Time: 33:24)

Now, let us talk about the Dynamic Characteristics of Instruments. We have defined
Dynamic Characteristics as the characteristics that you must consider when the
instrument is measuring a medium which changes with time. So, these are the attributes
associated with dynamic measurements and, we must consider these whole to just the
performance of an instruments or the quantity of measurement when the measuring
medium is changing rapidly with time.
Similar to static characteristic we have few Desirable Dynamic Characteristics and we
have few corresponding Undesirable Dynamic Characteristics. The Desirable dynamic
Characteristics are Speed of response which, is defined as the Rapidity with which an
instrument responds to changes in the measured medium or measure input. So, speed of
response is the rapidity with which an instrument response to changes in the measuring
medium, while, Lag is Delay in response.
So, speed of response is a Desirable dynamic Characteristics while obviously, Lag which
represents delay in response is Undesirable dynamic Characteristic. Similar to static
error, Dynamic Error is defined as True Value minus Value indicated by the instrument
and of course, it is in dynamic environment. So, Dynamic Error is the difference between
the True Value and the Value indicated by the instrument when the instrument is being
used to measure a dynamic medium. And Fidelity is the Degree to which an instrument

indicates the changes in measured variable without the dynamic error. So, this is a
faithful reproduction.
So, Fidelity is defined as degree to which an instrument indicates the changes in
measured variable without the dynamic error. So, we have Speed of Response and
Fidelity under Desirable static Characteristics and you have Lag and Dynamic error
under

Undesirable

static

Characteristics.

So,

the

corresponding

Undesirable

Characteristic to Speed of Response is Lag and similarly, the Dynamic error is


corresponding to Desirable Characteristic Fidelity.
(Refer Slide Time: 37:02)

Now, let us talk about Generalized Model of Instruments. We need a mathematical


model to describe the behavior of an instrument. If, we develop a mathematical model
that relates the input output relationships then you can solve that equation for known
inputs to get what will be the output of the instrument. In other words the dynamic
response of the instrument can be studied, if we have a suitable mathematical model for
my instrument. An ordinary differential equation of nth order with constant coefficients
can be considered to be a generalized model for any instrument.
So, we can consider an nth order differential equation ordinary differential equation with
constant coefficients to be a generalized model for an instrument. So, this is an nth order
in output and nth order in input differential equation that can be considered to be a
Generalized Model for an Instrument. Let us define q0 as output quantity, q in as input

quantity and a, b are constant coefficients which are essentially combination of system
parameters may be combination of physical parameters which are all being assumed to
be constant. So, that all the as and bs are constants.
So, and this is an equation where on the left we have an nth order OD in output q0 and
nth order OD in input qin. This can be taken as a very general model for any instrument.
Say, if I solve this equation for known inputs I can obtained q0 or output. So, you can
study the dynamic response of this instrument with help of this model.
(Refer Slide Time: 40:33)

Now, it is not really necessary to represent the input as nth order OD. We can express the
output the unknown output as nth order OD and we will know the functional form of the
input. Because, it may be a Step input, it may be a Ramp input, it may be a Sinusoidal
input it may be a combination of these inputs. So, we will consider a generalized model
of an instrument which is as follows. We have an nth order OD in output or in q0 and
this is the input with known functional form. So, if I solve these equations with known
Input and Initial condition I will get output. So, the dynamic response can be studied
with help of such equation.

(Refer Slide Time: 42:23)

Now, let us rewrite this equation here. We have an nth order OD in q0. So, this is the
general mathematical model for my instrument where, q0 is output quantity qi is input
quantity and a, b are constant coefficients which are combination of system parameters.
Now, if an instrument is described by nth order differential equation we say that the
order of the instrument is n. So, if the instrument is described by second order OD or
ordinary differential equation we call the order of the instrument to be 2 for first order
instrument the instrument will be described by a first order OD.
Similarly, at zeroth order OD which is nothing, but an algebraic equation will represent a
zero order instrument. So, let us first see, how the zero order instruments model will look
like. So, when we are talking about zero order instruments let us retain only the zero
order term in this equation. So, we have we will retain only this term from the left hand
side of the model equation.
So, the zero order instruments will be described by if, we rearrange call this K. Which is
static sensitivity also called gain. So, this becomes K into qi. So, q0 K qi is the model for
a zero order instrument. Look at this equation this is an algebraic equation the output can
be found out for known functional form of input. Let us say Step input or Ramp if we
know the value of the gain or static sensitivity. It may be noted that this zero order
instrument will show perfect dynamics the output will immediately follow they will not
be any lag any distortion there is no time term involved here.

So, we can say zero order instrument, if we can have a zero order instrument it will
follow perfect dynamics they will not be any lag at all. It is not easy to get a zero order
instrument, but let us consider a Displacement measuring potentiometer which can be
approximated as a zero order instrument let us draw the diagram.
(Refer Slide Time: 47:40)

Let this length be L. If, the resistance is distributed linearly along length L, so if the
resistance is distributed linearly we can write which is again gain its unit will be say
volts per unit of length may be inch. So, you see the displacement measuring
potentiometer is model by output which is a zero order instrument. Similarly, a first
order instrument will be described by a first order OD. So, again if we look at the
previous differential equation just say, generalize model for an instrument. We will
retain only the First order term. So, we will retain this as well as this for a first order
instrument.

(Refer Slide Time: 50:58)

So, if we retain the First order terms which is a first order differential equation first order
ordinary differential equation. If, we rearrange the equation as follows divide this
equation throughout by the a0. So, let us make this term coefficient less. We have forgot
the dq0 dt term here. If, you remember your class on process control this actually gives
Time constant term whereas, this is static sensitivity or gain.
So, we can write the equation as where. So, if we rearrange this we will get your familiar
transfer function. So, first order system will follow this model. For example, a Mercuryin-glass Thermometer to be more specific there ordinary Mercury-in-glass Thermometer
that means, a Mercury-in-glass Thermometer without any cover will behave like a first
order instruments. So, it can be shown by writing an energy balance equation or heat
balance equation that the governing equation is first order and we will look like this.
So, if I want to know. If, I put my Mercury-in-glass Thermometer suddenly from say a
temperature of 30 degree Celsius to a temperature of 40 degree Celsius. How the
instrument will respond to the this step input of magnitude 40 minus 30 equal to Ti? All I
need to do is solve this, so given the step input of magnitude Ti. I now have the transfer
function or I have now the ordinary differential equations they describes the mercury-inglass thermometer provided this tau and K which are combination of system parameters
are known. I can determine what will be the output of the Mercury-in-glass Thermometer
where the mercury level will rest in the scale.

(Refer Slide Time: 56:34)

Similarly, a Second order Instrument will be described by a second order OD. So, again
let us make this term coefficient less. We define the terms you are familiar with when
you read the process control part of the course usual static sensitivity, undammped
natural frequency damping ratio. So, these are three parameters which will if we know
then a second order instrument can be solved for any given input. U-tube manometer is a
good example of 2nd order instrument. How these instruments whether it is first order
instruments or second order instruments will behave when these instruments are
subjected to a known input.
You are familiar with those responses because, the response of first order systems
response of second order systems for varies inputs have been discussed in detail in the
process control part of the course. So, the same methodologies same principles can be
apply to determine what will be the dynamic response of an instrument whether whatever
its order be very easily.

Process Control and Instrumentation


Prof. D. Sarkar
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture - 39
Instrumentation: General Principles of Measurement Systems (Contd.)
(Refer Slide Time: 00:26)

In todays class we will talk about Error Analysis. While talking about performance
characteristics of instruments we have talked about static error and dynamic error. Now,
since no measurement can be perfect and may particular measurement will always been
error this error can be a various forms. So, we should have some estimate of how much
error there can be in our measurement.

(Refer Slide Time: 01:00)

So, let us now talk about error analysis in some detail Types of Error in Measurements.
The Types of Error in Measurement can be broadly classified into 3 categories 1 the
Gross error or human error 2 Systematic or Determinate Error and 3hree Random or
Indeterminate Error.
So, all the errors that can be involved on in our measurement can broadly be classified
intothis3 categories namely Gross error Systematic error and Random error. Now, Gross
Error is caused by human mistake. So, this is Wrong reading or wrong using of
instruments gross error which is caused by human mistakes cannot be eliminated
completely, but it can be minimized by taking proper care and appropriate uses of
instruments.

(Refer Slide Time: 02:22)

Next let us talk about Systematic or Determinate Error as the name suggests Systematic
error affect all the readings in a particular fashion. This is caused due to shortcomings of
the instrument such as defective or worn parts in the instrument improper design of
instrument loading effect etcetera.
However, systematic errors are usually recognizable. So, when determined systematic
error can be corrected by careful calibration of the instrument. So, let us repeat again
systematic error are those error which affect all the readings in a particular fashion this is
caused due to shortcomings of the instrument such as defective or worn parts improper
design of instrument loading effect etcetera. When you can recognize a determined
systematic error this type of error can be corrected by careful calibration of the
instrument systematic error May be of constant or proportional nature.

(Refer Slide Time: 03:57)

So, if we look at this figure is a float of x variable versus y variable. We can consider x
variable to be the input to the instrument and y variable as the output of the instrument.
So, this can be considered as a calibration curve. So, if we look at the blue line let us say
blue line should be the correct characteristic of the instrument. If, we get the behavior of
the x variable and y variable that is the input and output relationship is represented by a
red line. We see that there is an error associated with the ready but, these error is a
constant type whereas, if the input output relationship is characterized by the blue line
will have a Proportional error. However, as the graph suggest that if, we know how much
of constant error or how much of proportional error is associated within instrument we
can always introduce a suitable correction factor.
So, it is rewriteable easy to handle systematic error. We can classify the Systematic
errors under three different categories: 1 Instrumental error 2 Environmental error and 3
observational error. This can also be considered as source of systematic errors. So,
systematic errors can be of Instrumental error Environmental error or Observational
error.

(Refer Slide Time: 06:11)

Now, let us look at Instrumental error. The Instrumental error is inherent with the
measuring instrument because, of their mechanical structure like bearing friction,
irregular spring tension, stretching of spring, etcetera. So, instrumental error is inherent
with the measuring device. Instrumental error can be avoided by Careful design of
instrument and by careful selection of the components of the instruments it can also be
avoided by calibrating the instrument frequently against known standards.
So, Instrumental error is inherent with the measuring instrument because, of their
mechanical structure, so just bearing fraction, irregular spring tension, stretching of
spring, etcetera. And the best way to avoid Instrumental error will be the Careful design
of the instrument and proper selection of the components of the instrument.

(Refer Slide Time: 07:17)

Environmental error this is due to external condition affecting the measurement such as
change in temperature, change in humidity, change in atmospheric pressure, etcetera. So,
Environmental error is due to external conditions effecting the measurement such as
change in temperature, change in pressure, change in humidity, etcetera. To avoid the
environmental error we should provide suitable control environment for the instrument.
So, providing controlled environment for the instrument is necessary to avoid the
environmental error. Environmental error can also be avoided by sealing certain
components in the instruments. Observational error is introduced by the observer the
most common observational error is parallax error and estimation error while reading the
scale.

(Refer Slide Time: 08:39)

Next let us come to Random or Indeterminate Error. As the name suggests random errors
affect readings in a random way. So, unlike systematic error which effects all the
readings in systematic fashion random errors affect readings in a random way. It cannot
be determined easily we do not have any control over random errors and these are due to
unknown causes.
So, causes for random error are not always easily known. Random errors cannot be
eliminated completely it is difficult to eliminate completely the random errors because,
the sources of random error are usually not always known. However, the random errors
can be minimized by increasing the number of readings that means, multiple trials we
take several readings of the same value of the measuring variable. So, we repeat the
number of measurements and then use statistical means to obtain best approximation of
the true value. So, will take a particular reading several times this is call multiple trials
and then we will use some statistical principles to obtain best approximation of the true
value. This way we can minimize the random error in our measurements, but we can
never completely eliminated.

(Refer Slide Time: 10:36)

Let us now, talk about some Basic Statistics 1st Average. So, if we take a particular
reading N times and each time I have the major values in the carried as xi. So, I have
values x1, x2, x3, up to xN, and then some them up and then divide by the total number
of measurements. So, this is known as average of n readings or mean frequently
represented as mu or x bar. Deviation is defined as difference between a particular value
and the mean.
So, we have to statistics as of now average which is the sum of N readings divided by N.
It is frequently stated that, the mean value is the most probable value of a set of readings
and that is why it is a very important role in statistical error analysis the deviation of the
individual readings from the mean value is obtain as di equal to xi minus mu where xi is
the i-th reading and mu is the mean. We know one to have an idea about the deviation
that is why the individual readings are far away from the mean value or not. So, if we
one to have some idea whether the individual readings are far away from the mean value
or not can you use Mean of deviation.
So, if this is xi minus mu is mean. So, this will represent Mean of deviations.
Unfortunately we cannot use the mean of deviation, because if you follow this the mean
of deviation is always equal to 0. So, instead we use Variance or mean squared deviation
which is defined as this. So, Variance is mean squared deviation. So, we find out the
individual deviation square sum of over N readings and divide by number of readings N.

So, Variance or mean square deviation can be used to find out how far away from the a
particular reading is how far away from the mean value a closely related statistic is
Sample standard deviation which is square root of variance. We can have a biased
estimate of standard deviation or unbiased estimate of standard deviation. If, we want to
have a biased estimate of standard deviation we divide this by N if we want to use
unbiased estimate we should divide it by N minus 1. If, we have a large number of
readings typically more than 30 or so, it does not make much difference whether we
divide this by N or N minus 1 the term standard deviation which is square root of
variance is often used as a measure of uncertainty in a set of measurements.
So, we have Average or mean value then, the individual Deviation which is difference
between a particular value and the mean then the Variance or means square deviation
and then standard deviation which is square root of sigma square. Now, let us talk about
the probability distribution function.
(Refer Slide Time: 17:28)

Suppose we have a sufficiently accurate Pressure gauge, and where measuring a certain
pressure with this pressure gauge several times. Let us say the given pressure gauge the
given pressure is say 10 kilopascal. So, a measuring a 10 kilopascal pressure using a
pressure gauge and let us say for example, the Data looks like this, you can have even
more number already let us say we end like this. So, we are measuring a 10 kilopascal

pressure with a pressure gauge which, can be consider as a sufficiently accurate


instrument and I measure the same pressure several times and I get the readings like this.
So, the Lowest is 9.80 and the Highest is 10. 45. Now, let us Define an interval of let us
say 0.05 kilopascal and we want to determine, How many readings fall in each interval
of 0.05 kilopascal? So, let us Define a quantity called Z which is number of readings in
each interval divided by Total Number of readings divided by width of interval.
So, we have these data from repeated measurement of a given pressure obtain kilopascal
using this Pressure gauge and now, if I plot the values using an interval of 105 and
calculate Z which is define as follows Number of readings in each interval by Total
number of readings divided by width of interval. We can plot bar graph with high Z for
each interval which will be a histogram. So, maybe I can get a histogram like this, may
be something like this where this is and this is Z. So, here plot it x versus Z where x is
the readings scale readings.
Now, area of a particular bar is numerically equal to the probability that is specific
reading will fall in the associated interval. the area of the entire histogram has to be 1 or
100 percent because, there is definite probability or there is a probability 1 that a
particular reading will fall anywhere, within the histogram. That is it will fall somewhere
between the lower value and the upper value that is somewhere between 9.80 and 10.45
it this belongs to if this is 10. 45. So, it will lie between anywhere, between here to here
that you can say with probability 1.
Now, if I take infinite number of readings here, I have taken a finite number of readings
each with an infinite number of significant t digits we can make chosen intervals as small
as we want. And still have each interval contain a finite number of readings, in this
particular plot we have taken an interval width of 0.5. Now, in this example as small as
we want, and still can expect to have definite number of readings that will fall in each
interval.
Now, if we do this then these steps in this graph will be smaller and smaller and the
graph will approach a smooth curve in the limiting case. So, all we are saying is if we,
increase the number of readings that it will large value if we repeat this experiments
infinite number of times. We can make the intervals extremely small and even then we
can expect that some readings will fall in each interval. In that case the steps will be

smaller and smaller and this plot this plot will be like, a smooth graph or smooth curve in
the limiting case. Now, if we take this limiting abstract case as a mathematical model for
the real physical situation then the function Z equal to fx is call the probability density
function. So, if we take large number of readings this histogram can be represented by
smooth curve and the functional relationship Z equal to fx can be considered as the
probability density function.
(Refer Slide Time: 29:03)

So, From the definition of the Z, the Probability of reading lying between a and b which
can be written as the Probability that x lies between a and b, is the area under the curve
so. So, the probability of reading lying between a and b is the area under the curve. So,
this said region this probability information is sometimes given in terms of Cumulative
Distribution Function the probability information can also be given in terms of
cumulative distribution function. Which is defined as, the Cumulative Distribution
Function capital F x equal to probability that reading is less than, any chosen value of x.

(Refer Slide Time: 32: 36)

So, this will be represented as, now most random errors have a Gaussian distribution or
normal distribution. Since, Normal Distribution or Gaussian distribution is one of the
most use full density function. There are other probability density functions as well. In
fact, there are many, but the normal distribution function or the Gaussian distribution
function, is one of the most useful probability density function.
The normal distribution function or the Gaussian distribution function is define as this, 1
over sigma into square root of 2 pi into e 2 the power minus x minus mu square divided
by 2 sigma square. Where mu is the mean value and sigma square is the variance which
we define previously. So, Gaussian distribution can look like this, the fact that most
random errors may have a Gaussian distribution is the consequence of a very important
theorem, call the central limit theorem. When you overlay many random distributions,
each with an arbitrary probability distribution different mean value and a finite variance
the resulting distribution is Gaussian.

(Refer Slide Time: 34: 44)

So, the Normal Distribution or Gaussian distribution which is defined as, where mu is
mean and sigma square is variance. So mu can be obtained as, and the variance. So,
given the probability density function P x mean is, defined as follows and the variances
is defined as this equation. So, if you look at the Gaussian or Normal Distribution
function it has two characteristic parameters. One is sigma which is variance; another is
mu which is mean. So, the shape of the Gaussian or normal distribution function is
determine by mu and sigma.
Now, we can have a Gaussian Distribution like this we can also have a Gaussian
Distribution like this. Obviously, both has same mean but, different variance or different
standard deviation. This distribution has small standard deviation whereas, this has large
standard deviation both has same mean. So, when the standard deviation is small it
indicates that, there is a high probability a reading will be close to mu or mean.
So, we have a distribution with small standard deviation it indicates that, there is a high
probability that the reading will be close to mean value mu, while if, the standard
deviation is large, the readings will be more scattered along the mean. It may be pointed
out here, that if you look at this mathematical definition of the Gaussian Distribution it
says that, there is a small probability that even a very large reading will also occur. But, a
real distribution will always have their tails cut off. So, instead of this a real distribution
will have will be, where their tails will be cut off.

However, most of the or many measurements the errors the random errors associated
with many measurements can be represented by a Normal Distribution Function. Now,
having defined some of the basic statistics let us, now see. How the error can be
propagated in a particular measurements? A particular instrument can have various
components. So, a particular instrument can be chain of components, the question we ask
is if we know the error or uncertainty associated with each component.
What will be the overall error or what will be the overall uncertainty measurement? You
can also think of a situation where, a particular measurement involves various other
measurements. So, particular experiment can involve various measurements. And we
calculate a value of a particular variable, from the measure from these measure values.
Now, if I know the uncertainty or error associated with each measurement. Can I
compute the overall error or uncertainty measurements.
(Refer Slide Time: 42: 52)

So, let us now see how we do that? So, the question you ask is, if an experiment involves
several measurements. Say, using several instruments and if, I know the individually
inaccuracy or individual error uncertainty. How do I combine this individual
uncertainties or errors to get an estimate of overall accuracy? So, now, consider the
problem of Computing a variable Y which is known to be a function of n independent
variables. So, we want to compute Y which is a function of n independent variables x1,
x2 up to xn.

So, we may are measuring x1 x2 up to xn. If we know the errors or uncertainties


associated with x1 x2 up to xn we want to compute the overall accuracy in Y. So, we say
Y is a function of x1 up to xn. For a small change in the independent variable x1 x2 up to
xn from a given say operating point. A Taylor series expansion will give an good
approximation for the corresponding change in Y. So, which can be represented as where
you can only the First-order terms. So, if there are change is small change is in the
independent variables x1 x2 up to xn around a given operating point. The corresponding
change in Y can be obtain from a Taylor series expansion of these function, which is
written as this where we are written only the First-order terms.
We can think of these partial derivatives, as the sensitivity of Y to changes in the
particular x. So, these are sensitivity of Y to changes in particular x. So, this is sensitivity
to change sensitivity of Y to change in x1 sensitivity of Y to change in x2 and. So, on
and so 4th. If partial derivative is large we say Y is very sensitive to that particular x.
Now, Consider delta Y to be the uncertainty in measured value Y. Which we represent as
let us, say similarly delta x1 to be uncertainty in measured value x1 which we
represented as u x1. Similarly, delta xn is to be considered as uncertainty in measured
value xn which will represent as U xn.
(Refer Slide Time: 50: 49)

Then we can write as, Uy is we have just, use this definition in this equation. Now, the
maximum value of uncertainty will be when all the uncertainties happen to have the

same sign. So, the maximum value of uncertainty can be taken as, would be obtained
when all the uncertainties happen to have the same sign. And this would be the worst
possible case; however, the probability of such an occurrence is generally very small.
Therefore, the more realistic way is to square both sides to give equal weight age to both
positive and negative values of uncertainties. So, more realistic way of representing
uncertainty will be, to square the terms, here we have Neglected terms like cross terms
like.
So, this will be More realistic uncertainty. So, we can now take overall uncertainty as,
where N is this x1 up to x2 up to xn, an number of independent variables. So, if you have
an estimate of the individual uncertainties, we can determine the overall uncertainty.
(Refer Slide Time: 54:20)

But, is the inverse problem which is more important. What do you mean by inverse
problem is as follows? When you originally plan an experiment we must decide, how
much accurate our measurement will be. So, we set our goal that our measurement must
meet this much of accuracy, and accordingly we need to know what will be the allowable
uncertainties in individual measurements.
Now, this problem is not. So, easy to solve because there can be various combinations of
these uncertainties which will give the same over all accuracy. So, we have to get started
somewhere. So, at this stage we can make use of something called Method of equal
effects, where we assume that all the instruments or all the individual components

contribute equally to the overall error. In other words the individual components
contribute equally to the overall error.
So, if we look at this equation, under the assumption of Method of equal effects. We can
now, write or so we can now, have an estimate of; what should be the accuracy of
individual component to meet the requirement of overall accuracy? If you can find
instruments which meet all these nets we have at least one solution to a problem. If one
or more requirements cannot be meet we should check if some of our instruments are
better than that the above equation requires. If, that is the case, then you can relax the
requirement that we cannot meet ultimately the overall accuracy must be meet.

Process Control and Instrumentation


Prof. D. Sarkar
Department of Chemical Engineering
Indian Institute of Technology, Kharagpur
Lecture - 40
Instrumentation: General Principles of Measurement Systems (Contd.)
In our previous class, we have discussed how to combine different component errors to
get an estimate of overall uncertainty in the measurement; in other words, since a
particular measurement can be made of various experiments. And each experiment can
have individual errors associated with it, so how do I combine this errors or uncertainties
to get an overall estimate of the uncertainty in my measurement.
Similarly, a particular instrument can a varies components and this individual
components may be associated with various uncertainties so how do I combine these
uncertainties to get an estimate of overall uncertainty in my measurement. So, today let
us take an example and see whatever we have discussed in the last class, how we can
apply this.
(Refer Slide Time: 01:28)

So, let us taken the example, the resistance of a certain size of copper wire is given as R
equal to R 0 into 1 plus alpha into T minus 20, so is a familiar equation where R is the
resistance of the copper wire, R 0 is the resistance at 20 degree Celsius. And given here

as R 0 is 6 ohm plus minus 0.2 percent, so this is the error plus minus 0.2 percent is the
error associated with R 0. Alpha which is temperature coefficient of resistance is given
as 0.004 per degree Celsius plus minus 1 percent.
If the temperature is an measured with an accuracy of plus minus 1 degree Celsius,
calculate the resistance of the wire at 40 degree Celsius and it is uncertainty. So, the
resistance of a wire is given as R equal to R 0 into 1 plus T minus 20, so this equation
tells us that if I know the resistance R 0 at T equal to temperature equal to 20 degree
Celsius from the knowledge of alpha I can calculate the resistance at 40 degree Celsius.
We need to find out, what is the uncertainty in this computation, for resistance at
temperature 40 degree Celsius if R 0 alpha and temperature are given with a specified
uncertainty. So, let us try to solve this problem, so if you remember our previous days
lecture, we calculated view overall as.
(Refer Slide Time: 03:51)

So, you will make use of this equation to find out what is the uncertainty in the
computation of R, which is resistance at 40 degree Celsius.

(Refer Slide Time: 04:31)

So, let us try to solve the problem, what is known is resistance at T equal to 20 degree
Celsius, let us say I want I have alpha equal to 0.004 degree, 4 per degree Celsius plus
minus 2 percent is the temperature co-efficient of the resistance. And temperature T is
measured with accuracy of plus minus 1 degree Celsius, so by use of this equations, we
can find out what is the value of the nominal resistance.
So, just we have to put values are R 0 alpha and T in this equation so if we do that R at T
equal to 40 Celsius will be 6 plus 4, which is the value of alpha 40 minus 20, if you
compute this it will come it is 6.48 ohm. Now, will make use of the equation that we
discussed in our last class to find out the uncertainty in the computation of resistance, let
us denote it by U R will be equal delta R delta R 0 U R 0 square plus delta R delta alpha
U alpha square plus delta R delta T U T square, and square root of this sum.
We wrote this, because we know from here that R is the function of R 0 alpha and T, so
while following whatever will be in our last class, we can obtain an expression for
uncertainty in R as this, where U R 0 is the uncertainty in R 0. U alpha is the uncertainty
in alpha and U T is the uncertainty in temperature, these values are given and this special
derivatives, we can find out from the functional relationship of R equal to R 0 into 1 plus
alpha into T minus 20, so let us do that now.

(Refer Slide Time: 09:19)

So, we need del R del R 0 which is nothing but, del del R 0 into R 0 into 1 plus alpha
into T minus 20, which will come out is 1 plus alpha into T minus 20. So, if I put the
numerical values 1 plus alpha is 0.004 and T is 40 minus 20, it will come as 1.08. Next
we need delta R delta alpha, again if we compute this it will be R 0 T minus 20, after
putting the numerical values, it will come is 20 into 6 120, finally del R delta T will be R
0 alpha.
So, 6 times 0.004 which is 0.024, so you now have values for delta R del R 0, delta R del
alpha and del R del T, we now need to know U R 0 U alpha and U T. U R 0, you
consider 0.2 percent R 0 is given as 6 ohm plus minus 0.2 percent. So, 0.2 percent of 6
ohm so 6 into 0.02 which is 0.012 ohm, U alpha is 2 percent of 0.004, so it will be 0.004
0.02, which will be 8 into 10 to the power minus 5 per degree Celsius.
If u remember, we said R 0 is given as 0.2 percent, alpha is given as 0.004 per degree
Celsius plus minus 2 percent and temperature is measured with plus minus 1 degree
Celsius accuracy, so 40 plus minus 1 degree Celsius. So, you obtain the values of U R 0,
U alpha and U T from this given values, U T is straight away known as 1 degree Celsius.
So, we now just kept to put all these values into the equation we wrote that means, in
((Refer Time: 13:30)) this equation.

So, now have values for all these terms, so if we put this values, it will be ((Refer Time:
13:56)) del R del R 0 is 1.08 and U R 0 is 0.012, del R del alpha and U alpha, we
computed del R del alpha is 120 and U alpha is 8 into 10 to the power minus 5 plus delta
del T and U T, we computed delta del T as 0.024 and U T as 1. So, 024 1 square, this
will be 0.0289 ohm, so the uncertainty in the computation of resistance at 40 degree
Celsius is 0.0289 ohm.
If you want to express this in terms of percentage, we can do that 0.0289 divided by 6.48,
remember 6.48 is the resistance and the nominal case into 100 percent, this will come is
0.45 percent. So, we now know how to compute the overall uncertainty in the
measurement, let us take an another example.
(Refer Slide Time: 16:46)

The governing equation for the capilary tube viscometer is well known, so let us consider
a capilary tube viscometer. The governing equation is the well know Hagen-Poiseuely
equation, which is Q pi D to the power 4 128 theta into L, where Q is volumetric flow
rate of fluid in the capilary. D is diameter of the capilary, theta is coefficient of dynamic
viscosity, L is length of capilary tube and delta P is pressure drop across two ends of the
capilary tube.
Now, the question you ask is, if Q, L, D and P are measured with an uncertainty of plus
minus 1 percent, how accurately the dynamic viscosity eta is known. So, you are

considering the capilary tube viscometer and performing an experiment to determine the
viscosity eta, the coefficient of dynamic viscosity eta, to determine eta you must have the
knowledge of volumetric flow rate of the fluid in the capilary. The diameter of the
capilary and the length of capillary tube, as well as the pressure drop is missing here
((refer Time: 20:05)), so this pressure drop.
So, this is the having Poiseuely equation, which can be used to determine the dynamic
viscosity eta from the knowledge of volumetric flow rate, pressure drop diameter and
length of the tube. So, the question you ask is, if the volumetric flow rate, the length of
the tube, diameter of the tube and the pressure drop are measured with an uncertainty of
plus minus of 1 percent, how accurately eta is known. We can follow the same approach
that we just followed to solve the previous problem, to solve this problem as well.
(Refer Slide Time: 21:13)

Eta is pi D to the power of 4 128 Q, L, delta P, just rearrange diagonal the HagenPoiseuely equation, so eta is the function of D, Q, L and delta P. So, the overall
uncertainty can be computed as L, now we can find out this partial derivatives from this
relationship; and all these values are specified as 1 percent that means, 0.01. So,
following the same approach we can determine, the overall uncertainty in the
measurement of dynamic viscosity eta.

(Refer Slide Time: 24:02)

If you find out all these partial derivatives, the final equation will look like, after putting
the numerical values you will get as or 4.36 percent. So, this is the uncertainty in the
measurement of dynamic viscosity, now if we ask again that the uncertainty in the
measurement of diameter is reduced to say plus minus 0.1 percent by using very
improved instruments. So, what will be it is effect on the overall uncertainty, first we
solved that when all the measurements of Q, D, L and delta P are available with plus
minus 1 percent accuracy.
We have the overall accuracy as 4.36 percent, now we say only the diameter is measured
with much more improved instrumentation, so that the uncertainty associated with the
measurement of the diameter is reduced to 0.1 percent. So, what will be the effect on the
overall uncertainty, if you do the same calculation while U D will now be 0.001, while
all other Us will be 0.1, so which is...
So, the uncertainty is reduced to 1.78 percent, if you calculate the percentage improvement
will be it was 4.36, which is about 59.17 percent. So, more than 59 percent improvement
in the overall measurement of dynamic viscosity, just by reducing the uncertainty in the
diameter measurement from 1 percent to 0.1 percent. If you look at the relationship eta
with the diameter Q L and delta P, you see heater where is with fourth power of diameter
that is the reason we see here very sharp decrease in uncertainty, when we reduce the
uncertainty in the diameter.

Now, let us close our discussion on error analysis by having some more discussion on the
probability distribution function, that we talked about in our last class. We say that the
measurement states that contain only random errors, conform to the Gaussian distribution,
so measurements states that contain only random errors easily conformed to Gaussian
distribution, also known as normal distribution.
(Refer Slide Time: 29:58)

We wrote the probability density function, as the normal distribution function, where mu
is the mean of the measurements and sigma is standard deviation, the maximum
probability occur at x equal to mu, the mean. And what will be the value of this
probability, you can obtain this if I put x equal to mu here, so the value of this probability
will be 1 by sigma square root of pi, because this entered value will be equal to 1.
So, the smaller value of sigma will produce larger values of this probability, and this is
expected intuitively also, so the probability x into mu is sometimes called measure of
precision of the data. We now one to determine, the likelihood that a certain data points
will fall within a specified deviation from the mean of all the data points. So, what you ask
is, we wish to determine the likelihood that certain data points will fall within a specified
deviation from the mean of all the data points. We can find this out using the definition of
the distribution function, the probability that a measurement will fall within a certain range
x 1 of the mean heading mu will be...

(Refer Slide Time: 33:31)

So, this is the probability that a measurement will fall within mu plus x 1 and mu minus x
1, so the probability that a particular measurement will fall between mu minus x 1 to mu
plus x 1, can be obtained from the definition of the distribution function, so if I make this
substitution, it can be shown that P is where... Now, the values of this Gaussian normal
error function which is and the integral are given or available in the form of tables, the
Gaussian normal error function.
And the integral of the Gaussian function are available as standard tables, we can make
use of that table to find out what is the probability, the particular measurement will fall
within say plus minus 1 standard deviation or plus minus 2 or plus minus 3 standard
deviation.

(Refer Slide Time: 36:47)

So, how do we do that, so let us ask ourselves, what is probability that are measured or a
measurement will fall within 1, 2 and 3 standard deviation of the mean value. So, we need
to determine what is the probability that a particular measurement will fall within mu plus
minus sigma, sigma is standard deviation, so we can make use how about we learnt here.
This is same as, so you want to find out what is the probability that a particular
measurement will fall within 1 standard deviation of the mean value, let us denote that by
this.
So, all you need to do is for eta 1 equal to 1, we need to look another table what is the
value of this integral, if you look at the table this value is 0.34134, which will come is
0.6827. Similarly, the probability that a particular measurement will fall within plus minus
2 sigma of the mean value, can be computed as 2 into the value of these, when eta 1 equal
to 2, which can be found from the table, whose value can be obtained as 0.47725, this will
be 0.9545.
Finally, P 3 can be similarly obtained as value of this, when eta 1 equal to 3 from the table
you can obtain the value as 0.49865, which will be 0.9973, so what we conclude is that the
probability, that a measurement will fall within plus minus sigma of the mean value is
68.27 percent. Similarly, if the measurement the probability that a measurement will fall
within 2 sigma of the mean value is 95.45 percent and the probability that a particular
measurement will fall within 3 standard deviation of the mean value is 99.73 percent.

All these computations are of course, based on the fact that the measurement states
conformed to normal distribution or Gaussian distribution, so this ends our discussion on
error analysis, will now move on to transducers.
(Refer Slide Time: 42:05)

Transducers are devices that transform signals in one form to a more convenient form, so
is a device which will receive a signal in one form, and will convert to another form which
is more suitable for the purpose of measurement. Please note that is not just conversion of
energy associated with that signal, it is not just conversion of energy in one form to
another, we can take a... Let us now talk about transducers, transducers are devices that
transform signals in one form to a more convenient form.
So, it is a device that receives signal in one form and converts that signal to another form,
which is more suitable for the purpose of measurement, it is not just conversion of energy.
Let us take example of diaphragm, you imagine a thin metallic diaphragm or thin non
metallic diaphragm, which will produce a displacement on application of pressure. So, the
diaphragm receives pressure as input and gives displacement as a output, so what is a
transducers, note that both displacement and pressure are manifestation of the energy but
displacement is more convenient from measurement point of view.
Transducers can be a various types, mechanical types, electrical types, optical types,
acoustic types extra, obviously electrical transducers are always preferred, because there
are a several advantages associated with electrical transducers. For example, the signal can

be conditioned easily may be the signal can be modified or amplified or modulated very
easily, it is also easy to transfer signal from one place to another, so electrical transducers
will left easy remote operation.
(Refer Slide Time: 46:16)

For the purpose of this course, we are concerned with electrical transducers or
electromechanical transducers that producers an electrical output, due to an input of
mechanical displacement or strain. So, in this course will mainly talk about few electrical
transducers or electromechanical transducers that produces an electrical output, when it
receives mechanical displacement or strain as input; additionally will also talk about just
one pneumatic transducers briefly.
The mechanical strain or displacement may be produced by primary sensor due to various,
input physical variables such as temperature, pressure, flow, extra. So, during act of
measurement primary sensors that receives input signal, from physical variable such as
temperature pressure flow extra may produce a mechanical strain or displacement. We are
interested in converting these mechanical strain or displacement into an electrical signal,
let us say this is a diaphragm, this is the diaphragm will talk about diaphragm in more
detail later.
When this receives pressure as input it deflects, so there is a displacement, if we measure
the displacement from the centre of this diaphragm, we have this distance as maximum
deflection of the diaphragm. We are interested in converting this displacements signal to

an electrical output, similarly of a force acts here it will defect, so this displacement has to
be measured. We are interested in some electromechanical transducers, which receives
these displacement as input and gives us an electrical output.
So, this mechanical displacement or strain can be an output from various primary sensors,
that receives temperature, pressure, flow rate, extra as input, so this mechanical
displacement or strain is received by some transducers as input and gives us an electrical
output. So, will learn few such transducers, which takes mechanical displacement or strain
as input and gives us electrical output.
(Refer Slide Time: 50:23)

So, will briefly discuss mostly electro mechanical transducers, but will talk about one
pneumatic transducers also and the pneumatic transducers will briefly talk about a flapper
nozzle system, is an important pneumatic displacement measuring transducer, is an
integral part of all pneumatic control systems. And electromechanical transducers will
briefly discuss four different types of transducers, namely linear variable differential
transducer, also known as LVDT, then resistance strain gauge, capacitive type transducer
and piezo-electric transducer.
The linear variable differential transducer or LVDT is an inductance type transducer, it is
working principle, is based on the fact the magnetic characteristics of an electrical circuit
changes due to motion of an object. The working principle of resistance strain gauge is

based on the fact, that if a conductor is stretched or strained it is resistance will change,
and it is easy to measure this change in resistance, so you can measure the strain.
Similarly, capacitive type transducers is based on the fact that, there is a change in
capacitance between two plates due to motion. Finally, piezo-electric transducer is based
on the fact that, an electrical charge is produced when a crystalline material, such as quartz
and barium titanate is distorted.
(Refer Slide Time: 53:02)

So, let us start with flapper nozzle system, flapper nozzle system is a pneumatic
transducer, we know the pneumatic control system operates with air. The signal is
transmitted in the form of variable air pressure, in the range of 3 to 5 psi, we have read
more about this during discussion on control part of this course. Early days we had only
pneumatic control systems, but with the advent of modern electronics, many pneumatic
control systems have now been replaced by electronic control systems.
However, even these days many industrial actuators are pneumatic in nature, there are
certain advantages associated with pneumatic control system, they are very safe, because
you handled layer only. It is chief, it also generates more torque to it is own weight
compared to electrical actuators however, the disadvantage of pneumatic control systems
is it is slower response.

(Refer Slide Time: 54:31)

This is a flapper nozzle system is the basis of all pneumatic transmitters, the flapper nozzle
system consist of a fixed flow restriction or orifice and a variable restrictor nozzle and
flapper. So, it is a very simple system consist of an orifice, which provides a fixed flow
restriction and a nozzle and flapper, which acts as a variable restrictor. Air at a fixed
pressure, let us denote the pressure by P s, flows through a nozzle first a restriction in the
tube.
So, the air at fixed pressure, flows through this restriction and through this nozzle, so due
to the phrases of a flapper, there will be a back pressure that will alter the output pressure
P 0, also called as signal pressure. Altering the gap between nozzle and flapper, alters the
resistance to airflow and hence, the output pressure increase in x will lower the resistance
that means, increasing the gap between the flapper and nozzle will lower, the flow
resistance and fall in the output pressure.
So, thus P 0 which can be varies the measured using a good pressure measuring instrument
and P 0 can be calibrated with the gap that exists between the flapper and the nozzle that
means, the displacement can be calibrated with the output pressure P 0. So, let us repeat
once again, the flapper nozzle system consists of a fixed flow resistance and a variable
restrictor, orifice is the fixed flow resistance, and a nozzle and flapper works as a variable
restrictor.

Here at fixed pressure, flows through this restriction and then, through the nozzle, because
of the pressure of the flapper are back pressure will be developed and has the gap between
the nozzle and the flapper changes this back pressure will change. Increase in the gap
between nozzle and flapper will lower the resistance and fall the output pressure, so the
output pressure can be directly calibrated, in terms of the gap between the flapper and the
nozzle that means, distance.
(Refer Slide Time: 58:08)

So, if I plot the output pressure verses the gap, you will see between the range of 3 to 15
psi pressure we have an approximately linear relationship, and that is the 3 to 15 psi
pressure is the working range which is linear. So, will end our discussion today with the
flapper nozzle system, and in the next class will talk about the other electromechanical
transducers like LVDT, capacitive type, piezo-electric type and the resistance strain gauge.

Вам также может понравиться