Professional Documents
Culture Documents
ISA Transactions
journal homepage: www.elsevier.com/locate/isatrans
Parameter estimation of integrating and time delay processes using single relay
feedback test
Rames C. Panda a, , V. Vijayan a , V. Sujatha a , P. Deepa a , D. Manamali b , A.B. Mandal a
a
Chemical Engineering Department, CLRI (CSIR), Adyar, Chennai 600 020, India
Department of Instrumentation, Anna University, MIT Campus, Chrompet, Chennai 44, India
article
info
Article history:
Received 25 January 2011
Received in revised form
16 April 2011
Accepted 21 June 2011
Available online 20 July 2011
Keywords:
Relay feedback
Identification
Autotuning
Integrating process
PID controller
abstract
Autotuning using relay feedback is widely used to identify low order integrating plus dead time (IPDT)
systems as the method is simple and is operated in closed-loop without interrupting the production
process. Oscillatory responses from the process due to ideal relay input are collected to calculate ultimate
properties of the system that in turn are used to model the responses as functions of system model
parameters. These theoretical models of relay response are validated. After adjusting the phase shift,
input and output responses are used to find land mark points that are used to formulate algorithms for
parameter estimation of the process model. The method is even applicable to distorted relay responses
due to load disturbance or measurement noise. Closed-loop simulations are carried out using model based
control strategy and performances are calculated.
2011 ISA. Published by Elsevier Ltd. All rights reserved.
1. Introduction
Processes with long time delays are found in many chemical and
biochemical processes. Chemical and biochemical process systems
with a huge storage facility using inlet/exit valves can be approximated as IPDT [1] systems for designing model based control systems. Recently, these systems have drawn much attention as their
openloop step responses show increasing behavior and need to be
controlled for safe operation. Liquid level with a surge tank, water
level in steam boiler drum and pressure of boiling liquid are examples of such processes. Many researchers have attempted controller design for such processes. Tyreus and Luyben [2] presented
PI tuning rules of IPDT systems and in a separate attempt [3] discussed temperature control of an openloop unstable reactor. Ho
et al. [4] obtained stabilized response by using a derivative block
before IPDT process. Rice and Cooper [5] designed tuning rules exclusively for self-regulating systems to improve closed-loop performance. Scali and Rachid [6], Zhang et al. [7], Chien et al. [8],
Dwyer [9] also discussed control techniques for integrating systems. Kaya [10] considered parameter estimation using relay response under load disturbances. Recently, Liu and Gao [11] derived
theoretical models for relay responses of IPDT systems using biased/unbiased relay. Their presentation did not include IPDT systems with a zero in the numerator, relay identification under load
variation and they used IMC strategy for control. Many practical
systems have drawn much attention as they behave as an IPDT system with a zero in the numerator. Generally, the presence of an
integrator term
KP eDs
s
0019-0578/$ see front matter 2011 ISA. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.isatra.2011.06.004
530
are not clearly explained/understood in earlier research. Hence integrating plus dead time systems need to be studied for safe operation and control of process. In order to get desired performances
of the process output, a well-designed and properly tuned controller is needed. Any controller that is designed nearer to its ultimate frequency gives better performance. Model based controllers
require exact model parameters for controller tuning. Autotuning
can be used to identify model structures [12] and estimate for tuning controller parameters. It has got two steps, identification and
controller design phase. In this paper relay feedback approach is
used to identify model parameters of integrating processes from
which controller parameters are calculated using proper model
based tuning rules. Three typical model structures for integrating
systems, namely, integrator plus dead time (IPDT ), Integrating second order plus dead time (ISOPDT ) and Integrating second order
plus dead time with a zero (ISOPDTZ ), are considered for autotuning in this work. Emphasis has been given on ISOPDTZ type systems
as they pose many difficulties during control. The entire paper is
organized as follows: at first, mathematical models for relay responses are developed in Section 2. The relay responses obtained
from theoretical equations are validated against experimental response and parameter estimation algorithms are formulated using
land mark points from response curves in Section 3. In Section 4,
PID tuning algorithm is presented and discussed. In Section 5, typical examples are considered whose relay response models are derived, validated, identified and are treated for closed-loop control
using the autotuning theory developed in previous sections.
2. Development of analytical expressions
According to Astrom and Hagglund [13] and Yu [14], when a
process is subjected to relay inputs (height, h), the closed-loop
process output starts oscillating around its steady operating state
and lags behind the input by radians. The oscillatory response
can be used to calculate ultimate properties of the system (ultimate
gain, Ku = 4ha ; where, a = amplitude of response and ultimate
0.25(s+1)e1.0s
s(2.0s+1)
that as soon as a
(1)
where a1 = P Pz .
P
After shifting the input signal and synchronizing with the
output signal (Fig. 2), the 1st segment of the relay response is the
step response without time delay and can be given as
y1 = KP [uDp + t a1 P (1 et /P )].
(2)
2KP [t a1 P (1 et /P )]
= KP [uDP + (1 2)t
DP a1 P {(1 2) et /P (eDP /P 2)}].
(3)
y 3 = KP
uDp + t + DP +
Pu
2
a1 P 1 e
Pu
2KP t +
a1 P 1 e(t +DP )/P
2
+ 2KP t a1 P 1 et /P
= KP uDp + (1 2 + 2)t + DP +
Pu
2
Pu
+ 2e
(t +DP )/P
t /P
2e
(4)
Pu
ym = KP uDP + t DP
where c =
2
1 + ePu /2P
a1 P 1 et /P .c
(5)
(6)
yn = KP uDP + t + DP + (1)n
Pu
KP eDP s
,
s
with n = even.
(7)
K eDP s
For ISOPDT system, GP = s(P s+1) , the relay response equations can
P
be formulated as
yn = KP uDP + t + (1)n P DP
+ P et /P
2
1+
ePu /2P
Model structure
KP eDP s
s
K eDP s
yn = KP h[uDp + t + DP Pu /2]
2et /P1
1+ePu /2P1
2et /P1
(1+ePu /2P1 )
s(P1 s+1)
yn = KP h uDp + t DP Pu /2 P1 1
KP (P3 s+1)eDP s
s(P1 s+1)
yn = KP h uDp + t DP Pu /2 a1 P1 1
) .
531
Table 1
Analytical expressions for relay responses.
Pu
3.2. Identification of transfer functions
2
(8)
[
KP =
pu /2
] [
ydt
udt .
(9)
(10)
(11)
Time delay (DP ) can be found (DP = Pu /4) from initial part of the
relay response. The above two equations can be solved to get DP
and then substituting this value in the response equation, KP can
be found for any arbitrary response data (y) for a time value (t ).
Thus one can have
y
KP =
(12)
h(uDp + t Pu /4)
532
Table 2
Algorithms for identification of model structures along with their parameters.
Model structure
KP eDP s
s
K eDP s
Model parameters
Algorithm
KP and DP
DP = Pu /4 and KP =
KP , P1 and DP
s(P1 s+1)
KP (Pz s+1)eDP s
s(P1 s+1)
P1 from P1 ln
KP , P1 , P3 and DP
(13)
(14)
P1 1
1 + ePu /2P1
= uDp 2DP .
(15)
Parameters in the RHS of the above equation are known and can be
substituted to find out P1 Eq. (14) can be used to find out process
gain (KP )
KP =
h uDp D P1 1
(16)
Thus Eqs. (15) and (16) can be solved to get (P1 ) and (KP ).
(c) ISOPDTZ : These types of processes exhibit two types of
responses depending on the time delay value of the transfer
function. It has four unknown parameters, namely, process gain
(KP ), process time constant (P1 ), numerator zero (Z ) and time
delay (DP ). After observing and analyzing the response (Fig. 2), the
boundary conditions can be formulated using land mark points as:
In case the responses have rounded peaks (non-monotonic
response), one can have, the slope at the optimum point (minimum
or maximum) must be zero
dym
=0
(17)
(18)
(19)
dt
t =tPeak
(20)
kP
P1 (1 + e0.5Pu /P1 )
= 0.
(21)
a
KP h
h(Pu /4)
2
1+ePu /2P1
a
h(Pu /4)
= tPeak and KP =
2 (e
D P1
a
tPeak /P1 e(D +tPeak )/P1 )
(1+e0.5Pu /P1 )
1
= 0 and
P1 and P3 from tPeak P1 ln (1+e2a
0.5Pu /P1 )
t
a
(tpeak +D )
peak
a1 P1 c e
/p1 e
/p1 = KP h + D D + tpeak + a1 p1 (ce(D +tpeak )/p1 1) =
2e(0.5Pu D )/P1
y|t =0,Pu /2
(b) ISOPDT systems: The responses in this case have been observed
to be non-monotonic. There are three unknown parameters,
namely, process gain (KP ), process time constant (P1 ) and time
delay (DP ). The boundary conditions that can be formulated for
identification of these parameters are
+ D .
(22)
Pu
4
Pu
.
(23)
D + tpeak + a1 p1 (ce(D +tpeak )/p1 1) =
4
Substituting condition (20) into the model equation (6) gives
KP =
h uDp D P1 1
P
/
2P1
u
1 +e
(24)
|GP GC (j)| =
2
2
(2k
+ 4k + 1)(DP )
1 + (DP /2)2
2
1 + (k /(2k2 + 4k + 1))2 (DP /2)2
DP
GP GC (j) = tan1 ((2k + 1)DP ) + tan1
2
2
k
tan1
DP DP
2k2 + 4k + 1
2
where k = /DP with as an user defined parameter.
(25)
(26)
533
Table 3
Ultimate properties and estimated transfer functions for different examples.
Example
Amplitude, a
Period of oscillation, Pu
6s
Ex-1 = 0.0506e
s
0.3289
26
0.2s
Ex-2 = 1.s0e
(s+1)
0.4338
3.9986
Ex-3 =
0.5374
3.9772
0.4750
11.9994
Ex-4 =
1.0(s+1)e0.5s
s(2s+1)
0.25(s+1)
s(2s+1)
0.0506e6s
s
1.0e0.2s
s(0.9998s+1)
1.0353(1.0027s+1)e0.5s
s(2.0060s+1)
0.252(s+1)
s(2s+1)
Fig. 3. Plot of performance (IAE) vs. lambda () during (a) set point change (optimum resulted is 0.124 while corresponding IAE is 7.3) (b) load disturbance (optimum
resulted is 0.07695 while corresponding IAE is 46.17) with Ex-2 and present control scheme.
D s
1+
+
I s D s + 1
1
(27)
534
Fig. 4. Comparison of closed-loop responses of liquid level control for a boiler drum (ISOPDTZ ) process using different PID tuning rules.
Table 4
Comparison of performances of present PID tuning scheme with similar available schemes.
Process
0.547(0.418s+1)e0.1s
s(1.06s+1)
0.25(s+1)e1.0s
s(2.0s+1)
= CEx-1
= CEx-2
Controller parameters
Performance
KC
MSER
MSEL
Present
Gu et al. [17]
0.3
0.3
2.2093
1.3821
158.93
4.6308
1.43
1.42
18.372
29.391
13.198
14.185
Present
Gu et al. [17]
0.3
0.3
1.3957
1.5243
158.04
14.742
2.786
3.448
24.803
32.642
17.95
17.76
535
Fig. 7. Relay feedback responses for an IPDT process [Gp (s) = 0.2 exp(7.4s)/s]
with measurement noise, NSR = 0.01.
Fig. 5. Sensitivity of present controller for a perturbation in dead time (+35%
solid line, nominal valuedash dot line, 35% dashed line) of the process model
CEx-2.
Fig. 6. Use of output biased relay to restore the correct shape under load
disturbance for an IPDT system (transfer function: 0.2 exp(7.4s)/s).
(s+1)e5s
s(s+1)5
10
(28)
536
(s+1)e5s
s(s+1)5
Method
Present
0.945e7s
s
0.7319e5s
3.1953s(2.0051s+1)
1.0027(2.006s+1)e5s
s(1.0353s+1)
= 320
MSE = 3.9 103
Tao and Gao [11]
0.9664e10.97s
s
= 955
MSE = 1.09 102
= 294
MSE = 0.003
1.018e8.5278s
s(2.5293s+1)
6. Conclusions
Fig. 8. Comparison of closed-loop control of liquid level (real time and simulated
level).
will hold good for analyzing the stability of the closed-loop system
(while the openloop process is unstable). Let,
GP (s) =
GC (s) =
KP (z s + 1)eDP s
s(P s + 1)
KC [I D s2 + I s + 1]
I s
and
(29)
(30)
(31)
where
References
p 0 = KC KP
(32)
p1 = KC KP (z + I DP )
(33)
p2 = KC KP ((I z )DP + I (D z )) + I
p 3 = KC KP
DP I (z D ) I z D +
I P
(34)
KP KC
p4 = KP KC DP I z D ,
(35)
(36)
(37)
K12 (s) =
pL0
K21 (s) =
pU0
K22 (s) =
pU0
pU1 s
pL1 s
pU2 s2
pL2 s2
pL3 s3
pU3 s3
pL4 s4
(38)
pU4 s4
(39)
+
+ .
(40)
As the process parameters (KP , P , DP and z ) are perturbed,
+
pU1 s
pL2 s2
pL3 s3
pU4 s4
[1] Chien IL, Fruehauf PS. Consider IMC tuning to improve performance. Chem Eng
Prog 1990;3341.
[2] Tyreus BD, Luyben WL. Tuning PI controllers for integrator/dead time
processes. Ind Eng Chem Res 1992;31:26258.
[3] Luyben WL. Tuning proportional-integral-derivative controllers for integrator/dead time processes. Ind Eng Chem Res 1996;35:34803.
[4] Ho WK, Feng EB, Gan OP. A novel relay auto-tuning technique for processes
with integration. Control Eng Pract 1996;4(7):9238.
[5] Rice B, Cooper DJ. Design and tuning of PI controllers for integrating (non-self
regulating) processes. In: Proc. ISA 2002 annual meeting. 424. 2002. P057.
[6] Scali C, Rachid A. Analytical design of proportional-integral-derivative
controller for inverse response processes. Ind Eng Chem Res 1998;37:13729.
[7] Zhang WD, Xu XM, Sun YX. Quantitative performance design for inverse
response processes. Ind Eng Chem Res 2000;39:205661.
[8] Chien IL, Chang YC, Chen BS, Chuang CY. Single PID controller tuning method
for processes with inverse response plus dead time or large overshoot
response plus dead time. Ind Eng Chem Res 2003;42:446177.
[9] Dwyer AO. Handbook of PI and PID controller tuning rules. 3rd ed. London:
Imperial College Press; 2009.
[10] Kaya Ibrahim. Parameter estimation if integrating processes using relay
feedback control under static load disturbance. Ind Eng Chem Res 2006;
45(13):472631.
[11] Liu Tao, Gao Furong. Identification of integrating and unstable processes from
relay feedback. Comput Chem Eng 2008;32(12):303856.
[12] Panda RC, Yu Cheng-Ching. Analytical expressions of relay feedback responses.
J Process Control 2003;13(3):489501.
[13] Astrom KJ, Hagglund T. Automatic tuning of simple regulators with
specifications on phase and gain margins. Automatica 1984;20:64551.
[14] Yu CC. Autotuning of PID controllers. London: Springer-Verlag; 2006.
[15] Panda RC, Yu Cheng-Ching. Shape factor of relay response curves and its use
in autotuning. J Process Control 2005;15(8):893906.
537
[19] Chidambaram M, Padma Sree R. A simple method of tuning PID controllers for integrator/dead time processes. Comput Chem Eng 2003;27:
211215.
[20] Panda RC. Synthesis of PID tuning rules using desired closed-loop response.
Ind Eng Chem Res 2008;47(22):868492.
[21] Luyben WL. Identification and tuning of integrating processes with dead time
an inverse response. Ind Eng Chem Res 2003;42:30305.