Professional Documents
Culture Documents
ISA Transactions
journal homepage: www.elsevier.com/locate/isatrans
art ic l e i nf o a b s t r a c t
Article history: In this paper, a PID controller design method for the integrating processes based on frequency response
Received 15 May 2013 matching is presented. Two approaches are proposed for the controller design. In the rst approach, a
Received in revised form double feedback loop conguration is considered where the inner loop is designed with a stabilizing
23 August 2014
gain. In the outer loop, the parameters of the PID controller are obtained by frequency response
Accepted 30 August 2014
Available online 11 November 2014
matching between the closed-loop system with the PID controller and a reference model with desired
This paper was recommended for publica- specications. In the second approach, the design is directly carried out considering a desired load-
tion by Prof. A.B. Rad disturbance rejection model of the system. In both the approaches, two low frequency points are
considered for matching the frequency response, which yield linear algebraic equations, solution of
Keywords: which gives the controller parameters. Several examples are taken from the literature to demonstrate
PID controller
the effectiveness and to compare with some well known design methods.
Integrating process
& 2014 ISA. Published by Elsevier Ltd. All rights reserved.
Dead-time
Frequency response matching
Double feedback loops
Load-disturbance rejection
http://dx.doi.org/10.1016/j.isatra.2014.08.020
0019-0578/& 2014 ISA. Published by Elsevier Ltd. All rights reserved.
176 M.N. Anwar, S. Pan / ISA Transactions 55 (2015) 175187
frequency point of the process for tuning the PID controller. However, control the integrating processes.
these methods do not always provide satisfactory controllers for KI
wide variety of processes. Wang and Cluett [20] have presented a C s K P KDs 2
s
method for PID controller design by matching the frequency
where, Kp, K1, and KD are the proportional, the integral and the
response of the desired and actual closed-loop responses at two
derivative constants of the controller that are to be determined. In
specied frequency points. Vijayan and Panda [21] have designed the
Fig. 1, i is the input, y is the process output, r is the output of the
PID controllers for low order processes in double feedback loops to
set-point lter F(S), e is the error, u is the controller output, d is the
achieve stability and improved performance in such a way that the
disturbance, n is the measurement noise and x is the excitation to
inner feedback loop stabilizes the process using the ZN tuning rule
the process. From Fig. 1, the transfer functions from r to y and d to
and the outer loop (tuned by the IMC-PID rule) improves the set-
y may be written as, respectively,
point response. They [22] have also shown reduction of the over-
shoot for various types of processes by using suitable set-point lters. C sGP s
Gr;y s 3
In this paper, a PID controller design technique is proposed for 1 C sGP s
the integrating processes, which is based on frequency response
matching of the actual and the desired closed-loop systems at two GP s
Gd;y s 4
low frequency points. Two approaches are proposed. In the rst 1 C sGP s
approach, with a double feedback loop conguration, the inner The rst step of the design procedure is to choose or construct a
loop contains a stabilizing gain. In the outer loop, the parameters suitable reference model M r;y s which incorporates the desired
of the PID controller are obtained by the frequency response specications of the overall control system from r to y. The
matching of the actual model and a desired reference model at dynamics of the reference model should be such that it can be
two low frequency points. In the second approach, a reference achievable by the process dynamics with an implementable
model is chosen for a desired load-disturbance rejection and then controller. For example, dead-time of a process cannot be elimi-
the frequency response of the control system is matched with that nated in the closed-loop response by any controller. Hence, the
of the reference model to achieve the desired load-disturbance reference model is required to contain the dead-time term of the
rejection. process. Similarly, the inverse response of the process is to be
The desired closed-loop reference model is chosen considering included in the reference model. The proposed method offers the
the design requirement and the process dynamics. The method advantage to consider the industrial specications such as settling
gives a set of linear algebraic equations, solution of which gives the time, peak overshoot, etc. in time domain or gain margin, phase
controller parameters. Though, the frequency response matching margin, band width, etc. in frequency domain or damping factor,
is carried at two low frequency points only, a good set-point undamped natural frequency in complex domain. A good number
response as well as load disturbance rejection are achieved. Both of procedures for choosing a feasible reference transfer function
the proposed approaches are applicable to a wide range of may be found in [23]. However, from such industrial specications,
integrating processes with low or high order that may have we can always analytically construct a reference model of order up
dead-time and/or inverse response dynamics. Effectiveness of to 2 and it works well for the most of the industrial processes.
the proposed method is demonstrated through examples taken The reference model is constructed as
from the literature and favorably compared with some of the
N mr s Ls
methods prevalent in the literature. M r;y s e 5
Dmr s
The rest of the paper is organized as follows. In Section 2, the
proposed design method is presented in detail. In Section 3, the where, Nmr s=Dmr s is in the form of a rational polynomial. It is
method is demonstrated through examples. Conclusion is shown obvious that M r;y s is required to have a minimum time-delay as
in Section 4. that of the process and the non-minimum phase zero of the
process, if any.
For achieving the desired set-point response of the closed-loop
system, the frequency responses of Gr;y s and M r;y s are matched
and may be written as
2. The design method
C sGP s
Gr;y s M r;y ss j 6
An integrating process is considered as shown in Fig. 1 with the 1 C sGP ss j
following transfer function. where, the LHS expression is equivalent with the RHS expression
in terms of frequency response and the unknown parameters of
N p s Ls
GP s e 1 the controller C(s)are to be evaluated. The relation (6) may be
sDp s written as
M r;y s
where, N p s=sDp s is a rational transfer function and L is the time C sGP ss j M o ss j 7
delay of the system. The zeros of Dp(s) are all in the LHS of the 1 M r;y ss j
s-plane. The PID controller with the following form is used to where, M o s may be called as equivalent open-loop reference
model. Eq. (7) gives
d
M o s sDp sN mr s
C ss j Hss j
GP s s j Dmr s Nmr se Ls N p ss j
e u x y
i
F(s)
r
C(s) GP(s) 8
If H(s) could be implemented ideally, it would give the response
n
of Gr;y s exactly same as that of the desired model M r;y s. But the
term e Ls in the expression of H(s) is not suitable for practical
implementation. The Pade approximation of e Ls in terms of a
Fig. 1. Unity negative output feedback conguration. rational polynomial gives rise to a rational expression for H(s),
M.N. Anwar, S. Pan / ISA Transactions 55 (2015) 175187 177
where, Nmd s Ls
M d;y s e 18
2 3 2 3 Dmd s
1 0 0 2 3 H R 0
60 KP
6 10 0 7
7
6 7
6 H I 0 7 where N md s=Dmd s is a rational transfer function and the
A6 7; x 6
4 K 7
5 ; and ; b 6 7 reference model M d;y s must have one zero at origin so as to
61 0 0 7 I 6 H R 1 7
4 5 KD 4 5 achieve load-disturbance rejection. The numerator of M d;y s is
0 11 1 H I 1
required to contain a minimum time-delay as e Ls and non-
minimum phase zeros of the process, if any.
Directly from Eq. (15), we get two values of Kp as:
Following the proposed design method, Eq. (6) is to be
K P1 H R 0 ; K P2 H R 1 modied as
It is observed from various examples, that K P1 K P2 and we GP s
Gd;y s M d;y ss j 19
may consider the value of Kp as any one of K P1 or K P2 or an average 1 C sGP ss j
of these.
Then, to evaluate the remaining parameters, KI and KD, Eq. (15) and Eq. (19) may be simplied as
may be simplied as
1 1
C ss j
A1 x1 b1 16 M d;y s GP s s j
178 M.N. Anwar, S. Pan / ISA Transactions 55 (2015) 175187
N p sDmd s Dp sN md s 3. Simulation results
Hss j 20
N p sN md se Ls s j
Software simulation has been carried for various examples of
The rest of the design procedure is as per Eqs. (9)(16) to the processes taken from the literature. Set-point responses are
determine the values of K P ; K I and K D . shown for unit-step input. Load-disturbance rejections are shown
for unit-step disturbance applied at the process input after the set-
point responses attain the steady-state. Controller outputs are
2.1. Selection of frequency points shown for similar conditions.
The performances in terms of peak overshoot (OS), settling
Generally, the industrial processes show dominantly low-pass time (ts), integral square error (ISE) for the set-point responses;
dynamics in terms of the frequency response. For such cases, the maximum process output (yp), settling time (ts) and ISE for the
low frequency region is more important and frequency response load-disturbance responses; gain margin (GM), phase margin (PM)
matching at -0 ensures the desired steady-state response. As and maximum sensitivity (Ms, which is a measure of the robust-
per Eqs. (13) and (14) the frequency points for matching purpose
ness and is dened as M s max 1 CjGj 1 ) are com-
are required to be sufciently small. With this consideration both 0rr1
the frequency points are selected very small values. The theoretical puted. A lower value of the maximum sensitivity is preferred in
range of the frequency for the frequency response is from 0 to 1 the range of 1.22.0 [1].
and it is meaningless to nd a small number with respect to this Further, to show the robustness of the proposed control
innite range. Here, the small values of frequency points are systems 10% changes in the gain and the time-delay of the
proposed to choose with respect to the bandwidth frequency processes are simultaneously considered and the corresponding
(b) of the desired reference model, where the bandwidth may responses for the perturbed processes are shown. To study the
be treated as an indication to the effective range of frequency effect of the measurement noise the Gaussian noise with a
response. This has been further elaborated in the illustration of variance of 0.005 is introduced after the set-point response attains
Example 1 in Section 3. the steady-state and the corresponding responses are plotted.
All the results are compared with some of the methods
prevalent in the literature.
2.2. Set-point lter
3.1. Example 1
In the process industries, load-disturbance rejection is more
important than the set-point tracking and a high load-disturbance An integrating second order plus dead-time (ISOPDT) process is
rejection sometime results into a highly oscillatory set-point taken from Ali and Majhi [6] as given by
response associated with large peak overshoot. In such case, a
set-point lter, F(s) may be incorporated in the control system, as e 4s
GP s
shown in Figs. 1 and 2, to reduce the peak overshoot as well as the ss 1
oscillation in the set-point response. This gives rise to a two- First approach: The gain K is taken as 0.15 with the peak
degree-of-freedom controller, where F(s) works as a set-point overshoot as 23.9% and the settling time as 37.2 s for the inner
weighting without degrading the load-disturbance rejection con- closed-loop response. For the design of the PID controller in the
trol. The most simple structure of the set-point lter may be taken outer loop the reference model M r;y s is chosen as given by
as Fs 1=s 1, where the time constant is to be chosen
e 4s
judiciously. A large value of reduces the peak overshoot and M r;y s
2s 1
oscillation at the cost of the speed of the set-point response. For
the integrating processes, it is observed that improved result The frequency points are chosen as 0 0.01 rad/s and 1
occurs for 0 o o t s =4, where ts is the settling time of the 0.02 rad/s and the proposed design method yields the PID con-
closed-loop control system without the lter. troller C1 as
The peak overshoot (OS) and the settling time (ts) of the closed- 0:1666
loop control system without the lter and of the lter for each C1s 0:6661 3:037s
s
example are shown in Table 1. The value of is adjusted through
The set-point lter has been found as
simulation to have improved set-point response. It is observed
from Table 1 that the values of have been obtained within ts/4 of 1
Fs
the corresponding examples. 2:2s 1
Second approach: For the purpose of design, the reference
Table 1 model M d;y s for load-disturbance rejection is chosen as
Set-point lter selection.
120se 4s
Example Method Set-point response before employing of Filter M d;y s
5s 115s 1
the lter 1/(s 1)
which gives the peak amplitude of the process output as 4.53 and
OS (%) ts (s) ts/4 (s) the settling time as 78.6 s for unit step load-disturbance. The
Example 1 Proposed C1 35 33.7 8.4 2.2
frequency points are chosen as 0 0.01 rad/s and 1 0.02 rad/s
Proposed C2 43.4 59.5 14.9 12 for frequency response matching and the proposed design method
Example 2 Proposed C1 7.3 40.5 10.1 1.5 yields the PID controller C2 as follows.
Proposed C2 89.2 57.8 14.5 11.5
0:0083
Example 3 Proposed C1 34.5 59.5 14.9 2.5 C2s 0:20 0:357s
Proposed C2 74 81.4 20.4 17 s
Example 4 Proposed C1 17.5 11.2 2.8 1.5
The set-point lter has been found as
Proposed C2 63.0 8.5 2.1 2
Example 5 Proposed C1 18.1 14.2 3.6 12 1
Proposed C2 75.9 24.8 6.2 5 Fs
12s 1
M.N. Anwar, S. Pan / ISA Transactions 55 (2015) 175187 179
1.4 10
Proposed C1
1.2
8 Proposed C2
Ali and Majhi
1
6
SIMC
Proposed C1
Amplitude
Amplitude
0.8
Proposed C2
4
0.6
Ali and Majhi
SIMC
2
0.4
0.2 0
0 -2
0 50 100 150 200
Time (sec) Time (sec)
Fig. 3. Process output for Example 1; (a) Set-point response, (b) Load-disturbance response.
0.3
0.5
Proposed C1
0.25 Proposed C2 Proposed C1
Ali and Majhi Proposed C2
0
0.2 Ali and Majhi
SIMC
SIMC
Amplitude
Amplitude
0.15 -0.5
0.1
-1
0.05
-1.5
0
-0.05 -2
0 50 100 150 200
Time (sec) Time (sec)
Fig. 4. Controller output for Example 1; (a) During set-point response, (b) During load-disturbance response.
10 1.5
Proposed C1
8 Proposed C2
Ali and Majhi
1
6 SIMC
Amplitude
Amplitude
4
0.5
2 Proposed C1
Noise Signal
0
Proposed C2
0
-2
Ali and Majhi
0 50 100 150 200 SIMC
Time
-0.5
0 20 40 60 80 100 120 140
Fig. 5. Process output for perturbed system in Example 1.
Time
Table 2
Performance comparison for Example 1.
Method Controller parameters GM (dB) PM (deg) Ms Set-point response Load-disturbance response Set-point lter F(s)
Proposed C1 (double loop) 0.6661 0.1666 3.037 7.05 63.4 1.33 2.6 25.8 6.818 6.25 44.2 196.2 1/(2.2s 1)
Proposed C2 (single loop) 0.20 0.0083 0.357 5.92 41.5 1.40 0.5 43.9 10.35 6.70 102.5 402.4 1/(12s 1)
Ali and Majhi [6] (single loop) 0.19 0.0084 0.53 5.10 51.8 1.43 26 59.8 7.899 6.32 77.2 400.8
SIMC [8] (single loop) 0.13 0.0039 0.126 9.36 46.8 1.26 30 81.5 8.87 8.76 155.5 1214
Table 3
Parameters of the PID controllers designed with various sets of frequency points for Example 1.
S. no. 0 1 KP KI KD S. no. 0 1 KP KI KD
1 0.005 0.01 0.6665 0.1667 3.033 1 0.0002 0.0004 0.20 0.0083 0.333
2 0.01 0.02 0.6661 0.1666 3.037 2 0.002 0.004 0.20 0.0083 0.350
3 0.04 0.08 0.6578 0.1667 3.044 3 0.005 0.01 0.20 0.0083 0.366
4 0.1 0.2 0.6528 0.1670 3.079 4 0.01 0.02 0.20 0.0083 0.357
5 0.04 0.08 0.20 0.0084 0.399
1.6
1.5
1.4
1.2 1.4
1 1.3
Amplitude
Amplitude
Proposed C1
0.8 Proposed C2
1.2
Ali and Majhi
0.6 Chidambaram and Sree
Proposed C1 1.1
0.4
Proposed C2
Ali and Majhi 1
0.2
Chidambaram and Sree
0 0.9
0 50 100 150 200
Time (sec) Time (sec)
Fig. 7. Process output for Example 2; (a) Set-point response, (b) Load-disturbance response.
of frequency values, i.e., 0 0.01 rad/s and 1 0.02 rad/s is The set-point lter has been found as
taken for all the examples. This means, 0 is taken in the range 1
of 0.27.2% of the bandwidth frequencies of the corresponding Fs
1:5s 1
reference models. However, in general the choice of frequency
points in the range of 0.110% of the bandwidth frequency has Second approach: The reference model M d;y s for the load-
been observed to give good results for the most of the processes. disturbance rejection is chosen as
4s
M d;y s e 6s
3.2. Example 2 5s 16:3s 1
which gives the peak amplitude of the process output as 0.26 and
An integrating plus dead-time (IPDT) process is considered the settling time as 35.5 s for unit step load-disturbance. The
from the literature [6] as given by design procedure yields the PID controller C2 as
0:0506 6s 0:24
GP s e C2s 4:32 9:46s
s s
First approach: The gain K is taken as 1.5 that gives the peak The set-point lter has been found as
overshoot as 1.5% and the settling time as 24.8 s for the inner
closed-loop system. In order to design the outer loop the reference 1
Fs
model M r;y s with the settling time as 42 s is chosen as 11:5s 1
The proposed methods have been compared with the methods
e 6s
M r;y s of Chidambaram and Sree [2], Ali and Majhi [6] and the various
6s 1
simulation results are shown in Figs. 710 and Table 4.
The obtained PID controller C1 is as given below. The controller C1 gives the best set-point response while C2
0:0833 gives the best load-disturbance response. Controller outputs of C1
C1s 0:7231 2:33s and C2 are equally better than the other methods. The GMPM
s
M.N. Anwar, S. Pan / ISA Transactions 55 (2015) 175187 181
6
0.5
Proposed C1
Proposed C1
Proposed C2
4 0 Proposed C2
Ali and Majhi
Ali and Majhi
Chidambaram and Sree
Chidambaram and Sree
-0.5
2
Amplitude
Amplitude
-1
0
-1.5
-2
-2
-4 -2.5
0 50 100 150
Time (sec) Time (sec)
Fig. 8. Controller output for Example 2; (a) During set-point response, (b) During load-disturbance response.
2 transfer function as
e 4s
1.5 GP s
s4s 1
Amplitude
1
First approach: The inner loop gain is selected as 0.15 which
gives the peak overshoot as 44.5% and the settling time as 87.4 s
Proposed C1
for the inner closed-loop system. The reference model is chosen as
Proposed C2
0.5
Ali and Majhi
1
Chidambaram and Sree M r;y s e 4s
0:6s 13s 1
0
0 50 100 150 200
The design procedure gives the PID controller C1as
Time
0:132
Fig. 9. Process output for perturbed system in Example 2. C1s 0:457 6:16s
s
1
rejection is chosen as
0.5 120s
Proposed C1
M d;y s e 4s
Noise Signla 5s 120s 1
0 Proposed C2
Ali and Majhi The PID controller C2 has been obtained as
Chidambaram and Sree
-0.5 0:0083
0 20 40 60 80 100 120 C2s 0:2348 0:540s
Time
s
Fig. 10. Process output for Example 2 with the measurement noise. The set-point lter has been found as
1
Fs
17s 1
and Ms (the maximum sensitivity) are the best with C1. C2 is giving The proposed methods have been compared with the methods
the minimum ISE for the load-disturbance response while for set- of Vijayan and Panda [21] and Shamsuzzoha and Lee [11] and the
point response ISE of C2 is comparable with the others. Distinct various simulation results are shown in Figs. 1114 and Table 5.
improvement by the controllers C1 and C2 may be observed for The controller C1 gives the GMPM and the Ms comparable
the perturbed process. The effect of measurement noise is reduced with the others and the minimum ISE for the set-point response.
considerably in the process output by the proposed methods as C2 gives a comparable load-disturbance response. The controller
well as by the other methods in a similar way. output by C2 is remarkably good. For the perturbed process, the
performance by C1 and C2 are comparable. Here, the method of
Shamsuzzoha and Lee [10] gives the best load-disturbance
3.3. Example 3 response for the nominal as well as perturbed processes with
the help of a PID controller cascaded with a lead compensator. The
An example of an integrating second order plus dead-time effect of the measurement noise is found to be negligible in the
(ISOPDT) process is taken from [21] where the process has the process output.
182 M.N. Anwar, S. Pan / ISA Transactions 55 (2015) 175187
Table 4
Performance comparison for Example 2.
Method Controller parameters GM (dB) PM (deg) Ms Set-point response Load-disturbance response Set-point lter F(s)
Proposed C1 (double) 0.7231 0.0833 2.33 12.16 68.1 1.44 0 28.5 10.23 1.42 41.35 2.46 1/(1.5s 1)
Proposed C2 (single loop) 4.32 0.24 9.46 2.78 26.4 1.80 5.15 43.3 10.24 1.33 39.5 0.9083 1/(11.5s 1)
Ali and Majhi [6] (single loop) 3.39 0.17 9.96 4.09 42.5 1.53 51 50.6 10.62 1.34 41.4 1.26
Chidambaram and Sree [2] (single loop) 4.06 0.15 10.97 2.72 36.6 1.78 59 69.4 10.02 1.32 64.1 1.15
1.4
8
Proposed C1
1.2
Proposed C2
6
Vijayan and Panda
1
Shamsuzzoha and Lee
Amplitude
Proposed C1
Amplitude
0.8 4
Proposed C2
0.6
Vijayan and Panda
Shamsuzzoha and Lee 2
0.4
0
0.2
0 -2
0 50 100 150 200
Time (sec) Time (sec)
Fig. 11. Process output for Example 3; (a) Set-point response, (b) Load-disturbance response.
0.4
Proposed C1 0.5
Proposed C2 Proposed C1
0.3 Vijayan and Panda Proposed C2
0
Vijayan and Panda
Shamsuzzoha and Lee
Shamsuzzoha and Lee
Amplitude
0.2
Amplitude
-0.5
0.1 -1
0 -1.5
-0.1 -2
0 50 100 150 200
Time (sec) Time (sec)
Fig. 12. Controller output for Example 3; (a) During set-point response, (b) During load-disturbance response.
8 controller C2 as
Proposed C1
1:000
6 Proposed C2 C2s 3:5184 1:950s
s
Vijayan and Panda
Amplitude
1
2 Fs
2s 1
0 The proposed methods have been compared with the methods
of Shamsuzzoha and Lee [11], Gu et al. [14] and Luyben [17] and the
-2 various simulation results are shown in Figs. 1518 and Table 6.
0 50 100 150 200
The controller C2 gives the best set-point response while C1
Time gives the best load-disturbance response. With comparable GM,
Fig. 13. Process output for perturbed system in Example 3. PM and Ms the controller C1 generates the best controller output
during load-disturbance. Favorable response may be observed by
the proposed controllers for the perturbed process. The effect of
1.5 measurement noise is reduced considerably in the process output
by the proposed methods as well as by the other methods in a
similar way.
1
Amplitude
Table 5
Performance comparison for Example 3.
Proposed C1 (double loop) 0.457 0.132 6.160 7.15 62.6 1.33 5.5 49.2 7.794 6.05 95.1 264.5 1/(2.5s 1)
Proposed C2 (single loop) 0.2348 0.0083 0.540 4.91 22.8 1.66 2.4 46.4 12.60 6.8 128.2 542.7 1/(17s 1)
Vijayan and Panda [21] (double loop with 0.49 0.122 4.733 10.52 62.7 1.20 1.01 18.1 7.977 6.8 125.4 414.3 1/(2.1s 1)
K 0.1318)
Shamshuzzoha and Lee [11] (single loop; PID with 0.359 0.029 0.97 3.35 29.1 1.67 1.01 23.2 9.905 4.17 31.5 74.29 1/
lead compensator) (32.81s2 12.13s 1)
1.2 1.6
Proposed C1
1 1.5
Proposed C2
0.8 1.4
Shamsuzzoha and Lee
Gu et al
Amplitude
Luyben
Amplitude
0.6 1.3
0.4 1.2
Proposed C1
Proposed C2
0.2 1.1
Shamsuzzoha and Lee
Gu et al
0 1
Luyben
-0.2 0.9
0 10 20 20 30 40 50
Time (sec) Time (sec)
Fig. 15. Process output for Example 4; (a) Set-point response, (b) Load-disturbance response.
184 M.N. Anwar, S. Pan / ISA Transactions 55 (2015) 175187
2 2
Proposed C1 Proposed C1
1.5
Proposed C2 Proposed C2
1.5 Shamsuzzoha and Lee Shamsuzzoha and Lee
1
Gu et al Gu et al
Luyben 0.5 Luyben
1
Amplitude
Amplitude
0
0.5 -0.5
-1
0
-1.5
-0.5 -2
0 10 20 30
Time (sec) Time (sec)
Fig. 16. Controller output for Example 4; (a) During set-point response, (b) During load-disturbance response.
1
disturbance rejection is chosen as
0.6
Proposed C1
0.4 Noise Signal disturbance responses as well as the controller outputs than the
Proposed C2 other methods. The best GMPM and Ms are obtained by C1. The
0.2 Shamsuzzoha and Lee minimum set-point-ISE is given by the method of Luyben [17],
Gu et al
0 however, the response is slow. The proposed methods give an
Luyben
improvement over the other methods for perturbed process. The
-0.2
effect of the measurement noise is found to be negligible in the
-0.4 process output.
0 10 20 30 40 50
Time
Fig. 18. Process output for Example 4 with the measurement noise. 4. Conclusion
Table 6
Performance comparison for Example 4.
Method Controller parameters GM (dB) PM (deg) Ms Set-point response Load-disturbance Set-point lter F(s)
response
Proposed C1 (double loop) 3.128 0.9823 3.996 6.43 59.3 1.33 0 10.1 1.409 1.4 10.8 0.5734 1/(1.5s 1)
Proposed C2 (single loop) 3.5184 1.00 1.95 5.53 24.9 1.54 0 6.45 1.289 1.31 8.5 0.1813 1/(2s 1)
Shamsuzzoha and Lee 2.43 0.667 1.786 7.01 38.9 1.32 0 8.25 1.603 1.42 7.95 0.3794 (1.0963s 1)/(2.7528s2 3.6543s 1)
(single loop)
Gu et al. [14] (single loop) 2.088 0.541 1.436 8.77 38.7 1.27 0 9.2 1.678 1.49 8.2 0.5535 (1.1599s 1)/(2.6597s2 3.8664s 1)
Luyben [17] (single loop) 1.61 0.28 1.8515 7.08 60.8 1.29 0 12.8 2.407 1.51 14.6 1.236 (1.725s 1)/(6.612 s2 5.75 1)
2.5 2.2
Proposed C1
2 2 Proposed C2
Jeng and Lin
1.8 Pai et al
1.5
Luyben
Amplitude
Amplitude
1.6
1
1.4
0.5 Proposed C1
Proposed C2 1.2
Jeng and Lin
0 1
Pai et al
Luyben
-0.5 0.8
0 10 20 30 100 150
Time (sec) Time (sec)
Fig. 19. Process output for Example 5; (a) Set-point response, (b) Load-disturbance response.
2 0.5
Proposed C1 Proposed C1
1.5 Proposed C2
0 Proposed C2
Jeng and Lin
Pai et al Jeng and Lin
1 Pai et al
Luyben
Amplitude
-0.5
Amplitude
Luyben
0.5
-1
0
-1.5
-0.5
-1 -2
0 10 20 100 150
Time (sec) Time (sec)
Fig. 20. Controller output for Example 5; (a) During set-point response, (b) During load-disturbance response.
2
2.5
2 1.5
1.5
Amplitude
1
Amplitude
1 Proposed C1
Proposed C1 0.5 Noise Signal
0.5 Proposed C2 Proposed C2
Jeng and Lin 0 Jeeng and Lin
0 Pai et al Pai et al
Luyben Luyben
-0.5 -0.5
0 10 20 30 40 50 60 0 10 20 30 40 50
Time Time
Fig. 21. Process output for perturbed system in Example 5. Fig. 22. Process output for Example 5 with the measurement noise.
186 M.N. Anwar, S. Pan / ISA Transactions 55 (2015) 175187
Table 7
Performance comparison for Example 5.
Method Controller parameters GM (dB) PM (deg) Ms Set-point response Load-disturbance response Set-point lter F(s)
Proposed C1 (double loop) 1.118 0.370 1.926 6.36 65.5 1.35 0 12.2 3.485 2.14 18.2 4.482 1/(2s 1)
Proposed C2 (single loop) 1.36 0.16 1.20 2.51 26.5 1.87 1.5 11.4 3.826 2.07 55.5 3.62 1/(5s 1)
Jeng and Lin [16] (single loop) 0.950 0.16 1.10 4.87 41.8 1.45 45.4 14.8 3.885 2.136 24.2 5.171 1/(0.057 1)
Pai et al. [15] (single loop) 1.267 0.219 1.171 3.14 28.4 1.73 84.5 19 4.251 2.098 18 3.379
Luyben [17] (single loop) 0.867 0.0361 0.963 5.50 55.2 1.40 14 48.4 3.12 2.195 89.6 17.75
there is no requirement of elaborate frequency response analysis Using the rst order Pade approximation of e Ls in the
or any mathematically involved optimization technique. Overall, denominator, Eq. (27) may be simplied as
the method is mathematically simple and computational burden is
very small. The method is applicable to wide range of integrating K D s2 K P s K I N p se Ls
processes that may be of low order or high order and that may s2 Dp s
s j
have dead-time and non-minimum phase zeros.
Various types of examples of the integrating processes such as n L=2s 2n e Ls
2
h
i 28
integrating plus dead-time (IPDT) process, integrating second s L=2 s2 2 2
nL
1 s L 2 2
n n
s j
order plus dead-time (ISOPDT) process, ISOPDT process with
non-minimum phase zero, high order integrating plus dead-time In Eq. (28), LHS has two poles at origin (type-2 transfer
process with non-minimum phase zero are taken for design of function) while the RHS has only one pole at origin (type-1
controllers using the proposed method. It is observed from the transfer function). Then, for feasible matching, types of both sides
examples that the proposed rst approach gives better perfor- of Eq. (28) are to be same. This may be obtained by choosing KI 0
mance than the proposed second approach. which makes LHS as type-1. It is observed, as expected, that even
when the Pade approximation is not used in Eq. (27), the proposed
method gives K I 0. In a similar way, it can be shown that for any
Appendix A order of the reference model, KI comes out to be 0.
For the two frequency points 0 and 1, Eqs. (13) and (14) give
the following four equations References
K P H R 0 21 [1] Astrom KJ, Hagglund T. PID controllers theory design and tuning. Research
Triangle Park (NC): Instrument Society of America; 1995.
KI [2] Chidambaram M, Sree RP. A simple method of tuning PID controller for
K D 0 H I 0 22 integrating/dead time processes. Comput Chem Eng 2003;27:2115.
0 [3] Chen D, Seborg DE. PI/PID controller design based on direct synthesis and
disturbance rejection. Ind Eng Chem Res 2002;41:480722.
K P H R 1 23 [4] Poulin E, Pomerleau A. PID tuning for integrating and unstable processes. IEE
Proc- Control Theory Appl 1996;143:42935.
[5] Visioli A. Optimal tuning of PID controllers for integral and unstable processes.
KI
K D 1 H I 1 24 IEE Proc - Control Theory Appl 2001;148:1804.
1 [6] Ali A, Majhi S. PID controller tuning for integrating processes. ISA Trans
2010;49:708.
Eqs. (21)(24) may be rearranged in compact form to have [7] Rivera DE, Morari M, Skogestad S. Internal model control 4. PID controller
2 3 2 3 design. Ind Eng Chem Process Des Dev 1986;25:25265.
1 0 0 2 3 H R 0 [8] Skogestad S. Simple analytic rules for model reduction and PID controller
6 0 1 7 KP 6 7
6 0 07
7 6 H 7 tuning. J Process Control 2003;13:291309.
6 764 K 5 6 I 0 7 25 [9] Arbogast JE, Cooper DJ. Extension of IMC tuning correlations for non-self
61 0 0 7 I 6 H R 1 7
4 5 K 4 5 regulating (integrating) processes. ISA Trans 2007;46:30311.
0 11 1 D
H I 1 [10] Shamsuzzoha M, Lee M. IMC-PID controller design for imrproved disturbance
rejection of time-delayed processes. Ind Eng Chem Res 2007;46:207791.
[11] Shamsuzzoha M, Lee M. PID controller design for integrating processes with
which follows Eq. (15). time delay. Korean J Chem Eng 2008;25:63745.
[12] Chia TL, Lefkowitz I. Internal model-based control for integrating processes.
ISA Trans 2010;49:51927.
[13] Hua W, Xiao GI, Li X. An analytical method for PID controller tuning with
Appendix B specied gain and phase margins for integral plus time delay processes. ISA
Trans 2011;50:26876.
[14] Gu D, Ou L, Wang P, Zhang W. Relay feedback autotuning method for
A desired reference model of second order is considered as
integrating processes with inverse response and time delay. Ind Eng Chem
given by Res 2006;45:311932.
[15] Pai NS, Chang SC, Huang CT. Tuning PI/PID controllers for integrating
n 2
M r;y s e Ls 26 processes with deadtime and inverse response by simple calculations.
s2 2n s n 2 J Process Control 2010;20:72633.
[16] Jeng JC, Lin SW. Robust proportional-integral-derivative controller design for
Then, Eq. (7) may be written as stable/integrating processes with inverse response and time delay, Ind Eng
Chem Res 2012;51:265265.
e Ls
2n [17] Luyben WL. Tuning proportional-integral-derivative controllers for integrator/
KI N p s Ls s2 2n s 2n dead time processes. Ind Eng Chem Res 1996;35:34803.
KP KDs e 2n [18] Tyrus BD, Luyben WL, Tuning PI. Controllers for integrator/dead time pro-
s sDp s s j 1 2e
Ls
s2 2n s n s j cesses. Ind Eng Chem Res 1992;31:26258.
[19] Ziegler JG, Nichols NB. Optimum settings for automatic controllers. Trans.
27 ASME 1942;64:75968.
M.N. Anwar, S. Pan / ISA Transactions 55 (2015) 175187 187
[20] Wang L, Cluett WR, Tuning PID. Controllers for integrating processes. IEE Proc [23] Chen CT. Analog and digital control systems design: transfer-function, state-
Control Theory Appl 1997;144:38592. space, and algebraic methods. Philadelphia: Sanders College Pub; 1993.
[21] Vijayan V, Panda RC. Design of PID controllers in double feedback loops for [24] Pan S, Pal J. Reduced order modelling of discrete-time systems. Appl Math
SISO systems with set-point lters. ISA Trans 2012;51:51421. Model 1995;19:1338.
[22] Vijayan V, Panda RC. Design of a simple setpoint lter for minimizing [25] Astrom KJ, Hang CC, Persson P, Ho WK. Towards intelligent PID control.
overshoot for low order processes. ISA Trans 2012;51:2716. Automatica 1992;28:19.