ISA Transactions 63 (2016) 242–255

Contents lists available at ScienceDirect

ISA Transactions
journal homepage: www.elsevier.com/locate/isatrans

Research Article

Tuning of IMC based PID controllers for integrating systems

with time delay
D.B. Santosh Kumar, R. Padma Sree n
Department of Chemical Engineering, AU College of Engineering (A), Visakhapatnam 530003, India

art ic l e i nf o a b s t r a c t

Article history: Design of Proportional Integral and Derivative (PID) controllers based on IMC principles for various types
Received 30 October 2015 of integrating systems with time delay is proposed. PID parameters are given in terms of process model
Received in revised form parameters and a tuning parameter. The tuning parameter is IMC filter time constant. In the present
9 March 2016
work, the IMC filter (Q) is chosen in such a manner that the order of the denominator of IMC controller is
Accepted 30 March 2016
Available online 14 April 2016
one less than the order of the numerator. The IMC filter time constant (λ) is tuned in such a way that a
This paper was recommended for publica- good compromise is made between performance and robustness for both servo and regulatory problems.
tion by Panda Rames. To improve servo response of the controller a set point filter is designed such that the closed loop
response is similar to that of first order plus time delay system. The proposed controller design method is
Keywords: applied to various transfer function models and to the non-linear model equations of jacketed CSTR to
Integrating systems
demonstrate its applicability and effectiveness. The performance of the proposed controller is compared
Unstable system
with the recently reported methods in terms of IAE and ITAE. The smooth functioning of the controller is
Time delay
PID controller determined in terms of total variation and compared with recently reported methods. Simulation studies
IMC method are carried out on various integrating systems with time delay to show the effectiveness and superiority
Ms value of the proposed controllers.
& 2016 ISA. Published by Elsevier Ltd. All rights reserved.

1. Introduction
Integrating systems are the processes which contain at least
one pole at the origin. They are non-self regulating, which means
that when they are disturbed from the equilibrium operating point
by any environment disturbance/change in input conditions, the
process output varies continuously with time at certain speed. The
phenomenon is very disadvantageous and dangerous in most
occasions. Therefore, efficient control of such kind of processes is
always a challenging task. Different types of integrating systems
exist depending upon the number of poles present at the origin
and location of other poles in transfer function. Accordingly inte-
grating systems are classified as stable First Order plus Time Delay
systems with an Integrator (FOPTDI) Unstable First Order plus
Time Delay systems with an Integrator (UFOPTDI), Pure Integrating
plus Time Delay (PIPTD) systems and Double Integrating plus Time
Delay (DIPTD) systems.
Industrial processes such as composition control loop of a high
purity distillation column [1], bottom level control in a distillation
column [2], storage tank with a pump at the outlet [3], many level
control problems [4], an isothermal continuous copolymerization
n
Corresponding author.
E-mail address: padvan@gmail.com (R. Padma Sree).
reactor [5], the heating of well insulated batch systems [6], totally
heat integrated distillation columns [7] and high pressure steam
flowing to a steam turbine generator in a power plant [8], exhibit
Pure Integrator plus Time Delay transfer function models.
First order systems with an integrator and with/without zero
are frequently encountered in process industries. The occurrence
of such transfer function models is reported for the liquid storage
tanks [9], paper drum dryer cans [4] and a jacketed Continuous
Stirred Tank Reactor (CSTR) carrying out an exothermic reaction
[10]. For First Order plus Time Delay systems with an Integrator
(FOPTDI) and with a positive zero, the system gives inverse
response. The inverse response becomes deeper as the zero moves
towards the origin on the real axis, which is a tough challenge for
process control.
Double integrating systems exist in processes such as aerospace
control systems, vertical take-off of airplanes [11], DC motors and
high speed disk drives [12], oxygen control in feed batch (filament
fungal) fermentation reactors [13].
PID controllers are widely implemented in many of the che-
mical process industries because they are very simple to tune, easy
to understand and robust in control. PID control is the most
common control algorithm used in industry and has been uni-
versally accepted in industrial control. The popularity of PID con-
trollers can be attributed partly to their robust performance in a
wide range of operating conditions and partly to their functional

http://dx.doi.org/10.1016/j.isatra.2016.03.020
0019-0578/& 2016 ISA. Published by Elsevier Ltd. All rights reserved.
 D.B.S. Kumar, R. Padma Sree / ISA Transactions 63 (2016) 242–255
simplicity, which allows engineers to operate them in a simple,
straight forward manner.
In literature various methods are proposed to tune PID con-
trollers for integrating systems with time delay. They are empirical
method [14,15], Internal Model Control (IMC) method [2,16–22],
direct synthesis method [23–28], equating coefficient method
[29,30], two degree of freedom (2DOF) control scheme [31–33],
stability analysis method [34–37] and optimization method [38–
43]. Various methods to control different types of integrating
processes are given by Visioli and Zhang [44].
Seshagiri Rao et al. [25] proposed PID controller tuning rules for
integrating processes with time delay using direct synthesis
method. For PIPTD system the closed loop is assumed to be second
order system with first order numerator and time delay. For stable/
unstable FOPTDI system and DIPTD systems, the closed loop
transfer function is assumed to be third order with second order
numerator and time delay. From the closed loop transfer function,
using process transfer function controller transfer function is
obtained. First order Pade's approximation is used for time delay
in the denominator of controller transfer function, so that PID
controller parameters are obtained in terms of process model
parameters.
Panda [19] proposed a PID controller for pure integrating plus
time delay system based on IMC method. To avoid singularity
problem, Laurent series is used to derive PID controller settings.
The given tuning rules are logically applied to low order open loop
unstable process plant model with time delay. The tuning factor λ
was selected based on faster/sluggish response. The controller is
robust, stable and can be implemented easily on a real time
processes.
Nageswara Rao and Padma Sree [20] proposed a PID controller
based on IMC principles for integrating systems with time delay.
They have used first order Pade's approximation for process time
delay in the process itself. For that process, IMC controller is
designed and IMC filter is selected such that the numerator order
is equal to or less than the denominator order. From IMC con-
troller, PID controller is designed. PID parameters are given in
terms of model parameters. Anil and Padma Sree [43] proposed
tuning rules for PID controllers using differential evolution algo-
rithm to minimize Integral Time weighted Absolute Error (ITAE).
To reduce overshoot in servo problem, set point weighting is also
suggested.
Ajmeri and Ali [27] have proposed parallel control structure
that decouples servo problem and regulatory problem for PIPTD,
DIPTD and FOPTDI systems. For servo problem, proportional and
derivative (PD) controller is used and for regulatory problem, PID
controller is used. The controller is implemented as parallel form
of PD/PID controllers. Analytical tuning rules are proposed for PD
and PID controllers based on direct synthesis method. The tuning
parameters are tuned in such a way to achieve the desired
robustness. Ajmeri and Ali [27] have reported the PD/PID para-
meters for a maximum magnitude of sensitivity function, Ms ¼ 2
for PIPTD systems. The performance of the method is reported in
terms of Integral Square Error (ISE), Integral Absolute Error (IAE),
TV and settling time.
Shamsuzzoha [45] proposed analytical tuning rules for closed
loop PI/PID controller for stable and integrating process with time
delay. This method requires a closed loop step set point experi-
ment using a proportional (P) only controller with gain K co : In P
mode, a step change is given to the system such that the overshoot
is 30% .On the basis of this simulation results for a first order plus
time delay processes, simple correlations are derived to give PI/PID
controller settings. The controller gain (K C =K C o ) is the only func-
tion of the overshoot observed in the set point experiment. The
controller integral (τI ) and derivative time (τD ) is mainly a function
of the time of reach the first peak (t P ).
243
Cho et al. [28] proposed simple analytical PID controller tuning
rules for unstable process based on direct synthesis method. The
closed loop transfer function model is assumed to be third order
with second order numerator and time delay term. From the
closed loop transfer function, using process transfer function,
controller transfer function is obtained. The process time delay is
approximated using Taylor's series expansion up to two terms. The
tuning rules are derived and PID settings are tuned such that
trade-off between performance and robustness is achieved.
Lee et al. [26] revised the tuning rules of SIMC method pro-
posed by Skogestad [47] for PID controller. They suggested mod-
ifications of model reduction techniques proposed by Skogestad
[47].
Jin and Liu [21] have proposed 2DOF scheme using IMC prin-
ciples with an extra set point filter for PIPTD, DIPTD and FOPTDI
systems. They have designed IMC controller with numerator order
one greater than the denominator order. The conventional PID
controller is designed and implemented as parallel form PID
controller. An optimization problem is formulated with an objec-
tive function of IAE for regulatory problem with robustness
(Ms ¼2) as a constraint. The servo performance is improved by
using an extra set point filter. Analytical tuning rules are reported
for PID parameters and for an extra set point filter. The perfor-
mance is reported in terms of IAE and TV for both servo and
regulatory problem. Nageswara Rao and Padma Sree [33] proposed
a design to two degree of freedom PID controller for double
integrator with time delay systems and unstable First Order plus
Time Delay systems with an integrator. DIPTD systems and
UFOPTDI systems are stabilized in the inner loop by using PD
controller. To the stabilized system, PID controllers are designed by
equating the denominator of the stabilized system with the
numerator of the outer loop PID controller. Using the phase angle
criteria for the combined stabilized system and outer loop PID
controller, the cross over frequency (ωC ) is obtained. The ultimate
value of controller gain (outer loop) is calculated by using gain
margin (GM) criteria. Usually the GM of 1.5–2.5 is used to obtain
the design value of controller gain of the outer loop controller. Liu
et al. [51] have proposed two degree of freedom control structure
in which set point tracking and load disturbance loops are
decoupled. Set point tracking controller is PD controller and the
parameters are obtained analytically by specifying the Integral
Squared Error (ISE). By proposing the desired closed loop com-
plementary sensitivity function for rejecting load disturbances,
disturbance estimator is designed. Robust stability analysis for the
proposed control structure is provided in the presence of the
process multiplicative uncertainty. Liu and Gao [52] proposed
modified IMC based controller design for step and ramp type load
disturbance. The set point tracking is decoupled with the load
disturbance rejection with separate loops for both. The tuning
parameter is the closed loop time constant for load disturbance
rejection and tuned to meet a good trade-off between perfor-
mance and closed loop stability.
Review of literature reveals that though there are many
methods available to design PID controllers for integrating sys-
tems, still there is a scope to improve the performance and
robustness of the PID controller for integrating systems. Many
authors proposed a complicated structure with more than one
controller for control of integrating processes [12,27,51,52].
Therefore in the present work, design of PID controller with a
compensator for integrating systems with time delay to enhance
the performance for both servo and regulatory problems using
IMC principles is proposed. If the process is represented by a
perfect model with no modeling errors and if the model is inver-
tible, then the IMC controller is the inverse of the process model
and no IMC filter is required. But in the presence of modeling
errors and if the process model contains non-invertible parts like
244 D.B.S. Kumar, R. Padma Sree / ISA Transactions 63 (2016) 242–255

time delay and non-minimum phase zeros, IMC filters are chosen From IMC controller, the conventional feedback PID controller
to account for the non-invertible part of the process transfer is obtained by
function and to account for the robustness for parametric uncer- Q
tainty. Therefore, the selection of IMC filter depends on the process GC ¼ ð6Þ
1  GP Q
transfer function model.
IMC controller is the inverse of the invertible part (of the Substituting Eqs. (1) and (5) in Eq. (6), the PID controller for the
process transfer function) multiplied by IMC filter. IMC filter con- process is given by
 
sists of both numerator and denominator dynamics. The order of qðq þ 1Þ γ 0 q þ 1
the numerator is equal to the number of unstable poles (and/or GC ¼ h    i ð7Þ
0 2
K 0 P λ q þ 1  γ 0 q þ 1 e  ϵq
integrators). Usually the denominator order of the IMC filter is
chosen in such a way that the order denominator of the IMC Using first order Pade's approximation for time delay, the
controller is the same or more than the order of the numerator, so controller is written as
that it is realizable.  
In the present work, the order of the IMC filter is selected in qðq þ 1Þ γ 0 q þ 1 ð1 þ 0:5ϵqÞ
GC ¼ h  0  2   i ð8Þ
such a way that the denominator of IMC controller is one order K 0P λ q þ1 ð1 þ 0:5ϵqÞ  γ 0 q þ 1 ð1  0:5ϵqÞ
less than the numerator order. Actually such IMC controller if
implemented in IMC structure gives an error message as it is not Rearranging the above equation, the following equation for GC
realizable. But in the present work, from the IMC controller, PID is obtained.
controller is designed and implemented in the conventional    
q γ 0 q2 þ 1 þ γ 0 q þ 1 ½1 þ 0:5εq
feedback control loop. The closed loop transfer function model, GC ¼
2   
0 2 0 0 0
K 0P ½q3 ð0:5ελ Þ þ q2 λ þ λ ε þ 0:5γ 0 ε þ q 0:5ε þ 2λ  γ 0 þ 0:5ε 
which is the non-invertible
part of the process transfer function
multiplied by the IMC filter y
¼ Gp þ f , is still realizable. By fol- ð9Þ
yr
lowing this procedure the order of the closed loop transfer func- In order to represent Gc as PID controller, the constant term of
0
tion is reduced and the response is fast. The method is applied to denominator ϵ þ 2λ  γ 0 is equated to zero. Thus the value of γ 0 is
PIPTD, DIPTD, stable/unstable FOPTDI systems. The method has a obtained as
single tuning parameter λ, IMC filter time constant which is to be
γ 0 ¼ ε þ 2λ0 : ð10Þ
selected in such a way that the controller gives good performance
and is robust for parameter uncertainty. GC is written as
The closed loop response of integrating processes result in large 
 
1 þ γ 0 1 þγ γ 0 q þ γ 0 þ1 1 q þ1 ½1 þ 0:5ϵq
0
overshoot and to reduce the overshoot set point filter is suggested. ð Þ
In literature many authors used set point filter [21,22,28,48] to GC ¼
 02
ð11Þ
02 0 0:5ϵλ
improve servo performance. In the present work also set point K 0P λ þ λ ϵ þ 0:5γ 0 ϵ 02 0 qþ1
λ þ λ ϵ þ 0:5γ ϵ
0

filter is designed in such a way that the closed loop response is

similar to the response of a FOPTD system with unity gain [21]. this is a PID controller with a first order filter as given below


1 ð1 þ α0 qÞ
Gc ¼ K c 1 þ 0 þ τ0D q  0  ð12Þ
τI q 1þβ q
2. Proposed method
where
2.1. First Order plus Time Delay system with Integrator (FOPTDI) 1 þ γ0
K 0p K c ¼
 ð13Þ
λ þ λ0 ϵ þ 0:5γ 0 ϵ
02

The process transfer function of FOPTDI system is given by

 
K P e  Ls τ0I ¼ γ 0 þ1 ð14Þ
GP ðsÞ ¼ ð1Þ
sðτs þ 1Þ
γ0
τ0D ¼ ð15Þ
Using τ ¼ ε and q ¼ τs; the above transfer function model is
L 1 þ γ0
written as
02
0:5ϵλ
K 0 e  εq β0 ¼ ð16Þ
GP ¼ P
qðq þ 1Þ
ð2Þ λ þ λ0 ϵ þ 0:5γ 0 ϵ
02

where K 0P ¼ K P τ : α0 ¼ 0:5ϵ ð17Þ

The invertible part of the above process transfer function (Eq.
¼ λτ; γ 0 ¼ γττ0I ¼ ττI ; τ0D ¼ ττD ; α0 ¼ ατ, ¼ βτ:
0 0
and λ β
(2)) is given by

K 0P The closed loop servo response is given by

GP  ¼ ð3Þ  
qðq þ 1Þ y γ s þ1  Ls
¼ f ðsÞGP þ ¼   e ð18Þ
The non-invertible part of the process is given by yr λs þ 1 2
GP þ ¼ e  ε q ð4Þ ðλs þ 1Þ
A set point filter F ðsÞ ¼ γ s þ 1 is introduced to improve the servo
ð Þ
performance such that the closed loop response resembles the
The IMC controller is given by
response of the first order system with time delay.
0
1 qðq þ 1Þ ðγ 0 q þ1Þ The tuning parameter, IMC filter time constant (λ ) is tuned in
Q ¼ GP  f ðqÞ ¼ ð5Þ
K 0P ðλ0 q þ 1Þ2 such a way that the maximum magnitude of sensitivity function
(Ms) is equal to 2. The tuning rules for FOPTDI system that yields
 D.B.S. Kumar, R. Padma Sree / ISA Transactions 63 (2016) 242–255 245

Ms ¼2 for 0:1 r ε r 0:5 are given by In Eq. (30), using Ls ¼ p and transfer function model is written
as
K c K p τ ¼ 81:33ϵ  70:66ϵ þ 6:22ϵ þ 8:011
3 2
ð19Þ
K 0P e  p
τI GP ðsÞ ¼ ð31Þ
¼  8:333ϵ3 þ 10:21ϵ2  2:695ϵ þ2:056 ð20Þ p2
τ
where
τD
¼ 2:65ϵ3 þ 2:955ϵ2  0:748ϵ þ 0:515 ð21Þ K 0P ¼ K P L2 ð32Þ
τ
α The invertible part of the above process transfer function (Eq.
¼ 0:5ϵ ð22Þ (31)) is given by
τ
K 0P
β GP  ¼ ð33Þ
¼ 0:483ϵ3 0:522ϵ2 þ 0:214ϵ þ 0:015 ð23Þ p2
τ
The non-invertible part of the process is given by
λ
¼  4:166ε3 þ 5:107ε2  1:847ε þ 0:528 ð24Þ GP þ ¼ e  p ð34Þ
τ
When ε 4 0:5, if PID is tuned to stabilize the system for Ms ¼2, The IMC controller is given by
0
the performance is sluggish as the tuning parameter λ is large.  
p2 γ 0 p2 þ γ 0 p þ 1
Therefore for larger time delay to time constant ratio, PID is to be Q ¼ GP   1 f ðqÞ ¼ 0 1 0 2 3 ð35Þ
tuned to larger Ms (i.e. Ms 4 2). The PID controller is tuned for 0:6
KP λ pþ1
r ε r 1 in such way that good trade-off is made between perfor- 0 γ γ
where λ ¼ λL; γ 01 ¼ L21 and γ 02 ¼ L2
mance and robustness and tuning rules are given for 0:6 r ε r 1 as The conventional feedback PID controller is obtained by using
a function of ε. The tuning rules for stable FOPTD system with an Eq. (6) and is given by
integrator for 0:6 r ε r 1 are given below.  
p2 γ 01 p2 þ γ 02 p þ 1
K c K p τ ¼ 40:66ε3  93:98ε2 þ 68:58ε 12:46 ð25Þ GC ¼ h  0 3   i ð36Þ
K 0P pλ þ 1  γ 01 p2 þ γ 02 p þ 1 e  p
τI
¼ 0:34ϵ þ 1:956 ð26Þ Using first order Pade's approximation for time delay, the PID
τ
controller is written as
τD  
¼ 0:068ε þ 0:496 ð27Þ p2 γ 01 p2 þ γ 02 p þ 1 ð1 þ 0:5pÞ
τ GC ¼ h  0  3   i ð37Þ
K 0P pλ þ 1 ð1 þ 0:5pÞ  γ 01 p2 þ γ 02 p þ 1 ð1  0:5pÞ
α
¼ 0:5ε ð28Þ
τ

p2 ðγ 01 p2 þ γ 02 p þ 1Þð1 þ 0:5pÞ
GC ¼ h


 i ð38Þ
03 02 þ 1:5 02 þ 0:5
K 0P 0:5λ p4 þ 3γ λ γ 1' p3 þ 3λ02 þ 1:5λ0 þ 0:5γ 02  γ 01 p2 þ ð3λ0 þ 0:5  γ 02 þ 0:5Þp

β To derive PID controller from Eq. (38), the coefficients of p and

¼ 0:065ε þ 0:078 ð29Þ p2 in denominator of Eq. (38) are equated to zero. Thus the value of
τ
γ 01 and γ 02 is obtained as
γ 02 ¼ 3λ0 þ 1 ð39Þ

Remark 1: For First Order plus Time Delay system with an γ 01 ¼ 3λ02 þ1:5λ0 þ 0:5γ 02 ð40Þ
Integrator and a negative zero, PID controller with a first order
filter is designed

 for the system without a zero and an additional Rearranging Eq. (38), the following equation for PID controller
PID filter 1 þ1 Ps is used. This is implemented on the FOPTDI is obtained.
system with a zero.
0 
Remark 2: For a system with positive zero K P ð1sðτs PsÞe
 Ls γ 02 p3 γγ 10 p þ 1 þ γ10 p ð1 þ 0:5pÞ
þ 1Þ , PID con- GC ¼
 
 ð41Þ
2 2
K P e  θs
troller is designed for a system sðτs þ 1Þ where θ is effective time 03 02
K 0P λ þ 1:5λ þ 0:5γ 01 p3 0:5λ
03
pþ1
03 02
delay which is the sum of process time delay and numerator λ þ 1:5λ þ 0:5γ 1
0

time constant ‘P’. This is obtained by approximating

Rearranging the above equation,
ð1  PsÞ ¼ e  Ps . The designed PID controller is implemented on
0 
the actual system with a positive zero. γ 02 γ 1 ð0:5p þ 1Þ
GC ¼
03  10 p þ1 þ 0 p 

K 0P λ þ 1:5λ þ 0:5γ 01 γ 2 γ2
02 03
0:5λ
pþ1
λ þ 1:5λ0 þ 0:5γ 01
03 2

2.2. Double Integrating plus Time Delay (DIPTD) system ð42Þ

The process transfer function DIPTD system is given by This is a PID controller with a first order filter (i.e. Eq. (12)),
where
γ 02
K P e  Ls Kc ¼
 ð43Þ
03 02
GP ð s Þ ¼ ð30Þ K 0P λ þ 1:5λ þ 0:5γ 01
s2
246 D.B.S. Kumar, R. Padma Sree / ISA Transactions 63 (2016) 242–255

τ0I ¼ γ 02 ð44Þ From IMC controller, the conventional feedback PID controller
is obtained by Eq. (6) and is given by
γ 01  
τ0D ¼ ð45Þ p γ0p þ 1
γ 02 GC ¼ h  0  2   i ð60Þ
K 0P λ p þ 1  γ 0 p þ 1 e  p
03
0:5λ
β0 ¼ ð46Þ Using first order Pade's approximation for time delay, the PID
λ þ 1:5λ02 þ 0:5γ 01
03
controller is written as
 
p γ 0 p þ 1 ð1 þ 0:5pÞ
α0 ¼ 0:5 ð47Þ GC ¼ h  0   i ð61Þ
2
K 0P λ p þ 1 ð1 þ 0:5pÞ  ð1  0:5pÞ γ 0 p þ 1
where τ 0I ¼ τLI ; τ0D ¼ τLD ; α0 ¼ αL , and β ¼ βL
0

The tuning parameter, IMC filter time constant (λ) is tuned in Rearranging the above equation,
such a way that the maximum magnitude of sensitivity function
ðγ 0 p þ1Þ½1 þ 0:5p
(Ms) is equal to 2. The tuning rules for DIPTD system that yields GC ¼
2    ð62Þ
02
0
Ms ¼2 (for a value of λ ¼ 2:82Þ are given by K 0P ½p2 ð0:5 Þ þp λ λ0 þ λ0 þ 0:5γ 0 þ 2λ0 þ 1  γ 0 

K C K P L2 ¼ 0:1864 ð48Þ To represent Gc as PID controller, the constant of denominator

0
(2λ þ 1  γ 0 ) of Eq. (62) is equated to zero. Thus the value of γ 0 is
τI obtained as
¼ 9:46 ð49Þ
L
γ 0 ¼ 2λ0 þ 1: ð63Þ
τD
¼ 3:469 ð50Þ Eq. (63) is rearranged to give PID controller with a first order
L
filter.
α 
¼ 0:5 ð51Þ  0  0:5γ 0
L γ þ 0:5 ð0:5 þγ Þ0 p þ1 þ 1
ð0:5 þ γ Þp
0

GC ¼ h
i ð64Þ
β 02 0
K P λ þ λ þ 0:5γ
0 0 0:5λ
02
p þ 1
¼ 0:2209 ð52Þ 02 0
λ þ λ þ 0:5γ
0
L
The closed loop servo response is given by Eq. (64) is PID controller with a first order lag filter as given
 2  below.
y γ s þ γ 2 s þ 1  Ls

¼ f ðsÞGP þ ¼ 1  3 e ð53Þ 1 1
yr λs þ 1 Gc ¼ K c 1 þ 0 þ τ0D p  0  ð65Þ
τI p 1þβ p
2 
λ s2 þ 2λs þ 1
A set point filter F ðsÞ ¼ γ s2 þ γ s þ 1 is introduced to improve the where
ð1 2 Þ  
servo performance such that the closed loop response resembles 0:5 þ γ 0
the response of the first order system with time delay. Kc ¼
 ð66Þ
02 0
K P 0 λ þ λ þ 0:5γ 0

2.3. Pure Integrating plus Time Delay (PIPTD) system  

τI 0 ¼ 0:5 þ γ 0 ð67Þ
The process transfer function for PIPTD system is given by
0:5γ 0
τD 0 ¼   ð68Þ
KP 0:5 þ γ 0
GP ðsÞ ¼ e  Ls ð54Þ
s
The lag filter time constant is
In the Eq. (54), using Ls ¼ p and the transfer function is written
02
as 0:5λ
β0 ¼
02 0
 ð69Þ
K0 e p λ þ λ þ 0:5γ 0
GP ðsÞ ¼ P ð55Þ
p
where τ0I ¼ τLI ; τ0D ¼ τLD , β ¼ βL and λ ¼ λL
0 0

where The tuning parameter, IMC filter time constant (λ) is tuned in
such a way that the maximum magnitude of sensitivity function
K 0P ¼ K P L ð56Þ (Ms) is equal to 2. The tuning rules for PIPTD system that yields
0
The invertible part of the process transfer function (Eq. (55)) is Ms ¼2 (for a value of λ ¼ 1:62Þ are given by
given by K C K P L ¼ 0:7448 ð70Þ
K 0P τI
GP  ¼ ð57Þ ¼ 4:74 ð71Þ
p L
The non-invertible part of the process is given by τD
¼ 0:4473 ð72Þ
p L
GP þ ¼ e ð58Þ

The IMC controller for the above process is given by β

¼ 0:2062 ð73Þ
  L
p γ0p þ 1

Q ¼ GP1 f ðpÞ ¼ 0  0 ð59Þ The set point filter is selected as λγ ss þ
K P λ p þ 12
1
þ 1 , so that the closed loop
response is resembles the response of First Order plus Time Delay
where γ 0 ¼ γL and λ ¼ λL
0
system.
 D.B.S. Kumar, R. Padma Sree / ISA Transactions 63 (2016) 242–255 247



2.4. Unstable First Order plus Time Delay systems with Integrator The part of denominator,  0:5ϵλ
03
q2 
02 0
(UFOPTDI) system 3λ þ 1:5λ ϵ þ 0:5ϵγ 02  γ 01


λ0 þ 1:5λ0 ϵ þ 0:5ϵγ 01
3 2  0 
02 0 q  1 of Eq. (84) is equated to ðq  1Þ 1 þ β q
3λ þ 1:5λ ϵ þ 0:5ϵγ 02  γ 01
The process transfer function of Unstable First Order plus Time
Delay systems with Integrator is given by and comparing the corresponding coefficients of q2 and q, the
following equations are obtained for β.
K P e  Ls
GP ð s Þ ¼ ð74Þ  0:5ϵλ
03
sðτs  1Þ β0 ¼ ð85Þ
02 0
Using τL ¼ ε and q ¼ τs in Eq. (74), the transfer function model is 3λ þ 1:5λ ϵ þ 0:5ϵγ 02  γ 01
written as
3 
0 02
 λ þ 1:5λ ϵ þ 0:5ϵγ 01
K 0P e  εq 1β ¼
0
ð86Þ
GP ¼ ð75Þ 02 0
qðq  1Þ 3λ þ 1:5λ ϵ þ0:5ϵγ 02  γ 01
0
where From the Eq. (87), β is given by
K 0P ¼ K P τ ð76Þ 03
ðλ þ 1:5λ þ 0:5εγ 01 Þ
02
β0 ¼ 1 þ 02 0
ð87Þ
The invertible part of the process transfer function is given by 3λ þ1:5λ ε þ 0:5εγ 02  γ 01
K 0P By using Eqs. (85) and (87), γ 01 is obtained as
GP  ¼ ð77Þ
qðq  1Þ
02 0 03 02 03
3λ þ 1:5λ ε þ 0:5εγ 02 þ λ þ 1:5λ ε þ 0:5ελ Þ
The non-invertible part of the process is given by γ 01 ¼ ð88Þ
1  0:5ε
GP þ ¼ e  εq ð78Þ
Then Eq. (85) is written as
The IMC controller for the above process transfer function is  
 γ 0 q2 þ γ 02 q þ 1 ðq  1Þð1 þ 0:5ϵqÞ
given by Gc ¼
2 1   0  ð89Þ
0 0
  K 0P q 3λ þ1:5λ ϵ þ 0:5ϵγ 02  γ 01 ðq 1Þ β q þ 1
qðq  1Þ γ 01 q2 þ γ 02 q þ 1
Q ¼ GP   1 f ðqÞ ¼  0 3 ð79Þ
K 0P λ qþ1 Upon pole zero cancellation, Gc is given by

0 
γ
The PID controller for the process is given by Eq. (6). Sub-  γ 02 γ 10 q þ γ10 q þ 1 ð1 þ 0:5ϵqÞ
stituting Eqs. (79) and (75) in Eq. (6), the following equation for Gc Gc ¼
0 2 2
0
2

0  ð90Þ
K 0P 3λ þ 1:5λ ϵ þ 0:5ϵγ 02  γ 01 β q þ 1
is obtained
 
qðq 1Þ γ 01 q2 þ γ 02 q þ 1 This is a PID controller with a first order filter as given in Eq.
GC ¼ h    i ð80Þ (12).where
0 3
K 0P λ q þ 1  γ 01 q2 þ γ 02 q þ1 e  ϵS
 γ 02
K 0c ¼
 ð91Þ
Using first order Pade's approximation for time delay, the Eq. K 0P
02 0
3λ þ 1:5λ ϵ þ 0:5ϵγ 02  γ 01
(81) is written as
 
qðq  1Þ γ 01 q2 þ γ 02 q þ 1 ð1 þ 0:5ϵqÞ τ0I ¼ γ 02 ð92Þ
GC ¼ h 0 3   i ð81Þ
K 0P λ q þ 1 ð1 þ 0:5ϵqÞ  γ 01 q2 þ γ 02 q þ 1 ð1  0:5ϵqÞ
γ 01
τ0D ¼ ð93Þ
Rearranging Eq. (82), the following equation is obtained γ 02

 
qðq  1Þ γ 01 q2 þ γ 02 q þ 1 ð1 þ 0:5ϵqÞ
GC ¼ h


  0  i ð82Þ
0 3 03 02 02 0
K 0P 0:5ϵλ q4 þ λ þ1:5λ ϵ þ 0:5ϵγ 01 q3 þ 3λ þ 1:5λ ϵ þ 0:5ϵγ 02  γ 01 q2 þ 3λ þ ϵ  γ 02 q

 0 
To derive PID settings, the coefficient of q term 3λ þ m  γ 02 in 03
 0:5ϵλ
denominator of Eq. (82) is equated to zero. Thus the value of γ 02 is β0 ¼ ð94Þ
02 0
obtained as 3λ þ 1:5λ ϵ þ 0:5ϵγ 02  γ 01

γ 02 ¼ 3λ0 þ ϵ ð83Þ
α0 ¼ 0:5ε ð95Þ

Rearranging Eq. (77), the following equation is obtained for Gc.

and τ0I ¼ ττI ; τ0D ¼ ττD ; α0 ¼ ατ, β ¼ βτ
0

 
 γ 01 q2 þ γ 02 q þ 1 qðq  1Þð1 þ 0:5ϵqÞ
GC ¼
2  

03  ð84Þ
λ þ 1:5λ0 ϵ þ 0:5ϵγ 01
2
0 0 03
 0:5ϵλ
K 0P q 3λ þ 1:5λ ϵ þ 0:5ϵγ 02  γ 01 02 0 q2
0 2 0 q  1
3λ þ 1:5λ ϵ þ 0:5ϵγ 2  γ 1
0 0 3λ þ 1:5λ ϵ þ 0:5mϵ  γ 1
0
248 D.B.S. Kumar, R. Padma Sree / ISA Transactions 63 (2016) 242–255

Table 1

h i 
PID parameters for different methods K c 1 þ τ1I s þ τD s αs þ1
βs þ 1 .

Transfer function Method KC τI τD α β Set point filter

0:2e  s Proposed 7.415 7.8 1.9487 0.5 0.1863 0:7s þ 1

Case study 1 sð4s þ 1Þ 3:8s þ 1
Jin and Liu [21] 3.686 10.39 2.473 – – 9:7915s2 þ 6:2664s þ 1
25:6994s2 þ 10:392s þ 1
Case study 2 Jacketed CSTR [10] Proposed 2.7 16.72 3.195 0.5 0.2087 0:17s þ 1
4:4s þ 1
Jin and Liu [21] 1.108 15.27 4.5147 – – 22:63s2 þ 9:51s þ 1
68:95s2 þ 15:27s þ 1
e  0:5s Proposed 4.095 7.64 1.4087 0.5385 0.2473 2:35s þ 1
Case study 3 sðs þ 1Þðs þ 2Þðs þ 3Þ 5:777s þ 1
Lee et al. [26] 3.6823 7.363 1.3916 – – 3:44s þ 1
5:5s þ 1
Case study 4 0:2e  7:4s Proposed 0.5169 34.1 3.2985 – 1.4836 17:25s þ 1
30:4s þ 1
s
Lee et al. [26] 0.4226 26.2 1.145 – – 15:63s þ 1
25s þ 1
Jin and Liu [21] 0.384 35.788 – – – 14:2s þ 1
35:788s þ 1
es Proposed 0.1883 9.4 3.4489 0.5 0.2199 7:84s2 þ 5:6s þ 1
Case study 5 s2 32:42s2 þ 9:4s þ 1
Lee et al. [26] 0.1275 7.5649 3.7878 – – 7:16s2 þ 5:35s þ 1
28:62s2 þ 10:7s þ 1
a
Case study 6 Jacketed CSTR [46] Proposed 30,6517.5 9.4 3.448 0.5 0.22 7:84s2 þ 5:6s þ 1
32:42s2 þ 9:4s þ 1
Lee et al. [26] 20,7620 7.565 3.7878 – – 7:16s2 þ 5:35s þ 1
28:6s2 þ 16:7s þ 1
e  0:2s Proposed 1.9421 2.9 1.5459 0.1 0.0488 0:81s2 þ 1:8s þ 1
Case study 7 sðs  1Þ 4:4832s2 þ 2:9s þ 1
Cho et al. [28] 0.8594 4.4 2.7 – – 1:93s2 þ 2:8s þ 1
11:88s2 þ 4:4s þ 1
Liu et al.b [51] 1.4738 2.1752 1.683 – – s2 þ s þ 1
0:36s2 þ 1:2s þ 1

a
Along with the PID controller with filter an additional filter ð766:0752s
1
þ 1Þ is considered.
b
Two loop control scheme. Set point tracking controller is PD controller with Kc ¼1 and τD ¼ 2. The controller parameters given in the above table are for disturbance
rejection.

1.4 λ
¼ 165ε3 þ 120ε2  19:45ε þ 1:24 ð101Þ
1.2
τ
ð λs þ 1 Þ
2

1 A set point filter F ðsÞ ¼ γ s2 þ γ s þ 1 is introduced to improve the

ð1 2 Þ
servo performance such that the closed loop response resembles
0.8
the response of the first order system with time delay.
y

0.6 Proposed
0.4 Jin and Liu
3. Simulation results
0.2

0 In this section, simulation results on various types of integrat-

0 10 20 30 40 50 60 70 80 90 100 ing transfer function models are reported. The performance of the
controller is measured in terms of Integral Absolute Error (IAE)
40 and smooth functioning of the controller is given in terms of total
30 variation (TV) is sum of the differences between the current
20 controlled output and previous controlled output, over a time
equal to settling time of the process. In the present work, PID
u

10
controllers are tuned in such a way that they will give the same Ms
0 value as given by the literature reported methods, so that perfor-
-10 mance comparison can be made for the same robustness level. For
0 10 20 30 40 50 60 70 80 90 100 stable systems, if Ms value is between 1.4 and 2, the controller is
t said to be robust. But integrating processes can be considered as a
Fig. 1. The response of transfer function model 0:2e  s
(case study 1). sub class of unstable systems. Hence the Ms value can exceed 2 for
sð4s þ 1Þ
robust control. The present methods are also implemented for
The tuning parameter, IMC filter time constant (λ) is tuned in higher order systems. The higher order systems are reduced to
such a way that the maximum magnitude of sensitivity function integrating systems with time delay using model reduction tech-
(Ms) is equal to 2. The tuning rules for unstable FOPTDI system that niques. To the reduced model, PID controller is designed based on
yields Ms ¼2 for 0:1 r ε r0:4 are given by the present method and implemented on the original system. Jin
and Liu [21] implemented PID controller in parallel mode, there-
K c K p τ ¼ 1381ε3 þ 1317ε2  413:5ε þ 43:25 ð96Þ fore in the present work also it is implemented in parallel mode.
τI
¼  495ε3 þ 360ε2  57:35ε þ 3:72 ð97Þ 3.1. Stable First Order plus Time Delay system with Integrator
τ
(FOPTDI)
τD
¼  396:3ε3 þ 314:4ε2  57:13ε þ 3:433 ð98Þ
τ A stable First Order plus Time Delay with Integrator (FOPTDI)
α system GP ¼ sð4s0:2þ 1Þe  s [21] is considered. The proposed method is
¼ 0:5ε ð99Þ compared with Jin and Liu [21] method. The IMC filter time con-
τ
stant λ is tuned to the value of 1.4 such that maximum magnitude
β of sensitivity function (Ms) is 1.98 which is reported by Jin and Liu
¼  0:287ε2 þ 0:366ε  0:012 ð100Þ
τ [21]. The PID controller parameters of the proposed method and of
 D.B.S. Kumar, R. Padma Sree / ISA Transactions 63 (2016) 242–255 249

Table 2
Performance comparison of different methods.

Case study Method Ms Phase margin Z Gain margin Z Servo problem Regulatory problem

IAE ITAE TV IAE ITAE TV

1 Proposed 1.98 29.15° 2.01 3.51 8.667 23.13 1.05 61.15 1.917
Jin and Liu [21] 1.99 28.56° 2.03 4.136 12.9 37.528 2.81 170.3 1.69
2 Proposed 1.95 28.68° 1.98 31.59 346.1 45.05 27.13 455.3 8.00
Jin and Liu [21] 1.96 28.32° 1.98 47.96 536.9 9.42 60.37 918.32 7.478
3 Proposed 1.97 29.37° 2.02 3.984 12.89 6.75 1.87 15.0 1.634
Lee et al. [26] 1.97 29.41° 2.03 4.385 18.68 4.46 2.03 15.35 1.71
4 Proposed 2 28.27° 1.95 18.45 261.49 0.8417 65.96 15,346 1.796
Lee et al. [26] 2 29.39° 2.0 21.45 404.1 0.6317 68.39 15,851 2.05
Jin and Liu [21] 2 28.97° 2.0 21.58 306.73 0.322 93.26 21,973 1.915
5 Proposed 2.01 28.78° 1.98 4.8 18.438 0.4861 49.92 3479 2.121
Lee et al. [26] 2.01 28.85° 1.99 6.1 36.76 0.2458 68.962 4796 2.785
6 Proposed 2.0 28.8° 1.99 12.61 86.66 3165.8 0.0332 0.2882 15.61
Lee et al. [26] 2.0 28.8° 1.99 16.6 124.76 3529 0.0654 0.7844 37.8
7 Proposed 1.94 30.1° 2.06 1.423 1.74 5.63 1.4932 41.75 2.8
Cho et al. [28] 1.94 29.8° 2.06 2.097 3.92 2.138 5.1199 150.58 3.2882
Liu et al. [51] – – – 1.4056 1.3645 4.5409 1.7446 48.118 3.7599

Table 3 1.4
Stability regions for model parameters (Kharitonovs Theorem).
1.2
Transfer function Method K P (%) L (%) τ (%)
1
Case study 1 0:2e  s
sð4s þ 1Þ
Proposed 7 53 7 25 7 33
Jin and Liu [21] 7 55 7 62 7 55
0.8
Response

Case study 2 Jacketed CSTR [10] Proposed 7 63 7 26 7 32

Jin and Liu [21] 7 53 7 70 7 55 -Proposed
0.6
Case study 3 e  0:5s Proposed 7 56 7 24 7 34 Jin and Liu
sðs þ 1Þðs þ 2Þðs þ 3Þ
Lee et al. [26] 7 57 7 58 7 55
0:2e  7:4s Proposed 7 21 7 44 – 0.4
Case study 4 s
Lee et al. [26] 7 58 7 52 –
Case study 5 es Proposed 7 46 7 40 – 0.2
s2
Lee et al. [26] 7 30 7 66 –
Case study 6 Jacketed CSTR [46] Proposed 7 46 7 40 – 0
Lee et al. [26] 7 32 7 65 – 0 10 20 30 40 50 60 70 80 90 100
Case study 7 e  0:2s Proposed 7 42 7 60 7 53% time
sðs  1Þ
Cho et al. [28] 7 35 7 68 7 55%
Fig. 3. Response of the system in case study 1 for  20% uncertainty in Kp, L and τ.

1.4 respectively. The proposed method gives less overshoot and peak
error for servo and regulatory problems respectively. The con-
1.2 troller action for the present method is smooth (measured in
terms of TV) for servo problem compared with the method of Jin
1
and Liu [21]. The proposed controller gives fast response compared
0.8 to the controller by Jin and Liu [21].
Response

The performance of the system for parameter uncertainty is

0.6 checked by Kharitonov's theorem (refer to Appendix A) and the
Proposed
range of process parameters for which the controller (designed for
0.4 Jin and Liu
nominal model parameters) can stabilize the system are given in
0.2 Table 3. The performance of the controller designed by proposed
method and the controller designed by the method of Jin and Liu
0 [21] for þ20% uncertainty simultaneously in Kp, τ and L is shown
0 10 20 30 40 50 60 70 80 90 100
t
in Fig. 2 and for  20% uncertainty simultaneously in Kp, τ and L is
shown in Fig. 3. The performance of both the controller in terms of
Fig. 2. Response of the system in case study 1 for þ20% uncertainty in Kp, L and τ. IAE, ITAE and TV for 7 20% uncertainty in Kp, τ and L individually
is also reported in Table 4. The proposed controller performs
Jin and Liu [21] are listed in Table 1. The response of the system for better than the controller designed by the method of Jin and Liu
unit step change in set point at t ¼0 s and for the unit step change [21] even in the presence of uncertainty in process parameters.
in load at t ¼50 s with the proposed controller is shown in Fig. 1.
The servo and regulatory response of the system with the con-
3.2. Non-linear Jacketed CSTR
troller designed by Jin and Liu [21] is also shown in Fig. 1.
The performance of both the controllers is reported in terms of
Consider a non-linear Jacketed CSTR carrying first order irre-
IAE, ITAE and the smoothness of the controller which is given in
versible exothermic reaction proposed by Hovd and Skogestad [10]
terms of total variation (TV) for both servo and regulatory pro-
and the model equations are
blems (refer to Table 2).When compared with the Jin and Liu [21]
method, the performance of the proposed method is superior with dC A F
¼ ðcAF  cA Þ  r ð102Þ
less IAE and ITAE values for servo and regulatory problems dt V
250 D.B.S. Kumar, R. Padma Sree / ISA Transactions 63 (2016) 242–255

Table 4
Performance comparison under parameter uncertainty for case study 1.

Problem Error criteria Method K P ; τ and L KP τ L

þ 20%  20% þ20%  20% þ 20%  20% þ20%  20%

Servo IAE Proposed 3.26 3.82 3.34 3.94 3.67 3.5 3.42 3.57
JL 4.13 4.52 4.13 4.41 3.93 4.21 4.0 4.13
ITAE Proposed 6.55 12.28 8.11 11.78 9.26 9.86 8.1 9.77
JL 15.24 17.92 15.25 14.05 12.9 15.23 12.9 13.03
TV Proposed 28.62 21.3 27.2 22.36 23.1 27.76 27.4 21.71
JL 40.07 37.62 40 37.62 38.2 39.83 39.4 37.8
Regulatory IAE Proposed 1.05 1.05 1.05 1.05 1.05 1.05 1.05 1.05
JL 2.82 2.86 2.82 2.82 2.82 2.83 2.81 2.82
ITAE Proposed 61.14 61.5 61.1 61.4 61.2 61.14 61.1 61.4
JL 170.2 173.5 170.3 170.8 170.2 170.9 170.3 170.3
TV Proposed 3.8 2.38 2.4 2.71 1.87 3.43 2.55 2.48
JL 3.14 2.4 2.8 2.64 2.8 2.67 2.9 2.54

Table 5
Parameter values for jacketed CSTR [10]. 597

Parameter Value 596

595
Volume, V 50 m3
Feed flow rate, F 5 m3/min 594
Feed Temperate,TF 500 K
T
Feed Concentration, CAF 15.61 K mol/m3 593 Proposed
Product of heat Transfer coefficient and heat transfer area 2600 kJ/min K Jin and Liu
(hA) 592
Specific Heat, Cp 1.8 kJ/kg K 591
Heat of reaction,  ΔH  20,000 kJ/K mol
Universal gas constant, R 8.314 kJ/K mol 590
Frequency factor, Ko 680,000 min  1 0 10 20 30 40 50 60 70 80
Density, ρ 800 kg/m3 50
Temperature, Tc 401.6 K
40
30
CAF

dcB F 20
¼  cB þr ð103Þ
dt V
10
dT F ΔHr hA 0
¼ ðT F  T Þ   ðT  T C Þ ð104Þ
dt V ρC P ρV C P 0 10 20 30 40 50 60 70 80
Time in min
r ¼ KcA e  RT
E
ð105Þ
594
The process parameters are given in Table 5. The above non- 593.5
linear equations are linearized around the operating point Proposed
593
C A ¼10 K mol/m3, C BS ¼5.62 K mol/m3 and T S (steady state reactor Jin and Liu
592.5
temperature) ¼590 K and the transfer function model is given by
592
T

TðsÞ 0:07803s þ 0:007803 591.5

¼ ð106Þ
C AF ðsÞ s3 þ0:1766s2 þ 0:00735s  0:00031 591
On simplifying Eq. (106), the following equation is obtained 590.5

TðsÞ 0:007803ðs þ 0:1Þ 590

¼ ð107Þ 0 10 20 30 40 50 60 70 80
C AF ðsÞ ðs þ 0:1Þðs þ 0:0805Þðs  0:0039Þ
20
Upon pole zero cancellation, Eq. (107) reduces to,
18
TðsÞ 0:07803
¼ ð108Þ
CAF

16
C AF ðsÞ 0:0805  0:0039ð12:4224s þ 1Þð256:41033s  1Þ
14
In the above equation, the unstable time constant (256.4103) is
very large and can be considered into process gain. On simplifying 12
0 10 20 30 40 50 60 70 80
the equation, stable FOPTDI transfer function is obtained.
Time in min
TðsÞ 0:9693
¼ ð109Þ Fig. 4. (a) The servo response of Jacketed CSTR (non-linear model simulation) (case
C AF ðsÞ sð12:4224s þ 1Þ study 2). (b) The regulatory response of response of Jacketed CSTR [non-linear
A measurement delay of one minute is considered and the model simulation] (case study2).

transfer function is written as

A PID controller is designed for the above transfer function
TðsÞ 0:9693e  1s model of jacketed CSTR using Eqs. (102)–(105) by tuning λ in such
¼ ð110Þ
C AF ðsÞ sð12:4224s þ 1Þ
a way that it gives same Ms value as Jin and Liu [21]. The PID
 D.B.S. Kumar, R. Padma Sree / ISA Transactions 63 (2016) 242–255 251

4
1.4
3.5 Proposed
1.2 Jin and Liu
3
1 Lee et al.
2.5
0.8
2
y

y
0.6 Proposed
Lee et al. 1.5
0.4
1
0.2
0.5
0
0 5 10 15 20 25 30 35 40 45 50
0
6 0 50 100 150 200 250 300 350 400
1.5
4

2 1
u

0 0.5

u
-2
0 5 10 15 20 25 30 35 40 45 50 0
t
-0.5
0 50 100 150 200 250 300 350 400
0.3  7:4s
Fig. 6. The response of transfer function model GP ¼ 0:2
s e (case study 4).
0.25
0.2 Proposed
10
Lee et al.
0.15
y

0.1 8
Proposed
0.05
6 Lee et al.
0
y

-0.05
0 5 10 15 20 25 30 4

1 2

0.5 0
0 50 100 150
u

0 1.5
1
-0.5
0 5 10 15 20 25 30 0.5
t
u

0
e  0:5s
Fig. 5. (a) The servo response of transfer function model sðs þ 1Þðs þ 2Þðs þ 3Þ (case study
e  0:5s -0.5
3). (b) Regulatory response of transfer function model sðs þ 1Þðs þ 2Þðs þ 3Þ (case study 3).
-1
0 50 100 150
es
Fig. 7. The response of transfer function model (case study 5).
controller and filter parameters of the proposed method and of Jin s2

and Liu [21] are listed in Table 1.The servo response of the pro-
posed controller with the non-linear process model (Eqs. (102)–
(105)) for a step change in operating temperature from 590 to
595 K are shown in Fig. 4(a).The servo response of the system with
the controller designed by Jin and Liu [21] is also shown in Fig. 4
(a). The performance of both the controllers is reported in terms of
IAE, ITAE and the smoothness of the controller is given in terms of
total variation (TV) for servo problem (refer to Table 2). When
compared with the Jin and Liu [21] method, the performance of
the proposed method is superior with less IAE, ITAE values (refer
to Table 2) and less overshoot. The proposed controller gives fast
response compared to the controller by Jin and Liu [21].
The regulatory performance of the proposed controller and
controller designed by Jin and Liu [21] for a change of feed con-
centration from 15.61 to 20 K mol/m3 is reported in Table 2 in
terms of IAE, ITAE and the response is shown in Fig. 4(b). The
response of the proposed controller gives less peak error. The
proposed controller gives fast response compared to the controller
by Jin and Liu [21].
The stability regions for the model parameters (considering the
transfer function model in Eq. (110)) for which the controller can
stabilize the system are obtained by using Kharitonov's theorem
and reported in Table 3.
3.3. Higher order system
e  0:5s
The transfer function model sðs þ 1Þðs þ 2Þðs þ 3Þ [41] is considered.
 1:077s
The model is reduced to FOPTDI model 0:167e [32]. The PID
sð1:863s þ 1Þ
parameters for the proposed method and Lee et al. [26] method
are given in Table 1. The response of the proposed controller for a
unit step change in set point is shown in Fig. 5(a). The servo
response of the system with the controller designed by Lee et al.
[26] is also shown in the Fig. 5(a). The performance of both the
controllers is given in terms of IAE, ITAE and the smoothness of the
controller which is given in terms of total variation (TV) for servo
problem (refer to Table 2). The IMC filter time constant λ is tuned
to the value of 2.35 to obtain the same robustness level of
252 D.B.S. Kumar, R. Padma Sree / ISA Transactions 63 (2016) 242–255

maximum magnitude of sensitivity function (Ms) 1.97 of Lee et al.

[26]. When compared with Lee et al. [26] method, the perfor- 347.5

mance of the proposed method is superior with less IAE and ITAE 347
values and less overshoot. The proposed controller gives fast
346.5
response compared to the controller by Lee et al. [26]. The reg-
ulatory response of the proposed controller for a 0.1 change in load 346

T,K
is shown in Fig. 5(b). The regulatory response of the system with 345.5
the controller designed by Lee et al. [26] is also shown in the Fig. 5
345 Proposed
(b). When compared with the Lee et al. [26] method, the perfor-
Lee et al.
mance of the proposed method is superior with less IAE and ITAE 344.5
values for both servo and regulatory problems. For regulatory
344
problem the proposed method gives less peak error. The controller 0 10 20 30 40 50 60
action is smooth for regulatory problem for the present method 2000
compared with the method of Lee et al. [26]. The proposed con-
1500
troller gives fast response compared to the controller by Lee et al.
1000

TJ, K
[26].
The stability regions for the model parameters (considering the 500
FOPTDI transfer function model) for which the controller can 0
stabilize the system are obtained by using Kharitonov's theorem -500
and reported in Table 3. 0 10 20 30 40 50 60
Time, s

3.4. Pure Integrator plus Time Delay (PIPTD) system 344.008

 7:4s 344.006
The transfer function model 0:2 s e [21,26,42] is considered. Proposed
For this system PID controller with a lag filter is designed by using 344.004 Lee et al.
the present method ((Eqs. (70)–73)) for Ms ¼2. The PID controller 344.002
and filter parameters for various methods are reported in Table 1.
T,K

344
The response of the system for unit step change in set point at
t¼0 s for the unit step change in load at t¼200 s with the pro- 343.998

posed controller is shown in Fig. 6. The servo and regulatory 343.996

performances of the system with the controller designed by Lee 343.994
et al. [26] and Jin and Liu [21] is also shown in Fig. 6.
343.992
The performance of the three controllers is given in terms of 0 10 20 30 40 50 60
IAE, ITAE and the smoothness of the controller is given in terms of
325
total variation (TV) for both servo and regulatory problems (refer
to Table 2). The controller action is smooth for regulatory problem
for the present method compared with the method of Lee et al. 320
TJ, K

[26] and Jin and Liu [21]. When compared with the reported
methods, the performance of the proposed method is superior 315
with less IAE, ITAE values for both servo and regulatory problems.
The present method gives less overshoot and peak error for servo 310
0 10 20 30 40 50 60
and regulatory problems respectively. The proposed controller
Time, s
gives fast response compared to the controller by the Lee et al. [26]
and Jin and Liu [21]. Fig. 8. (a) The servo response of Jacketed CSTR (non-linear model simulation) (case
study 6). (b) The regulatory response of Jacketed CSTR (non-linear model equations
The stability regions for the model parameters (Kp, L) for which
of CSTR) (case study 6).
the controller can stabilize the system are obtained by using
Kharitonov's theorem and reported in Table 3. 3.5. Double Integrator plus Time Delay (DIPTD) system
s
Table 6 The transfer function model es2 [26] is considered. For this
Parameter values for jacketed CSTR [46].
system PID controller with a lead lag filter is designed by using the
Parameter Value present method ((Eqs. (48)–52)) for Ms ¼2. The PID controller and
filter parameters for various methods are reported in Table 1. The
Volume, V 1 m3 response of the system for unit step change in set point at t¼ 0 s
Feed flow rate, F 0.00065 m3/s
Feed Temperate, To 300 K
and for the unit step change in load at t ¼60 s with the proposed
Feed Concentration, CAo 7.5 K mol/m3 controller is shown in Fig. 7. The servo and regulatory perfor-
Product of overall heat Transfer coefficient and heat transfer 1.4 kJ/s K mance of the system with the controller designed by Lee et al. [26]
area (UA)
is also shown in Fig. 7. The performance of both the controllers is
Specific Heat, Cp 3.5 kJ/kg K
Heat of reaction,  ΔH 50,000 kJ/K mol given in terms of IAE, ITAE and the smoothness of the controller is
Universal gas constant, R 8.345 kJ/K mol K given in terms of total variation (TV) for both servo and regulatory
Frequency factor, K o 1.8  107 s  1 problems (refer to Table 2). When compared with the Lee et al.
Density, ρ 850 kg/m3
[26] method, the performance of the proposed method is superior
Activation energy E 69,000 kJ/K mol
with less IAE, ITAE values for servo and regulatory problems. The
 D.B.S. Kumar, R. Padma Sree / ISA Transactions 63 (2016) 242–255 253

proposed method gives less peak error for regulatory problem. The The above transfer function model, a model time delay of 1 s is
proposed controller gives fast response compared to the controller considered and the transfer function is written as
by the Lee et al. [26]. Comparison of TV value shows that the
TðsÞ 6:143  10  7 ð1 þ 766:0752sÞ  1s
proposed controller action is smoother than Lee et al. [26] ¼ e ð117Þ
T J ðsÞ s2
controller.
The stability regions for the model parameters (Kp, L) for which For this process PID controller is designed using Eqs. (48)–(52).
the controller can stabilize the system are obtained by using The PID parameters for the proposed method and literature
Kharitonov's theorem and reported in Table 3. reported method is given in Table 1. The response of the proposed
controller with the non-linear process model ((Eqs. (111) and 112))
for a step change in reactor temperature from 344 to 346 K is
3.6. Non-linear Jacketed CSTR
shown in Fig. 8(a).The servo response of the Jacketed CSTR with
the controller designed by Lee et al. [26] is also shown in Fig. 8(a).
Consider a jacketed CSTR carrying out first order irreversible
The performance of both the controllers is given in terms of IAE,
exothermic reaction [46] and the model equations are
ITAE and the smoothness of the controller is given in terms of Total


dC A F   E Variation (TV) for servo problem (refer to Table 2). When com-
¼ C Ao C A K O C A exp  ð111Þ
dt V RT pared with the Lee et al. [26] method, the performance of the
proposed method is superior with less IAE and ITAE values and


dT F ð  ΔHÞ E UA less overshoot. The proposed controller gives fast response com-
¼ ðT O  T Þ þ K O C A exp  þ ðT  TÞ ð112Þ
dt V ρC P RT ρVC P J pared to the controller by Lee et al. [26]
The regulatory response of the proposed controller for a load
Process parameters are given in Table 6. The above non-linear change in reactor jacket temperature from 317.4 to 310 K is shown
equations are linearized around unstable operating point in Fig. 8(b).The regulatory response of the system with the con-
C As ¼3.734 K mol/m3, TS ¼344 K and TJS ¼317.4 K and transfer troller designed by Lee et al. [26] is also shown in Fig. 8(b). The
function model is given by performance of both the controllers is given in terms of IAE, ITAE
and the smoothness of the controller is given in terms of total
TðsÞ 0:0004706s þ 6:143  10  7 variation (TV) for servo problem (refer to Table 2). The perfor-
¼ ð113Þ
T J ðsÞ s2  0:0004483s  4:055  10  7 mance of the proposed method is superior with less IAE, ITAE
values for both servo and regulatory problems. The proposed
Rearranging Eq. (113)
method gives less overshoot and less peak error when compared
TðsÞ 6:143  10  7 ð1 þ766:0752sÞ with the method reported by Lee et al. [26]. The TV values for
¼ ð114Þ servo and regulatory problems are less than the reported method.
T J ðsÞ ðs  0:8992  10  3 Þðs þ 0:4509  10  3 Þ
The proposed controller gives smoother control action. The pro-
posed controller gives fast response compared to the controller by
TðsÞ 6:143  10  7 ð1 þ 766:0752sÞ
¼ Lee et al. [26].
T J ðsÞ ð0:8992  10  0:4509  10  3 Þð1112:099s  1Þð2217:79s þ 1Þ
 3
The stability regions for the model parameters [considering the
ð115Þ transfer function model in Eq. (117)] for which the controller can
stabilize the system are obtained by using Kharitonov's theorem
In the above transfer function, stable and unstable time con-
and reported in Table 3.
stants (1112.099, 2217.79 respectively) are very large and can be
considered in to process gain. On simplifying the transfer function
3.7. Unstable First Order plus Time Delay system with Integrator
model, DIPTD system with a zero is obtained.
(UFOPTDI)
TðsÞ 6:143  10  7 ð1 þ766:0752sÞ
¼ ð116Þ e
The transfer function model sðs
 0:2s
T J ðsÞ s2  1Þ [28,51] is considered. The PID
controller parameters are reported in Table 1. The response of the
system for unit step change in set point at t ¼0 s for the unit step
2.5 change in load at t¼40 s with the proposed controller is shown in
Proposed
2 Cho et al. Fig. 9. The servo and regulatory performances of the system with
Liu et al. the controller designed by Cho et al. [28] and Liu et al. [51] is also
1.5
shown in Fig. 9. Liu et al. [51] decoupled set point tracking pro-
y

1 blem with regulatory problem using two degree of freedom con-

0.5
trol scheme. For servo response PD controller with a set point filter
is used and for regulatory problem PID controller is used. The PID
0 setting for both the controllers are listed in Table 1.
0 5 10 15 20 25 30 35 40 45 50
t The performance of the three controllers is given in terms of
3 IAE and the smoothness of the controller is given in terms of total
variation (TV) for both servo and regulatory problems (refer to
2
Table 2). The IMC filter time constant λ is tuned to the value of
1 0.9 to obtain the same robustness level of maximum magnitude of
u

sensitivity function (Ms) 1.94 of Cho et al. [28]. The performance of

0
the proposed method is superior to Cho et al. [28] with less IAE,
-1 ITAE values for both servo and regulatory problems. The control
-2 scheme proposed by Liu et al. [51] gives better performance for
0 5 10 15 20 25 30 35 40 45 50 servo problem compared to the proposed controller and the reg-
t
ulatory response of the proposed controller is better than Liu et al.
 0:2s
Fig. 9. The response of transfer function model Gp ¼ sðs
e
 1Þ (case study 7). [51]. The proposed method gives less overshoot and peak error for
254 D.B.S. Kumar, R. Padma Sree / ISA Transactions 63 (2016) 242–255

servo and regulatory problems respectively. The controller action a2 ¼ cL2 ττ I þ aτβcL2 þ 0:5LττI β þ cL2 τI τD αK C K P ðA6Þ
is smooth for regulatory problem for the present method com-


pared with the method of Cho et al. [28]. a3 ¼ K c K P cL2 τI α þ cL2 τI τD  0:5LτI τD α þ caL2 τI
þ 0:5LτI τ þ 0:5LaτI β þ τI βτ ðA7Þ
4. Conclusions

a4 ¼ K c K P cL2 α þ cL2 τI  0:5LτI α  0:5LτI τD þ τI τD α
PID controller design using IMC principles for PIPTD, DIPTD, þ 0:5LaτI þ τI τ þ aτI β ðA8Þ
stable FOPTDI with/without a zero and unstable FOPTDI systems is


proposed. Tuning rules are given for stable/unstable FOPTDI sys- a5 ¼ K c K P cL2  0:5Lα  0:5LτI þ τI α þ τI τD þ aτI ðA9Þ
tem, PIPTD system and DIPTD system for Ms ¼2 in terms of model
parameters. First order Pade's approximation is used for time
a6 ¼ K c K P ðα þ τI  0:5LÞ ðA10Þ
delay to design PID controller. The controller is tuned by a single
tuning parameter λ which governs the trade-off between the a7 ¼ K c K P ðA11Þ
performance and robustness of the control system. For larger
values of time delay to time constant, λ is tuned in such a way that c ¼ 1=12 ðA12Þ
a trade-off is made between performance and robustness and the
The Kharitonov's polynomials are given below. Here and aiþ ai
tuning rules for stable FOPTDI system are given. In all the case
are the lower bound and upper bound of ai respectively.
studies, PID controllers are tuned in such a way that they will give
the same maximum magnitude of sensitivity function Ms value as a1þ s6 þ a2þ s5 þa3 s4 þa4 s3 þ a5þ s2 þ a6þ s þ a7 ¼ 0 ðA13Þ
given by the literature reported methods, so that performance
comparison is made for the same robustness level. The designed a1 s6 þ a2 s5 þa3þ s4 þa4þ s3 þ a5 s2 þ a6 s þ a7þ ¼ 0 ðA14Þ
controller performed well for both disturbance rejection and servo
problem and reduced overshoot and peak error. The proposed a1þ s6 þ a2 s5 þa3 s4 þa4þ s3 þ a5þ s2 þ a6 s þ a7 ¼ 0 ðA15Þ
controllers are implemented on various transfer function models,
non-linear model equations of jacketed CSTR and the process with a1 s6 þ a2þ s5 þa3þ s4 þa4 s3 þ a5 s2 þ a6þ s þ a7þ ¼ 0 ðA16Þ
higher order dynamics. The higher order system is reduced to
For fixed value of kP and τ, a perturbation in time delay L i.e.
integrating processes with time delay system using model reduc-
(L  ΔL)rL r(L þ ΔL) is substituted in the above coefficients and
tion techniques. To the reduced model, PID controller is designed
Kharitonov's polynomials are checked for stability by Routh–Hur-
based on the present method and implemented on the original
witz method [49,50]. Similar procedure is repeated to find stability
higher order system. The controller designed by the present
regions for kP and τ.
method gives fast response. The performance comparison in terms
For PIPTD system, since a1 ¼0 as α ¼0 in Eq. (A2), the Khar-
of IAE, ITAE and TV show that the proposed controller performs
itonov's polynomials are given below.
better than the recently reported methods.
a2þ s6 þ a3þ s5 þa4 s4 þa5 s3 þ a6þ s2 þ a7þ s ¼ 0 ðA17Þ

Appendix A a2 s6 þ a3 s5 þa4þ s4 þa5þ s3 þ a6 s2 þ a7 s ¼ 0 ðA18Þ

Kharitonov's theorem a2þ s6 þ a3 s5 þa4 s4 þa5þ s3 þ a6þ s2 þ a7 s ¼ 0 ðA19Þ

The closed loop characteristic equation is given by a2 s6 þ a3þ s5 þa4þ s4 þa5 s3 þ a6 s2 þ a7þ s ¼ 0 ðA20Þ
GðsÞ ¼ 1 þ Gc GP ¼ 0 ðA1Þ For fixed value of kP ; a perturbation in time delay L i.e. (L  ΔL)
where, rLr (L þ ΔL) is substituted in the above coefficients and Khar-

 itonov's polynomials are checked for stability by Routh–Hurwitz
1 1 þ αs method [49,50]. Similar procedure is repeated to find stability
Gc ¼ k c 1 þ þ τ D s ðA2Þ
τI s 1 þ βs region for kP .

kP e  Ls
GP ¼ ðA3Þ
sðτs þ aÞ
References

Case (i): τ ¼ 0 and a ¼1, the process Gp is a pure integrating
system.
Case (ii): τ ¼ 1 and a ¼0, the process is Gp a double integrating
system.
Case (iii): a ¼1, the process Gp is an Stable First Order plus Time
Delay system with an Integrator.
Case (iv): a ¼  1, the process Gp is an Unstable First Order plus
Time Delay system with an Integrator.
Using second order Pade's approximation for time delay, the
characteristic equation is given by
GðsÞ ¼ a1 s6 þ a2 s5 þ a3 s4 þ a4 s3 þ a5 s2 þ a6 s þ a7
where,
a1 ¼ cL2 τI βτ
[1] Fuentes C, Luyben WL. Control of high purity distillation columns. Ind Eng
Chem Process Des Dev 1983;22:362.
[2] Chein IL, Fruehauf PS. Consider IMC tuning to improve performance. Chem
Eng Prog 1990;33:41.
[3] Ogunnaike BA, Ray WH. Process dynamics modeling and control. New York:
Oxford University Press; 1994. p. 666–7.
[4] Wang L, Cluette WR. PID tuning controllers for integrating processes. IEEE Proc
– Control Theory Appl 1997;144:385.
[5] Srividya R, Chidambaram M. Online controller tuning for integrator plus time
delay processes. Process Control Qual 1997;59:66.
[6] Monroy NK, Ramirez M. Control systems. . New Delhi: New Age International
(P) Ltd. Publishers; 2003. p. 144.
[7] Huang CC, Wang QG, Yang XP. Single loop controller design via IMC principles.
Automatica 2008;726:733.
[8] Howard W, Cooper I. Internal model based control for integrating processes.
ðA4Þ ISA Trans 2009;519:527.
[9] Bequette BW. Process control: modeling design and simulation. New Delhi:
Prentice Hall; 2003.
[10] Hovd M, Skogestad S. Pairing criteria for decentralized control of unstable
ðA5Þ plants. Ind Eng Chem Res 1994;33:2134–9.
 D.B.S. Kumar, R. Padma Sree / ISA Transactions 63 (2016) 242–255
[11] Filatov NM, Keuchel U, Unbehaven H. Dual control of unstable mechanical
plant. IEEE Control Syst 1996;16(4):31–7.
[12] Liu T, Cai YZ, Gu DY, Zhang WD. Analytical decoupling control design for
dynamic plants with time delay and double integrators. IEEE Proc Control
Theory Appl 2004;151(6):745–53.
[13] Bodizs L, Titica M, Faria N, Srinivasan B, Dochain D, Bonvin D. Oxygen control
for an industrial pilot scale fed batch filamentous fungal fermentation. J Pro-
cess Control 2007;17:595–606.
[14] Ziegler JG, Nichols NB. Optimum setting for automatic controllers. ASME Trans
1942;64:759–68.
[15] Cohen GH, Coon GA. Theoretical consideration of retarded control. Trans ASME
1953;75:827–34.
[16] Rivera DE, Morari M, Skogested S. Internal model control for PID controller
design. Ind Eng Chem Process Des Dev 1986;25:252–65.
[17] Lee Y, Lee J, Park S. PID controller tuning for integrating and unstable pro-
cesses with time delay. Chem Eng Sci 2000;55:3481–93.
[18] Shamsuzzoha M, Lee M. PID controller design for integrating processes with
time delay. Korean J Chem Eng 2008;25(4):637–45.
[19] Panda R. Synthesis of PID controller for unstable and integrating processes.
Chem Eng Sci 2009;64:2807–16.
[20] Nageswara Rao CV, Padma Sree R. IMC based controller design for integrating
systems with time delay. Ind Chem Eng 2010;52(3):194–218.
[21] Jin QB, Liu Q. Analytical IMC-PID design in terms of performance/robustness
trade-off for integrating processes from 2-Dof to 1-Dof. J Process Control
2014;24:22–32.
[22] Shamsuzzoha M, Lee M. A simple analytical controller design for integrating
and first order unstable time delay process. Theory App Chem Eng 2007;13
(1):125–8.
[23] Chen D, Seborg DE. PI/PID controller design based on direct synthesis and
disturbance rejection. Ind Eng Chem Res 2002;41:4807–22.
[24] Ch Anil, Padma Sree R. Design of PID controllers for FOPTD systems with an
integrator and with/without a zero. Indian Chem Eng 2005;47(4):235–42.
[25] Seshagiri Rao A, Rao VSR, Chidambaram M. Direct synthesis based controller
design for integrating processes with time delay. J Frankl Inst 2009;346:38–56.
[26] Lee J, Cho W, Edgar TF. Simple analytic PID controller tuning rules revisited.
Ind Eng Chem Res 2014;53:5038–47.
[27] Ajmeri M, Ali A. Direct synthesis based tuning of parallel control structure of inte-
grating systems. Int J Syst Sci 2015. http://dx.doi.org/10.1080/00207721.2013.871369.
[28] Cho W, Lee J, Edgar TF. Simple analytic proportional integral derivative (PID)
controller tuning rules for unstable processes. Ind Eng Chem Res
2014;53:5048–54.
[29] Chidambram M, Padma Sree R. A simple method of tuning PID controllers for
integrator/dead time processes. Comp Chem Eng 2003;27(2):211–5.
[30] Padma Sree R, Chidambaram M. A simple and robust method of tuning con-
trollers for integrator/dead time processes. J Chem Eng Jpn 2005;38(2):113–9.
[31] Majhi S, Mahantha C. Tuning of controllers for integrating time delay process.
IEEE Trans 2001;1:317–20 (Catalogue no. 01CH 37239).
[32] Kaya I. Two degree of freedom IMC structure and controller design for inte-
grating processes based on gain and face margin specification. IEE Process
Control Theory Appl 2004;154(4):481–7.
255
[33] Nageswara Rao CV, Padma Sree R. Two degree of freedom controller for
double integrating systems with time delay and unstable FOPTD systems with
an integral. Ind Chem Eng Res 2014;56(4):336–54.
[34] (a) Pouiln E, Pomerleau A. PID tuning for integrating and unstable process. IEE
Proc Control Theory Appl 1996;143(5):429–35;
(b) Tyreus BD, Luyben WL. Tuning PI controllers for integrator/dead time
process. Ind Eng Chem Res 1992;31:2625–8.
[35] Tyreus BD, Luyben WL, Tuning PI. controllers for integrator/dead time process.
Ind Eng Chem Res 1992;31:2625–8.
[36] Luyben WL. Design of proportional integral and derivative controllers for
integrating dead-time processes. Ind Eng Chem Res 1996;35:3480.
[37] Luyben WL, Luyben ML. Essentials of process control. New York: McGraw Hill;
1997.
[38] Visioli A. Optical tuning of PID controllers for integral and unstable processes.
IEEE Proc Control Theory Appl 2001;148:180–4.
[39] Bingual ZA. A new PID tuning technique using differential evolution for
unstable and integrating processes with time delay. In: Proceedings of ICONIP:
LNCS, vol. 3316; 1998. p. 254–60.
[40] Syrcos G, Kookos IK. PID controller tuning using mathematical programming.
Chem Eng Process 2005;44:41–9.
[41] Ali A, Majhi S. Integral criteria for optimal tuning of PI/PID controllers for
integrating processes. Asian J Control 2011;13:328–37.
[42] Arrieta O, Vilanova R, Visioli A. Proportional integral and derivative tuning for
servo/regulation control operation for unstable and integrating processes. Ind
Eng Chem Res 2011;50:3327–34.
[43] Ch Anil, Padma Sree R. Tuning of proportional integral derivative controllers
for integrating systems using differential evolution technique. Indian Chem
Eng 2011;53(4):239–60.
[44] Visioli A, Zhang QC. Control of integral process with dead time. Springer-
Verlag London Limited; 2011.
[45] Shamsuzzoha M. Closed-loop PI/PID controller tuning for stable and inte-
grating process with time delay. Ind Eng Chem Res 2013;52:12973–92.
[46] Wright RA, Kravaris C. Two degree of freedom output feedback controllers for
discrete time non-linear systems. Chem Eng Sci 2006;61:4676–88.
[47] Skogestad S. Simple analytic rules for model reduction and PID controller
tuning. J Process Control 2003;13:291–309.
[48] Vijayan V, Panda RC. Design of a simple set point filter for minimizing over-
shoot for low order processes. ISA Trans 2012;51:271–6.
[49] Sinha NK. Control systems. India: Weily Eastern Ltd.; 1994. p. 142–4.
[50] Kharitonov VL. Asymptotic stability of an equilibrium position of a family of
systems of linear differential equations. Differ Uravn 1978;14:1483–5.
[51] Liu T, Zhang W, Gu D. Analytical design of two-degree-of-freedom control
scheme for open-loop unstable processes with time delay. J Process Control
2005:559.
[52] Liu T, Gao F. Enhanced IMC design of load disturbance rejection for integrating
and unstable processes with slow dynamics. ISA Trans 2011:239.