Вы находитесь на странице: 1из 37

Chapter 2

Inventory and Flow Control

The inventory controller balances demand and supply.


Greg Shinskey, an influential control engineer from the Foxboro company

Control: To have power, direction or influence over (e.g. The pressure of the steam engine
is controlled by this valve. You try to control me as if I were your slave.) (Longmans
Dictionary of Contemporary English)

Inventory: List of all goods in a place (a detailed inventory of all jobs to be done) (Longmans
Dictionary of Contemporary English)

Goal: To develop the concept of inventory and flow control chemical process
systems

2.1 Introduction
Many basic ideas applicable to process control were formalized in the context of business
management by the former Secretary of Defence Mr. Robert McNamara when he was the
CEO of Ford Motor Company:

1. Define clear control (business) objectives,

2. Develop control strategies (plans) to achieve the objectives,

3. Systematically monitor progress against the plan and make corrections (gap analysis)

4. Adapt the objectives and the control strategies as new information, needs and oppor-
tunities arise.

These ideas can by represented in terms of the information flow (block) diagram shown
in Figure 2.3. The diagram shows a process that can be affected by external actions by a
controller. The result from applying these actions to the process are measured and evaluated

40
41

disturbance
variable (d)
manipulated
variable (m)
Process
controlled
variable (y)

Figure 2.1: Signal flow (block) diagram of a generic process with a manipulated variable
(m), a disturbance variable (d), and manipulated variable (y).

provide feedback signals that are used to update the control control strategy (Feedback 2).
However, and as importantly, a second feedback loop evaluates the desired performance
and adapts this as needed. On the one hand the performance requirements should be
ambitious to provide the best performance possible. On the other hand they should not be
too aggressive since this gives unrealistic controls and instability. In order to achieve the
best performance possible it is necessary to model the process, obtain good measurements
of key performance indicators and good estimates of the environment and forecasts about
what kind of disturbances may impact the system in the future. Such disturbances may
include changing market conditions and different feedstock. The control system should also
be robust so that it can behave properly even when unexpected upsets (faults) happen.
Designing control systems that achieve all this objectives at the same time is at present an
unsolved problem except in very specific, small scale problems.
Inventory and flow control provides a good point of departure to introduce some of
these ideas into the real of process control and automation and to define the basic ideas of
feedforward and feedback control. The applications within our sphere of interest include
supply chains, reactor control, mass balance control, etc. Inventory control is not limited
to classical process control problems, however, and can be used in mechanical systems,
information systems, ecological systems, molecular dynamics, and finance.
Following the definition above we find that process control and automation concerns
the problem of how to manage the flows so that the critical inventories stay on target or
at least within certain bounds. To achieve this objective we must define how the system
given by equation (??) connects to the supervisory control and data acquisition system.
Equation (1.4.1) therefore includes a variable m called the manipulated variable or MV.
The presence of the MV allows us to change one or more flows using a control valve.
The variable y in equation (??) represents measured process variables, or PVs, such as
the pressure, flowrate or the temperature. The process variables are sometimes divided
into two distinct groups. PVs that are controlled to a setpoint (SP) are called controlled
variables, or CVs and variables that are measured but not controlled directly are referred
to as disturbance variables or DVs. Complex control systems are constructed by connecting
more such systems together into networks that may have 100s of MVs and 1000s of CVs.
Consider for example the continuous stirred tank reactor (CSTR) shown in Figure 2.2.
42

Supervisory Control and Data Acquisition (SCADA)

Temperature
Pressure
Flow
PT TT

Composition
FT
Feed CT
FT

Product
Control Valve

Figure 2.2: Schematic of a chemical reactor showing process flows (solid lones) and infor-
mation flow (broken lines) that connect the process to the supervisory control and data
acquisition (SCADA) system.

This figure, which shows the process and instrumentation diagram (P&ID), shows a diagram
of the process, the process flows as well as how the instruments and control valves are
connected to the automatic control system. In this particular case there are four flows
that connect the reactor to other equipment. Three of these can be controlled and they
represent degrees of freedom that must be set by the control system. The flowrate of
reactants, products and cooling water represent degrees of freedom that are controlled by
signals sent by the supervisory control system. In addition, there are five measurements
that indicate the state of the system. These include the flowrates of reactants and products,
the reactor pressure, temperature and outlet composition.
In a CSTR the objective may be to adjust flows to maintain the pressure, temperature
and reactor composition at given reference values or setpoints. These setpoints are chosen
so that production requirements are satisfied. The setpoints may change occasionally, but
in general they are left constant so that the reactor operates at a stable steady state for
considerable periods of time. The control problem for a stirred batch reactor (SBR) is a
little different. Such systems do not reach steady state and control valves may open and
close as chemicals are loaded initially according to a given recipe. Reactions then take place
while some chemicals, such as initiators or inhibitors are added while the reaction proceeds.
The temperature and pressure may be monitored and controlled and finally, when certain
conditions are satisfied, the reaction is complete. The reactor is then emptied out and a
new batch can be initiated.
43

Feedback 1

Desired Performance Objectives

Control strategy Process Measure Evaluate (Gap analysis)

Feedback 2

Figure 2.3: The flow diagram shows the structure of an adaptive control system.

MT MC

FT
min mout

Figure 2.4: Process and Instrumentation Diagram for a push system, where the production
rate is set upstream.

2.2 Mass Balance Control


We now show how we apply inventory and flow control to the overall mass balance in a
chemical process
dM
= ṁin − ṁout , M (0) = M0 (2.2.1)
dt

Control Objective: Adjust the flows so that the total mass (CV) converges to a constant

3
CV
Flow/Mass

DV
2 MV

1
Time
0 50 100 150

Figure 2.5: The response to setpoint change at t = 75 and disturbance change at t = 100.
44

MC MT

FT
min mout

Figure 2.6: Process and Instrumentation Diagram for a pull system, where the production
rate is set by downstream demand.

setpoint (SP) denoted by M ∗ .

The mass balance equation tells us how to achieve the objective. If there is too little
mass in the tank, then we need to change the net supply so that dM/dt is positive. If
there is too much then we need to choose the net supply so that dM/dt is negative. On
simple way to achieve this objective would be to set the net supply proportional to the error
between the measure mass in the tank and the set point so that

ṁin − ṁout = Kc (M ∗ − M ) (2.2.2)

where Kc is a positive constant called the proportional gain. The correction is fast if kc is
large.
The control equation has 2 degrees of freedom. It is possible to control the mass balance
in a number of different ways. We can choose the inlet flow to be the MV and the outlet
flow to be a DV or vice versa. We use whichever approach suits the present needs the best.

2.2.1 Push System - Supply sets production rate


In this example we allow disturbances (DV) enter the system via flowrate changes upstream.
The objective is to control the mass hold-up (the CV) using the flow rate out of the system
(the MV). The inventory control (2.2.2) with these assignments gives

ṁout = ṁin − Kc (M ∗ − M ) (2.2.3)


|{z} | {z }
feedforward feedback

This control system is illustrated schematically in Figure 2.1. The figure shows a valve
which adjusts the output flow rate, a measurement (MT) which measures the mass in the
tank, a measurement (FT) of the input flowrate, and a controller (MC) which represents
the control equation (??). The controller consists of two terms and works as follows:

1. The feedforward term measures the input flowrate and adjusts the output flowrate
before an error is developed to compensate for any change in the input flowrate.
45

2. The feedback term measures the setpoint error and adjusts the output flow until the
error is equal to zero.

2.2.2 Pull System - Demand sets the production rate


The same ideas apply when the input flow controls the inventory and when the demand
acts as a source of disturbances. In this case the inventory control becomes:

ṁin = ṁout + Kc (M ∗ − M ) (2.2.4)


| {z } | {z }
feedforward feedback

The controller consists of two components. The feedback component compares the CV
to the SP and creates an error which is used to provide a proportional adjustment. The
feedforward component measures the DV and feeds the signal directly to the MV according
to the idea that mass and energy flows to and from a system must balance at steady state.
How do we use such a control system in practical applications, with a real chemical
process rather than a mathematical construct? We have seen this situation in some of the
previous examples. In the pull system the inventory control comprised four items:

1. A measurement of the total mass M

2. A setpoint for the mass M ∗

3. A variable used for control ṁin

4. A disturbance variable ṁout

Figure 2.5 shows a simulation of the feedback control system with Kc = 0.1. The setpoint
os initially equal to 3. At time t = 100 it changes from 3 to 1. A small change in the MV
can be seen as the CV converges to the new setpoint. At time t = 100 the inlet flowrate
chages from 2 to 1. The feedforward component of the controller responds immediately and
there is no effect on the CV.

2.3 Energy Balance Control


In many chemical and bio-chemical processes it is appropriate to neglect changes in the
kinetic and the potential energy. We can then write the energy balance so that

X n
dU
= ṁi Ĥi + Q̇ + Ẇ (2.3.1)
dt
i=1

where

ṁ is the rate of mass transfer along connection i

Ĥ = U + P V̂ is the specific enthalpy


46

Stirring (Work)

Feed
1

2
Product
3 4
Cooling water

Figure 2.7: Schematic diagram of a stirred tank reactor.

Q̇ is the rate of heat transfer

Ẇ is the rate of work being added

The work is defined so that


dV
Ẇ = −P + Ẇs (2.3.2)
dt
where P is the pressure, V is the volume, and Ẇs is the rate of shaft work. The energy
balance controller can therefore be expressed so that
n
X
ṁi Ĥi + Q̇ + Ẇ = Kc (U ∗ − U ) (2.3.3)
i=1

The degrees of freedom are

ṁi the mass flow

Q̇ the heat flow


dV
dt the rate of volume change

Ẇs the rate of shaft work

In most chemical process control problems one or more of these terms can be manipulated
to manage the energy content of the system.

2.3.1 Energy balance control in mixer


Consider a mixer with two flows with different temperature that need to be mixed to achieve
the desired temperatures. This system is a pull-system, since we need to adjust feedrates in
order to meet a specific demand. An example of such a system consists of a shower, where
47

we adjust flow rates of hot and cold water to achieve desired flowrate and temperature.

Control Objective: Adjust the flow ṁ2 so that the flow out has the desired temperature
T ∗ . The flow ṁ1 is adjusted so that the mass hold-up is constant. The throughput ṁ3
varies.

The mass and energy balances for this system are


dM
= ṁ1 + ṁ2 − ṁ3
dt
dU
= ṁ1 Ĥ1 + ṁ2 Ĥ2 − ṁ3 Ĥ3
dt
There are three degrees of freedom corresponding to the three flows. the mass balance
controller satisfies.
ṁ1 + ṁ2 − ṁ3 = Kc (M ∗ − M )
The energy balance control system is modeled by the equation

ṁ1 Ĥ1 + ṁ2 Ĥ2 − ṁ3 Ĥ3 = Kc (U ∗ − U )

We will now assume that the mass balance controller has converged so that

ṁ1 + ṁ2 = ṁ3

Solving the energy balance controller for the control variable ṁ2 gives the controls

Ĥ3 − Ĥ1 Kc
ṁ2 = ṁ3 + (U ∗ − U )
Ĥ1 − Ĥ2 Ĥ1 − Ĥ2
For liquids we have U = U0 + cV ρ(T − T0 ) and H = H0 + cP ρ(T − T0 ) where subscript 0
denotes the reference state. Substituting these into the control law gives the control strategy
in terms of the temperatures so that
T3 − T1 Kc
ṁ2 = ṁ3 + (T ∗ − T )
T1 − T2 T1 − T2
ṁ1 = −ṁ2 + ṁ3

2.3.2 Energy balance control in a chemical reactor


Consider the stirred tank reactor shown in Figure ??. We want to develop a reactor control
system that removes heat so that energy follows a predetermined setpoint. Following the
procedure established in the previous chapter, we first write down the balance equation for
the extensive quantity we want to control, then we derive the inventory controller.

Control Objective: Adjust the rate of heat removal Q so that the energy content of the
reactor is stabilized at a constant setpoint U ∗ .
48

Like in the previous example, we assume that the mass holdup is controlled using a mass-
balance control system so that we have

ṁF = ṁP

where subscript F represents the feed and P represents the product.


By ignoring the work due to volume change and stirring, we find that the energy balance
of this system can be modeled by the equation
dU
= ṁĤF − ṁĤP + Q̇ (2.3.4)
dt

where Q̇ is the rate of heat-removal and the subscripts P, F have been dropped. There are
2 degrees of freedom corresponding to {ṁ, Q̇} available for control. Inventory control using
Q as control variable therefore reduces to

Q̇ = ṁ(−ĤF − ĤP ) + Kc (U ∗ − U ) (2.3.5)

This control system solves the energy balance control system exactly.
In some applications, we find that Q and ĤP are the disturbance variables. We can
then develop the energy balance control using the throughput ṁ so that


ṁ = − + Kc (U ∗ − U ) (2.3.6)
ĤF − ĤP

Example 2.3.1. Drum boiler power unit: Drum boilers are used to generate steam in a
wide range of applications, such as:

• heating houses by transferring heat from furnaces to radiators

• heating and cooling units in chemical plants

• generating electricity in fossil and nuclear power plants

Control of steam generating systems is difficult because many variables interact and
safety and environmental constraints must be simultaneously satisfied.
The schematic in Figure 2.13 illustrates how a steam-boiler in a typical power plant
works. Fuel and air is mixed and sent to the furnace for combustion.

Control Objective: The objective of the boiler problem is to coordinate the flow of fuel,
combustion air, and feedwater to meet the following demands:

1. Deliver steam to satisfy varying demand.

2. Stabilize the steam pressure at given levels.

3. Use the fuel as efficiently as possible.


49

Waste Heat
5 FT
Saturated steam
CONTROL LT
2

1
Cold water supply

Air supply 3
4 Fuel Supply

Figure 2.8: The water supply control system in a drum-boiler consists of a number of
components. A flowmeter measures the demand rate for steam. This system ensures that
the water supply i increased if the level is too low and decreased if the level is too low,
using a feedforward signal that allows the controller to adjust the feed water flow rate, and
a feedback signal.
50

4. Keep NOX at or below environmentally acceptable levels.

5. Maintain boiler water level within acceptable design constraints.

In the main part of the book, we will derive a control system which achieves all these
objectives. The main tenets of the theory can be illustrated using the mass balance system.
First we write the overall mass balance corresponding to the conservation law (??) so that
dM
= ṁw − ṁs
dt | {z }
|{z} rate of supply
accumulation

where
• M is the mass of water in the boiler

• ṁw is the mass flow rate in

• ṁs is the demand for steam


The objective of the mass balance control system is to choose the inflow of water so that
the mass is kept close to a reference value M ∗ (Objective number 5 above).
In this example ṁw is called a control, or manipulated variable, since this is a variable
we use to control the system. The inventory is called the output variable since it can be
used to indicate how close we are to the reference target. Finally, ṁs is called a disturbance
variable since it is set by the demand for electricity.
The mass balance control objective can be achieved by adjusting the net rate of addition
of water so that it is proportional to the deviation from the setpoint

ṁ − ṁ = −Kc (M − M ∗ )
| w {z }s
rate of supply

The positive constant Kc is called the proportional gain. The idea here is that the net rate
of supply is positive if there is too little water in the boiler and negative is there is too
much.
We can write the control equation by separating out the disturbance variable and the
control variable so that
ṁw = ṁs − Kc (M − M ∗ )
|{z} | {z }
feedforward feedback
This equation balances the water supply to the boiler against the demand so that the
objective is met.
A similar approach can be used to develop the energy balance control system. In this
case we use the energy balance (??) with Ẇ = 0 to get the control system

Q̇ = (ṁĤ)s − (ṁĤ)w − Kc (U − U ∗ )
| {z } | {z }
feedforward feedback
51

CA0
CA CB

Concentration

Residence time = (Dilution rate)-1

Figure 2.9: Residence time distribution for reaction A → B.

These control systems are based on the conservation of mass and energy, and are unique
because they are only ones that will satisfy the objective.
A number of practical questions arise when we use conservation laws to derive control
systems:

• How do we measure the water and energy inventories?

• How do we control the flow?

• How do we control the energy input?

• How to we manage the coupling between the mass and energy balance control systems?

• Are the controllers very sensitive to errors?

• Is it sufficient to use a proportional controller or should some other function be used?

• What if the information is delayed?

• Can these ideas be used in a general context when there are interactions and chemical
transformations?

• What happens when we have a network of feedback systems?

These questions will be addressed in this book. Look at Chengtao’s paper

2.4 Control of Chemical Reactions


Systems with reaction kinetics allow for one extra degree of freedom since it also matters
how long time the the material resides in the process. In this section we will see how
52

chemical reaction kinetics impose limitations on which concentrations are achievable in


reactive system. Consider again the stirred tank reactor with stoichiometry

A→B

The reaction is first order in A so that

pA = −V k1 cA , pB = V k1 cA

The plot of the concentration of the different constituents as functions of the residence time,
τR = V /F for a given inlet concentration cA0 of A is shown in Figure ??. This plot shows
as discussed before that we can control concetration of A within the range 0 < cA < cA0 .
We will consider two different objectives. In both cases we assume that the volume is well
controlled and that the density is constant so that we can write

F0 = F

where subscript 0 refers to the feedrate.

Control objective for reactor: Consider the problem of controlling the composition of
reactant A (the CV) close to a constant setpoint c∗A . Use the throughput rate F as a as the
manipulated variable (the MV). The inlet composition c0 is considered to be a disturbance.

From the mo balance developed in the previous chapter it follows that the controller is
written so that
F0 cA0 − F cA − V k1 cA = −K(NA − NA∗ ) (2.4.1)
Using F as the control variable and setting F0 = F gives
1
F = (V k1 cA − Kc (NA − NA∗ )) (2.4.2)
cA0 − cA
We can solve for F as long as cA 6= cA0 . The control is not well defined if cA = cA0 which
corresponds to the case where there is no reaction. This means that we cannot control
composition using throughput in non-reacting systems.
We now re-develop the control using a different choice of manipulated variable.

Control objective; Use the inlet concentration cA0 as the manipulated variable to control
the concentration of A to its setpoint. The throughput F is now a disturbance variable.

The controller takes the form


 
1 
c∗A0 = F cA + Vr k1 cA − Kc (NA − NA∗ )
F | {z } | {z }
feedforward feedback
53

We can solve for the control variable as long as F 6= 0. Which means that we lose control-
lability of the system if there is no throughput.
These examples illustrated how we use inventory control to control a concentration in
a stirred tank reactor system. Different choices of manipulated variables led to different
structures for the control system and finally we showed how the system might lose control-
lability.

2.4.1 System with input multiplicity


Many control systems have a property called input-multiplicity. In practice, this means that
that there is more than one way to achieve the control objectives. Input multiplicities are
especially prevalent in problems where we have to generate a control path. For example,
when we drive a car from one point to another, there is often a choice of different paths
that can be chosen. One path may use the least fuel, another path may take the shortest
time, and a third one may be the shortest distance. Sometimes, all three coincide; but they
often don’t. To find the best solution, we therefore need to define the system as well as the
objective that needs to be satisfied. Do we want the shortest time, the least fuel, or maybe
some combination of the two.
Example 2.4.1. Composition control in a chemical reactor Consider again the stirred
tank reactor as shown in Figure 2.7. The reaction has the stoichiometry

A→B→C

The reaction is first order in A and B so that

pA = −V k1 cA , pB = V k1 cA − V k2 cB

The reactor is well mixed, the volume V and the density ρ are assumed to be constant.
A plot of the concentration of the different constituents as functions of the residence time,
τR = V /F for a given inlet concentration cA0 of A was shown in Figure 1.18. The plot and
discussion showed that there was a unique relationship between the the concentration of A
and the flowrate whereas this was not true for B.

Control objective A: Consider the problem of controlling the composition of reactant A


close to a constant setpoint c∗A Use the throughput rate F as a control variable.

The control was developed in the example above where we showed


1
F = (V k1 cA − Kc (NA − NA∗ ) (2.4.3)
cA0 − cA
We can solve for F as long as cA 6= cA0 . This control is well defined. Suppose now that we
are interested in controlling the concentration of B. We state the objective

Control objective B: Consider the problem of controlling the composition of reactant B


54

close to a constant setpoint c∗B . Use the throughput rate F as a control variable.

The control expression is quite easy to derive and we get


1
F = (V k1 cA − V k2 cB + Kc (NB − NB∗ )) (2.4.4)
cB
The control law is well-defined.
However there is a problem in that that the result from applying the control is ambigu-
ous. The problem can be understood by solving the for the steady state. We find

F/V
cA = cA0
F/V + k1
k1 F/V
cB = cA0
(F/V + k1 )(F/V + k2 )
k1 k2
cC = cA0
(F/V + k1 )(F/V + k2 )

Figure 1.18 shows that the inventory of B increases as a function of residence time until it
reaches a maximum where
dcB
=0

where τ = V /F is the residence time. We used

k1 = 1, k2 = 1/6, cA0 = 10, F = [50, ∞), V = 500

Thereafter B decreases since it is getting more and more converted to C. For each setpoint

0 < c∗B < cmax


B

there are two different flows that give zero setpoint error. The high flow rate corresponds to
low conversion since the reaction does not have time to proceed very far. The low flow rate
corresponds to high conversion since the residence time is larger than that corresponding
to the maximum. The controller will reach one of these points, although which is reached
depends on the choice of the initial conditions.
The only way to cure the problem is to give the controller better information about
which point it needs to control to. The easiest is to chose a control variable which has a
monotone dependence and does not display multiplicity, for example the concentration of
A and convert the desired setpoint for B into an equivalent setpoint for A.
2
The problem of input multiplicity can arise in flash systems, distillation, ,chemical re-
actors and biological systems.
55

2.4.2 Systems with Output Multiplicity


Some chemical process systems have a property called output multiplicity or multiple steady
states. In these these systems the steady state output may be ambiguous since it can show
different steady states for the same set of input variables. The property is often illustrated
by the first order dissociation reaction with Arrhenius rate expression

A → B + C, rA = −V k0 e−E/RT cA

The heat released by the reaction can be written so that

Qgen = V (−∆H0 )k0 e−E/RT cA

where −∆H0 is called the heat of reaction and represents the difference in enthalpy of
products and reactants. In the case of an endothermic reaction we have −∆H0 < 0 and
in the case of an exothermic reaction we have −∆H > 0, indicating weather the products
have lower or higher enthalpy than the reactants.
We assume that the volume, density, and heat-capacity are all constant. The component
and energy balances are written
dV cA
= F c0 − F cA − V ke−E/RT cA
dt
dρcP T
= F ρcP (T0 − T ) + (−∆H)V ke−E/RT cA + U A(Tj − T )
dt
The inventory control can now be developed. We can for example consider the following
objective.

Control objective: Use the jacket temperature Tj to control the temperature.

The controller can then be written


1
Tj = T (F ρcP (T0 − T ) + (−Ĥ)V ke−E/RT cA )
UA
Other objectives can also be considered. We can for example control the concentration of
A using the flowrate as the manipulated variable.
The steady state is found by solving the following coupled nonlinear equations

0 = F c0 − F cA − V ke−E/RT cA
0 = F (T0 − T + V (−∆H0 )k0 e−E/RT cA ) + Q

These coupled nonlinear equations can be solved numerically.


56

6
x 10
14

12

Heat Generated/Removed
10
C
8

B
6

4
A
2

-2
50 60 70 80 90 100 110 120 130 140 150
Temperature

Figure 2.10: Heat generated by reaction and heat removed is balanced at three points,
showing multiple steady states.

Example 2.4.2. Consider the following example


k0 = 10
−∆H = 100
ρcp = 1
E/R = 3000
V = 500
F = 1
U A = 0.11
cA0 = 1
T0 = 350
T j0 = 300
We need to do a simulation here

2.4.3 Systems that do not converge to steady state


In a managed forest there may be harvesting of hares and foxes. In this case, we may write
the balance equations so that
dH
= aH − bHF − uH
dt
dF
= −cF + d(bHF ) − uF
dt
57

Inventory
Feed forward inputs
fi p + Z
-
Setpoint
Error Z*
e= Z-Z*

Feed forward Feedback


Controller Controller

Bias Feedback

+
fb fb
+

Controller output
f

Figure 2.11: The control algorithm consists of two components. The feedforward component
reacts to external disturbances, the feedback part reacts to the errors.

Here uH and uF are the rates at which the hares and the foxes are harvested.
By careful management the fish and wildlife service can now control the populations.
This way they enter different cycles or become steady by allowing more or less fox and
hare hunters in the forest and carefully balancing harvesting against growth and even flight
across the boundary of the forest. The idea here is that we compare the hare and fox
populations with their respective setpoints and then we adjust the hunting licences up or
down proportional to the errors. Thus, if there is an excess of hares in the forest then we
do not simply balance hunting versus the net growth, but we also hunt more. Similarly, if
there are two few foxes in the forest then we issue proportionally fewer hunting licences.
In this way we can stabilize the inventories of hares and foxes in the forest and control the
cyclic behavior which arises in the absence of control.
The corresponding inventory controllers become

uH = aH − bHF + K1 (H − H ∗ )
uF = cF + dbF H − K2 (F − F ∗ )

Using this control strategy we can stabilize the Fox and Hare populations at any desired
level by adjusting the harvesting rate uH , uF . On
58

Connections to other systems

fk
y* Φ(m,d)
+ + m
Feedback Control Chemical
- Calculation Process
-
Measurements y
p d
Measurement system
y

The Control System

Figure 2.12: Inventory control system.

2.5 Structure of the Inventory Control Algorithm


Going back to the inventory balance and the examples above we see how we can apply
feedback control to general systems
dZ
=p+φ (2.5.1)
dt
where
n
X
φ= fi
i=1
where fi is flow number i to the process. In this case, the objective is to control inventory
to its setpoint. The net supply rate p + φ controls the direction of the process. If p + φ is
positive, then the inventory increases and conversely if p + φ is negative then the inventory
decreases. If p + p + φ = 0 then the inventory remains constant. This approach gives
Choose m so that : φ(m, d) + p = Kc (Z ∗ − Z)
The number Kc is positive and it is called the proportional feedback gain. The proportional
gain does not have to be constant. It could be a function of other variables and a number of
other possibilities exist. The most important point is that the controller has to be chosen
so that it does not reverse the sign so that we get positive rather than negative feedback.
The strategy is shown schematically in Figure 2.11.
The inventory control system has the form:
φ(m, d) = −p + Kc (Z ∗ − Z)
To use this expression in practice, it is necessary to connect the control system to the
process using measurements (y) and manipulated variables (m) as shown in Figure 2.12. The
59

different blocks in the figure serve different purposes. The measurement systems calculate
the current inventory Z, the disturbances d, and the current rate of production p. The
calculated inventory is compared to the setpoint, and the error is used in the feedback
control law. Finally, the control input is generated by subtracting the production p and
inverting the function φ(m, d) to find the control input.
The accuracy of the control depends on how well all these signals are estimated how
quickly the information in processed. Later in the book, we will address these questions
and show that the feedback control system can function very well even if the information
is not very accurate.
Example 2.5.1. Mass balance control. We have a reactor with the overall mass balance
dM
= ṁin − ṁout
dt
We measure the level h so that
h = M/(Aρ)
where A is the area and ρ is the density.
The flow ṁin is the DV. The flow out is defined so that

ṁout = kv m h

where kv is the valve constant and m is the MV. m = 0 indicates that the valve is closed
and m = 1 indicates that the valve is open fully open.
The objective is to control to a the setpoint M ∗ . The control equation (??) can then
be written √
ṁin − mkv h = Kc (M ∗ − M )
Solving this for the MV gives
1
m= √ (ṁin − Kc ρ(M ∗ − M ))
kv h
where
M = Aρh, M ∗ = Aρh∗
relates the inventory to the measured variable h.

2.6 Multivariable Control Systems


We say that a control system is multivariable if the conservation laws interact. Sometimes,
this gives a more involved control strategy. The problem has been encountered in many
earlier examples, but it has not led to greater difficult in defining the control law. The
population balance example developed demonstrates interaction between the populations.
This interaction was taken into account when we developed the control systems for hares
and the foxes. We saw that the feedforward terms shared information. Systems where two
or more control control systems share information are said to be multivariable.
60

Information sharing is relatively straightforward if it only occurs in the feedforward


terms in the control law, as was the case in the fox and hare problem as well as the other
problems we have encountered so far. It becomes more difficult when it happens in the
feedback path as the following examples show. In these cases, we may have to solve for the
control variables simultaneously to achieve the control objectives. The method is quite easy
to develop. Suppose we have n inventories Zi to be controlled simultaneously. Problems of
this type are said to be multivariable. We now arrange the inventories in a vector Z so that

Z = (Z1 , ..., Zn )T

The corresponding conservation laws are written


dZ
=φ+p (2.6.1)
dt
where p = (p1 , ..., pn )T is the vector of production/consumption terms and φ = (φ1 , ..., φn )T
is the vector of net flows and each net flow is written so that
n
X
φj = fj,i (u, d)
i=1

We have indicated that each flow and production may depend on the inventory levels, as
well as manipulated variables and disturbance variables.
The inventory control law is therefore given by
dZ ∗
φ+p− = Kc (Z − Z ∗ ) (2.6.2)
dt
In this case Kc can be a matrix of proportional gains so that
 
K11 ... K1n
 
K=  . ... . 

Kn1 ... Knn

Where K should be symmetric and positive definite. Most often we achieve this objective
by setting the off-diagonal terms to zero. A more complete discussion follows in a later
chapter.

2.6.1 Multivariable Reactor Control


Consider a stirred tank chemical reactor with mass and energy balance
dM
= ṁin − ṁout
dt
dU
= (Ĥ ṁ)in − (Ĥ ṁ)out + Q
dt
61

The mass balance is coupled with the energy balance, but not vice versa.

Control objective: Adjust the input mass flow ṁin and the heat input Q so that the
mass and energy converge their setpoints M ∗ and U ∗

To achieve these objectives, we must manipulate the mass and energy inventories simulta-
neously. The mass balance controller gives

ṁin − ṁout = −K1 (M − M ∗ )

Using the same ideas we get the energy balance controller

Q̇ + (ṁĤ)in − (ṁĤ)out = −K2 (U − U ∗ )

The feedback gains for the mass and the energy balance controllers need not be the same,
indicating the closed loop response time for one loop can be fast while the other is slow or
vice versa.
We now follow the approach for the single variable control method and move the distur-
bance variables to the right hand side of the equation so that we get a composite controller
with a feedback and feedforward term so that
      
1 0 ṁ 1 K (M − M ) ∗
   out  =   ṁin +  1 
Ĥout 1 Q̇ Ĥout −K2 (U − U ∗ )

The matrix on the left is easy to invert. We have:


 −1  
1 0 1 0
  = 
Ĥout 1 −Ĥout 1

We can therefore write the energy and mass balance control so that

ṁout = ṁin + K1 (M − M ∗ )
Q̇ = (−Ĥout + Ĥin )ṁin + Ĥout K1 (M − M ∗ ) − K2 (U − U ∗ )

2.6.2 Steam Utility


In many chemical processes and powerplants, steam is used as a medium to heat processes
and drive turbines to generate electricity. The schematic of a typical boiler is illustrated in
Figure 2.13.

Control objectives: Supply a varying demand of saturated steam at constant pressure


while maintaining the total mass of water in boiler at its setpoint.

According to Gibbs phase, rule we know that the number of degrees of freedom (nF ) in
62

Waste Heat
5
Saturated steam

1
Cold water supply

Air supply 3
4 Fuel Supply

Figure 2.13: The steam boiler supplies saturated steam

a thermodynamic system is equal to two plus the number of components (nc ) minus the
number of phases (nP ) so that

nF = nc − nP + 2 = 1 (2.6.3)

That means that we have specified the state of the saturated steam leaving the boiler
completely if we specify one thermodynamic variable, e.g. the internal energy. We can
therefore stabilize the pressure by controlling the total mass and the internal energy of the
boiler.
We now develop the inventory controller for the energy and the mass in the boiler
system. The mass balance satisfies the equation
dM
= ṁ1 − ṁ2 (2.6.4)
dt
The energy balance satisfies
dU
= ṁ1 Ĥ1 − ṁ2 Ĥ2 + Q (2.6.5)
dt
There are 2 differential equations and 3 independent flow variables (M, U, ṁ1 , ṁ2 , Q so that
leaves 2 DOFs. The mass flow of steam out of the boiler is set by the demand for steam
so this is a disturbance variable. This leaves two control variables. The water to the boiler
has to be set so that water inventory is kept constant and the fuel rate has to be controlled
so that energy of the system is maintained at a given setpoint. Energy is directly related
to the pressure (recall the phase rule) so that controlling energy is equivalent to controlling
the pressure. We follow the standard procedure for inventory control and put the control
63

variables on the left of the equality sign and the disturbance and feedback variable on the
right. We therefore have the following inventory control system for the boiler

ṁ1 = ṁ2 − K1 (M − M ∗ )
ṁ1 Ĥ1 + Q = ṁ2 Ĥ2 − K2 (U − U ∗ )

The mass balance controller corresponds exactly to the pull system and can be used as
before. The energy balance controller depends on the mass balance control system, and
this is what we call multivariable control. In this case, there is one way interaction since
the energy balance depends on the mass balance but not vice versa. In the next example,
we see a case of two way interaction.
With one way interaction it is quite easy to get an expression for the interacting control
system since we can simply use the result from the mass balance controller and substitute
it into the energy balance to get the final result for the energy balance controller so that

Q = (ṁ2 Ĥ1 − Ĥ1 )ṁ2 − K1 Ĥ1 (M − M ∗ ) − K2 (U − U ∗ )

A typical set of enthalpy values for a typical Rankine cycle power plant is given in Table
2.14. What we see here is that the enthalpy of water is so much less than that of the steam
that the interaction between the mass and energy balance can be neglected. We can replace
energy feedback by pressure feedback so that we get the control system

Q = (ṁ2 Ĥ1 − Ĥ1 )ṁ2 − K2 (P − P ∗ )

where Q is the command signal that goes the fuel flow rate. Note that this control signal
has units of energy (kJ or more often BTU) and must be converted to a fuel flow before
it it used. The air flow rate is given by a ratio control so that the air and fuel mix is set
so that we get complete combustion, in the case of a coal plant this means a 2:1 oxygen to
carbon. EPA regulations set caps on NOX so that it is necessary to run the plant with a
slight excess (about 2%)to ensure that NOX is below these limits. This is normally achieved
using an O2 analyzer at the stack.
The steam is further heated in a sequence of superheaters where the temperature is
brought as high as possible to stay within the limits set by the turbine manufacturer. The
reason the temperature should be as high as possible is motivated by the Carnot efficiency
Tmax − Tmin
η=
Tmax
which states that efficiency is maximized when the temperature is as high possible. A
change in a few degrees can translate into big fuel savings.
We will get back to the practical implementation of these control schemes later in the
chapter when we introduce the idea of measurements and manipulated variables. We will
relate the mass and energy hold-up to level and temperature or pressure measurements
using thermodynamic correlations and introduce the concept of uncertainties in the data.
For example, in the boiler example there is uncertainty in the exact combustion heat of
64

Stream State enthalpy


ṁ1 Inlet water 25◦ C 104.9 kJ/kg
ṁ2 Saturated steam at 20MPa 2409.7 kJ/kg

Figure 2.14: Thermodynamic data for boiler control system

coal. In the next Chapter we will discuss how interconnected processes and entire chemical
plants can be controlled in a systematic manner so that all degrees of freedom are specified
and no flow conflicts arise.
Example 2.6.1. Thermostats in molecular dynamics Consider a body with mass M
moving in a gravitational field. The height measured from the ground is equal to h. The
first law of thermodynamics gives
d(K + P )
=0
dt
K = 12 M v 2 is the kinetic energy, v = dh/dt is the velocity, P = M gh is the potential energy
and g is the gravitational constant. Using these definitions we have
 2 !
d 1 dh
M + M gh = 0
dt 2 dt

Taking the mass as constant gives


 
d2 h dh dh
M 2 + Mg =0
dt dt dt
Hence
d2 h
M = −M g
dt2
The equation expresses the observed fact (by Sir Isaac Newton) that the force is equal to
the mass times acceleration. We write this law of nature, which we derived from thermo-
dynamics, so that
F = Ma
where a = d2 h/dt2 and F = −M g.
It is often convenient to re-write the conservation law for a moving particle in a potential
field as a coupled system of first order equations, called Hamilton’s equations so that
dh p
=
dt M
dp
= −p + F
dt

Where p = M v denotes the momentum. The main advantage of this formulation is that it
is coordinate free.
65

Hamilton’s equations form the basis for a field of study called Non-Equilibrium Molecular
Dynamics (NEMD). In this approach the position and momentum of each molecule in a
mixture is modeled. F represents the force that acts on the particle. It is general a function
of the distance to other particles. In order to control the system we also introduce a friction
parameter α and the strain rate χ.
For an MD system with N molecules then we have one equation for each molecule so
that the complete system becomes
dri pi
= + χri
dt M
dvi Fi
= − + χvi − αvi
dt Mi
dV
= 3χV
dt
where the vector ri denotes the position and the vector vi = pi /Mi denotes the velocity of
molecule number i in a three dimensional space. Thus, there are 6 differential equations
for each and every molecule in a system to keep track of its momentum and position and
potentials that indicate how strongly their dynamics are influenced by other molecules in
the system.
The internal energy of a system of N particles is defined to be the total kinetic and
potential energies for the entire system so that
n
1X
U= Mi vi2
2
i=1

From this equation we can develop an equation for the internal energy
dU X dvi
= Mi
dt dt
i

Hence
dU X
= −Fi + Mi χvi − αMi vi
dt
i

The inventory controller for the friction rate in an MD simulation


P
−Fi + χvi dU ∗
α= i + K(U − U ∗ ) −
Mi v i dt
Such an expression is called a thermostat since we effectively control the temperature of the
system by controlling the internal energy. Similar controller can be developed for the strain
rate. Thermostats are used in NEMD simulations to investigate the behavior of different
ensembles. The controller above stabilizes the so-called macro-canonical ensemble {U, V, N }
(constant energy, volume and number of molecules).
2
66

2.7 Control of Quality Variables


In many applications we need control quality variables, like temperature, pressure, com-
position, pH or some other function of the inventories rather than the inventory itself. In
order to apply the theory developed so far we must then write the variable we are interested
in controlling as a function of one or more inventories so that

y = h(Z1 , Z2 , ..., Zn )

where Z is the inventory and y is the variable we wish to control. We call the quality
variable y the output.
For example, if we are interested in controlling the mol fraction xA of A rather than the
total hold-up of A then we must define the relationship for the mol-fraction so that
NA
xA =
N
where NA is the molar hold-up of component A and and N is the total molar hold-up. In
this example y corresponds to xA , Z1 to NA and Z2 to N . In another application we may
be interested in controlling pressure in a system which can be modeled as an ideal gas with
constant volume and temperature so that
NR
P =T
V
This example is simpler since the variable of interest is a function of one inventory only,
namely the molar hold-up N .
The balance equations are now written
dZ
= p(Z) + φ
dt
y = h(Z)

The first equation gives the balance equations we are familiar with. The second equation
relates the output y to the inventory using the function h(·). Examples of output func-
tions include temperature, pressure, composition, molecular weight and level in a tank.
Sometimes the output function is directly related the quality of the product and must be
controlled. Other times the choice of output function may be dictated by the availability
of the measurements.
We now derive the theoretical basis for controlling output functions. Suppose that the
output variable y is a function of n inventories so that

y = h(Z1 , ..., Zn )

and that the setpoint for y is given by the variable y ∗ . The objective is to derive the
controller which controls the error e = y ∗ −y to zero.
67

Using the chain rule we write


n
dy X ∂y dZi
=
dt ∂Zi dt
i=1

By using the inventory balances introduced above we get the control expression for y exactly
as it was developed for the inventory control:
 
Xm n
∂y X dY ∗
fji + pi  = −K(y − y ∗ ) +
∂Zi dt
i=1 j=1

This expression can now be solved for the control flows exactly in the same manner as we
did for controlling the inventories.
In many process control applications it is important to control the composition rather
than the inventory of a chemical species. The same ideas apply once we note that in well
mixed system the relationship between inventory and composition is given by can write

V cA = NA

where NA is the number of moles of component A, V is the total volume and cA is the
composition. We then note that we can derive the differential equation for composition so
that
dNA dcA dV
=V + cA (2.7.1)
dt dt dt
The term on the right hand side vanishes if the volume is constant.

2.7.1 Pressure control


Suppose that we want to control the pressure in a push system where the fluid can be
modeled as an ideal gas with constant temperature. The resulting control system is shown
in Figure ?? with the mass measurement is replace by a pressure measurement.
We first need to develop the differential equation for the pressure. The ideal gas law
gives
P V = N RT
Hence
R
dP = (T dN + N dT )
V
The temperature variations are equal to zero so that we can write dT = 0. From the mass
balance (2.2.1 with n = 2 and the relationship M = Mw N we write

dP RT dM
=
dt V Mw dt
RT
= (ṁin − ṁout )
V Mw
68

Stirring (Work)

F1
1

cA0, F2 cA, F3
2 3

Figure 2.15: The mixer supplies a stream with constant concentration

We now chose the control variable so the pressure is equal to the setpoint P ∗ , namely
V Mw
ṁout = ṁin + K (P − P ∗ )
RT
If the temperature varies considerably then and it is necessary to achieve accurate pres-
sure control the we need to include the energy balance as well. We note that for an idea
gas we have
dU = cV dT
By using the expression for the pressure given above we get
 
dP R dM dU
= T + P cV
dt V Mw dt dt

We now get a composite control law which involves the energy and mass balances.

2.7.2 Composition control in a mixer


Consider the system explained in Figure 2.15 where the problem is the following

Control objective for mixer: Control input flow rate F2 so that the output concentration
cA is kept at the setpoint c∗A when the flow rate F3 varies. The volume of the mixer is
constant and the density differences among the different streams are negligible.

The assumptions of constant volume and density gives

F1 + F2 = F3 (2.7.2)

F3 is the disturbance variable and F2 is the control variable. There are 2 degrees of freedom.
It remains to the develop the expression for the feedback controller.
69

The control objective requires us to write the differential equation for the composition
of A. Using expression (2.7.1) with constant volume we get

dcA 1 dNA
=
dt V dt
We now need the the conservation law for A. It is written
dNA
= F2 cA0 − F3 cA
dt
Combining these equations we get the balance for composition so that
dcA 1
= (F2 cA0 − F3 cA ) (2.7.3)
dt V
We can therefore write the control equation (2.7.2) so that

cA Kc V
F2 = F3 − (cA − c∗A ) (2.7.4)
cA0 cA0
This equation is well-defined, meaning that we can solve for F2 as long as cA0 6= 0. If
cA0 = 0 then it is impossible to control the composition of A.

Challenges: Develop a control system which controls composition using dilution with the
pure stream F1 . In this case you need the balance (2.7.2). Develop a control system which
controls composition using stream F1 without assuming constant density. In this case you
use a mass or mole balance instead of the volume balance (2.7.2). Develop a mixer control
system with a time varying composition setpoint.
70

CV -Flow measurement SP - Flow setpoint

FC

FT MV - Valve position

Figure 2.16: The P&ID of a flow controller.

Appendix: Measurements and Flow Control


In this chapter we show how feedback control is used to reduce the effect disturbances
and model uncertainty by using feedback control. The idea is very simply this: The flow
controller adjusts the valve position until the correct flow is achieved as seen in the P&ID
in Figure 2.16. The setpoint (SP) provides the desired flow rate. It is compared with
the measured flow (CV for controlled variable) and the manipulated variable (MV) is then
adjusted until the error between the SP and the CV is eliminated.
One very typical example is provided by the hot-water valve in a shower. In this case
the user turns the valve one way or the other until the right temperature is achieved.
In some case adjustments need to made because supply pressure may change. The desired
temperature may also change as the showering process proceeds. The entire system therefore
consist of three interlinked components, the valve, the temperature sensitive nerves in the
skin and finally the brain which acts as a control system. These signals are connected
through mechanical devices, fluid flow, neuron transport and muscular activities. Happy
showering!

2.7.3 Flow Measurement


In the chemical plant there exist many different types of flow transmitters (FT). The most
common FT use an orifice to establish a pressure difference which is correlates with the flow
rate in a unique manner (Figure 2.17). This correlation depends on a number of factors,
such as the density, viscosity, temperature, pressure of the fluid as well as the geometry of
the pipe, the orifice and how the pressure sensors are placed relative to the orifice.
The Bernoulli equation provides a simple correlation between pressure and mass flowrate
so that p
f = CEπ 2rρ∆P
The parameter C is the orifice flow coefficient, r the radius of the orifice hole, ρ the density
of the fluid, ∆P = P1 − P2 > 0 the pressure difference across the orifice and E is the
71

FT   Descrip/on  
A102    
Point  Iden/fica/on  

Figure 2.17: The orifice flow meter correlates flowrate with the pressure difference across
an orifice. The figure on the right illustrates how the measurement is represented in the
piping and instrumentation diagram. The type of measurement is indicated, in this case
it is a flow transmitter (FT). A unique point identifier is provided to show exactly which
measurement it refers to in the P&ID.

expansion coefficient. For an incompressible fluid E = 1. For isentropic expansion of an


ideal gas we have  
1 − P2 /P1  
E =1− 0.41 + 0.35β
cp /cv
where 0.2 < β < 0.75 depends on the type of gas being measured. The measurement is
calibrated by adjusting the orifice flow coefficients E and β.
The activity of relating the pressure difference between the pressure difference and fluid
flow is referred to as calibration. In some cases it is sufficient to calibrate within ± 5 %.
In other applications it may be necessary to achieve much more accurate results and better
measurement technology may be required.

2.7.4 Valves
Many different kinds of devices control flows in chemical processes. These include variable
speed pumps, compressors, conveyor belts or even trucks. In this section we consider a flow
control system that determines the flow by adjusting the resistance to flow. It acts like a
variable resistor in an electrical circuit. A diagram of such a valve is provided in Figure 2.18.
This valve work on the principle that the flow through valve is a function of the pressure
difference ∆Pv = P1 − P2 across the valve so that

f = mf (∆P )

where m is the manipulated variable (MV) so that m = 0 corresponds to a closed valve and
m = 100% maximizes the flowrate. We may for example have a square root dependency so
that √
f = kv m ∆P
where kv is the valve characteristic, which depends on the type of fluid, valve geometry and
pressure temperature.
72

Handle  
Valve  stem  

Valve  seat  

Figure 2.18: The figure on the left shows the working principle of a globe valve. The handle
can be turned clockwise or anti-clockwise to open or close the valve. The valve opening is
often expressed in terms of a percentage value with 100% representing the maximum flow.
The figure below shows the P&ID symbol for the manual valve. The automatic control
valve is shown on the right. In this case the valve opening is controlled by supplying air
pressure to a diaphragm attached to the valve stem. The corresponding P&ID symbol is
shown below.
73

Figure 2.19: The valve characteristic shows the qualitiative relationship between valve open-
ing and flow for different types of valves. The quantitative relationship is found by calibrat-
ing the flow versus opening for a given fluid and pressure difference.

Flow is always in the direction of decreasing pressure so that the following inequality
holds
∆f ∆P ≥ 0 ∆P 2 , 0 > 0
The inequality shows that an incremental incremental increase in pressure gives an incre-
mental increase in flow.
Example 2.7.1. Linear valve: Suppose that

f = kv m∆P

with the characteristic

kv = 0.15litre/kPa % sec, ∆P = 4kP a

Question: How much should the valve open to achieve 30 l/sec.


Answer: We invert the valve equation for m. Given f = 30 gives
1
m= f = 50% (2.7.5)
∆Pv
Inverting the valve characteristic does not work well in applications where the pressure
varies or the characteristic is not very accurately known 1
74

f∗
m0
Setpoint
Bias
+ Measurement
- f
e
Kc

m Manipulated Variable

Figure 2.20: Feedback control reduces sensitivity with respect to uncertain valve parameters.

m0
Controller Valve

f∗ + e m f
Kc kv ∆Pv
- +
n
+

Figure 2.21: The blockdiagram of thecontrol system shows how the signals are transmitted
in the control loop.
75

2.7.5 Sensitivity Reduction


Accurate flow control can be achieved by using a flow measurement with an automatic
control system which adjusts the MV until the error between the desired and measured
flow is eliminated. The flow controller (FC) shown in Figure 2.16 can be realized by a
proportional controller
m = m0 + Kc e, Kc ≥ 0 (2.7.6)
where e = f ∗ − f is the error between the setpoint and the measured flow, m0 is a bias
variable that provides an estimate of the desired valve position and Kc is the controller
gain.
We now analyze how the proportional controller, valve and measurement work together.
It is now helpful to consider closed loop block diagram in Figure 2.21 . We assume that the
valve and the flow measurement are linear and that there is no noise. The control equation
and the flow equations are given by the expressions

m = m0 + Kc e
f = kv m∆P

where f ∗ is the setpoint (SP) for the flow, Kc the feedback gain and m0 the best estimate
of the valve position from equation (2.7.5). If the feedback gain is equal to zero then there
is no correction from the measurement and we get m = m0 which probably give the wrong
flow due to the factors mentioned above. The question we want to ask is whether the use
of feedback control reduces the sensitivity with respect to errors made in choice of m0 .
From the expressions above

e = f∗ − f
= f ∗ − Kp (m0 + Kc e), Kp = kv ∆P

where Kp is the process gain. Eliminating the manipulated variable m from these equations
gives the closed loop equation for the error
1 Kp
e= f∗ − m0 (2.7.7)
1 + Kp Kc 1 + Kp Kc

This expression shows how the error between the setpoint and the measurement depends
on the control parameter Kc , the setpoint f ∗ and the bias m0 . We get

e → 0 as Kc → ∞

Thus, we can make the error small by choosing large controller gain. However,choosing
very large controller gains in equation may ,be counterproductive as the controller becomes
sensitive to computational delay and noise.
1
Some valves use a valve positioner to measure the position of the valve stem and then the stem is adjusted
up or down until the correct valve position is obtained.
76

One way around the problem is to use a nonlinear controller gain

m = m0 + Kc |e|e

This expression gives a large gain when error is large and small gain when the error is
small. This approach is effective in flow loops and more generally in control systems that
have short delay and fast dynamics. However, this approach will not reduce the error to
zero.
Feedback control and dynamical systems theory has application domains extending be-
yond what one initially might suspect. The principles provide the theoretical basis for
designing automated systems that operate complex chemical processes such that the en-
ergy, and raw materials are utilized in the most economical and efficient ways. They can
be used to explain the inner workings of the cells in our bodies. They can be used to model
instabilities in supply-chains and to model the interlinked global economy. They provide
basis for life-cycle analysis and can be used to develop automated systems to help physicians
perform better in the operating room or help managers make better decisions in a complex
global economy. A good understanding of process control therefore provides a good basis
for understanding how complex dynamics can arise in a broad range of systems which one
would not necessarily think of as control systems.
A number of other examples should be developed
Example 2.7.2. Production of Electronic and Solar Grade Silicon
Example 2.7.3. Population balance (Ecology)
Example 2.7.4. Biological Process (Metabolic Pathway)
Example 2.7.5. Plant wide control and Real time optimization
Example 2.7.6. Drug Injection
Example 2.7.7. Batch Control

Вам также может понравиться