You are on page 1of 8

Third International Conference on

Advances in Control and Optimization of Dynamical Systems


March 13-15, 2014. Kanpur, India

Robust controller design for First order Plus Time Delay systems using Kharitonov
Theorem
N.Varun Chowdary, M.Chidambaram*
Department of Instrumentation, M.I.T, Chromepet, Chennai
*Department of Chemical Engg, Indian Institute of Technology, Madras

Abstract: In this paper, a new method using Kharitonov theorem and an algorithm in MATLAB for stability boundary locus is applied
to compute all the stabilizing PI controller gains for a stable first order plus time delay system which satisfies both robust stability and
robust performance criteria. The PI settings designed using SIMC rule is also shown in the common stability region. Similarly for an
unstable FOPTD system all stabilizing PID controller is identified which satisfies both robust stability and P.M specification. The PI
settings designed for unstable FOPTD system using Dominant Pole Placement and Equating Coefficient method is shown on
plane to compare robustness. It is also shown that to design robust controller for Unstable FOPTD, PID controller should be preferred
to PI controller.
Keywords: Kharitonov Theorem, Robust control, PID controllers, Stabilization, First order plus Time Delay

1. INTRODUCTION considered stable FOPTD but also maintain a pre-specified gain


This paper is concerned about the design of a robust PI controller margin (G.M) and phase margin (P.M) specifications. The PI
for first order plus dead time systems. The problem is motivated settings designed using SIMC rule is marked in the common
by industrial control applications where quite often the plant is stability region to show the robustness. Furthermore, PID
modeled as a first-order system with time delay and the controller is designed for the unstable FOPTD process which
controller is either of the PI or PID type)[8]. Most industrial robustly stabilizes the system in the considered interval. The PI
systems are still controlled by PID controllers. Therefore there settings designed using Dominant Pole Placement and Equating
has been a lot of research on designing the appropriate PI(D) Coefficient methods are marked in the plane to compare
controllers especially for systems under some nonlinearities, robustness.
perturbations or time-variant behavior. The primary and essential
2. KHARITONOV THEOREM
requirement of all applications is the stability of closed control
loop. But due to the presence of Model Uncertainty the closed 2.1 For real coefficients
loop stability and performance may not be satisfactory. Model This theorem investigates the stability characteristics of the
Uncertainty can be caused due to the presence of one or more of interval systems via four vortex polynomials with real
the following: non-linear effects when a linear model is used, coefficients varying in a bounded range. For the interval system
high order dynamics when model neglects such performance, P(s)= a0+a1s+a2s2+a3s3+...+ansn (1)
slow varying parameters such as heat transfer coefficient or where aI ϵ [ail , aiu], where ail is the lower limit and aiu is the upper
some unknown phenomenon. limit and i=1,2,...n, the stability of the set could be found by
Hence there is a need for robust stability analysis and applying the Routh criterion to the following polynomials
controller synthesis for uncertain systems especially control- P1(s) = a0l+a1ls+a2us2+ a3us3+.... (2)
systems with parametric uncertainty. In [13], [17] kharitonov P2(s) = a0l+a1us+a2us2+ a3ls3+.... (3)
theorem was used to compare the robustness of the various P3(s) = a0u+a1ls+a2ls2+ a3us3+.... (4)
tuning rules. In [3], the Kharitonov theorem for P4(s) = a0u+a1us+a2ls2+ a3ls3+.... (5)
Uncertain/Interval plants is exploited for the purpose of
characterizing all PID controllers that stabilize an uncertain 2.2 For complex coefficients
plant. In [4], a graphical method was proposed for computing all The complex kharitonov theorem for the polynomial ( ) is
feasible gain and phase margin specifications-oriented robust shown from the equations (8) to (15) [7]
PID controllers to stabilize uncertain systems. Complete sets of
stabilizing PI/PID controllers have been computed using similar ( ) ( ) ( ) ( )
theorems for varying time delay systems [5] , multiple time ( ) (6)
delay systems[10] etc. In [8-9] Hermite-Biehler was used to find [ ] [ ] [ ] (7)
the robust PI settings. In [6] robust PID controller for unstable ( ) ( ) ( ) ( )
system was designed which satisfies GM and PM specification ( ) (8)
and also the optimal gains which minimizes ISE, IAE in the ( ) ( ) ( ) ( )
common region were also found out. Though lot of robust PID ( ) (9)
controllers design are there for delay time process, no method is ( ) ( ) ( ) ( )
designed with simple &less analytical calculations. Moreover ( ) (10)
boundary locus approach used in [3],[4],[5],[11],[12] requires ( ) ( ) ( ) ( )
one to calculate and separately which increases the ( ) (11)
analytical calculation for higher order system. Hence there is a ( ) ( ) ( ) ( )
need of simple algorithm for delay (or) higher order systems. ( ) (12)
This paper deals with proposal of new method based on ( ) ( ) ( ) ( )
kharitonov theorem and an algorithm in MATLAB is used to ( ) (13)
find graphically all feasible PI controllers gains, directly in the ( ) ( ) ( ) ( )
- parameter plane, which could not only robust stabilize the ( ) (14)

978-3-902823-60-1 © 2014 IFAC 184 10.3182/20140313-3-IN-3024.00103


2014 ACODS
March 13-15, 2014. Kanpur, India

( ) ( ) ( ) ( ) TABLE 1. CLASSES OF K, L AND T


( ) (15)
The complex coefficient polynomial ( ) is Hurwitz stable if
and only if all the eight vertex polynomials are stable.

3. METHOD PROPOSED
Consider a first order plus delay transfer function of the form
( )
( )
( )
3.1 Existing method
According to [5], to design the robust PID controller of the
form c(s)=( ) using complex kharitonov theorem for
a Gain margin(G.M), Phase Margin(P.M) it is required to take
the characteristic equation of the form (1+ G.M*c(s)*G(s)*
)=0 and substituting s=jw. Then the equation is separated 4. STABILITY, GAIN MARGIN, PHASE MARGIN
into real and imaginary parts and equated to zero. By solving CALCULATION
both the equations keeping as constant, are obtained To draw the graph between and for a fixed using
in terms of w. By sweeping over the admissible values of w stability boundary locus the methods given in [11], [12] requires
stability boundary locus is obtained. determination of ( ), ( ), ( ) ( ). In
Hence the above method is computationally intensive to order to simplify the design of robust PID controller the below
calculate and . To overcome this new method is proposed to given calculations and the program written in MATLAB based on
design a robust PID controller with a given specification which the calculation is used.
overcomes the analytical calculation for each system.
3.2 Method proposed 4.1 Stability and Gain Margin Calculation
Consider the transfer function of plant ( )
( )
(16) |G.M*Gc(s)*G(s)|=1 (21)
( ) Where G.M is the required Gain Margin, Gc(s) is the controller
Where N(s) = k*exp(-L*s),D(s)=T*s 1 (17) transfer function and G(s) is the plant model
K=[ ] , L =[ ], T= [ ] Gc(s)= (kds2+kps+ki)/s (22)
Consider the numerator N(s)=k*exp(-L*s) (18) Let s=jw
Substituting s=jw, then we get Gc(jw)=(kdj2w2+kpjw+ki)/jw (23)
N(s)=k*[cos(L*w)-j sin(L*w)] (19) ( )
( ) ( ) Where [ *cos( *w), *cos( *w)] |Gc(jw)|= = ( )
(24)
[ *sin( *w), *sin( *w)] (20) ( ) (25)
Instead of writing Kharitonov polynomials for N(s), by ( )

observing from the complex Kharitonov polynomials ( ) ( ) (26)


Case 1: ( ) ( ) (27)
For ( ): is minimum and is minimum ( ) -180- G(s) (28)
is minimum when k is minimum ( ) and L is maximum( )
is minimum when k is maximum(k2) and L is maximum(L1) ( ) -90- G(s) (29)
Case 2:
( ( )) (30)
For ( ): is minimum and is maximum
is minimum when k is minimum ( ) and L is maximum( ) ( ) (31)
( )
is maximum when k is minimum(k2) and L is minimum(L1)
Where ( )
Case 3:
For ( ): is maximum and is minimum Substituting (31) in (25),
is maximum when k is maximum( ) and L is minimum( ) ( ) ( )
(32)
is minimum when k is maximum(k2) and L is maximum(L1)
Case 4: ( ) ( )
(33)
For ( ): is maximum and is maximum
is maximum when k is maximum ( ) and L is minimum( ) ( )
is maximum when k is minimum(k2) and L is minimum(L1) (34)
Consider the denominator D(s) = T*s 1 ( )
(35)
Since T= [ ], additional two cases are possible for the case From(31),
1 to case 4.
From the above discussion, it is clear that the below specified (36)
( )
eight cases are sufficient to analyze G(s) for stability. In the rest 4.2 Phase Margin Calculation
of this paper the validity of these results are verified and the At w=w1,
controller is designed based on it. |Gc(jw1)*G(jw 1)|=1 (37)
|Gc(jw1)|*|G(jw 1)|=1 (38)
Ac *A1=1 whereA1 =|G(jw)|; (39)
Ac=|Gc(jw1)| (40)

185
2014 ACODS
March 13-15, 2014. Kanpur, India

Gc(jw1)= (41)
EXAMPLE 6.1: STABLE FOPTD
= desired phase margin at w1(dpm) –phase margin of Let ( ) 4*exp(-2*s)/(4*s+1)
uncompensated system at w1
Where Gain k [ ] L [1.6, 2.4],T [3.2,4.8]
Gc(jw 1)= G.M≥3, P.M≥ 30
We know that Case1: ( ) 3.2*exp(-1.6*s)/(3.2*s+1) (48)
Gc(s) = (kds2+kps+ki)/s (43) Case 2: ( ) 3.2*exp(-1.6*s)/(4.8*s+1) (49)
Gc(jw 1)= (-kdw12+kpjw1+ki)/jw1 = [ ] (44) Case 3: ( ) 3.2*exp(-2.4*s)/(3.2*s+1) (50)
Case 4: ( ) 3.2*exp(-2.4*s)/(4.8*s+1) (51)
Comparing equations on both the sides, we get Case 5: ( ) 4.8*exp(-1.6*s)/(3.2*s+1) (52)
(45) Case 6: ( ) 4.8*exp(-1.6*s)/(4.8*s+1) (53)
Case 7: ( ) 4.8*exp(-2.4*s)/(3.2*s+1) (54)
(46)
Case 8: ( ) 4.8*exp(-2.4*s)/(4.8*s+1) (55)
(47)
6.1.1 Stability Boundary Locus
By substituting “gm=1” in the stability and gain margin code,
5. MATLAB CODE
the stability boundary locus is obtained. In order to avoid
frequency gridding problem adjust “i” such that “w” is just above
5.1 Stability and Gain Margin code
phase cross over frequency. Similarly draw the boundary locus
This code is valid only for process having non-zero phase cross
for all the cases (48) to (55) by changing “sys” in the code. The
over frequency.
curves obtained are as shown in Fig.2
s=zpk('s');
sys=(3.2*exp(-1.6*s))/(4.8*s+1); 6.1.2 Gain Margin Boundary Locus
d=0; % d=0 for PI controller
w=0; % w is frequency By substituting “gm=3” in the stability and gain margin
gm=3;% for stability gm=1 code, following the same procedure as mentioned in stability
for i=1:40 boundary locus the gain margin boundary locus is obtained.
[mag,phase]=bode(sys,w); 6.1.3 Phase Margin Boundary locus
t1=(-90-phase)*(pi/180);
y(i)=(gm*(mag)); By substituting “dpm=30” in the phase margin code and by
u(i)=(sin(t1))/(y(i)); following the same procedure as mentioned in stability
v(i)=(d*(w^2)+(u(i)*w/tan(t1))); boundary locus, the phase margin boundary locus is obtained.
w=w+0.05; 6.1.4 Robust PI Controller Settings
end
As shown in Fig.3, the gain margin curve for case 1 lies
plot(u,v,'m') %u(i)is ( ),v(i) is ( )
inside the stability curve. It also holds for all the cases .Hence it is
5.2 phase margin code sufficient that by choosing the points in the common region of all
the gain and phase margin curves as shown in Fig.5, the robust
This code is valid only for process having non-zero gain cross PID controller setting which satisfies both robust stability and
over frequency. robust performance specification can be obtained.
s=zpk('s'); EXAMPLE 6.2: UNSTABLE FOPTD
sys=(3.2*exp(-1.6*s))/(4.8*s+1); Let ( ) 4*exp(-2*s)/(4*s+1)
d=0; %d=0 for PI controller where Gain k [ ] L [1.6, 2.4],T [3.2,4.8]
dpm=30; P.M≥ 10
w=0;
Case1: ( ) 3.2*exp(-1.6*s)/(3.2*s-1) (56)
for i=1:50
Case 2: ( ) 3.2*exp(-1.6*s)/(4.8*s-1) (57)
[mag,phase]=bode(sys,w);
Case 3: ( ) 3.2*exp(-2.4*s)/(3.2*s-1) (58)
t=(dpm-phase-180)*pi/180;
Case 4: ( ) 3.2*exp(-2.4*s)/(4.8*s-1) (59)
b(i)=((d*(w^2))-((w/mag)*sin(t)));
Case 5: ( ) 4.8*exp(-1.6*s)/(3.2*s-1) (60)
a(i)=cos(t)/mag;
Case 6: ( ) 4.8*exp(-1.6*s)/(4.8*s-1) (61)
w=w+0.05;
end Case 7: ( ) 4.8*exp(-2.4*s)/(3.2*s-1) (62)
plot(a,b,'m') % a(i)is ( ), b(i) is ( ) Case 8: ( ) 4.8*exp(-2.4*s)/(4.8*s-1) (63)

6.2.1 STABILITY BOUNDARY LOCUS


6. ROBUST CONTROLLER DESIGN
A. PI CONTROLLER
Consider the control system shown in Fig 1.
By following the same procedure as mentioned above, stability
boundary locus is obtained which is shown in Fig.6. From the
figure it is clear there is no common region for all the cases, so it
is not possible to design robust PI controller for 20% increase in
Fig.1 Feedback control system gains. So the choice is PID controller.

Where C(s) is the PI(D) controller, G(s) is the plant

186
2014 ACODS
March 13-15, 2014. Kanpur, India

B. PID CONTROLLER From the Fig.12 it is clear that for k [ ] L [1.6, 2.4],
Derivative gain ( ) is taken as 0.3. By substituting “d=0.3” in T [3.2,4.8] the system is stable , in addition to stability the
the MATLAB code, stability boundary locus for PID controller system has P.M ≥10ᵒ as shown in table 3. By changing the limits
is shown in Fig.7. In order to get the single best point which is of K, L, T and applying the same procedure, robust PID
robustly stable in the considered interval, choose any controller for that system can be designed. If there is no common
optimization technique which minimizes ISE(or) IAE etc.[6]. region then increase kd and apply the same procedure.
To show the validity of the proposed method for unstable
FOPTD only phase margin is taken. 7.4 Comparison of PI controller designed for unstable FOPTD
6.2.2 PHASE MARGIN BOUNDARY LOCUS using DOMINANT POLE PLACEMENT and EQUATING
For this process P.M≥10ᵒ. Hence by substituting “dpm=10” in COEFFICIENT methods
the MATLAB code, phase margin boundary locus is obtained. The settings of PI controller designed using DOMINANT
6.2.3 ROBUST PID CONTROLLER SETTINGS POLE PLACEMENT method is kp=0.3557, ki=0.012. [18]
The common region of stability and Phase Margin boundary The settings of PI controller designed using EQUATING
locus is shown in Fig.9. The point which is taken in the common COEFFICIENT method is kp=0.4912, ki=0.014. [19]
region is the robust PID controller which satisfies both the From the Fig. 13 it is clear that controller designed using
stability and P.M specification. Dominant Pole placement technique is more robust than Equating
Coefficient method, but both methods are not robust in the
7. VALIDATION OF PROPOSED METHOD considered interval. Furthermore, Simulation studies shows that,
7.1 Servo and Regulatory Response of Stable FOPTD in order to design a robust controller for unstable FOPTD system,
PID controller should be preferred to PI controller.
The transfer function model ( ) 4*exp(-2*s)/(4*s+1) is
considered. The PID parameters taken are kp=0.15, ki=0.04, kd ( )
=0. For unit step change in the input at t=0 and for unit The Equating coefficient tuning rule for ( )
( )
disturbance at t=100, the servo and regulatory response of all the is
cases for the fixed PI settings is as shown in the Fig.10
Where ,
From the Fig.10 it is clear that for k [ ] L [1.6, 2.4],
T [3.2,4.8] the system is stable , in addition to stability the
system has G.M≥3 and P.M ≥30ᵒ as shown in table 2. Just by
changing the limits of K, L, T and applying the same procedure, The dominant pole placement tuning rule is
robust PID controller for that system can be designed. ( )

7.2 Comparison of PI controller designed for stable FOPTD ( )


using SIMC technique
From the Fig. 11 it is clear that the PI controller designed
using SIMC technique by changing closed loop time constant ( )
falls in the common stability region, hence the controller designed
using SIMC technique is stable in the considered interval. It is
also clear that with increase in robustness increases but the
response will be sluggish.
( )
The SIMC rules for ( ) are =
( ) ( )
( ( ))

7.3 Servo and Regulatory Response of Unstable FOPTD


The transfer function model ( ) 4*exp(-2*s)/(4*s-1) is
considered. The PID parameters taken are kp=0.36, ki=0.004, kd
=0.3. For unit step change in the input at t=0 and for unit
disturbance at t=500, the servo and regulatory response of all the
cases for the fixed PID settings is as shown in the Fig.12.

Fig.2. Stability curves of all the 8 cases (stable FOPTD, =0 is fixed)

187
2014 ACODS
March 13-15, 2014. Kanpur, India

Fig.3. Stability and Gain margin Boundary Locus for the pi controller (stable FOPTD)

Fig.4. Gain and Phase margin boundary locus of all 8 cases (stable FOPTD)

Fig.5.Gain& Phase margin boundary locus common region of 8 cases (stable FOPTD)

188
2014 ACODS
March 13-15, 2014. Kanpur, India

Fig.6.Stability curves of all the 8 cases (unstable FOPTD, =0 is fixed)

Fig.7. Stability curves of all the 8 cases (unstable FOPTD, =0.3 is fixed)

Fig.8. Stability and Phase margin boundary locus of 8 cases (unstable FOPTD, =0.3 is fixed)

Fig.9.Common region of Stability and Phase margin Boundary Locus (unstable FOPTD, =0.3 is fixed)

189
2014 ACODS
March 13-15, 2014. Kanpur, India

Fig.10. Servo and Regulatory response of 8 cases ( =0.15, =0.04, ) -stable FOPTD

Fig.11.SIMC PI controller gains in stability boundary locus of 8 cases

Fig.12. Servo and Regulatory response of 8 cases ( =0.36, =0.004, ) -unstable FOPTD

Fig.13. Comparison of DOMINANT POLE PLACEMENT and EQUATING COEFFICIENT method


PI controllers

190
2014 ACODS
March 13-15, 2014. Kanpur, India

8. CONCLUSIONS 13. V. Nageswara Rao , A. Seshagiri Rao, and R. Padma Sree,


In this work, Kharitonov theorem is used to propose new rule for Design of PID Controllers for Pure Integrator Systems with
the class of delay time process which reduces the time taken to Time Delay, International Journal of Applied Science and
design robust PID controller. The validity of the new rule is Engineering 2011. 9, 4: 241-260
applied to both stable and unstable FOPTD process and 14. Padma Sree, R. and Chidambaram, M. 2005. A Simple
simulation studies shows that the controller designed based on method of tuning PID controllers for integrator/dead-
proposed method makes the system stable over the time processes. Journal of Chemical Engineeringof Japan,
predetermined range and obey the performance specifications. It 38: 113-119.
is also shown that, to design a robust controller for unstable 15. Seshagiri Rao, A., Rao, V. S. R. and Chidambaram, M.
FOPTD, PID controller should be preferred to PI controller. 2009. Direct synthesis based controller design for
integrating processes with time delay. J. Franklin Inst, 346:
In future work, optimization methods like particle swarm, 38-56
Genetic algorithm, method of inequalities etc. will be applied on 16. D. D. Siljak, “Generalization of the Parameter
the common region of all cases to find single PI(D) setting which PlaneMethod,” 1966,11(1): 63-71
minimizes ISE (or) IAE (or) ITAE and also reduces the 17. Ch. Anil and R. Padma Sree“Design of PID Controllers for
convergence time significantly. FOPTD Systems with an Integrator and with/without a
Zero” Indian Chem. Engr., Section A, Vol. 47, No. 4
REFERENCES 18. M. chidambaram: APPLIED PROCESS CONTROL, pages
1. Chien, L. and Fruehauf, P. S. 1990. Consider IMC tuning to 96-102
improve performance. Chem Eng Progr, 10: 33-41. 19. R.vilanova,A.visoli: PID CONTROL IN THIRD
2. Fuentes, C. and Luyben, W. L. 1983.Control of high purity MILLENIUM,ADVANCES IN INDUSTRIAL CONTROL
distillation columns.Ind Eng Chem Process Des Dev,22: , pages 97-101
362.
3. Y. J. Huang and Y. J. Wang, “Robust PID Tuning Strategy
for Uncertain Plants Based on the Kharitonov Theorem”, TABLE 2. GAIN MARGIN AND PHASE MARGIN OF
ISA Transactions, 2000, 39(4): 419-431 VARIOUS TRANSFER FUNCTIONS
4. Y. J. Huang and Y. J. Wang, “Robust PID Controller
Design for Non-minimum Phase Time Delay Systems,” ISA
Transactions, 2001,40(1): 31-39.
5. WANG Yuan-Jay, HUANG Jian-Jie, SHIH Shang-Hong,
LAI Yan-Chang, DING Jia-Jun, “ Graphical Computation
of Gain and Phase Margin Specifications-oriented Robust
PID Controllers for Uncertain Systems with Time-varying
Delay”, ISA transactions journal of process control,
Volume 21, Issue 4, April 2011, pages 475-478
6. Yuan-Jay Wang “Determination of all feasible robust PID
controllers for open-loop unstable plus time delay processes
with gain margin and phase margin specifications” ISA
Transactions (2014)
7. S. P. Bhattacharyya, H.chapellat, L.H.keel: ROBUST TABLE 3. PHASE MARGIN OF VARIOUS TRANSFER
CONTROL, The Parametric Approach. Prentice Hall FUNCTIONS
PTR(1995)
8. G. J. Silva, A. Datta and S. P. Bhattacharyya., "PI
stabilization of first order systems with time-delay",
Automatica, Vol. 37, No.12, pp.2025–2031, 2001.
9. Sami Elmadssia, Karim Saadaoui, and Mohamed Benrejeb,
“PI Controller Design for Time Delay Systems Using an
Extension of the Hermite-Biehler Theorem”, Journal of
Industrial Engineering, Volume 2013, Article ID 813037, 6
pages.
10. Asma Karoui, Rihem Farkh, Moufida Ksouri , “PID
controller design for multiple time delays system”,
International Conference on Control, Engineering &
Information Technology (CEIT'13),Proceedings
Engineering & Technology - Vol.1, pp. 146-152.
11. Nusret Tan and Ibrahim Kaya, “Computation of stabilizing
PI and PID controllers using the stability boundary locus”,
ISA transcations on Energy Conversion and Management
47 (2006) 3045–3058
12. Celaleddin Yeroglu and Nusret Tan “Design of Robust PI
Controller for Vehicle Suspension System”, Journal of
Electrical Engineering & Technology, Vol. 3, No. 1, pp.
135~142, 2008

191

You might also like