Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

Journal of Process Control 108 (2021) 112–124

Contents lists available at ScienceDirect

Journal of Process Control


journal homepage: www.elsevier.com/locate/jprocont

Control of nonlinear jacketed continuous stirred tank reactor using


different control structures

Mohammad Atif Siddiqui a,c , , Md. Nishat Anwar b , S.H. Laskar a
a
Department of EIE, National Institute of Technology Silchar, Silchar, India
b
Department of EE, National Institute of Technology Patna, Patna, India
c
Department of EE, Integral University, Lucknow, India

article info a b s t r a c t

Article history: This study presents the control of the nonlinear model of a continuous stirred tank reactor (CSTR)
Received 24 April 2021 through different control structures. The structures considered are single feedback configuration (SFC),
Received in revised form 5 November 2021 cascade control configuration (CCC) and sliding mode control (SMC). To design the controllers, the
Accepted 6 November 2021
dynamic behaviour of CSTR with a recirculating jacket heat transfer system is modelled into a third
Available online 25 November 2021
order unstable transfer function. Efforts have been made for the first time to control the CSTR also
Keywords: through parallel control configuration (PCC). Model matching technique in low frequency region has
CSTR been used to design the controllers of SFC, CCC and PCC. The SMC controller parameters are obtained
Single feedback configuration using root-locus technique and by minimizing the performance index. To analyse the performance of
Cascade control configuration CSTR in these structures, simulations are conducted using the nonlinear model equations of the CSTR.
Parallel control configuration
The performance is also compared with recently reported work. It is observed that CSTR has been
Sliding mode control
satisfactorily controlled using these structures and the best performance has been obtained using CCC.
© 2021 Elsevier Ltd. All rights reserved.

1. Introduction a pre-specified maximum sensitivity value. To further improve


the performance of the CSTR, Vanavil et al. [12] have used and
Continuous stirred tank reactor (CSTR) is an important com- designed a PID controller along with a lead–lag filter based on a
ponent in many industrial processes [1,2]. The maximum benefit second order plus dead time (SOPDT) model. Later on, Novella-
from CSTR is obtained when operated about a definite equilib- Rodriguez et al. [13] have considered the unstable SOPDT model
rium point associated with the optimal productivity or output of of CSTR and proposed stabilizing PID controller in terms of model
a process [1]. The control challenges in operating at the optimal parameters. In these mentioned literature either one state or two
point are its highly non-linear dynamics, unmeasured states, zero state model of CSTR have been considered, neglecting the energy
dynamics and one or more relative degrees [3,4]. These problems around the jacket which leads to three state model of CSTR. It
become more challenging in the presence of system uncertainty is also to note that only single loop control structure has been
and/or external disturbance [5,6]. These issues are addressed utilized to control the CSTR.
by many researchers [4–10] considering the different dynamical Many researchers [4,14–17] have also considered the third
models of CSTR with different types of the controller in the state of CSTR i.e. the dynamics of energy around the jacket and
various control structure. then modelled the CSTR. Considering the third state enriches the
Kumar and Sree [7] have considered a double integrating with dynamic behaviour of the CSTR model but it adds complexity
inverse response dynamical model of CSTR and controlled it with in control using single feedback configuration. With the use of
single loop configuration using PID controller. The parameter of cascade control configuration (CCC) in the CSTR having additional
the PID controller has been derived using the internal model feedback of jacket temperature will improve the load disturbance
control (IMC) principle. Their method has shown improved per- rejection performance [15]. This advantage of the cascade con-
formance when compared with the method proposed by Lee trol configuration has been utilized by Lee et al. [18] to control
et al. [11]. Begum et al. [9] have controlled CSTR using an un- the CSTR by considering the unstable SOPDT model. Here, the
stable first order plus dead time (UFOPDT) model to achieve parameters of the PID controller of two loops of CCC have been
derived through the internal model control (IMC) scheme. The
∗ Corresponding author at: Department of EIE, National Institute of robust performance is obtained by Saraf et al. [17] where they
Technology Silchar, Silchar, India. controlled the CSTR by considering an unstable SOPDT model in
E-mail addresses: atif.nits@gmail.com (M.A. Siddiqui), nishat@nitp.ac.in a CCC. They have achieved improved performance due to the
(M.N. Anwar), shlaskar@ei.nits.ac.in (S.H. Laskar). simultaneous auto tuning of the cascade controllers.

https://doi.org/10.1016/j.jprocont.2021.11.005
0959-1524/© 2021 Elsevier Ltd. All rights reserved.
M.A. Siddiqui, M.N. Anwar and S.H. Laskar Journal of Process Control 108 (2021) 112–124

Further performance improvements have been obtained using


a complex three-degree-of-freedom CCC by Liu et al. [19]. They
have considered the unstable SOPDT model of CSTR and con-
troller design is based on the complementary sensitivity function.
Bhaskaran and Rao [20] applied a predictive control method
on a CSTR in CCC with three controllers and three filters. The
controllers and filters are designed using an unstable CSTR model,
IMC scheme, direct synthesis approach and predictive scheme.
The method has shown enhanced performance over the meth-
ods of Yin et al. [21] and Dasari et al. [22]. Though the per-
formance of the CSTR is improved using CCC under nominal
conditions but ensuring a satisfactory performance even under
model discrepancies is a challenging task.
The sliding mode control (SMC) derived from variable struc-
ture control (VSC) have robustness against parametric uncertain-
ties and un-modelled dynamics. Few researchers [3,23–25] have
used the SMC technique in the control of the CSTR and achieved
robust closed-loop performance. Camacho et al. [23] have used
the SMC concept to control the CSTR by considering a stable Fig. 1. CSTR with cooling jacket.
FOPDT model with inverse response dynamics. They have shown
robust performance against modelling errors, noise and distur-
bances. Further improvements in the robust performance have
order systems. These methods are recently developed and are
been obtained using a complex sliding mode predictive controller
well-suited candidates for CSTR.
by Garcia-Gabin et al. [24], where they modelled CSTR as a stable
In this regard, all the above strategies have been applied to
second order model with inverse response characteristics. A ter-
control the CSTR by considering the unstable third order model
minal sliding mode control has been proposed by Zhao et al. [3]
(which is a maiden application to third order unstable model).
to control the CSTR by considering its second order model. They
The control structures utilized in this work are single feedback
have shown improved disturbance rejection and fast setpoint
configuration, cascade control configuration, parallel control con-
tracking in comparison with the method of Camacho et al. [23] at
figuration and sliding mode control configuration. To design the
the cost of high control efforts. A less load on the controller with
controller of these structures, the closed-loop transfer function of
stable closed-loop performance over multiple operating points is
the CSTR has been modelled into an unstable third order model by
achieved by Sinha and Mishra [25] for CSTR by combining the considering jacket make-up flowrate as the manipulated variable.
SMC scheme with event triggering control. The controllers of the single feedback configuration (SFC), cascade
A summary of the literature related to the control of CSTR has control configuration (CCC) and parallel control configuration
been reported in Table 1. (PCC) are designed by selecting the desired closed-loop transfer
Based on the aforementioned literature survey, the following function either for load disturbance rejection or setpoint tracking.
conclusions have been drawn: The controller is then approximated using model-matching in
the frequency domain. On the other hand, the SMC controller
1. In literature, CSTR has been controlled by many structures
is designed by utilizing the parameters of the unstable third
such as SFC, CCC, SMC, etc, but none of the works shows
order model, root-locus technique and minimizing the perfor-
a comparative analysis of the control of CSTR through
mance index using a recently developed metaheuristic algorithm,
different structures.
i.e. grasshopper optimization algorithm. Finally, the comparative
2. It is to note that none of the literature has considered analysis of these different structures i.e. SFC, CCC, SMC and PCC,
an unstable third order transfer function of CSTR for con- has been conducted using the nonlinear model equations of the
troller design. The reason may be, majority of the controller CSTR. It is observed that all these structures have successfully
design methods are based on the first or second order stabilized the unstable nonlinear jacketed CSTR.
stable/unstable system model. The salient features of the proposed work are:
(i) The unstable third order model of CSTR (Jacket makeup
Researchers usually designed controllers based on the lower
flowrate as a manipulated variable) has been considered first time
order model owing to reduce the mathematical complexity that
to design the controller
may arise due to (i) higher order model, (ii) limitation imposed
(ii) CSTR has been controlled using third order unstable model
by the controller order (PI or PID), (iii) approximation technique
with SFC, CCC, PCC and SMC.
involvement to obtain low order controller. To overcome these
(iii) The PCC has been applied first time on the CSTR model and
issues, in the last decade, few researchers came up with unified
in this work we have considered a third order unstable model of
approaches which are applicable on a wide range of processes CSTR.
(like lower order, higher order, stable, unstable, etc.) [27–30]. (iv) A detailed comparative analysis of all the aforementioned
Anwar and Pan [27] and Anwar et al. [28] suggested an approach structures has been carried out on CSTR.
in the single loop that is applicable to higher order systems in The outline of the work is as follows: System description
addition to low order model with diverse dynamics. Siddiqui and mathematical modelling of CSTR are discussed in Section 2.
et al. [30] and Raza and Anwar [31] have applied PCC to high Section 3 deals with the study of controller design techniques.
as well as low order stable and unstable processes. Recently, The simulation study is presented in Section 4 followed by the
Siddiqui et al. [32] extended the work of Anwar and Pan [27] and conclusion in Section 5.
designed the controllers of CCC to control the unstable cascade
processes. It is to note that all these methods utilizes approximate 2. System description and mathematical modelling of CSTR
frequency response matching in low frequency region [33–36].
Siddiqui et al. [37] have overcome the limitation of SMC to FOPDT A continuous stirred tank reactor is shown in Fig. 1. The CSTR
and proposed a sliding mode control strategy for unstable higher convert feed input into a product by allowing the fluid stream
113
M.A. Siddiqui, M.N. Anwar and S.H. Laskar Journal of Process Control 108 (2021) 112–124

Table 1
Literature on CSTR control.
Control Description CSTR model used Control approach Controller Complexity Inference
structure
used
Single Kumar and Double integrating Internal model control PID Complex due to use of Have overshoot and
feedback Sree [7] with inverse (IMC) principle two additional filter undershoot
configura- response model
tion Lee et al. [11] Double integrating Internal model control PID Complex due to complex Large control action
(SFC) with inverse (IMC) principle setpoint filter
response model
Begum et al. [9] UFOPD model Maximum sensitivity based PID Simple Sluggish response
Novella-Rodriguez Unstable SOPDT Frequency domain-based PD or PID Simple Settling time
et al. [13] techniques moderate
Proposed SFC Unstable third Model- matching in PI Simple No overshoot
order with inverse frequency domain
response
Cascade Chidambaram and Unstable SOPDT Model reference PID Complex due to addition No overshoot but
Control Con- Yugender [16] of random numbers sluggish response
figuration Lee et al. [18] Unstable SOPDT IMC scheme PID with Simple Oscillatory response
(CCC) model setpoint
filter
Liu et al. [19] Unstable SOPDT Complementary sensitivity PID with Complex due to complex Deprived regulator
dynamics function, pre-filters structure performance
Bhaskaran and Unstable SOPDT IMC scheme and direct Predictive Complex due three Overshoot and
Rao [20] synthesis control controller and three undershoot
method filter
Jeng [26] Stable third order IMC approach PI, PID Simple Sluggish
with inverse
response
Proposed CCC Unstable third Model- matching in PI with filter Simple Smooth closed-loop
order frequency domain performance
Camacho et al. Stable FOPDT Plant model parameters SMC Simple Overshoot and
Sliding
[23] model undershoot
Mode
Control Garcia-Gabin et al. Stable second Linear model predictive SMC Complex due to Presence of chattering
[24] order model control combining SMC with
predictive control
Zhao et al. [3] Second order Terminal sliding mode SMC Complex High control action
model control
Sinha and Second order Event triggered sliding SMC Complex due to Less control action
Mishra [25] complex model mode combining SMC with
event triggering control
Proposed SMC Unstable third Model parameters and SMC Simple Low overshoot with
order minimization of fast settling time
performance index
Parallel As per author’s knowledge, literature is not available
Control Con-
figuration
Proposed PCC Unstable third Model- matching in PID with Simple Smooth regulatory
order frequency domain filter performance

continuously fed into the reactor. The product gets perfectly where F is the inlet flow rate, T is the temperature of the reactor,
mixed in the CSTR and the exit stream contains the fluid of Ca is the concentration of component A, V is the volume of the
uniform temperature and concentration. The jacket is mounted reactor, k0 is the frequency factor, Caf is the feed concentration,
around the reactor, which also has the feed and exit streams. The R is the ideal gas constant, Ea is the activation energy, Cp is
jacket too has the perfect mix and maintains the reactor tem- the specific heat capacity, −∆H is the heat of reaction, ρ is the
perature at the desired level. The dynamic equations that govern
density, Ar is the area of heat transfer, Tf is the feed temperature,
the reaction inside the CSTR are represented by the nonlinear
U is the heat transfer coefficient. By assuming perfect mixing, an
ordinary differential equations as
( ) additional energy balance around the cooling jacket is obtained
dCa F ( −E a as
Ca ,
)
= Caf − Ca − k0 exp (1)
dt V RT
dTj Fjf ( UAr (
T − Tj .
) )
= Tjf − Tj + (3)
and dt Vj Vj ρj Cpj

−∆ H
( ) ( )
dT F ( −E a UAr ( where Tj represents the cooling jacket temperature, Fjf is the
T − Tj .
) )
= Tf − T + k0 exp Ca −
dt V ρ Cρ RT V ρ Cp jacket make-up flowrate, Tjf is the jacket inlet coolant temper-
(2) ature.
114
M.A. Siddiqui, M.N. Anwar and S.H. Laskar Journal of Process Control 108 (2021) 112–124

a11 a12 a13 ∂ f1 /∂ x1 ∂ f1 /∂ x2 ∂ f1 /∂ x3


⎡ ⎤ ⎡ ⎤

A = ⎣a21 a22 a23 ⎦ = ⎣∂ f2 /∂ x1 ∂ f2 /∂ x2 ∂ f2 /∂ x3 ⎦ =


⎢ ⎥ ⎢ ⎥

a31 a32 a33 ∂ f3 /∂ x1 ∂ f3 /∂ x2 ∂ f3 /∂ x3


Fs
⎡ ⎤
− − k0 exp (−Ea /RTs ) −Cas k0 exp (−Ea /RTs ) Ea /RTs2
( )
0 (6)
⎢ V
⎢ (−∆H ) (−∆H )

F UAr UAr ⎥
⎢ ρ C k0 exp (−Ea /RTs ) Cas k0 exp (−Ea /RTs ) Ea /RTs2
( )
⎢ − − + ⎥
⎢ p V V ρ Cp ρ Cp ( V ρ Cp

)⎥
⎣ UAr Fjfs UAr ⎦
0 − +
Vj ρj Cpj Vj Vj ρj Cpj

Box I.

The generalized state space representation of a system is given The steady state operating point is considered as Cas =
as 0.066 lbmol/ft3 and Ts = 101.1 ◦ F [1,38]. The linearized state
.
x = Ax + Bu space model around the steady state operating point for CSTR
. (4) with jacket make-up flowrate of 800 ft3 /h is obtained as
y = Cx + Du
−7.9909 −0.013674 0 0
⎡ ⎤ ⎡ ⎤
For Eqs. (1)–(3), the state space representation with the states
A = ⎣ 2922.9 4.5564 1.4582 ⎦ , B = ⎣ 0
⎢ ⎥ ⎢ ⎥
(x), input (u) and outputs (y) are represented in the following ⎦
deviation variables: 0 4.7482 −5.8977 −3.2558
x1 Ca − Cas
[ ] [ ] [ ] [ ]
0 1 0 0
x= x2 = T − Ts C = ,D =
x3 Tj − Tjs 0 0 1 0
[ ] [ ]
u1 Fjf − Fjfs (5) (10)
u= =
u2 Tjf − Tjfs
[ ] [ ] Using Eqs. (9) and (10), the transfer function of the CSTR with
y1 T − Ts jacket make-up flowrate as manipulating variable is obtained as:
y= =
y2 Tj − Tjs

Here, the subscript (s) indicates steady state value and the equa- T −4.747s − 37.94
tions are linearized around this steady state value. The method Gp (s) = = , (11)
Fjf s3 + 9.332s2 + 16.89s − 34.35
based on the Taylor series approximation has been used to get
the linear model of CSTR. and the transfer function of CSTR with jacket inlet temperature
The elements of the Jacobian matrix are Eq. (6) is given in Box I as the manipulating variable is obtained as
The elements of matrix B are
T 46.66s + 372.9

0 0
⎤ Gp (s) = = . (12)
b11 b12 ∂ f1 /∂ u1 ∂ f1 /∂ u2 Tjf s3 + 40.18s2 + 122.8s + 75.4
⎡ ⎤ ⎡ ⎤
⎢ ⎥
0 ) 0 ⎥ The transfer function in Eq. (11) has two stable poles at −5.25 ±
B = ⎣b21 b22 ⎦ = ⎣∂ f2 /∂ u1 ∂ f2 /∂ u2 ⎦ = ⎢ ( (7)
⎢ ⎥ ⎢ ⎥ ⎢
⎣ Tjfs − Tjs Fjfs ⎦

1.27i and one unstable pole at +1.175, whereas, the poles of the
b31 b32 ∂ f3 /∂ u1 ∂ f3 /∂ u2 V V j j model in Eq. (12) have three stable poles at −0.84, −2.43 and
The elements of matrix C and D are −36.9.
[ ] [ ] When the recirculating flowrate is altered from 800 ft3 /h to
c11 c12 c13 0 1 0
C = = 270.105 ft3 /h, the CSTR transfer function is obtained as
c21 c22 c23 0 0 1 T 15.75s + 125.9
[ ] Gp (s) = = . (13)
0 Tjf s3 + 18.99s2 + 50.05s + 0
D= (8)
0 The poles of this model are located at 0, −3.16, −15.82 and
The transfer function of CSTR from state space model in Eq. (4) is thereby indicating integrating dynamics. It is observed that the
obtained as dynamic behaviour of CSTR changes under different operating
conditions. These dynamic behaviour can also be captured using
Gp (s) = C (sI − A)−1 B + D. (9) different system identification packages available in Python [39]
The CSTR model used in Bequette [1,38] has been considered for and MATLAB [40–42].
modelling and controller design. A summary of modelling of CSTR In the literature, many researchers have investigated the con-
described in Bequette [1,38] is presented here in brief. trol of CSTR with third order stable and integrating dynamical
The operating parameters and constants of CSTR are as fol- model [26,43–45]. However, the control of CSTR with third order
lows: Ea = 32, 400 Btu/lbmol, −∆H = 39 000 Btu/lbmol, k0 = unstable model is difficult and as per the author’s best knowledge,
16.96 × 1012 h−1 , V /F = 0.25 h, UA = 6600 Btu/h ◦ F, R = there is no literature available that investigates the control of
1.987 Btu/lbmol ◦ F, ρ Cρ = 53.25 Btu/ft2 ◦ F, F = 340 ft3 /h, Tjf = this unstable model of CSTR. In this manuscript, the controllers
0 ◦ F, Caf = 0.132 lbmol/ft3 , Tf = 60 ◦ F, Vj /V = 0.25, ρj Cρ j = of the different structures are designed considering the unstable
55.6 Btu/ft3 ◦ F and V = 85 ft3 . transfer function of CSTR as mentioned in Eq. (11).
115
M.A. Siddiqui, M.N. Anwar and S.H. Laskar Journal of Process Control 108 (2021) 112–124

Selecting one frequency point as ω1 and dissociating real and


imaginary parts yields the following equation which is grouped
in a matrix form as

( −Re [a2 (ω1 )] )


⎡ ⎤[ ]
1 Kp
−1
− Im [a2 (ω1 )]
⎣ ⎦
0 Ki
ω1
−Re [b2 (ω1 )]
[ ]
= . (19)
Fig. 2. Bock diagram of SFC. −Im [b2 (ω1 )]
Solving Eq. (19) by selecting the value of ω1 close to 0.1 percent of
the bandwidth of the desired model will determine the controller
3. Study of controller design techniques
parameters. A thorough description of the design method, as well
This section deals with the synthesis of controllers for different as the selection of frequency points, is available in [27,28].
structures such as single feedback configuration, cascade control
configuration, parallel control configuration and sliding mode 3.2. Controller design of CCC
control. The controllers of the SFC, CCC and PCC are designed
based on model matching in the frequency domain [27,28,30,32]. A cascade control configuration is a multi-loop approach used
The SMC controller is designed based on the model parameters to achieve better disturbance rejection. The CCC usually con-
and minimization of the performance index [37]. A brief descrip- sists of two loops i.e. primary loop and the secondary loop. The
tion of the controller design technique of these structures are secondary loop is formed by using an additional sensor and a
presented as follows: controller which thereby improves the load disturbance property.
The cascade control configuration is shown in Fig. 3, where,
3.1. Controller design of SFC Gc1 (s), Gp1 (s), Gc2 (s) and Gp2 (s) are the primary controller, primary
plant, secondary controller and secondary plant, respectively. A
Fig. 2 shows the single feedback configuration with the CSTR setpoint filter (1/(β s + 1)) is used to reduce the overshoot in
model as Gp (s) and the controller as Gc (s). In the figure, r and d are the setpoint response. The setpoint command is indicated by r,
the setpoint and load disturbance, respectively. y is the process whereas, the output of the cascade system is indicated by y(s).
output, e is the error, uc is the controller output and x is the input The manipulated input is uc (s), the output of the secondary loop is
to the CSTR. The structure of the controller Gc is considered as a y2 (s) and load disturbances are D1 and D2 . The input to the inner
PI controller with transfer function as loop and Gp2 is indicated by r2 (s) and x, respectively.
Ki The schematic diagram representing the application of CCC to
Gc (s) = Kp + , (14) CSTR is shown in Fig. 4. The reactor temperature controller is
s
Gc1 and the jacket temperature controller is Gc2 . The controller
Here, Kp and Ki are the proportional and integral gains of the
Gc1 is used to track the reactor temperature, while Gc2 is utilized
controller.
to reject the disturbances (D1 and D2 ). The additional sensor of
To design the controller, a desired closed-loop transfer func-
the secondary loop measures the jacket temperature which is
tion or model which comprises an anticipated performance spec-
the output of the inner loop. Both the measurements i.e. jacket
ification for the regulatory performance is selected. The desired
temperature (y2 (s)) and reactor temperature (y(s)) along with the
model Md,y is selected as:
temperature setpoint are used to adjust the jacket make-up flow
sK rate (uc (s)).
Md,y = . (15)
(λs + 1)n To design the controllers of the CCC, the third order model
of the CSTR is converted into a cascaded plant having outer
Here, the desired model Md,y contains a tuning parameter λ used 1
and inner loop processes as Gp1 = s2 +10.51s +25.99
and Gp2 =
to provide an adjustment between robustness and performance,
−4.747s−34.94
n is the order of the desired model and K = 1/Ki [27,29,30,32]. s−1.175
,
respectively. First, the secondary controller Gc2 (s) is
Furthermore, one zero is placed at the origin to achieve rejection designed with an objective to force the inner loop to track the
of the load disturbance signal. desired load disturbance rejection. Hence, a desired disturbance
After determining the desired model to achieve better distur- rejection transfer function that comprises an anticipated load dis-
bance rejection as in Eq. (15), the controller is designed to match turbance performance specification is selected for the inner loop.
the response of the closed-loop control system with the desired To achieve the above-said objective, the concept of the model
load disturbance model. It is mathematically expressed as matching in the frequency domain has been used to synthesis
⏐ ⏐ the PI controllers [29,32,46]. After synthesizing Gc2 , the Gc1 is
y(s) ⏐ Gp (s)
= Md,y (s)⏐s=jω .

⏐ ⏐
⏐ = ⏐ (16) designed by considering a plant that consists of Gp1 along with Gc2
d(s) ⏐ s=jωn
1 + Gc (s)Gp (s) ⏐ n
and Gp2 . The desired model having setpoint tracking performance
s=jω
n

Re-arranging Eq. (16) as component is chosen and the controller is designed by utilizing
⏐ the concept of model matching at a low-frequency point. The
1 1 ⏐ detailed design procedure for the primary controller (Gc1 ), the
Gc (s)|s=jωn = − ⏐ . (17)
Md,y (s) Gp (s) ⏐s=jω secondary controller (Gc2 ) and the setpoint filter is mentioned
n
n
in [32].
Ki (λ(jωn )+1) 1
Substituting Gc as PI and assuming (jωn )
= a2 and Gp (jωn )
=
b2 , Eq. (17) can be written as 3.3. Controller design of PCC
Ki
Kp + = Ki (Re [a2 ] + jIm [a2 ]) − (Re [b2 ] + jIm [b2 ]) , (18) The parallel control configuration shown in Fig. 5 is a two-
jωn degree freedom configuration where the two responses i.e. set-
here, Re indicates real part and Im indicates imaginary part. point and load disturbance response are decoupled.
116
M.A. Siddiqui, M.N. Anwar and S.H. Laskar Journal of Process Control 108 (2021) 112–124

Fig. 3. Block diagram of the cascade control configuration.

Fig. 4. Process flow diagram for CSTR in a cascade control configuration.

Fig. 5. Block diagram of PCC.

The output (y) of the PCC is given as Assuming the perfect modelling of plant i.e. Gp = Gm , the
[ ][ ] [ ] Eq. (20) is written as
Gp 1 + Gcl Gm Gcs Gm Gp
y= r+ d (20)
Gm 1 + Gcl Gp 1 + Gcs Gm 1 + Gcl Gp

Here, Gp is the actual plant and its model is represented as Gm . (21)


Gcs and Gcl are the controllers of the PCC having control signals as Eq. (21) is further simplified as
u11 and u22, respectively. r is the setpoint input and d is the load
[ ] [ ]
Gcs Gm Gm
disturbance input. y= r+ d. (22)
1 + Gcs Gm 1 + Gcl Gm
117
M.A. Siddiqui, M.N. Anwar and S.H. Laskar Journal of Process Control 108 (2021) 112–124

Fig. 7. Block diagram of SMC.

has robustness against uncertainties and load disturbances. The


sliding mode control scheme is shown in Fig. 7. Here, Gp is
the model of the CSTR, r is the setpoint input and d is the
load disturbance. In this manuscript, the SMC controller design
technique proposed by Siddiqui et al. [37] has been applied to the
unstable third order model of CSTR. Their method was proposed
for unstable second order plus dead time model. However, the
CSTR model is having a zero at −8 in addition to three poles
at −5.25 ± 1.27i, +1.175. So to apply the method of Siddiqui
Fig. 6. Schematic diagram of CSTR controlled through PCC.
et al. [37], a suitable model of CSTR is obtained using first order
Taylor series approximation (e−θ s ≈ 1 − θ s).
The transfer function of CSTR is
From Eq. (22), it is observed that the setpoint response is de-
−37.94(1 + 0.1251s)
coupled from the load-disturbance response. This advantageous Gp (s) = . (26)
feature of PCC allows separate control of setpoint and load dis- (s − 1.175)(s2 + 10.5s + 29.18)
turbance response using Gcs and Gcl , respectively. The schematic By approximating the unstable pole and zero using e−θ s ≈ 1 − θ s,
diagram representing the application of PCC to CSTR is shown in Eq. (26) is written as
Fig. 6.
The two controllers of PCC are considered as −37.94
Gp (s) = . (27)
2
Kd1 s + Kp1 s + Ki1 −1.175(e−0.8511s )(e−0.1251s )(s2 + 10.5s + 29.18)
Gcs = ( ) (23) Further Eq. (27) is simplified as
s Tf 1 s + 1
2
Kd2 s + Kp2 s + Ki2 −37.94
Gcl = . (24) Gp (s) = . (28)
−1.175e−(0.8511+0.1251)s (s2 + 10.5s + 29.18)
( )
s Tf 2 s + 1
Here, Kp1 and Kp2 are the proportional gain, Ki1 and Ki2 are the Using first order Taylor series approximation, Eq. (28) may be
integral gain, Kd1 and Kd2 are the derivative gain and Tf 1 and Tf 2 written as
are the filter time constant. −37.94
Gp (s) = . (29)
The controllers (Gcs and Gcl ) are designed using the model 1.147s3 + 10.87s2 + 21.12s − 34.28
matching concept in the frequency domain to achieve the desired
The parameters of Eq. (29) are used to synthesize the SMC con-
setpoint and load disturbance performance. For model matching,
troller. The design objective of the SMC is to pull the system from
the desired closed-loop transfer function for setpoint tracking
its initial state to a surface called a sliding surface (ss(t)). The
or load disturbance rejection is selected and matched in a low
sliding surface ss(t) is selected as
frequency region with the designed closed-loop transfer function.
t
d2

The desired closed-loop transfer function for load disturbance is d
selected as shown in Eq. (15). The desired closed-loop transfer ss(t) = k1 e(t) + k2 e(t)dt + k3 e(t) + e(t). (30)
0 dt dt 2
function for setpoint specification is selected as
Here, e(t) = y(t)-r(t) is the error, r(t) is the setpoint input, y(t) is
(β0 s + 1)
Mr ,y (s) = , (25) the system output, k1 , k2 , k3 are the gains.
(λs s + 1)n1 After selecting the sliding surface, the SMC law is developed
where, n1 is the order of the desired reference model, λs repre- to satisfy the conditions ss(t) = 0 and dtd (ss(t)) = 0. The SMC
sents the desired system speed. The term (β0 s + 1) is considered law is a combination of two components: a discontinuous compo-
in the desired closed-loop transfer function to avoid the unstable nent, U d (t) and a continuous component, Ucc (t). The sliding mode
pole-zero cancellation between the plant and the setpoint con- control law Utotal may be written as
troller (Gcs ). The value of β0 is selected as β0 = τp1 eθ1 /τp1 (λ1 /τp1 +
Utotal (t) = Ud (t) + Ucc (t). (31)
1)n1 − 1, which is described in detail in [31,32]. Once the desired
closed-loop transfer functions have been selected, the frequency The continuous and the discontinuous components are expressed
response matching at two low frequency points is performed to as
yield the controller parameters of the parallel control structure.
For a detailed description of the method to design Gcs and Gcl , one ss(t)
Ucc (t) = f (y(t), r(t), e(t)) and Ud (t) = KD , respectively.
may refer to Siddiqui et al. [30,31,47]. |ss(t)| + δ
(32)
3.4. SMC controller design
In Eq. (32), Ucc (t) is the functions of the system output, set-
The sliding mode control is basically a particular type of vari- point input and error. The discontinuous component, Ud (t) has
able structure control system. It is a robust control scheme that two adjusting parameters δ and KD responsible for reducing the
118
M.A. Siddiqui, M.N. Anwar and S.H. Laskar Journal of Process Control 108 (2021) 112–124

Table 2
Controller settings and performance evaluation.
Method Controller δoutput% δco% ymax
r Tjf Tf
−0.01 4.89
CCC Gc2 = −0.01 + Gc1 = 131.1 + 3.96 3.44 104.3 94.1 112
s s
s 29s + 1
Desired model = Desired model =
(1.5s + 1)3 (35s + 1)1
−0.3723 s
SFC Gc = −1.65 + , Desired model = , 4.64 3.68 101.1 94 111.5
s (1.5s + 1)3
−0.708s2 − 1.599s − 0.2168 −5.966s2 − 1.788s − 0.1074
PCC Gcl = Gcs = 5.72 2.03 123.1 93.2 111.9
[s(0.4797s + 1)] [s(4.66s + 1)]
23.5
SMC Gc = 18.66 + + 9.35s + s , KD = 199.12, δ = 112.2
2
3.71 6.53 108.9 96 108.5
s
−0.46
SHAMS Gc = −0.94 + , KCO = −1, ov ershoot = 0.4, tp = 20, b = 7.31 4.21 2.38 101.1 88.7 115.1
( s ) ( )
1 1
JENG Gc2 = −0.964 1 + Gc2 = 1.284 1 + + 0.727s 7.06 6.44 101.1 99.7 111.6
0.522s 2.432s

undesirable chattering phenomenon and guiding the state trajec- Table 3


tory towards the sliding surface, respectively. The parameters of Performance indices.

the continuous component are derived in terms of CSTR model Methods IAE ISE ITAE IAE improvement of TV
CCC with respect to
parameters using the root locus technique. On the other hand,
δ and KD are tuned by minimizing a performance index using CCC 91.96 3093 4508 – 243.67
SFC 258.4 4640 12 910 180.99% 329.01
the grasshopper optimization technique. A detailed description PCC 446.7 7177 23 000 385.75% 254.31
of designing the continuous and discontinuous component of the SMC 99.08 2708 5279 7.74% 2020.4
SMC control law is available in [37]. SHAMS 189.1 4294 14 850 105.63% 213.74
JENG 529.2 11 390 21 070 475.46% 618.39
4. Simulation results and discussion

In this section, the aforementioned controller design methods jacket make-up flowrate response is illustrated in Fig. 10. From
have been applied to the nonlinear CSTR model (Eqs. (1)–(3)) these figures, it is observed that the reactor temperature response
given in Bequette [1,38]. To design the controller, a linearized obtained by the CCC has the smallest overshoot with fast setpoint
third order unstable model (Eq. (11)) has been considered. The tracking. Whereas, the reactor temperature obtained by JENG has
design parameter of these structures and obtained controller a sluggish response. The settling time of SHAMS and SMC are
parameters are given in Table 2. The methods mentioned in this
similar, but the later one has overshoot.
work are compared with the methods of Shamsuzzoha [48] and
The disturbances in the jacket inlet coolant temperature Tjf are
Jeng [26]. The work of Shamsuzzoha has been applied on the third
quickly rejected by CCC whereas, sluggish disturbance rejection is
order unstable model of the CSTR and the controller parameters
observed in SHAMS, PCC, SFC and SMC. Also, better disturbance
are given in Table 2. It is to note that Jeng has designed the
rejection in feed temperature Tf is observed in CCC. The jacket
controller by considering a stable third order model of CSTR. In
inlet coolant temperature response of SHAMS, PCC, SFC and JENG
this article, the work of Shamsuzzoha and Jeng is referred to as
have a long settling time with a slightly large overshoot. A slightly
SHAMS and JENG, respectively.
higher jacket make-up flow rate is required by SMC, SHAMS and
The performance of these methods have been compared in
JENG to maintain the desired reactor temperature. Performance
terms of (i) integral absolute error (IAE) (ii) integral square error
indices obtained by CCC, SFC, PCC, SMC, SHAMS and JENG are
(ISE) (iii)
∑integral time absolute error (ITAE) (iv) total variation
∞ listed in Tables 2 and 3. From these tables, it is observed that the
(TV = i=1 |ui+1 − ui |), here ui is the discretized manipulated
variable (v) IAE improvement (vi) peak value (ymax ) (vii) δoutput% least value of IAE and ITAE value is observed for CCC and the ISE
and δco% value. The δoutput% and δco% values are defined as the value is least for SMC. IAE improvement of CCC over the method
standard deviation of the feed temperature and jacket make-up of JENG is highest whereas, it is lowest for SMC.
flowrate, respectively, with respect to the magnitude of set point Furthermore, to validate the efficacy of the different
change under noisy conditions [49]. IAE improvement of the CCC approaches under noisy conditions, a white noise having noise
in terms of total IAE over the other structures is calculated as power equal to 0.1, seed equal to 0 and sampling time equal
to 0.1 s on the process output is considered. Simulation is con-
IAEof other structure − IAEof CCC
IAE improv ement = × 100%. (33) ducted and the impact of noise on the reactor temperature
IAEof CCC and jacket( make-up flowrate is quantified and expressed as
δoutput% = σsystem output /r ∗ 100 and δco% = (σcontroller output /r) ×
)
To evaluate the closed-loop performance, simulations are per-
formed in MATLAB 2013a SIMULINK environment by utilizing an 100, respectively. Where σsystem output and σsystem output is the stan-
®
Intel core™ i5-8250U CPU @ 1.60 GHz and 8 GB RAM desktop. dard deviation of the reactor temperature and jacket make-up
The SIMULINK block diagram of only cascade control (CCC) of flowrate, respectively, with respect to the amplitude of the set-
CSTR is shown in Fig. 8 due to space limitations. Simulations are point signal [29]. The reactor temperature and jacket make-up
conducted on the nonlinear model Eqs. (1)–(3) of the CSTR by flowrate response under noisy conditions are shown in Figs. 11
applying a setpoint command of 101.1 ◦ F at t = 0. The distur- and 12, respectively. From these figures, it is observed that all
bance change has been considered in Tjf from 0 ◦ F to 40 ◦ F at the methods are able to control CSTR under noisy conditions. The
t = 100 and Tf from 60 ◦ F to 80 ◦ F at t = 150. The resulting values of δoutput% and δco% are listed in Table 2 where it is observed
reactor temperature is shown in Fig. 9 and the corresponding that SMC and CCC have lesser values of δoutput% compared to JENG,
119
M.A. Siddiqui, M.N. Anwar and S.H. Laskar Journal of Process Control 108 (2021) 112–124

Fig. 8. Cascade control configuration for CSTR (a) in the differential equations form and (b) SIMULINK model.

PCC, SFC and SHAMS whereas the lowest δco% value is observed introducing perturbation of ±50% in the density (ρ and ρj ), spe-
for PCC. cific heat capacity (Cp and Cpj ), area of heat transfer (Ar ), heat
To analyse the robustness of these methods in different struc- transfer coefficient (U) and operating volumes (V and Vj ). The
tures (SFC, CCC, PCC and SMC), simulations are conducted by corresponding perturbed responses of SFC, CCC, PCC and SMC are
120
M.A. Siddiqui, M.N. Anwar and S.H. Laskar Journal of Process Control 108 (2021) 112–124

Fig. 9. Reactor temperature response.

Fig. 10. Jacket make-up flowrate response.

Fig. 11. Reactor temperature response under noisy conditions.

shown in Fig. 13. From the response, it is clear that the methods in 5. Conclusion
different structures are capable of controlling the nonlinear CSTR
In this work, different control schemes have been investi-
under perturbed conditions. For −50% perturbation, the method gated to control a nonlinear CSTR model. Four different control
structure that includes (i) single feedback (ii) cascade control (iii)
for SMC shows sluggish response, otherwise, all other methods
parallel control (iv) sliding mode control have been adopted to
showed good robustness to parametric change. control a third order CSTR model with unstable dynamics. The
121
M.A. Siddiqui, M.N. Anwar and S.H. Laskar Journal of Process Control 108 (2021) 112–124

Fig. 12. Jacket make-up flowrate response under noisy conditions.

Fig. 13. Perturbed responses for SFC, CCC, PCC and SMC.

controller in SFC, CCC and PCC are designed using model match- is better in terms of reactor temperature tracking, quickly reject-
ing technique in frequency domain. The parameter of SMC con- ing the disturbances and maintaining the smooth jacket make-up
troller are obtained utilizing root locus technique and minimized flow rate. The regulatory performance of SMC is quite good but
performance index in terms of CSTR model parameters. The com- demands high jacket make-up flowrate. In noisy environment,
parative performance analysis of these methods along with the the performance of PCC is better as the fluctuation in the jacket
method of Jeng and Shamsuzzoha is performed. All these methods make-up flowrate is minimum. The overall performance of the
give satisfactory performance, however, the performance of CCC CCC is better over the other structure/methods. So, a suitable
122
M.A. Siddiqui, M.N. Anwar and S.H. Laskar Journal of Process Control 108 (2021) 112–124

structure to control the nonlinear jacket CSTR may be selected [20] A. Bhaskaran, A.S. Rao, Predictive control of unstable time delay series
as per the requirement. cascade processes with measurement noise, ISA Trans. 99 (2020) 403–416,
http://dx.doi.org/10.1016/j.isatra.2019.08.065.
[21] C.Q. Yin, H.T. Wang, Q. Sun, L. Zhao, Improved cascade control system for
CRediT authorship contribution statement a class of unstable processes with time delay, Int. J. Control. Autom. Syst.
17 (2019) 126–135, http://dx.doi.org/10.1007/s12555-018-0096-8.
Mohammad Atif Siddiqui: Conceptualization, Methodology, [22] P.R. Dasari, L. Alladi, A. Seshagiri Rao, C.K. Yoo, Enhanced design of cascade
control systems for unstable processes with time delay, J. Process Control.
Writing – review & editing. Md. Nishat Anwar: Methodology,
45 (2016) 43–54, http://dx.doi.org/10.1016/j.jprocont.2016.06.008.
Reviewing, Supervision. S.H. Laskar: Reviewing, Supervision. [23] O. Camacho, R. Rojas, W. García, Variable structure control applied to
chemical processes with inverse response, ISA Trans. 38 (1999) 55–72,
Declaration of competing interest http://dx.doi.org/10.1016/S0019-0578(99)00005-1.
[24] W. García, J.E. Normey-Rico, E.F. Camacho, Sliding mode predictive control
of a delayed CSTR, IFAC Proc. 6 (2006) 246–251, http://dx.doi.org/10.3182/
The authors declare that they have no known competing finan- 20060710-3-it-4901.00041.
cial interests or personal relationships that could have appeared [25] A. Sinha, R.K. Mishra, Control of a nonlinear continuous stirred tank reactor
to influence the work reported in this paper. via event triggered sliding modes, Chem. Eng. Sci. 187 (2018) 52–59,
http://dx.doi.org/10.1016/j.ces.2018.04.057.
[26] J.C. Jeng, Simultaneous closed-loop tuning of cascade controllers based
References directly on set-point step-response data, J. Process Control. 24 (2014)
652–662, http://dx.doi.org/10.1016/j.jprocont.2014.03.007.
[1] B.W. Bequette, Process Control Modelling Design and Simulation, Prentice [27] M.N. Anwar, S. Pan, A frequency response model matching method for
Hall Professional, 2002, http://dx.doi.org/10.1017/CBO9781107415324.004. PID controller design for processes with dead-time, ISA Trans. 55 (2015)
[2] M. Guay, D. Dochain, M. Perrier, Adaptive extremum seeking control of 175–187, http://dx.doi.org/10.1016/j.isatra.2014.08.020.
continuous stirred tank bioreactors with unknown growth kinetics, Auto- [28] M.N. Anwar, M. Shamsuzzoha, S. Pan, A frequency domain PID controller
matica 40 (2004) 881–888, http://dx.doi.org/10.1016/j.automatica.2004.01. design method using direct synthesis approach, Arab. J. Sci. Eng. 40 (2015)
002. 995–1004, http://dx.doi.org/10.1007/s13369-015-1582-4.
[3] D. Zhao, Q. Zhu, J. Dubbeldam, Terminal sliding mode control for con- [29] M.A. Siddiqui, M.N. Anwar, S.H. Laskar, M.R. Mahboob, A unified approach
tinuous stirred tank reactor, Chem. Eng. Res. Des. 94 (2015) 266–274, to design controller in cascade control structure for unstable, integrating
http://dx.doi.org/10.1016/j.cherd.2014.08.005. and stable processes, ISA Trans. 114 (2020) 331–346, http://dx.doi.org/10.
[4] P.B. Sistu, B.W. Bequette, Nonlinear predictive control of uncertain pro- 1016/j.isatra.2020.12.038.
cesses: Application to a CSTR, AIChE J. 37 (1991) 1711–1723, http://dx.doi. [30] M.A. Siddiqui, M.N. Anwar, S.H. Laskar, Tuning of PIDF controller in parallel
org/10.1002/aic.690371114. control structure for integrating process with time delay and inverse
[5] W.A. Cebuhar, V. Costanza, Nonlinear control of CSTR’s, Chem. Eng. Sci. 39 response characteristic, J. Control. Autom. Electr. Syst. 31 (2020) 829–841,
(1984) 1715–1722, http://dx.doi.org/10.1016/0009-2509(84)80105-0. http://dx.doi.org/10.1007/s40313-020-00603-x.
[6] H. Seki, Y. Naka, Optimizing control of CSTR/distillation column processes [31] A. Raza, M.N. Anwar, Direct synthesis based pid controller design for
with one material recycle, Ind. Eng. Chem. Res. 47 (2008) 8741–8753, time-delayed unstable processes in parallel control structure, J. Control.
http://dx.doi.org/10.1021/ie800183a. Autom. Electr. Syst. 30 (2019) 879–891, http://dx.doi.org/10.1007/s40313-
[7] D.B.S. Kumar, R. Padma Sree, Tuning of IMC based PID controllers for 019-00513-7.
[32] M.A. Siddiqui, M.N. Anwar, S.H. Laskar, Enhanced control of unstable
integrating systems with time delay, ISA Trans. 63 (2016) 242–255, http:
cascade systems using direct synthesis approach, Chem. Eng. Sci. 232
//dx.doi.org/10.1016/j.isatra.2016.03.020.
(2021) 116322, http://dx.doi.org/10.1016/j.ces.2020.116322.
[8] J. Lee, W. Cho, T.F. Edgar, Simple analytic PID controller tuning rules
[33] S. Pan, J. Pal, Reduced order modelling of discrete-time systems, Appl.
revisited, Ind. Eng. Chem. Res. 53 (2014) 5038–5047, http://dx.doi.org/10.
Math. Model. 19 (1995) 133–138, http://dx.doi.org/10.1016/0307-904X(94)
1021/ie4009919.
00010-4.
[9] K.G. Begum, A.S. Rao, T.K. Radhakrishnan, Maximum sensitivity based
[34] M.N. Anwar, S. Pan, A new PID load frequency controller design method in
analytical tuning rules for PID controllers for unstable dead time processes,
frequency domain through direct synthesis approach, Int. J. Electr. Power
Chem. Eng. Res. Des. 109 (2016) 593–606, http://dx.doi.org/10.1016/j.
Energy Syst. 67 (2015) 560–569, http://dx.doi.org/10.1016/j.ijepes.2014.12.
cherd.2016.03.003.
024.
[10] R. Bacci di Capaci, M. Vaccari, C. Scali, G. Pannocchia, Enhancing MPC
[35] A. Raza, M.N. Anwar, A unified approach of PID controller design for unsta-
formulations by identification and estimation of valve stiction, J. Process
ble processes with time delay, J. Cent. South Univ. 27 (2020) 2643–2661,
Control. 81 (2019) 31–39, http://dx.doi.org/10.1016/j.jprocont.2019.05.020.
http://dx.doi.org/10.1007/s11771-020-4488-6.
[11] J. Lee, W. Cho, T.F. Edgar, Simple analytic pid controller tuning rules
[36] A. Kumar, M.N. Anwar, Decentralized load–frequency controller design
revisited, Ind. Eng. Chem. Res. 53 (2014) 5038–5047, http://dx.doi.org/10.
for a single as well as multi-area power system, Iran. J. Sci. Technol. -
1021/ie4009919.
Trans. Electr. Eng. 44 (2020) 309–326, http://dx.doi.org/10.1007/s40998-
[12] B. Vanavil, K.K. Chaitanya, A.S. Rao, Improved PID controller design for 019-00246-y.
unstable time delay processes based on direct synthesis method and [37] M.A. Siddiqui, M.N. Anwar, S.H. Laskar, Sliding mode controller design for
maximum sensitivity, Internat. J. Systems Sci. 46 (2015) 1349–1366, http: second-order unstable processes with dead-time, J. Electr. Eng. 71 (2020)
//dx.doi.org/10.1080/00207721.2013.822124. 237–245, http://dx.doi.org/10.2478/jee-2020-0032.
[13] D.F. Novella-Rodríguez, B. del M. Cuéllar, J.F. Márquez-Rubio, M.Á. [38] B.W. Bequette, Behavior of a CSTR with a recirculating jacket heat transfer
Hernández-Pérez, M. Velasco-Villa, PD–PID controller for delayed systems system, Proc. Am. Control Conf. 4 (2002) 3275–3280, http://dx.doi.org/10.
with two unstable poles: a frequency domain approach, Int. J. Control. 92 1002/9783527809080.cataz04234.
(2019) 1196–1208, http://dx.doi.org/10.1080/00207179.2017.1386326. [39] G. Armenise, M. Vaccari, R.B. Di Capaci, G. Pannocchia, An open-source
[14] L.P. Russo, B.W. Bequette, Effect of process design on the open-loop behav- system identification package for multivariable processes, in: 2018 UKACC
ior of a jacketed exothermic CSTR, Integr. Process Des. Control. Pergamon. 2 12th Int. Conf. Control (Control), 2018, pp. 152–157, http://dx.doi.org/10.
(1994) 123–128, http://dx.doi.org/10.1016/b978-0-08-042358-6.50023-4. 1109/CONTROL.2018.8516791.
[15] L.P. Russo, B.W. Bequette, Effect of process design on the open-loop [40] H. Garnier, M. Gilson, V. Laurain, B. Ni, Developments for the CONTSID tool-
behavior of a jacketed exothermic CSTR, Comput. Chem. Eng. 20 (1996) box, IFAC Proc. 45 (2012) 1553–1558, http://dx.doi.org/10.3182/20120711-
417–426, http://dx.doi.org/10.1016/0098-1354(95)00031-3. 3-BE-2027.00231.
[16] M. Chidambaram, C. Yugender, Model reference cascade control of nonlin- [41] P.C. Young, C.J. Taylor, Recent developments in the CAPTAIN toolbox
ear systems: application to an unstable CSTR, Chem. Eng. Commun. 113 for matlab, IFAC Proc. 16 (2012) 1838–1843, http://dx.doi.org/10.3182/
(1992) 15–29, http://dx.doi.org/10.1080/00986449208936001. 20120711-3-BE-2027.00096.
[17] V. Saraf, F. Zhao, B.W. Bequette, Relay autotuning of cascade-controlled [42] B. Ninness, A. Wills, A. Mills, UNIT: A freely available system identification
open-loop unstable reactors, Ind. Eng. Chem. Res. 42 (2003) 4488–4494, toolbox, Control Eng. Pract. 21 (2013) 631–644, http://dx.doi.org/10.1016/
http://dx.doi.org/10.1021/ie0011268. j.conengprac.2012.10.007.
[18] Y. Lee, S. Oh, S. Park, Enhanced control with a general cascade control [43] C.T. Chen, S.T. Peng, A sliding mode control scheme for uncertain non-
structure, Ind. Eng. Chem. Res. 41 (2002) 2679–2688, http://dx.doi.org/10. minimum phase CSTRs, J. Chem. Eng. Japan. 39 (2006) 181–196, http:
1021/ie010157f. //dx.doi.org/10.1252/jcej.39.181.
[19] T. Liu, W. Zhaang, D. Gu, IMC-based control strategy for open-loop unstable [44] J.C. Jeng, A model-free direct synthesis method for PI/PID controller design
cascade processes, Ind. Eng. Chem. Res. 44 (2005) 900–909, http://dx.doi. based on disturbance rejection, Chemom. Intell. Lab. Syst. 147 (2015)
org/10.1021/ie049203c. 14–29, http://dx.doi.org/10.1016/j.chemolab.2015.08.004.

123
M.A. Siddiqui, M.N. Anwar and S.H. Laskar Journal of Process Control 108 (2021) 112–124

[45] W. Zhang, Y. Cui, X. Ding, An Improved Analytical Tuning Rule of a [47] A. Kumar, M.N. Anwar, Parallel control structure scheme for load frequency
Robust PID Controller for Integrating Systems with Time Delay Based on controller design using direct synthesis approach, Int. J. Electr. Comput.
the Multiple Dominant Pole-Placement Method, (n.d.), http://dx.doi.org/10. Eng. 10 (2020) 47–60, http://dx.doi.org/10.11591/ijece.v10i1.pp47-60.
3390/sym12091449. [48] M. Shamsuzzoha, Closed-loop PI/PID controller tuning for stable and
[46] M.N. Anwar, S. Pan, A new PID load frequency controller design method in integrating process with time delay, Ind. Eng. Chem. Res. 52 (2013)
frequency domain through direct synthesis approach, Int. J. Electr. Power 12973–12992, http://dx.doi.org/10.1021/ie401808m.
Energy Syst. 67 (2015) 560–569, http://dx.doi.org/10.1016/j.ijepes.2014.12. [49] M. Bošković, T.B. Šekara, M.R. Rapaić, Novel tuning rules for PIDC and
024. PID load frequency controllers considering robustness and sensitivity to
measurement noise, Int. J. Electr. Power Energy Syst. 114 (2020) 105416,
http://dx.doi.org/10.1016/j.ijepes.2019.105416.

124

You might also like