Professional Documents
Culture Documents
Robust PID controller design for time delay processes with peak of
maximum sensitivity criteria
Mohammad Shamsuzzoha
Department of Chemical Engineering, King Fahd University of Petroleum and Minerals, Dhahran 31261, Saudi Arabia
© Central South University Press and Springer-Verlag Berlin Heidelberg 2014
Abstract: The motivation of this work is to obtain single PI/PID tuning formula for different types of processes with enhanced
disturbance rejection performance. The proposed tuning formula consistently gives better performance in comparison to several
well-known methods at the same degree of robustness for stable, integrating and unstable processes. For the selection of the
closed-loop time constant (τc), a guideline is provided over a broad range of time-delay/time-constant ratios on the basis of the peak
of maximum sensitivity (Ms). An analysis has been performed for the uncertainty margin with the different process parameters for the
robust controller design. It gives the guideline of the Ms-value settings for the PI controller designs based on the process parameters
uncertainty. Furthermore, a relationship has been developed between Ms-value and uncertainty margin with the different process
parameters (k, τ and θ). Simulation study has been conducted for the broad class of processes and the controllers are tuned to have the
same degree of robustness by measuring the maximum sensitivity, Ms, in order to obtain a reasonable comparison.
Key words: PI/PID controller tuning; internal model control (IMC) method; unstable delay process; integrating delay process;
disturbance rejection
y
i si 1
Gp qf r pA i 1 fr (6)
r ( c s 1) r
m
y
i si 1
(1 Gp q )Gp 1 pA i 1 G (7)
d ( c s 1) r
p
m
Fig. 1 Block diagram of IMC and classical feedback control
systems: (a) IMC structure; (b) Feedback control structure
The numerator expression i si 1 in Eq. (6)
i 1
causes an excessive overshoot in the servo response,
For the nominal case Gp G p , the set-point and which can be eliminated by introducing the set-point
disturbance responses in the IMC control structure can filter fr to compensate for the overshoot in the servo
be simplified as response.
y Gp qf r r (1 G p q)Gp d (1) From the above design procedure, a stable,
closed-loop response can be achieved by using the IMC
According to the IMC parameterization [2], the controller. The ideal feedback controller that is
process model G p is factored into two parts: equivalent to the IMC controller can be expressed in
G p pm pA (2) terms of the internal model G p and the IMC controller
q as:
where pm is the portion of the model inverted by the
controller, pA is the portion of the model not inverted by q
Gc (8)
the controller and pA(0)=1. The noninvertible part 1 G p q
usually includes the dead time and/or right half plane
Substituting Eqs. (2) and (5) into Eq. (8) gives the
zeros and is chosen to be all-pass.
ideal feedback controller:
To get a superior response for unstable processes or
m
stable processes with poles near zero, the IMC controller
( i s i 1)
q should satisfy following conditions. If the process Gp
pm1 i 1
has unstable poles or poles near zero at z1, z2,…, zm, then ( c s 1) r
Gc (9)
q should have zeroes at z1, z2,…, zm, and (1−Gpq) should m
pA ( i s 1) i
also have zeroes at z1, z2,…, zm.
1 i 1
Since the IMC controller q is designed as ( c s 1) r
q pm1 f , the first condition is satisfied automatically.
The second condition can be fulfilled by designing the The resulting controller in Eq. (9) is physically
IMC filter (f) as realizable but it does not have the standard PI/PID form.
The desired form of the controller can be obtained by
im1 i s i 1
f (3) using the proper approximation of the dead time term for
( c s 1) r example Taylor series expansion. In this work, both
where τc is an adjustable parameter which controls the simplicity and approximation error due to dead time term
tradeoff between the performance and robustness; r is has been considered carefully during the PI/PID
selected to be large enough to make the IMC controller controller design.
(semi-)proper; αi is determined by Eq. (4) to cancel the
poles near zero in Gp. 3 PI/PID controller design
pA (im1 i s i 1)
1 Gp q 1 0 (4) This section illustrates the PI/PID controller design
s z1 , zm ( c s 1) r sz ,, zm method for several representative cases which are
1
3780 J. Cent. South Univ. (2014) 21: 3777−3786
frequently used by the process control engineer in real process becomes the series-form of the controller as
processes. 1 Ds 1
Gc K c 1 (17)
Is D N s 1
3.1 First-order plus dead time process
First order plus dead time (FOPDT) process is where τD is the derivative time. In the simulation study,
representative model and commonly used in the chemical filter parameter N is typically around 100 and can be
process industries. utilized to make the series-PID controller with derivative
filter. The implementation of the series-PID structure in
ke s
Gp (10) Eq. (17) is shown in Fig. 2. In the simulation of the
s 1 second order process, the robustness margins have been
where k is the process gain, τ is the time constant, and θ computed with τD/N=0. The PID setting of the proposed
is the time delay, the IMC filter structure selected is method is for the series form of the PID controller. If
s 1 required, it can be easily converted to the ideal (parallel)
f (11)
( c s 1) 2 form of the PID.
Table 1 PI/PID controller setting for proposed and other methods with performance matrix
Setpoint Load disturbance
Case Process Method τc Ms Kc τI VT
EIA: (y) EIA (y) VT
Proposed 2.46 1.60 4.57 4.85 3.1 5.9 1.06 1.37
e s
C1 SIMC 1.0 1.60 5.0 8.0 2.5 5.6 1.60 1.16
10s 1
LEE et al [3] 1.0 1.60 5.12 10.25 2.17 5.58 2.0 1.10
Proposed 19.37 1.70 0.304 39.63 30.21 0.51 131.9 1.74
0.2e7.4s
C2 SIMC 7.4 1.70 0.338 59.2 28.8 0.49 174.5 1.55
s
TL — 1.67 0.33 64.7 29.13 0.46 195 1.50
e 0.5 s Proposed 1.36 6.0 1.646 8.25 6.72 12.0 5.01 7.30
C3
s 1 LEE et al [9] 1.4 6.0 1.668 8.67 6.77 12.04 5.20 7.24
0.5 s Proposed 2.9 2.2 3.22 9.50 6.37 7.59 2.95 2.59
e
C4
(5s 1)(2 s 1)(0.5s 1) YANG et al [10] 1.5 2.2 2.564 10.98 8.57 7.24 4.28 2.95
10e s Proposed 2.57 1.65 0.942 5.55 20.39 1.48 5.89 2.44
C5
(20 s 1)(20 s 1) SIMC 1.0 1.65 1.0 8.0 20.6 1.44 8.0 2.44
Note: For case C4, τD=2.25 and 1.82 is used for proposed and YANG et al [10] methods, respectively; For case C5, τD=20 is used for both proposed and SIMC
methods, response is without setpoint filter for both the methods. A set-point filter fr is suggested to enhance servo response, for PI controller
f r ( c s 1) /( s 1); for PID f r ( c s 1) /( 2 s 2 s 1).
3782 J. Cent. South Univ. (2014) 21: 3777−3786
Fig. 5 Responses of PID-control for high order unstable Fig. 6 MV plots of PI-control for first-order process e−s/(10s+1)
process Gp=e−0.5s/(5s−1)(2s+1)(0.5s+1) (C4) for both setpoint (C1) for both setpoint (a) and load disturbance (b) of magnitude
(a) and load disturbance (b) of magnitude 1 at t=0 1 at t=0
performance of the control system. As τc decreases, the is important to obtain the relationship between Ms-value
closed-loop response becomes faster and can be unstable. and parametric uncertainty in the control system design.
On the other hand, as τc increases, the closed-loop It is because these uncertainties play an important role in
response becomes sluggish and more stable. A good control system and cause poor performance or even
tradeoff is obtained by choosing τc to give Ms-value in instability of closed-loop control systems. A typical first
the range of 1.2−2.0 for stable process. An analysis of order delay process (e−θs/(10s+1)) is considered for this
the τc selection has been conducted and plot of Ms verses analysis for various dead time to lag time ratios by
θ/τ for different values of τc= γθ, where γ=1.0, 2.0 and changing θ while fixing τ=10). The Kharitonov’s
3.0 is shown in Fig. 9. The figure clearly shows that τc=θ theorem is used to obtain the uncertainty margin in the
is not the proper choice because for the lag dominant process parameter and further it is verified by the
process, it gives tight controller setting. For τc=3θ, it simulation for different θ/τ ratio. The percentages of the
gives smooth and robust setting because Ms lies between uncertainty margin in different parameters have been
1.61 to 1.25. A good tradeoff between robustness and analyzed for different Ms-value. Figure 10 shows the
performance can be achieved for τc=2θ where it gives variation in dead time margin for different Ms-values.
Ms=2.0 for lag dominant process and Ms=1.26 for delay The figure clearly indicates that for a fixed Ms-value, as
dominant process. the θ/τ ratio increases, the dead time margin also
increases. The variation in the process gain uncertainty
6 Robustness study for the different Ms-values is shown in Fig. 11. The trend
in Fig. 11 is reverse and it shows that for the fixed
This section presents the analysis of the control Ms-value, as θ/τ ratio increases, the percentage in the
system design for system affected by parametric gain margin decreases. The uncertainty in the process
uncertainty. It indicates the maximum uncertainty margin time constant τ is shown in Fig. 12. The uncertainty
in different process parameters for the fixed Ms-value. It margin in the τ is lower than the original value whereas
3784 J. Cent. South Univ. (2014) 21: 3777−3786
Acknowledgement
The author would like to acknowledge the support
Fig. 12 Variation of uncertainty margin in process time constant
provided by King Abdulaziz City for Science and
(τ) with θ/τ ratio for different Ms-values Technology (KACST) through the “KACST Annual
Program” at King Fahd University of Petroleum &
Table 2 Maximum uncertainty margin in θ, k and τ for different
Minerals (KFUPM) for funding this work through
value of Ms project number AT-32-41.
Uncertainty Uncertainty Uncertainty
Ms
margin in θ/% margin in k/% margin in τ/%
References
1.4 278 322 79.4
[1] KANO M, OGAWA M. The state of art in chemical process control
1.6 183 211 72.0 in Japan: Good practice and questionnaire survey [J]. Journal of
1.8 136 157 66.1 Process Control, 2010, 20: 969−982.
[2] RIVERA D, MORARI M, SKOGESTAD S. Internal model control. 4.
2.0 108 126 61.2 PID controller design [J]. Ind Eng Chem Process Des Dev, 1986, 25:
252−265.
7 Conclusions [3] LEE Y, PARK S, LEE M, BROSILOW C. PID controller tuning for
desired closed–loop responses for SI/SO systems [J]. AIChE, 1998,
44: 106−115.
1) A simple analytical design method for the PI/PID [4] SKOGESTAD S. Simple analytic rules for model reduction and PID
controller is proposed, which is based on the IMC controller tuning [J]. Journal of Process Control, 2003, 13: 291−309.
principle in order to improve the disturbance rejection [5] SHAMSUZZOHA M, LEE M. IMC–PID controller design for
performance. improved disturbance rejection of time–delayed processes [J]. Ind
Eng Chem Res, 2007, 46: 2077−2091.
2) Another important feature of the proposed
[6] SMITH C L, CORRIPIO A B, MARTIN J. Controller tuning from
methodology is that it deals stable, integrating and simple process models [J]. Instrum Technol, 1975, 22(12): 39
unstable process in a unified way. [7] CHEN D, SEBORG D. PI/PID controller design based on direct
3) In conclusion, the final tuning formula for the synthesis and disturbance rejection [J]. Industrial and Engineering
proposed PI/PID tuning rule is summarized as Chemistry Research, 2002, 41: 4807−4822.
[8] MORARI M, ZAFIRIOU E. Robust process control [M]. New Jersey:
Kc ; I ; D 2 ; and Prentice-Hall Englewood Cliffs, 1989: 85−111.
k (2 c ) [9] LEE Y, LEE J. PARK S. PID controller tuning for integrating and
unstable processes with time delay [J]. Chemical Engineering
c
2
Science, 2000, 55: 3481−3493.
1 1 e
[10] YANG X, WANG Q, HANG C, LIN C. IMC-based control system
design for unstable processes [J]. Industrial and Engineering
4) For the first order and integrating process with Chemistry Research, 2002, 41: 4288−4294.
[11] WANG Y, CAI W. Advanced proportional-integral-derivative tuning
time delay, the resulting tuning rule is PI where τD=0.
for integrating and unstable processes with gain and phase margin
5) The design method is based on the disturbance specifications [J]. Industrial and Engineering Chemistry Research,
rejection and a set-point filter is recommended to 2002, 41: 2910−2914.
eliminate the overshoot in set-point response mainly in [12] TAN W, MARQUEZ H, CHEN T. IMC design for unstable processes
unstable and integrating process. with time delays [J]. Journal of Process Control, 2003, 13: 203−213.
[13] LIU T, ZHANG W, GU D. Analytical design of
6) For a given process model, parameter τc is a
two-degree-of-freedom control scheme for open-loop unstable
single adjustable parameters in the tuning rule and it is
process with time delay [J]. Journal of Process Control, 2005, 15:
used to obtain a compromise between performance and 559−572.
robustness. The recommended selection of closed loop [14] JUNG C S, SONG H K, HYUN C J. A direct synthesis tuning
3786 J. Cent. South Univ. (2014) 21: 3777−3786
method of unstable first-order-plus-time-delay processes [J]. Journal [24] ALFARO V M, VILANOVA R. Optimal robust tuning for 1DoF
of Process Control, 1999, 9: 265−269. PI/PID control unifying FOPDT/SOPDT models [C]// in IFAC
[15] MAJHI S, Atherton D. Obtaining controller parameters for a new Conference on Advances in PID Control PID’12, Brescia (Italy):
smith predictor using autotuning [J]. Automatica, 2000, 36: IFAC, 2012.
1651−1658. [25] JIN Q B, LIU Q. IMC-PID design based on model matching
[16] KWAK H, SUNG S, LEE I B, PARK J. Modified smith predictor approach and closed-loop shaping [J]. ISA Transactions, 2014, 53:
with a new structure for unstable processes [J]. Industrial and 462−473.
Engineering Chemistry Research, 1999, 38: 405−411. [26] JIN Q B, LIU Q. Analytical IMC-PID design in terms of
[17] ZHANG W, GU D, WANG W, XU X. Quantitative performance performance/robustnesstradeoff for integrating processes: From
design of a modified smith predictor for unstable processes with time 2-Dof to 1-Dof [J]. Journal of Process Control , 2014, 24: 22−32.
delay [J]. Industrial and Engineering Chemistry Research , 2004, 43: [27] JENG J C, LIN S W. Robust Proportional-integral-derivative
56−62. controller design for stable/integrating processes with inverse
[18] TYREUS B, LUYBEN W. Tuning PI controllers for integrator/dead response and time delay [J]. Ind Eng Chem Res, 2012, 51:
time processes [J]. Ind Eng Chem Res, 1992: 2625−2628. 2652−2665.
[19] SHAMSUZZOHA M, LEE M. Analytical design of enhanced [28] ARRIETA O, VILANOVA R. Simple servo/regulation
PID·filter controller for integrating and first order unstable processes proportional-integral-derivative (PID) tuning rules for arbitrary
with time delay [J]. Chemical Engineering Science, 2008, 63: Ms-based robustness achievement [J]. Ind Eng Chem Res, 2012, 51:
2717−2731. 2666−2674.
[20] ALCANTARA S, VILANOVA R, PEDRET C. PID control in terms [29] NORMEY-RICO AND J E, GUZMÁN J L. Unified PID tuning
of robustness/performance and servo/regulator trade-offs: A unifying approach for stable, integrative, and unstable dead-time processes [J].
approach to balanced autotuning [J]. Journal of Process Control, Ind Eng Chem Res, 2013, 52: 16811−16819.
2013, 23: 527−542. [30] ZIEGLER J G. NICHOLS N B. Optimum settings for automatic
[21] LEE J, CHO W, EDGAR T F. Simple analytic PID controller tuning controllers [J]. Trans ASME, 1942, 64: 759−768.
rules revisited [J]. Industrial & Engineering Chemistry Research, [31] SHAMSUZZOHA M, SKOGESTAD S. The setpoint overshoot
2014, 53: 5038−5047. method: A simple and fast closed-loop approach for PID tuning [J].
[22] CHO W, LEE L, EDGAR T. Simple analytic Journal of Process Control, 2010, 20: 1220−1234.
proportional-integral-derivative (PID) controller tuning rules for [32] HU W, XIAO G. Analytical proportional-integral (PI) controller
unstable processes [J]. Industrial & Engineering Chemistry Research, tuning using closed-loop setpoint response [J]. Ind Eng Chem Res,
2014, 53: 5048−5054. 2011: 2461-2466.
[23] TORRICO B C, CAVALCANTE M U, BRAGA A P, [33] SHAMSUZZOHA M. Closed-loop PI/PID controller tuning for
NORMEY-RICO J E, ALBUQUERQUE A A. Simple tuning rules stable and integrating process with time delay [J]. Ind Eng Chem Res,
for dead-time compensation of stable, integrative, and unstable 2013, 52: 12973−12992.
first-order dead-time processes [J]. Industrial & Engineering (Edited by DENG Lü-xiang)
Chemistry Research, 2013, 52: 11646−11654.