Results in Control and Optimization 5 (2021) 100070

Contents lists available at ScienceDirect

Results in Control and Optimization

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

Development of tuning free SISO PID controllers for First Order Plus
Time Delay (FOPTD) and First Order Lag Plus Integral Plus Time
Delay model (FOLIPD) systems based on partial model matching
and experimental verification
Raju Yerolla, Chandra Shekar Besta ∗
Department of Chemical Engineering, National Institute of Technology Calicut, Kerala, 673601, India

ARTICLE INFO ABSTRACT

Keywords: This work aims to introduce a tuning-free PID controller for stable FOPTD and FOLIPD systems.
PID controller The objective was achieved using a partial model-matching strategy; matching the coefficients
Stable system of corresponding powers of s of the closed loop response of the model to the desired closed loop
FOPTD model
transfer function. Closed-loop performance is evaluated by considering a unit step change in the
FOLIPD model
process model and compared the results with other methods reported earlier in the literature.
Setpoint tracking
Input and output performances were evaluated using error analyses and total variation. The
robustness of the proposed method was assessed via incorporation of uncertainties in the process
model parameters. The proposed method was also experimentally validated for FOPTD model
using a temperature controlled lab. It was concluded that the method was analytically derived
and model based with a key advantage that the resultant PID required no tuning parameter.

1. Introduction

The PID (Proportional Integral Derivative) controller is employed mainly in process industry applications to provide robust
performance for stable, unstable, and nonlinear processes. PID controllers are ubiquitous in the process control industry because of
their limited number of tuning parameters and ease of implementation than advanced control [1]. More than half of the controller
strategies are still using PID control, yet finding optimal values is not easy without an effective method [2]. Mostly dynamic processes
are defined in terms of the FOPTD (First Order plus Time Delay) model for designing the PID controller.
Although many control theory developments, the PID controller is widely used in the process control industry [3]. For more
than two decades, industries have attracted IMC (Internal Model Control) based PID tuning methods due to improved performance
and a single tuning parameter. IMC-PID tuning methods [3–8], and DS (Direct Synthesis) methods [9,10] are the two prominent
examples of typical tuning methods to obtain the desired closed-loop performance. The IMC-based PID controller provides a better
setpoint response but a sluggish response to disturbances, especially in small time delay to time constant ratio processes.
In the literature, traditional PID control tuning methods include trial and error method, Ziegler–Nichols step response method,
Ziegler–Nichols frequency response method [11], Relay Tuning method [12], and Cohen–Coon method [13]. The computational
approach, often used for data modeling and cost-function optimization has been used for PID tuning. Practical implementation
strategies require optimization of the cost function.
Following the emergence of traditional PID, better performance of complex systems controller design requires new tuning control
strategies. Besharati Rad proposed a new system of auto-tuning [14]. This tuning method is fast and is used for tuning PID controllers

∗ Corresponding author.
E-mail address: [email protected] (C.S. Besta).

https://doi.org/10.1016/j.rico.2021.100070
Received 4 May 2021; Received in revised form 25 August 2021; Accepted 5 November 2021
Available online 10 November 2021
2666-7207/© 2021 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
R. Yerolla and C.S. Besta Results in Control and Optimization 5 (2021) 100070

that do not have an automatic feature. Koivo and Tanttu conducted a survey on the PID controller of SISO (Single Input Single
Output) for various configuration modes and MIMO (Multiple Input Multiple Output) [15]. Researchers have contributed to the
development of tuning techniques such as auto-tuning [16], model matching method [17], and genetic algorithm method [18]. The
assumptions made by the above methods are invested in a well-known transfer function, and PID parameters are attained by solving
a set of linear equations. L Aguirre developed an algorithm for control design based on pade approximation, matching coefficients,
and Markov parameters [18]. Jones proposes genetic algorithm techniques as an alternative to digital PID controllers [18]. This
genetic algorithm is attractive, even in digital algorithms, as the same basic method can still be easily applied to PID controllers with
complex flexible plants with highly interactive dynamics. Alberto Lava suggested a relay-based algorithm for automatic PID tuning
by defining a single point of a process frequency response [19]. When the process considered is closed to reality, the algorithm
inclines to provide satisfactory performance and accuracy. Astrom analyzed various adaptive strategies such as scheduling gains
and automated tuning [12]. This tuning strategy depends on a maximum peak resonance specification that prompts simple tuning
parameter expressions. PID tuning rules such as time domain-based design and fuzzy logic tuning were proposed by Antonio and
Visioli [20], respectively. Comino noted that PID controllers are poorly tuned, and efforts have been made to deal with the problem
systematically [21]. In his contribution that incorporates PID tuning strategies commonly utilized and to explored a few recent
effective strategies. Robust and optimum PI and PID controller tuning have been suggested [21]. Lennartson presented an analytical
design strategy and a RIMC (Robust Internal Model Control) evaluation procedure. The IMC tuning parameter change is substituted
by specific tuning parameters relevant to crucial medium frequency and higher frequency robustness characteristics [22].
The proposed method focuses on designing the PID controller design for the FOPTD and FOLIPD (First Order Lag Integrating plus
Time Delayed model) models and provides the advantage of tuning free PID controller design while being analytically derived and
model-based. The closed-loop control performance was evaluated using error analysis, total variation, and robustness by introducing
the uncertainty in the process model parameters.

2. Proposed method

Let us consider the generalized transfer function of 𝐺𝑝 (𝑠) of the process:

1
𝐺𝑝 (𝑠) = (1)
𝑔0 + 𝑔1 𝑠 + 𝑔2 𝑠2 + 𝑔3 𝑠3 + ⋯
Considered the PID controller transfer function, 𝐺𝐶 (𝑠) and given by:
𝐾𝑖
𝐺𝑐 (𝑠) = 𝐾𝑝 + + 𝐾𝑑 𝑠 (2)
𝑠
𝐾
Where 𝐾𝑃 = 𝐾𝑐 , 𝐾𝑖 = 𝜏 𝑐 and 𝐾𝑑 = 𝐾𝑐 𝜏𝑑 are PID controller settings called propositional gain, integral gain, and derivative gain.
𝑖
The closed-loop transfer function of the process, 𝐺𝑝 (𝑠) and controller 𝐺𝑐 (𝑠) is given by
𝐺𝑐 (𝑠) 𝐺𝑝 (𝑠)
𝐺𝑐𝑙 (𝑆) = (3)
1 + 𝐺𝑐 (𝑠) 𝐺𝑝 (𝑠)
The transfer function, 𝐺𝑟 (𝑠, 𝜎) of the reference model is given by:
1
𝐺𝑟 (𝑠, 𝜎) = (4)
1 + 𝑠𝜎 + 𝛼2 (𝑠𝜎)2 + 𝛼3 (𝑠𝜎)3 + 𝛼4 (𝑠𝜎)4 + ⋯
Where 𝛼2 , 𝛼3 , 𝛼4 are weighting factor, and 𝜎 is series coefficient in general reference model, and by equating Eq. (3) and Eq. (4)
then the controller 𝐺𝑐 (𝑠) is given by:
𝐺𝑟 (𝑠, 𝜎)
𝐺𝑐 (𝑠) = (5)
𝐺𝑝 (𝑠)(1 − 𝐺𝑟 (𝑠, 𝜎))
𝑔0 + 𝑔1 𝑠 + 𝑔2 𝑠2 + 𝑔3 𝑠3
𝐺𝑐 (𝑠) = (6)
s𝜎(1 + 𝛼2 (𝑠𝜎) + 𝛼3 (𝑠𝜎)2 + 𝛼4 (𝑠𝜎)3 )
Expanding the controller equation (6), by Maclaurin’s series
( ( ))
𝑔0 𝑔1 𝑔2 𝑔0 𝜎 2 𝛼3 −𝜎 2 𝛼22
𝐺𝑐 (𝑠) = 𝑠𝜎
+ 𝜎
− 𝛼2 𝑔0 − 𝑠 𝛼2 𝑔1 − 𝜎
+ 𝜎
( ( ) ( ( ) )) (7)
𝑔3 𝑔1 𝛼3 𝜎 2 −𝛼22 𝜎 2 𝑔0 𝛼2 𝜎 𝛼3 𝜎 2 −𝛼22 𝜎 2 −𝛼4 𝜎 3 +𝛼2 𝛼3 𝜎 3
−𝑠2 𝛼2 𝑔2 − 𝜎
+ 𝜎
− 𝜎
+⋯

Equating the parallel PID transfer function in Eq. (2) with Eq. (7) can get PID controller settings for the generalized reference
model.
𝑔
𝐾𝑝 = 1 − 𝛼2 𝑔0 (8)
𝜎
𝑔0
𝐾𝑖 = (9)
𝜎 ( 2 )
2
𝑔0 𝜎 𝛼3 − 𝜎 𝛼2 2
𝑔
𝐾𝑑 = −[𝛼2 𝑔1 − 2 + ] (10)
𝜎 𝜎

2
R. Yerolla and C.S. Besta Results in Control and Optimization 5 (2021) 100070

Fig. 1. By varying 𝜃∕𝜏 from 0 to 10 corresponding values of 𝜎∕𝜏 for FOPTD stable system.

( 2 2
) ( ( ) )
2
𝑔3 𝑔1 𝛼3 𝜎 − 𝛼2 𝜎 𝑔0 𝛼2 𝜎 𝛼3 𝜎 2 − 𝛼22 𝜎 2 − 𝛼4 𝜎 3 + 𝛼2 𝛼3 𝜎 3
0 = −𝛼2 𝑔2 + − + (11)
𝜎 𝜎 𝜎

2.1. The proposed method for FOPTD stable model

Consider the FOPTD stable process model

𝐾𝑒−𝜃𝑠
𝐺𝑝 (𝑠) = (12)
(1 + 𝜏𝑠)
The process model in Eq. (12) is expanded and reduced by using Maclaurin’s series, and the process model is given by
1
𝐺𝑝 (𝑠) = 2
(13)
1 0.166𝜃 3 +0.5𝜃 2 𝜏 3
𝐾
+ 𝜃+𝜏
𝐾
𝑠 + 0.5𝜃𝐾+𝜃𝜏 𝑠2 + 𝐾
𝑠 +⋯

The following relations will be obtained using the partial model matching method to the process model in Eq. (13) to the controller
Eqs. (8)–(10).
𝜃 + 𝜏 𝛼2
𝐾𝑝 = − (14)
𝐾𝜎 𝐾
1
𝐾𝑖 = (15)
𝐾𝜎
−𝛼2 𝜎 (𝜃 + 𝜏) + 0.5𝜃 2 + 𝜏𝜃 − 𝛼3 𝜎 2 + 𝛼22 𝜎 2
𝐾𝑑 = (16)
𝐾𝜎
( ) ( ) ( )
0 = 𝜎 3 2𝛼2 𝛼3 − 𝛼23 + 𝜎 2 𝛼22 − 𝛼32 (𝜏 + 𝜃) − 𝜎 0.5𝜃 2 + 𝜏𝜃 𝛼2 + 0.1667𝜃 3 + 0.5𝜏𝜃 2 (17)

By using Eq. (17), the value of 𝜎 can be calculated by taking 𝛼 (𝛼1 = 1, 𝛼2 = 0.5, 𝛼3 = 0.15, 𝛼4 = 0.03) values from the reference
model [1] by varying 𝜃∕𝜏 from 0 to 10. Corresponding values of 𝜎∕𝜏 are plotted in Fig. 1.
From Fig. 1, the value of 𝜎 = 1.3452𝜃 + 0.0504𝜏 is obtained, this value contributes to a reduction of overshoot, and substituting
these values in Eqs. (14)–(16) will get the PID controller settings for FOPTD stable model.
0.3274𝜃 + 0.9748𝜏
𝐾𝑝 = (18)
𝑘(1.3452𝜃 + 0.0504𝜏)
1
𝐾𝑖 = (19)
𝑘(1.3452𝜃 + 0.0504𝜏)
0.00835𝜃 2 − 0.0249𝜏 2 + 0.31635𝜃𝜏
𝐾𝑑 = (20)
𝑘(1.3452𝜃 + 0.0504𝜏)
Where 𝐾𝑝 , 𝐾𝑖 , and𝐾𝑑 are the PID controller setting for FOPTD stable model called proportional gain, integral gain, and derivative
gain, respectively. The limitation of the proposed method, ratio of a time delay to a time constant is up to 2.5.

3
R. Yerolla and C.S. Besta Results in Control and Optimization 5 (2021) 100070

Fig. 2. By varying 𝜃∕𝜏 from 0 to 10 corresponding values of 𝜎∕𝜏 for FOLIPD stable system.

2.2. The proposed method for the FOLIPD model

Consider the FOLIPD stable process model

𝐾𝑒−𝜃𝑠
𝐺𝑝 (𝑠) = (21)
𝑠(1 + 𝜏𝑠)
The process model in Eq. (21) is expanded and reduced by using Maclaurin’s series, and the process model is given by
1
𝐺𝑝 (𝑠) = (22)
1 0.5𝜃 2 +𝜃𝜏 3 0.1667𝜃 3 +0.5𝜃 2 𝜏 4
𝐾
𝑠 + 𝜃+𝜏
𝐾
𝑠2 + 𝐾
𝑠 + 𝐾
𝑠 +⋯
The following relations will be obtained using the partial model matching method to the process model Equation (22) to the
controller Eqs. (8)–(10).
1
𝐾𝑝 = (23)
𝐾𝜎
𝐾𝑖 = 0 (24)
−𝛼2 𝜎 + 𝜏 + 𝜃
𝐾𝑑 = (25)
𝐾𝜎
−𝛼2 (𝜏 + 𝜃) 0.5𝜃 2 + 𝜃𝜏 𝛼3 𝜎 2 𝛼 𝜎 2
0= + − + 2 (26)
𝐾 𝐾𝜎 𝐾𝜎 𝐾𝜎
By using Eq. (26), the value of 𝜎 will be calculated by taking 𝛼 (𝛼1 = 1, 𝛼2 = 0.5, 𝛼3 = 0.15, 𝛼4 = 0.03) values from the reference
model [1] by varying 𝜃∕𝜏 from 0 to 10, found the corresponding values of 𝜎∕𝜏 are plotted as shown in Fig. 2.
From Fig. 2, the value of 𝜎 = 1.4832𝜃 + 0.415𝜏 is obtained, this value contributes to a reduction of overshoot. Substituting this
value throughout Eqs. (23)–(25) will provide the PID controller settings for FOLIPD stable model as follows.
1
𝐾𝑝 = (27)
𝐾(1.4832𝜃 + 0.415𝜏)
𝐾𝑖 = 0 (28)
0.2854𝜃 + 0.7923𝜏
𝐾𝑑 = (29)
𝐾(1.4832𝜃 + 0.415𝜏)
Where 𝐾𝑝 , 𝐾𝑖 and 𝐾𝑑 are the PID controller setting and are called proportional gain, integral gain, and derivative gain.

3. Simulation results

Simulation studies are carried out using the MATLAB/Simulink 2019b package. The proposed method is focuses on dealing
with set point tracking application. Considered a unit step change to evaluate the closed loop performance, the output performance
evaluated by computing Integral Time Absolute Error, and the input performance is evaluated by computing Total Variation. The
experimental validation of the proposed PID controller design is carried out using the TC (Temperature Control) lab device. The
proposed controller settings are also working for load rejection (regulatory problem) by detuning controller settings.

4
R. Yerolla and C.S. Besta Results in Control and Optimization 5 (2021) 100070

Fig. 3. Setpoint tracking response for a unit step change in example 1. (𝑘 = 1, 𝜏 = 1, 𝜃 = 0.5).

Table 1
ITAE, TV and overshoot values for FOPTD stable system example
1. (𝑘 = 1, 𝜏 = 1, 𝜃 = 0.5).
Setpoint tracking
Method ITAE TV Overshoot
CC method 1.395 2.916 6.3
ZN method 0.870 2.284 4.0
Proposed method 0.607 1.574 1.0

Table 2
Robust Analysis for FOPTD stable system example 1. (𝑘 = 1, 𝜏 = 1, 𝜃 = 0.5).
Method Uncertainty in 𝑘𝑝 - 20% Uncertainty in 𝜏 - 20% Uncertainty in 𝜃- 20%
ITAE TV Overshoot ITAE TV Overshoot ITAE TV Overshoot
CC method 1.549 2.916 6.3 6.215 2.916 9.1 1.288 2.916 6.0
ZN method 0.867 2.284 4.2 1.390 2.284 6.5 0.891 2.284 3.8
Proposed method 0.607 1.574 1.1 0.935 1.574 2.4 0.634 1.574 1.0

3.1. Simulation results for the FOPTD model

3.1.1. Example 1
Consider FOPTD stable process model with 𝐾 = 1, 𝜏 = 1, 𝜃 = 0.5 for the proposed method, the PID controller settings are
𝐾𝑐 = 1.5747, 𝐾𝑖 = 1.3831, 𝐾𝑑 = 0.1872. The PID controller setting by Ziegler’s Nicholas method [11] are 𝐾𝑐 = 2.2842, 𝐾𝑖 = 2.6576, 𝐾𝑑 =
0.4908 and by Cohen Coon method [13] are 𝐾𝑐 = 2.9167, 𝐾𝑖 = 2.8333, 𝐾𝑑 = 0.4861. For a unit set point change in the process, the
closed loop performance is plotted in Fig. 3. The proposed method results have been compared with Ziegler’s Nicholas method, and
Cohen Coon method. The comparison results show that the proposed method gave smooth and better closed loop control performance
in terms of settling time, and overshoot, and error analysis which is listed in Table 1. Computing ITAE and TV evaluate the input
and output control performance, and these values are small for proposed method compared to other methods. The Robustness is
assessed by performing the uncertainties in the process model parameters. However, the PID controller settings used were the same.
Fig. 4 shows the 20% uncertainties in the process model parameters. The proposed method still shows better closed loop control
performance. The corresponding errors, overshoot and total variation listed in the Table 2.

3.1.2. Example 2
Consider FOPTD stable process model with 𝐾 = 1, 𝜏 = 5, 𝜃 = 1, for the proposed method, the PID controller settings are 𝐾𝑐 =
3.2566, 𝐾𝑖 = 0.6261, 𝐾𝑑 = 0.6058. The PID controller setting by Ziegler’s Nicholas method [11] are 𝐾𝑐 = 5.10, 𝐾𝑖 = 2.7456, 𝐾𝑑 = 2.3683
and PID controller settings by Cohen Coon method [13] are 𝐾𝑐 = 6.9167, 𝐾𝑖 = 3.0417, 𝐾𝑑 = 2.4269. For a unit setpoint change in the
process, the closed loop performance is plotted in Fig. 5. The proposed method results have been compared with Ziegler’s Nicholas
method, and Cohen Coon method. The comparison results show the proposed method gave smooth and better closed loop control
performance in terms of settling time, and overshoot, and error analysis which is listed in the Table 3. Computing ITAE and TV

5
R. Yerolla and C.S. Besta Results in Control and Optimization 5 (2021) 100070

Fig. 4. Setpoint tracking under 20% parameter uncertainty, for example 1. (𝑘 = 1, 𝜏 = 1, 𝜃 = 0.5). (a) Uncertainty in K (b) Uncertainty in 𝜏 (c) Uncertainty in 𝜃.

Table 3
ITAE, TV, and Overshoot values for FOPTD stable system
example 2. (𝑘 = 1, 𝜏 = 5, 𝜃 = 1).
Setpoint tracking
Method ITAE TV Overshoot
CC method 6.634 3.256 8.4
ZN method 6.303 5.10 5.5
Proposed method 3.154 6.910 0.6

evaluate the input and output control performance, and these values are small for proposed method compared to other methods.
Robustness is assessed by performing the uncertainties in the process model parameters. However, the PID controller settings used
were the same. Fig. 6 shows the 20% uncertainties in the process model parameters. The proposed method still shows better closed
loop control performance. The corresponding errors, overshoot and total variation listed in the Table 4.

3.2. Simulation results for the FOLIPD model

3.2.1. Example 3
Consider FOLIPD stable process model with 𝐾 = 1, 𝜏 = 1, 𝜃 = 4, for the proposed method, the PID controller settings are
𝐾𝑐 = 0.1575, 𝐾𝑖 = 0, 𝐾𝑑 = 0.3046. The PID controller setting by the Sree method[23] are 𝐾𝑐 = 0.2175, 𝐾𝑖 = 0.0217, 𝐾𝑑 = 0.690 and
by Hagglund method [24] were 𝐾𝑐 = 0.0938, 𝐾𝑖 = 0.0019, 𝐾𝑑 = 0.230. For a unit set point change in the process, the closed loop
performance is plotted in Fig. 7. The results of the proposed method have been compared with the Sree method and Hagglund

6
R. Yerolla and C.S. Besta Results in Control and Optimization 5 (2021) 100070

Fig. 5. Setpoint tracking response for a unit step change in example 2. (𝑘 = 1, 𝜏 = 5, 𝜃 = 1).

Table 4
Robust Analysis for FOPTD stable system example 2. (𝑘 = 1, 𝜏 = 5, 𝜃 = 1).
Method Uncertainty in 𝑘𝑝 - 20% Uncertainty in 𝜏 - 20% Uncertainty in 𝜃- 20%
ITAE TV Overshoot ITAE TV Overshoot ITAE TV Overshoot
CC method 7.342 6.916 8.4 4.437 6.916 6.6 7.759 6.916 8.7
ZN method 6.157 5.10 5.5 8.70 5.10 4.8 6.427 5.10 5.7
Proposed method 3.140 3.256 0.8 5.539 3.256 0.5 3.325 3.256 8.1

Table 5
ITAE, TV, and Overshoot values for FOLIPD stable system
example 3. (𝑘 = 1, 𝜏 = 1, 𝜃 = 4).
Setpoint tracking
Method ITAE TV Overshoot
Hagglund method 591.2 0.093 1.3
Sree method 241.5 0.217 5.9
Proposed method 39.59 0.159 0.2

Table 6
Robust Analysis for FOLIPD stable system example 3. (𝑘 = 1, 𝜏 = 1, 𝜃 = 4).
Method Uncertainty in 𝑘𝑝 - 20% Uncertainty in 𝜏 - 20% Uncertainty in 𝜃- 20%
ITAE TV Overshoot ITAE TV Overshoot ITAE TV Overshoot
Hagglund method 492.4 0.093 1.3 594.2 0.093 1.3 62.59 0.093 1.4
Sree method 199.1 0.217 7.2 251 0.217 6.4 270.1 0.217 9.0
Proposed method 42.7 0.159 1.1 43.21 0.159 0.4 62.59 0.159 1.3

method. The comparison results show that the proposed method gave smooth and better closed loop control performance in terms
of settling time, overshoot, and error analysis listed in Table 5. Computing ITAE and TV evaluate the input and output control
performance, and these values are small for proposed method compared to other methods Robustness is assessed by performing
the uncertainties in the process model parameters. However, the PID controller settings used were the same. Fig. 8 shows the
20% uncertainties in the process model parameters. The proposed method still shows better-closed loop control performance. The
corresponding errors, overshoot, and total variation listed in Table 6.

3.2.2. Example 04
Consider FOLIPD stable process model with 𝐾 = 1.2, 𝜏 = 1, 𝜃 = 4.8, for the proposed method, the PID controller settings are
𝐾𝑐 = 0.1106, 𝐾𝑖 = 0, 𝐾𝑑 = 0.2391. The PID controller setting by the Sree method [23] are 𝐾𝑐 = 0.1518, 𝐾𝑖 = 0.0126, 𝐾𝑑 = 0.5417
and PID controller setting by Hagglund method [24] are 𝐾𝑐 = 0.0650, 𝐾𝑖 = 0.0011, 𝐾𝑑 = 0.1819. For a unit set point change in the
process, the closed loop performance is plotted in Fig. 9. The proposed method results have been compared with the Sree method and
Hagglund method. The comparison results show the proposed method gave smooth and better closed loop control performance in

7
R. Yerolla and C.S. Besta Results in Control and Optimization 5 (2021) 100070

Fig. 6. Setpoint tracking under 20% parameter uncertainty, for example, 2. (𝑘 = 1, 𝜏 = 5, 𝜃 = 1). (a) Uncertainty in K (b) Uncertainty in 𝜏 (c) Uncertainty in 𝜃.

Table 7
ITAE, TV, and Overshoot values for FOLIPD stable system
example 4. (𝑘 = 1.2, 𝜏 = 1, 𝜃 = 4.8).
Setpoint tracking
Method ITAE TV Overshoot
Hagglund method 709.5 0.065 0.9
Sree method 322.6 0.151 5.8
Proposed method 54.2 0.11 0.2

terms of settling time, overshoot, and error analysis, listed in Table 7. Computing ITAE and TV evaluate the input and output control
performance, and these values are small for proposed method compared to other methods. Robustness is assessed by performing
the uncertainties in the process model parameters. However, the PID controller settings used were the same. Fig. 10 shows the
20% uncertainties in the process model parameters. The proposed method still shows better closed-loop control performance. The
corresponding errors, overshoot, and total variation listed in Table 8.

4. Experimental validation of proposed PID controller for the FOPTD model

The validation of proposed controller is assessed by using a TC lab device [25] as show in Fig. 11. TC lab is a standard hardware
benchmark for testing different controller settings and is a printed circuit board shield that connects to an Arduino microcontroller.
The proposed method’s evaluation has been carried over in six steps: dynamic step test, process model identification, model
verification, controller design, and controller testing using different methods against the proposed method. Firstly, it conducted

8
R. Yerolla and C.S. Besta Results in Control and Optimization 5 (2021) 100070

Fig. 7. Setpoint tracking response for a unit step change in example 3 (𝑘 = 1, 𝜏 = 1, 𝜃 = 4).

Table 8
Robust Analysis for FOLIPD stable system example 4. (𝑘 = 1.2, 𝜏 = 1, 𝜃 = 4.8).
Method Uncertainty in 𝑘𝑝 - 20% Uncertainty in 𝜏 - 20% Uncertainty in 𝜃- 20%
ITAE TV Overshoot ITAE TV Overshoot ITAE TV Overshoot
Hagglund method 597.9 0.065 1.1 714.1 0.065 1.0 730.2 0.65 1.3
Sree method 259.2 0.151 7.2 332.9 0.151 6.3 360.6 0.151 9.1
Proposed method 58.55 0.11 1.0 58.54 0.11 0.4 86.87 0.11 1.1

Table 9
ITAE, TV, and Overshoot values for FOPTD identified process.
(𝑘 = 6.55, 𝜏 = 140, 𝜃 = 15).
Setpoint tracking
Method ITAE TV Overshoot
CC method 1516 1.9381 9.1
ZN method 1819 1.4024 6.0
Proposed method 732.6 0.7926 0.4

Table 10
Robust Analysis for FOPTD IDENTIFIED process. (𝑘 = 6.55, 𝜏 = 140, 𝜃 = 15).
Method Uncertainty in 𝑘𝑝 - 20% Uncertainty in 𝜏 - 20% Uncertainty in 𝜃- 20%
ITAE TV Overshoot ITAE TV Overshoot ITAE TV Overshoot
CC method 6542 1.9381 15.3 1661 1.9381 7.4 13980 1.9381 14.4
ZN method 1414 1.4024 7.4 2563 1.4024 5.3 2416 1.4024 9.5
Proposed method 734.2 0.7926 1.3 1108 0.7926 0.3 1015 0.7926 1.3

a dynamic step test to determine the process parameters using [26]. Secondly, real-time implementation of different controller
settings to evaluate the effectiveness of the proposed controller.
Firstly, the standard step test has conducted on the TC Lab device [25] and found FOPTD parameters using [26] are 𝐾 =
6.55, 𝜏 = 140, 𝜃 = 15, for the present method the PID controller settings are 𝐾𝑐 = 0.7926, 𝐾𝑖 = 0.0056, 𝐾𝑑 = 0.9988. The PID controller
setting by Ziegler’s Nicholas method are 𝐾𝑐 = 1.4010, 𝐾𝑖 = 0.0486, 𝐾𝑑 = 10.1047, and PID controller setting by Cohen Coon method
are 𝐾𝑐 = 1.9381, 𝐾𝑖 = 0.0548, 𝐾𝑑 = 10.3694. For a unit set point change in the process, the closed loop performance is plotted in
Fig. 12. The proposed method results has been compared with Ziegler’s Nicholas method, and Cohen Coon method. The comparison
results shows that the proposed method gave smooth and better closed loop control performance in terms of settling time, and
overshoot, and error analysis which is listed in the Table 9. Robustness is assessed by performing the uncertainties in the process
model parameters. However, the PID controller settings used were the same. Fig. 13 shows the 20% uncertainties in the process
model parameters. The proposed method still shows better closed loop control performance. The corresponding errors, overshoot
and total variation listed in Table 10.

9
R. Yerolla and C.S. Besta Results in Control and Optimization 5 (2021) 100070

Fig. 8. Setpoint tracking under 20% parameter uncertainty for example 3. (𝑘 = 1, 𝜏 = 1, 𝜃 = 4) (a) Uncertainty in K (b) Uncertainty in 𝜏 (c) Uncertainty in 𝜃.

Fig. 9. Setpoint tracking response for a unit step change in example 4(𝑘 = 1.2, 𝜏 = 1, 𝜃 = 4.8).

10
R. Yerolla and C.S. Besta Results in Control and Optimization 5 (2021) 100070

Fig. 10. Setpoint tracking under 20% parameter uncertainty, for example 4 (𝑘 = 1.2, 𝜏 = 1, 𝜃 = 4.8). (a) Uncertainty in K (b) Uncertainty in 𝜏 (c) Uncertainty
in 𝜃.

Fig. 11. Temperature control lab setup.

11
R. Yerolla and C.S. Besta Results in Control and Optimization 5 (2021) 100070

Fig. 12. Setpoint tracking response for unit step change for the identified process (𝑘 = 6.55, 𝜏 = 140, 𝜃 = 15).

Fig. 13. Setpoint tracking under 20% parameter uncertainty for identified process (𝑘 = 6.55, 𝜏 = 140, 𝜃 = 15). (a) Uncertainty in K (b) Uncertainty in 𝜏 (c)
Uncertainty in 𝜃.

12
R. Yerolla and C.S. Besta Results in Control and Optimization 5 (2021) 100070

Fig. 14. Temperature tracking response: Real-time implementation of different controller settings on TC lab.

Fig. 15. Manipulating variable versus time behavior: Real-time implementation of different controller settings on TC lab.

Secondly, real-time implementation has been performed to evaluate the controller’s closed-loop performance and experimental
validation of proposed PID controller settings (Ziegler’s Nicholas, Cohen Coon, and Proposed). Fig. 14 shows a comparison set point
tracking response between the proposed Ziegler’s Nicholas and Cohen Coon methods by changing the temperature in a sequence.
The proposed method shows better control performance in overshoot and setting time than the other methods. Fig. 15 shows
the corresponding process behavior. Hence, by simulation and real-time implementation results, the proposed method gave the
controller’s superior closed loop performance compared with other methods reported earlier in the literature.

5. Conclusion

Tuning free PID controller settings were proposed for FOPTD and FOLIPD models for setpoint tracking applications. The closed-
loop control performance (errors, total variation, and overshoot) substantiates the proposed PID settings better than other methods
reported earlier in the literature. Post introduction of uncertainties in the process model parameters validates the proposed method’s
robustness. Experimental verification of proposed PID controller settings results validates the proposed method is robust and
preferable. The proposed method’s advantages were tuning free, explicitly uses the process model, and easy to use.

13
R. Yerolla and C.S. Besta Results in Control and Optimization 5 (2021) 100070

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared
to influence the work reported in this paper.

Acknowledgments

Authors greatly acknowledge the computational facilities and simulation tools provided through Faculty Seed Research Grant
2019, by National Institute of Technology Calicut.

Appendix. Finding 𝝈 value for FOPTD and FOLIPD systems

Finding the 𝜎 value for FOPTD system

The value of 𝜎 calculated by using 𝛼 (𝛼1 = 1, 𝛼2 = 0.5, 𝛼3 = 0.15, 𝛼4 = 0.03) values from the reference model [1].
The following equation is obtained after substituting the 𝛼 values in Eq. (17)
( )
0 = −0.005𝜎 3 + 0.1𝜎 2 (𝜏 + 𝜃) − 0.5𝜎 0.5𝜃 2 + 𝜏𝜃 + 0.1667𝜃 3 + 0.5𝜏𝜃 2

Dividing equation above on both sides with 𝜏 3

( ) ( )( ) ( ) ( ( )2 ) ( )2 ( )3
𝜎 𝜎 𝜃 𝜎 𝜃 𝜃 𝜃 𝜃
0 = −0.005 + 0.1 1+ − 0.5 0.5 + + 0.5 + 0.1667
𝜏 𝜏 𝜏 𝜏 𝜏 𝜏 𝜏 𝜏
By changing the value of 𝜃∕𝜏 from 0 to 10. Corresponding values for 𝜎∕𝜏 can be obtained. Which were plotted as show in Fig. 1.
The plotted figure in the form of straight line 𝑦 = 𝑚𝑥 + 𝑐, where 𝑦 = 𝜎𝜏 , 𝑥 = 𝜃𝜏 . Substituting the value of slop in the equation gives
𝜎 = 1.3452𝜃 + 0.0504𝜏.
Finding the 𝜎 value for FOLIPD system
The value of 𝜎 calculated by using 𝛼 (𝛼1 = 1, 𝛼2 = 0.5, 𝛼3 = 0.15, 𝛼4 = 0.03) values from the reference model [1].
The following equation is obtained after substituting the 𝛼 values in Eq. (26)

0 = 0.5𝜎 (𝜏 + 𝜃) + 0.5𝜃 2 + 𝜏𝜃 − 0.15𝜎 2 + 0.25𝜎 2

Dividing the above equation on both sides with 𝜏 2

( )2 ( ) ( )( ) ( ) ( )2
𝜎 𝜎 𝜎 𝜃 𝜎 𝜃
0 = 0.1 − 0.5 − 0.5 + 0.5
𝜏 𝜏 𝜏 𝜏 𝜏 𝜏
By changing the value of 𝜃∕𝜏 from 0 to 10. Corresponding values for 𝜎∕𝜏 can be obtained. Which were plotted as show in Fig. 2.
The plotted figure in the form of straight line 𝑦 = 𝑚𝑥 + 𝑐, where 𝑦 = 𝜎𝜏 , 𝑥 = 𝜃𝜏 . Substituting the value of slop in the equation gives
𝜎 = 1.4832𝜃 + 0.415𝜏.

References

[1] Shigemasa T, Takagi Y, Ichikawa Y, Kitomori T. A practical reference model for control system design. Trans Soc Instrum Control Eng 1983;19(7):592–4.
http://dx.doi.org/10.9746/sicetr1965.19.592.
[2] Skogestad S. Simple analytic rules for model reduction and PID controller tuning. J Process Control 2003;13(4):291–309. http://dx.doi.org/10.1016/S0959-
1524(02)00062-8.
[3] Rivera DE, Morarl M, Skogestad S. Internal model control: Pid controller design. Ind Eng Chem Process Des Dev 1986;25(1):252–65. http://dx.doi.org/
10.1021/i200032a041.
[4] Horn IG, Arulandu JR, Gombas CJ, Vanantwerp JG, Braatz RD. Improved filter design in internal model control. Ind Eng Chem Res 1996;35(10):3437–41.
http://dx.doi.org/10.1021/ie9602872.
[5] Lee Y, Park S, Lee M, Brosilow C. PID Controller tuning for desired closed-loop responses for SI/SO systems. AIChE J 1998;44(1):106–15. http:
//dx.doi.org/10.1002/aic.690440112.
[6] Shamsuzzoha M, Lee M. IMC - PID Controller design for improved disturbance rejection of time-delayed processes. Ind Eng Chem Res 2007;46(7):2077–91.
http://dx.doi.org/10.1021/ie0612360.
[7] Skogestad S. Simple analytic rules for model reduction and PID controller tuning. Model Identif Control 2004;25(2):85–120. http://dx.doi.org/10.4173/
mic.2004.2.2.
[8] Vivek S, Chidambaram M. An improved relay auto-tuning of PID controllers for critically damped SOPTD systems. Chem Eng Commun
2012;199(11):1437–62. http://dx.doi.org/10.1080/00986445.2012.668863.
[9] Chen D, Seborg DE. PI/PID controller design based on direct synthesis and disturbance rejection. Ind Eng Chem Res 2002;41(19):4807–22. http:
//dx.doi.org/10.1021/ie010756m.
[10] Martin J, Corripio AB, Smith CL. How to select controller modes and tuning parameters from simple process models. ISA Trans 1976;15(4):314–9.
[11] Ziegler JG, Nichols NB. Optimum settings for automatic controllers. ASME Trans 1942;42(6):94–100.
[12] Åström KJ, Hägglund T, Hang CC, Ho WK. Automatic tuning and adaptation for PID controllers - a survey. Control Eng Pract 1985;1(4):699–714.
http://dx.doi.org/10.1016/0967-0661(93)91394-C.
[13] Cohen GH, Coon GA. Theoretical consideration of retarded control. Trans ASME 1953;75:827–34.
[14] Besharati Rad A, Lo WL, Tsang KM. Self-tuning PID controller using Newton–Raphson search method. IEEE Trans Ind Electron 1997;44(5):717–25.
http://dx.doi.org/10.1109/41.633479.
[15] Koivo HN, Tanttu JT. Tuning of PID controllers: Survey of SISO and MIMO techniques. In: IFAC symposia series. vol. 7, 1991, p. 75–80. http:
//dx.doi.org/10.1016/s1474-6670(17)51299-9.

14
R. Yerolla and C.S. Besta Results in Control and Optimization 5 (2021) 100070

[16] Ruano AEB, Fleming PJ, Jones DI. Connectionist approach to PID auto-tuning. IEE Proc D Control Theory Appl 1992;139(3):279–85. http://dx.doi.org/
10.1049/ip-d.1992.0037.
[17] Aguirre LA. PID Tuning based on model matching. Electron Lett 1992;28(25):2269–71. http://dx.doi.org/10.1049/el:19921460.
[18] Porter B, Jones AH. Genetic tuning of digital PID controllers. Electron Lett 1992;28(9):843–4. http://dx.doi.org/10.1049/el:19920533.
[19] Leva A. Pid auto-tuning algorithm based on relay feedback. IEE Proc D Control Theory Appl 1993;140(5):328–38. http://dx.doi.org/10.1049/ip-
d.1993.0044.
[20] Visioli A. Tuning of PID controllers with fuzzy logic. IEEE Proc Control Theory Appl 2001;148(1):1–8. http://dx.doi.org/10.1049/ip-cta:20010232.
[21] Cominos P, Munro N. PID Controllers: Recent tuning methods and design to specification. IEEE Proc Control Theory Appl 2002;149(1):46–53. http:
//dx.doi.org/10.1049/ip-cta:20020103.
[22] Lennartson B, Kristiansson B. Evaluation and tuning of robust PID controllers. IET Control Theory Appl 2009;3(3):294–302. http://dx.doi.org/10.1049/iet-
cta:20060450.
[23] Sree R, MC. Control of unstable systems. Alpha Science Int’l Ltd; 2006.
[24] Astrom KJ, TH. Advanced PID controllers. Instrument Society of America; 2006.
[25] Park J, Martin RA, Kelly JD, Hedengren JD. Benchmark temperature microcontroller for process dynamics and control. Comput Chem Eng 2020;135:106736.
http://dx.doi.org/10.1016/j.compchemeng.2020.106736.
[26] Sundaresan KR, Krishnaswamy PR. Estimation of time delay time constant parameters in time, frequency, and Laplace domains. Can J Chem Eng
1978;56(2):257–62. http://dx.doi.org/10.1002/cjce.5450560215.

15