Вы находитесь на странице: 1из 19

ARTICLE IN PRESS

Journal of the Franklin Institute 346 (2009) 38–56


www.elsevier.com/locate/jfranklin

Direct synthesis-based controller design for


integrating processes with time delay
A. Seshagiri Rao, V.S.R. Rao, M. Chidambaram
Department of Chemical Engineering, Indian Institute of Technology Madras, Chennai-36, India
Received 28 November 2005; received in revised form 5 January 2008; accepted 4 June 2008

Abstract

Using the direct synthesis method, a PID controller in series with a lead/lag compensator is
designed for control of open loop integrating processes with time delay. Set-point weighting is
considered for reducing the undesirable overshoot. Guidelines are provided for selection of the
desired closed loop tuning parameter in the direct synthesis method and set point weighting
parameter. The method gives significant load disturbance rejection performances. Illustrative
examples are considered to show the performances of the proposed method. Significant improvement
is obtained when compared to recently reported methods.
r 2008 The Franklin Institute. Published by Elsevier Ltd. All rights reserved.

Keywords: Integrating process; Time delay; Direct synthesis method; Set point weighting

1. Introduction

The dynamic response of many processes such as heating boilers, batch chemical
reactors, liquid storage tanks, and a liquid level system with a pump attached to the
outflow is very slow with large dominant time constant. Time delay is common in all the
process industries due to transportation delays, recycle loops, composition analysis loops,
etc. Chien and Fruehauf [1] studied bottom level control in a distillation column and
observed that the resulting open loop process shows a large time constant. These types of
processes can be approximated as integrating processes with time delay for the purpose of
designing the controllers instead of controlling in the original large time constant form.

Corresponding author. Fax: +91 44 235 0509.


E-mail address: chidam@iitm.ac.in (M. Chidambaram).

0016-0032/$32.00 r 2008 The Franklin Institute. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.jfranklin.2008.06.004
ARTICLE IN PRESS
A. Seshagiri Rao et al. / Journal of the Franklin Institute 346 (2009) 38–56 39

Designing controllers based on the integrating processes with time delay gives superior
closed loop performance than designing based on first-order plus time delay process for
nominal and model mismatch conditions [1]. In the open literature, several PI and PID
controllers design methods have been proposed [2–6] for control of integrating processes
with time delay. Internal model control (IMC)-based control design methods [7–9] are also
proposed to improve the closed loop performance. However, when there is large time
delay, control of integrating processes is difficult because of the limitations imposed by the
time delay on system performance and stability. To overcome this difficulty, modified
Smith predictors have been proposed [10–17] to enhance the closed loop performance as
the original Smith predictor cannot reject load disturbance for integrating processes with
large time delay.
Recently, Liu et al. [18] have proposed a modified Smith predictor scheme with three
controllers for improved performance for integrating processes with time delay. Their
method involves a set-point tracking controller, a disturbance estimator and a third
controller for stabilizing the integrating process without time delay. Later, Lu et al. [19]
have proposed a double two-degree-of-freedom control scheme with four controllers for
controlling integrating and unstable processes.
There also exists industrial processes such as aerospace control systems, DC motors and
high-speed disk drives whose dynamics show the characteristics of double integrator types
(i.e. the process model transfer function is in the form of (keys/s2or keys/s2(ts+1)) [20].
The controller design methods for these types of processes have been addressed by
Skogestad [7], Hang et al. [17] and Liu et al. [20].
Most of the methods discussed above use either the modified form of the IMC method
or the modified form of Smith predictor structure, whose structures are complicated. Also,
the methods require design of more number of controllers with more number of tuning
parameters. Practically, a simple control structure with a simple controller is desirable as it
is very easy for the operator to tune. In the present work, using the direct synthesis method
for set-point tracking, a simple controller design method is proposed with only one
controller in a single feedback loop for all classes of integrating processes with time delay.
However, with the conventional controllers, there may be problems like large overshoot
and settling time. To avoid large overshoot, set-point weighting is suggested in the
controller [21]. Hence, in the present work, set-point weighting is also considered.

2. Proposed method

Typically, the classes of integrating processes with time delay can be represented by any
of the following transfer function models:
Pure integrating process plus time delay (IPTD),

keys
Gp ¼ (1)
s
integrating plus first-order plus time delay (IFOPTD),

keys
Gp ¼ (2)
sðts þ 1Þ
ARTICLE IN PRESS
40 A. Seshagiri Rao et al. / Journal of the Franklin Institute 346 (2009) 38–56

integrating plus unstable first-order plus time delay (IUFOPTD),


keys
Gp ¼ (3)
sðts  1Þ
pure double-integrating plus time delay (DIPTD),
keys
Gp ¼ (4)
s2
double-integrating plus first-order plus time delay (DIFOPTD),
keys
Gp ¼ (5)
s2 ðts þ 1Þ
If the integrating process is of higher order, then it can be reduced to the form of any of
the above-mentioned process models by using the relevant identification method. The
simple feedback block diagram showing the process and the controller is shown in Fig. 1,
where Gp is the transfer function of the integrating process and Gc is the transfer function
of the controller.
For designing of the controller Gc, the direct synthesis method is considered here. This
synthesis method is chosen here because the desired output behavior of the closed loop can
be specified as a trajectory model based on the process to design the required form of the
controller [22]. The closed loop relation for set-point changes is given by
y Gc Gp
¼ (6)
yr 1 þ G c G p
From Eq. (6), the controller is given by
1 ðy=yr Þ
Gc ¼ (7)
G p 1  ðy=yr Þ
According to the direct synthesis method, the closed loop trajectory model should be
specified for designing the controller. Thus, the controller can be written as
1 ðy=yr Þd
Gc ¼ (8)
G p 1  ðy=yr Þd
where (y/yr)d is the desired closed loop trajectory for set-point changes. With this, the
controller is designed for different types of integrating processes (Eqs. (1)–(5)) after
specifying the desired closed loop transfer function. In the present work, the controller is
designed for all the above integrating processes after specifying the desired closed loop
transfer function based on the type of the integrating process.

ydi ydo

+ +
yr u + + y
Gc Gp
+
-

Fig. 1. Feedback control structure.


ARTICLE IN PRESS
A. Seshagiri Rao et al. / Journal of the Franklin Institute 346 (2009) 38–56 41

3. Controller design

Based on the nature of the integrating process, the desired closed loop transfer function
is chosen and correspondingly the controller structure is derived. In fact, for all the cases
PID controller in series with lead/lag compensator is obtained. The detailed design
procedure is explained below.

3.1. Case (i)

If the process is of Eq. (1), i.e. Gp ¼ keys/s, the desired closed loop transfer function is
considered as
 
y ðZs þ 1Þeys
¼ (9)
yr d ðls þ 1Þ2
From Eqs. (1) and (9) the controller is obtained as
s ðZs þ 1Þ
Gc ¼ (10)
kp ½ðls þ 1Þ2  ðZs þ 1Þeys 
With first-order Padé approximation for the time delay [eys ¼ (10.5ys)/(1+0.5ys)],
Eq. (10) reduces after simplification to
ðZs þ 1Þð0:5ys þ 1Þ
Gc ¼ (11)
k½0:5yl2 s2 þ ðl2 þ yl þ 0:5yZÞs þ ð2l þ y  ZÞ
Eq. (11) can be further simplified by applying Z ¼ 2l+y to a conventional PID
controller in series with a first-order filter as
 
1 1
G c ¼ kc 1 þ þ td s (12a)
ti s ðtf s þ 1Þ
where
2l þ 1:5y
kc ¼
kðl þ 2ly þ 0:5y2 Þ
2

ti ¼ 2l þ 1:5y
yl þ 0:5y2
td ¼
2l þ 1:5y
0:5yl2
tf ¼ 2 (12b)
l þ 2yl þ 0:5y2

3.2. Case (ii)

If the process is of Gp ¼ keys/s(ts+1) (Eq. (2)), the desired closed loop transfer
function is considered as
 
y ðZ s2 þ Z1 s þ 1Þeys
¼ 2 (13)
yr d ðls þ 1Þ3
ARTICLE IN PRESS
42 A. Seshagiri Rao et al. / Journal of the Franklin Institute 346 (2009) 38–56

From Eqs. (2) and (8) the controller is obtained as

sðts þ 1Þ ðZ2 s2 þ Z1 s þ 1Þ
Gc ¼ (14)
kp ½ðls þ 1Þ3  ðZ2 s2 þ Z1 s þ 1Þeys 
Using first-order Padé approximation for the time delay, after simplification, the
controller is obtained as

ðts þ 1ÞðZ2 s2 þ Z1 s þ 1Þð0:5ys þ 1Þ


Gc ¼
k½0:5yl3 s3 þ ðl þ 1:5yl þ 0:5yZ2 Þs2 þ ð3l2 þ 1:5yl  Z2 þ 0:5yZ1 Þs þ ð3l þ y  Z1 Þ
3 2

(15)
Eq. (15) can be approximated as a PID controller in series with lead/lag compensator in
the form of (refer Appendix for details)
 
1 ðas þ 1Þ
G c ¼ kc 1 þ þ td s (16a)
ti s ðbs þ 1Þ
where
Z1
kc ¼ 2
kð3l þ 1:5yl þ 0:5yZ1  Z2 Þ
t i ¼ Z1
Z
td ¼ 2
Z1
0:5yl3
a ¼ 0:5y; b¼
tð3l2 þ 1:5yl þ 0:5yZ1  Z2 Þ
ð0:5y  tÞl3 þ ð3t2  1:5ytÞl2 þ 3yt2 l þ 0:5y2 t2
in which Z1 ¼ 3l þ y and Z2 ¼
tð0:5y þ tÞ
(16b)

3.3. Case (iii)

If the process is of Gp ¼ keys/s(ts1) Eq. (3), the controller design procedure is the
same as explained in case (ii) and the controller parameters are obtained as
Z1
kc ¼
kðZ2  3l2  1:5yl  0:5yZ1 Þ
t i ¼ Z1
Z
td ¼ 2
Z1
0:5yl3
a ¼ 0:5y; b¼
tðZ2  3l2  1:5yl  0:5yZ1 Þ
ð0:5y þ tÞl3 þ ð3t2 þ 1:5ytÞl2 þ 3yt2 l þ 0:5y2 t2
in which Z1 ¼ 3l þ y and Z2 ¼
tðt  0:5yÞ
(17)
ARTICLE IN PRESS
A. Seshagiri Rao et al. / Journal of the Franklin Institute 346 (2009) 38–56 43

3.4. Case (iv)

If the process is of double integrating with time delay, i.e. Gp ¼ keys/s2 (Eq. (4)) the
desired closed loop transfer function is assumed as given in Eq. (13). Thus the controller is
obtained by substituting Eqs. (4) and (13) in Eq. (8) and approximating the time delay as a
first-order Padé as
sðZ2 s2 þ Z1 s þ 1Þð0:5ys þ 1Þ
Gc ¼
k½0:5yl3 s3 þ ðl þ 1:5yl þ 0:5yZ2 Þs2 þ ð3l2 þ 1:5yl  Z2 þ 0:5yZ1 Þs þ ð3l þ y  Z1 Þ
3 2

(18)
Eq. (18) can be approximated in the form of a PID controller in series with lead/lag
compensator as Eq. (16a) by taking
Z1 ¼ 3l þ y and Z2 ¼ 3l2 þ 1:5yl þ 0:5yZ1 (19a)
With that the controller parameters are obtained as
Z1
kc ¼
kðl þ 1:5yl2 þ 0:5yZ2 Þ
3

t i ¼ Z1
Z
td ¼ 2
Z1
0:5yl3
a ¼ 0:5y; b¼ (19b)
ðl þ 1:5yl2 þ 0:5yZ2 Þ
3

3.5. Case (v)

If the process is of Gp ¼ keys/s2(s+1) (Eq. (5)), the desired closed loop transfer
function is considered as
 
y ðZ s2 þ Z1 s þ 1Þeys
¼ 2 (20)
yr d ðls þ 1Þ4
By substituting Eqs. (5) and (20) in Eq. (8), and using first-order Padé approximation for
time delay, the controller is obtained as
sðts þ 1ÞðZ2 s2 þ Z1 s þ 1Þð0:5ys þ 1Þ
Gc ¼
k½0:5yl4 s4 4
þ ðl þ 2yl 3
Þs3 þ ð4l3 þ 3yl2 þ 0:5yZ2 Þs2 þ ð6l2 þ 2yl  Z2 þ 0:5yZ1 Þs þ ð4l  Z1 þ yÞ
(21)
By taking Z1 ¼ 4l+y and Z2 ¼ 6l2+20l+0.5yZ1, the controller Eq. (21) can be
converted into the form of
ðZ2 s2 þ Z1 s þ 1Þ
Gc ¼
ks
ðts þ 1Þð0:5ys þ 1Þ
 (22)
½0:5yl4 s2 þ ðl þ 2ylÞs þ ð4l3 þ 3yl2 þ 0:5yZ2 Þ
4
ARTICLE IN PRESS
44 A. Seshagiri Rao et al. / Journal of the Franklin Institute 346 (2009) 38–56

Eq. (22) can be written in the form of a PID controller is series with compensator as
 
1 ða2 s2 þ a1 s þ 1Þ
G c ¼ kc 1 þ þ td s (23a)
ti s ðb2 s2 þ b1 s þ 1Þ
where
Z1
kc ¼
kð4l3 þ 3yl2 þ 0:5yZ2 Þ
t i ¼ Z1
Z
td ¼ 2
Z1
a2 ¼ 0:5yt; a1 ¼ 0:5y þ t
0:5yl4 l4 þ 2yl3
b2 ¼ ; b1 ¼ (23b)
4l3 þ 3yl2 þ 0:5yZ2 4l3 þ 3yl2 þ 0:5yZ2

3.6. Set-point weighting

To reduce large overshoot, set-point weighting is suggested [21] and thus the PID
controller in Eqs. (12a), (16a), and (23a) should be implemented in the form of
  Z 
1 de
uðtÞ ¼ kc ðyr  yÞ þ e dt þ td (24)
ti dt
where e ¼ yry, in which e is the set-point weighting parameter. The set-point weighting
parameter should be selected in such a way that the resulting controller should not give
large overshoot responses. Based on extensive simulations conducted on different types of
integrating processes, it is suggested that e can be taken in the range of 0.3–0.4.

3.7. Selection of the tuning parameter

The primary requirement for selection of l is that the resulting controller gains should be
positive for positive values of k. Hence, to get positive values of controllers gain (kc) for
cases (ii) and (iii), the constraints to be followed are Zo3l2+1.5yl+0.5yZ1 and
Z243l2+1.5yl+0.5yZ1, respectively. Further, l should be selected in such a way that
the resulting controller gives good nominal as well as robust control performances. After
carrying out several simulation studies, it was observed that the initial value of the tuning
parameter can be taken as equal to half of the time delay of the process. If good control
performances are not achieved with this value, then the tuning parameter can be increased
gradually from this value till good nominal and robust control performances are achieved.
Quantitatively, the range of the tuning parameter that gives good control performances is
0.8y–3y.

3.8. Remarks in controller design

It is to be noted that the designed controllers give good control performances for the
corresponding integrating processes with time delay by choosing suitable values of l.
ARTICLE IN PRESS
A. Seshagiri Rao et al. / Journal of the Franklin Institute 346 (2009) 38–56 45

However, it is observed that the designed value of b is too large for obtaining robust
performances of the closed loop system. Also, because of high value of b, the phase lag
imposed by the term (bs+1) in the controller is more and thus the designed controller with
this value of b is not able to give robust control performances. When sensitivity and
complementary sensitivity functions are plotted against the frequency, the peak values of
both the sensitivity and complementary sensitivity functions are significantly high, which
results in low gain and phase margins of the open loop system than the required values
(gain margin should be 41.7 and phase margin should be 4351 for robust control of a
process) [23]. Hence, to get good nominal and robust closed loop performances, the value
of b should be considered less than that obtained by the design procedure. Based on many
simulation studies, it is observed that taking ‘0.1b’ instead of b gives good compromise
between nominal performances and robust control performances. Thus, in the present
work, the value of b obtained is modified as ‘0.1b’ for simulation studies.

4. Stability and robustness

For any closed loop control system, it is necessary to analyze the stability and robustness
for uncertainties in the process and for the load disturbances. When the bound on the
process multiplicative uncertainty equals to zero, the closed loop system is stable, because
the closed loop system is robustly stable if and only if the closed loop system is stable when
the multiplicative uncertainty equals to 0 and
kl m ðjoÞTðjoÞko1 8o 2 ð1; 1Þ (25)
where T(s ¼ jo) is the complementary sensitivity function and lm(s ¼ jo) is the bound on
the process multiplicative uncertainty [24]. The process uncertainty can be represented as
 
G p ðjoÞ  G m ðjoÞ
l m ðjoÞ ¼  
 (26)
G m ðjoÞ
where Gm(jo) is the model of the integrating process. To show the stability and robustness
analysis more qualitatively, let us consider the IPTD process Eq. (1) for which the
complementary sensitivity function of the closed loop with the designed controller Eqs.
(12a) and (12b) is
kkc ð1 þ ti s þ ti td s2 Þeys
TðsÞ ¼ (27)
sðtf s þ 1Þ þ kkc ð1 þ ti s þ ti tD s2 Þeys
where the controller parameters kc,ti,td and tf are the functions of the tuning parameter ‘l’.
If uncertainty exists in the time delay, then the tuning parameter should be selected such
that
1
kTðjwÞk1 o (28)
jeDys  1j
If the uncertainty exists in the gain, then the tuning parameter should be selected in such
a way that
1
kTðjoÞk1 o (29)
jDkj=k
ARTICLE IN PRESS
46 A. Seshagiri Rao et al. / Journal of the Franklin Institute 346 (2009) 38–56

Also for ensuring that the closed loop performance is robust, the constraints to be
followed by the sensitivity and complementary sensitivity functions are [24]
kl m ðjoÞTðjoÞ þ wm ðjoÞð1  TðjoÞÞko1 (30)
where wm(jo) is the uncertainty bound on the sensitivity function, which is given by
1T(jo). Hence the tuning parameter has to be selected such that the resulting controller
should satisfy the robust stability and robust performance constraints (Eqs. (25) and (30)).
Similar stability and robustness analysis can be done for the remaining types of integrating
processes with time delay.

5. Simulation study

For the purpose of simulation study, different integrating processes with time delay are
considered and the control performances obtained by using the proposed method are
compared with the recently reported methods.

5.1. Example-1

Consider an IPTD process with large time delay [18]:


1
G p ðsÞ ¼ e5s
s
By using Eqs. (12a) and (12b), the controller parameters obtained for the proposed
method are kc ¼ 0.2, ti ¼ 17.5, td ¼ 2.143 and tf ¼ 0.0714. The tuning parameter is
considered as l ¼ y ¼ 5 and the set-point weighting parameter is chosen as e ¼ 0.4. With
these controller settings, the performances of the closed loop control system are evaluated
by giving a unit step input in the set point and a negative step input of 0.1 in the load (ydi)
at t ¼ 150 s. For comparison, the methods proposed by Skogestad [7], Liu et al. [18] and
Lu et al. [19] are considered. Fig. 2 shows the responses when there are no mismatches in
1.4

1.2

1
Response, y

0.8

0.6

0.4

0.2

0
0 50 100 150 200 250 300
Time

Fig. 2. Responses for perfect model for example-1. Solid—proposed, dash—Liu et al. [18], dot—Lu et al., dash
dot—Skogestad.
ARTICLE IN PRESS
A. Seshagiri Rao et al. / Journal of the Franklin Institute 346 (2009) 38–56 47

the plant and model. From the responses it can be observed that Liu et al. [18] and Lu et al.
[19] methods give slightly better control performances. However, it is to be noted that in
practice there always exist plant model mismatch. Thus the control performances should
also be checked considering that there exists uncertainty in the process model. In the
present work, +20% perturbation in the process time delay (controller is designed
considering y ¼ 5 and simulation is done considering y ¼ 6) and the corresponding
responses are shown in Fig. 3. It can be observed that the proposed method gives better
control performances compared to those of Lu et al. [19], Liu et al. [18] and Skogestad [7].
A similar trend is observed when there are perturbations in the process gain. The method
of Lu et al. [19] shows completely unstable responses for infinitesimal perturbation in time
delay and thus is not shown in Fig. 3. It is to be noted that Liu et al. [18] already showed
that their method gives better control performances compared to that of Mataušek and
Micić [11]. For quantitative comparison the IAE performance criterion is considered here.
Table 1 shows the IAE values when there is no plant model mismatch and considering
there is +20% mismatch in process time delay.

5.2. Example-2

Consider an IFOPTD process [18]:


1
G p ðsÞ ¼ e6:567s
sð3:4945s þ 1Þ
The controller parameters obtained using Eqs. (16a) and (16b) are kc ¼ 0.1571,
ti ¼ 22.325, td ¼ 2.94, a ¼ 3.282 and b ¼ 0.0958, with the tuning parameters selected as
l ¼ 0.8, y ¼ 5.253. The set-point weighting parameter is chosen as e ¼ 0.4. Using these
controller settings the performance is evaluated by giving unit step input in the set point
and a negative step input of 0.1 in the load (ydi) at t ¼ 150 s. To show the improvement, the

1.4

1.2

0.8
Response, y

0.6

0.4

0.2

-0.2
0 50 100 150 200 250 300
Time

Fig. 3. Responses for a perturbation of +20% in process time delay for example-1. Solid—proposed, dash—Liu et al.
[18], dash dot—Skogestad.
ARTICLE IN PRESS
48 A. Seshagiri Rao et al. / Journal of the Franklin Institute 346 (2009) 38–56

Table 1
Total IAE values for servo and regulatory responses

Example-1 Perfect model +20% in time delay

Proposed 19.726 20.610


Lu et al. 12.443 –
Liu et al. [14] 12.380 27.075
Skogestad 59.176 60.960

Example-2 Perfect model +20% in time delay

Proposed 28.399 30.167


Liu et al. [14] 30.902 41.266
Kaya 47.730 47.841

Example-3 Perfect model +20% in gain and time delay

Proposed 4.921 4.843


Kaya 6.189 6.153

Example-4 Perfect model +20% in time delay and 20% in time constant

Proposed 1.751 1.782


Liu et al. [21] 3.149 3.226

Example-5 Perfect model +20% in time delay

Proposed 3.731 3.736


Liu et al. [15] 5.935 5.942
Hang et al. 5.345 5.498

Example-6 Perfect model +15% in time constant and time delay

Proposed 7.766 8.010


Liu et al. [15] 11.693 12.191

(–) Unstable responses.

methods proposed recently by Kaya [9] and Liu et al. [18] are considered. Fig. 4 shows the
responses for perfect model. It can be observed that the proposed method gives better
control performances. A perturbation of +20% in process time delay is considered and the
corresponding responses are shown in Fig. 5. It is clear that the proposed method is better.
Similar improvement in the performance is observed for perturbations in process gain and
time constant. Corresponding IAE values are shown in Table 1. From the IAE values it is
evident that the proposed method gives low IAE values.

5.3. Example-3

A higher-order integrating process with time delay is considered here [9]:


1
G p ðsÞ ¼ e0:5s
sðs þ 1Þðs þ 2Þðs þ 3Þ
ARTICLE IN PRESS
A. Seshagiri Rao et al. / Journal of the Franklin Institute 346 (2009) 38–56 49

1.4

1.2

1
Response, y

0.8

0.6

0.4

0.2

0
0 50 100 150 200 250 300
Time

Fig. 4. Responses for perfect model for example-2. Solid—proposed, dash—Liu et al. [18], dot—Kaya.

1.2

0.8
Response, y

0.6

0.4

0.2

-0.2
0 50 100 150 200 250 300
Time

Fig. 5. Responses for perturbations of +20% in process time delay for example-2. Solid—proposed, dash—Liu
et al. [18], dot—Kaya.

Kaya [9] has shown that this process can be identified as an IFOPTD process as
0:167
G p ðsÞ ¼ e1:077s
sð1:863s þ 1Þ
The proposed controller design technique is applied to the identified process and the
corresponding controller parameters are obtained using Eqs. (16a) and (16b) as kc ¼ 6.332,
ti ¼ 5.6, td ¼ 1.24, a ¼ 0.5385 and b ¼ 0.0187, with the tuning parameters selected as
ARTICLE IN PRESS
50 A. Seshagiri Rao et al. / Journal of the Franklin Institute 346 (2009) 38–56

l ¼ 1.4, y ¼ 1.507. The set-point weighting parameter is chosen as e ¼ 0.3. Using these
controller settings the performance is evaluated by giving unit step change in the set point
and a unit negative step input in the load (ydi) at t ¼ 30 s. Fig. 6 shows the responses for
perfect model conditions for the proposed method and that of Kaya [9]. The proposed
method gives better control performances. Perturbations of +20% in process gain and
time delay is considered, and the corresponding responses are shown in Fig. 7. It is evident
that the proposed method gives better control performances. The corresponding IAE
values are shown in Table 1. From the IAE values it can be observed that the proposed
method shows low values.

5.4. Example-4

Consider an IUFOPTD process [25]:


1
G p ðsÞ ¼ e0:2s
sðs  1Þ
For the proposed method, the controller settings obtained by using Eq. (17) are
kc ¼ 3.6615, ti ¼ 2, td ¼ 1.003, a ¼ 0.1 and b ¼ 0.00395 with the tuning parameters
selected as l ¼ 3, y ¼ 0.6. The set-point weighting parameter is chosen as e ¼ 0.4. Using
these controller settings the performance is evaluated by giving unit step change in the set
point and a unit negative step input in the load (ydi) at t ¼ 25 s. To show the improvement
the Liu et al. [25] method is considered here. Fig. 8 shows the responses for a perfect model.
Improved performance is obtained by the proposed method. Perturbations of +20% in
process time delay and 20% in unstable process time constant are considered and the
responses are shown in Fig. 9. Clearly, the proposed method gives improved performances.
The corresponding IAE values are shown in Table 1. Low IAE values are obtained by the
proposed method.

1.4

1.2

1
Response, y

0.8

0.6

0.4

0.2

0
0 10 20 30 40 50 60
Time

Fig. 6. Responses for perfect model for example-3. Solid—proposed, dash—Kaya.


ARTICLE IN PRESS
A. Seshagiri Rao et al. / Journal of the Franklin Institute 346 (2009) 38–56 51

1.4

1.2

1
Response, y

0.8

0.6

0.4

0.2

0
0 10 20 30 40 50 60
Time

Fig. 7. Responses for perturbations of +20% in process gain and time delay for example-3. Solid—proposed,
dash—Kaya.

1.4

1.2

1
Response, y

0.8

0.6

0.4

0.2

0
0 5 10 15 20 25 30 35 40 45 50
Time

Fig. 8. Responses for perfect model for example-4. Solid—proposed, dash—Liu et al. [25].

5.5. Example-5

A DIPTD [20] is considered here;

1 0:8s
G p ðsÞ ¼ e
s2
ARTICLE IN PRESS
52 A. Seshagiri Rao et al. / Journal of the Franklin Institute 346 (2009) 38–56

1.4

1.2

Response, y 1

0.8

0.6

0.4

0.2

0
0 5 10 15 20 25 30 35 40 45 50
Time

Fig. 9. Responses for perturbations of +20% in process time delay and 20% in process time constant for
example-4. Solid—proposed, dash—Liu et al. [25].

The controller parameters obtained for the proposed method are kc ¼ 0.6807, ti ¼ 4.4,
td ¼ 1.7091, a ¼ 0.4 and b ¼ 0.0107, with the tuning parameters selected as l ¼ 1.5,
y ¼ 1.2. The set-point weighting parameter is chosen as e ¼ 0.3. Using these values the
performance is evaluated by giving unit step change in the set point and a negative step
input of 0.1 in the load (ydi) at t ¼ 40 s. For comparison, method proposed by Hang et al.
[17] and Liu et al. [20] are considered. In the Hang et al. [17] method, 10% uncertainty is
assumed (b ¼ 0.12) and the tuning parameter (a) is considered as 0.8. With that, the
controller parameters for P1 and P2 are obtained as kc1 ¼ 0.2934, td1 ¼ 3.148,
kc2 ¼ 0.1877, td2 ¼ 3.936; set-point filter (F) and load filter (C) are obtained as
(1.259s+1)/(3.148s+1) and (4.736s+1)/(3.936s+1), respectively. Fig. 10 shows the
responses for the perfect model. The proposed method performs better. Fig. 11 shows
the responses for a perturbation of +20% in process time delay. Here also, the proposed
method gives improved performances. Similar performance improvement is observed for
perturbations in the process gain. The corresponding IAE values are shown in Table 1. It is
evident that the proposed method gives low IAE values.

5.6. Example-6

Consider a DIFOPTD [20]:


1
G p ðsÞ ¼ 2 es
s ðs þ 1Þ
The controller parameters calculated using Eqs. (23a) and (23b) are kc ¼ 0.256, ti ¼ 6.6,
td ¼ 2.7061, a2 ¼ 0.4, a1 ¼ 1.5, b2 ¼ 0.074 and b1 ¼ 0.361, with the tuning parameters
selected as l ¼ 1.4, y ¼ 1.4. The set-point weighting parameter is chosen as e ¼ 0.3. Using
these controller settings, the performance is evaluated by giving unit step input in the set
point, a negative step input of 0.1 in ydi at t ¼ 50 s and a negative step input of 0.1 in ydo at
t ¼ 100 s. Fig. 12 shows the responses for the perfect model and Fig. 13 shows the
ARTICLE IN PRESS
A. Seshagiri Rao et al. / Journal of the Franklin Institute 346 (2009) 38–56 53

1.4

1.2

1
Response, y

0.8

0.6

0.4

0.2

0
0 10 20 30 40 50 60 70 80
Time

Fig. 10. Responses for perfect model for example-5. Solid—proposed, dash—Liu et al. [20], dot—Hang et al.

1.4

1.2

1
Response, y

0.8

0.6

0.4

0.2

0
0 10 20 30 40 50 60 70 80
Time

Fig. 11. Responses for a perturbation of +20% in time delay for example-5. Solid—proposed, dash—Liu et al.
[20], dot—Hang et al.

responses for perturbations of +15% in process time constant and time delay. Clearly, the
proposed method performs better. Corresponding IAE values are given in Table 1 and it
can be observed that the proposed method gives less IAE values.
The main advantage of the proposed method is that it is simple with conventional
feedback structure and hence is easy for the operator. The previous methods make use of
more complicated structures for controlling the integrating processes with either small or
large time delay. Also, the methods use more number of controllers and hence more degree
ARTICLE IN PRESS
54 A. Seshagiri Rao et al. / Journal of the Franklin Institute 346 (2009) 38–56

1.4

1.2

Response, y 1

0.8

0.6

0.4

0.2

0
0 50 100 150
Time

Fig. 12. Responses for perfect model for example-6. Solid—proposed, dash—Liu et al. [20].

1.4

1.2

1
Response, y

0.8

0.6

0.4

0.2

0
0 50 100 150
Time

Fig. 13. Responses for perturbations of +15% in time constant and time delay for example-6. Solid—proposed,
dash—Liu et al. [20].

of freedom. In addition, the methods have more number of tuning parameters, which is
very difficult for the operator to tune online on trial and error basis. In the proposed
method, there are two parameters to be selected, viz. set-point weighting parameter (e) and
the desired closed loop time constant (l). However, it is already shown by simulation that
the set-point weighting can be taken as 0.3–0.4. Hence, only one tuning parameter, i.e. l,
needs to be tuned. Guidelines are also provided for selection of this tuning parameter and
hence the proposed method is simple to implement online and easy to tune.
ARTICLE IN PRESS
A. Seshagiri Rao et al. / Journal of the Franklin Institute 346 (2009) 38–56 55

6. Conclusions

A simple method of controller design for integrating processes with time delay using
direct synthesis method is proposed. The method has a single controller and makes use of
two tuning parameters. Guidelines are provided for the selection of the tuning parameters.
Good nominal and robust control performances are achieved with the designed controller.
Significant improvement is obtained when compared to recently reported methods in the
open literature. Several simulation examples are considered to show the advantage of the
proposed method.

Appendix A

In Eq. (15), letting Z1 ¼ 3l+y in the constant term of the denominator, Eq. (15) reduces
to
ðts þ 1ÞðZ2 s2 þ Z1 s þ 1Þð0:5ys þ 1Þ
Gc ¼ (A.1)
khs½ð0:5yl3 =hÞs2 þ ðl3 þ 1:5yl2 þ 0:5yZ2 =hÞs þ 1
where
h ¼ 3l2 þ 1:5yl  Z2 þ 0:5yZ1 (A.2)
Now, factorizing the second-order denominator as
 3 
0:5yl3 2 l þ 1:5yl2 þ 0:5yZ2
s þ sþ1
h h
¼ ðts þ 1Þðbs þ 1Þ (A.3)
Upon equating the corresponding coefficients in Eq. (A.3), and substituting h from
Eq. (A.2) we get
0:5yl3
bt ¼
3l2 þ 1:5yl  Z2 þ 0:5yZ1
and
l3 þ 1:5yl2 þ 0:5yZ2
bþt¼ 2
(A.4)
3l þ 1:5yl  Z2 þ 0:5yZ1
From Eq. (A.4) Z2 and b are obtained as
ð0:5y  tÞl3 þ ð3t2  1:5ytÞl2 þ 1:5yt2 l þ 0:5yZ1 t2
Z2 ¼ (A.5)
0:5yt þ t2

0:5yl3
b¼ (A.6)
tð3l2 þ 1:5yl  Z2 þ 0:5yZ1 Þ
Thus, the controller Eq. (A.1) can be written as
ðZ2 s2 þ Z1 s þ 1Þð0:5ys þ 1Þ
Gc ¼ (A.7)
khsðbs þ 1Þ
ARTICLE IN PRESS
56 A. Seshagiri Rao et al. / Journal of the Franklin Institute 346 (2009) 38–56

This is a PID controller in series with a lead/lag filter and the controller parameters are
given in Eqs. (16a) and (16b).

References

[1] I.L. Chien, P.S. Fruehauf, Consider IMC tuning to improve performance, Chem. Eng. Prog. 10 (1990) 33–41.
[2] H.J. Kwak, S.W. Sung, I.-B. Lee, On-line process identification and autotuning for integrating processes,
Ind. Eng. Chem. Res. 36 (12) (1997) 5329–5338.
[3] K. Kookos, A.I. Lygeros, K.G. Arvanitis, On-line PI controller tuning for integrator/dead time processes,
Eur. J. Control 5 (1999) 19–31.
[4] E. Poulin, A. Pomerleau, PI settings for integrating processes based on ultimate cycle information, IEE
Trans. Control Syst. Tech. 7 (4) (1999) 509–511.
[5] Y.-G. Wang, W.-J. Cai, Advanced proportional-integral-derivative tuning for integrating and unstable
processes with gain and phase margin specifications, Ind. Eng. Chem. Res. 41 (2002) 2910–2914.
[6] M. Chidambaram, R.P. Sree, A simple method of tuning PID controllers for integrator/dead-time processes,
Comp. Chem. Eng. 27 (2003) 211–215.
[7] S. Skogestad, Simple analytic rules for model reduction and PID controller tuning, J. Process Control 13
(2003) 291–309.
[8] W. Zhang, X. Xu, Y. Sun, Quantitative performance design for integrating processes with time delay,
Automatica 35 (1999) 719–723.
[9] I. Kaya, Two-degree-of-freedom IMC structure and controller design for integrating processes based on gain
and phase-margin specifications, IEE Proc. Control Theory Appl. 151 (4) (2004) 401–407.
[10] M.R. Mataušek, A.D. Micić, A modified Smith predictor for controlling a process with an integrator and
long dead time, IEEE Trans. Autom. Control 41 (8) (1996) 1199–1203.
[11] M.R. Mataušek, A.D. Micić, On the modified Smith predictor for controlling a process with an integrator
and long dead time, IEEE Trans. Autom. Control 44 (8) (1999) 1603–1606.
[12] Y.-C. Tian, F. Gao, Control of integrator processes with dominant time delay, Ind. Eng. Chem. Res. 38
(1999) 2979–2983.
[13] H.J. Kwak, S. Whan, I.-B. Lee, Modified Smith predictor for integrating processes: comparisons and
proposition, Ind. Eng. Chem. Res. 40 (2001) 1500–1506.
[14] I.-L. Chien, S.C. Peng, J.H. Liu, Simple method for integrating processes with long dead time, J. Process
Control 12 (2002) 391–404.
[15] J.E. Normey-Rico, E.F. Camacho, A unified approach to design dead-time compensators for stable and
integrative processes with dead time, IEEE Trans. Autom. Control 47 (2) (2002) 299–305.
[16] I. Kaya, Obtaining controller parameters for a new PI-PD Smith predictor using autotuning, J. Process
Control 13 (2003) 465–472.
[17] C.C. Hang, Q.-G. Wang, X.-P. Yang, A modified Smith predictor for a process with an integrator and long
dead time, Ind. Eng. Chem. Res. 42 (2003) 484–489.
[18] T. Liu, Y.Z. Cai, D.Y. Gu, W.D. Zhang, New modified Smith predictor scheme for integrating and unstable
processes with time delay, IEE Proc. Control Theory Appl. 152 (2) (2005) 238–246.
[19] X. Lu, Y.-S. Yang, Q.-G. Wang, W.-X. Zheng, A double two-degree-of-freedom control scheme for
improved control of unstable delay processes, J. Process Control 15 (5) (2005) 605–614.
[20] T. Liu, X. He, D.Y. Gu, W.D. Zhang, Analytical decoupling control design for dynamic plants with time
delay and double integrators, IEE Proc. Control Theory Appl. 151 (6) (2004) 745–753.
[21] K.J. Åström, T. Hägglund, PID Controllers, second ed., ISA Publishers, Research Triangle Park, NC, 1995.
[22] B.A. Ogunnaike, W.H. Ray, Process Modeling, Dynamics and Control, Oxford University Press, New York,
1994.
[23] M.A. Johnson, M.H. Moradi, PID Control: New Identification and Design Methods, Springer, London,
2005.
[24] M. Morari, E. Zafiriou, Robust Process Control, Prentice Hall, Englewood Cliffs, 1989.
[25] T. Liu, W. Zhang, D. Gu, Analytical design of two-degree-of-freedom control scheme for open loop unstable
processes with time delay, J. Process Control 15 (5) (2005) 559–572.

Вам также может понравиться