Studies in Systems, Decision and Control 264

Kishore Bingi · Rosdiazli Ibrahim ·

Mohd Noh Karsiti · Sabo Miya Hassan ·
Vivekananda Rajah Harindran

Fractional-order
Systems and PID
Controllers
Using Scilab and Curve Fitting Based
Approximation Techniques
Studies in Systems, Decision and Control

Volume 264

Series Editor
Janusz Kacprzyk, Systems Research Institute, Polish Academy of Sciences,
Warsaw, Poland
The series “Studies in Systems, Decision and Control” (SSDC) covers both new
developments and advances, as well as the state of the art, in the various areas of
broadly perceived systems, decision making and control–quickly, up to date and
with a high quality. The intent is to cover the theory, applications, and perspectives
on the state of the art and future developments relevant to systems, decision
making, control, complex processes and related areas, as embedded in the fields of
engineering, computer science, physics, economics, social and life sciences, as well
as the paradigms and methodologies behind them. The series contains monographs,
textbooks, lecture notes and edited volumes in systems, decision making and
control spanning the areas of Cyber-Physical Systems, Autonomous Systems,
Sensor Networks, Control Systems, Energy Systems, Automotive Systems,
Biological Systems, Vehicular Networking and Connected Vehicles, Aerospace
Systems, Automation, Manufacturing, Smart Grids, Nonlinear Systems, Power
Systems, Robotics, Social Systems, Economic Systems and other. Of particular
value to both the contributors and the readership are the short publication timeframe
and the world-wide distribution and exposure which enable both a wide and rapid
dissemination of research output.
** Indexing: The books of this series are submitted to ISI, SCOPUS, DBLP,
Ulrichs, MathSciNet, Current Mathematical Publications, Mathematical Reviews,
Zentralblatt Math: MetaPress and Springerlink.

More information about this series at http://www.springer.com/series/13304

Kishore Bingi Rosdiazli Ibrahim
• •

Mohd Noh Karsiti Sabo Miya Hassan

• •

Vivekananda Rajah Harindran

Fractional-order Systems
and PID Controllers
Using Scilab and Curve Fitting Based
Approximation Techniques

123
Kishore Bingi Rosdiazli Ibrahim
Institute of Autonomous Systems Department of Electrical
Universiti Teknologi PETRONAS and Electronic Engineering
Perak, Malaysia Universiti Teknologi PETRONAS
Perak, Malaysia
Mohd Noh Karsiti
Department of Electrical Sabo Miya Hassan
and Electronic Engineering Department of Electrical
Universiti Teknologi PETRONAS and Electronics Engineering
Perak, Malaysia Abubakar Tafawa Balewa University
Bauchi, Nigeria
Vivekananda Rajah Harindran
Instrumentation and Control
PETRONAS Group Technical Solutions
Petaling Jaya, Malaysia

ISSN 2198-4182 ISSN 2198-4190 (electronic)

Studies in Systems, Decision and Control
ISBN 978-3-030-33933-3 ISBN 978-3-030-33934-0 (eBook)
https://doi.org/10.1007/978-3-030-33934-0
© Springer Nature Switzerland AG 2020
This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt from
the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, expressed or implied, with respect to the material contained
herein or for any errors or omissions that may have been made. The publisher remains neutral with regard
to jurisdictional claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
This work is dedicated to:
My parents Narasimharao and Sudharani,
wife Sai Ramya Sri and brother Subbu for
their encouragement, care, love and support.
—Kishore Bingi

My beloved wife and best friend, Lidia,

and my princesses Azra, Auni, Ahna and
Ayla, my love will always be with you.
—Rosdiazli Ibrahim

My much-loved family, Wan Fatimah, Nuar

Noh, Redzwan Noh and Amirah Farhanah.
—Mohd Noh Karsiti

My beloved wife, Amina and daughters

Amatullah and Aishatu for their support
and love.
—Sabo Miya Hassan

My parents, wife and friendly colleagues

for their unfailing support.
—Vivekananda Rajah Harindran
Preface

This work aims to design and develop an improved PID control strategies based on
set-point weighting, structural modifications and fractional ordering for achieving
adequate set-point tracking and disturbance rejection through smoother control
action. In line with this, two fractional-order PID controllers, i.e., fractional-order
set-point weighted PID and fractional-order PI-PD controller for process control to
achieve adequate set-point tracking and disturbance rejection through smoother
control action has been developed. Furthermore, the performance of the proposed
fractional-order control strategies has been evaluated on real-time pH neutralization
and pressure process plants over set-point tracking and disturbance rejection for
smoother control action.
For practical realization or equivalent circuit implementation of the fractional-
order systems and controllers, a novel curve fitting-based integer-order approxima-
tion techniques for fractional-order controllers using exact frequency response data
has been developed. On the other hand, for effective implementation of
fractional-order PI/PID controllers, the Scilab-based toolbox has been developed. The
developed Scilab-based toolbox is the first toolbox for fractional-order systems
developed in open-source software. The toolbox includes the definitions of fractional-
order parameters, approximation techniques, fractional-order differentiator/
integrator, fractional-order based systems, and PI/PID controllers. The toolboxes
also allow time and frequency domain as well as stability analysis of the fractional-
order systems and controllers. The toolbox allows the analysis of the dynamic
behavior of these chaotic systems for various commensurate and non-commensurate
orders using Scilab.
Therefore, with a total of five chapters, the book is structured in such a way that
the sequential flow is maintained. Thus, after the introduction in Chaps. 1 and 2
gives the review on the PID controller and its modifications. Then, a review of these
modified PID control strategies such as SWPID, PI-PD and FOPID controllers is
provided. The chapter then provides the design of fractional-order set-point
weighted control strategies in standard, industrial, ideal and parallel configurations.
The simulation study on control of real-time pH neutralization process and pressure
control using the developed fractional-order control strategies has also been

vii
viii Preface

presented in the chapter. Chapter 3 presents the development of novel curve

fitting-based approximation algorithms for fractional-order parameters. The chapter
also presents the simulation study on various fractional order-based systems and
controllers on the performance of the proposed curve-fitting approximation tech-
niques. Chapter 4 develops the Scilab-based toolbox for fractional-order systems
and PID controllers. The toolbox includes the definitions of fractional-order
parameters, approximation techniques, fractional-order differentiator/integrator,
fractional-order based systems, and PI/PID controllers. Furthermore, the toolbox
also provides the time and frequency domain as well as stability analysis of the
fractional-order systems. The chapter then presents the simulation study on various
fractional-order based systems and controllers. Chapter 5 presents the toolbox for
fractional-order chaotic systems. The toolbox includes fractional-order Van der Pol
and duffing oscillators and fractional-order Lorenz, Chen and Rössler’s systems.
Then chapter analyzes the dynamic behavior of these chaotic systems for various
commensurate and non-commensurate orders using Scilab. In the Appendix, the
tuning of fractional-order set-point weighted controller parameters using the
accelerated particle swarm optimization algorithm is presented.
We would like to thank Dr. Thomas Ditzinger, Editorial Director of
Interdisciplinary and Applied Sciences Engineering and Prof. Dr. Janusz Kacprzyk,
Series Editor of Studies in Systems, Decision and Control for acceptance of the
publication of this book with Springer. We also want to thank the production team
of Springer, in particular, the project coordinator Ms. Saranya Kalidoss for her
professional and kind assistance.
The book was supported by Universiti Teknologi PETRONAS through the
Award of Yayasan Universiti Teknologi PETRONAS Fundamental Research
Grants with Nos 0153AA-H16 and 015LC0-053.
Last but not the least, the authors would like to express our gratitude to
Dr. Irraivan Elamvazuthi, Ms. Saranya, Mr. Shihab, Mr. Wasif Umar, Dr. Nurlidia,
Dr. Tran Duc, Mr. Sohail, Mr. Moslem, Mr. Emad, Mr. Seshu, Mr. Erman,
Ms. Anika, Mr. Fuad, Mr. Atef, Mr. Rama Krishna, Mr. Arun, Ms. Manasa,
Ms. Niharika, Mr. Malla Konda Reddy, Ms. Madiah, Ms. Fatihah, Ms. Swathi and
Mr. Arun Kumar Kakumanu for their encouragement in the research.

Seri Iskandar, Malaysia Kishore Bingi

Seri Iskandar, Malaysia Rosdiazli Ibrahim
Seri Iskandar, Malaysia Mohd Noh Karsiti
Kabale, Uganda Sabo Miya Hassan
Petaling Jaya, Malaysia Vivekananda Rajah Harindran
September 2019
Acknowledgements

This monograph is based on series of conference proceedings and articles that we

have been published during my Ph.D. study. Therefore, it has been necessary at
times to reuse some works that we previously published in various papers and
publications. Even though, in many cases, the work has been modified and
rewritten for this monograph, copyright permission from several publishers is
acknowledged as follows.
Acknowledgement is given to the Springer Science & Business Media to
reproduce material from the following papers:
• Bingi, K., Ibrahim, R., Karsiti, M. N., Hassan, S. M., Harindran, V. R.:
Real-time control of pressure plant using 2DOF fractional-order PID controller.
Arab. J. Sci. Eng. 44(3), 2091–2102 (2019)
• Bingi, K., Ibrahim, R., Karsiti, M. N., Hassan, S. M.: Fractional order set-point
weighted PID controller for pH neutralization process using accelerated PSO
algorithm. Arab. J. Sci. Eng. 43(6), 2687–2701 (2018)
Acknowledgement is given to the Polish Academy of Sciences to reproduce
material from the following paper:
• Bingi, K., Ibrahim, R., Karsiti, M. N., Hassan, S. M., Harindran, V. R.:
A comparative study of 2DOF PID and 2DOF fractional order PID controllers
on a class of unstable systems. Arch. Control Sci. 28(4), pp. 635–682 (2018)
Acknowledgement is given to the Walter de Gruyter to reproduce material from
the following paper:

ix
x Acknowledgements

• Bingi, K., Ibrahim, R., Karsiti, M. N., Hassan, S. M., Harindran, V. R.:
Frequency Response Based Curve Fitting Approximation of Fractional-order
PID Controllers. Int. J. Appl. Math. Comput. Sci, 29(2), 311–326 (2019)

Seri Iskandar, Malaysia Kishore Bingi

Seri Iskandar, Malaysia Rosdiazli Ibrahim
Seri Iskandar, Malaysia Mohd Noh Karsiti
Kabale, Uganda Sabo Miya Hassan
Petaling Jaya, Malaysia Vivekananda Rajah Harindran
September 2019
Contents

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

Part I Fractional-order Set-Point Weighted Controllers

and Approximation Techniques
2 Fractional-order Set-Point Weighted Controllers . . . . . . . . ....... 9
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ....... 9
2.1.1 Review on Modified PID Control Strategies . . . ....... 13
2.1.2 Review on Application of Fractional-order PID
Controllers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.1.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.2 Fractional-order Set-Point Weighted PID Controller . . . . . . . . . . 20
2.2.1 Standard Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.2.2 Ideal Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.2.3 Parallel Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.2.4 Industrial Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.2.5 Equivalent Configurations . . . . . . . . . . . . . . . . . . . . . . . 27
2.2.6 Parameters of Various Controllers . . . . . . . . . . . . . . . . . . 31
2.2.7 Conversion of Controller Parameters . . . . . . . . . . . . . . . . 31
2.3 Fractional-order PI-PD Controller . . . . . . . . . . . . . . . . . . . . . . . 32
2.3.1 PI-PD Controller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.3.2 PI¸ -PDl Controller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.3.3 Parameters of Various Controller Configurations . . . . . . . 37
2.3.4 Conversion of Controller Parameters . . . . . . . . . . . . . . . . 38
2.4 Simulation Study on pH Neutralization Process . . . . . . . . . . . . . 39
2.4.1 P&ID of pH Neutralization Process Plant . . . . . . . . . . . . 39
2.4.2 Modeling of pH Neutralization Process . . . . . . . . . . . . . . 42

xi
xii Contents

2.4.3 Schematics of Experimental Setup . . . . . . . . . . . . . . . . . 44

2.4.4 Performance of SWPI¸ Dl Controller . . . . . . . . . . . . . . . . 44
2.4.5 Performance of PI¸ -PDl Controller . . . . . . . . . . . . . . . . . 51
2.4.6 Comparison of SWPI¸ Dl and PI¸ -PDl Controllers . . . . . 59
2.5 Simulation Study on Pressure Process Plant . . . . . . . . . . . . . . . . 62
2.5.1 PI&D of Pressure Process Plant . . . . . . . . . . . . . . . . . . . 63
2.5.2 Schematics of Experimental Setup . . . . . . . . . . . . . . . . . 64
2.5.3 Performance of SWPI¸ Dl Controller . . . . . . . . . . . . . . . . 65
2.5.4 Performance of PI¸ -PDl Controller . . . . . . . . . . . . . . . . . 78
2.5.5 Comparison of SWPI¸ Dl and PI¸ -PDl Controllers . . . . . 87
2.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
3 Approximation Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
3.2 Curve Fitting Based Approximation Algorithm . . . . . . . . . . . . . . 102
3.3 Transfer Function Estimation Algorithm . . . . . . . . . . . . . . . . . . . 105
3.4 Numerical and Stability Analysis . . . . . . . . . . . . . . . . . . . . . . . . 105
3.5 Performance Analysis of Approximation Algorithms . . . . . . . . . . 106
3.5.1 Curve Fitting Based Approximation . . . . . . . . . . . . . . . . 107
3.5.2 Transfer Function Estimation Algorithm . . . . . . . . . . . . . 110
3.6 Approximation Table for Fractional-order Differentiator . . . . . . . 114
3.7 Simulation Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
3.7.1 Fractional-order Differentiator . . . . . . . . . . . . . . . . . . . . . 119
3.7.2 Fractional-order Integrator . . . . . . . . . . . . . . . . . . . . . . . 123
3.7.3 Fractional-order PID Controller . . . . . . . . . . . . . . . . . . . . 126
3.7.4 Fractional-order Filter . . . . . . . . . . . . . . . . . . . . . . . . . . 128
3.7.5 Fractional-order Transfer Function . . . . . . . . . . . . . . . . . 130
3.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

Part II Scilab Based Toolbox for Fractional-order Systems

and PID Controllers
4 Scilab Based Toolbox for Fractional-order Systems and PID
Controllers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
4.2 Fractional-order Differintegral Operator . . . . . . . . . . . . . . . . . . . 139
4.3 Approximation of Fractional-order Operator . . . . . . . . . . . . . . . . 140
4.3.1 Oustaloup Approximation . . . . . . . . . . . . . . . . . . . . . . . . 140
4.3.2 Refined Oustaloup Approximation . . . . . . . . . . . . . . . . . 141
4.3.3 Matsuda Approximation . . . . . . . . . . . . . . . . . . . . . . . . . 142
4.3.4 Curve Fitting Approximation Algorithm . . . . . . . . . . . . . 143
Contents xiii

4.3.5 Transfer Function Estimation . . . . . . . . . . . . . . . . . . . . . 143

4.3.6 Power Series Expansion Techniques . . . . . . . . . . . . . . . . 144
4.4 Fractional-order Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
4.4.1 Fractional-order Differentiator . . . . . . . . . . . . . . . . . . . . . 145
4.4.2 Fractional-order Integrator . . . . . . . . . . . . . . . . . . . . . . . 146
4.5 Fractional-order Transfer Function . . . . . . . . . . . . . . . . . . . . . . . 147
4.6 Fractional-order Filters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
4.6.1 Fractional-order Low-Pass Filter . . . . . . . . . . . . . . . . . . . 149
4.6.2 Fractional-order High-Pass Filter . . . . . . . . . . . . . . . . . . . 150
4.6.3 Fractional-order Band-Pass Filter . . . . . . . . . . . . . . . . . . 151
4.6.4 Fractional-order All-Pass Filter . . . . . . . . . . . . . . . . . . . . 152
4.7 Fractional-order PID Controllers . . . . . . . . . . . . . . . . . . . . . . . . 153
4.7.1 PID Controller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
4.7.2 I. Podlubny’s Fractional-order PID Controller . . . . . . . . . 153
4.7.3 M. Tenoutit’s Fractional-order PID Controller . . . . . . . . . 154
4.7.4 Luo’s Fractional-order PI Controller . . . . . . . . . . . . . . . . 155
4.7.5 IMC Based Fractional-order PID Controller . . . . . . . . . . . 156
4.7.6 Modified Fractional-order PI Controller . . . . . . . . . . . . . . 157
4.7.7 Modified Fractional-order PID Controller . . . . . . . . . . . . 158
4.8 Numerical and Stability Analysis . . . . . . . . . . . . . . . . . . . . . . . . 159
4.8.1 Time and Frequency Domain Analysis . . . . . . . . . . . . . . 159
4.8.2 Stability Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
4.9 Simulation Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
4.9.1 Fractional-order Operators and Transfer Functions . . . . . . 166
4.9.2 Fractional-order PID Controllers . . . . . . . . . . . . . . . . . . . 177
4.9.3 Fractional-order Filters . . . . . . . . . . . . . . . . . . . . . . . . . . 195
4.10 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
5 Scilab Based Toolbox for Fractional-order Chaotic Systems . . . . . . . 213
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
5.2 Fractional-order Chaotic Oscillators . . . . . . . . . . . . . . . . . . . . . . 214
5.2.1 Fractional-order Van der Pol Oscillator . . . . . . . . . . . . . . 214
5.2.2 Fractional-order Duffing Oscillator . . . . . . . . . . . . . . . . . 219
5.3 Fractional-order Lorenz’s Chaotic System . . . . . . . . . . . . . . . . . 227
5.3.1 Commensurate Orders . . . . . . . . . . . . . . . . . . . . . . . . . . 231
5.3.2 Non-commensurate Orders . . . . . . . . . . . . . . . . . . . . . . . 233
5.4 Fractional-order Chen’s Chaotic System . . . . . . . . . . . . . . . . . . . 237
5.4.1 Commensurate Orders . . . . . . . . . . . . . . . . . . . . . . . . . . 238
5.4.2 Non-commensurate Orders . . . . . . . . . . . . . . . . . . . . . . . 240
5.5 Fractional-order Rössler’s Chaotic System . . . . . . . . . . . . . . . . . 241
5.5.1 Commensurate Orders . . . . . . . . . . . . . . . . . . . . . . . . . . 242
5.5.2 Non-commensurate Orders . . . . . . . . . . . . . . . . . . . . . . . 242
xiv Contents

5.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245

Appendix: Tuning of Controller Parameters Using Accelerated

Particle Swarm Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . 247
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
Acronyms

AG Agitator
AI Analog Input
AIC pH Internal Controller
AO Analog Output
APSO Accelerated Particle Swarm Optimisation
AT pH Sensor
CFE Continued Fraction Expansions
CRONE Commande Robuste d’Ordre Non Entier
CSTR Continuous Stirred-tank Reactor
CT Conductivity Transmitter
DISO Dual-input Single-output
EnPID Enhanced PID
FCV Flow Control Valve
FIR Finite Impulse Response
FIS Fuzzy Inference System
FIT Fractional Integration Tool
FOC Fractional-order Controller
FOD Fractional-order Differentiator
FOI Fractional-order Integrator
FOMCON Fractional Order Modeling and Control
FOPDT First-order Plus Dead Time
FOS Fractional-order Systems
FOTF Fractional-order Transfer Function
FPGA Field Programmable Gate Array
FSST Fractional State Space Tool
FT Flow Transmitter
GL Grünwal-Letnikov
HV Hand Valve
IAE Integral Absolute Error
IIR Infinite Impulse Response

xv
xvi Acronyms

IMC Internal Model Control

ISE Integral Squared Error
ITAE Integral Time Absolute Error
ITSE Integral Time Squared Error
LS Level Sensor
MAE Mean Absolute Error
MATLAB Matrix Laboratory
MPC Model Predictive Control
MSE Mean Squared Error
Ninteger Non-Integer
NPID Non-linear PID
P&ID Process and Instrumentation Diagram
PCI Peripheral Component Interconnect
PCV Process Control Valve
PIC Pressure Indicating Controller
PI¸ Dl Fractional-order PID
PID Proportional Integral Derivative
PIDsd PID-deadtime
PI¸ -PDl Fractional-order PI-PD
PI-PD Proportional Integral—Proportional Derivative
PPI Predictive PI
PSE Power Series Expansion
PSO Particle Swarm Optimisation
PT Pressure Transmitter
RL Riemann-Liouville
RMSE Root Mean Square Error
Scilab Scientific Laboratory
SISO Single-input Single-output
SK Sanathanan-Koerner
SWPI¸ Dl Fractional-order Set-point Weighted PID
SWPID Set-point Weighted PID
TID Tilt-Integral-Derivative
TIOO Two-input One-output
VE Vessel Tank
ZN Ziegler-Nichols
Chapter 1
Introduction

1.1 Introduction

Proportional-Integral-Derivative (PID) controllers are the most widely used con-

troller in industry [1]. The PID controller is a three action controller that has a long
history in the field of process and automation [2]. The three actions Proportional
(P), Integral (I) and Derivative (D) are based on present (P), past (I) and future (D)
control errors. These actions are defined as follows [3]:

• P—Proportional action is proportional to the current/present control error,

• I—Integral action is proportional to the integral of past values of control error and
• D—Derivative action is proportional to the derivative of future values of control
error.

The conventional PID controllers are applicable to many control problems, and
often achieve adequate performance without any improvement and can perform
poorly in some control applications [1]. The basic problem with PID control is that
with constant parameters and no direct knowledge of the overall process and thus
performance is compromised [4]. Another problem with the PID controllers is that
they are linear and in particular symmetric. Thus, the performance of PID in non-
linear systems is variable [3]. In the PID controller, the proportional and derivative
actions in the forward paths cause rapid alteration of the control signal during the
set-point change. These effects are called proportional and derivative kick effects
[5]. Furthermore, the control action of these conventional PID controllers is quite
oscillatory which degrades the actuators faster. Therefore, these controllers need to
be modified for process control applications.
On the other hand, the fractional calculus is an emerging field in mathematics with
deep applications in the areas of engineering, physics, finance, chemistry, and social
sciences [6]. In the last two decades, applications of fractional calculus in control
of engineering problems has attracted by many researchers [7]. This is because
fractional-order controllers (FOCs) based on fractional calculus methods are less
sensitive to parameter changes in the controller and the controlled systems [8, 9].
At the very beginning of differential calculus, in 1695 Leibniz raised a question
to L’Hôpital as “Can the meaning of derivatives with integer-order be generalized
© Springer Nature Switzerland AG 2020 1
K. Bingi et al., Fractional-order Systems and PID Controllers,
Studies in Systems, Decision and Control 264,
https://doi.org/10.1007/978-3-030-33934-0_1
2 1 Introduction

to derivatives with non-integer-orders?” L’Hôpital was replied by another question

to Leibniz: “What if the order will be 21 ?” then Leibniz replied as “It will lead to
a paradox, from which one-day useful consequences will be drawn.” The question
raised by Leibniz for a non-integer-order derivative was an ongoing topic for more
than 300 years, and now it is known as fractional calculus [10, 11]. Fractional
calculus is a generalization of ordinary differentiation and integration to arbitrary
non-integer-order. In differential calculus, the nth-order differentiation of f (x) is
denoted as d dfx(x)
n
n where n is a positive integer. Thus, the extension of this theory
α
to fractional calculus where αth-order differentiation of f (x) is denoted as d d fx(x)
α .

Here, α could be rational or irrational, positive or negative, real or complex [12, 13].
Until recent times, fractional calculus was considered as a mathematical theory
without any applications, but in the last few decades, there has been an explosion of
research activities on the application of fractional calculus to very diverse scientific
fields [6, 8, 13]. These applications include fractional-order control of engineering
systems, optimal control of fractional-order dynamic systems, fundamental-order
modeling of mechanical, electrical, thermal, biomedical, signal and image processing
and other engineering and scientific applications [7, 9].
The fractional-order PID (PIλ Dμ ) controller is an extension of conventional PID
which has been achieved by fractional-ordering the integral and derivative actions.
Thus, the controller has two more parameters than the conventional PID controller.
Therefore, two more specifications to be met, which improves the performance of the
overall system can be improved. The controller has been extensively used in many
engineering control applications to achieve more robust and stable performance [8,
14–16]. This is because fractional-order controllers are less sensitive to parameter
changes and can easily attain the property of iso-damping; a desirable property of
the system referring to a state where the open-loop phase Bode plot is flat [8, 11].
The PIλ Dμ controller achieves better performance compared to conventional PID
in the following situations/applications:
• Performs betters and achieves adequate tracking response for systems like
– Higher-order systems [17],
– Highly nonlinear systems [18],
– High time-delay processes [19] and
– Non-minimum phase systems [8].
• Achieves more robust and stable performance [20, 21].
• Attain the property of iso-damping easily.
Over the years many other modifications of PID has been proposed by researchers.
Among these variants, SWPID, PI-PD and PIλ Dμ received the most attention in
the area of process control [3, 10, 22–24]. This is because, compared to conven-
tional PIDs, the SWPIDs and PI-PDs have the advantage of handling both set-point
tracking and load regulation separately while the fractional-order controllers provide
more robust and stable control performance. However, a key issue with the practi-
cal realization or equivalent circuit implementation of fractional-order controllers
1.1 Introduction 3

in a finite-dimensional integer-order system is the approximation of fractional-order

parameters for the desired frequency range and the order of approximation [13, 25].
For an effective approximation of fractional-order integrator/differentiator,
researchers have proposed several frequency-domain approximation techniques such
as Oustaloup, Matsuda, Carlson, continued fractional and power series expansion
techniques [11, 14, 26]. Among the frequency domain approximation techniques,
Outstaloup’s approximation is the most popular and widely used. However, in cer-
tain situations, the Oustaloup’s approximation cannot fit around the frequency range
of interest. Moreover, the modified Oustaloup produces higher-order transfer func-
tion. Similarly, the power series expansion techniques produces higher-order transfer
function [10, 27]. On the other hand, the fitting range of continued fractional expan-
sion techniques are very small and the quality of fitting is unsatisfactory [27]. Other
approximation algorithms based on stability boundary locus, vector fitting method,
time moments approach and frequency distribution mode are quite complex hence
difficult to implement. Thus, a novel curve fitting approximation is needed that can
be easy to implement and fit the entire desired frequency range.
For the implementation or practical realization of fractional-order controllers,
researchers have developed few toolboxes such as CRONE, Ninteger, FOMCON [6,
8, 9, 13, 25, 28]. A common feature of these toolboxes is that they are based on
MATLAB/Simulink software. A key disadvantage of these toolboxes is that there
is no direct option to approximate higher-order systems. Moreover, the available
techniques are mostly limited to Outsaloup and refined Oustaloup approximation
approaches. Another improvement needed on the toolboxes is in the graphical,
numerical and stability analysis. It should be noted that MATLAB is a powerful
software that provides many numerical computations. However, despite many advan-
tages, the software is expensive and generally have limited licenses in terms of time
and number of installations [29]. On the other hand, Scilab is extremely effective
and available as a free open-source software package for scientific computations. It
includes hundreds of general and specialized functions for numerical computations,
2D and 3D visualization and data analysis [30–33].
Therefore, research into fractional-order PID control strategies, approximation
techniques and open-source software toolboxes for real-time process control, with
the view to mitigating the proportional and derivative effects for achieving adequate
set-point point tracking and disturbance rejection through smoother control action is
inevitable.
Thus, this book will focus on the development of fractional-order based con-
trol strategies for real-time process control to achieve adequate set-point tracking
and disturbance rejection performance through smoother control action. The con-
trol methodologies to be developed will be based on set-point weighting, structural
modification, and fractional-ordering of PID controllers. In this book, an attempt
has been made for the real-time implementation of proposed fractional-order con-
trollers for the control of pH neutralization and pressure process plants. The book
will further concentrate on developing the integer-order approximation techniques
based on curve fitting approaches using frequency response data. This book will
also concentrate on developing Scilab based toolboxes for the implementation of
4 1 Introduction

the fractional-order PID controllers and systems. Lastly, a Scilab based toolbox for
fractional-order oscillators and chaotic systems has also been developed.

1.2 Summary

In this chapter, to simulate the interest to the readers, a background on conventional

PID controllers, fractional calculus, fractional-order PID controllers, approximation
techniques and toolboxes for implementation of fractional-order controllers has been
presented. Finally, the scope of this book has been provided.

References

1. Ang, K.H., Chong, G., Li, Y.: PID control system analysis, design, and technology. IEEE Trans.
Control Syst. Technol. 13(4), 559–576 (2005)
2. Åström, K.J., Hägglund, T.: Advanced PID control. ISA-The Instrumentation, Systems, and
Automation Society (2006)
3. Vilanova, R., Visioli, A.: PID Control in the Third Millennium. Springer (2012)
4. Visioli, A.: Practical PID Control. Springer Science & Business Media (2006)
5. Alfaro, V.M., Vilanova, R.: Model-reference robust tuning of 2DoF PI controllers for first-and
second-order plus dead-time controlled processes. J. Process Control 22(2), 359–374 (2012)
6. Machado, J.T., Kiryakova, V., Mainardi, F.: Recent history of fractional calculus. Commun.
Nonlinear Sci. Numer. Simul. 16(3), 1140–1153 (2011)
7. Ortigueira, M.D.: Fractional Calculus for Scientists and Engineers, vol. 84. Springer Science
& Business Media (2011)
8. Shah, P., Agashe, S.: Review of fractional PID controller. Mechatronics 38, 29–41 (2016)
9. De Oliveira, E.C., Tenreiro Machado, J.A.: A review of definitions for fractional derivatives
and integral. Math. Probl. Eng. (2014)
10. Caponetto, R.: Fractional Order Systems: Modeling and Control Applications. World Scientific
(2010)
11. Monje, C.A., Chen, Y., Vinagre, B.M., Xue, D., Feliu-Batlle, V.: Fractional-Order Systems and
Controls: Fundamentals and Applications. Springer Science & Business Media (2010)
12. Xue, D., Chen, Y., Atherton, D.P.: Linear Feedback Control: Analysis and Design with MAT-
LAB. Siam (2007)
13. Krishna, B.T.: Studies on fractional order differentiators and integrators: a survey. Signal Pro-
cess. 91(3), 386–426 (2011)
14. Li, H., Luo, Y., Chen, Y.: A fractional order proportional and derivative (FOPD) motion con-
troller: tuning rule and experiments. IEEE Trans. Control Syst. Technol. 18(2), 516–520 (2009)
15. Padula, F., Visioli, A.: Tuning rules for optimal PID and fractional-order PID controllers. J.
Process Control. 21(1), 69–81 (2011)
16. Sharma, R., Rana, K.P.S., Kumar, V.: Performance analysis of fractional order fuzzy PID
controllers applied to a robotic manipulator. Expert Syst. Appl. 41(9), 4274–4289 (2014)
17. Das, S., Saha, S., Das, S., Gupta, A.: On the selection of tuning methodology of FOPID
controllers for the control of higher order processes. ISA Trans. 50(3), 376–388 (2011)
18. Petráš, I.: Fractional-Order Nonlinear Systems: Modeling, Analysis and Simulation. Springer
Science & Business Media (2011)
19. Luo, Y., Chen, Y.: Stabilizing and robust fractional order PI controller synthesis for first order
plus time delay systems. Automatica 48(9), 2159–2167 (2012)
References 5

20. Margarita, R., Sergei, V.R., José, A.T.M., Juan, J.T.: Stability of fractional order systems. Math.
Probl. Eng. (2013)
21. Tavazoei, M.S., Haeri, M.: A note on the stability of fractional order systems. Math. Comput.
Simul. 79(5), 1566–1576 (2009)
22. Mudi, R.K., Dey, C.: Performance improvement of PI controllers through dynamic set-point
weighting. ISA Trans. 50(2), 220–230 (2011)
23. Sahu, R.K., Panda, S., Rout, U.K., Sahoo, D.K.: Teaching learning based optimization algorithm
for automatic generation control of power system using 2-DOF PID controller. Int. J. Electr.
Power Energy Syst. 77, 287–301 (2016)
24. Zou, H., Li, H.: Improved PI-PD control design using predictive functional optimization for
temperature model of a fluidized catalytic cracking unit. ISA Trans. 67, 215–221 (2017)
25. Li, Z., Liu, L., Dehghan, S., Chen, Y., Xue, D.: A review and evaluation of numerical tools for
fractional calculus and fractional order controls. Int. J. Control. 90(6), 1165–1181 (2017)
26. Tepljakov, A., Petlenkov, E., Belikov, J.: Application of Newton’s method to analog and dig-
ital realization of fractional-order controllers. Int. J. Microelectron. Comput. Sci. 3(2), 45–52
(2012)
27. Valério, D., Trujillo, J.J., Rivero, M., Machado, J.T., Baleanu, D.: Fractional calculus: a survey
of useful formulas. Eur. Phys. J. Spec. Top. 222(8), 1827–1846 (2013)
28. Freeborn, T.J.: A survey of fractional-order circuit models for biology and biomedicine. IEEE
J. Emerg. Sel. Top. Circuits 3(3), 416–424 (2013)
29. Sohal, J.S.: Improvement of artificial neural network based character recognition system, using
SciLab. Optik 127(22), 10510–10518 (2016)
30. Campbell, S.L., Chancelier, J.P., Nikoukhah, R.: Modeling and Simulation in SCILAB.
Springer, New York (2006)
31. Bunks, C., Chancelier, J.P., Delebecque, F., Goursat, M., Nikoukhah, R., Steer, S.: Engineering
and Scientific Computing with Scilab. Springer Science & Business Media (2012)
32. Magyar, Z., Žáková, K.: Scilab based remote control of experiments. IFAC Proc. Vol. 45(11),
206–211 (2012)
33. Rohit, M.T., Ashish, M.K.: Digital Image Processing Using SCILAB. Springer, Cham (2018)
 Part I
Fractional-order Set-Point Weighted
Controllers and Approximation Techniques
Chapter 2
Fractional-order Set-Point Weighted
Controllers

2.1 Introduction

The advantage of the PID controller is that it can deal with important practical issues
such as actuator saturation, integrator windup. The controller is the most widely used
in the process industry for the control of temperature, flow, pressure, force, position,
speed, concentration and pH [1–5]. In fact, based on the recent survey published in
IEEE control system magazine [6], it is clear that the PID controllers are among the
top choice of control technologies in terms of industrial impact as compared to model
predictive, nonlinear, adaptive, robust, hybrid, decentralized, discrete and intelligent
control strategies. The summary of the survey is shown in Fig. 2.1. The survey results
from the figure show that the PID controller has the highest 100% industrial-impact
compared to all other control strategies.
This is because of its advantages of simplicity in design, ease of tuning and
implementation [7–10]. However, the instantaneous step change in the set-point
with a PID controller produces an excessive change in the controller output. This is
caused by the derivative action since it takes the derivative of the error, which is very
large during step change. Hence, the overall error will be very high. Furthermore, for
disturbance rejection, a fast response will be achieved with a high-gain PID controller
but with an oscillatory response [11–14]. On the other hand, the controller also
produces undesired oscillations and rapid changes in the control signal. Therefore,
reducing these undesired oscillations of a control signal is another common control
objective [15]. This is because the undesired oscillations in the control signals degrade
the actuators faster.
Over time, many modifications of PID has been proposed by various researchers.
This is due to the failure of conventional PID structure to achieve robust performance
under conditions such as the change in process dynamics, variation in set-point, high
external disturbance, long deadtime. Among the various modifications of PID are Set-
point Weighted PID (SWPID), Predictive PI (PPI), Fractional-order PID (FOPID or
PIλ Dμ ), Non-linear PID (NPID), PI-PD, Enhanced PID (EnPID) and PID-deadtime

© Springer Nature Switzerland AG 2020 9

K. Bingi et al., Fractional-order Systems and PID Controllers,
Studies in Systems, Decision and Control 264,
https://doi.org/10.1007/978-3-030-33934-0_2
10 2 Fractional-order Set-Point Weighted Controllers

Percentage of Industrial-impact 100

High-impact
Low-impact
80

60

40

20

0
PID Control

Model Predictive Control

System Identification

Process Data Analytics

Soft Sensing

Fault Detection and Identification

Decentralized Control

Intelligent Control

Discrete-event systems

Nonlinear Control

Adaptive Control

Robust Control

Hybrid Dynamical Systems

Fig. 2.1 Industrial impact of low-level and advanced control strategies

(PIDτd ) controller. The features and limitations of these modified PID controllers
are given in Table 2.1.
Of these PID variants from the table, the SWPID, PI-PD and PIλ Dμ received the
most attention recently in the area of process control. This is because, compared to
conventional PIDs, the SWPIDs and PI-PDs have the advantage of handling both
set-point tracking and load regulation separately. On the other hand, the fractional-
order controllers provide more robust and stable control performance. This is because
fractional-order PID is more flexible and less sensitive to parameter changes in the
controller and the systems to be controlled [7, 16]. For example, consider the con-
ventional integrator and differentiator in PID as s1n and s n , n ∈ R+ respectively, the
positive effects of these control actions in a closed-loop system are defined as follows:
• the integral action eliminates the steady-state errors which can be deduced by an
infinite gain at zero frequency and
• the derivative action increases the relative stability which is achieved by introduc-
ing π2 phase lead.
On the other hand, the negative effects of these control actions are as follows:
• the integral action decreases the relative stability by introducing π2 phase lag and
• the derivative action increases the sensitivity to high-frequency noise by increasing
gain with the slope of 20 dB/dec.
2.1 Introduction 11

Table 2.1 Features and limitations of several modified PID controllers

Controller Features Limitations References
SWPID Capable of achieving fast Produces undesired [10, 11, 17–27]
disturbance rejection without oscillations and rapid changes
significant increase in in control signal also increases
overshoot during set-point the number of controller
tracking. Also, reduces both parameters to the conventional
proportional and derivative PID
kick effects
PPI Capable of time Unable to achieve robust [14, 28–30]
delay/deadtime compensation performance in the presence of
even for long time delay high external disturbance and
process noise
PI-PD Capable of avoiding derivative Produces oscillatory control [31–41]
kick effect during set-point signal, high overshoot during
change and has the same set-point tracking and slow
number of controller disturbance rejection
parameters as that of
conventional PID
NPID High robustness against Causes the problems of [42–45]
uncertainties, delays and noise. integral windup, stability and
Thus, achieving fast set-point high noise rejection. Also,
tracking response and prevent doubles the number of
oscillatory behavior controller parameters as that of
conventional PID
PIDτd Capable of having deadtime Failed to achieve robust [28, 29]
compensation with performance in the presence of
functionality similar to model high external disturbance
based controllers. Also,
achieves satisfactory
performance over disturbance
rejection
PIλ Dμ More robust and stable Produces high overshoot [7, 46, 59, 68]
performance even for during set-point tracking and
higher-order systems change in set-point which is
characterized by high caused by kick effects. Also,
non-linearities and delays. difficulties may raise on
Also, attains the property of approximation and
iso-dampinga easily implementation of controllers
a A desirable property of the system referring to a state where the open-loop phase Bode plot is

flat [7]

From these effects, it can be concluded that satisfactory performance can be

achieved by combining and compromising between these effects. Furthermore, it can
also be noted that the gain and phase of integrator/differentiator are only in the orders
of 20 dB/dec and π2 . Thus, more powerful and flexible designs than the conventional
can be achieved with the use of fractional-order integrator and differentiator. With
fractional-order integrator ( s1α ) and differentiator (s α ), α ∈ R, the gain and phase
can be achieved in the orders of α×20 dB/dec and απ 2
respectively. An example of
12 2 Fractional-order Set-Point Weighted Controllers

Bode Diagram
100

Magnitude (dB)
50

0
= 0.1
-50 = 0.2
= 0.3
= 0.4
-100
90 = 0.5
= 0.6
= 0.7
Phase (deg)

= 0.8
= 0.9
45
=1

0
10-3 10-2 10-1 100 101 102 103
Frequency (rad/s)

Fig. 2.2 Bode plot of fractional-order differentiator for orders 0 < α < 1

Bode Diagram
100
Magnitude (dB)

50

0
= 0.1
-50 = 0.2
= 0.3
= 0.4
-100 = 0.5
0
= 0.6
= 0.7
Phase (deg)

= 0.8
= 0.9
-45 =1

-90
10-3 10-2 10-1 100 101 102 103
Frequency (rad/s)

Fig. 2.3 Bode plot of fractional-order integrator for orders 0 < α < 1

such flexibility with fractional-order differentiator and integrator for 0 < α < 1 are
shown in Figs. 2.2 and 2.3 respectively. From the figures, it can also be observed that
for α = 1, the characteristics of the fractional-order differentiator and integrator is
the same as the conventional differentiator and integrator.
2.1 Introduction 13

2.1.1 Review on Modified PID Control Strategies

This section reviews three modified or improved configurations of the conventional

PID controller. This section is divided into three subsections. In the first subsection,
the review on SWPID control strategy is presented. Then, the review on the PI-PD
controller followed by the review on FOPID controller is presented.

2.1.1.1 Set-Point Weighted PID Controller

The set-point weighing strategy for PID controller has been attempted by many
researchers [10, 17, 21, 23, 25, 47, 48]. Generally, these strategies focus on achieving
an adequate set-point tracking performance while maintaining the simplicity of the
PID controller. The other feature of this control strategy is that it can reduce the
derivative kick effect during the set-point change. The derivative kick effect is the
spike in the control signal due to an abrupt change of the set-point.
On the other hand, the implementation of the control strategy can also be done
in two ways i.e., single-input-single-output (SISO) and dual-input-single-output
(DISO) configurations. Thus, maintaining the advantage of PIDs easy of implemen-
tation. Furthermore, based on the set-point weighting parameters (b and c) of various
controller structures can be achieved as given in Table 2.2. On the set-point weighing
strategies, for instance, the strategy reported in [21] is a component separated type
SISO configuration. The developed control strategy has been implemented for a class
of unstable time-delay processes including real-time processes like isothermal con-
tinuous stirred-tank reactor (CSTR) and CSTR with non-ideal mixing. Furthermore,
the controller parameters including the set-point weighting parameters were tuned
using heuristic optimization algorithms. The problem with this method is that it does
not solve the problem of derivative kick effect during the set-point change. Further-
more, the researcher has not evaluated the performance over variation in set-point.
Another disadvantage of this method is that the control signal rapidly changes during

Table 2.2 Parameters of controllers derived from SWPID controller

Controller Set-point weighting parameters
Category Type b c
PID P 1 0
PD 1 1
PI 1 0
PID 1 1
SWPID SWPD 1 0<c<1
SWPI 0<b<1 1
SWPID 0<b<1 0<c<1
SWPI –D 0<b<1 0
14 2 Fractional-order Set-Point Weighted Controllers

transient response which causes the high overshoot and oscillatory response in the
behavior.
To avoid the problem of derivative kick effect, the authors of [13] has proposed
the selection of derivative set-point weighting (c) as either 0 or 1. Thus, to avoid
the extreme instantaneous change in the controller output during set-point change
c is usually set to zero. With the use of this method, the same authors in [48–51]
has developed robust tuning algorithms based on maximum sensitivity function and
model-reference optimization procedures. The developed tuning procedures have
been validated for a simulation study on stable and unstable time-delay processes
and integrating processes. However, for practical use, the control action has quite
oscillatory behavior.
In [11, 18], the use of dynamic set-point weighting for PI controller has been
proposed based on instantaneous process trend and immersion invariance frame-
work respectively. The effectiveness of the proposed techniques was evaluated in the
simulation environment for high-order processes. However, both the techniques are
applicable only for the PI controller. Thus, neglecting the issue of derivative kick
effect. On the other hand, the authors of [8, 10] proposed the conversion of controller
parameters between various controller structures. Furthermore, the use of the fuzzy
algorithm for tuning of controller parameters including set-point weighting has been
proposed in [22].
For the implementation of SWPID controller, various equivalent configurations
such as pre-filter, feedforward, feedback type have been proposed in [52]. A com-
pletely different type of structure based on feedforward configuration has been pro-
posed for magnetic levitation system in [19]. However, the proposed configuration
has failed to achieve adequate disturbance rejection performance.

2.1.1.2 PI-PD Controller

The PI-PD controller has been used for process control due to excellent control
performance and robustness [33, 34, 36]. The advantage of this control strategy is
that the inner PD feedback controller converts the integrating or unstable plant to a
stable open-loop system. Then, the cascaded PI controller controls the overall stable
process [14]. Another advantage of the controller is that it has the same number of
parameters as that of PID. Furthermore, the controller has the capability of reducing
derivative kick effect without producing high overshoot during the set-point change.
Research on improvement of PI-PD controller performance has been going on
for some time. For instance, in [32], the author proposed a model-based design of
PI-PD controller for the control of unstable and integrating processes. Moreover,
the controller parameters were obtained using the relay-based auto-tuning approach
reported by the same author in [31]. The extension of the developed control strategy
for stable processes has been reported in [33]. A similar type of relay-based PI-PD
controller structure is also reported in [35, 36]. However, the proposed methods
are only applicable to model-based approaches. Furthermore, in all the works the
2.1 Introduction 15

implementation is done only in the simulation environment and also not evaluated
for variation in set-point.
In [34], an improved version of the PI-PD controller has been proposed using
a predictive functional optimization algorithm. The control strategy achieved bet-
ter performance over PID and conventional PI-PD. However, the major industrial
constraints on the algorithm like output range must be within the limits and the
inputs boundary range should be based on the output demand are not considered.
In a related development, the authors of [39] implemented the cuckoo optimization
algorithm based PI-PD controller for the control of stable and unstable systems.
Similarly, the authors of [37, 38] implemented the cascaded PI-PD controller
for automatic generation control of power systems. In both, the parameters of the
controller were tuned using fractal search, pattern search, and grey wolf optimiza-
tion techniques. In all these reported works, the PI-PD controller achieved better
performance over conventional PID but produces an oscillatory control action. Fur-
thermore, there is no reported work on robust performance analysis of the developed
control strategy in the real-time environment.

2.1.1.3 Fractional-order PID Controller

In last two decades, there is a continuous growth in the design and development
of fractional calculus for feedback control problems as shown in Fig. 2.4 [7, 16,
53–55]. From the figure, it can be seen that over the years, researchers are devel-
oping several fractional-order forms of conventional systems and controllers. For
example, Igor Podlubny was the first person reported on fractional-order systems
and fractional-order PID controller (FOPID or PIλ Dμ ). Here, the fractional-order
form of the conventional PID has achieved by fractional-ordering the integral and
derivative actions with λ and μ respectively.
Thus, the control action of fractional-order PID (PIλ Dμ ) from conventional PID
is obtained by fractional ordering the integral and derivative actions with λ and μ as
follows:  
Ki Kd sμ
U (s) = K p + λ + E(s), 0 < λ, μ < 1 (2.1)
s α Kd sμ + 1

where
• λ is the order of integration and
• μ is the order of differentiation.
Similarly, the control actions of standard and industrial configuration of PIλ Dμ
controller with fractional-order integral and derivative actions is given as
 
1 Td s μ
U (s) = K p 1 + + , 0 < λ, μ < 1 (2.2)
Ti s λ Td s μ + 1
16 2 Fractional-order Set-Point Weighted Controllers

Fig. 2.4 Evolution of fractional-order based controllers and other related techniques

  
1 Td s μ + 1
U (s) = K p 1 + , 0 < λ, μ < 1 (2.3)
Ti s λ αTd s μ + 1

Observing (2.1), (2.2) and (2.3), it can be noted that the PIλ Dμ controller has
two more parameters (i.e., λ and μ) than the conventional PID. Therefore, optimal
selection of K p , K i , K d , λ and μ in (2.1) will improve the overall system performance.
This is because the fractional-order controllers are less sensitive to parameter changes
of the controller and the system to be controlled. Furthermore, the controller also
attain the property of iso-damping easily.
2.1 Introduction 17

The fractional-order form of several advanced controllers such as sliding mode

control, optimal control, model reference adaptive control and model predictive con-
trol has been proposed by researchers. Another important component in the feed-
back loop is the noise filter. The performance of conventional filters can also be
improved with the use of fractional-order filters. Observing all the fractional-order
based controllers and filters, it can be noted that the development of these fractional-
order controllers (FOC) is only happened because of the numerical solutions for
fractional-order derivatives defined by Grünwald-Letnikov, Riemann-Liouville, and
Caputo [56, 57]. However, a key issue with the practical realization or equivalent
circuit implementation of FOCs in a finite-dimensional integer-order system is the
approximation of the fractional-order operators (fractional-order differentiator and
integrator) for the desired frequency range. Over the years, researchers proposed sev-
eral approximation techniques including continued fraction, power series expansions
and curve fitting approaches [57–61, 68].
The researchers also proposed several MATLAB and Simulink based toolboxes
for the implementation of fractional-order controllers. These toolboxes include
CRONE, Non-Integer (NINTEGER), fractional state space tool (FSST) and Frac-
tional Order Modeling and Control (FOMCON) [53]. Apart from these toolboxes,
several fractional-order functions also available from MATLAB central file exchange
as given in Fig. 2.4. Researchers also developed the fractional-order numerical func-
tions in other software packages like Wolfram Mathworld and Sysquake [54]. These
functions also include time and frequency domain analysis as well as numerical and
stability analysis of the controllers [55]. Thus, it can be concluded that there is a con-
tinuous growth in the application of fractional-order controllers for feedback control
problems.
The fractional-order PID (PIλ Dμ ) controller has been extensively used in many
engineering control applications to achieve more robust and stable performance [7,
62–67]. This is due to the fact that fractional-order controllers are less sensitive to
parameter changes and can easily attain the property of iso-damping [7, 68]. With
the fractional-order PID controller, various specifications of the controller can be
achieved as given in Table 2.3.
The generalized transfer function of PIλ Dμ controller proposed by Podlubny has to
meet two more specifications which improve or enhance the performance of the over-
all systems. Over the years, many forms of fractional-order PI and PID controllers
have been proposed by researchers. The PIλ Dμ class of controllers are referred to
as a generalized or standard form of the controller derived from the conventional
PID. However, as argued by Luo et al. in [69–72], the structure is not appropriate in
ensuring the robust performance of the closed-loop system. Thus, the authors pro-
posed a novel structure called FO[PI] (or [PI]λ ). The proposed structure was imple-
mented on a class of fractional-order time-delay systems, motion control systems,
and fixed-wing unmanned aerial vehicle. In all these cases, the control parameters
were obtained using frequency domain analysis and the application is based on the
simulated environment.
In a similar way, a number of authors proposed a different structure called (PI)n for
a class of stable time-delay systems [73–75]. Here, to determine the controller param-
18 2 Fractional-order Set-Point Weighted Controllers

Table 2.3 Parameters of controllers derived from PIλ Dμ controller

Controller Fractional-order parameters
Category Type λ μ
PID P 0 0
PI 1 0
PD 0 1
PID 1 1
PIλ Dμ PIλ 0<λ<1 0
PDμ 0 0<μ<1
PIλ Dμ 0<λ<1 0<μ<1

eters, output feedback approach was adopted [73] using time moments approach in
[75] and steepest descent method in [74]. These approaches indicated that the per-
formance of the controller is more robust to gain variants. Similarly, to the frequency
domain approach application here is also limited to the simulation environment. On
the other hand, the nonlinear form of the fractional-order PID controller is a com-
bination of NPID in [42–45] and FOPID controllers [76]. This type of controller
can improve the set-point tracking properties of the given closed-loop system with a
smaller design effort. However, it is very difficult to tune the eight controller param-
eters.
In a related development, the internal model controller (IMC) of fractional-order
PID (IMC-PIα Dα ) reported in [4, 77, 78] is an extension of PIλ Dμ in which γ s α + 1 is
added to make proper transfer function and derived with an assumption of λ = μ = α.
The one of the objective of this class of PIλ Dμ controller is to reduce the controller
parameters. Thus, a modified version of FOPID controller reported in [79–81] has
only three controller parameters. These proposed structures exhibits a flat phase
band-limited lead contribution, and improves the robustness of the system.

2.1.2 Review on Application of Fractional-order PID

Controllers

The industrial perspective of exploiting the expertise of systems and control engineer-
ing has a significant wide spectrum. The objective of the industrial process control is
to develop high-performance feedback systems, adaptive and self-tuning controllers
taking care of the changes in process dynamics, nonlinear, robust and optimal con-
trol policies to meet a predefined set of performance criteria, fault-tolerant schemes,
and safety-critical applications. Example of such sectors fall in this field are sensor
and transducer manufacturing processes, chemical process industries, robot manu-
facturing industries, energy industries processing all forms of resources and energy
transmission industries, marine and automotive sectors and metal forming industries
[56, 82, 83]. The application of fractional-order controllers in some of these sec-
2.1 Introduction 19

Table 2.4 List of patents related to fractional-order systems and controllers

Year Patent number Title Reference
1994 US5371670A Three-parameter tunable tilt-integral-derivative [97]
(TID) controller
2004 US6678670B2 Non-integer order dynamic systems [98]
2006 WO2006112976A2 Electrical component with fractional-order [99]
impedance
2009 US7599752B2 Tuning methods for fractional-order controllers [100]
2016 US9305706B2 Fractional-order capacitor [101]

tors has been studied by various researchers. An example of such applications are
fractional-order control of bioreactor temperature [4, 84, 85], robotic manipulator
[15, 86, 87], load frequency [88, 89], power control [90–92], DC motor [93], auto-
matic voltage regulator [64, 94], wind turbine generator [64, 89], level control [95],
twin-rotor system [15, 67, 87], rail vehicles, electro-hydraulic systems, electrical
drives and position control [62, 96].
In all these applications, the FOPID controller achieved better robust and stable
performance compared to conventional PID. However, in the area of process control,
like in the case of PID, the controller has both proportional and derivative kick
effects. Furthermore, the key challenges in the implementation of the fractional-order
controller involve the approximation of fractional-order operators and computational
tools for analysis, simulation and implementation.
It should be noted that the patents and hardware prototypes can be seen as impor-
tant factors in industrial applications and automation. Thus, the list of available
patents related to fractional-order controllers are given in Table 2.4. From the table,
it can be seen that there are only a few patents related to fractional-order systems and
controllers. However, these patents have also been useful in the sense of achieving
an ideal implementation of fractional-order controllers.
On the other hand, for hardware prototypes, the authors of [102–104] developed
an analog and digital realizations of fractional-order integrator and differentiator
using field-programmable gate array (FPGA). Furthermore, Tepljakov developed a
hardware prototype for FOPID controller using 8-bit Atmel AVR and 32-bit STMi-
croelectronics STM32F407 microcontroller for the digital implementation of the
controller [105]. However, the prototype is designed mainly for laboratory experi-
ments. For industrial use, a suitable form-factor and a suitable choice of components
like manual controls and display are necessary to meet industrial standards.

2.1.3 Summary

Looking at the available literature, it is confirmed that the PID controllers are the
most widely used and has the highest industrial impact compared to all other robust,
20 2 Fractional-order Set-Point Weighted Controllers

advanced and intelligent control strategies. This is because of its advantages of sim-
plicity in design, ease of tuning and implementation. However, during the instanta-
neous step change in the set-point with a PID controller produces an excessive change
in the controller output. Furthermore, for disturbance rejection, a fast response will
be achieved with a high gain PID controller but with an oscillatory response. On the
other hand, reducing undesired oscillations of a control signal is another common
control objective. This is because the undesired oscillations in the control signals
degrade the actuators faster. Thus, the overall objective is now to achieve adequate
set-point tracking and effective load regulation through smooth control action. There-
fore, to mitigate these issues, this work proposes the fractional-ordering of SWPID
and PI-PD control strategies. The proposed control strategy is expected to achieve
good set-point tracking and load regulation performance through smooth control
action over conventional controllers.

2.2 Fractional-order Set-Point Weighted PID Controller

In the first part of this section, the design of fractional-order set-point weighted PID
(SWPIλ Dμ ) controller in standard, industrial, parallel and ideal configurations of
PID are presented [106]. Then, the implementation of the controller in dual-input
single-output and single-input single-output configurations are presented. Lastly, the
conversion of controller parameters from one form to other inducing the effect of
variation on these controller parameters is presented.

2.2.1 Standard Structure

Consider the closed-loop control system of Fig. 2.5, the control signal (U (s)) of a
standard configuration of a conventional PID controller is defined as follows:
 
1 Td s
U (s) = K p 1 + + E(s) (2.4)
Ti s αTd s + 1

Fig. 2.5 Representation of simple closed-loop process control system

2.2 Fractional-order Set-Point Weighted PID Controller 21

where
• K p is the proportional gain,
• Ti is the integral time constant,
• Td is the derivative time constant and
• α is the derivative filter constant.
Thus, from U (s) in Eq. (2.4) it can be seen that the three control actions are
functions of error signal (E(s)) defined as

E(s) = R(s) − Y (s) (2.5)

where R(s) is the reference signal and Y (s) is the process output.
The control signal of set-point weighted PID (SWPID) is obtained by treating the
set-point R(s) and the process output Y (s) separately. In order to achieve this, the
control signal of PID in Eq. (2.4) can be written with error terms associated with
each of the proportional, integral and derivative actions as follows:
 
1 Td s
U (s) = K p E p (s) + E i (s) + E d (s) (2.6)
Ti s αTd s + 1

In the above equation, E p (s), E i (s) and E d (s) are the error terms associated with
each of the proportional, integral and derivative actions. Therefore, the control action
of set-point weighted PID controller is obtained by weighting the reference signal of
the proportional and derivative actions of U (s) in Eq. (2.6) with set-point weights b
and c respectively as:

E p (s) = b R(s) − Y (s), 0 < b < 1 (2.7)

E i (s) = R(s) − Y (s) (2.8)

E d (s) = c R(s) − Y (s), 0 < c < 1 (2.9)

where b is the set-point weight on the proportional action and c is the set-point weight
on the derivative action.
From E i (s) given in Eq. (2.8), it should be noted that to avoid steady-state control
error, the error associated with the integral action is not weighted. Thus, by substi-
tuting E p (s), E i (s) and E d (s) from Eqs. (2.7), (2.8) and (2.9) respectively in U (s)
of Eq. (2.6), the control action can be written as
 
1 cTd s
U (s) = K p b + + R(s)
Ti s αTd s + 1
  (2.10)
1 Td s
− Kp 1 + + Y (s), 0 < b, c < 1
Ti s αTd s + 1
22 2 Fractional-order Set-Point Weighted Controllers

Fig. 2.6 Block diagram of standard configuration of SWPIλ Dμ controller

From the above control action, the control signal of fractional-order set-point
weighted PID (SWPIλ Dμ ) is obtained by fractional-ordering the integral and deriva-
tive actions. Thus, the control action of SWPIλ Dμ controller from the control signal
of SWPID in Eq. (2.10) with fractional-order integral and derivative actions of orders
λ and μ respectively as follows:
 
1 cTd s μ
U (s) =K p b + + R(s)−
Ti s λ αTd s μ + 1
  (2.11)
1 Td s μ
Kp 1 + + Y (s), 0 < λ, μ < 1
Ti s λ αTd s μ + 1

where λ is the order of integration and μ is the order of differentiation.

From the control signal of SWPIλ Dμ controller in Eq. (2.11), it can be seen that
the controller is synonymous to having two fractional-order PID (PIλ Dμ ) controllers;
one for set-point tracking and the other for disturbance rejection. The block diagram
implementation of Eq. (2.11) is given in Fig. 2.6. Here, for small values of weight b,
reduces overshoot but results in a slower response to set-point changes. Furthermore,
the weight c is usually set to zero in order to avoid derivative kick effect in the control
action. In this case Eq. (2.11) reduces to
   
1 1 Td s μ
U (s) = K p b + R(s) − K p 1 + + Y (s) (2.12)
Ti s λ Ti s λ αTd s μ + 1

Therefore, Eq. (2.12) is implemented in a closed loop control structure of Fig. 2.7.
In the figure, Csp (s) and C y (s) are the respective controllers applied to R(s) for set-
point tracking and Y (s) for disturbance rejection are given in Eqs. (2.13) and (2.14)
respectively.  
1
Csp (s) = K p b + (2.13)
Ti s λ
2.2 Fractional-order Set-Point Weighted PID Controller 23

Fig. 2.7 Implementation of SWPIλ Dμ controller

 
1 Td s μ
C y (s) = K p 1 + + (2.14)
Ti s λ αTd s μ + 1

Thus, the designed SWPIλ Dμ controller will respond to set-point changes and
load disturbances separately. Furthermore, the fractional-order control action of the
controller is smoother and avoids undesired oscillations.

2.2.2 Ideal Structure

The standard SWPIλ Dμ controller in Eq. (2.12) reduces to ideal controller for α = 0.
Therefore, the control signal of ideal SWPIλ Dμ is given as
   
1 1 μ
U (s) = K p b + R(s) − K p 1 + + Td s Y (s) (2.15)
Ti s λ Ti s λ

Thus, the ideal SWPIλ Dμ controller with external first-order filter with time con-
stant T f is given as
      
1 1 μ 1
U (s) = K p b + R(s) − K p 1 + + Td s Y (s)
Ti s λ Ti s λ Tf s + 1
(2.16)
The block diagram implementation of Eq. (2.16) is given in Fig. 2.8. Figure 2.7
can also be used for implementation in a similar way to standard controller. Thus,
the respective set-point weighted and disturbance rejection controllers Csp (s) and
C y (s) are given as   
1 1
Csp (s) = K p b + (2.17)
Ti s λ Tf s + 1
24 2 Fractional-order Set-Point Weighted Controllers

Fig. 2.8 Block diagram of ideal SWPIλ Dμ controller with external filter

  
1 μ 1
C y (s) = K p 1 + + Td s (2.18)
Ti s λ Tf s + 1

2.2.3 Parallel Structure

Similarly, from Eq. (2.12), the control signal of parallel SWPIλ Dμ controller is given
as follows:
   
Ki Ki Kd sμ
U (s) = bK p + λ R(s) − K p + λ + Y (s) (2.19)
s s α Kd sμ + 1

where
• K p is the proportional gain,
• K i is the integral gain,
• K d is the derivative gain and
• α p is the derivative filter constant.
It should be noted that, unlike the standard controller where the K p appears in all
actions, here, K p is independent of integral and derivative actions. The block diagram
implementation of Eq. (2.19) is given in Fig. 2.9. The closed-loop implementation
of the controller is done using the block diagram in Fig. 2.7. Thus, the respective
controllers Csp (s) and C y (s) for R(s) and Y (s) are given as
2.2 Fractional-order Set-Point Weighted PID Controller 25

Fig. 2.9 Block diagram of

parallel SWPIλ Dμ controller

Ki
Csp (s) = bK p + (2.20)
sλ
Ki Kd sμ
C y (s) = K p + + (2.21)
sλ α Kd sμ + 1

2.2.4 Industrial Structure

From the closed-loop control system of Fig. 2.5, the control signal (U (s)) for indus-
trial configuration of PID controller is defined as follows:
  
1 Td s + 1
U (s) = K p 1+  E(s) (2.22)
Ti s αTd s + 1

From the equation, it can be seen that this type of controller is synonymous
to having PI and PD controllers in series for easy implementation in industry. In
the above control signal, the  notation is used to differentiate with the standard
parameters of PID and it is for analysis purpose only.
Therefore, in a similar way to standard PID, the control signal of industrial con-
figuration of SWPID controller is derived as
    
1 c Td s + 1
U (s) = K p 
b +  R(s)−
Ti s α  c T  s + 1
   d  (2.23)
1 Td s + 1
K p 1 +  Y (s), 0 < b , c < 1
Ti s α  Td s + 1
26 2 Fractional-order Set-Point Weighted Controllers

Fig. 2.10 Block diagram of industrial configuration of SWPIλ Dμ controller

The same assumption of c = 0 for avoiding the derivative kick effect is done
here. Therefore, the simplified form of the controller is given as
     
1 1 Td s + 1
U (s) = K p  
b +  R(s) − K p 1 +  Y (s) (2.24)
Ti s Ti s α  Td s + 1

From the above control signal, the fractional-order form of the set-point weighted
PID controller is obtained by fractional ordering the integral and derivative actions
of the controller with λ and μ as follows:
 
1
U (s) = K p b + R(s)
Ti s λ
   μ  (2.25)
1 Td s + 1
− K p 1 +  λ Y (s), 0 < λ , μ < 1
Ti s α  Td s μ + 1

The block diagram implementation of industrial SWPIλ Dμ controller structure

based on Eq. (2.25) is shown in Fig. 2.10. Equation (2.25) is implemented in a similar
way to Eq. (2.11) using Fig. 2.7. The first terms of standard SWPIλ Dμ in Eq. (2.12)
and industrial SWPIλ Dμ in Eq. (2.25) clearly shows that both configuration have
similar Csp (s) controller. The difference between the two configurations is in the
second terms of the equations i.e., C y (s), which for the industrial configuration is an
interactive PID given in Eq. (2.26).
   
1 Td s μ + 1
C y (s) = K p 1+ (2.26)
Ti s λ α  Td s μ + 1
2.2 Fractional-order Set-Point Weighted PID Controller 27

Fig. 2.11 Feedforward type

of SWPIλ Dμ controller

2.2.5 Equivalent Configurations

The standard implementation of standard, ideal, parallel and industrial configurations

of SWPIλ Dμ controller shown in Fig. 2.7 is a dual-input and single-output (DISO)
controller which can be decomposed into Csp (s) and C y (s) controllers. The controller
can also be in other configurations such as feedforward, feedback, pre-filter and
component separated type. These additional configurations take the form of single-
input-single-output (SISO) controllers as explained in the following subsections.

2.2.5.1 Feedforward Configuration

In feedforward configuration, the SWPIλ Dμ controller is decomposed into a

fractional-order forward controller C(s) and a fractional-order feedforward con-
troller C f (s) as shown in Fig. 2.11.
From the figure, the control signal (U (s)) is given as

U (s) = C f (s)R(s) + C(s)E(s) (2.27)

By substituting E(s) = R(s) − Y (s) in Eq. (2.27), we get

 
U (s) = C f (s)R(s) + C(s) R(s) − Y (s)
  (2.28)
= C f (s) + C(s) R(s) − C(s)Y (s)

Therefore, by comparing Eqs. (2.12) and (2.28), the controllers transfer functions
can be deduced to  
1
C f (s) + C(s) = K p b + (2.29)
Ti s λ
 
1 Td s μ
C(s) = K p 1 + + (2.30)
Ti s λ αTd s μ + 1
28 2 Fractional-order Set-Point Weighted Controllers

Fig. 2.12 Feedback type of

SWPIλ Dμ controller

Thus, by substituting Eqs. (2.30) in (2.29), the transfer function of C f (s) is derived
as  
Td s μ
C f (s) = K p (b − 1) − (2.31)
αTd s μ + 1

2.2.5.2 Feedback Configuration

Similar to feedforward configuration, the feedback configuration of controller can

be decomposed into a fractional-order forward controller C(s) and a fractional-order
feedback controller Cb (s) as shown in Fig. 2.12.
From the figure, the control signal (U (s)) can be written as

U (s) = C(s)E(s) − Cb (s)Y (s)

 
= C(s) R(s) − Y (s) − Cb (s)Y (s) (2.32)
 
= C(s)R(s) − C(s) + Cb (s) Y (s)

Therefore, by comparing Eqs. (2.12) and (2.32), the controllers transfer functions
can be deduced to  
1
C(s) = K p b + (2.33)
Ti s λ
 
1 Td s μ
C(s) + Cb (s) = K p 1 + + (2.34)
Ti s λ αTd s μ + 1

Substituting Eq. (2.33) in Eq. (2.34), the transfer function of Cb (s) is derived as
 
Td s μ
Cb (s) = K p (1 − b) + (2.35)
αTd s μ + 1
2.2 Fractional-order Set-Point Weighted PID Controller 29

Fig. 2.13 Pre-filter type of SWPIλ Dμ controller

2.2.5.3 Pre-filter Configuration

In this configuration, the controller is decomposed into a fractional-order forward

controller C(s) and a fractional-order pre-filter F(s) on the reference signal as shown
in Fig. 2.13. From the figure, the control signal is given as
 
U (s) = C(s) F(s)R(s) − Y (s)
(2.36)
= C(s)F(s)R(s) − C(s)Y (s)

By comparing Eqs. (2.12) and (2.36), the controllers transfer functions can be
simplified as  
1 Td s μ
C(s) = K p 1 + + (2.37)
Ti s λ αTd s μ + 1
 
1
C(s)F(s) = K p b + (2.38)
Ti s λ

Thus, the transfer function of filter (F(s)) is given as

bαTi Td s λ+μ + bTi s λ + αTd s μ + 1

F(s) = (2.39)
(α + 1)Ti Td s λ+μ + Ti s λ + αTd s μ + 1

2.2.5.4 Component Separated Type

In component separated type, the proportional, integral and derivative actions of the
controller are built separately as shown in Fig. 2.14. Therefore, from the figure, the
control signal is given as

  1  
U (s) = K p (1 − α) R(s) − Y (s) + λ
R(s) − Y (s)
Ti s

Td s μ  
+ (1 − β) R(s) − Y (s) , 0 < α, β < 1
T f sμ + 1
(2.40)
30 2 Fractional-order Set-Point Weighted Controllers

Fig. 2.14 Component-separated type of SWPIλ Dμ controller

where
• α and β are the set-point weight on the proportional derivative actions and
• T f is the derivative filter time constant.
The control signal (U (s)) given in Eq. (2.40) can be written in the form of con-
trollers applied to R(s) for set-point tracking and Y (s) for disturbance rejection as
follows:
 
1 Td s μ
U (s) = K p (1 − α)+ λ + (1 − β) R(s)
Ti s T f sμ + 1
  (2.41)
1 Td s μ
− Kp 1 + + Y (s)
Ti s λ T f sμ + 1

Therefore, the respective controllers Csp (s) and C y (s) as defined in Fig. 2.7
applied to R(s) for set-point tracking and Y (s) for disturbance rejection are given as
 
1 Td s μ
Csp (s) = K p (1 − α) + + (1 − β) (2.42)
Ti s λ T f sμ + 1
 
1 Td s μ
C y (s) = K p 1 + + (2.43)
Ti s λ T f sμ + 1

In order to avoid steady state error, β is set to one. In this case, Csp (s) is reduced
to  
1
Csp (s) = K p (1 − α) + (2.44)
Ti s λ
2.2 Fractional-order Set-Point Weighted PID Controller 31

Table 2.5 Parameters of controllers derived from standard SWPIλ Dμ controller

Controller Set-point weighting parameters Fractional-order parameters
Category Type b c λ μ
PID P 1 0 0 0
PD 1 1 0 1
PI 1 0 1 0
PID 1 1 1 1
SWPID SWPD 0<b<1 0<c<1 0 1
SWPI 0<b<1 0<c<1 1 0
SWPID 0<b<1 0<c<1 1 1
SWPI–D 0<b<1 0 1 1
PIλ Dμ Iλ -PDμ 0 0 0<λ<1 0<μ<1
Iλ Dμ -P 0 1 0<λ<1 0<μ<1
PIλ -Dμ 1 0 0<λ<1 0<μ<1
PIλ Dμ 1 1 0<λ<1 0<μ<1
SWPIλ Dμ SWPIλ -Dμ 0<b<1 0 0<λ<1 0<μ<1
SWIλ Dμ -P 0 0<c<1 0<λ<1 0<μ<1
SWPIλ Dμ 0<b<1 0<c<1 0<λ<1 0<μ<1

2.2.6 Parameters of Various Controllers

In this section, the parameters of various controllers that are derived from the
SWPIλ Dμ controller and the effect of changing these parameters on the steady-state
and transient response of the system is presented. Therefore, the tunable set-point
weighting parameter (b and c) and fractional-order parameters (λ and μ) of various
controllers derived from SWPIλ Dμ are given in Table 2.5. From the table, respective
planes for the PIλ Dμ and SWPIλ Dμ for various configurations are shown in Fig. 2.15.
The PIλ Dμ plane in Fig. 2.15a is formed by four control points P, PI, PD, and PID.
In this plane, there are two tuning knobs λ and μ to enhance the performance. This
performance can be further be improved by adding two more additional tuning knobs
b and c in the SWPIλ Dμ plane with boundaries Iλ -PDμ , PIλ -Dμ , Iλ Dμ -P and PIλ Dμ
as shown in Fig. 2.15b. Effectively, this makes the tunable knobs to four i.e., λ, μ,
b and c. The effect of changing these parameters on the steady-state and transient
response of the system is summarized in Table 2.6.

2.2.7 Conversion of Controller Parameters

The conversion of controller parameters from standard SWPIλ -Dμ controller given
in Sect. 2.2.1 to industrial one presented in Sect. 2.2.4 can be deduced by comparing
both the control signals as shown in Fig. 2.16. In the figure, F and F  denotes the
32 2 Fractional-order Set-Point Weighted Controllers

Fig. 2.15 Graphical representation of controller structures

Table 2.6 Effect of variation in parameters of standard SWPIλ -Dμ controller

Parameter Variation Rise time Settling Overshoot Steady-state Stability
time error
Kp Increasing Reduces Minor Increases Reduces Degrades
change
Ti Decreasing Reduces Increases Increases Eliminates Degrades
Td Increasing Minor Reduces Reduces Minor Improves
change change
b Decreasing Increases Reduces Reduces Eliminates Improves
λ Decreasing Minor Reduces Reduces Increases Degrades
change
μ Decreasing Minor Reduces Reduces Minor Improves
change change

conversion factors. From the relations, it should be noted that the conversion from
α2 T 2
standard to industrial and vice versa is only possible if 1 − (4+2α)T
Ti
d
+ T 2d > 0 and
i
Ti
α < 1 + Td
respectively.

2.3 Fractional-order PI-PD Controller

In this section, the design of the fractional-order PI-PD (PIλ -PDμ ) controller in
two single-loop control configurations is presented. The advantage of this controller
is the integer-order part of the controller has the ability to achieve good set-point
tracking and load regulation performance while the fractional-order part reduces
the oscillations in the control signal. Next, the parameters of the various controller
configurations are deduced. Finally, the conversion of these controller parameters
from one configuration to another is presented.
2.3 Fractional-order PI-PD Controller 33

Fig. 2.16 Conversion of

SWPIλ -Dμ controller
parameters

Fig. 2.17 PI-PD controller

Type 1

2.3.1 PI-PD Controller

Consider the controller transfer functions of PI and PD controllers of PI-PD as defined

as follows:
ki
C P I (s) = k p + (2.45)
s

C P D (s) = 1 + kd s (2.46)

where, k p , ki and kd are the gains of PI-PD controller.

From Eq. (2.46), it can be noted that the proportional gain of the PD controller
must be one in order to have a zero steady-state error. This type of controller can be
implemented in two single-loop control configurations as shown in Figs. 2.17 and
2.18.
Observing the implementation block diagram of PI-PD controller Type 1 which
is referred to as PI-PD1 in Fig. 2.17, the control signal (U (s)) is derived as
34 2 Fractional-order Set-Point Weighted Controllers

Fig. 2.18 PI-PD controller Type 2

U (s) = C P I (s)E(s)
 (2.47)
= C P I (s) R(s) − C P D (s)Y (s)

where C P I (s) and C P D (s) are the PI and PD controllers defined in Eqs. (2.45) and
(2.46) respectively. Thus, by substituting both the control actions in Eq. (2.47), we
get   
ki  
U (s) = k p + R(s) − 1 + kd s Y (s)
s
    (2.48)
ki ki
= kp + R(s) − (k p + ki kd ) + + k p kd s Y (s)
s s

From the above control signal, it should be noted that the controller is equivalent to
SWPID controller. Therefore, the transfer functions of Csp (s) and C y (s) controllers
applied to R(s) for set-point tracking and Y (s) for disturbance rejection are given as

ki
Csp (s) = k p + (2.49)
s
ki
C y (s) = (k p + ki kd ) + + k p kd s (2.50)
s
Thus, the PI-PD controller has the advantage of achieving set-point weighted
behavior and load disturbance rejection with same number of parameters as that of
PID. Similarly, from Fig. 2.18, the control signal (U (s)) of Type 2 PI-PD (PI-PD2)
controller is given as follows:

U (s) = C P I (s)E(s) − C P D (s)Y (s)

 (2.51)
= C P I (s) R(s) − Y (s) − C P D (s)Y (s)
2.3 Fractional-order PI-PD Controller 35

After substituting C P I (s) and C P D (s) from Eqs. (2.45) and (2.46) in Eq. (2.51), we
get  
ki   
U (s) = k p + R(s) − Y (s) − 1 + kd s Y (s)
s
    (2.52)
ki ki
= kp + R(s) − (k p + 1) + + kd s Y (s)
s s

Likewise, the control signal in Eq. (2.52) is similar to SWPID controller. Here, the
controller Csp (s) with respective to R(s) is for set-point tracking and the controller
C y (s) with respect to Y (s) is for disturbance rejection are given as follows:

ki
Csp (s) = k p + (2.53)
s
ki
C y (s) = (k p + 1) + + kd s (2.54)
s
Therefore, in a similar way to SWPID, the PI-PD controller responds to set-point
changes and load disturbances separately. As mentioned earlier, the PI-PD retains
the simplicity of PID by maintaining the same number of parameters of the PID
controller. However, to achieve smoother control action with the PI-PD controller,
the proposed design of fractional-order PI-PD (PIλ -PDμ ) is presented in the following
subsection.

2.3.2 PIλ -PDμ Controller

The controller transfer functions for fractional-order PI (PIλ ) and fractional-order

PD (PDμ ) controllers of PIλ -PDμ controller are obtained by fractional-ordering the
conventional PI and PD controllers given in Eqs. (2.45) and (2.46) with parameters
λ and μ respectively as follows

ki
C P I λ (s) = k p + , 0<λ<1 (2.55)
sλ

C P Dμ (s) = 1 + kd s μ , 0 < μ < 1 (2.56)

The controller will be implemented in a similar way to the integer-order PI-PD

controller as given in Figs. 2.19 and 2.20. Therefore, the control signal of Type 1
PIλ -PDμ (PIλ -PDμ 1) controller based on Fig. 2.17 is given as

U (s) = C P I λ (s)E(s)
 (2.57)
= C P I λ (s) R(s) − C P Dμ (s)Y (s)
36 2 Fractional-order Set-Point Weighted Controllers

Fig. 2.19 PIλ -PDμ controller Type 1

Fig. 2.20 PIλ -PDμ controller Type 2

Hence, by substituting C P I λ (s) and C P Dμ (s) from Eqs. (2.55) and (2.56) in
Eq. (2.57), we get
  
ki  μ

U (s) = k p + λ R(s) − 1 + kd s Y (s)
s
    (2.58)
ki ki
= k p + λ R(s) − (k p + ki kd ) + λ + k p kd s μ Y (s)
s s

The transfer functions of Csp (s) and C y (s) controllers applied to R(s) for set-point
tracking and Y (s) for disturbance rejection are given as

ki
Csp (s) = k p + (2.59)
sλ
ki
C y (s) = (k p + ki kd ) + + k p kd s μ (2.60)
sλ
2.3 Fractional-order PI-PD Controller 37

Thus, the controller responds to set-point changes and load disturbances sepa-
rately. Furthermore, the fractional-order part of the control action is smoother and
avoids undesired oscillations. It can also be noted that the controller is simple and
has the same number of parameters of fractional-order PID. On the other hand, the
control signal of Type 2 PIλ -PDμ (PIλ -PDμ 2) controller based on Fig. 2.18 is given
as
U (s) = C P I λ (s)E(s) − C P Dμ (s)Y (s)
 (2.61)
= C P I λ (s) R(s) − Y (s) − C P Dμ (s)Y (s)

Similarly, by substituting C P I λ (s) and C P Dμ (s) from Eqs. (2.55) and (2.56) in
Eq. (2.61), we get
 
ki   
U (s) = k p + λ R(s) − Y (s) − 1 + kd s μ Y (s)
s
    (2.62)
ki ki μ
= k p + λ R(s) − (k p + 1) + λ + kd s Y (s)
s s

Thus, the controller transfer functions of Csp (s) and C y (s) controllers applied to
R(s) for set-point tracking and Y (s) for disturbance rejection are given in Eqs. (2.63)
and (2.64) respectively.
ki
Csp (s) = k p + λ (2.63)
s
ki
C y (s) = (k p + 1) + + kd s μ (2.64)
sλ

Observing the control actions, it can be noted that the developed PIλ -PDμ con-
trollers have the capability of handling set-point changes and load disturbances sepa-
rately through smoother control action in both the cases. Furthermore, the controller
maintains the same number of parameters of the PIλ Dμ controller.

2.3.3 Parameters of Various Controller Configurations

In a similar way to the parameters of various controller configurations derived from

SWPIλ Dμ controller in Sect. 2.2.7, here also, the parameters of various controllers
will be derived from the developed PIλ -PDμ controller. Thus, the tunable fractional-
order parameters λ and μ of various controllers derived from PIλ -PDμ controller are
given in Table 2.7.
The graphical representation of various controllers derived from PIλ -PDμ on a
λ-μ plane is given in Fig. 2.21. The PIλ -PDμ plane in the figure is formed by four
control points P-P, PI-P, P-PD, and PI-PD. In this plane, there are two tuning knobs
λ and μ to enhance the control performance of the PI-PD controller.
38 2 Fractional-order Set-Point Weighted Controllers

Table 2.7 Parameters of controllers derived from PIλ -PDμ controller

Controller λ μ
Category Type
PI-PD P-P 0 0
P-PD 0 1
PI-P 1 0
PI-PD 1 1
PIλ -PDμ PIλ -P 0< λ <1 0
PIλ -PD 0< λ <1 1
P-PDμ 0 0< μ <1
PI-PDμ 1 0< μ <1
PIλ -PDμ 0< λ <1 0< μ <1

Fig. 2.21 Graphical

representation of controller
structures

2.3.4 Conversion of Controller Parameters

In this subsection, the conversion relations are developed to convert the SWPIλ Dμ
controller parameters presented in Sect. 2.2.3 to PIλ -PDμ controller parameters.

Therefore,
 Eq. (2.58) can be rewritten by multiplying and dividing the term k p +
ki kd as follows:
  
kp   ki
U (s) = k p + ki kd + λ R(s)
k p + ki kd s
  (2.65)
  ki μ
− k p + ki kd + λ + k p kd s Y (s)
s

Therefore, by comparing the above control signal and parallel configuration of

SWPIλ Dμ controller in Sect. 2.2.1, the relations are given in Fig. 2.22. The left-hand
2.3 Fractional-order PI-PD Controller 39

Fig. 2.22 Conversion of

controller parameters

side of equations is SWPIλ -Dμ parameters while the right-hand side is PIλ -PDμ
parameters. Observing the figure, it can be seen that the conversion from SWPIλ -Dμ
to PIλ -PDμ is only possible if K 2p > 4K i K d .

2.4 Simulation Study on pH Neutralization Process

In the first part of the section, the description of pH neutralization process plant is
given with the help of Process and Instrumentation Diagram (P&ID). In the second
section, the fundamental modeling of neutralization process for the titration of acetic
acid with sodium hydroxide is presented. Then, the schematics of the complete
entire experimental setup is presented. The subsequent three subsections present the
performance evaluation of the designed SWPIλ Dμ and PIλ -PDμ controllers on the
pH neutralization process plant.

2.4.1 P&ID of pH Neutralization Process Plant

The P&ID of the pH neutralization process representing the flow is shown in Fig. 2.23
[107, 108]. From the figures, it can be seen that the process is performed in a con-
tinuous stirred-tank reactor (CSTR) where a neutralizing reagent (acid) is added to
the base to regulate its pH to an acceptable value. In the process, the desired amount
of acid and base streams from the source tanks VE100 and VE110 are pumped into
chemical mixing tank VE120 by using the pumps P100 and P110 respectively. To
regulate the flow rate of acid and base indicated by the flow transmitters FT120 and
FT121, two flow control valves FCV120 and FCV121 are used.
40 2 Fractional-order Set-Point Weighted Controllers

AT AG
122 120 AG Agitator
AIC pH Internal Controller
AT pH Sensor
CT Conductivity Transmitter
FCV Flow Control Valve
FT Flow Transmitter
AIC HV Hand Valve
122 LS Level Sensor
P Pump
VE Vessel Tank
VE120

HV121
HV110 HV100
CT LS AT LS FCV120 LS CT
Water 110 110 141 140 100 100 Water
FCV121
Supply Supply

FT FT
121 120

HV113 HV103

VE110 VE140 VE100

P110 P100
HV102
HV112 HV142

Fig. 2.23 P&ID of pH neutralization process plant

HOST PC Panel

I/O Interface
Board

Digital I/O Card

PCI-1751

AI Card
Flow Transmitter
PCI-1713U
FT121
AO Card
PCI-1720U

MATLAB Server
(HOST PC)

Flow Control Valve

FCV121

Pump
P110

Fig. 2.24 Experimental setup of pH neutralization process plant (Rear View)

Furthermore, the acid and base concentrations are monitored continuously using
two conductivity transmitters CT110 and CT110 respectively. The pH transmitter
AT122 is used to measure the pH value of the mixed solution. Whereas hand valve
HV121 is used to convey the neutralized solution to the storage tank and HV102,
HV112 and HV142 are used to drain out the acid, storage, and base tanks respectively.
The level sensors in acid, storage, and base tanks are LS100, LS110 and LS140
respectively are used to measure the level of solution tanks. The agitator AG120 in
chemical mixing tank is used for mixing the solution homogeneously. On the other
hand, the front and rear views of the real-time pH neutralization process plant used
in the experiment are given in Figs. 2.25 and 2.24 respectively.
2.4 Simulation Study on pH Neutralization Process 41

Agitator
AG120

pH Sensor
AT122

Vessel Tank
VE120

Conductivity Transmitter Flow Control Valve

CT100 FCV120

Conductivity Transmitter
CT110

Flow Transmitter
FT120

Vessel Tank
VE140
Vessel Tank
VE100

Pump
Vessel Tank
P100
VE110

Fig. 2.25 Experimental setup of pH neutralization process plant (Front View)

Flow Control Valve FCV120

400

350

300
Flowrate (L/h)

250

200

150

100

50

0
10 20 30 40 50 60 70 80 90 100
Valve Opening (%)

Fig. 2.26 Characteristics of electro-pneumatic control valve FCV 120

The characteristics of both the flow control valves FCV120 and FCV121 electro-
pneumatic control valves are shown in Figs. 2.26 and 2.27 respectively. From the
characteristics, it can be observed that the type of behavior is a nonlinear increasing
sensitivity type and the maximum input flowrates are around 390 and 230 L/h.
42 2 Fractional-order Set-Point Weighted Controllers

Flow Control Valve FCV121

250

200
Flowrate (L/h)

150

100

50

0
10 20 30 40 50 60 70 80 90 100
Valve Opening (%)

Fig. 2.27 Characteristics of electro-pneumatic control valve FCV 121

2.4.2 Modeling of pH Neutralization Process

The dynamical behavior of the CSTR is governed by the incoming flow-rates (Fa ,
Fb ) and outgoing flow-rate (Fa + Fb ). Therefore, the dynamical model is given by
mass balance equations as:

d
V C[H + ] = Fa Ca − (Fa + Fb )C[H + ] (2.66)
dt
d
V C[O H − ] = Fb Cb − (Fa + Fb )C[O H − ] (2.67)
dt
where
• V is the volume of tank VE120,
• Ca and Cb are the concentrations of acid and base and
• C[H + ] and C[O H − ] are the concentrations of non-reactant components H + and
O H −.
Consider the neutralization process of acetic acid (C H3 C O O H ) titrated with
sodium hydroxide (N a O H ). In this process, the control objective is to regulate the
pH by controlling Fb and at a constant Fa . Therefore, the titration process model is
developed based on the following assumptions:
2.4 Simulation Study on pH Neutralization Process 43

1. all ions are completely soluble,

2. there is perfect mixing in the chemical mixing tank and
3. the reactions in the process are instantaneous and isothermal.
The first stage is to adequately model the titration process by observing the fol-
lowing chemical reactions occurring in the mixing tank:

C H3 C O O H + N a O H → C H3 C O O N a + H2 O (2.68)

H + + O H −  H2 O (2.69)

C H3 C O O H → H + + C H3 C O O − (2.70)

N a O H → N a+ + O H − (2.71)

From Eqs. (2.69) and (2.70) the ionic product of water (K w ) and the dissociation
constant of acetic acid (K a ) are given in Eqs. (2.72) and (2.73) respectively.

K w = [H + ][O H − ] = 1 × 10−14 (at 25 ◦ C) (2.72)

[H + ][C H3 C O O − ]
Ka = (2.73)
[C H3 C O O H ]

The next stage is to derive the mass balance equations of non-reactant components
in the system that relates the concentrations of dissociation products as follows:

C[H + ] = [C H3 C O O H ] + [C H3 C O O − ] (2.74)

C[O H − ] = [N a + ] (2.75)

The final stage is the derivation of the electroneutrality principle. According to

the principle, the sum of all positively charged ions must be equal to the sum of all
negatively charged ions in aqueous solutions. Therefore, the charge balance is given
as follows:
[H + ] + [N a + ] = [O H − ] + [C H3 C O O − ] (2.76)

In summary, by substituting Eqs. (2.72) through (2.75) in Eq. (2.76) a polynomial

equation of cubic order in [H + ] is obtained as follows:

[H + ]3 + [H + ]2 (K a + C[O H − ])+
(2.77)
[H ](K a C[O H ] − K w − K a C[H + ]) − K a K w = 0
+ −

The pH value, which indicates the level of concentration of hydrogen ion [H + ]

is calculated as
p H = − log10 [H + ] (2.78)
44 2 Fractional-order Set-Point Weighted Controllers

Fig. 2.28 Schematic diagram of the complete experimental setup

2.4.3 Schematics of Experimental Setup

The schematic diagram of the complete experimental setup is shown in Fig. 2.28.
The developed fractional-order control strategies will be implemented in the host
PC using MATLAB/Simulink software which is accessed through ‘Remote Desktop
Connection’ in the control room. The host PC is interfaced with the plant through
Peripheral Component Interconnect (PCI) cards 1713U, 1720U and 1751 as shown in
Table 2.8. The PCI-1713U is a 100 kS/s, 12-bit, 32-ch isolated analog input (AI) card
used to access signal from all the transmitters. On the other hand, the PCI-1720U is
a 12-bit, 4-channel isolated analog output (AO) card used to send the control signal
to flow control valves. The last card PCI-1751 is a 48-channel digital I/O and 3-
channel counter card used to operate the pumps, agitator and the process plant either
Remotely or Locally (R/L).

2.4.4 Performance of SWPIλ Dμ Controller

In this subsection, the design of SWPIλ Dμ controller parameters will be presented.

Then, the results of the proposed standard and industrial configuration of the designed
controller in comparison with PID, PIλ Dμ and SWPI-D for set-point tracking, vari-
ation in set-point and disturbance rejection will be presented. In all these cases, as
mentioned earlier, the numerical analysis will be done in terms of rise time (tr ), set-
tling time before and after disturbance (ts1 and ts2 ) and percentage overshoot (%OS).
Furthermore, it should be noted that a disturbance (D(s)) of 25% is injected at
250 s for the purpose of disturbance rejection evaluation. Moreover, the ability of the
controllers to track variation in set point will also be analyzed.
2.4 Simulation Study on pH Neutralization Process 45

Table 2.8 Channel connections of PCI cards to the devices of pH neutralization plant
PCI card Channel Device
PCI-1713U (Analog Input) 0 CT100
1 FT120
2 CT110
3 FT121
4 AT122
PCI-1720 (Analog Output) 0 FCV120
1 FCV121
PCI-1751 (Digital Input/Output) 0 P100
1 P110
2 AG120
3 WS-PNL

2.4.4.1 Controller Parameters

The parameters of the pH model developed earlier in Sect. 2.4.2, are given in
Table 2.9, while the simulation result on the variation of titration curve for the pH
process due to different concentrations of solutions is shown in Fig. 2.29. From the
curve, it can be seen that the behavior of the process is highly nonlinear. The S-
shaped curve is very sensitive around the equilibrium range (i.e., 6–12 pH), and thus
a small perturbation of the input at this point causes a significant change of the output.
Another observation made on the curve is that change in concentrations of the acid
and base results in the dynamic behavior of the process.
On the other hand, for the tuning of developed SWPIλ Dμ controller using APSO
algorithm as discussed in Appendix A, the optimization parameters for the algo-
rithm are also given in Table 2.9. Therefore, the tuned parameters of the controllers
compared with performance index ITAE in Eq. (A.8) are given in Table 2.10.

2.4.4.2 Standard SWPIλ Dμ Controller

The comparison of closed-loop response of the system for set-point tracking and dis-
turbance rejection with standard configuration of PID in Eq. (2.4), PIλ Dμ in Eq. (2.2),
SWPID in Eq. (2.10) and SWPIλ Dμ in Eq. (2.12) is shown in Fig. 2.30.
From the figure, it can be seen that the set-point tracking ability of SWPIλ -Dμ is
better when compared to those of PID, PIλ Dμ and SWPI-D. The figure also reveals
that the action of SWPIλ -Dμ gives a smoother control signal while other controllers
produced undesired oscillations. This is more visible in Fig. 2.31 of the zoomed plot
of regions of interest A, B, C and D. To numerical observation of Table 2.11 shows that
the SWPIλ -Dμ has less overshoot of 0.4802% compared to SWPI-D, PIλ Dμ and PID
with overshoots of 0.9536, 4.4836 and 4.9182% respectively. From the table, it can
46 2 Fractional-order Set-Point Weighted Controllers

Table 2.9 Model and optimization parameters

Category Parameter Value
Model parameters V 50 L
Fa 10 m3 /h
Ca 0.5 mol/L
Cb 0.5 mol/L
Ka 1.86 × 10−5
Kw 10−14
Optimization parameters D1 3 (K p , Ti , Td )
D2 2 (λ, μ)
D3 1 (b)
N 50
bir d Step 50
αo 0.5
γ 0.7
β 0.7

Fig. 2.29 Titration curve variation under various concentrations

be seen that the SWPIλ -Dμ settled faster compared to the other controllers at 20.7780
and 363.5095 s before and after disturbance respectively. However, PIλ Dμ responds
faster with a rise time of 15.8972 s as against 17.0750, 18.8494 and 19.5630 s of PID,
SWPIλ -Dμ and SWPI-D controllers. This is because the weighting factor b which is
less than one for the two set-point weighted controllers SWPI-D and SWPIλ -Dμ .
2.4 Simulation Study on pH Neutralization Process 47

Table 2.10 Tuned controller parameters for standard form of controllers

Controller Kp Ti Td λ μ b
PID 39.7059 1.2836 0.0171 – – –
PIλ Dμ 39.7059 1.2836 0.0171 0.9767 0.5811 –
SWPI-D 39.7059 1.2836 0.0171 – – 0.9516
SWPIλ -Dμ 39.7059 1.2836 0.0171 0.9767 0.5811 0.9516

7.5
A
B
7
pH

Setpoint
PID
6.5 PI λ D μ
SWPI-D
SWPI λ -Dμ
6
0 50 100 150 200 250 300 350 400 450 500
(a) Time (s)

40
D
Control signal (u(t))

30

C
20
PID
PI λ D μ
10
SWPI-D
SWPI λ -Dμ
0
0 50 100 150 200 250 300 350 400 450 500
(b) Time (s)

Fig. 2.30 Performance comparison of standard SWPIλ Dμ controller for set-point tracking and
disturbance rejection

The faster settling times 363.5675 and 363.5095 s of the PIλ Dμ and SWPIλ -Dμ
controllers after the disturbance compared to 370.9695 and 370.9194 s of PID and
SWPI-D is an indication of the good load regulation ability of the fractional-order
controllers. It is worth noting that, the disturbance rejection controller C y (s) for PID
and SWPI-D is the same while that of the PIλ Dμ and SWPIλ -Dμ is the same. Hence,
the closeness of the ts2 of PID and SWPI-D in one hand and PIλ Dμ and SWPIλ -Dμ
on the other hand.
To observe the derivative kick effect of the compared controllers, the plant is
simulated to a variable set-point signal and the result is shown in Fig. 2.32. From
48 2 Fractional-order Set-Point Weighted Controllers

(a) (b)
7
7.2
6.8
pH

7
6.6
6.8
6.4
20 40 60 80 100 250 300 350

(c) (d)
25
Control signal (u(t))

20 30

15
20

10
10
5
20 40 60 80 100 250 300 350
Time (s) Time (s)

Fig. 2.31 Regions of interest A, B, C and D of Fig. 2.30

Table 2.11 Performance analysis of standard form of controllers

Controller tr ts1 ts2 %OS J I T AE
PID 17.0750 110.1329 370.9695 4.9182 0.8131
PIλ Dμ 15.8972 101.4693 363.5675 4.4836 0.6930
SWPI-D 19.5630 22.1078 370.9194 0.9536 0.2903
SWPIλ -Dμ 18.8494 20.7780 363.5095 0.4802 0.2624

the responses, it can be seen that during the set-point change, the performance of
SWPIλ -Dμ is better when compared to the other controllers. This is clearly seen in
the zoomed plot of regions of interest A and B in Fig. 2.33.
Additionally, during the set-point change, the derivative kick effect observed in
the control signals of both PID and PIλ Dμ is reduced in SWPI-D and SWPIλ -Dμ .
This is visible from the zoomed plot of regions C and D also in Fig. 2.33. As seen in
the zoomed figure, the control signals of PID and PIλ Dμ are both very narrow, sharp
and reached a peak of 100% and around 40% respectively. On the other hand, the
peak of the control signals of SWPI-D and SWPIλ -Dμ is both around 25%.
2.4 Simulation Study on pH Neutralization Process 49

7.5

7
A B
Setpoint
pH

6.5 PID
PI λ D μ
6 SWPI-D
SWPI λ -Dμ
5.5
0 100 200 300 400 500 600 700
(a) Time (s)

100
Control signal (u(t))

PID D
80
PI λ D μ
60 SWPI-D
SWPI λ -Dμ
40
C
20

0
0 100 200 300 400 500 600 700
(b) Time (s)

Fig. 2.32 Performance comparison of standard SWPIλ Dμ controller for variable set-point tracking

2.4.4.3 Industrial SWPIλ Dμ Controller

Following the similar fashion to the standard PIDs in Sect. 2.4.4.2, the tuned controller
parameters of the industrial PIDs compared based on the tuning procedure given in
Appendix A and the APSO algorithm with ITAE in Eq. (A.8) as error function are
given in Table 2.12.
The comparison of closed-loop response of the system for set-point tracking and
disturbance rejection with industrial configuration of PID in Eq. (2.22), PIλ Dμ in
Eq. (2.3), SWPID in Eq. (2.23) and SWPIλ Dμ in Eq. (2.25) is shown in Fig. 2.34.
From the figure, just as observed in the case of standard PID forms, the set-point
tracking ability of SWPIλ -Dμ is better when compared to those of PID, PIλ Dμ and
SWPI-D. The smoother control signal is also observed with SWPIλ -Dμ compared to
the oscillatory response of the other controllers. This becomes more clear if regions
of interests A, B, C and D of Fig. 2.35 are observed.
The numerical observation of Table 2.13 further confirms the advantages of the
proposed approach over the compared controllers. From the table, SWPIλ -Dμ has
less overshoot of 0.4395% compared to SWPI-D, PIλ Dμ and PID with overshoots
of 0.9732, 2.8758 and 2.9329% respectively.
50 2 Fractional-order Set-Point Weighted Controllers

(a) (b)
6.8 6.6

6.6
6.4
6.4
pH

6.2
6.2

6 6

250 300 350 500 520 540 560 580

(c) (d)
15
Control signal (u(t))

80
10
60
5
40
0
20
-5
250 300 350 500 520 540 560 580
Time (s) Time (s)

Fig. 2.33 Regions of interest A, B, C and D of Fig. 2.32

Table 2.12 Tuned controller parameters for industrial form of controllers

Controller Kp Ti Td λ μ b
PID 35.9386 1.7955 0.0390 – – –
PIλ Dμ 35.9386 1.7955 0.0390 0.9164 0.9570 –
SWPI-D 35.9386 1.7955 0.0390 – – 0.9835
SWPIλ -Dμ 35.9386 1.7955 0.0390 0.9164 0.9570 0.9835

The SWPIλ -Dμ also settled faster than other controllers at 26.3520 and 395.4278 s
before and after disturbance respectively. However, PIλ Dμ responds faster with a rise
time of 19.3370 s as against 20.8046, 22.4501 and 23.8224 s of PID, SWPIλ -Dμ and
SWPI-D respectively. This is due to the reason explained earlier of the weighting
factor b being less than unity for the set-point weighted controllers.
As observed in the case of standard configurations, faster settling times 397.1517
and 395.4278 s of the PIλ Dμ and SWPIλ -Dμ controllers after the disturbance com-
pared to 405.7697 and 404.9420 s of PID and SWPI-D indicates good distur-
bance rejection capability of the fractional-order configuration. The slightly different
behavior of control signal of the PIλ Dμ is observed here compared to the standard
2.4 Simulation Study on pH Neutralization Process 51

7.5
A
B
7
pH

Setpoint
PID
6.5
PI λ D μ
SWPI-D
SWPI λ -Dμ
6
0 50 100 150 200 250 300 350 400 450 500
(a) Time (s)

40
D
Control signal (u(t))

30

C
20
PID
10 PI λ D μ
SWPI-D
SWPI λ -Dμ
0
0 50 100 150 200 250 300 350 400 450 500
(b) Time (s)

Fig. 2.34 Performance comparison of industrial SWPIλ Dμ controller for set-point tracking and
disturbance rejection

form. The derivative kick effect of the compared controllers is shown in Fig. 2.36.
From the responses, it can be seen that for the set-point change SWPIλ -Dμ performed
better than the other controllers. This is clearly seen in the zoomed plot of regions
of interest A and B in Fig. 2.37.
The derivative kick effect observed in the control signals of both PID and PIλ Dμ
at set-point change is reduced in SWPI-D and SWPIλ -Dμ . This is further highlighted
in the zoomed plot of regions of interest C and D in Fig. 2.37. From this figure, the
control signals of PID and PIλ Dμ are very narrow, sharp and both reached a peak of
100%. On the other hand, the control signals of both SWPI-D and SWPIλ -Dμ peaked
at around 25%.

2.4.5 Performance of PIλ -PDμ Controller

This subsection presents the performance evaluation of the proposed PIλ -PDμ
controller developed in Sect. 2.3 on real-time pH neutralization process given in
Sect. 2.4.2. First, the design of PIλ -PDμ controller parameters will be presented.
52 2 Fractional-order Set-Point Weighted Controllers

(a) (b)
7
7.2
6.8
pH

7
6.6

6.8 6.4

20 40 60 80 100 250 300 350

(c) (d)
25
Control signal (u(t))

20 30

15 20

10 10

5
20 40 60 80 100 250 300 350
Time (s) Time (s)

Fig. 2.35 Regions of interest A, B, C and D of Fig. 2.34

Table 2.13 Performance analysis of industrial form of controllers

Controller tr ts1 ts2 %OS J I T AE
PID 20.8046 119.3300 405.7697 2.9329 0.9797
PIλ Dμ 19.3370 83.9811 397.1517 2.8758 0.9294
SWPI-D 23.8224 28.6699 404.9420 0.9732 0.9126
SWPIλ -Dμ 22.4501 26.3520 395.4278 0.4395 0.8322

Then, the results of the proposed controllers will be presented in three phases. In the
first phase, the performance of the integer-order PI-PD controller on the real-time
pressure process will be given. Then in the second phase, the performance of the
fractional-order PI-PD controller will be presented. Lastly, a comparison between
the integer and fractional-order PI-PD controllers will be presented.
In all these cases, as mentioned earlier, the performance has been evaluated for
set-point tracking, variation in set-point and disturbance rejection will be presented.
Furthermore, the numerical analysis will be done in terms of rise time (tr ), settling
time before and after disturbance (ts1 and ts2 ) and percentage overshoot (%OS).
Furthermore, it should be noted that a disturbance (D(s)) of 25% is injected at
2.4 Simulation Study on pH Neutralization Process 53

7.5

7
A B
Setpoint
pH

6.5
PID
PI λ D μ
6 SWPI-D
SWPI λ -Dμ
5.5
0 100 200 300 400 500 600 700
(a) Time (s)

100
PID D
Control signal (u(t))

80 PI λ D μ
SWPI-D
60 SWPI λ -Dμ
40
C
20

0
0 100 200 300 400 500 600 700
(b) Time (s)

Fig. 2.36 Performance comparison of industrial SWPIλ Dμ controller for variable set-point tracking

250 s for the purpose of disturbance rejection evaluation. Moreover, the ability of the
controllers to track variation in set point will also be analyzed.

2.4.5.1 Controller Parameters

The tuned controller parameters of PIλ -PDμ given in Sect. 2.3 can be obtained from
the designed SWPIλ Dμ controller in Sect. 2.4.4 using the conversion relations given
in Appendix A and Fig. A.1. Therefore, the parameters of PID, PI-PD and PIλ -PDμ
controllers are given in Table 2.14.

2.4.5.2 PI-PD Controllers

The comparison of the response of the system with various integer-order controllers
as given in Sect. 2.3.1 will be presented here. Thus, the performance of PID, PI-PD1
and PI-PD2 in Eqs. (2.4), (2.48) and (2.52) respectively for set-point tracking and
disturbance rejection is given in Fig. 2.38. The zoomed-in view of regions of interest
54 2 Fractional-order Set-Point Weighted Controllers

(a) (b)
6.8 6.6

6.6
6.4
6.4
pH

6.2
6.2

6 6

250 300 350 500 520 540 560 580

(c) (d)
15
Control signal (u(t))

80
10
60
5
40
0
20
-5
250 300 350 500 520 540 560 580
Time (s) Time (s)

Fig. 2.37 Regions of interest A, B, C and D of Fig. 2.36

Table 2.14 Controller parameters of PI-PD controllers

Controller P I D λ μ
PID 39.7059 30.9332 0.6790 – –
PI-PD 39.1697 30.9332 0.0173 – –
PIλ -PDμ 39.1697 30.9332 0.0173 0.9767 0.5811

A, B, C and D of the figure is highlighted in Fig. 2.39 while the numerical assessment
of the figure is given in Table 2.15.
Observing the Figs. 2.38 and 2.39 and the Table 2.15, it can be seen that the set-
point tracking ability of PI-PD2 controller is outperformed those of PID and PI-PD1.
The PI-PD1 and PI-PD2 controllers have the fastest rise time and settling time of
16.8577 and 57.0507 s respectively. Furthermore, the controller produced the least
overshoot of 2.2464%. However, for disturbance rejection, the performance of all
the compared controllers are satisfactory with PID has the fastest settling time of
370.9301 s.
Observing the control signals of all the compared controllers, it can be seen that
the PI-PD1 produced smoother action as compared to the oscillatory actions of both
PID and PI-PD2. The smooth control action is due to the structure of the PI-PD.
2.4 Simulation Study on pH Neutralization Process 55

7.5
A
B
7
pH

Setpoint
6.5 PID
PI-PD1
PI-PD2
6
0 50 100 150 200 250 300 350 400 450 500
(a) Time (s)

40
D
Control signal (u(t))

30

C
20
PID
10 PI-PD1
PI-PD2
0
0 50 100 150 200 250 300 350 400 450 500
(b) Time (s)

Fig. 2.38 Performance comparison of PI-PD controllers for set-point tracking and disturbance
rejection

Similarly, the performance comparison of the PID, PI-PD1 and PI-PD2 controllers
for variable set-point tracking is given in Fig. 2.40 while the regions of interest are
further highlighted in Fig. 2.41. From the figures, it can be observed that, during step
change, PID controller experiences a very narrow, sharp and peak of 100% while both
the PI-PD controllers reached a peak of around 30%. Therefore, it can be concluded
that the derivative kick effect of the PID controller is avoided with the use of PI-PD
controllers.

2.4.5.3 PIλ -PDμ Controllers

In a similar way to the performance of integer-order controllers presented in

Sect. 2.4.5.2, the comparison of the response of the system with fractional-order
controllers as given in Sect. 2.3.2 will be analyzed. Therefore, the performance of
PIλ Dμ , PIλ -PDμ 1 and PIλ -PDμ 2 given in Eqs. (2.1), (2.58) and (2.62) respectively
for set-point tracking and disturbance rejection is given in Fig. 2.42. The zoomed-in
view of regions of interest A, B, C and D of the figure is highlighted in Fig. 2.43
while the numerical observations are given in Table 2.16.
56 2 Fractional-order Set-Point Weighted Controllers

(a) (b)
7
7.2
6.8
pH

7
6.6
6.8
6.4
0 50 100 250 300 350
(c) (d)
20
Control signal (u(t))

30
15

20
10

10
5
20 40 60 80 100 250 300 350
Time (s) Time (s)

Fig. 2.39 Regions of interest A, B, C and D of Fig. 2.38

Table 2.15 Performance analysis of integer-order controllers

Controller tr ts1 ts2 %OS
PID 17.0742 111.5825 370.9301 4.5570
PI-PD1 17.5341 95.5005 372.1216 3.3269
PI-PD2 16.8577 57.0507 371.8033 2.2464

From the figures and the table, it can be seen that unlike in the case of integer-
order where PI-PD2 controller outperformed the other controllers, the performance
of the two fractional PI-PD controllers (PIλ -PDμ 1 and PIλ -PDμ 2) outperformed that
of PIλ Dμ . Observing the table, the two PIλ -PDμ controllers have each produced %OS
of around 2.2% while that of the PIλ Dμ is 4.4836%.
Furthermore, the PIλ -PDμ 2 controller settled faster at 52.1932 s when compared to
71.4198 and 101.4690 s of PIλ -PDμ 1 and PIλ Dμ controllers respectively. However,
in terms of rise time, the PIλ Dμ controller has the fastest rise time of 15.8972 s.
Furthermore, the controller also settled faster after a disturbance with a settling time
of 363.2637 s.
2.4 Simulation Study on pH Neutralization Process 57

7.5 Setpoint
PID
PI-PD1
7 PI-PD2
A B
pH

6.5

5.5
0 100 200 300 400 500 600 700
(a) Time (s)

PID
100 PI-PD1
Control signal (u(t))

D PI-PD2
80

60

40
C
20

0
0 100 200 300 400 500 600 700
(b) Time (s)

Fig. 2.40 Performance comparison of PI-PD controllers for variable set-point tracking

Similarly, the performance comparison of the various controller for variable set-
point tracking is given in Fig. 2.44 while the regions of interest are further highlighted
in Fig. 2.45. From the figure, it can be observed that, during step change, PIλ Dμ
controller experiences a sharp peak of 40% while the PIλ -PDμ controllers reached a
peak of around 30% as in the PI-PD controllers.

2.4.5.4 Comparison of PI-PD and PIλ -PDμ Controllers

The comparison of the response of the system with various integer-order controllers
(PID, PI-PD1 and PI-PD2) as given in Sect. 2.3.1 and fractional-order controllers
(PIλ Dμ , PIλ -PDμ 1 and PIλ -PDμ 2) as given in Sect. 2.3.2 for set-point tracking and
disturbance rejection is given in Fig. 2.46. The regions of interest A, B, C and D of
the figure are further highlighted in Fig. 2.47. The numerical assessment of the figure
in terms of step response characteristics is given in Table 2.17.
Observing the figures, it can be seen that the proposed fractional-order PI-PD
controllers performed better compared to all other integer-order controllers. From
the numerical analysis, it can be seen that both the proposed PIλ -PDμ controllers have
58 2 Fractional-order Set-Point Weighted Controllers

(a) (b)
6.8 6.6

6.6
6.4
6.4
pH

6.2
6.2

6 6

250 300 350 500 520 540 560 580

(c) (d)
15
Control signal (u(t))

80
10
60

40
5
20
0
250 300 350 500 520 540 560 580
Time (s) Time (s)

Fig. 2.41 Regions of interest A, B, C and D of Fig. 2.40

produced the least overshoots of around 2.2%. Furthermore, the PIλ -PDμ 2 controller
settled faster at 52.1932 s. However, in terms of rise time, the PIλ Dμ controller has
the fastest tr of 15.8972 s.
Furthermore, the comparison of the response for disturbance rejection shows that
all the compared controllers are satisfactory with PIλ Dμ controller has the fastest
settling time of 363.2637 s. Observing the control signals, it can be seen that the
proposed controllers have a smoother control action compared to the oscillatory
actions of integer-order controllers.
Similarly, the comparison of the response of the system for variable set-point
tracking is given in Fig. 2.48 while the regions of interest A, B, C and D of the
figure are further highlighted in Fig. 2.49. From the figures, it can be seen that during
set-point change both the integer-order and fractional-order PI-PD controllers are
free from derivative kick effects. On the other hand, the PID and PIλ Dμ controllers
experience very sharp narrow peaks of 100% and 40% respectively.
2.4 Simulation Study on pH Neutralization Process 59

7.5
A
B
7
pH

Setpoint
6.5 PI D
PI -PD 1
PI -PD 2
6
0 50 100 150 200 250 300 350 400 450 500
(a) Time (s)

40
D
Control signal (u(t))

30

C
20 PI D
PI -PD 1
PI -PD 2
10

0
0 50 100 150 200 250 300 350 400 450 500
(b) Time (s)

Fig. 2.42 Performance comparison of PIλ -PDμ controllers for set-point tracking and disturbance
rejection

2.4.6 Comparison of SWPIλ Dμ and PIλ -PDμ Controllers

This section presents the comparison of both the proposed SWPIλ Dμ and PIλ -PDμ
controllers as given in Sects. 2.2.3 and 2.3 on the pH neutralization process plant given
in Sect. 2.4.2 is presented. First, the performance of all compared controllers will be
evaluated for set-point tracking and disturbance rejection. Then, the performance
will be evaluated for all the controllers to track variation in set point.

2.4.6.1 Set-Point Tracking and Disturbance Rejection

The controller parameters of various compared controllers is shown in Table 2.18.

Thus, the set-point tracking and disturbance rejection performance of the system
with various controllers including PID, SWPI-D, PI-PD, PIλ Dλ , SWPIλ -Dμ and PIλ -
PDμ is shown in Fig. 2.50. The regions of interest A, B, C and D of the figure are
further highlighted in Fig. 2.51. Observing the figures, it can be seen that the proposed
fractional-order controllers (SWPIλ -Dμ , PIλ -PDμ 1 and PIλ -PDμ 2) performed better
compared to all other controllers.
60 2 Fractional-order Set-Point Weighted Controllers

(a) (b)
7
7.2
6.8
pH

7
6.6
6.8
6.4
0 50 100 250 300 350

(c) (d)
20
Control signal (u(t))

30
15

20
10

10
5
20 40 60 80 100 250 300 350
Time (s) Time (s)

Fig. 2.43 Regions of interest A, B, C and D of Fig. 2.42

Table 2.16 Performance analysis of fractional-order controllers

Controller tr ts1 ts2 %OS
PIλ Dμ 15.8972 101.4690 363.2637 4.4836
PIλ -PDμ 1 17.1534 71.4198 373.4480 2.2243
PIλ -PDμ 2 16.8406 52.1932 372.2233 2.2045

The numerical assessment of the figure for step-response characteristics is given

in Table 2.19. From the table, it can be noted that the proposed controllers SWPIλ -Dμ ,
PIλ -PDμ 1 and PIλ -PDμ 2 has the least overshoots of 0.4805, 2.2243 and 2.2045%
respectively. Furthermore, it can also be observed that the controller SWPIλ -Dμ and
PIλ -PDμ 2 also has the fastest settling times of 20.7780 and 52.1932 s respectively.
However, in terms of rise time, the fractional-order controllers PIλ Dμ and PIλ -PDμ 2
has the fastest rise time of 15.8972 and 16.8406 s respectively.
On the other hand, the disturbance rejection performance of all the compared
controllers is satisfactory. The shorter settling times 363.5675 and 363.5095 s of the
PIλ Dμ and SWPIλ -Dμ controllers after the disturbance compared to 370.9695 and
370.9194 s of PID and SWPI-D is an indication of the good load regulation ability
of the fractional-order controllers.
2.4 Simulation Study on pH Neutralization Process 61

Setpoint
7.5 PI D
PI -PD 1
7 PI -PD 2
A B
pH

6.5

5.5
0 100 200 300 400 500 600 700
(a) Time (s)

60
Control signal (u(t))

PI D
PI -PD 1
40 PI -PD 2 D

C
20

0
0 100 200 300 400 500 600 700
(b) Time (s)

Fig. 2.44 Performance comparison of PIλ -PDμ controllers for variable set-point tracking

It is worth noting that the disturbance rejection controller for PID and SWPI-D is
the same while that of the PIλ Dμ and SWPIλ -Dμ is the same. Hence, the closeness
of ts2 of around 370 s for PID and SWPI-D on one hand and PIλ Dμ and SWPIλ -Dμ
of around 363 s on the other hand. Furthermore, from the zoomed figures C and D,
it can be seen that the control actions of the proposed fractional-order controllers
SWPIλ -Dμ , PIλ -PDμ 1 and PIλ -PDμ 2 is smooth compared to the oscillatory signals
of PID, PI-PD1, PI-PD2, and SWPI-D.

2.4.6.2 Variable Set-Point Tracking

To observe the derivative kick effect of all the compared controllers including PID,
SWPI-D, PI-PD, PIλ Dλ , SWPIλ -Dμ and PIλ -PDμ , the plant is simulated to a variable
set-point signal and the result is shown in Fig. 2.52. The zoomed-in view for regions
of interest A, B, C and D of the figure is given in Fig. 2.53. From the responses, it can
be seen that during the set-point change, the performance of proposed fractional-
order controllers is better when compared to the other controllers. This is clearly
seen in the zoomed plot A and B in Fig. 2.53.
62 2 Fractional-order Set-Point Weighted Controllers

(a) (b)
6.8 6.6

6.6
6.4
6.4
pH

6.2
6.2

6 6

250 300 350 500 520 540 560 580

(c) (d)
15
Control signal (u(t))

40

10 30

5 20

10
0
250 300 350 500 520 540 560 580
Time (s) Time (s)

Fig. 2.45 Regions of interest A, B, C and D of Fig. 2.44

Additionally, during the set-point change, the derivative kick effect observed in
the control signals of both PID and PIλ Dμ controllers is reduced in SWPI-D, PI-PD,
SWPIλ -Dμ and PIλ -PDμ controllers. This is visible from the zoomed plot of regions
C and D in Fig. 2.53. As seen in the zoomed figure, the control signals of PID and
PIλ Dμ are both very narrow and sharp and reached a peak of 100% and around 40%,
respectively. On the other hand, the peak of the control signals of SWPI-D, PI-PD
and SWPIλ -Dμ is about 25%.

2.5 Simulation Study on Pressure Process Plant

In the first part of this section, a description of the pressure process plant is given with
the help of P&ID while the second part gives the schematics of the entire experimental
setup. Then, the subsequent three subsections present the performance evaluation of
the designed SWPIλ Dμ and PIλ -PDμ controllers on the pressure process plant.
2.5 Simulation Study on Pressure Process Plant 63

7.5
A

B
7
pH

Setpoint
PI-PD1
6.5 PI-PD2
PI -PD 1
PI -PD 2
6
0 50 100 150 200 250 300 350 400 450 500
(a) Time (s)

40
D
Control signal (u(t))

30

C
20 PI-PD1
PI-PD2
PI -PD 1
10 PI -PD 2

0
0 50 100 150 200 250 300 350 400 450 500
(b) Time (s)

Fig. 2.46 Performance comparison of PI-PD and PIλ -PDμ controllers for set-point tracking and
disturbance rejection

2.5.1 PI&D of Pressure Process Plant

The real-time pressure process plant used in the experiment is shown in Fig. 2.54
while the P&ID of process representing the flow is shown in Fig. 2.55 [109]. The
plant consists of a buffer tank VL 202 which is connected to a centralized air com-
pressor through a hand valve HV 202 and a process control valve PCV 202. The
characteristics of the electro-pneumatic control valve PCV 202 is shown in Fig. 2.56.
From the characteristics, it can be observed that the type of behavior is a nonlinear
increasing sensitivity type and the maximum input air pressure is around 6 bar. Thus,
this will guide in the choice of the set point.
Furthermore, a pressure transmitter PT 202 is used to measure the pressure in the
tank. The pressure readings from PT 202 are fed into pressure indicating controller
PIC 202 in terms of voltage in the range of 0–5 V. The control signal is used to control
the opening of PCV 202 to adjust the pressure in accordance with the set-point. On
the other side of the tank, the downstream process valve is set at a fixed opening to
allow a constant flow from the pressure tank during the experiment.
64 2 Fractional-order Set-Point Weighted Controllers

(a) (b)
7
7.2
6.8
pH

7
6.6
6.8
6.4
0 50 100 250 300 350
(c) (d)
20
Control signal (u(t))

30
15

20
10

10
5
20 40 60 80 100 250 300 350
Time (s) Time (s)

Fig. 2.47 Regions of interest A, B, C and D of Fig. 2.46

Table 2.17 Performance analysis of PI-PD and PIλ -PDμ controllers

Category Controller tr ts1 ts2 %OS
Integer-order controllers PID 17.0742 111.5825 370.9301 4.5570
PI-PD1 17.5341 95.5005 372.1216 3.3269
PI-PD2 16.8577 57.0507 371.8033 2.2464
Fractional-order controllers PIλ Dμ 15.8972 101.4690 363.2637 4.4836
PIλ -PDμ 1 17.1534 71.4198 373.4480 2.2243
PIλ -PDμ 2 16.8406 52.1932 372.2233 2.2045

2.5.2 Schematics of Experimental Setup

The schematic diagram of the complete experimental setup is shown in Fig. 2.57.
The proposed control strategy will be implemented in the host PC using MAT-
LAB/Simulink software which is accessed through ‘Remote Desktop Connection’ in
the control room. As in the case of the pH neutralization process plant, here also the
host PC is interfaced with the pilot plant through Peripheral Component Interconnect
(PCI) cards 1713U, 1720U and 1751.
2.5 Simulation Study on Pressure Process Plant 65

Setpoint
7.5 PI-PD1
PI-PD2
7 PI -PD 1
A B PI -PD 2
pH

6.5

5.5
0 100 200 300 400 500 600 700
(a) Time (s)

60
PI-PD1
Control signal (u(t))

PI-PD2
40 PI -PD 1
D
PI -PD 2

20 C

0
0 100 200 300 400 500 600 700
(b) Time (s)

Fig. 2.48 Performance comparison of PI-PD and PIλ -PDμ controllers for variable set-point tracking

The PCI-1713U is a 100 kS/s, 12-bit, 32-ch isolated AI card used to access signal
from PT 202. On the other hand, the PCI-1720U; a 12-bit, 4-channel isolated AO
card, is used to send the control signal to PCV 202. The last card PCI-1751 is a
48-channel digital I/O and 3-channel counter card used to operate the process plant
either Remotely or Locally (R/L).

2.5.3 Performance of SWPIλ Dμ Controller

This subsection presents the performance evaluation of the proposed SWPIλ Dμ con-
troller presented in Sect. 2.2.3 on pressure process plant in Sect. 2.5. First, the design
of SWPIλ Dμ controller parameters will be presented. Then, the results of proposed
parallel and series configuration of the designed SWPIλ Dμ controller in comparison
with PID, PIλ Dμ and SWPI-D will be presented.
In all these cases, the performance has been evaluated for set-point tracking,
variation in set-point and disturbance rejection will be presented. Furthermore, the
numerical analysis will be done in terms of rise time (tr ), settling time before and
66 2 Fractional-order Set-Point Weighted Controllers

(a) (b)
6.8 6.6

6.6
6.4
6.4
pH

6.2
6.2

6 6

250 300 350 500 520 540 560 580

(c) (d)
15 30
Control signal (u(t))

25
10
20

15
5
10

0 5
250 300 350 500 520 540 560 580
Time (s) Time (s)

Fig. 2.49 Regions of interest A, B, C and D of Fig. 2.48

Table 2.18 Controller parameters of proposed controllers on pH neutralization process

Category Controller Kp Ki Kd b λ μ
Integer-order controllers PID 39.7059 30.9332 0.6790 – – –
PI-PD1 39.1697 30.9332 0.0173 – – –
PI-PD2 39.1697 30.9332 0.0173 – – –
SWPI-D 39.7059 30.9332 0.6790 0.9516 – –
Fractional-order controllers PIλ Dμ 39.7059 30.9332 0.6790 – 0.9767 0.5811
PIλ -PDμ 1 39.1697 30.9332 0.0173 – 0.9767 0.5811
PIλ -PDμ 2 39.1697 30.9332 0.0173 – 0.9767 0.5811
SWPIλ -Dμ 39.7059 30.9332 0.6790 0.9516 0.9767 0.5811
2.5 Simulation Study on Pressure Process Plant 67

7.5
A
B
7
pH

6.5

6
0 50 100 150 200 250 300 350 400 450 500
(a) Time (s)

40
D
Control signal (u(t))

30 Setpoint
PID
C
20 PI D
SWPI-D
PI-PD2
10 PI-PD1
SWPI -D
0 PI -PD 2
0 50 100 150 200 250 300 350 400 PI450
-PD 1 500
(b) Time (s)

Fig. 2.50 Performance comparison of SWPIλ Dμ and PIλ -PDμ controllers for set-point tracking
and disturbance rejection

after disturbance (ts1 and ts2 ) and percentage overshoot (%OS). Furthermore, it should
be noted that a disturbance (D(s)) of 25% is injected at 500 s for the purpose of
disturbance rejection evaluation. Moreover, the ability of the controllers to track
variation in set point will also be analyzed.

2.5.3.1 Controller Parameters

The conventional PID controller parameters (K p , K i , and K d ) are tuned using the
Ziegler-Nichols (ZN) tuning approach. Therefore, the sustained uniform oscillations
of PT202 for an ultimate gain of K u = 51 is shown in Fig. 2.58. From the response,
the ultimate period (Pu ) is calculated as 20 s.
In order to select the best values of set-point weighting and fractional-order param-
eters (b, λ and μ), several values of these parameters within the range 0 to 1 are used.
Therefore, the effect on variation of b on the plant performance given in Fig. 2.59
shows that as the value of b decreases the %OS decreases but increases in tr . This
effect is shown numerically in Table 2.20. Furthermore, it can be seen from the con-
68 2 Fractional-order Set-Point Weighted Controllers

Fig. 2.51 Regions of interest A, B, C and D of Fig. 2.50

Table 2.19 Performance analysis of proposed controllers on pH neutralization process

Category Controller tr ts1 ts2 %OS
Integer-order controllers PID 17.0750 110.1329 370.9695 4.9182
PI-PD1 17.5341 95.5005 372.1216 3.3269
PI-PD2 16.8577 57.0507 371.8033 2.2464
SWPI-D 19.5630 22.1078 370.9194 0.9536
Fractional-order controllers PIλ Dμ 15.8972 101.4693 363.5675 4.4836
PIλ -PDμ 1 17.1534 71.4198 373.4480 2.2243
PIλ -PDμ 2 16.8406 52.1932 372.2233 2.2045
SWPIλ -Dμ 18.8494 20.7780 363.5095 0.4805
2.5 Simulation Study on Pressure Process Plant 69

7.5

7
A B
pH

6.5

5.5
0 100 200 300 400 500 600 700
Setpoint
(a) Time (s) PID
PI D
SWPI-D
Control signal (u(t))

100 PI-PD2
D
PI-PD1
SWPI -D
50 PI -PD 2
PI -PD 1
C

0
0 100 200 300 400 500 600 700
(b) Time (s)

Fig. 2.52 Performance comparison of SWPIλ Dμ and PIλ -PDμ controllers for variable set-point
tracking

trol signal that as b decreases the proportional kick effect decreases. Hence, the best
value of b is selected as 0.8.
Observing the effect on the variation of λ on the plant performance as given in
Fig. 2.60 and Table 2.21, it can be seen that the maximum value of λ for the steady-
state is 1.0 while the minimum value is 0.8. Values lower than the minimum leads
to high steady-state error. Similarly, from Fig. 2.61, it can be clearly seen that as the
values of μ drips away from unity, the performance of the system in terms of tr ,
ts and %OS improves. This also leads to reduced oscillation in the control signal.
Hence, the best values for λ and μ are chosen as 0.95 and 0.5 respectively.

2.5.3.2 Parallel SWPIλ Dμ Controller

The parameters of compared parallel configurations of PID, PIλ Dμ , SWPID and

SWPIλ Dμ controllers obtained from the previous Sect. 2.5.3.1 are given in Table 2.22.
The comparison of the responses of the system with these PID, PIλ Dμ , SWPID and
SWPIλ Dμ controllers for set-point tracking ability, disturbance rejection capability,
and smoother control action are given in Fig. 2.62. For more clarity regions A, B, C,
and D of the figure are highlighted in Fig. 2.63. Furthermore, the numerical analysis
of the controller performance is also given in Table 2.23.
70 2 Fractional-order Set-Point Weighted Controllers

Fig. 2.53 Regions of interest A, B, C and D of Fig. 2.52

Outlet gas to
downstream process Digital I/O Card
PCI-1751

Process Control AI Card

Valve PCI-1713U
PCV 202
AO Card
Inlet gas from PCI-1720U
centralized
compressor I/O Interface
Board
Hand Valve
HV 202

Buffer Tank
VL 202

Pressure Transmitter MATLAB Server

PT 202 (HOST PC)

Fig. 2.54 Experimental setup of real-time pressure process plant

2.5 Simulation Study on Pressure Process Plant 71

Fig. 2.55 P&ID of pressure process plant

4
Pressure (bar)

0
0 10 20 30 40 50 60 70 80 90 100
Valve Opening (%)

Fig. 2.56 Characteristics of electro-pneumatic control valve PCV 202

72 2 Fractional-order Set-Point Weighted Controllers

Fig. 2.57 Schematic diagram of the complete experimental setup

3.1

3.05

2.95
Pressure (bar)

2.9

2.85

2.8

2.75

2.7

2.65

2.6
400 450 500 550 600 650 700 750 800
Time (s)

Fig. 2.58 Sustained uniformed oscillation of PT202 for ZN closed-loop tuning

From both figures and table, it can be seen that the SWPIλ Dμ controller produces
the least %OS of 3.6283% as compared to 8.2935, 5.1853, 3.9360% of the PID,
SWPID, and PIλ Dμ respectively. From the settling time (ts1 ) of the controllers, it
can be seen that at 83.2167 s the proposed approach settles faster than the other con-
trollers. However, in terms of rise-time (tr ), the proposed controller came third with
a value of 7.3698 s. This indicated that there is a trade-off between good overshoot
and settling time in one hand and speed of response (rise time) on the other.
The disturbance rejection of all controllers PID, PIλ Dμ , SWPID, and SWPIλ Dμ
controllers were found to be satisfactory with the SWPID having the fastest recovery
with ts2 of 537.9967 s and PIλ Dμ having the slowest recovery of 582.2311 s. Observ-
ing the control signals of all controllers compared, it can be seen that the proposed
SWPIλ Dμ generates smoother signal compared to the oscillatory signals of others.
2.5 Simulation Study on Pressure Process Plant 73

3
Pressure (bar)

0 Setpoint
0 20 40 60 80 100 120 140 160 180 200
b=1.0
(a) Time (s)
b=0.9
100 b=0.8
b=0.7
Control Signal

b=0.6
b=0.5
50 b=0.4

0
0 20 40 60 80 100 120 140 160 180 200
(b) Time (s)

Fig. 2.59 Effect of parameter b on the plant performance

Table 2.20 Numerical assessment on effect of variation of b

Parameter Value tr ts %OS
b 1.0 6.4044 95.8871 10.9402
0.8 8.8206 104.2408 5.4508
0.6 12.9751 108.0727 5.0218
0.4 16.2399 104.5702 5.0060

To evaluate the ability of various controllers for variable set-point tracking, the
plant with various controllers is made to track variable reference signal as shown in
Fig. 2.64. For more clarity, regions A, B, C and D of the figure are further highlighted
in Fig. 2.65.
From both figures, it can be observed during set-point change (i.e., at 500 s),
the two fractional-order controllers (i.e., PIλ Dμ and SWPIλ Dμ ) produced smaller
overshoot as against those of two integer-order controllers (i.e., PID and SWPID).
Observing the control signals of the controllers, it can be seen that PIλ Dμ and PID
have higher derivative kick effect making the signal to reach 80% of the valve open-
ing. However, this effect is significantly reduced in the SWPID and SWPIλ Dμ .
74 2 Fractional-order Set-Point Weighted Controllers

Pressure (bar)

2.5

2
20 40 60 80 100 120 140 160 180 200
(a) Time (s)

Setpoint
100 =1
Control Signal (%)

=0.975
80 =0.95
=0.925
=0.9
60
=0.8

40

20 40 60 80 100 120 140 160 180 200

(b) Time (s)

Fig. 2.60 Effect of parameter λ on the plant performance

Table 2.21 Numerical assessment on effect of variation of fractional-order controller parameters

λ and μ
Parameter Value tr ts %OS
λ 1.0 6.4911 95.3112 5.0151
0.975 6.6421 77.9867 4.4215
0.95 6.6913 79.5911 3.2994
0.925 6.9599 72.2900 3.4077
0.9 17.8056 70.0560 2.6667
0.8 32.7617 198.1350 0.2777
μ 0.9 7.0383 88.4318 5.4255
0.7 7.1283 99.3004 5.0350
0.5 6.9372 93.6639 4.6373
0.3 6.7716 85.7019 4.7881

2.5.3.3 Series SWPIλ Dμ Controller

The controller parameters of compared series configuration of PID in Eq. (2.22),

PIλ Dμ in Eq. (2.3), SWPID in Eq. (2.24) and SWPIλ Dμ in Eq. (2.25) controllers
2.5 Simulation Study on Pressure Process Plant 75

Pressure (bar)

3
Setpoint
= 0.9
2.5 = 0.7
= 0.5
= 0.3
2
0 50 100 150 200 250 300
(a) Time (s)

100
Control Signal

50

0
0 50 100 150 200 250 300
(b) Time (s)

Fig. 2.61 Effect of parameter μ on the plant performance

Table 2.22 Controller parameters for parallel form of controllers

Controller Kp Ki Kd α b λ μ
PID 30.0 3.0 1.0 0.025 – – –
PIλ Dμ 30.0 3.0 1.0 0.025 – 0.95 0.50
SWPI-D 30.0 3.0 1.0 0.025 0.80 – –
SWPIλ -Dμ 30.0 3.0 1.0 0.025 0.80 0.95 0.50

will be obtained using the parallel configuration of SWPIλ Dμ previous Sect. 2.5.3.2.
Therefore, based on the tuning procedure reported in Appendix A and Fig. A.1, the
parameters of the series PID’s are given in Table 2.24.
The responses of the system with the compared controllers for set-point tracking
ability and disturbance rejection capability are given in Fig. 2.66. The regions of
interest A, B, C, and D are further highlighted in Fig. 2.67 while the numerical
results are given in Table 2.25.
From both figures and table, it can be seen that the SWPIλ Dμ controller produces
the least %OS of 2.9943% as compared to 12.5007, 9.1338, 4.1832% of the PID,
SWPID and PIλ Dμ respectively. Observing the settling times ts1 and ts2 of all the
compared controllers, it can be seen that at 33.4194 s the proposed approach settled
76 2 Fractional-order Set-Point Weighted Controllers

3
Pressure (bar)

B Setpoint
2 PID
A
PI D
1 SWPI-D
SWPI -D
0
0 100 200 300 400 500 600 700 800 900 1000
(a) Time (s)

100
PID
Control Signal

PI D
SWPI-D
D
50 SWPI -D

0
0 100 200 300 400 500 600 700 800 900 1000
(b) Time (s)

Fig. 2.62 Performance comparison of parallel SWPIλ Dμ controller for set-point tracking and
disturbance rejection

much faster than the others. However, in terms of tr , the proposed SWPIλ Dμ con-
troller came fourth with a value of 17.2004 s. The results here followed the similar
pattern to the parallel configuration of SWPIλ Dμ controller in Sect. 2.5.3.2.
A noticeable improvement in the performance of the series configuration of
SWPIλ Dμ controller compared to the parallel is that of the faster settling time of
33.4194 s and less overshoot of 2.9943% as against of 83.2167 s and 3.6283% of the
parallel configuration of SWPIλ Dμ controller.
The SWPID has the fastest recovery from disturbance effect settling at 541.8869 s
while PIλ Dμ has the slowest recovery settling at 841.0198 s. However, the disturbance
rejection ability of all controllers was found to be satisfactory. Observing the con-
trol signals of all controllers, it can be seen that the proposed SWPIλ Dμ generates
smoother signal compared to the oscillatory signals of others.
The performance comparison of various controllers PID, PIλ Dμ , SWPID and
SWPIλ Dμ for variable set-point tracking has been evaluated in the same as to the
previous section. The plant with various controllers is made to track the variable
reference signal as shown in Fig. 2.68. The regions A, B, C and D of the figure are
highlighted in Fig. 2.69.
From the two figures, it can be observed that, during the set-point change, the two
fractional-order controllers (i.e., PIλ Dμ and SWPIλ Dμ ) produced smaller overshoot
2.5 Simulation Study on Pressure Process Plant 77

(a) (b)
3.2
Pressure (bar)

3 3

2.8
2.5
2.6

2 2.4
0 50 100 150 200 500 550 600

(c) (d)
100
40
Control Signal

80
30

60
20

40 10

0 50 100 150 200 500 550 600

Time (s) Time (s)

Fig. 2.63 Regions of interest A, B, C and D of Fig. 2.62

Table 2.23 Performance analysis for parallel form of controllers

Controller tr ts1 ts2 %OS
PID 6.7511 98.1610 544.3894 8.2935
PIλ Dμ 6.9613 90.5565 582.2311 3.9360
SWPI-D 8.4787 94.5839 537.9967 5.1853
SWPIλ -Dμ 7.3698 83.2167 553.4759 3.6283

as against those of PID and SWPID as in the case of the parallel configuration.
Observing the control signals of the controllers, it can be seen that the PID has a
higher derivative kick effect reaching 100% of the valve opening. On the other hand,
this effect is significantly reduced in the PIλ Dμ , SWPID and SWPIλ Dμ .
78 2 Fractional-order Set-Point Weighted Controllers

4
Pressure (bar)

3 Setpoint
B PID
2
A PI D
1 SWPI-D
SWPI -D
0
0 100 200 300 400 500 600 700 800 900 1000
(a) Time (s)

PID
100
PI D
Control Signal

SWPI-D
SWPI -D
50

C D

0
0 100 200 300 400 500 600 700 800 900 1000
(b) Time (s)

Fig. 2.64 Performance comparison of parallel SWPIλ Dμ controller for variable set-point tracking

2.5.4 Performance of PIλ -PDμ Controller

In this subsection, first, the design of PIλ -PDμ controller parameters based will
be presented on the tuning procedure given in Appendix A. Then, the results of
the proposed controllers will be presented in three phases. In the first phase, the
performance of the integer-order PI-PD controller given in Sect. 2.3.1 on the real-
time pressure process will be given. Then in the second phase, the performance of
the fractional-order PI-PD controller given in Sect. 2.3.2 will be presented. Lastly,
a comparison between the integer and fractional-order PI-PD controllers will be
presented.
In all these cases, as mentioned earlier, the performance has been evaluated for
set-point tracking, variation in set-point and disturbance rejection will be presented.
Furthermore, the numerical analysis will be done in terms of rise time (tr ), settling
time before and after disturbance (ts1 and ts2 ) and percentage overshoot (%OS).
Furthermore, it should be noted that a disturbance (D(s)) of 25% is injected at
500 s for the purpose of disturbance rejection evaluation. Moreover, the ability of the
controllers to track variation in set point will also be analyzed.
2.5 Simulation Study on Pressure Process Plant 79

(a) (b)
4
Pressure (bar)

3.5
2.5

3
2
0 50 100 150 200 500 550 600
(c) (d)
100 100
Control Signal

80 80

60 60

40 40

0 50 100 150 200 500 550 600

Time (s) Time (s)

Fig. 2.65 Regions of interest A, B, C and D of Fig. 2.64

Table 2.24 Controller parameters for series form of controllers

Controller Kp Ti Td α b λ μ
PID 29.8995 9.9665 0.0585 0.4271 – – –
PIλ Dμ 29.8995 9.9665 0.0585 0.4271 – 0.95 0.50
SWPI-D 29.8995 9.9665 0.0585 0.4271 0.8028 – –
SWPIλ -Dμ 29.8995 9.9665 0.0585 0.4271 0.8028 0.95 0.50

2.5.4.1 Controller Parameters

The parameters of the proposed PIλ -PDμ controller can be obtained from the designed
SWPIλ Dμ controller parameters as given in Sect. 2.5.3.1 using the conversion rela-
tions given in Fig. 2.22. Therefore, the parameters of PI-PD and PIλ -PDμ controllers
are given in Table 2.26.
80 2 Fractional-order Set-Point Weighted Controllers

3
Pressure (bar)

B Setpoint
2 PID
A
PI D
1 SWPI-D
SWPI -D
0
0 100 200 300 400 500 600 700 800 900 1000
(a) Time (s)

100 PID
PI D
Control Signal

SWPI-D
D SWPI -D
50

0
0 100 200 300 400 500 600 700 800 900 1000
(b) Time (s)

Fig. 2.66 Performance comparison of series SWPIλ Dμ controller for set-point tracking and dis-
turbance rejection

2.5.4.2 PI-PD Controllers

The comparison of the response of the system with various integer-order controllers
(PID, PI-PD1, and PI-PD2) for set-point tracking and disturbance rejection is given
in Fig. 2.70 while the zoomed-in view of regions of interest A, B, C and D of the figure
is given in Fig. 2.71. The numerical assessment of the figure is given in Table 2.27.
From the figure and the table, it can be observed that the set-point tracking ability
of PI-PD1 controller is outperformed those of PID and PI-PD2 in terms of all the
parameters. Comparing the control signals of the controllers, it can be seen that the
PI-PD1 produced smoother action as compared to the oscillatory actions of both PID
and PI-PD2. The smooth control action is due to the structure of the PI-PD.
Similarly, the performance comparison of the various controller for variable set-
point tracking is given in Fig. 2.72 while the regions of interest are further high-
lighted in Fig. 2.73. From the figure, it can be observed that, during step change,
PID controller experiences derivative kick effect while both the PI-PD1 and PI-PD2
controllers experienced none of this effect.
2.5 Simulation Study on Pressure Process Plant 81

(a) (b)
3.5
Pressure (bar)

3
3
2.8
2.5
2.6

2 2.4
0 50 100 150 200 500 550 600
(c) (d)
100
40
Control Signal

80
30

60 20

10
40
0
0 50 100 150 200 500 550 600
Time (s) Time (s)

Fig. 2.67 Regions of interest A, B, C and D of Fig. 2.66

Table 2.25 Performance analysis for series form of controllers

Controller tr ts1 ts2 %OS
PID 7.6653 95.1112 543.0661 12.5007
PIλ Dμ 6.7738 93.1063 841.0198 4.1832
SWPI-D 10.9104 100.4979 541.8869 9.1338
SWPIλ -Dμ 17.2004 33.4194 686.2195 2.9943

2.5.4.3 PIλ -PDμ Controllers

In a similar way to the integer-order controllers given in previous section, the com-
parison of the response of the system with various fractional-order controllers PIλ Dμ ,
PIλ -PDμ 1 and PIλ -PDμ 2 given in Eqs. (2.1), (2.58) and (2.62) for set-point tracking
and disturbance rejection is given in Fig. 2.74 while the zoomed-in view of regions of
interest A, B, C and D of the figure is given in Fig. 2.75. The numerical observations
are given in Table 2.28.
From the figure and the table, it can be seen that unlike in the case of integer-
order where PI-PD1 controller outperformed the other controllers, the performance
82 2 Fractional-order Set-Point Weighted Controllers

4
Pressure (bar)

3 Setpoint
B PID
2 PI D
A
SWPI-D
1
SWPI -D
0
0 100 200 300 400 500 600 700 800 900 1000
(a) Time (s)

PID
100
PI D
Control Signal

SWPI-D
SWPI -D
50

C D

0
0 100 200 300 400 500 600 700 800 900 1000
(b) Time (s)

Fig. 2.68 Performance comparison of series SWPIλ Dμ controller for variable set-point tracking

of the two fractional PI-PD controllers (PIλ -PDμ 1 and PIλ -PDμ 2) outperformed that
of PIλ Dμ .
The two PIλ -PDμ controllers have each produced an overshoot of around 3.5%
while that of the PIλ Dμ is 4.0737%. In terms of rise time, each of the PIλ -PDμ 1
and PIλ -PDμ 2 controllers produced around 6.6 s. This is faster compared to 7.17 s of
PIλ Dμ . Similar pattern is also observed for both settling times as given in the table.
Similarly, the performance comparison of the various controller for variable set-
point tracking is given in Fig. 2.76 while the zoomed-in view of regions of interest
A, B, C and D of the figure is given in Fig. 2.77. From the figure, it can be observed
that, during step change, the PIλ Dμ experienced derivative kick effect while both the
PIλ -PDμ 1 and PIλ -PDμ 2 controllers experienced none of this effect.

2.5.4.4 Comparison of PI-PD and PIλ -PDμ Controllers

The responses of the system with type 1 (PI-PD1 and PIλ -PDμ 1) and type 2 (PI-
PD2 and PIλ -PDμ 2) is given in Fig. 2.78 while the regions of interest of the figure is
highlighted in Fig. 2.79. For the numerical results of these figures see Table 2.29.
2.5 Simulation Study on Pressure Process Plant 83

(a) (b)
4
Pressure (bar)

3.5
2.5

3
2
0 50 100 150 200 500 550 600
(c) (d)
100 100
Control Signal

80 80

60 60

40 40

0 50 100 150 200 500 550 600

Time (s) Time (s)

Fig. 2.69 Regions of interest A, B, C and D of Fig. 2.68

Table 2.26 Controller parameters of PI-PD controllers

Controller P I D λ μ
PID 30.0 3.0 1.0 – –
PI-PD 29.8997 3.0000 0.0334 – –
PIλ -PDμ 29.8997 3.0000 0.0334 0.95 0.5

From the figures, considering the type 1 controllers, it can be seen that the PIλ -
PDμ 1 controller performed better compared to PI-PD1. The numerical comparison
shows that the PIλ -PDμ 1 controller produced the least overshoot of 3.5303% while
PI-PD1 has the overshoot of 5.4441%. Furthermore, the PIλ -PDμ 1 controller settled
faster at 86.1941 s while PI-PD1 at 91.7818 s. However, in terms of tr , both the
controller has the same rise time of around 6.6 s. The disturbance rejection of both
the controllers (PI-PD1 and PIλ -PDμ 1) are similar and satisfactory with PI-PD1 has
the fastest settling time of around 240 s.
On the other hand, observing the performance of PI-PD2 and PIλ -PDμ 2, it can
be seen that the PIλ -PDμ 2 performed much better when compared to PI-PD2. The
numerical comparison also shows that PIλ -PDμ 2 controller has the fastest rise and
settling times and very less overshoot of 6.6958 s, 78.1124 s and 3.5257% respectively
84 2 Fractional-order Set-Point Weighted Controllers

Setpoint
PID
3.5 PI-PD2
Pressure (bar)

PI-PD1
3

2.5
A B
2
0 50 100 150 200 250 300 350 400
(a) Time (s)

100
PID
Control Signal

PI-PD2
D PI-PD1
50

0
0 50 100 150 200 250 300 350 400
(b) Time (s)

Fig. 2.70 Performance comparison of PI-PD controller for set-point tracking and disturbance
rejection
(a) (b)
3.2
3.5
Pressure (bar)

3 2.8

2.6
2.5
2.4
20 40 60 80 100 200 220 240 260 280
(c) (d)
100 50

40
Control Signal

80
30

20
60
10

40 0
0 50 100 200 220 240 260 280
Time (s) Time (s)

Fig. 2.71 Regions of interest A, B, C and D of Fig. 2.70

2.5 Simulation Study on Pressure Process Plant 85

Table 2.27 Performance analysis of integer-order controllers

Controller tr ts1 ts2 %OS
PID 7.7953 103.9629 240.9393 15.7821
PI-PD2 8.0101 105.0095 239.8613 16.6024
PI-PD1 6.6580 91.7818 239.3725 5.4441

4.5
Pressure (bar)

4 A

3.5
Setpoint
3 PID
PI-PD2
2.5 B PI-PD1
2
0 50 100 150 200 250 300 350 400
(a) Time (s)

100 PID
PI-PD2
Control Signal

PI-PD1

50

C D

0
0 50 100 150 200 250 300 350 400
(b) Time (s)

Fig. 2.72 Performance comparison of PI-PD controller for variable set-point tracking

when compared to PI-PD2 controller with a rise time of 8.0101 s, settling time of
150.0095 s and overshoot of 16.6024%.
The performance of PI-PD2 controller has been significantly improved by frac-
tional ordering. From the control signals, it can be observed that the PIλ -PDμ 2 con-
troller produced smoother control action compared to the oscillatory signal of PI-
PD2. The undesired oscillations may increase the deterioration rate of the valve.
Similarly, the responses of the system with type 1 and type 2 for variation in
set-point is given in Fig. 2.80 and the regions of interest of the figure are highlighted
in Fig. 2.81. From the figures it can be seen the both PIλ -PDμ controllers produced
smoother control action while the PI-PD controllers produced undesired oscillations.
Furthermore, it can be seen that all the PI-PD controllers are free from derivative
kick effects with fractional-order controllers having smoother control action.
86 2 Fractional-order Set-Point Weighted Controllers

(a) (b)
3.5 4
Pressure (bar)

3 3.5

2.5 3

0 50 100 200 220 240 260 280

(c) (d)
100 100
Control Signal

80 80

60 60

40 40

0 50 100 200 220 240 260 280

Time (s) Time (s)

Fig. 2.73 Regions of interest A, B, C and D of Fig. 2.72

3
Pressure (bar)

Setpoint
2.5 PI D
A PI -PD 2
B
PI -PD 1
2
0 50 100 150 200 250 300 350 400
(a) Time (s)

100
PI D
Control Signal

PI -PD 2
PI -PD 1
D
50

0
0 50 100 150 200 250 300 350 400
(b) Time (s)

Fig. 2.74 Performance comparison of PIλ -PDμ controller for set-point tracking and disturbance
rejection
2.5 Simulation Study on Pressure Process Plant 87

(a) (b)
3.2
3
Pressure (bar)

3
2.8
2.8
2.6
2.6
2.4
50 100 150 200 220 240 260 280
(c) (d)
100 50

40
Control Signal

80
30
60 20

10
40
0
0 50 100 200 220 240 260 280
Time (s) Time (s)

Fig. 2.75 Regions of interest A, B, C and D of Fig. 2.74

Table 2.28 Performance analysis of fractional-order controllers

Controller tr ts1 ts2 %OS
PIλ Dμ 7.1764 86.5618 260.1200 4.0737
PIλ -PDμ 2 6.6958 78.1124 256.1254 3.5257
PIλ -PDμ 1 6.6820 86.1941 252.2575 3.5303

2.5.5 Comparison of SWPIλ Dμ and PIλ -PDμ Controllers

In this subsection, the comparison of both the proposed controllers on the pressure
process plant is presented. First, the performance of all compared controllers will
be evaluated for set-point tracking and disturbance rejection. Then, the ability of all
the controllers to track variation in set point will be analyzed. In both the cases, as
mentioned earlier, a disturbance (D(s)) of 30% is injected at 200 s for the purpose
of disturbance rejection evaluation. Furthermore, the performance of the controllers
to track variation in set point will also be analyzed.
88 2 Fractional-order Set-Point Weighted Controllers

Pressure (bar) 4

3.5 A
Setpoint
3
PI D
PI -PD 2
2.5 B
PI -PD 1
2
0 50 100 150 200 250 300 350 400
(a) Time (s)

100 PI D
PI -PD 2
Control Signal

PI -PD 1

50

C D

0
0 50 100 150 200 250 300 350 400
(b) Time (s)

Fig. 2.76 Performance comparison of PIλ -PDμ controller for variable set-point tracking

(a) (b)
3.2
4
Pressure (bar)

2.8
3.5
2.6

2.4 3

0 50 100 200 220 240 260 280

(c) (d)
100
80
Control Signal

80

60
60

40 40

0 50 100 200 220 240 260 280

Time (s) Time (s)

Fig. 2.77 Regions of interest A, B, C and D of Fig. 2.76

2.5 Simulation Study on Pressure Process Plant 89

3.5
Pressure (bar)
3
Setpoint
2.5 PI-PD2
A B PI -PD 2
2
PI-PD1
1.5 PI -PD 1
0 50 100 150 200 250 300 350 400
(a) Time (s)

100 PI-PD2
PI -PD 2
Control Signal

PI-PD1
D PI -PD 1
50

0
0 50 100 150 200 250 300 350 400
(b) Time (s)

Fig. 2.78 Performance comparison of PI-PD and PIλ -PDμ for set-point tracking and disturbance
rejection

(a) (b)
3.2
3.5
Pressure (bar)

3 2.8

2.6
2.5
2.4
20 40 60 80 100 200 220 240 260 280
(c) (d)
100 50

40
Control Signal

80
30

20
60
10

40 0
0 50 100 200 220 240 260 280
Time (s) Time (s)

Fig. 2.79 Regions of interest A, B, C and D of Fig. 2.78

90 2 Fractional-order Set-Point Weighted Controllers

Table 2.29 Performance analysis of PI-PD and PIλ -PDμ controllers

Category Controller tr ts1 ts2 %OS
Integer-order controllers PID 7.7953 103.9629 240.9393 15.7821
PI-PD2 8.0101 105.0095 239.8613 16.6024
PI-PD1 6.6580 91.7818 239.3725 5.4441
Fractional-order controllers PIλ Dμ 7.1764 86.5618 260.1200 4.0737
PIλ -PDμ 2 6.6958 78.1124 256.1254 3.5257
PIλ -PDμ 1 6.6820 86.1941 252.2575 3.5303

Table 2.30 Controller parameters of proposed controllers

Category Controller Kp Ki Kd b λ μ
Integer-order controllers PID 30.00 3.00 1.00 – – –
SWPI-D 30.00 3.00 1.00 0.80 – –
PI-PD2 29.8997 3.00 0.0334 – – –
PI-PD1 29.8997 3.00 0.0334 – – –
Fractional-order controllers PIλ Dμ 30.00 3.00 1.00 – 0.95 0.50
SWPIλ -Dμ 30.00 3.00 1.00 0.80 0.95 0.50
PIλ -PDμ 2 29.8997 3.00 0.0334 – 0.95 0.50
PIλ -PDμ 1 29.8997 3.00 0.0334 – 0.95 0.50

Table 2.31 Performance analysis of proposed controllers

Category Controller tr ts1 ts2 %OS
Integer-order controllers PID 6.7511 98.1610 244.3894 8.2935
SWPI-D 8.4787 94.5839 237.9967 5.1853
PI-PD2 8.0101 105.0095 239.8613 16.6024
PI-PD1 6.6580 91.7818 239.3725 5.4441
Fractional-order controllers PIλ Dμ 6.9613 90.5565 282.2311 3.9360
SWPIλ -Dμ 7.3698 83.2167 253.4759 3.1089
PIλ -PDμ 2 6.6958 78.1124 256.1254 3.5257
PIλ -PDμ 1 6.6820 86.1941 252.2575 3.5303

2.5.5.1 Set-Point Tracking and Disturbance Rejection

The parameters of various compared integer-order and fractional-order controllers

are given in Table 2.30. The comparison of the response of the system with all the
compared controllers for set-point tracking and disturbance rejection is given in
Fig. 2.82. The regions A, B, C and D of the figure are highlighted in Fig. 2.83.
Observing the figures, it can be seen that the proposed fractional-order controllers
are performed better compared to all other integer-order controllers. The numerical
assessment of the figure given in Table 2.31 shows that the proposed controllers have
2.5 Simulation Study on Pressure Process Plant 91

4.5
Pressure (bar)
4 A

3.5 Setpoint
PI-PD2
3 PI -PD 2
2.5 B PI-PD1
PI -PD 1
2
0 50 100 150 200 250 300 350 400
(a) Time (s)
100 PI-PD2
PI -PD 2
Control Signal

PI-PD1
PI -PD 1
50

C D

0
0 50 100 150 200 250 300 350 400
(b) Time (s)

Fig. 2.80 Performance comparison of PI-PD and PIλ -PDμ for variable set-point tracking

(a) (b)

3.5 4
Pressure (bar)

3 3.5

2.5 3

0 50 100 200 220 240 260 280

(c) (d)
100
70
Control Signal

80
60

60 50

40
40
30
0 50 100 200 220 240 260 280
Time (s) Time (s)

Fig. 2.81 Regions of interest A, B, C and D of Fig. 2.80

92 2 Fractional-order Set-Point Weighted Controllers

3.5
Pressure (bar)

2.5
A B
Setpoint
2 PID
0 50 100 150 200 250 300 350 400
PI D
(a) Time (s) SWPI-D
100 PI-PD2
PI-PD1
80 SWPI -D
Control Signal

PI -PD 2
60 PI -PD 1
D
40

20 C

0
0 50 100 150 200 250 300 350 400
(b) Time (s)

Fig. 2.82 Performance comparison of SWPIλ Dμ and PIλ -PDμ controllers for set-point tracking
and disturbance rejection

(a) (b)
3.5 3.2
Pressure (bar)

3 2.8

2.6
2.5
2.4
20 40 60 80 100 200 250 300
(c) (d)
100
40
Control Signal

80
30

60 20

10
40
0
0 50 100 200 250 300
Time (s) Time (s)

Fig. 2.83 Regions of interest A, B, C and D of Fig. 2.82

2.5 Simulation Study on Pressure Process Plant 93

4.5

Pressure (bar) 4 A
3.5

3
Setpoint
2.5 B PID
PI D
2 SWPI-D
PI-PD2
0 50 100 150 200 250 300 350 400
PI-PD1
(a) Time (s)
SWPI -D
100 PI -PD 2
C D PI -PD 1
Control Signal

80

60

40

20
0 50 100 150 200 250 300 350 400
(b) Time (s)

Fig. 2.84 Performance comparison of SWPIλ Dμ and PIλ -PDμ controllers for variable set-point
tracking

(a) (b)

3.5 4
Pressure (bar)

3
3.5

2.5
3
2
20 40 60 80 100 200 250 300

(c) (d)
100 100
Control Signal

80 80

60 60

40 40

0 50 100 200 250 300

Time (s) Time (s)

Fig. 2.85 Regions of interest A, B, C and D of Fig. 2.84

94 2 Fractional-order Set-Point Weighted Controllers

the least overshoot of around 3.5%. Furthermore, the controllers also performed
better with least tr and ts1 of around 6.7 s and 82.5 s respectively.
On the other hand, the disturbance rejection performance from the figure shows
that all the compared controllers are similar and satisfactory with the integer-order
controller have the fastest settling time of around 240 s. This is an indication of the
good load regulation ability of the fractional-order controllers. Furthermore, from the
control signals, it can be observed that the proposed fractional-order controllers (PIλ -
PDμ 1, PIλ -PDμ 2 and SWPIλ -Dμ ) produced smoother control signal compared to the
oscillatory signal of integer-order controllers (PID, PI-PD1, PI-PD2, and SWPI-D).
The undesired oscillations and rapid changes in the control signal may increase the
deterioration rate of the control valve.

2.5.5.2 Variable Set-Point Tracking

The comparison of the response of the system with all the compared controllers for
variable set-point tracking is shown in Fig. 2.84. The regions A, B, C and D of the
figure are highlighted in Fig. 2.85.
From the figure, it can be seen that during set-point change, the proposed
fractional-order controllers (PIλ -PDμ 1, PIλ -PDμ 2 and SWPIλ -Dμ ) performed bet-
ter when compared to the other controllers. This is clearly seen in the zoomed plot
of regions of interest B of Fig. 2.85.
Additionally, observing the control signals, it can be seen that during the set-point
change the control signals of PID and PIλ Dμ are very narrow and sharp and both
reached a peak of 100%. On the other hand, the control actions of both integer-order
and fractional-order SWPI-D and PI-PD controllers are peaked at 75%. Furthermore,
smoother control actions of proposed fractional-order controllers are observed com-
pared to the oscillatory signal of integer-order controllers. Thus, the reduction in
derivative kick effect of the proposed controllers thereby reducing the deterioration
rate of the valve. This is due to the setting of derivative set-point weight c which is
zero.

2.6 Summary

This chapter has reviewed the PID control strategy and its modifications. The
fractional-order PID is an extension of PID which has been extensively used in many
control engineering applications to achieve more robust and stable performance. Fur-
thermore, the key challenge in achieving the objective of adequate set-point tracking
and disturbance rejection through smoother control action with the conventional PID
has been highlighted. Therefore, the design of two fractional-order control strategies
namely SWPIλ Dμ and PIλ -PDμ for the control of highly non-linear, sensitive and
dynamic processes is presented. The designed SWPIλ Dμ controller has been devel-
oped for standard, industrial, ideal and parallel configurations and also implemented
2.6 Summary 95

in single-input-single-output and dual-input-single-output configurations. Similarly,

the PIλ -PDμ controller has been developed in two single-loop feedback congratula-
tions. Furthermore, in both the control strategies, the parameters of various control
configurations, the effect of changing these parameters and conversion relations for
controller parameters have been developed.
On the other hand, the results on real-time pH neutralization and pressure process
plants show the proposed controllers have adequate set-point tracking and distur-
bance rejection through smoother control action compared to PID, PI-PD, set-point
weighted PID, and fractional-order PID controllers. Furthermore, during a set-point
change, the control action of all compared controllers is sharp and reached a peak of
100% while the proposed controllers around 40%. Thus, the reduction in derivative
kick effect of the proposed controllers thereby reducing the deterioration rate of the
valve. This is due to the setting of derivative set-point weight c which is zero. The
step-response characteristics also show that the proposed controllers have the least
overshoot and fastest setting time. However, the proposed controllers have the slow-
est rise-time. This is because of the proportional set-point weighting factor b which
is less than one. In terms of disturbance rejection, the proposed controller shows
quick recovery from the effect of disturbance. This is an indication of the good load
regulation ability of the fractional-order controllers.

References

1. Wang, Q.G., Ye, Z., Cai, W.J., Hang, C.C.: PID Control for Multivariable Processes. Springer
(2008)
2. Grimholt, C., Skogestad, S.: Optimal PI and PID control of first-order plus delay processes
and evaluation of the original and improved SIMC rules. J. Process Control 70, 36–46 (2018)
3. Haugen, F.: Comparing PI tuning methods in a real benchmark temperature control system.
Model. Identif. Control 31(3), 79–91 (2010)
4. Pachauri, N., Rani, A., Singh, V.: Bioreactor temperature control using modified fractional-
order IMC-PID for ethanol production. Chem. Eng. Res. Des. 122, 97–112 (2017)
5. Hermansson, A.W., Syafiie, S.: Model predictive control of pH neutralization processes: a
review. Control Eng. Pract. 45, 98–109 (2015)
6. Samad, T.: A survey on industry impact and challenges thereof [technical activities]. IEEE
Control Syst. Mag. 37(1), 17–18 (2017)
7. Shah, P., Agashe, S.: Review of fractional PID controller. Mechatronics 38, 29–41 (2016)
8. Alfaro, V.M., Vilanova, R.: Conversion formulae and performance capabilities of two-degree-
of-freedom PID control algorithms. In: Proceedings of 2012 IEEE 17th International Confer-
ence on Emerging Technologies & Factory Automation, Kraków, Poland, 17–21 Sept 2012
9. Ang, K.H., Chong, G., Li, Y.: PID control system analysis, design, and technology. IEEE
Trans. Control Syst. Technol. 13(4), 559–576 (2005)
10. Alfaro, V.M., Vilanova, R.: Model-Reference Robust Tuning of PID Controllers. Springer
(2016)
11. Mudi, R.K., Dey, C.: Performance improvement of PI controllers through dynamic set-point
weighting. ISA Trans. 50(2), 220–230 (2011)
12. Visioli, A.: Practical PID Control. Springer (2006)
13. Alfaro, V.M., Vilanova, R.: Model-reference robust tuning of 2DoF PI controllers for first-and
second-order plus dead-time controlled processes. J. Process Control 22(2), 359–374 (2012)
96 2 Fractional-order Set-Point Weighted Controllers

14. Åström, K.J., Hägglund, T.: Advanced PID Control. ISA-The Instrumentation, Systems, and
Automation Society (2006)
15. Azarmi, R., Tavakoli-Kakhki, M., Sedigh, A.K., Fatehi, A.: Analytical design of fractional
order PID controllers based on the fractional set-point weighted structure: case study in twin
rotor helicopter. Mechatronics 31, 222–233 (2015)
16. De Oliveira, E.C., Tenreiro Machado, J.A.: A review of definitions for fractional derivatives
and integral. Math. Probl. Eng. (2014)
17. Sahu, R.K., Panda, S., Rout, U.K.: DE optimized parallel 2-DOF PID controller for load
frequency control of power system with governor dead-band nonlinearity. Int. J. Electr. Power
Energy Syst. 49, 19–33 (2013)
18. Mantz, R.J.: A PI controller with dynamic set-point weighting for nonlinear processes. IFAC
Proc. Vol. 45(3), 512–517 (2012)
19. Ghosh, A., Krishnan, T.R., Tejaswy, P., Mandal, A., Pradhan, J.K., Ranasingh, S.: Design and
implementation of a 2-DOF PID compensation for magnetic levitation systems. ISA Trans.
53(4), 1216–1222 (2014)
20. Jin, Q.B., Liu, Q.: Analytical IMC-PID design in terms of performance/robustness tradeoff
for integrating processes: from 2-Dof to 1-Dof. J. Process Control 24(3), 22–32 (2014)
21. Rajinikanth, V., Latha, K.: Setpoint weighted PID controller tuning for unstable system using
heuristic algorithm. Arch. Control Sci. 22(4), 481–505 (2012)
22. Rodriguez-Martinez, A., Garduno-Ramirez, R.: 2 DOF fuzzy gain-scheduling PI for combus-
tion turbogenerator speed control. IFAC Proc. Vol. 45(3), 276–281 (2012)
23. Pachauri, N., Singh, V., Rani, A.: Two degree of freedom PID based inferential control of
continuous bioreactor for ethanol production. ISA Trans. 68, 235–250 (2017)
24. Bianchi, F.D., Mantz, R.J., Christiansen, C.F.: Multivariable PID control with set-point weight-
ing via BMI optimisation. Automatica 44(2), 472–478 (2008)
25. Sahu, R.K., Panda, S., Rout, U.K., Sahoo, D.K.: Teaching learning based optimization algo-
rithm for automatic generation control of power system using 2-DOF PID controller. Int. J.
Electr. Power Energy Syst. 77, 287–301 (2016)
26. Vilanova, R., Visioli, A.: PID Control in the Third Millennium. Springer (2012)
27. Papadopoulos, K.G.: PID Controller Tuning Using the Magnitude Optimum Criterion.
Springer (2015)
28. Ingimundarson, A., Hägglund, T.: Performance comparison between PID and dead-time com-
pensating controllers. J. Process Control 12(8), 887–895 (2002)
29. Tan, K.K., Tang, K.Z., Su, Y., Lee, T.H., Hang, C.C.: Deadtime compensation via setpoint
variation. J. Process Control 20(7), 848–859 (2010)
30. Larsson, P., Hägglund, T.: Comparison between robust PID and predictive PI controllers with
constrained control signal noise sensitivity. IFAC Proc. Vol. 45(3), 175–180 (2012)
31. Kaya, I.: Obtaining controller parameters for a new PI-PD Smith predictor using autotuning.
J. Process Control 13(5), 465–472 (2003)
32. Kaya, I.: A PI-PD controller design for control of unstable and integrating processes. ISA
Trans. 42(1), 111–121 (2003)
33. Kaya, I.: PI-PD controllers for controlling stable processes with inverse response and dead
time. Electr. Eng. 98(1), 55–65 (2016)
34. Zou, H., Li, H.: Improved PI-PD control design using predictive functional optimization for
temperature model of a fluidized catalytic cracking unit. ISA Trans. 67, 215–221 (2017)
35. Tsai, K.I., Tsai, C.C.: Design and experimental evaluation of robust PID and PI-PD tempera-
ture controllers for oil-cooling machines. In: 9th World Congress on Intelligent Control and
Automation, Taipei, Taiwan 21–25 June 2011
36. Padhy, P.K., Majhi, S.: Relay based PI-PD design for stable and unstable FOPDT processes.
Comput. Chem. Eng. 30(5), 790–796 (2006)
37. Padhy, S., Panda, S.: A hybrid stochastic fractal search and pattern search technique based
cascade PI-PD controller for automatic generation control of multi-source power systems in
presence of plug in electric vehicles. CAAI Trans. Intell. Technol. 2(1), 12–25 (2017)
References 97

38. Padhy, S., Panda, S., Mahapatra, S.: A modified GWO technique based cascade PI-PD con-
troller for AGC of power systems in presence of plug in electric vehicles. Int. J. Eng. Sci.
Technol. 20(2), 427–442 (2017)
39. Nema, S., Kumar Padhy, P.: Identification and cuckoo PI-PD controller design for stable and
unstable processes. Trans. Inst. Meas. Control 37(6), 708–720 (2015)
40. Majhi, M.P.V.S., Mahanta, C.: Fuzzy proportional integral-proportional derivative (PI-PD)
controller. In: Proceedings of the 2004 American Control Conference, vol. 5, pp. 4028–4033,
IEEE (2004)
41. Silva, G.J., Datta, A., Bhattacharyya, S.P.: PID Controllers for Time-Delay Systems. Springer
(2007)
42. Liu, G.P., Daley, S.: Optimal-tuning nonlinear PID control of hydraulic systems. Control Eng.
Pract. 8(9), 1045–1053 (2000)
43. Su, Y.X., Sun, D., Duan, B.Y.: Design of an enhanced nonlinear PID controller. Mechatronics
15(8), 1005–1024 (2005)
44. Ye, J.: Adaptive control of nonlinear PID-based analog neural networks for a nonholonomic
mobile robot. Neurocomputing 71(7–9), 1561–1565 (2008)
45. Prakash, J., Srinivasan, K.: Design of nonlinear PID controller and nonlinear model predictive
controller for a continuous stirred tank reactor. ISA Trans. 48(3), 273–282 (2009)
46. Xue, D., Chen, Y., Atherton, D.P.: Linear Feedback Control: Analysis and Design with MAT-
LAB. Siam (2007)
47. Adar, N.G., Kozan, R.: Comparison between real time PID and 2-DOF PID controller for
6-DOF robot arm. Acta Phys. Pol. A 130(1), 269–271 (2016)
48. Alfaro, V.M., Vilanova, R.: Robust tuning of 2DoF five-parameter PID controllers for inverse
response controlled processes. J. Process Control 23(4), 453–462 (2013)
49. Alfaro, V.M., Vilanova, R.: Robust tuning and performance analysis of 2DoF PI controllers
for integrating controlled processes. Ind. Eng. Chem. Res. 51(40), 13182–13194 (2012)
50. Alfaro, V.M., Vilanova, R.: Simple robust tuning of 2DoF PID controllers from a perfor-
mance/robustness trade-off analysis. Asian J. Control 15(6), 1700–1713 (2013)
51. Vilanova, R., Alfaro, V.M., Arrieta, O.: Simple robust autotuning rules for 2-DoF PI con-
trollers. ISA Trans. 51(1), 30–41 (2012)
52. Araki, M., Taguchi, H.: Two-degree-of-freedom PID controllers. Int. J. Control Autom. Syst.
1(4), 401–411 (2003)
53. Li, Z., Liu, L., Dehghan, S., Chen, Y., Xue, D.: A review and evaluation of numerical tools
for fractional calculus and fractional order controls. Int. J. Control 90(6), 1165–1181 (2017)
54. Machado, J.T., Kiryakova, V., Mainardi, F.: Recent history of fractional calculus. Commun.
Nonlinear Sci. Numer. Simul. 16(3), 1140–1153 (2011)
55. Margarita, R., Sergei, V.R., José, A.T.M., Juan, J.T.: Stability of fractional order systems.
Math. Probl. Eng. (2013)
56. Krishna, B.T.: Studies on fractional order differentiators and integrators: a survey. Signal
Process. 91(3), 386–426 (2011)
57. Valério, D., Trujillo, J.J., Rivero, M., Machado, J.T., Baleanu, D.: Fractional calculus: a survey
of useful formulas. Eur. Phys. J. Spec. Top. 222(8), 1827–1846 (2013)
58. Petráš, I.: Fractional-Order Nonlinear Systems: Modeling, Analysis and Simulation. Springer
Science & Business Media (2011)
59. Caponetto, R.: Fractional Order Systems: Modeling and Control Applications. World Scien-
tific (2010)
60. Vinagre, B.M., Podlubny, I., Hernandez, A., Feliu, V.: Some approximations of fractional
order operators used in control theory and applications. Fract. Calc. Appl. Anal. 3(3), 231–
248 (2000)
61. Petrás, I.: Fractional derivatives, fractional integrals, and fractional differential equations in
Matlab. Engineering education and research using MATLAB, IntechOpen (2011)
62. Monje, C.A., Vinagre, B.M., Feliu, V., Chen, Y.: Tuning and auto-tuning of fractional order
controllers for industry applications. Control Eng. Pract. 16(7), 798–812 (2008)
98 2 Fractional-order Set-Point Weighted Controllers

63. Chen, Y., Bhaskaran, T., Xue, D.: Practical tuning rule development for fractional order
proportional and integral controllers. J. Comput. Nonlinear Dyn. 3(2), 021403 (2008)
64. Zamani, M., Karimi-Ghartemani, M., Sadati, N., Parniani, M.: Design of a fractional order
PID controller for an AVR using particle swarm optimization. Control Eng. Pract. 17(12),
1380–1387 (2009)
65. Li, H., Luo, Y., Chen, Y.: A fractional order proportional and derivative (FOPD) motion
controller: tuning rule and experiments. IEEE Trans. Control Syst. Technol. 18(2), 516–520
(2009)
66. Padula, F., Visioli, A.: Tuning rules for optimal PID and fractional-order PID controllers. J.
Process Control 21(1), 69–81 (2011)
67. Sharma, R., Rana, K.P.S., Kumar, V.: Performance analysis of fractional order fuzzy PID
controllers applied to a robotic manipulator. Expert Syst. Appl. 41(9), 4274–4289 (2014)
68. Monje, C.A., Chen, Y., Vinagre, B.M., Xue, D., Feliu-Batlle, V.: Fractional-Order Systems
and Controls: Fundamentals and Applications. Springer Science & Business Media (2010)
69. Luo, Y., Chen, Y.: Fractional order [proportional derivative] controller for a class of fractional
order systems. Automatica 45(10), 2446–2450 (2009)
70. Luo, Y., Chao, H., Di, L., Chen, Y.: Lateral directional fractional order (PI) α control of a
small fixed-wing unmanned aerial vehicles: controller designs and flight tests. IET Control
Theory Appl. 5(18), 2156–2167 (2011)
71. Luo, Y., Chen, Y.: Fractional-order [proportional derivative] controller for robust motion con-
trol: tuning procedure and validation. In: American Control Conference, St. Louis, Missouri,
USA, 10–12 June 2009
72. Luo, Y., Chen, Y.Q., Wang, C.Y., Pi, Y.G.: Tuning fractional order proportional integral con-
trollers for fractional order systems. J. Process Control 20(7), 823–831 (2010)
73. Tenoutit, M., Maamri, N., Trigeassou, J.C.: An output feedback approach to the design of
robust fractional PI and PID controllers. IFAC Proc. Vol. 44(1), 12568–12574 (2011)
74. Lachhab, N., Svaricek, F., Wobbe, F., Rabba, H.: Fractional order PID controller (FOPID)-
Toolbox. In: European Control Conference. Zurich, Switzerland, 17–19 July 2013
75. Tenoutit, M., Maamri, N., Trigeassou, J.C.: A time moments approach to the design of robust
fractional PID controllers. In: 8th International Multi-conference on Systems, Signals &
Devices, Sousse-Tunisia, 22–25 March 2011
76. Merrikh-Bayat, F., Mirebrahimi, N.: Introduction to the nonlinear PIλ Dμ control. In: 2011
IEEE International Conference on Control System, Computing and Engineering, Penang,
Malaysia, 25–27 Nov 2011
77. Tavakoli-Kakhki, M., Haeri, M.: Fractional order model reduction approach based on retention
of the dominant dynamics: application in IMC based tuning of FOPI and FOPID controllers.
ISA Trans. 50(3), 432–442 (2011)
78. Bettayeb, M., Mansouri, R.: Fractional IMC-PID-filter controllers design for non integer order
systems. J. Process Control 24(4), 261–271 (2014)
79. Feliu-Batlle, V., Perez, R.R., Rodriguez, L.S.: Fractional robust control of main irrigation
canals with variable dynamic parameters. Control Eng. Pract. 15(6), 673–686 (2007)
80. Feliu-Batlle, V., Rivas-Perez, R., Castillo-Garcia, F.J.: Fractional order controller robust to
time delay variations for water distribution in an irrigation main canal pool. Comput. Electron.
Agric. 69(2), 185–197 (2009)
81. El-Khazali, R.: Fractional-order PIλ Dμ controller design. Comput. Math. Appl. 66(5), 639–
646 (2013)
82. Freeborn, T.J.: A survey of fractional-order circuit models for biology and biomedicine. IEEE
J. Emerg. Sel. Top. Circuits 3(3), 416–424 (2013)
83. Freeborn, T.J., Maundy, B., Elwakil, A.S.: Fractional-order models of supercapacitors, bat-
teries and fuel cells: a survey. Mater. Renew. Sustain. Energy 4(3), 1–9 (2015)
84. Jain, M., Rani, A., Pachauri, N., Singh, V., Mittal, A.P.: Design of fractional order 2-DOF
PI controller for real-time control of heat flow experiment. Eng. Sci. Technol. Int. J. 22(1),
215–228 (2019)
References 99

85. Pachauri, N., Singh, V., Rani, A.: Two degrees-of-freedom fractional-order proportional-
integral-derivative-based temperature control of fermentation process. J. Dyn. Syst. Meas.
Control-Trans. ASME 140(7), 071006 (2018)
86. Angel, L., Viola, J.: Fractional order PID for tracking control of a parallel robotic manipulator
type delta. ISA Trans. 79, 172–188 (2018)
87. Sharma, R., Gaur, P., Mittal, A.P.: Performance analysis of two-degree of freedom fractional
order PID controllers for robotic manipulator with payload. ISA Trans. 58, 279–291 (2015)
88. Sondhi, S., Hote, Y.V.: Fractional order PID controller for load frequency control. Energy
Conv. Manag. 85, 343–353 (2014)
89. Zamani, A., Barakati, S.M., Yousofi-Darmian, S.: Design of a fractional order PID controller
using GBMO algorithm for load-frequency control with governor saturation consideration.
ISA Trans. 64, 56–66 (2016)
90. Lamba, R., Singla, S.K., Sondhi, S.: Fractional order PID controller for power control in
perturbed pressurized heavy water reactor. Nucl. Eng. Des. 323, 84–94 (2017)
91. Kumar, N., Tyagi, B., Kumar, V.: Deregulated multiarea AGC scheme using BBBC-FOPID
controller. Arab. J. Sci. Eng. 42(7), 2641–2649 (2017)
92. Debbarma, S., Saikia, L.C., Sinha, N.: Automatic generation control using two degree of
freedom fractional order PID controller. Int. J. Electr. Power Energy Syst. 58, 120–129 (2014)
93. Tepljakov, A., Gonzalez, E.A., Petlenkov, E., Belikov, J., Monje, C.A., Petráš, I.: Incorporation
of fractional-order dynamics into an existing PI/PID DC motor control loop. ISA Trans. 60,
262–273 (2016)
94. Pan, I., Das, S.: Frequency domain design of fractional order PID controller for AVR system
using chaotic multi-objective optimization. Int. J. Electr. Power Energy Syst. 51, 106–118
(2013)
95. Kumar, G., Arunshankar, J.: Control of nonlinear two-tank hybrid system using sliding mode
controller with fractional-order PI-D sliding surface. Comput. Electr. Eng. 71, 953–965 (2018)
96. Tepljakov, A., Alagoz, B.B., Yeroglu, C., Gonzalez, E., HosseinNia, S.H., Petlenkov, E.:
FOPID controllers and their industrial applications: a survey of recent results. IFAC-
PapersOnLine 51(4), 25–30 (2018)
97. Lurie, B.J.: Three-parameter tunable tilt-integral-derivative (TID) controller. US Patent
5,371,670, 6 Dec 1994
98. Abbisso, S., Caponetto, R., Diamante, O., Porto, D., Di Cola, E., Fortuna, L.: Non-integer
order dynamic systems. US Patent 6,678,670, 13 Jan 2004
99. Bohannan, G., Hurst, S., Spangler, L.: Electrical component with fractional order impedance.
US Patent 11/372,232, 30 Nov 2006
100. Chen, Y.: Tuning methods for fractional-order controllers. US Patent 7,599,752, 6 Oct 2009
101. Almadhoun, M.N., Elshurafa, A., Salama, K., Alshareef, H.: Fractional order capacitor. US
Patent 9,305,706, 6 Apr 2016
102. Rana, K.P.S., Kumar, V., Mittra, N., Pramanik, N.: Implementation of fractional order inte-
grator/differentiator on field programmable gate array. Alex. Eng. J. 55(2), 1765–1773 (2016)
103. Muñiz-Montero, C., García-Jiménez, L.V., Sánchez-Gaspariano, L.A., Sánchez-López, C.,
González-Díaz, V.R., Tlelo-Cuautle, E.: New alternatives for analog implementation of
fractional-order integrators, differentiators and PID controllers based on integer-order inte-
grators. Nonlinear Dyn. 90(1), 241–256 (2017)
104. Tolba, M.F., AboAlNaga, B.M., Said, L.A., Madian, A.H., Radwan, A.G.: Fractional order
integrator/differentiator: FPGA implementation and FOPID controller application. AEU-Int.
J. Electron. Commun. 98, 220–229 (2019)
105. Tepljakov, A.: Fractional-Order Modeling and Control of Dynamic Systems. Springer (2017)
106. Bingi, K., Ibrahim, R., Karsiti, M.N., Hassan, S.M., Harindran, V.R.: A comparative study of
2DOF PID and2DOF fractional order PID controllerson a class of unstable systems. Arch.
Control Sci. 28(4), 635–682 (2018)
107. Bingi, K., Ibrahim, R., Karsiti, M.N., Hassan, S.M.: Fractional order set-point weighted PID
controller for pH neutralization process using accelerated PSO algorithm. Arab. J. Sci. Eng.
43(6), 2687–2701 (2018)
100 2 Fractional-order Set-Point Weighted Controllers

108. Bingi, K., Ibrahim, R., Karsiti, M.N., Chung, T.D., Hassan, S.M.: Optimal PID control of pH
neutralization plant. In: 2nd IEEE International Symposium on Robotics and Manufacturing
Automation. Universiti Teknologi PETRONAS, Malaysia 25–27 Sept 2016
109. Bingi, K., Ibrahim, R., Karsiti, M.N., Hassan, S.M., Harindran, V.R.: Real-time control of
pressure plant using 2DOF fractional-order PID controller. Arab. J. Sci. Eng. 44(3), 2091–
2102 (2019)
Chapter 3
Approximation Techniques

3.1 Introduction

A key issue with the practical realization or equivalent circuit implementation of

fractional-order PID controllers in a finite-dimensional integer-order system is the
approximation of the fractional-order parameters for the desired frequency range
and the order of approximation. Thus, for the effective approximation of fractional-
order integrator and differentiator in the PIλ Dμ controller, researchers have pro-
posed several frequency-domain approximation techniques [1–3]. The summary of
these approximation techniques including the MATLAB commands for implement-
ing these algorithms is given in Table 3.1. However, in certain situations, these meth-
ods cannot fit around the desired frequency range and also have limitations on order
of approximation, the accuracy of frequency and time responses [4, 5].
Among the frequency domain approximation techniques available in the literature,
Oustaloup’s approximation is the most popular and widely used. This approximation
is based on a recursive distribution of poles and zeros in the desired frequency range
using frequency response fitting [6]. In certain situations, the Oustaloup’s approxi-
mation cannot fit around the frequency range of interest [16]. Thus, a modified or
refined version has been proposed [6, 13, 17–20]. However, the modified approxi-
mation technique produces a higher-order approximated model. Thus, to reduce the
order of approximation, [21] proposed a fixed-pole approximation technique.
Similarly, Carlson [8] proposed an approximation technique using Newton itera-
tive method for continued fractional expansion (CFE) of the fractional-order differen-
tiator. However, the technique is limited to fewer values of fractional-order parameter
[8, 22]. Furthermore, Matsuda proposed an approximation technique using the gain
of the fractional-order transfer function [9]. However, in this method, if the order of
the approximation is chosen as an odd number, the approximated transfer function
will be improper i.e., there will be one more zero than poles [2, 22, 23]. Researchers
also proposed various power series expansion (PSE) techniques based on Taylor
series, Maclaurin Series, finite impulse response (FIR) and infinite impulse response

© Springer Nature Switzerland AG 2020 101

K. Bingi et al., Fractional-order Systems and PID Controllers,
Studies in Systems, Decision and Control 264,
https://doi.org/10.1007/978-3-030-33934-0_3
102 3 Approximation Techniques

Table 3.1 Summary of approximation techniques for fractional-order parameters

Category Technique Syntaxa Reference
Continued fractional CFE Low, CFE High nid() [7]
expansion
Carlson [8]
Matsuda [9]
Power series FIR Filter dfod1(), [10]
expansion dfod2()
IIR Filter dfod3() [11]
Curve Fitting Oustaloup ousta_fod() [12]
Refined Oustaloup new_fod() [13]
Charef – [14]
Vector Fitting – [15]
Others Impulse Response irid_fod(), [12]
irid_folpf()
Step Response srid_fod()
a The ‘Syntax’ column denotes the command for the approximation technique available from MAT-

LAB file exchange

(IIR) [9–11]. However, it is proven that the CFE methods are frequently converging
more rapidly than PSE methods [2].
In the related development, Charef proposed an approximation technique [14]
where accuracy is determined by properly selecting the maximal permissible error.
However, the order of approximation involves a significant amount of trials and errors
[24, 25]. Thus, an extension of this method was proposed by [17]. This extended
method is focused on improving the accuracy of the original proposal. Other approx-
imation algorithms based on stability boundary locus [5], vector fitting method [15],
time moments approach [26], state-space approach [27, 28] and frequency distribu-
tion mode [18] have been proposed too. A key issue with these methods is that they
are quite complex hence difficult to implement.
Therefore, here a simple curve fitting based approximation techniques using exact
frequency response data of fractional-order operators (differentiator/integrator) has
been proposed. The approach is expected to achieve better approximation compared
with the commonly used Oustaloup, refined Oustaloup and Matsuda [12]. To demon-
strate the performance of the proposed approach, a simulation study will be conducted
on a class of fractional-order systems and controllers.

3.2 Curve Fitting Based Approximation Algorithm

The proposed curve fitting based approximation algorithm for fractional-order dif-
ferentiator s α will be done in three stages. The first stage of the approximation is
obtaining the frequency response data (frd) of s γ . This is computed by substitut-
3.2 Curve Fitting Based Approximation Algorithm 103

ing s = jω and then evaluating the function for different value of ω ∈ (ωl , ωh ) as
follows:
s γ = ( jω)γ |ω=(ωl ,...,ωh ) (3.1)

The next stage is to derive the integer-order transfer function of the fractional-order
differentiator G d (s) = s α from the data obtained in Eq. (3.1) using Sanathanan-
Koerner (SK) least square iterative method [29]. Thus, the integer-order transfer
function of fractional-order differentiator (G d (s)) is defined as follows:


N
pn s n
P(s) n=0 Pψ(s)
G d (s) = ≈ ≈ (3.2)
Q(s) 
N 1 + Qφ(s)
1+ qn s n

n=1

where
• P and Q are the coefficients and
• ψ(s) and φ(s) are the monomial functions.
Thus, the coefficients and monomial functions are defined as follows:
⎡ ⎤ ⎡ ⎤
p0 q1
⎢ p1 ⎥ ⎢ q2 ⎥
⎢ ⎥ ⎢ ⎥
P = ⎢ . ⎥; Q = ⎢ . ⎥ (3.3)
⎣ .. ⎦ ⎣ .. ⎦
pN qN

ψ(s) = [1, s, . . . , s N ]; φ(s) = [s, s 2 , . . . , s N ] (3.4)

The objective is to identify P and Q of the integer-order transfer function G d (s)

in the desired frequency range ω ∈ (ωl , ωh ). This is achieved by minimizing the
difference between data samples (H (s)) obtained from Eq. (3.1) and G(s) using the
following Levy’s linearized cost function with the SK’s least square iteration method.

h 2
P( jωk ) Q τ ( jωk )
arg min − τ −1 H ( jωk ) (3.5)
P,Q τ −1
Q ( jωk ) Q ( jωk )
k=l

where τ = 1, . . . , T is the iteration step.

From the above equation, it should be noted that the unbiased curve fitting is
achieved, as Q τ −1 ( jωk ) approaches to Q τ ( jωk ). Furthermore, to obtain the minimal
realization of G d (s), the matching poles and zeros will be canceled. The final stage
is to convert the obtained state-space model as given in Eq. (3.6) after pole zero
cancellation into a transfer function model in the form of:
104 3 Approximation Techniques

s X (s) = AX (s) + BU (s)

(3.6)
Y (s) = C X (s) + DU (s)

Y (s)
−1
G d (s) ≈ = C s IN − A B + D (3.7)
U (s)

where
• X (s), Y (s) and U (s) are the state, output and control vectors respectively,
• A, B, C and D are the state, input, output, feedforward matrices respectively.
Therefore, the above procedure for the proposed integer-order approximation of
fractional-order differentiator s α based on the curve fitting of frequency response
data with the use of MATLAB inbuilt commands will be implemented as follows:
1. Obtain the frequency response data for integer-order part of s α within the desired
frequency range ω ∈ (ωl , ωh ) using MATLAB inbuilt function frd().
2. Obtain the exact frd of s α by powering the data obtained in previous step with α.
3. Choose the order of the approximate N for the integer-order model.
4. Obtain the state space model of exact function response data based on SK’s least
square iteration method by using MATLAB inbuilt function fitfrd().
5. Convert the state space model to transfer function using MATLAB inbuilt com-
mand ss2tf().
The MATLAB commands for implementing the proposed algorithm is given as
follows:

function Gp = curveFitting(alpha,N,wl,wh)
w = logspace(log10(wl),log10(wh));
s = tf(’s’); FRD = frd(s,w);
FRD.ResponseData = FRD.ResponseData.ˆalpha;
Gp = fitfrd(FRD,N);
[num,den] = ss2tf(Gp.A,Gp.B,Gp.C,Gp.D);
Gp = tf(num,den);
end

This proposed curve fitting approximation is simple and expected to fit the entire
frequency range of interest. Furthermore, for effective approximation, the order of
fractional-order differentiator s α is limited to [−1, 1] range [9]. Thus, s α is divided
as follows:
s α = s α s α−α (3.8)

For example, the 3.2th order of fractional-order differentiator can be written using
the above equation as s 3.2 = s 3 × s 0.2 . Therefore, the integer-order approximation
of fractional-order differentiator s 0.2 will be obtained using the proposed curve fit-
ting based approximation technique while the integer-order differentiator s 3 will be
implemented directly.
3.3 Transfer Function Estimation Algorithm 105

3.3 Transfer Function Estimation Algorithm

Similarly, the proposed transfer function estimation algorithm using the Sanathanan-
Koerner (SK) least square iterative method presented in the previous section can
also be implemented in MATLAB using inbuilt function tfest() [30]. Thus, for
obtaining the integer-order transfer of fractional-order differentiator s α using the
exact frequency response data is as follows:
1. Obtain the frequency response data for integer-order part of s α within the desired
frequency range ω ∈ (ωl , ωh ) using the function frd() or idfrd().
2. Obtain the exact frequency response data of s α by powering the data obtained in
the previous step with α.
3. Choose the order of the approximate N for the integer-order model.
4. Obtain the transfer function model of frequency response data using the tfest()
function.
Thus, the MATLAB commands for estimating the integer-order transfer function
using the above-proposed algorithm is given as follows:

function Gp = functionEstimation(gam,N,wb,wh)
w = logspace(log10(wb),log10(wh),10000);
s=tf(’s’); FRD = idfrd(s,w);
FRD.ResponseData = FRD.ResponseData.ˆgam;
Gp = tfest(FRD,N,N);
end

3.4 Numerical and Stability Analysis

The performance evaluation for the accuracy of approximation techniques will be

done in both frequency and time domains. The frequency domain analysis include
Bode plot response while for the numerical assessment of the response, the mean
absolute error (MAE) defined in Eq. (3.9) is chosen. This type of error function is
the most suitable for non-time based problems.

1
n
MAE = |yi − xi | (3.9)
n i=1

where yi is the approximated frequency response data, xi is the exact frequency

response data and n is the total size of the data.
Furthermore, to evaluate the effectiveness of the proposed approaches, a time
domain comparison will also be done. Therefore, the exact step response of the
106 3 Approximation Techniques

fractional-order integrator 1/s α is obtained from the inverse Laplace transform of

integer-order integrator 1/s n as:

−1 1 t n−1
L = , n∈N (3.10)
sn (n − 1)!

As in the case of Eq. (3.10), the inverse Laplace transform of fractional integrator
1/s α is defined as: 
−1 1 t α−1
L = , 0<α<1 (3.11)
sα (α)

where (α) = (α − 1)!. From Eq. (3.11), the step response is computed as

1 tα tα
L −1 = = (3.12)
s α+1 (α + 1) α (α)

Furthermore, the stability analysis of the approximated transfer function (G(s))

will be done using H2 and H∞ -norms as defined in Eqs. (3.13) and (3.14) respectively.
  ∞
1
||G(s)||2 = trace[G( jω)∗ G( jω)]dω (3.13)
2π −∞

||G(s)||∞ = sup |G( jω)| (3.14)

ω∈R

From the equations, it can be noted that for a stable system the H2 -norm is the
average system gain over all frequencies while H∞ -norm is the peak gain of the
frequency response. However, for an unstable system, these norms will be infinite.

3.5 Performance Analysis of Approximation Algorithms

This section evaluates the performance of both the developed curve fitting based
approximation techniques developed. In the first part of this section, the performance
of curve fitting approximation has been evaluated on a fractional-order differentia-
tor of order α = 0.1. Then, for the same system, the performance of the transfer
function estimation algorithm has been evaluated. Next, an approximated transfer
functions for orders (α = 0.1, 0.2, . . . , 0.9) has been developed using both the pro-
posed approaches. Lastly, a simulation study will be conducted on fractional-order
based controllers and systems. The selected systems and controller are differentia-
tor, integrator, PID controller and higher-order transfer function. In all the cases, the
obtained results from the proposed approaches will be compared with Oustaloup,
refined Oustaloup and Matsuda approximations for MAE’s of frequency and step
responses. Furthermore, the stability analysis of the approximated transfer functions
3.5 Performance Analysis of Approximation Algorithms 107

will also be done using H2 and H∞ -norms as defined in Eqs. (3.13) and (3.14)
respectively.

3.5.1 Curve Fitting Based Approximation

To study the proposed curve fitting approach given in Sect. 3.2, the fractional-order
differentiator, s 0.1 is used. For this term, the desired frequency range ω is chosen as
(10−2 , 102 ). However, to study the effect of variation of N , the order of approximation
N is chosen as 4, 5 and 6. Thus, the approximated transfer functions using the
proposed approach for orders 4, 5 and 6 are given as follows:

1.714s 4 + 75.22s 3 + 248.1s 2 + 83.38s + 1.935

s 0.1 ≈ (3.15)
s 4 + 58.57s 3 + 244.1s 2 + 103.5s + 3.237

1.777s 5 + 123.9s 4 + 873.4s 3 + 909.9s 2 + 137.7s + 1.914

s 0.1 ≈ (3.16)
s 5 + 90.81s 4 + 785.4s 3 + 985s 2 + 182.9s + 3.335

1.831s 6 + 183.7s 5 + 2279s 4 + 5199s 3 + 2427s 2 + 203s + 1.9

s 0.1 ≈ (3.17)
s 6 + 128.7s 5 + 1919s 4 + 5142s 3 + 2818s 2 + 283.7s + 3.424

The Bode plots of the proposed approach in comparison to Oustaloup, refined

Oustaloup and Matsuda for orders 4, 5 and 6 are presented in Figs. 3.1, 3.2 and 3.3
respectively. From the figures, it can be observed that the proposed approach is more
accurate within the desired frequency range than the other methods for the order of
approximations 5 and 6. The MAE of various techniques is compared for magnitude
and phase responses of s 0.1 in Table 3.2.
From the table and for the order of approximations 5 and 6, the proposed technique
has the least MAE. However, for the 4th order approximation, Oustaloup, and refined
Oustaloup approximation techniques are better with the least errors. Thus, for the
higher order of approximations, the proposed technique gives a better approximation
than the other approaches.
The step responses of the proposed approach in comparison to Oustaloup, refined
Oustaloup and Matsuda for the orders 4, 5 and 6 are presented in Figs. 3.4, 3.5 and 3.6
respectively. From the figure, similar observation of better accuracy of the proposed
approach as in the case of frequency plot is made. Furthermore, the numerical assess-
ment of the step response in Table 3.3 also confirmed that the proposed approach has
produced less error when compared to other approaches. On the other hand, the sta-
bility analysis from the table also shows that all the approximated transfer functions
are stable with a finite value of H∞ -norm.
108 3 Approximation Techniques

10 Exact
Oustaloup
Magnitude (dB)

5 Refined Oustaloup
Matsuda
0 Proposed Method

-5

-10
10-3 10-2 10-1 100 101 102 103

15
Phase (deg)

10

0
10-3 10-2 10-1 100 101 102 103
Frequency (rad/sec)

Fig. 3.1 Bode response of s 0.1 using curve fitting for N = 4

Table 3.2 Comparison of MAE’s for Bode response of s 0.1

Technique Order ω1 ∈ (10−1 , 10) ω2 ∈ (10−2 , 102 )
Magnitude Phase Magnitude Phase
Oustaloup 4 0.0053 0.4936 0.0091 2.3204
Refined 0.0084 0.2879 0.0154 0.6566
Oustaloup
Matsuda 0.0107 0.4887 0.0290 1.0742
Proposed 0.0068 0.3982 0.0227 0.6822
Oustaloup 5 0.0027 0.2530 0.0131 2.3867
Refined 0.0116 0.1664 0.0096 0.5042
Oustaloup
Matsuda 0.0035 0.2407 0.0112 0.6956
Proposed 0.0020 0.1433 0.0124 0.2614
Oustaloup 6 0.0007 0.3016 0.0154 2.4373
Refined 0.0119 0.0887 0.0075 0.5241
Oustaloup
Matsuda 0.0018 0.0795 0.0067 0.3352
Proposed 0.0011 0.0541 0.0058 0.1415
3.5 Performance Analysis of Approximation Algorithms 109

10 Exact
Oustaloup
Magnitude (dB)

5 Refined Oustaloup
Matsuda
0 Proposed Method

-5

-10
10-3 10-2 10-1 100 101 102 103

15
Phase (deg)

10

0
10-3 10-2 10-1 100 101 102 103
Frequency (rad/sec)

Fig. 3.2 Bode response of s 0.1 using curve fitting for N = 5

Table 3.3 Comparison of MAE’s for step response of s 0.1

Technique Order t1 ∈ (0, 125) t2 ∈ (0, 250) H∞ -norm Stability
Oustaloup 4 0.0215 0.0441 1.5849 Stable
Refined 0.0547 0.0809 1.7425 Stable
Oustaloup
Matsuda 0.0144 0.0122 1.8280 Stable
Proposed 0.0061 0.0199 1.7145 Stable
Oustaloup 5 0.0237 0.0455 1.5849 Stable
Refined 0.0540 0.0813 1.7425 Stable
Oustaloup
Matsuda 0.0110 0.0120 1.8905 Stable
Proposed 0.0053 0.0097 1.7773 Stable
Oustaloup 6 0.0247 0.0463 1.5849 Stable
Refined 0.0538 0.0814 1.7425 Stable
Oustaloup
Matsuda 0.0103 0.0110 1.9462 Stable
Proposed 0.0071 0.0059 1.8315 Stable
110 3 Approximation Techniques

10 Exact
Oustaloup
Magnitude (dB)

5 Refined Oustaloup
Matsuda
0 Proposed Method

-5

-10
10-3 10-2 10-1 100 101 102 103

15
Phase (deg)

10

0
10-3 10-2 10-1 100 101 102 103
Frequency (rad/sec)

Fig. 3.3 Bode response of s 0.1 using curve fitting for N = 6

3.5.2 Transfer Function Estimation Algorithm

In a similar way, to study the transfer function estimation algorithm given in Sect. 3.3,
the fractional-order differentiator of order α = 0.1 is used. For the integer-order
transfer function estimation of this term, the desired frequency range ω is chosen
as (10−2 , 102 ) and order of approximation N is chosen as 4, 5 and 6. Thus, the
approximated transfer function using the proposed approach for orders 4, 5 and 6
are given as follows:

1.758s 4 + 109.7s 3 + 527.2s 2 + 194.7s + 4.272

s 0.1 ≈ (3.18)
s 4 + 81.96s 3 + 504.9s 2 + 242.2s + 7.147

1.824s 5 + 177s 4 + 1839s 3 + 2326s 2 + 330.3s + 4.043

s 0.1 ≈ (3.19)
s 5 + 124.8s 4 + 1600s 3 + 2501s 2 + 444.4s + 7.103

1.881s 6 + 259.7s 5 + 4740s 4 + 14170s 3 + 6852s 2 + 492.8s + 3.919

s 0.1 ≈
s 6 + 175.3s 5 + 3850s 4 + 13770s 3 + 8010s 2 + 701.2s + 7.152
(3.20)
3.5 Performance Analysis of Approximation Algorithms 111

1
Exact
Oustaloup
0.9 Refined Oustaloup
Matsuda
Proposed Method
0.8
Amplitude

0.7

0.6

0.5

0.4
0 50 100 150 200 250
Time (s)

Fig. 3.4 Step response of s 0.1 using curve fitting for N = 4

The Bode plots of the proposed approach in comparison to Oustaloup, refined

Oustaloup and Matsuda for orders 4, 5 and 6 are presented in Figs. 3.7, 3.8 and 3.9
respectively. From the figures, a similar observation of the better performance of the
proposed approach as in the case of curve fitting algorithm is made.
The MAE of various techniques is compared for magnitude and phase responses
of s 0.1 in Table 3.4. From the table and for the desired frequency of ω2 , the proposed
approach has the least MAE. However, for the desired frequency ω1 , Oustaloup and
refined Oustaloup approximation techniques are better with the least error. Thus, for
the high-frequency ranges, the proposed technique gives a better approximation than
the other approaches.
Furthermore, the step responses of the proposed approach in comparison to
Oustaloup, refined Oustaloup and Matsuda for the orders 4, 5 and 6 are presented in
Figs. 3.10, 3.11 and 3.12 respectively. From the figure, similar observation of better
accuracy of the proposed approach as in the case of frequency plot is made. The
numerical assessment of the step response in Table 3.5 also confirmed that for longer
time periods, the proposed approach gives a better approximation than the other
approaches. On the other hand, the stability analysis from the table also shows that
all the approximated transfer functions are stable with a finite value of H∞ -norm.
112 3 Approximation Techniques

1
Exact
Oustaloup
0.9 Refined Oustaloup
Matsuda
Proposed Method
0.8
Amplitude

0.7

0.6

0.5

0.4
0 50 100 150 200 250
Time (s)

Fig. 3.5 Step response of s 0.1 using curve fitting for N = 5

Table 3.4 Comparison of MAE’s for Bode response of s 0.1

Technique Order ω1 ∈ (10−1 , 10) ω2 ∈ (10−2 , 102 )
Magnitude Phase Magnitude Phase
Oustaloup 4 0.0053 0.4936 0.0091 2.3204
Refined 0.0084 0.2879 0.0154 0.6566
Oustaloup
Matsuda 0.0107 0.4887 0.0290 1.0742
Proposed 0.0081 0.4439 0.0134 0.3664
Oustaloup 5 0.0027 0.2530 0.0131 2.3867
Refined 0.0116 0.1664 0.0096 0.5042
Oustaloup
Matsuda 0.0035 0.2407 0.0112 0.6956
Proposed 0.0047 0.1613 0.0059 0.1491
Oustaloup 6 0.0007 0.3016 0.0154 2.4373
Refined 0.0119 0.0887 0.0075 0.5241
Oustaloup
Matsuda 0.0018 0.0795 0.0067 0.3352
Proposed 0.0016 0.0980 0.0022 0.0871
3.5 Performance Analysis of Approximation Algorithms 113

1
Exact
Oustaloup
0.9 Refined Oustaloup
Matsuda
Proposed Method
0.8
Amplitude

0.7

0.6

0.5

0.4
0 50 100 150 200 250
Time (s)

Fig. 3.6 Step response of s 0.1 using curve fitting for N = 6

Table 3.5 Comparison of MAE’s for step response of s 0.1

Technique Order t1 ∈ (0, 125) t2 ∈ (0, 250) H∞ -norm Stability
Oustaloup 4 0.0215 0.0441 1.5849 Stable
Refined 0.0547 0.0809 1.7425 Stable
Oustaloup
Matsuda 0.0144 0.0122 1.8280 Stable
Proposed 0.0048 0.0194 1.7576 Stable
Oustaloup 5 0.0237 0.0455 1.5849 Stable
Refined 0.0540 0.0813 1.7425 Stable
Oustaloup
Matsuda 0.0110 0.0120 1.8905 Stable
Proposed 0.0043 0.0082 1.8239 Stable
Oustaloup 6 0.0247 0.0463 1.5849 Stable
Refined 0.0538 0.0814 1.7425 Stable
Oustaloup
Matsuda 0.0103 0.0110 1.9462 Stable
Proposed 0.0071 0.0052 1.8809 Stable
114 3 Approximation Techniques

Exact
Oustaloup
5
Magnitude (dB)

Refined Oustaloup
Matsuda
0 Proposed Method

-5

-10
10-3 10-2 10-1 100 101 102 103

10
Phase (deg)

0
10-3 10-2 10-1 100 101 102 103
Frequency (rad/sec)

Fig. 3.7 Bode response of s 0.1 using transfer function estimation algorithm for N = 4

3.6 Approximation Table for Fractional-order

Differentiator

The approximated transfer functions for s α (α = 0.1, 0.2, . . . , 0.9) using both the
curve fitting approximation and transfer function estimation algorithm are given in
Tables 3.6 and 3.7 respectively [31].
These approximated transfer functions can be used directly to obtain the approx-
imation of fractional-order based systems and controllers. The comparison of the
accuracy of the proposed approaches with Oustaloup, refined Oustaloup and Mat-
suda is shown in Table 3.8. From the table and for the desired frequency of ω2 , both
the proposed approaches have the least MAE in all the values of α for both magnitude
and phase.
Furthermore, for the desired frequency of ω1 , the proposed approach outper-
formed all the others in phase response. However, for magnitude, the Oustaloup pro-
duced better approximation in all but two cases (α = 0.6, 0.9) where the proposed
approaches are better. It should be noted despite the Oustaloup being the best for ω1
magnitude, the proposed approaches were very close in all the cases. On the other
hand, from the time domain comparison, it can be noted that for longer time periods,
3.6 Approximation Table for Fractional-order Differentiator 115

Exact
Oustaloup
5 Refined Oustaloup
Magnitude (dB)

Matsuda
0 Proposed Method

-5

-10
10-3 10-2 10-1 100 101 102 103

10
Phase (deg)

0
10-3 10-2 10-1 100 101 102 103
Frequency (rad/sec)

Fig. 3.8 Bode response of s 0.1 using transfer function estimation algorithm for N = 5

Table 3.6 Approximation table for s α using curve fitting based approximation

sα Approximated transfer function using curve fitting

1.777s 5 + 123.9s 4 + 873.4s 3 + 909.9s 2 + 137.7s + 1.914
s 0.1
s 5 + 90.81s 4 + 785.4s 3 + 985s 2 + 182.9s + 3.335
3.233s 5 + 223s 4 + 1624s 3 + 1762s 2 + 275.3s + 3.725
s 0.2
s 5 + 116.8s 4 + 1279s 3 + 2011s 2 + 473s + 11.14
6.048s 5 + 413.7s 4 + 3111s 3 + 3513s 2 + 565.4s + 7.36
s 0.3
s 5 + 151s 4 + 2089s 3 + 4115s 2 + 1224s + 37.33
11.7s 5 + 794.7s 4 + 6166s 3 + 7236s 2 + 1198s + 14.71
s 0.4
s 5 + 197.2s 4 + 3438s 3 + 8478s 2 + 3181s + 125.9
23.59s 5 + 1594s 4 + 12740s 3 + 15520s 2 + 2632s + 29.63
s 0.5
s 5 + 262.6s 4 + 5753s 3 + 17720s 2 + 8368s + 429.6
50.26s 5 + 3381s 4 + 27790s 3 + 35050s 2 + 6070s + 59.93
s 0.6
s 5 + 361.4s 4 + 9918s 3 + 38090s 2 + 22550s + 1498
116s 5 + 7769s 4 + 65570s 3 + 85390s 2 + 15010s + 121.2
s 0.7
s 5 + 526.7s 4 + 18060s 3 + 86240s 2 + 63740s + 5453
305.7s 5 + 20410s 4 + 176400s 3 + 236400s 2 + 41890s + 244.3
s 0.8
s 5 + 857.9s 4 + 36660s 3 + 217000s 2 + 199500s + 21830
1092s 5 + 72700s 4 + 642000s 3 + 881700s 2 + 156000s + 489.3
s 0.9
s 5 + 1852s 4 + 98540s 3 + 720500s 2 + 819500s + 113800
116 3 Approximation Techniques

Exact
Oustaloup
5 Refined Oustaloup
Magnitude (dB)

Matsuda
0 Proposed Method

-5

-10
10-3 10-2 10-1 100 101 102 103

10
Phase (deg)

0
10-3 10-2 10-1 100 101 102 103
Frequency (rad/sec)

Fig. 3.9 Bode response of s 0.1 using transfer function estimation algorithm for N = 6

Table 3.7 Approximation table for s α using transfer function estimation algorithm

sα Approximated transfer function using transfer function estimation

1.824s 5 + 177s 4 + 1839s 3 + 2326s 2 + 330.3s + 4.043
s 0.1
s 5 + 124.8s 4 + 1600s 3 + 2501s 2 + 444.4s + 7.103
3.455s 5 + 361.7s 4 + 4476s 3 + 7173s 2 + 1276s + 17.44
s 0.2
s 5 + 171.8s 4 + 3183s 3 + 7732s 2 + 2170s + 51.73
6.807s 5 + 762.6s 4 + 11040s 3 + 22190s 2 + 4988s + 77.03
s 0.3
s 5 + 234.5s 4 + 6075s 3 + 22370s 2 + 1.001 × 104 s + 370.7
13.99s 5 + 1668s 4 + 2.782e04s 3 + 69160s 2 + 19720s + 343.9
s 0.4
s 5 + 320.5s 4 + 1.129 × 104 s 3 + 61150s 2 + 43410s + 2584
30.19s 5 + 3800s 4 + 71770s 3 + 216200s 2 + 77260s + 1479
s 0.5
s 5 + 442.9s 4 + 20660s 3 + 1.588 × 105 s 2 + 1.742 × 105 s + 16870
69.07s 5 + 9111s 4 + 191400s 3 + 681900s 2 + 298900s + 5913
s 0.6
s 5 + 627.5s 4 + 37900s 3 + 398300s 2 + 651000s + 101400
171.2s 5 + 23420s 4 + 536800s 3 + 2.2 × 106 s 2 + 1.1 × 106 s + 21150
s 0.7
s 5 + 933.1s 4 + 71490s 3 + 990800s 2 + 2.3 × 106 s + 557300
484.9s 5 + 68230s 4 + 1.6 × 106 s 3 + 7.7 × 106 s 2 + 4.6 × 106 s + 68530
s 0.8
s 5 + 1541s 4 + 147600s 3 + 2.6 × 106 s 2 + 8.3 × 106 s + 3.0 × 106
1855s 5 + 2.6 × 105 s 4 + 6.9 × 106 s 3 + 3.5 × 107 s 2 + 2.3 × 107 s + 1.9 × 105
s 0.9
s 5 + 3349s 4 + 3.9 × 105 s 3 + 8.7 × 106 s 2 + 3.7 × 107 s + 1.9 × 107
3.6 Approximation Table for Fractional-order Differentiator 117

Table 3.8 Numerical and stability analysis of approximated transfer functions using both the
proposed algorithms
sα Technique ω1 ∈ (10−1 , 10) ω2 ∈ (10−2 , 102 ) t1 ∈ t2 ∈ H∞ -norm Stability
Mag Phase Mag. Phase (0, 125) (0, 250)
s 0.1 Oustaloup 0.0027 0.2530 0.0131 2.3867 0.0237 0.0455 1.5849 Stable
Refined Oustaloup 0.0116 0.1664 0.0096 0.5042 0.0540 0.0813 1.7425 Stable
Matsuda 0.0035 0.2407 0.0112 0.6956 0.0110 0.0120 1.8905 Stable
Proposed 0.0047 0.1613 0.0059 0.1491 0.0043 0.0082 1.8239 Stable
s 0.2 Oustaloup 0.0043 0.5102 0.0400 4.7832 0.0265 0.0510 2.5119 Stable
Refined Oustaloup 0.0261 0.3213 0.0288 0.9244 0.0832 0.1065 3.0744 Stable
Matsuda 0.0075 0.4588 0.0313 1.3285 0.0286 0.0268 3.5911 Stable
Curve Fitting 0.0048 0.3045 0.0257 0.4257 0.0208 0.0175 2.6816 Stable
TF Estimate 0.0076 0.4074 0.0102 0.2617 0.0181 0.0172 3.4554 Stable
s 0.3 Oustaloup 0.0068 0.7757 0.0931 7.1992 0.0262 0.0450 3.9811 Stable
Refined Oustaloup 0.0445 0.4539 0.0498 1.2581 0.0943 0.1045 5.5103 Stable
Matsuda 0.0117 0.6336 0.0636 1.8410 0.0427 0.0363 6.8910 Stable
Curve Fitting 0.0069 0.4015 0.0479 0.5379 0.0345 0.0233 4.4710 Stable
TF Estimate 0.0072 0.6095 0.0125 0.3349 0.0300 0.0251 6.8073 Stable
s 0.4 Oustaloup 0.0091 1.0534 0.1952 9.6440 0.0272 0.0378 6.3096 Stable
Refined Oustaloup 0.0683 0.5553 0.0825 1.5059 0.0942 0.0916 10.0820 Stable
Matsuda 0.0158 0.7484 0.1104 2.1852 0.0511 0.0400 13.4464 Stable
Curve Fitting 0.0111 0.4452 0.0675 0.6272 0.0433 0.0263 7.4974 Stable
TF Estimate 0.0124 0.4709 0.0131 0.3715 0.0365 0.0296 13.9939 Stable
s 0.5 Oustaloup 0.0111 1.3481 0.3877 12.1269 0.0293 0.0319 10.0000 Stable
Refined Oustaloup 0.0991 0.6192 0.1268 1.6761 0.0881 0.0761 18.9737 Stable
Matsuda 0.0194 0.7919 0.1721 2.3275 0.0546 0.0398 26.9262 Stable
Curve Fitting 0.0151 0.4306 0.0845 0.6768 0.0476 0.0270 12.7730 Stable
TF Estimate 0.0124 0.4625 0.0124 0.4062 0.0393 0.0307 30.1873 Stable
s 0.6 Oustaloup 0.0121 1.6874 0.7450 14.6585 0.0314 0.0276 15.8489 Stable
Refined Oustaloup 0.1393 0.6424 0.2471 1.7973 0.0794 0.0617 37.1951 Stable
Matsuda 0.0216 0.7591 0.2433 2.2501 0.0548 0.0374 56.1234 Stable
Curve Fitting 0.0162 0.3525 0.0949 0.6501 0.0488 0.0265 22.0811 Stable
TF Estimate 0.0101 0.7294 0.0115 0.4299 0.0398 0.0294 69.0747 Stable
s 0.7 Oustaloup 0.0115 2.0579 1.3991 17.2459 0.0330 0.0256 25.1189 Stable
Refined Oustaloup 0.1916 0.6247 0.6104 1.9275 0.0703 0.0496 77.7765 Stable
Matsuda 0.0217 0.6520 0.3089 1.9530 0.0529 0.0340 124.8754 Stable
Curve Fitting 0.0149 0.2691 0.0950 0.5576 0.0483 0.0245 38.8116 Stable
TF Estimate 0.0125 0.6856 0.0105 0.4451 0.0393 0.0267 171.2372 Stable
s 0.8 Oustaloup 0.0088 2.4601 2.5821 19.8955 0.0338 0.0251 39.8107 Stable
Refined Oustaloup 0.2596 0.5840 1.4006 2.0564 0.0618 0.0402 182.9633 Stable
Matsuda 0.0187 0.4792 0.3369 1.4538 0.0499 0.0305 313.3221 Stable
Curve Fitting 0.0109 0.2380 0.0809 0.4085 0.0468 0.0222 74.6020 Stable
TF Estimate 0.0105 0.4763 0.0089 0.4404 0.0387 0.0235 484.8823 Stable
s 0.9 Oustaloup 0.0059 2.8934 4.6972 22.6126 0.0340 0.0205 63.0957 Stable
Refined Oustaloup 0.3478 0.5762 3.0330 2.2127 0.0543 0.0330 573.8763 Stable
Matsuda 0.0117 0.2553 0.2680 0.7866 0.0466 0.0273 1052.5 Stable
Curve Fitting 0.0055 0.1535 0.0499 0.2176 0.0451 0.0249 150.9561 Stable
TF Estimate 0.0057 0.2575 0.0059 0.3492 0.0401 0.0210 1855.5 Stable
118 3 Approximation Techniques

1
Exact
Oustaloup
0.9 Refined Oustaloup
Matsuda
Proposed Method
0.8
Amplitude

0.7

0.6

0.5

0.4
0 50 100 150 200 250
Time (s)

Fig. 3.10 Step response of s 0.1 using transfer function estimation algorithm for N = 4

the proposed approaches gives a better approximation than the other approaches.
Furthermore, it can also be noted that all the approximated transfer functions using
various techniques are stable with a finite value of H∞ -norm.

3.7 Simulation Study

In this section, a simulation study will be conducted on five examples of fractional-

order based controllers and systems. The selected systems are differentiator, inte-
grator, PID controller, filter and higher-order transfer function. The obtained results
from the proposed approaches will be compared with Oustaloup, refined Oustaloup
and Matsuda approximations in terms of MAE’s of frequency and step responses.
Furthermore, the comparison will also be done for the stability analysis of approx-
imated transfer functions. In all the cases, the selected desired frequency range ω
and the order of approximation N are chosen as (10−3 , 103 ) and 5 respectively. Fur-
thermore, for effective approximation, as mentioned earlier in Sect. 3.2, the order of
s α is limited to [−1, 1] range as defined in Eq. (3.8).
3.7 Simulation Study 119

1
Exact
Oustaloup
0.9 Refined Oustaloup
Matsuda
Proposed Method
0.8
Amplitude

0.7

0.6

0.5

0.4
0 50 100 150 200 250
Time (s)

Fig. 3.11 Step response of s 0.1 using transfer function estimation algorithm for N = 5

3.7.1 Fractional-order Differentiator

This example studies the performance of the proposed approach on a fractional-

order differentiator (G(s)) given in Eq. (3.21) and compares it with exact solution and
other approximation methods. The approximated transfer function using the proposed
curve fitting and transfer function estimation approaches are given in Eqs. (3.22) and
(3.23) respectively.
G(s) = s 0.26 (3.21)

7.147s 5 + 1836s 4 + 24930s 3 + 33930s 2 + 4811s + 29.74

G(s) ≈ (3.22)
s 5 + 597.2s 4 + 16080s 3 + 38500s 2 + 10220s + 163.4

8.808s 5 + 7141s 4 + 584900s 3 + 4540000s 2 + 2255000s + 32640

G(s) ≈
s 5 + 1585s 4 + 231800s 3 + 3409000s 2 + 3655000s + 149300
(3.23)
Similarly, the approximated transfer function of fractional-order differentiator
given in Sect. 3.7.1 using Oustaloup G O (s), refined Oustaloup G R (s) and Matsuda
G M (s) approximation techniques are given in Eqs. (3.24), (3.25) and (3.26) respec-
tively.
120 3 Approximation Techniques

1
Exact
Oustaloup
0.9 Refined Oustaloup
Matsuda
Proposed Method
0.8
Amplitude

0.7

0.6

0.5

0.4
0 50 100 150 200 250
Time (s)

Fig. 3.12 Step response of s 0.1 using transfer function estimation algorithm for N = 6

6.026s 5 + 1128s 4 + 1.253 × 104 s 3 + 8750s 2 + 384s + 1

G O (s) ≈ (3.24)
s 5 + 384s 4 + 8750s 3 + 1.253 × 104 s 2 + 1128s + 6.026

7.923s 13 + 1.029 × 104 s 12 + 1.664 × 106 s 11 + 1.832 × 107 s 10
+ 1.279 × 107 s 9 + 5.634 × 105 s 8 + 1566s 7 + 0.2747s 6
+ 3.039 × 10−6 s 5 + 2.122 × 10−12 s 4 + 9.347 × 10−20 s 3

+ 2.588 × 10−28 s 2 + 4.252 × 10−38 s
G R (s) ≈  13
s + 1885s 12 + 5.853 × 105 s 11 + 1.315 × 107 s 10
+ 1.883 × 107 s 9 + 1.705 × 106 s 8 + 1.01 × 104 s 7 + 5.758s 6
+ 0.0008956s 5 + 1.864 × 10−8 s 4 + 2.655 × 10−14 s 3

+ 2.398 × 10−21 s 2 + 1.362 × 10−29 s + 4.589 × 10−39
(3.25)
8.283s 5 + 2347s 4 + 2.872 × 104 s 3 + 1.982 × 104 s 2 + 749.1s + 1
G M (s) ≈ 5
s + 749.1s 4 + 1.982 × 104 s 3 + 2.872 × 104 s 2 + 2347s + 8.283
(3.26)
The Bode plots of the proposed approach in comparison to Oustaloup, refined
Oustaloup and Matsuda are presented in Fig. 3.13 while the numerical assessment of
3.7 Simulation Study 121

Exact
20 Oustaloup
Refined Oustaloup
Magnitude (dB)

10 Matsuda
Curve Fitting
0 TF Estimate

-10

-20
10-3 10-2 10-1 100 101 102 103

40

30
Phase (deg)

20

10

0
10-3 10-2 10-1 100 101 102 103
Frequency (rad/sec)

Fig. 3.13 Bode plot of fractional-order differentiator (G(s)) for different methods

the figure is given in Table 3.9. From the figure, it can be observed that the proposed
approaches are more accurate within the desired frequency range than the other
methods. Also, from the table, it can be seen that for the frequency range ω1 the
curve fitting approximation has the least MAE for both magnitude and phase of
0.0219 and 0.6364 respectively. On the other hand, for the frequency range ω2 and
ω3 , the transfer function estimation has the least errors for both magnitude and phase.
To further evaluate the performance of the proposed approach, the step responses
of all the compared techniques is shown in Fig. 3.14 while numerical assessment of
the figure is given in the same Table 3.10. In the figure, the exact time response of
the system G(s) will be obtained using Eq. (3.12) is given in Eq. (3.27).

0.26 × (0.26)
g(t) = (3.27)
t 0.26
From both figure and table, it can be observed that for the time period between 0
and 125 s, the Oustaloup has the least MAE of 0.0315 while for the time period from 0
to 250 s, the curve fitting approach has the least MAE of 0.0212. This is an indication
that for longer time periods, the proposed approach gives a better approximation
122 3 Approximation Techniques

Table 3.9 Comparison of MAE’s for Bode response of G(s)

Technique ω1 ∈ (10−1 , 10) ω2 ∈ (10−2 , 102 ) ω3 ∈ (10−3 , 103 )
Magnitude Phase Magnitude Phase Magnitude Phase
Oustaloup 0.0399 1.2785 0.0598 2.1321 0.1031 5.8981
Refined 0.0593 1.2832 0.0774 1.5645 0.1890 2.0613
Oustaloup
Matsuda 0.0442 1.4149 0.0722 2.3276 0.4328 2.9122
Curve 0.0219 0.6364 0.0662 1.4370 0.2470 2.5384
Fitting
TF Estimate 0.0360 1.3301 0.0377 0.6927 0.0472 0.3968

1
Exact
0.9 Oustaloup
Refined Oustaloup
0.8 Matsuda
Curve Fitting
TF Estimate
0.7
Amplitude

0.6

0.5

0.4

0.3

0.2

0.1
0 50 100 150 200 250
Time (s)

Fig. 3.14 Step response of fractional-order differentiator (G(s)) for different methods

than the other approaches. Furthermore, the stability analysis of G(s) for different
methods given in table shows that all the approximation techniques are stable with a
finite value of H∞ -norm.
3.7 Simulation Study 123

Table 3.10 Comparison of MAE’s for step response of G(s)

Technique t1 ∈ (0, 125) t2 ∈ (0, 250) H∞ -norm Stability
Oustaloup 0.0315 0.0213 6.0256 Stable
Refined Oustaloup 0.0493 0.0439 7.9227 Stable
Matsuda 0.0363 0.0266 8.2832 Stable
Curve Fitting 0.0368 0.0212 7.1467 Stable
TF Estimate 0.0515 0.0331 8.8083 Stable

3.7.2 Fractional-order Integrator

In this example, the proposed approaches are studied for a fractional-order integrator
(P(s)) given in Eq. (3.28). Here, a similar comparison is made as to the differentiator
example of Sect. 3.7.1. Therefore, the approximated transfer function model using
the proposed curve fitting and transfer function estimation techniques are given in
Eqs. (3.29) and (3.30) respectively.
1
P(s) = (3.28)
s 0.6

s 5 + 2134s 4 + 209400s 3 + 1.6 × 106 s 2 + 1.3 × 106 s + 88840

P(s) ≈ (3.29)
148.8s 5 + 46460s 4 + 929400s 3 + 1.8 × 106 s 2 + 373100s + 2353

s 5 + 5870s 4 + 3.2 × 106 s 3 + 2.9 × 108 s 2 + 3.7 × 109 s + 3.4 × 109

P(s) ≈
264.3s 5 + 3.2 × 105 s 4 + 6.0 × 107 s 3 + 1.7 × 109 s 2 + 5.2 × 109 s + 3.0 × 108
(3.30)
Similarly, the approximated transfer functions using Oustaloup PO (s), refined
Oustaloup PR (s) and Matsuda PM (s) techniques are given in Eqs. (3.31), (3.32) and
(3.33) respectively.

s 5 + 614.2s 4 + 2.239 × 104 s 3 + 5.129e04s 2 + 7384s + 63.1

PO (s) ≈ (3.31)
63.1s 5 + 7384s 4 + 5.129 × 104 s 3 + 2.239e04s 2 + 614.2s + 1
 13
s + 3392s 12 + 1.728 × 106 s 11 + 6.224 × 107 s 10
+ 1.426 × 108 s 9 + 2.068 × 107 s 8 + 1.99 × 105 s 7 + 209.6s 6
+ 0.0623s 5 + 2.129 × 10−6 s 4 + 4.857 × 10−12 s 3
−19 2 −27 −36

+ 7.017 × 10 s + 6.374 × 10 s + 3.436 × 10
PR (s) ≈ 
148.1s 13 + 1.819 × 105 s 12 + 1.938 × 107 s 11 + 1.338 × 108 s 10
+ 5.84 × 107 s 9 + 1.608 × 106 s 8 + 2795s 7 + 0.3065s 6
+ 2.12 × 10−6 s 5 + 9.255 × 10−13 s 4 + 2.548 × 10−20 s 3

+ 4.411 × 10−29 s 2 + 4.532 × 10−39 s
(3.32)
124 3 Approximation Techniques

Exact
40 Oustaloup
Refined Oustaloup
Magnitude (dB)

20 Matsuda
Curve Fitting
0 TF Estimate

-20

-40
10-3 10-2 10-1 100 101 102 103

0
Phase (deg)

-20

-40

-60

10-3 10-2 10-1 100 101 102 103

Frequency (rad/sec)

Fig. 3.15 Bode plots of fractional-order integrator (P(s)) for different methods

s 5 + 1807s 4 + 8.141 × 104 s 3 + 1.922 × 105 s 2 + 2.565 × 104 s + 156.4

PM (s) ≈
156.4s 5 + 2.565 × 104 s 4 + 1.922 × 105 s 3 + 8.141 × 104 s 2 + 1807s + 1
(3.33)

The frequency plots of the proposed approach in comparison to Oustaloup,

Refined Oustaloup and Matsuda are presented in Fig. 3.15 while the numerical assess-
ment of the figure is given in Table 3.11. From both figure and table, it can be observed
that compared to the other three techniques the proposed methods are more accurate
for the phase responses with the least error. However, for the magnitude response, the
curve fitting algorithm, refined Oustaloup and Oustaloup approximations are better
with MAE of 0.0282, 0.0112 and 0.0094 for ω1 , ω2 and ω3 respectively.
In the time domain, the step response of all the compared approaches is shown in
Fig. 3.16 while the numerical assessment is given in the same Table 3.12. It can be
observed from both response and table, that for the longest time range t ∈ (0, 250),
the proposed approach has the least MAE of 0.1146 while for the shorter time range
t ∈ (0, 125), Matsuda has the least MAE of 0.1280. This indicated that the pro-
posed approach achieves better approximation for longer time periods than the other
approaches. This is in agreement with the case of fractional-order differentiator given
3.7 Simulation Study 125

Table 3.11 Comparison of MAE’s for Bode response of P(s)

Technique ω1 ∈ (10−1 , 10) ω2 ∈ (10−2 , 102 ) ω3 ∈ (10−3 , 103 )
Magnitude Phase Magnitude Phase Magnitude Phase
Oustaloup 0.0311 1.6822 0.0117 3.3804 0.0094 14.0237
Refined 0.0353 1.6965 0.0112 2.0836 0.0153 2.7880
Oustaloup
Matsuda 0.0369 1.8590 0.0126 3.1571 0.0177 4.3198
Curve 0.0282 0.8503 0.0529 1.3945 0.0270 1.3961
Fitting
TF Estimate 0.2056 1.4566 0.0478 0.9195 0.0519 0.2173

40
Exact
35 Oustaloup
Refined Oustaloup
Matsuda
30 Curve Fitting
TF Estimate
25
Amplitude

20

15

10

0
0 50 100 150 200 250
Time (s)

Fig. 3.16 Step responses of fractional-order integrator (P(s)) for different methods

in Sect. 3.7.1. Furthermore, the stability analysis of various methods given in the table
shows that the approximation transfer functions of Oustaloup, Matsuda and proposed
approach are stable with proposed technique has the least H∞ -norm of 37.7568. From
the table, it can also be seen that the approximated transfer function using refined
Oustaloup is unstable.
126 3 Approximation Techniques

Table 3.12 Comparison of MAE’s for step response of P(s)

Technique t1 ∈ (0, 125) t2 ∈ (0, 250) H∞ -norm Stability
Oustaloup 0.2531 1.0345 63.0957 Stable
Refined 1.2426 2.1651 inf Unstable
Oustaloup
Matsuda 0.1280 0.3347 156.4489 Stable
Curve Fitting 0.8510 0.1146 37.7568 Stable
TF Estimate 1.1916 5.1254 11.4671 Stable

3.7.3 Fractional-order PID Controller

In this example, the proposed approaches are studied for the PIλ Dμ controller which
consists of both fractional-order differentiator and integrator given in Sects. 3.7.1
and 3.7.2 respectively. The transfer function of PIλ Dμ controller used for the study
is given as follows:
1
C(s) = 5 + 0.8 + 2s 0.5 (3.34)
s
For this example, the approximation will be done in two stages. First, the
fractional-order differentiator term (s 0.5 ) and the fractional-order integrator term
(1/s 0.8 ) will be approximated using the approximations given in Sect. 3.6. Then,
using the approximated transfer functions, the overall approximation of C(s) is done
by combining this integral and differential terms which will lead to the transfer func-
tion of C(s) given in Eqs. (3.35) and (3.36) for curve fitting and transfer function
estimation approaches respectively.

1.64 × 105 s 10 + 1.122 × 108 s 9 + 2.207 × 1010 s 8 + 9.906 × 1011 s 7
+ 1.546 × 1013 s 6 + 7.506 × 1013 s 5 + 1.384 × 1014 s 4 + 9.075 × 1013 s 3

+ 2.321 × 10 13 2
s + 1.953 × 10 12
s + 5.153 × 10 10
C(s) ≈ 
1379s 10 + 2.47 × 106 s 9 + 8.25 × 108 s 8 + 6.329 × 1010 s 7
+ 1.405 × 1012 s 6 + 8.811 × 1012 s 5 + 1.842 × 1013 s 4 + 1.202 × 1013 s 3

+ 2.418 × 1012 s 2 + 1.009 × 1011 s + 3.871 × 108
(3.35)
3.7 Simulation Study 127

50 Exact
Oustaloup
Magnitude (dB)

40 Refined Oustaloup
Matsuda
Curve Fitting
30
TF Estimate

20

10
10-3 10-2 10-1 100 101 102 103

50
Phase (deg)

-50

10-3 10-2 10-1 100 101 102 103

Frequency (rad/sec)

Fig. 3.17 Bode plots of fractional-order PID controller (C(s)) for different methods


5.598 × 105 s 10 + 1.431 × 109 s 9 + 1.212 × 1012 s 8 + 3.887 × 1014 s 7
+ 5.233 × 1016 s 6 + 2.829 × 1018 s 5 + 6.148 × 1019 s 4 + 4.541 × 1020 s 3

 + 1.164 × 1021 s 2 + 5.637 × 1020 s + 8.381 × 1019
C(s) ≈
2977s 10 + 1.606 × 107 s 9 + 2.213 × 1010 s 8 + 1.076 × 1013 s 7
+ 2.109 × 1015 s 6 + 1.626 × 1017 s 5 + 4.955 × 1018 s 4 + 4.893 × 1019 s 3

+ 1.582 × 1020 s 2 + 7.402 × 1019 s + 3.667 × 1018
(3.36)
The frequency plot of the approximated C(s) using the proposed approaches in
comparison to Oustaloup, refined Oustaloup and Matsuda are presented in Fig. 3.17
while the numerical assessment of the figure is given in Table 3.13. From both figure
and table, it can be observed that compared to the other three techniques the proposed
methods are more accurate for both magnitude and phase response. This is true for
all the desired frequency ranges ω1 , ω2 and ω3 . This can be seen more clearly from
the table.
For the magnitude response, the proposed approaches has the least MAE of 0.0805,
0.2932 and 0.3419 for ω1 , ω2 and ω3 respectively. Similarly, for the phase response,
the proposed approaches have the least MAE of 0.4521, 0.9308 and 0.2766 for ω1 , ω2
128 3 Approximation Techniques

Table 3.13 Numerical analysis of C(s) for various methods

Technique ω1 ∈ (10−1 , 10) ω2 ∈ (10−2 , 102 ) ω3 ∈ (10−3 , 103 )
Magnitude Phase Magnitude Phase Magnitude Phase
Oustaloup 0.1700 0.9766 0.4032 2.5189 1.4721 10.6800
Refined 0.2688 1.5433 0.6408 1.8742 2.2084 2.5266
Oustaloup
Matsuda 0.1916 1.1085 0.5167 2.4684 5.6817 4.0801
Curve 0.0805 0.4521 0.3908 1.3567 1.2892 1.5815
Fitting
TF Estimate 0.3555 0.7057 0.2932 0.9308 0.3419 0.2766

Table 3.14 Stability analysis of C(s) for various methods

Technique H∞ -norm Stability
Oustaloup 256.2519 Stable
Refined Oustaloup inf Unstable
Matsuda 1.2269 ×103 Stable
Curve Fitting 133.1226 Stable
TF Estimate 187.5695 Stable

and ω3 respectively. Furthermore, the stability analysis shows that the approximation
transfer functions of Oustaloup, Matsuda and proposed approach are stable and that
of refined Oustaloup is unstable. From the stability analysis given in Table 3.14, it can
also be seen that the proposed curve fitting and transfer function estimation methods
having the least H∞ -norms of 133.1226 and 187.5695 respectively.
The result obtained here shows an improvement to the separate integrator and
differentiator cases. This indicates that combining the two using the proposed
approaches yields an overall better result. Hence, the overall improvement in system
performance.

3.7.4 Fractional-order Filter

Apart from the fractional-order controllers, the proposed approaches can also be used
to approximate the other fractional-order systems. Example of such is the fractional-
order filter given in Eq. (3.37).

1
F(s) = (3.37)
0.5s 1.15 +1
3.7 Simulation Study 129

0
Magnitude (dB)

Exact
Oustaloup
Refined Oustaloup
-50
Matsuda
Curve Fitting
TF Estimate
-100
10-3 10-2 10-1 100 101 102 103

0
Phase (deg)

-50

-100

10-3 10-2 10-1 100 101 102 103

Frequency (rad/sec)

Fig. 3.18 Bode plots of fractional-order filter (F(s)) for different methods

Like in the previous case of the PIλ Dμ controller, the approximation of F(s)
will be done in three stages. First, based on Eq. (3.8), the fractional-order derivative
s 1.15 is divided into s × s 0.15 . Then s 0.15 will be approximated using the proposed
approach. Finally, by substituting this in Eq. (3.37), the overall approximated transfer
function of F(s) is determined. Thus, the approximated transfer function using both
the proposed approaches are given in Eqs. (3.38) and (3.39).

s 5 + 387.7s 4 + 6767s 3 + 11190s 2 + 2004s + 19.27

F(s) ≈
1.504s 6 + 355.3s 5 + 4579s 4 + 11780s 3 + 11810s 2 + 2008s + 19.27
(3.38)
s 5 + 938s 4 + 66610s 3 + 356600s 2 + 109400s + 1224
F(s) ≈
1.661s 6 + 1039s 5 + 51610s 4 + 247200s 3 + 391600s 2 + 109700s + 1224
(3.39)
The frequency plots of the approximated filter using the proposed approaches in
comparison to Oustaloup, refined Oustaloup and Matsuda are presented in Fig. 3.18
while the numerical comparison of the figure is given in Table 3.15. From the results,
it can be seen that, just as in the case of a PIλ Dμ controller, the proposed methods
are more accurate for both magnitude and phase response compared to the other
three techniques. Moreover, Oustaloup approximation is also performed better for
the magnitude response of F(s).
130 3 Approximation Techniques

Table 3.15 Numerical analysis of F(s) for various methods

Technique ω1 ∈ (10−1 , 10) ω2 ∈ (10−2 , 102 ) ω3 ∈ (10−3 , 103 )
Magnitude Phase Magnitude Phase Magnitude Phase
Oustaloup 0.0061 0.7071 0.0010 1.2853 0.0001 1.3831
Refined 0.0067 0.7490 0.0011 0.9792 0.0001 1.2997
Oustaloup
Matsuda 0.0066 0.7896 0.0012 1.4194 0.0002 1.7922
Curve 0.0025 0.3451 0.0007 1.2658 0.0001 1.2313
Fitting
TF Estimate 0.0058 0.6640 0.0008 0.6186 0.0001 0.4495

Table 3.16 Stability analysis of F(s) for various methods

Technique H2 -norm H∞ -norm Stability
Oustaloup 0.9773 1.0273 Stable
Refined Oustaloup 0.9841 1.0276 Stable
Matsuda 0.9772 1.0275 Stable
Curve Fitting 0.9784 1.0267 Stable
TF Estimate 0.9796 1.0289 Stable

On the other hand, the stability analysis is given in Table 3.16 also shows that all
the compared techniques are better and stable with H2 and H∞ -norms of around 0.98
and 1.02 respectively.

3.7.5 Fractional-order Transfer Function

To further study the performance of the proposed approach, a fractional-order transfer

function (FOTF) given in Eq. (3.40) is considered [12, 26]. The FOTF in polynomial
form with fractional powers is given as

s+1
R(s) = (3.40)
10s 3.2 + 185s 2.5 + 288s 0.7 + 1

Like in the previous two cases of PIλ Dμ and F(s), the approximation will be
done in three stages. First, based on Eq. (3.8), the fractional-order derivatives s 3.2
and s 2.5 will be divided as s 3 × s 0.2 and s 2 × s 0.5 respectively. Then s 0.2 , s 0.5 and
s 0.7 will be approximated using the approximations given in Sect. 3.6. Finally, by
substituting all these in Eq. (3.40), the overall approximated transfer function of
R(s) is determined. Thus, the approximated transfer functions of R(s) using both
the proposed approaches are given in Eqs. (3.41) and (3.42).
3.7 Simulation Study 131

s 16 + 5206s 15 + 7.573 × 106 s 14 + 3.565 × 109 s 13
+ 5.807 × 1011 s 12 + 3.69 × 1013 s 11 + 9.463 × 1014 s 10
+ 1.085 × 1016 s 9 + 5.936 × 1016 s 8 + 1.626 × 1017 s 7
+ 2.361 × 1017 s 6 + 1.854 × 1017 s 5 + 7.647 × 1016 s 4
+ 1.53 × 1016 s 3 + 1.359 × 1015 s 2 + 4.618 × 1013 s

+ 4.001 × 10 11
R(s) ≈  (3.41)
44.31s 18 + 2.31 × 105 s 17 + 3.299 × 108 s 16 + 1.355 × 1011 s 15
+ 2.312 × 1013 s 14 + 1.738 × 1015 s 13 + 5.26 × 1016 s 12
+ 7.555 × 1017 s 11 + 5.299 × 1018 s 10 + 1.788 × 1019 s 9
+ 3.458 × 1019 s 8 + 4.593 × 1019 s 7 + 4.372 × 1019 s 6
+ 2.484 × 1019 s 5 + 6.997 × 1018 s 4 + 8.796 × 1017 s 3

+ 3.93 × 1016 s 2 + 5.396 × 1014 s + 2.125 × 1012
 16
s + 14170s 15 + 5.995 × 107 s 14 + 9.665 × 1010 s 13
+ 7.069 × 1013 s 12 + 2.522 × 1016 s 11 + 4.501 × 1018 s 10
+ 4.14 × 1020 s 9 + 1.942 × 1022 s 8 + 4.43 × 1023 s 7
+ 4.868 × 1024 s 6 + 2.555 × 1025 s 5 + 6.026 × 1025 s 4
+ 6.579 × 1025 s 3 + 3.278 × 1025 s 2 + 6.25 × 1024 s

+ 1.166 × 1023
R(s) ≈  (3.42)
51.32s 18 + 7.189 × 105 s 17 + 2.933 × 109 s 16 + 4.274 × 1012 s 15
+ 2.914 × 1015 s 14 + 1.015 × 1018 s 13 + 1.835 × 1020 s 12
+ 1.785 × 1022 s 11 + 9.375 × 1023 s 10 + 2.536 × 1025 s 9
+ 3.471 × 1026 s 8 + 2.294 × 1027 s 7 + 6.909 × 1027 s 6
+ 1.133 × 1028 s 5 + 1.343 × 1028 s 4 + 8.991 × 1027 s 3

+ 2.554 × 1027 s 2 + 1.728 × 1026 s + 2.571 × 1024

The frequency plots of the approximated R(s) using the proposed approaches in
comparison to Oustaloup, refined Oustaloup and Matsuda are presented in Fig. 3.19
while the numerical comparison of the figure is given in Table 3.17. From the results,
it can be seen that for phase response the curve fitting approach has the least MAE of
1.2067 for ω1 while for ω2 and ω3 transfer function estimation algorithm has the least
MAE of 0.9718 and 0.3777 respectively. For magnitude response, all the compared
techniques have provided the best approximation with almost zero error. Furthermore,
the stability analysis from the Table 3.18 also shows that all the approximated transfer
functions are stable with the proposed approaches has the least H2 and H∞ -norms.
132 3 Approximation Techniques

Magnitude (dB) 0

-50
Exact
Oustaloup
-100
Refined Oustaloup
Matsuda
-150 Curve Fitting
TF Estimate
-200
10-3 10-2 10-1 100 101 102 103

-50
Phase (deg)

-100

-150

-200
10-3 10-2 10-1 100 101 102 103
Frequency (rad/sec)

Fig. 3.19 Bode plots of FOTF (R(s)) for different methods

Table 3.17 Numerical analysis of R(s) for various methods

Technique ω1 ∈ (10−1 , 10) ω2 ∈ (10−2 , 102 ) ω3 ∈ (10−3 , 103 )
Magnitude Phase Magnitude Phase Magnitude Phase
Oustaloup 0.0001 1.5838 0.0000 2.5255 0.0000 5.1760
Refined 0.0002 1.7686 0.0000 1.5011 0.0001 1.9960
Oustaloup
Matsuda 0.0001 1.7556 0.0000 2.6874 0.0000 2.6449
Curve 0.0001 1.2067 0.0002 2.4408 0.0001 2.6240
Fitting
TF Estimate 0.0013 3.4518 0.0004 0.9718 0.0003 0.3777

Table 3.18 Stability analysis of R(s) for various methods

Technique H2 -norm H∞ -norm Stability
Oustaloup 0.0141 0.3042 Stable
Refined Oustaloup 0.0225 1.2395 Stable
Matsuda 0.0164 0.5885 Stable
Curve Fitting 0.0138 0.1883 Stable
TF Estimate 0.0103 0.0453 Stable
3.8 Summary 133

3.8 Summary

In this chapter, the curve fitting based approximation has been developed for the prac-
tical realization of fractional-order parameters using SK least-square iterative method
with Levy’s cost function. Furthermore, the approximation table for fractional-order
differentiator has also been developed. The results from the simulation study on var-
ious fractional-order based systems and controllers show that the proposed approx-
imation technique performed better compared to Oustaloup, refined Oustaloup and
Matsuda approximations. This is true for both time and frequency-domain analysis
using Step and Bode responses respectively. Furthermore, the numerical analysis in
terms of MAE shows that the proposed approach shows better approximation for
high-frequency range and longer time periods. The stability analysis in terms of
H-norms also confirms that the proposed approach is better and stable.

References

1. Krishna, B.T.: Studies on fractional order differentiators and integrators: a survey. Signal Pro-
cess. 91(3), 386–426 (2011)
2. Vinagre, B.M., Podlubny, I., Hernandez, A., Feliu, V.: Some approximations of fractional order
operators used in control theory and applications. Fract. Calc. Appl. Anal. 3(3), 231–248 (2000)
3. Li, Z., Liu, L., Dehghan, S., Chen, Y., Xue, D.: A review and evaluation of numerical tools for
fractional calculus and fractional order controls. Int. J. Control 90(6), 1165–1181 (2017)
4. Djouambi, A., Charef, A., BesançOn, A.: Optimal approximation, simulation and analog real-
ization of the fundamental fractional-order transfer function. Int. J. Appl. Math. Comput. Sci.
17(4), 455–462 (2007)
5. Deniz, F.N., Alagoz, B.B., Tan, N., Atherton, D.P.: An integer order approximation method
based on stability boundary locus for fractional-order derivative/integrator operators. ISA
Trans. 62, 154–163 (2016)
6. Merrikh-Bayat, F.: Rules for selecting the parameters of Oustaloup recursive approximation for
the simulation of linear feedback systems containing PIλ Dμ controller. Commun. Nonlinear
Sci. Numer. Simul. 17(4), 1852–1861 (2012)
7. Valério, D., Da Costa, J.S.: Ninteger: a non-integer control toolbox for MatLab. In: Proceedings
of Fractional Differentiation and Its Applications, Bordeaux, July 2004
8. Tepljakov, A., Petlenkov, E., Belikov, J.: Application of Newton’s method to analog and dig-
ital realization of fractional-order controllers. Int. J. Microelectron. Comput. Sci. 3(2), 45–52
(2012)
9. Valério, D., Trujillo, J.J., Rivero, M., Machado, J.T., Baleanu, D.: Fractional calculus: a survey
of useful formulas. Eur. Phys. J. Spec. Top. 222(8), 1827–1846 (2013)
10. Petráš, I.: Fractional-Order Nonlinear Systems: Modeling, Analysis and Simulation. Springer
Science & Business Media (2011)
11. Caponetto, R.: Fractional Order Systems: Modeling and Control Applications. World Scientific
(2010)
12. Monje, C.A., Chen, Y., Vinagre, B.M., Xue, D., Feliu-Batlle, V.: Fractional-Order Systems and
Controls: Fundamentals and Applications. Springer Science & Business Media (2010)
13. Xue, D., Zhao, C., Chen, Y.: A modified approximation method of fractional order system.
In: Proceedings of the 2006 IEEE International Conference on Mechatronics and Automation,
Luoyang, Henan, China, 25–28 June 2006
14. Das, S.: Functional Fractional Calculus. Springer Science & Business Media (2011)
134 3 Approximation Techniques

15. Du, B., Wei, Y., Liang, S., Wang, Y.: Rational approximation of fractional order systems by
vector fitting method. Int. J. Control Autom. Syst. 15(1), 186–195 (2017)
16. Oustaloup, A., Levron, F., Mathieu, B., Nanot, F.M.: Frequency-band complex noninteger
differentiator: characterization and synthesis. IEEE Trans. Circuits Syst. I-Fundam. Theor.
Appl. 47(1), 25–39 (2000)
17. Meng, L., Xue, D.: A new approximation algorithm of fractional order system models based
optimization. J. Dyn. Syst. Meas. Control 134(4), 044504 (2012)
18. Wei, Y., Gao, Q., Peng, C., Wang, Y.: A rational approximate method to fractional order systems.
Int. J. Control Autom. Syst. 12(6), 1180–1186 (2014)
19. Krajewski, W., Viaro, U.: A method for the integer-order approximation of fractional-order
systems. J. Frankl. Inst.-Eng. Appl. Math. 351(1), 555–564 (2014)
20. Atherton, D.P., Tan, N., Yüce, A.: Methods for computing the time response of fractional-order
systems. IET Control Theory Appl. 9(6), 817–830 (2014)
21. Liang, S., Peng, C., Liao, Z., Wang, Y.: State space approximation for general fractional order
dynamic systems. Int. J. Syst. Sci. 45(10), 2203–2212 (2014)
22. Sheng, H., Chen, Y., Qiu, T.: Fractional Processes and Fractional-Order Signal Processing:
Techniques and Applications. Springer Science & Business Media (2011)
23. Yüce, A., Deniz, F.N., Tan, N.: A new integer order approximation table for fractional order
derivative operators. IFAC-PapersOnLine 50(1), 9736–9741 (2017)
24. Mitkowski, W., Oprze˛dkiewicz, K.: An estimation of accuracy of Charef approximation. In:
Domek, S., Dworak, P. (eds.) Theoretical Developments and Applications of Non-integer Order
Systems: 7th Conference on Non-integer Order Calculus and Its Applications. Lecture Notes
in Electrical Engineering, vol. 357, pp. 71–80. Springer (2016)
25. Oprze˛dkiewicz, K.: Approximation method for a fractional order transfer function with zero
and pole. Arch. Control Sci. 24(4), 447–463 (2014)
26. Khanra, M., Pal, J., Biswas, K.: Rational approximation and analog realization of fractional
order transfer function with multiple fractional powered terms. Asian J. Control 15(3), 723–735
(2013)
27. Poinot, T., Trigeassou, J.C.: A method for modelling and simulation of fractional systems.
Signal Process. 83(11), 2319–2333 (2003)
28. Krajewski, W., Viaro, U.: On the rational approximation of fractional order systems. In: 2011
16th International Conference on Methods and Models in Automation and Robotics, Miedzyz-
droje, Poland, 22–25 Aug 2011
29. Shi, G.: On the nonconvergence of the vector fitting algorithm. IEEE Trans. Circuits Syst.
II-Express Briefs 63(8), 718–722 (2016)
30. Ozdemir, A.A., Gumussoy, S.: Transfer function estimation in system identification toolbox
via vector fitting. IFAC-PapersOnLine 50(1), 6232–6237 (2017)
31. Bingi, K., Ibrahim, R., Karsiti, M.N., Hassan, S.M., Harindran, V.R.: Frequency response based
curve fitting approximation of fractional-order PID controllers. Int. J. Appl. Math. Comput.
Sci. 29(2), 311–326 (2019)
 Part II
Scilab Based Toolbox for Fractional-order
Systems and PID Controllers
Chapter 4
Scilab Based Toolbox
for Fractional-order Systems
and PID Controllers

4.1 Introduction

Fractional-order systems have been extensively used in many engineering applica-

tions [1, 2]. This is due to more accurate modeling and controlling possibilities
they offer over integer-order systems [3, 4]. Fractional-order systems are useful in
studying the dynamical behavior of the systems in the area of physics, biology and
electrical engineering. They are also used to study chaotic systems and many oth-
ers. Moreover, fractional-order PID (FOPID) controllers which are derived based
on fractional-ordering of the parameters, allow for robust and stable control perfor-
mance [5–8]. This is because they are less sensitive to parameter changes [9]. The
controllers can also easily attain the iso-damping; a desirable property referring to a
state where the open-loop phase Bode plot of the system is flat [10].
A key issue with the implementation of fractional-order systems and controllers
is the numerical tools. Researchers have developed few toolboxes such as Com-
mande Robuste d’Ordre Non Entier (CRONE) [11–14], Non-integer (NINTEGER)
[15, 16], Fractional Order Modeling and Control (FOMCON) [17–19], FOPID [20],
fractional-order transfer function (@fotf) [21], fractional state space tool (FSST)
[22, 23], Sysquake FOPID [24–26], Discrete FOPID [4, 27], fractional integration
tool (FIT) [28] for fractional order modeling and control applications. A summary
of these toolboxes explaining the features and limitations is given in Table 4.1.
The CRONE toolbox was developed by CRONE research team and is dedicated to
the application of fractional-order derivatives in science and engineering. The toolbox
allows for mathematical modeling, system identification and CRONE control design
[11, 12, 14, 29]. On the other hand, the NINTEGER toolbox developed by Valério
presents the implementation of non-integer PID controllers and CRONE controllers
[15]. The toolbox also allows for the approximation of non-integer order derivatives,
functions for model development and frequency response plots. Unlike the toolboxes
mentioned earlier, the FOMCON toolbox incorporates advanced features such as sys-
tem identification, fractional-order PID control, and real-time implementation. The
toolbox is an extension of @fotf toolbox developed in [21] for fractional-order sys-
tem identification and PID controller design. Similarly, the authors in [20] developed
a fractional-order PID (FOPID) toolbox for designing and tuning the controller using

© Springer Nature Switzerland AG 2020 137

K. Bingi et al., Fractional-order Systems and PID Controllers,
Studies in Systems, Decision and Control 264,
https://doi.org/10.1007/978-3-030-33934-0_4
138 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

Table 4.1 Summary of numerical toolboxes for fractional-order systems and PID controller
Name Features Software Syntaxa Downloadb Reference
CRONE Non-integer differentiation, Matlab/ – ✓ [29, 30]
differential equations, system Simulink
identifications, inverse
Laplace transform and
CRONE control
NINTEGER Fractional-order Matlab/ nipid() ✓ [15]
approximation techniques, Simulink
system identification function
and analysis, GUI for
controller design
FOPID FOPID design, Matlab – ✗ [20]
Approximation techniques,
Step, Nyquist, Bode, Nichols
plots
@fotf Overload functions were Matlab fotf() ✓ [21]
developed for fractional
transfer function, time and
frequency domain, stability
analysis
FSST Simulink blocks for Simulink – ✓ [22, 23]
stochastic and non-stochastic
fractional-order systems and
fractional Kalman filter.
FOMCON System identification in both Matlab/ fpid() ✓ [17, 31]
time and frequency domains, Simulink
FOPID design, tuning and
implementation, Step, Bode,
Nyquist and Nichols plots
Sysquake Integrative tool for FOPID Sysquake GUI ✗ [24–26]
FOPID design and analysis
DFOC Discrete version of FOPID Matlab DFOC() ✓ [4, 27]
controller
FIT Riemann-Liouville type of C++, Matlab – ✗ [28]
differentiation and integration
a The‘Syntax’ column denotes the MATLAB command for FOPID controller
b The‘Download’ column denotes whether the toolbox is publicly available for download either
from MATLAB file exchange or other websites/forums

the steepest descent method. On the other hand, a MATrix LABoratory (MATLAB)
script defining the discrete form of FOPID controller is given in [4]. A common
feature of these toolboxes is that they are based on MATLAB/Simulink software.
Apart from these toolboxes, researchers also developed MATLAB scripts for sev-
eral fractional-order based functions. These functions include fractional-order defi-
nitions, differentiator, integrator, inverse Laplace transforms, root locus and approx-
imation techniques. The list of these function retrieved from file exchange central of
MATLAB with the syntax is shown in Table 4.2.
MATLAB is a powerful software that provides many numerical computations.
However, despite many advantages, the software is expensive and generally have
limited licenses in terms of time and number of installations [32]. On the other hand,
4.1 Introduction 139

Table 4.2 List of fractional-order functions retried from MATLAB central

Category Method Script/Function name
Fractional-order Definitions Grünwald-Letnikov fderiv()
glfdiff()
Mittag-Leffler mlf()
Fractional-order Differentiator Fourier series fourier_diffint()
Power series dfod1(),dfod2()
Approximations Techniques Refined Oustaloup ora_foc()
Impulse response irid_fod()
Others Inverse Laplace invlap()
Root locus forlocus()

the Scientific laboratory (Scilab) is extremely effective and available as a free open-
source software package for scientific computations. It includes hundreds of general
and specialized functions for numerical computations, 2D and 3D visualization and
data analysis [33, 34]. The software is organized in libraries called toolboxes, which
cover areas such as simulation, optimization, control systems, and signal process-
ing. It also provides its own dynamic systems modeler and simulator called Xcos.
Recently, researchers have applied Scilab to solve ordinary and partial differential
equations [35]. Furthermore, it has also been used for digital image processing [36],
control and optimization [37–40]. Thus, there is a scope to develop an open-source
toolbox in this open-source software for fractional-order systems and PID controllers.

4.2 Fractional-order Differintegral Operator

The fractional-order differintegral operator; a combined differentiator/integrator, Dαt

of order α ∈ R that generalizes the notations for integrals (α < 0) and derivatives
(α > 0) is defined as
⎧ t
⎪
⎪ f (τ )dτ α α < 0,
⎨ 0
Dαt f (t) = f α(t) α = 0, (4.1)
⎪
⎪ d f (t)
⎩ α > 0.
dt α
The fractional-order differentiator in Eq. (4.1) can be defined in many forms such
as Grünwald-Letnikov (GL), Riemann-Liouville (RL), Caputo etc. Among these
forms, GL definition is the most straightforward and widely used and is defined as
follows:
1  (α)
t/ h
α
Dt f (t) ≈ α ω f (t − j h) (4.2)
h j=0 j
140 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

where h is the step size, ω j are the coefficients which can be calculated as follows:

α + 1 (α)
ω0(α) = 1; ω(α)
j = 1− ω j−1 , j = 1, 2, 3, . . . (4.3)
j

Based on the above definition, the αth derivative of f (t) can be evaluated in Scilab
using the following function with syntax df=grunwald(f,t,alpha,h).
function df=grunwald(f,t,alpha,h)
df(1)=0; w=1; f=f’;
for j=2:length(t)
w(j)=w(j-1)*(1-((alpha+1)/(j-1)));
end
w=w’;
for i=2:length(t)
df(i)=w(1:i)*[f(i:-1:1)];
end
df=df/hˆalpha;
endfunction

Also, the Laplace transform of fractional derivative in Eq. (4.2) at zero initial
conditions is given as
L{Dαt f (t); s} = s α F(s) (4.4)

where F(s) is the Laplace transform of f (t).

4.3 Approximation of Fractional-order Operator

The fractional-order operator s α in Eq. (4.4) can be approximated using any of

the approximation technique such as Oustaloup, refined Oustaloup, Matsuda, Curve
Fitting, Transfer Function Estimation, finite impulse, and infinite impulse responses.
The techniques will be briefly explained subsequently.

4.3.1 Oustaloup Approximation

The Oustaloup approximation of s α in the desired frequency range (ωl , ωh ) is

defined as:

N
s + ωk
s α ≈ ωhα , 0<α<1 (4.5)
k=1
s + ωk

where
4.3 Approximation of Fractional-order Operator 141

• α is the order of fractional-order derivative,

• N is the order of approximation,
• (ωl , ωh ) is the frequency range of interest and
• ωk and ωk are the zeros and poles.
The ωk and ωk of Eq. (4.5) are computed as:
 2k−1−α
ωh 2N
ωk = ωl (4.6)
ωl
 2k−1+α
ωh 2N
ωk = ωl (4.7)
ωl

Therefore, the function representing the Oustaloup approximation with syntax

G=oustaloup(alpha,N,wl,wh) can be written in Scilab as follows:
function G=oustaloup(alpha,N,wl,wh)
k=1:N; wu=sqrt(wh/wl);
wkz=wl*wu.ˆ((2*k-1-alpha)/N);
wkp=wl*wu.ˆ((2*k-1+alpha)/N);
G=zpk(-wkz,-wkp,whˆalpha,"c"); G=zpk2tf(G);
endfunction

4.3.2 Refined Oustaloup Approximation

The refined Oustaloup approximation of s α in the desired frequency range (ωl , ωh )

is defined as:
 α 
N
dωh ds 2 + bωh s s + ωk
sα ≈ , 0<α<1 (4.8)
b d(1 − α)s 2 + bωh s + dα k=−N
s + ωk

where ωk and ωk are the respective zeros and poles defined in Eqs. (4.6) and (4.7)
respectively.
In order to achieve good approximation, constants b and d in Eq. (4.8) are selected
as 10 and 9 respectively. Thus, in Scilab, the function defining this approximation
with syntax G=refOustaloup(alpha,N,wl,wh) is as follows:
function G=refOustaloup(alpha,N,wl,wh)
b=10; d=9; k=-N:N;
wu=sqrt(wh/wl);
K=(d*wh/b)ˆalpha;
wkz=wl*wu.ˆ((2*k-1-alpha)/N);
wkp=wl*wu.ˆ((2*k-1+alpha)/N);
G1=zpk(-wkz’,-wkp’,K,"c");
142 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

exec(’tf.sci’); s=tf(’s’);
G2=syslin(’c’,((d*sˆ2)+(b*wh*s))/((d*(1-alpha)*sˆ2)...
+(b*wh*s)+(d*alpha)));
G=G1*G2;
G=zpk2tf(G);
endfunction

In the above function, the tf.sci defining the transfer function variable ‘s’ can
be written as follows:
function s=tf(s)
s=poly(0,’s’);
s=syslin(’c’,s/((0*s)+1));
disp(’Continuous-time transfer function.’);
endfunction

4.3.3 Matsuda Approximation

The Matsuda approximation of s α is achieved in two steps. First, a rational model

of s α is obtained using continued fraction expansions method. Then, fitting of the
original function in a desired frequency points ω0 , ω1 , . . . , ω N is performed. Thus,
the approximated transfer function of s α is defined as:

s − ω0
s α ≈ d0 (ω0 ) + (4.9)
d1 (ω1 ) + d (ωs−ω
)+
1
s−ω2
2 2 ...

where
d0 (ω0 ) = |( jω0 )α | (4.10)

ω − ωk
dk+1 (ω) = , k = 0, 1, 2, . . . , N (4.11)
dk (ω) − dk (ωk )

where N is the sum of the total number of zeros and poles which is known as the
order of approximation. Here, N should be an even number, else the approximation
results in an improper transfer function.
Thus, based on the above algorithm, the approximation can be written with also
in Scilab with syntax G=matsuda(alpha,N,wl,wh) as following function.
function G=matsuda(alpha,N,wl,wh)
w=logspace(log10(wl),log10(wh),N+1);
k=20*log10(abs(w.ˆalpha)); K=abs(10.ˆ(k/20));
temp1=[];
for j=0:length(w)-1
temp(0+1,j+1)=K(j+1);
for i=1:j
4.3 Approximation of Fractional-order Operator 143

temp(i+1,j+1)=(w(j+1)-w(i))/(temp(i,j+1)-temp(i,i));
end
temp1=[temp1 temp(j+1,j+1)];
end
Gm=temp1(length(temp1));
for i=length(temp1)-2:-1:0
Gm=temp1(i+1)+(zpk([1 w(i+1)],1,1,"c")/Gm);
end
G=zpk2tf(Gm);
endfunction

4.3.4 Curve Fitting Approximation Algorithm

The procedure for obtaining the curve fitting approximation of s α using the frequency
response data is as follows:
1. Obtain the frequency response data for integer-order part of s α within the fre-
quency range (ωl , ωh ) using the function repfreq().
2. Obtain the exact frequency response data of s α by powering the data obtained in
the previous step with α.
3. Obtain the N th order transfer function model of response data using the function
frfit().
Based on the above procedure, the algorithm can be written as following function
with syntax G=curFit1(alpha,N,wl,wh)
function G=curFit1(alpha,N,wl,wh)
w=logspace(log10(wl),log10(wh));
s=poly(0,’s’); s=syslin(’c’,s/((0*s)+1));
F=repfreq(s,w);
f=F.ˆalpha; G=frfit(w,f,N);
endfunction

4.3.5 Transfer Function Estimation

Similarly, the procedure for estimating the integer-order transfer function of s α is as

follows:
1. Obtain the frequency response data for integer-order part of s α within the fre-
quency range using the function repfreq().
2. Obtain the exact frequency response data of s α by powering the data obtained in
the previous step with α.
3. Estimate the N th order transfer function model of response data using the function
frep2tf().
144 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

Thus, the above algorithm can be written as following function with syntax
G=curFit2(alpha,N,wl,wh)
function G=curFit2(alpha,N,wl,wh)
w=logspace(log10(wl),log10(wh),10000);
s = poly(0,’s’);
s=syslin(’c’,s/((0*s)+1));
F=repfreq(s,w);
f=F.ˆalpha;
[G,Err]=frep2tf(w,f,N);
endfunction

4.3.6 Power Series Expansion Techniques

The approximation of discretized fractional-order differentiator in the form of finite

impulse response (FIR) filter based on power series expansion is as follows:
 α
α 1  α 
0 DT X (z) = P S E 1 − z −1 n
T (4.12)
−α −1
≈T Rn (z )

where
• n is order of approximation or truncation,
• T is sampling period,
−1
• R is polynomial
 in the variable z of order n and  α
−1 α
• PSE{ 1 − z } denotes the power series expansion of the function 1 − z −1 .

 From Eq.
α
(4.12), it can be noted that the approximated power series expansion of
1 − z −1 in the form of FIR filter has only zeros. Therefore, the algorithm can be
written with syntax G=firPSE(alpha,N,T) as follows:
function G=firPSE(alpha,N,T)
if alpha>0 then numCoeff=cumprod([1,1-((alpha+1)./[1:N])]);
numCoeff=numCoeff(:,$:-1:1);
num=poly(numCoeff,’z’,’c’);
z=poly(0,’z’);
den=(Tˆalpha)*zˆN;
G=syslin(’d’,num,den);
end
if alpha<0 then denCoeff=cumprod([1,1-((-alpha+1)./[1:N])]);
denCoeff=denCoeff(:,$:-1:1);
den=poly(denCoeff,’z’,’c’);
z=poly(0,’z’);
num=(Tˆ-alpha)*zˆN;
4.3 Approximation of Fractional-order Operator 145

G=syslin(’d’,num,den);
end
endfunction

Similarly, the approximation of discretized fractional-order differentiator in the

form of infinite impulse response (IIR) filter based on power series expansion can
be written with syntax as G=iirPSE(alpha,N,T) follows:
function G=iirPSE(alpha,N,T)
num(1)=1; den(1)=1;
for i=1:N
num(i+1)=((-1)ˆi)*(gamma(abs(alpha)+1)./(gamma(i+1).* ...
gamma(abs(alpha)-i+1)));
den(i+1)=gamma(abs(alpha)+1)./(gamma(i+1).* ...
gamma(abs(alpha)-i+1));
end
num=num’; den=den’;
num=num(:,$:-1:1);
den=den(:,$:-1:1);
z=poly(0,’z’);
num=poly(num,’z’,’c’);
den=poly(den,’z’,’c’);
if alpha>0 then
G=((2/T)ˆalpha)*syslin(’d’,num,den);
end
if alpha<0 then
G=((2/T)ˆalpha)*syslin(’d’,den,num);
end
endfunction

4.4 Fractional-order Parameters

In this section, using the Scilab based approximation techniques given in Sect. 4.3,
the implementation of fractional-order differentiator and integrator will be briefly
explained subsequently.

4.4.1 Fractional-order Differentiator

The fractional-order differentiator (FOD) from Eq. (4.4) is defined as follows:

G d (s) = s α , 0 < α < 1 (4.13)

146 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

From Eq. (4.13), the implementation of G d (s) will be done with the use of approx-
imation algorithms with the syntax of Gd=fod(alpha,func, N,wl,wh),
where
• α is the order of differentiation,
• f unc is the approximation technique,
• N is the order of approximation and
• (ωl , ωh ) is the desired frequency range of interest.
The five available f unc options are “Oustaloup”, “RefinedOustaloup”,
“Matsuda”, “curveFitting” and “tfEstimate”. Therefore, the function defining the
fractional-order differentiator can be written with syntax G=fod(alpha,f,N,w)
as follows:
function G=fod(alpha,f,N,w)
wl=w(1); wh=w(2);
if f==’o’ then
exec(’oustaloup.sci’);
G=oustaloup(alpha,N,wl,wh);
elseif f==’r’ then
exec(’refOustaloup.sci’);
G=refOustaloup(alpha,N,wl,wh);
elseif f==’m’ then
exec(’matsuda.sci’);
G=matsuda(alpha,N,wl,wh);
elseif f==’cf1’ then
exec(’curFit1.sci’);
G=curFit1(alpha,N,wl,wh);
elseif f==’cf2’ then
exec(’curFit2.sci’);
G=curFit2(alpha,N,wl,wh);
else
disp(’Check the name of the approximation function’)
end
endfunction

4.4.2 Fractional-order Integrator

Similarly, the fractional-order integrator (FOI) from Eq. (4.4) is defined as follows:

1
G i (s) = , 0<α<1 (4.14)
sα
Thus, the foi function will be implemented with the help of fod with the similar
syntax as G=foi(alpha,f,N,w), where α is the order of integration. Therefore,
the function defining the foi is as follows:
4.4 Fractional-order Parameters 147

function G=foi(alpha,f,N,w)
exec(’fod.sci’);
G=1/fod(alpha,f,N,w);
endfunction

4.5 Fractional-order Transfer Function

A fractional-order system will be represented in the form of fractional-order transfer

function (FOTF) as follows:

Y (s) bm s γm + bm−1 s γm−1 + · · · + b1 s γ1 + b0

G(s) = = (4.15)
U (s) an s ηn + an−1 s ηn−1 + · · · + a1 s η1 + a0

where
• bi , ai ∈ R are the coefficients of numerator and denominator and
• γi , ηi ∈ R are the orders of numerator and denominator.
For the model in Eq. (4.15), the Scilab function can be established with the coef-
ficients and orders of numerator and denominator. Therefore, let n and np be the
coefficients and order of numerator respectively given as

n = [bm , bm−1 , . . . , b1 , b0 ]
(4.16)
np = [γm , γm−1 , . . . , γ1 , γ0 ]

Similarly, let us consider d and dp as the coefficients and order of denominator

also defined as
d = [an , an−1 , . . . , a1 , a0 ]
(4.17)
dp = [ηn , ηn−1 , . . . , η1 , η0 ]

Thus, the function syntax is G=fotf(n,np,d,dp,f,N,w). As mentioned

earlier f unc, N , ωl and ωh are the parameters of the approximation algorithm.
Consequently, the entire function in Scilab is as follows:
function G=fotf(n,np,d,dp,f,N,w)
s=poly(0,’s’); s=syslin(’c’,s/((0*s)+1));
[nr,nc]=size(n);[npr,npc]=size(np);
[dr,dc]=size(d);[dpr,dpc]=size(dp);
Gnum=0;Gden=0;
exec(’fod.sci’);
if nc==npc & dc==dpc then
for i=1:nc
npTemp(i)=fix(np(i));
np(i)=np(i)-npTemp(i);
Gnum=Gnum+n(i)*sˆnpTemp(i)*fod(np(i),f,N,w);
148 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

end
for i=1:dc
dpTemp(i)=fix(dp(i));
dp(i)=dp(i)-dpTemp(i);
Gden=Gden+d(i)*sˆdpTemp(i)*fod(dp(i),f,N,w);
end
G=syslin(’c’,Gnum/Gden);
else
disp(’Check the orders’)
disp(’Order of n is not equal to np or’)
disp(’the order of d is not equal to dp’)
end
endfunction

The result from the above function, G will be the approximated transfer function
of the FOTF given in Eq. (4.15). The transfer function data of G i.e., the coefficients
of numerator and denominator can be deduced using the following function:
function [n,d]=tfdata(G)
n=coeff(G.num);
d=coeff(G.den);
endfunction

where n and d are the coefficients of numerator and denominator.

Similarly, the zeros (z) and poles ( p) of the approximated transfer function G can
also be deduced using the following function:
function [z,p]=zeros_poles(G)
G=syslin(’c’,G);
z=roots(G.num);
p=roots(G.den);
endfunction

4.6 Fractional-order Filters

In this section, the design of fractional-order filters of order α with the use of approx-
imation techniques given in Sect. 4.3 are presented [41]. In all the cases, α is chosen
between 0 and 2. This is because the system will be stable if and only if 0 < α < 2.
On the other hand, for α = 2, the system will produce oscillatory behavior. For values
greater than 2, the system will be unstable.
4.6 Fractional-order Filters 149

4.6.1 Fractional-order Low-Pass Filter

The transfer function of fractional-order low-pass filter for 0 < α < 2 is given as
follows:
ωα
F(s) = α c α , 0 < α < 2 (4.18)
s + ωc

where
• α is the fractional-order parameter and
• ωc is the pole frequency.
The characteristics of the filter are obtained as
1
|F(s)|s= jω =   ω α  απ (4.19)
ω 2α
ωc
+2 ωc
cos 2
+1
 ω α 
−1 ω
sin απ 2
∠F(s)|s= jω = − tan  ω αc  απ (4.20)
ωc
cos 2 + 1

The peak frequency of the filter ω p is defined as the frequency at which the
magnitude is maximum or minimum (i.e., dωd
|F( jω)|ω=ω p = 0) and is calculated as
  
απ 1/α
ω p = ωc − cos (4.21)
2

Similarly, the right-phase frequency ωr p is defined as the frequency at which phase

is equal to ± π2 and is calculated as

ωc
ωr p =   1/α (4.22)
− cos απ2

The half-power frequency ωhp is defined as the frequency at which |F( jω)|ω=ωh =
|F( jω)|ω=ω p
√ and is obtained as
2

   1/α
απ απ
ωhp = ωc 1 + cos2 − cos (4.23)
2 2

Thus, the phase angle of filter at half-power frequency ωhp is defined as


−1
sin( απ )
∠F( jω)ω=ωhp = − tan 2 (4.24)
2 cos( απ
2
) + 1 + cos2 ( απ
2
)
150 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

Thus, the function with syntax [H,wp,wrp,wh,Hwh]=foLPF(alpha,wc,

f,N,w) representing the filter using Scilab is given as
function [H,wp,wrp,wh,Hwh]=foLPF(alpha,wc,f,N,w)
alphaTemp=fix(alpha);
exec(’fod.sci’);
s=poly(0,’s’); s=syslin(’c’,s/((0*s)+1));
alp=alpha-alphaTemp;
num=wcˆalpha;
den=((sˆalphaTemp)*fod(alp,f,N,w))+(wcˆalpha);
H=num/den;
wp=wc*(-cos(alpha*%pi/2))ˆ(1/alpha);
wrp=wc/(-cos(alpha*%pi/2))ˆ(1/alpha);
wh=wc*(sqrt(1+(cos(alpha*%pi/2)ˆ2))...
-cos(alpha*%pi/2))ˆ(1/alpha);
Hwh=-atan((sin((alpha*%pi)/2)))/(((2*cos(alpha*%pi/2)))...
+(sqrt(1+(cos(alpha*%pi/2)ˆ2))));
endfunction

Here, H denotes the approximated N th order transfer function of the filter.

4.6.2 Fractional-order High-Pass Filter

In a similar way, the fractional-order high-pass filter is given as follows:

ωcα s α
F(s) = , 0<α<2 (4.25)
s α + ωcα

The characteristics responses of the filter |F( jω)| and ∠F( jω) respectively are
defined as  α
ω
ωc
|F(s)|s= jω =   ω α  απ (4.26)
ω 2α
ωc
+2 ωc
cos 2
+1
 ω α  απ
απ ω
sin
− tan−1  ω αc
2
∠F(s)|s= jω =  απ (4.27)
2 ω
cos 2 + 1
c

As defined in Eqs. (4.21), (4.22) and (4.23), the peak, right-phase and half-power
frequencies are given in Eqs. (4.28), (4.29) and (4.30) respectively.
 1/α
ωc
ωp =  (4.28)
− cos απ
2
4.6 Fractional-order Filters 151
  1/α
απ
ωr p = − ωc cos (4.29)
2

    1/α
απ απ
ωhp = ωc 1 + cos2 + cos (4.30)
2 2

Thus, the phase angle of fractional-order high-pass filter at ωhp is defined as

 
απ −1
sin( απ )
∠F( jω)ω=ωhp = − tan 2 (4.31)
2 2 cos( 2 ) + 1 + cos2 ( απ
απ
)
2

Therefore, the function with syntax [H,wp,wrp,wh,Hwh]=foHPF(alpha,

wc,f,N,w) representing the filter using Scilab is given as
function [H,wp,wrp,wh,Hwh]=foHPF(alpha,wc,f,N,w)
alphaTemp=fix(alpha);
alp=alpha-alphaTemp;
exec(’fod.sci’);
s=poly(0,’s’); s=syslin(’c’,s/((0*s)+1));
num=(wcˆalp)*((sˆalphaTemp)*fod(alp,f,N,w));
den=((sˆalphaTemp)*fod(alp,f,N,w))+(wcˆalpha);
H=num/den;
wp=wc/(-cos(alpha*%pi/2))ˆ(1/alpha);
wrp=wc*(-cos(alpha*%pi/2))ˆ(1/alpha);
wh=wc*(sqrt(1+(cos(alpha*%pi/2)ˆ2))...
+cos(alpha*%pi/2))ˆ(1/alpha);
Hwh=(alpha*%pi/2)...
-atan((sin((alpha*%pi)/2)))/(((2*cos(alpha*%pi/2)))...
+(sqrt(1+(cos(alpha*%pi/2)ˆ2))));
endfunction

4.6.3 Fractional-order Band-Pass Filter

The fractional-order band-pass filter for 0 < α, β < 2 is given in Eq. (4.32). From
the equation, it should be noted that the response of fractional-order band-pass filter
can only be obtained if and only if α > β. Thus, the magnitude and phase responses
of the filter are given in Eq. (4.33).
ωcα s β
F(s) = , 0 < α, β < 2 (4.32)
s α + ωcα
 ω β
ωc
|F(s)|s= jω =   ω α  απ (4.33)
ω 2α
ωc
+2 ωc
cos 2
+1
152 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers
 ω α 
βπ −1 ωc
sin απ 2
∠F(s)|s= jω = − tan  ω α  απ (4.34)
2 ω
cosc 2
+1

Similar to Eqs. (4.21) and (4.28), the peak frequency of the fractional-order band-
pass filter ω p is defined as

 cos( απ )(2β − α) + α 2 + 4β(α − β) tan2 ( απ
 1
) α
2 2
ω p = ωc (4.35)
2(α − β)

Therefore, the function with syntax [H,wp]=foBPF(alpha,bita,wc,f,

N,w) representing the filter using Scilab is given as
function [H,wp]=foBPF(alpha,bita,wc,f,N,w)
s=poly(0,’s’); s=syslin(’c’,s/((0*s)+1));
exec(’fod.sci’);
alphaTemp=fix(alpha);
alp=alpha-alphaTemp;
bitaTemp=fix(bita);
bi=bita-bitaTemp;
num=(wcˆalpha)*((sˆbitaTemp)*fod(bi,f,N,w));
den=((sˆalphaTemp)*fod(alp,f,N,w))+(wcˆalpha);
H=num/den;
wptemp=sqrt(alphaˆ2+4*bita*(alpha-bita)...
*tan(alpha*%pi/2)ˆ2);
wp=wc*((cos(alpha*%pi/2)*((2*bita-alpha)+(wptemp)))...
/(2*(alpha-bita)))ˆ(1/alpha);
endfunction

4.6.4 Fractional-order All-Pass Filter

The fractional-order all-pass filter is defined as

s α − ωcα
F(s) = , 0<α<2 (4.36)
s α + ωcα

Furthermore, the magnitude and phase responses of the filter are obtained as
follows:   α 
 ω 2α
 ω − 2 ωωc cos απ +1

|F(s)|s= jω =  2α
2
 
c
α
(4.37)
ω
ωc
+ 2 ωωc cos απ 2
+1
 ω α   ω α 
−1 ω
sin απ
2 −1 ωc
sin απ 2
∠F(s)|s= jω = − tan  ω αc  − tan  ω α  απ (4.38)
ωc
cos απ
2
+ 1 ωc
cos 2
−1
4.6 Fractional-order Filters 153

Therefore, the function with syntax H=foAPF(alpha,wc,f,N,w) represent-

ing the filter using Scilab is given as
function H=foAPF(alpha,wc,f,N,w)
s=poly(0,’s’); s=syslin(’c’,s/((0*s)+1));
exec(’fod.sci’);
alphaTemp=fix(alpha);
alpha=alpha-alphaTemp;
num=((sˆalphaTemp)*fod(alpha,f,N,w))-(wcˆalpha);
den=((sˆalphaTemp)*fod(alpha,f,N,w))+(wcˆalpha);
H=num/den;
endfunction

4.7 Fractional-order PID Controllers

In this section, the Scilab based design and implementation of various forms of
fractional-order PI and PID based controllers with the use of approximation tech-
niques given in Sect. 4.3 are presented.

4.7.1 PID Controller

The transfer function of conventional PID controller (C(s)) is given by:

Ki
C(s) P I D = K p + + Kd s (4.39)
s
where, K p , K i and K d are the gains of proportional, integral and derivative actions.
The controller can be written in with syntax C=pid(Kp,Ki,Kd) as:
function C=pid(Kp,Ki,Kd)
s=poly(0,’s’);
s=syslin(’c’,s/((0*s)+1));
C=Kp+(Ki/s)+(Kd*s);
endfunction

Over the years, many forms of fractional-order PID based controllers have been
proposed by researchers as given in the following subsections.

4.7.2 I. Podlubny’s Fractional-order PID Controller

The fractional-order PID (FOPID or PIλ Dμ ) according to I. Podlubny’s definition

is obtained from Eq. (4.39) by fractional-ordering the integral action with λ and
derivative actions with μ as follows [5]:
154 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

Ki
C(s) F O P I D = K p + + K d s μ , 0 ≤ λ, μ ≤ 2 (4.40)
sλ
where λ is order of integration and μ is the order of differentiation. In the above
equation, s μ and 1/s λ are approximated using Oustaloup approximation as given in
Sect. 4.3.1. Thus, the controller can be written in Scilab with syntax C=fpid(Kp,
Ki,Kd,fop,f,N,w) as follows:
function C=fpid(Kp,Ki,Kd,fop,f,N,w)
s=poly(0,’s’); s=syslin(’c’,s/((0*s)+1));
exec(’fod.sci’);
lam=fop(1); mu=fop(2);
lambdaTemp=fix(lam);lam=lam-lambdaTemp;
muTemp=fix(mu);mu=mu-muTemp;
C=Kp+(Ki/(sˆlambdaTemp*fod(lam,f,N,w)))...
+(Kd*sˆmuTemp*fod(mu,f,N,w));
endfunction

From Eq. (4.40), the control transfer functions of FOPI (PIλ ) and FOPD (PDμ )
controllers are given as

Ki
C(s) F O P I = K p + ; C(s) F O P D = K p + K d s μ ; (4.41)
sλ
Thus, the controllers can be written with the use of fpid.sci function as fol-
lows:
function C=fpi(Kp,Ki,fop,f,N,w)
s=poly(0,’s’); s=syslin(’c’,s/((0*s)+1));
exec(’fpid.sci’);
C=fpid(Kp,Ki,0,[fop 0],f,N,w);
endfunction
function C=fpd(Kp,Kd,fop,f,N,w)
s=poly(0,’s’); s=syslin(’c’,s/((0*s)+1));
exec(’fpid.sci’);
C=fpid(Kp,0,Kd,[0 fop],f,N,w);
endfunction

4.7.3 M. Tenoutit’s Fractional-order PID Controller

The M. Tenoutit’s definition of fractional-order PID i.e., (PID)n is defined as [42]:


1 Ki
C(s)(P I D)n = K p + + K d s , 0≤n≤1 (4.42)
sn s

where n is the fractional-order parameter.

4.7 Fractional-order PID Controllers 155

Here, 1/s n will be approximated using Oustaloup approximation. Thus, the con-
troller can be written with syntax C=fpidTen(Kp,Ki,Kd,f,n,N,w) as:
function C=fpidTen(Kp,Ki,Kd,f,n,N,w)
s=poly(0,’s’); s=syslin(’c’,s/((0*s)+1));
exec(’fod.sci’);
nTemp=fix(n);n=n-nTemp;
C=(1/(sˆnTemp*fod(n,f,N,w)))*(Kp+(Ki/s)+(Kd*s));
endfunction

Similarly, M. Tenoutit’s fractional-order PI controller (PI)n and fractional-order

PD controller (PD)n for fractional-order parameter n are defined as:

1 Ki
C(s)(P I )n = n Kp + , 0≤n≤1 (4.43)
s s
1
C(s)(P D)n = n
K p + Kd s , 0 ≤ n ≤ 1 (4.44)
s
The controllers can be implemented with the use of above function by substituting
K d = 0 for (PI)n controller as C=fpidTen(Kp,Ki,0,f,n,N,w) and K i = 0 for
(PD)n controller as C=fpidTen(Kp,0,Kd,f,n,N,w).

4.7.4 Luo’s Fractional-order PI Controller

The Luo’s definitions of FO[PI] (or [PI]λ ) and FO[PD] (or [PD]μ ) controllers are
given in Eqs. (4.45) and (4.46) respectively [43].
 λ
Ki
C(s) F O[P I ] = Kp + , 0≤λ≤2 (4.45)
s
 μ
C(s) F O[P D] = K p + K d s , 0 ≤ μ ≤ 2 (4.46)

Equation (4.45) can be rewritten as follows:

 
Ki λ
C(s) F O[P I ] = K λp 1 + (4.47)
K ps

The 3rd order Binomial expansion of Eq. (4.47) is given by:

 
λK i λ(λ − 1) K i 2
C(s) F O[P I ] = K λp 1 + + +
K ps 2! K ps
  (4.48)
λ(λ − 1)(λ − 2) K i 3
3! K ps
156 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

Thus, the controller can be written with syntax C=fpiLuo(Kp,Ki,lam) as

function C=fpiLuo(Kp,Ki,lam)
s=poly(0,’s’); s=syslin(’c’,s/((0*s)+1));
C1=(lam*Ki)/(Kp*s);
C2=((lam*(lam-1))/2)*(Ki/(Kp*s))ˆ2;
C3=((lam*(lam-1)*(lam-2))/6)*(Ki/(Kp*s))ˆ3;
C=(Kpˆlam)*(1+C1+C2+C3);
endfunction

Similarly, Eq. (4.46) can be rewritten as follows:

 
Kd s μ
C(s) F O[P D] = K μp 1 + (4.49)
Kp

Therefore, the 3rd order Binomial expansion of Eq. (4.49) can be derived as:
 
μK d s μ(μ − 1) K d s 2
C(s) F O[P D] = K μp 1 + +
Kp 2! Kp
  (4.50)
μ(μ − 1)(μ − 2) K d s 3
+
3! Kp

Thus, the controller function can be written with syntax

C=fopdLuo(Kp,Kd,mu) as follows:
function C=fopdLuo(Kp,Kd,mu)
s=poly(0,’s’); s=syslin(’c’,s/((0*s)+1));
C1=(mu*Kd*s)/(Kp);
C2=((mu*(mu-1))/2)*((Kd*s)/Kp)ˆ2;
C3=((mu*(mu-1)*(mu-2))/6)*((Kd*s)/Kp)ˆ3;
C=(Kpˆmu)*(1+C1+C2+C3);
endfunction

4.7.5 IMC Based Fractional-order PID Controller

The fractional-order PID (FOPID) based on internal model controller (IMC-FOPID)

proposed by Tavakoli et al. is represented as [44]

Ki Kd sα
C(s) I MC−F O P I D = K p + α
+ α , 0≤α≤1 (4.51)
s γs + 1

where γ s α + 1 is added to make proper transfer function. Thus, the controller can
be written with syntax C=IMCfpid(Kp,Ki,Kd,gam,alpha,f,N,w) as fol-
lowing function. Here, s α will be approximated using Oustaloup approximation.
4.7 Fractional-order PID Controllers 157

function C=IMCfpid(Kp,Ki,Kd,gam,alpha,f,N,w)
s=poly(0,’s’); s=syslin(’c’,s/((0*s)+1));
exec(’fod.sci’);
alphaTemp=fix(alpha);alpha=alpha-alphaTemp;
Ctemp=(sˆalphaTemp)*fod(alpha,f,N,w);
C=Kp+(Ki/(Ctemp))+((Kd*Ctemp)/((gam*Ctemp)+1));
endfunction

4.7.6 Modified Fractional-order PI Controller

A modified form of FOPI controller (Mod FPI) proposed by V. Feliu-Batlle et al. is

represented as follows [45, 46]:

1 + Td s
C(s) Mod F P I = K p , 0≤α≤1 (4.52)
sα
where Td is the derivative time. Similar to the previous cases, the fractional-order
parameter s α will be approximated using Oustaloup technique. Thus, the controller
function with syntax C=modFPI(Kp,Td,alpha,f,N,w) defining the controller
is as follows:
function C=modFPI(Kp,Td,alpha,f,N,w)
s=poly(0,’s’); s=syslin(’c’,s/((0*s)+1));
exec(’fod.sci’);
alphaTemp=fix(alpha);alpha=alpha-alphaTemp;
Ctemp=(sˆalphaTemp)*fod(alpha,f,N,w);
C=Kp*((1+(Td*s))/Ctemp);
endfunction

To improve the steady state performance, the further modified form of FOPI
controller is as follows:
1 + Td s s + η
C(s) Mod F P I SS = K p , 0≤α≤1 (4.53)
sα s
where η is the minuscule which improves the steady-state behavior. The Scilab func-
tion defining the controller with syntax C=modFPIss(Kp,Td,alpha,eta,f,
N,w) is as follows:
function C=modFPIss(Kp,Td,alpha,eta,f,N,w)
s=poly(0,’s’); s=syslin(’c’,s/((0*s)+1));
exec(’fod.sci’);
alphaTemp=fix(alpha);alpha=alpha-alphaTemp;
Ctemp=(sˆalphaTemp)*fod(alpha,f,N,w);
C=Kp*((1+(Td*s))/Ctemp)*((s+eta)/s);
endfunction
158 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

4.7.7 Modified Fractional-order PID Controller

Based on the reported works by R. El-Khazali, the modified FOPID (Mod FPID)
with assumptions λ = μ and Ti = Td can be written as follows [47]:

(1 + Ti s μ )2
C(s) Mod F P I D = K c , 0≤μ≤1 (4.54)
sμ
Thus, the Scilab function with syntax C=modFPID(Kc,Ti,mu,f,N,w) defin-
ing the controller is as follows:
function C=modFPID(Kc,Ti,mu,f,N,w)
s=poly(0,’s’); s=syslin(’c’,s/((0*s)+1));
exec(’fod.sci’);
muTemp=fix(mu);mu=mu-muTemp;
Ctemp=(sˆmuTemp)*fod(mu,f,N,w);
C=Kc*(((1+(Ti*Ctemp))ˆ2)/Ctemp);
endfunction

where K c is the gain constant and Ti is the integral time. From both the modified
forms given in Eqs. (4.52) and (4.54), it should be noted that there are only three
controller parameters as in the case of classical PID. Thus, the summary of all the
developed fractional-order PI and PID controllers including their Scilab functions is
given in Table 4.3.

Table 4.3 Different variations of fractional-order PID controller

Controller Transfer function Syntax
Ki
PI Kp + s C = pid(Kp,Ki,0)
PD K p + Kd s C = pid(Kp,0,Kd)
Ki
PID Kp + s + Kd s C = pid(Kp,Ki,Kd)
PIλ Kp + Ki
sλ
C = fpid(Kp,Ki,0,fop,f,N,w)
PDμ K p + Kd sμ C = fpid(Kp,0,Kd,fop,f,N,w)
PIλ Dμ Kp + Ki
sλ
+ Kd sμ C = fpid(Kp,Ki,Kd,fop,f,N,w)
 Ki
(P I )n sn K p + s
1
C = fpidTen(Kp,Ki,0,f,n,N,w)
 Ki
(P I D)n sn K p + s + Kd s
1
C = fpidTen(Kp,Ki,Kd,f,n,N,w)
 λ
[PI]λ K p + Ksi C = fopiLuo(Kp,Ki,lam)
 μ
[PD]μ K p + Kd s C = fopdLuo(Kp,Kd,mu)
α
IMC-FOPID K p + Ks αi + γKs dα s+1 C = IMCfpid(Kp,Ki,Kd,gam,alpha,f,N,w)
Mod FPI K p 1+T
sα
ds
C = modFPI(Kp,Td,alpha,f,N,w)
1+Td s s+η
Mod FPI SS K p sα s C = modFPIss(Kp,Td,alpha,eta,f,N,w)
μ 2
Mod FPID K c (1+Ts μi s ) C = modFPID(Kc,Ti,mu,f,N,w)
4.8 Numerical and Stability Analysis 159

4.8 Numerical and Stability Analysis

In this section, the Scilab based implementation for numerical and stability analysis
of fractional-order systems and controllers are presented. The numerical analysis
includes both the time and frequency domain.

4.8.1 Time and Frequency Domain Analysis

The root locus plot of the transfer function model with the help of Scilab function
evans can be written as follows:
function rlocus(G,Kmax)
evans(G,Kmax);
sgrid();
endfunction

where K max is the maximum desired gain.

Similarly, the impulse and step responses of the transfer function model for a time
period t can be calculated with the extension of Scilab function csim as follows:
function impulse(G,t)
Gres=csim(’impulse’,t,G);
plot(t,Gres);
f=gcf();
xlabel("Time (sec)");
ylabel("Amplitude");
title("Impulse Response");
endfunction

function step(G,t)
Gres=csim(’step’,t,G);
f=gcf();
plot(t,Gres);
xlabel("Time (sec)");
ylabel("Amplitude");
title("Step Response");
endfunction

The stepinfo function with syntax sInfo=stepinfo(G,t,yf) is used

to calculate the overshoot (%), undershoot (%), peak, peak time (s), rise time (s) and
settling time (s) of the step response is given as follows:
function sinfo=stepinfo(G,t,yf)
y=csim(’step’,t,G); exec(’isfinite.sci’);
if y==[] then
160 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

disp(’Empty Data’)
else
TF=isfinite(y);
if TF==1 then ns=length(y);
[sinfo.Peak,ipeak]=max(abs(y));
sinfo.PeakTime=t(ipeak);
if yf==0 then
sinfo.Overshoot=%inf;
if and(y>=0) then
sinfo.Undershoot=0;
else
sinfo.Undershoot=%inf;
end
else
yrel=y/yf; sinfo.Overshoot=100*max(0,max(yrel-1));
sinfo.Undershoot=-100*min(0,min(yrel));
end
tolBand=0.02; err=abs(y-yf);
tol=tolBand*max(err); nn=length(err);
nn1=nn;
for i=1:nn
err1(nn1)=err(i); nn1=nn1-1;
end
err1=err1’;
isettle=find(err1>tol,1); nny=length(y);
nny1=nny;
for i=1:nny
y1(nny1)=y(i); nny1=nny1-1;
end
y1=y1’;
t1=t(length(t)):-(t(2)-t(1)):t(1);
if isettle==[] then
sinfo.SettlingTime=0;
elseif isettle==ns
sinfo.SettlingTime=%nan;
elseif y1(isettle)˜=y1(isettle+1)
y1Settle=yf+sign(y1(isettle)-yf)*tol;
sinfo.SettlingTime = t1(isettle) +...
(t1(isettle)-t1(isettle+1))/...
(y1(isettle)-y1(isettle+1))...
*(y1Settle-y1(isettle));
else
sinfo.SettlingTime=t1(isettle+1);
end
yLow=y(1)+0.1*(yf-y(1));
4.8 Numerical and Stability Analysis 161

iLow=1+find((y(1:ns-1)-yLow).*(y(2:ns)-yLow)<=0,1);
if iLow==[] then
tLow=%nan;
elseif iLow>1 && y(iLow)˜=y(iLow-1)
tLow=t(iLow)+(t(iLow)-t(iLow-1))/...
(y(iLow)-y(iLow-1))*(yLow-y(iLow));
else
tLow=t(iLow);
end
yHigh=y(1) + 0.9*(yf-y(1));
iHigh=1+find((y(1:ns-1)-yHigh).*(y(2:ns)-yHigh)<=0,1);
if iHigh==[] then
tHigh=%nan;
sinfo.SettlingMin=%nan; sinfo.SettlingMax=%nan;
else
if iHigh>1 && y(iHigh)˜=y(iHigh-1)
tHigh=t(iHigh)+(t(iHigh)-t(iHigh-1))/...
(y(iHigh)-y(iHigh-1))*(yHigh-y(iHigh));
else
tHigh=t(iHigh);
end
yRisen=y(iHigh:length(y));
sinfo.SettlingMin=min(yRisen);
sinfo.SettlingMax=max(yRisen);
end
sinfo.RiseTime=tHigh-tLow;
sinfo.SettlingMin=min(yRisen);
sinfo.SettlingMax=max(yRisen);
else
sinfo.RiseTime=%nan;
sinfo.SettlingTime=%nan;
sinfo.SettlingMin=%nan;
sinfo.SettlingMax=%nan;
sinfo.Overshoot=%nan; sinfo.Undershoot=%nan;
sinfo.Peak=%inf; sinfo.PeakTime=%inf;
end
end
endfunction

Here, G is the closed-loop transfer function, t is the time period and y f is the
final steady-state value. In the function, by default the settling time threshold limit is
chosen as 2%. Thus, values other than this can also be given by changing the variable
tolBand.
162 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

On the other hand, the frequency domain plots such as Bode, Nyquist, and Nichols
are available in Scilab with command bode, nyquist and black respectively.
Furthermore, the gain margin, phase margin and crossover frequencies from the Bode
plot of the response is calculated using the following function:
function [Gm,Pm,Wgm,Wpm] = margins(G)
[Pm,Wpm] = p_margin(G);
[Gm,Wgm] = g_margin(G);
if Wpm ˜= [] then
Wpm = Wpm*2*%pi;
end
if Wgm ˜= [] then
Wgm = Wgm*2*%pi;
end
clf();
show_margins(G,’bode’)
endfunction

where p_margin and g_margin are Scilab functions to calculate phase and gain
margins respectively.

4.8.2 Stability Analysis

The stability of the approximated transfer function can also be deduced by using
the following function with syntax S=stability(n,np,d,dp,f,N,w). The
function will also provide the list of stable and unstable poles of the system.
function S=stability(n,np,d,dp,f,N,w)
exec(’fotf.sci’);
G=fotf(n,np,d,dp,f,N,w);
G=syslin(’c’,G);
z=roots(G.num);
p=roots(G.den);
pr=real(p);
nu=0; ns=0; up=0; sp=0; temp=0;
for i=1:length(p)
if pr(i)>0 then
nu = nu+1; up(nu)=p(i);
elseif pr(i)<0
ns = ns+1; sp(ns)=p(i);
temp=1;
else
disp(’The system has no poles’)
end
4.8 Numerical and Stability Analysis 163

Table 4.4 List of functions developed in Scilab based toolbox for fractional-order systems and
controllers
Function Syntax
Grünwald-Letnikov df = grunwald(f,t,alpha,h)
Oustaloup G = oustaloup(alpha,N,wl,wh)
Refined Oustaloup G = refOustaloup(alpha,N,wl,wh)
Matsuda G = matsuda(alpha,N,wl,wh)
Curve Fitting G = curFit1(alpha,N,wl,wh)
Transfer Function Estimation G = curFit2(alpha,N,wl,wh)
Finite Impulse Response G = firPSE(alpha,N,T)
Infinite Impulse Response G = iirPSE(alpha,N,T)
Differentiator G = fod(alpha,f,N,w)
Integrator G = foi(alpha,f,N,w)
Transfer Function s = tf(‘s’)
Fractional-order Transfer Function G = fotf(n,np,d,dp,f,N,w)
Transfer Function Data [n,d] = tfdata(G)
Zeros and Poles [z, p] = zeros_poles(G)
PI C = pid(Kp,Ki,0)
PD C = pid(Kp,0,Kd)
PID C = pid(Kp,Ki,Kd)
PIλ C = fpid(Kp,Ki,fop,f,N,w)
PDμ C = fpid(Kp,Kd,fop,f,N,w)
PIλ Dμ C = fpid(Kp,Ki,Kd,fop,f,N,w)
(PI)n C = fpidTen(Kp,Ki,0,f,n,N,w)
(PD)n C = fpidTen(Kp,0,Kd,f,n,N,w)
(PID)n C = fpidTen(Kp,Ki,Kd,f,n,N,w)
[PI]λ C = fpiLuo(Kp,Ki,lam)
[PD]μ C = fpdLuo(Kp,Kd,mu)
IMC-FOPID C = IMCfpid(Kp,Ki,Kd,gam,alpha,f,N,w)
Mod FPI C = modFPI(Kp,Td,alpha,f,N,w)
Mod FPI SS C = modFPIss(Kp,Td,alpha,eta,f,N,w)
Mod FPID C = modFPID(Kc,Ti,mu,f,N,w)
Low-pass Filter [H,wp,wrp,wh,Hwh] = foLPF(alpha,wc,f,N,w)
High-pass Filter [H,wp,wrp,wh,Hwh] = foHPF(alpha,wc,f,N,w)
Band-pass Filter [H,wp] = foBPF(alpha,bita,wc,f,N,w)
All-pass Filter H = foAPF(alpha,wc,f,N,w)
Pole Zero Plot plzr(G)
Root locus rlocus(G,Kmax)
Step step(G,t)
(continued)
164 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

Table 4.4 (continued)

Function Syntax
Step Info sInfo=stepinfo(G,t,yf)
Impulse impulse(G,t)
Nyquist nyquist(G)
Bode bode(G)
Magnitude Plot gainplot(G)
Phase Plot phaseplot(G)
Asymptotic Bode Plot bode_asymp(G)
Peak Frequencies fr = freson(G)
Margins [Gm,Pm,Wgm,Wpm] = margins(G)
Nichols black(G), nicholschart()
Singular-value Sigma-plot svplot(G)
Stability S = stability(n,np,d,dp,f,N,w)
H2 -norm h2norm(G)
H∞ -norm h_norm(G)
L∞ -norm linf(G)

end
disp(p,’Pole(s) of the system:’)
if length(sp)>0 & temp==1 then
disp([sp],’Stable pole(s):’)
end
if length(up)>0 & up˜=0 then
disp([up],’Unstable pole(s):’)
end
if nu>0 then
disp(’Unstable System’); S=0;
else
disp(’Stable System’); S=1;
end
endfunction

The summary of all the developed functions in this toolbox including their Scilab
syntax is given in Table 4.4. Furthermore, the organization chart of the complete
toolbox is shown in Fig. 4.1.
 FOS&C Toolbox Using Scilab

Fractional-order Operators Analysis of Fractional-order Systems

Approximation Fractional-order
Definitions Techniques Parameters Time-domain Frequency-domain Stability

Grünwald-Letnikov Integrals
Derivatives Root H2 -norm,
Step Impulse Nyquist Bode Nichols
locus
4.8 Numerical and Stability Analysis

H∞ -norm,
L∞ -norm,

Transfer Finite Infinite Step info Margins

Refined Curve
Oustaloup Matsuda Function Impulse Impulse
Oustaloup Fitting
Estimation Response Response

Fractional-order Systems and Controllers

Fractional-order Systems Fractional-order PID Controllers

Fractional-order
Fractional-order Filters PIλ , (PI)n , Mod FPI,
Transfer Functions [PI]λ ,
IMC–FOPID PDμ , (PD)n , mod FPIss,
[PD]μ
PIλ Dμ (PID)n mod FPID

High-pass Low-pass Band-pass All-pass

Filter Filter Filter Filter
165

Fig. 4.1 Organization chart of the complete Scilab based toolbox for fractional-order systems and controllers
166 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

4.9 Simulation Study

In this section, a simulation study is conducted on several fractional-order based

operators, PID controller, transfer function models, and filters to evaluate the perfor-
mance of the developed Scilab based toolbox.

4.9.1 Fractional-order Operators and Transfer Functions

4.9.1.1 Grünwald-Letnikov

Consider the function f (t) = e−4t cos(2t + 1). The fractional-order derivative of
f (t) for orders α = 0.6, 0.7 and 0.8 for the time period t = 0 − π using Grünwald-
Letnikov’s definition is obtained with the following commands:
– –> t=0:0.01:%pi; alpha=[0.6 0.7 0.8]; h=0.001;
exec(’grunwald.sci’); f=exp(-4*t).*cos(2*t+1);
for i=1:length(alpha)
df=grunwald(f,t,alpha(i),h);
plot(t,df’)
end
xlabel(’Time (sec)’); ylabel(’Amplitude’);
legend([’$\alpha=0.6$’,’$\alpha=0.7$’,’$\alpha=0.8$’]);
Thus, the resultant plot from the above commands is shown in Fig. 4.2. Further-
more, for the same function, the fractional-order derivatives for different orders of
18

16

14

12

10

8
Amplitude

−2

−4

−6

−8
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3
Time (sec)

Fig. 4.2 Fractional-order derivatives of f (t) for orders α = 0.6, 0.7 and 0.8
4.9 Simulation Study 167

20

10

−10

−20

−30

−40
−1
3.5 3 −0.5
2.5 2 1.5 0
1 0.5 0.5
1
0
t

Fig. 4.3 Fractional-order derivatives of f (t) for different orders of α

α ∈ (−1, 1) for the time period of t = 0 − π shown in Fig. 4.3 can be obtained using
the following commands:
– –> t=0:0.01:%pi; alpha=[-1:0.1:1]; h=0.001; F=[];
exec(’grunwald.sci’); f=exp(-4*t).*cos(2*t+1);
for i=alpha
F=[F grunwald(f,t,i,h)];
end
surf(t,alpha,F’); xlabel(’t’);ylabel(’$\alpha$’);
zlabel(’$\mathcal{D}_tˆ{\alpha}f(t)$’);

Similarly, consider the sinusoidal function f (t) = sin(t). The fractional-order

derivatives of f (t) for different orders of α ∈ (0, 1) for the time period of t = 0 − 5π
shown in Fig. 4.4 can be obtained using the following commands:
– –> t=0:0.01:5*%pi; alpha=[0:0.1:1]; h=0.01; F=[];
exec(’grunwald.sci’); f=sin(t);
for i=alpha
F=[F grunwald(f,t,i,h)];
end
surf(t,alpha,F’); xlabel(’t’); ylabel(’$\alpha$’);
zlabel(’$\mathcal{D}_tˆ{\alpha}f(t)$’);

4.9.1.2 Approximation Techniques

Consider the fractional-order differentiator of order α = 0.8, the approximated trans-

fer function for order of approximation N = 5 and the desired frequency range
168 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

Fig. 4.4 Fractional-order derivatives of sinusoidal function for different orders of α

ω = (10−5 , 105 ) can be obtained using various approximation techniques presented

in Sect. 4.3 as follows:
– –> N=5; wl=10ˆ-5; wh=10ˆ5; alp=0.8;
exec(’oustaloup.sci’); G1=oustaloup(alp,N,wl,wh);
exec(’refOustaloup.sci’); G2=refOustaloup(alp,N,wl,wh);
exec(’matsuda.sci’); G3=matsuda(alp,2*N,wl,wh);
exec(’curFit1.sci’); G4=curFit1(alp,N,wl,wh);
exec(’curFit2.sci’); G5=curFit2(alp,N,wl,wh);
bode([G1;G2;G3;G4;G5],wl,wh,"rad")
legend(’Oustaloup’,’Refined Oustaloup’,’Matsuda’,...
’Curve Fitting1’,’Curve Fitting2’);
scf();
exec(’firPSE.sci’); G6=firPSE(alp,20*N,0.01);
exec(’iirPSE.sci’); G7=iirPSE(alp,20*N,0.01);
bode([G6;G7],"rad"); legend(’FIR-PSE’,’IIR-PSE’);
Thus, the approximated transfer function of s 0.6 using Oustaloup, refined
Oustaloup, Matsuda and both the curve fitting techniques from the above commands
are as follows:

1 + 63733.065s + 40216867s 2 + 2.538D + 08s 3

+ 16009022s 4
+ 10000s 5
G 1 (s) ≈  (4.55)
10000 + 16009022 s + 2.538D + 08s 2 + 40216867s 3

+ 63733.065s 4 + s 5
4.9 Simulation Study 169

8.093D − 66s + 5.158D − 49s 2 + 3.255D − 34s 3
+ 2.054D − 21s 4 + 1.296D − 10s 5 + 0.081758s 6
+ 515858.01s 7 + 3.255D + 10s 8 + 2.054D + 13s 9
+ 1.296D + 14s 10 + 8.176D + 12s 11 + 5.180D + 09s 12

 + 45958.306s 13
G 2 (s) ≈
2.524D − 57 + 4.040D − 42s + 6.404D − 29s 2 + 1.015D − 17s 3
+ 1.609D − 08s 4 + 0.2571923s 5 + 75818.957s 6
+ 5.676D + 09s 7 + 8.896D + 12s 8 + 1.410D + 14s 9
+ 2.234D + 13s 10 + 3.545D + 10s 11 + 619288.62s 12

+ s 13
 (4.56)
1 + 189966.01s + 1.292D + 08s 2 + 8.203D + 08s 3

+ 51196362s 4 + 28613.773s 5
G 3 (s) ≈  (4.57)
28613.773 + 51196362 s + 8.203D + 08s 2 + 1.292D + 08s 3

+ 189966.01s 4 + s 5

5209.5414 + 5284093.2s + 74629531s 2 + 77813607s 3

+ 15008004s 4 + 435054.73s 5
G 4 (s) ≈  (4.58)
752376.93 + 57692362 s + 88806404s 2 + 23776455s 3

+ 1074614.1s 4 + s 5

− 5.062D − 09 − 0.0000041s − 1.288D − 09s 2 − 2.317D − 14s 3

+ 8.300D − 20s 4 + 3.638D − 25s 5
G 5 (s) ≈ 
− 0.0000136 − 8.206D − 09s − 2.598D − 13s 2 + 4.312D − 19s 3

+ 5.758D − 24s 4 + 1.112D − 30s 5
(4.59)
Furthermore, the resultant bode plot of the approximated transfer function is
shown in Fig. 4.5. On the other hand, the resultant Bode plot using the power series
expansion techniques is shown in Fig. 4.6.
The above approximation techniques can also be implemented with the use of
fod.sci function as follows:
– –> N=5; wl=10ˆ-5; wh=10ˆ5; alp=0.8;
exec(’fod.sci’); w=[wl wh];
G1=fod(alp,’o’,N,w); G2=fod(alp,’r’,N,w);
G3=fod(alp,’m’,N,w); G4=fod(alp,’cf1’,N,w);
G5=fod(alp,’cf2’,N,w);
bode([G1;G2;G3;G4;G5],wl,wh,"rad")
legend(’Oustaloup’,’Refined Oustaloup’,’Matsuda’,...
’Curve Fitting1’,’Curve Fitting2’);
170 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

100 Oustaloup
Refined Oustaloup
Matsuda
Magnitude (dB)

50
Curve Fitting1
Curve Fitting2
0

−50

−6 −5 −4 −3 −2 −1 0 1 2 3 4 5 6
10 10 10 10 10 10 10 10 10 10 10 10 10
Frequency (rad/s)

80
Phase (degree)

60

40

20

0
−6 −5 −4 −3 −2 −1 0 1 2 3 4 5 6
10 10 10 10 10 10 10 10 10 10 10 10 10
Frequency (rad/s)

Fig. 4.5 Bode plot of fractional-order differentiator using various techniques

The results from the above commands are same as the results given in Eqs. (4.55),
(4.56), (4.57), (4.58), (4.59), and Fig. 4.5. Similarly, the approximation of fractional-
order integrator for order α = 0.75 using Oustaloup, refined Oustaloup, Matsuda
and both the curve fitting techniques can be done with the use of foi.sci function
as follows:
– –> N=5; wl=10ˆ-5; wh=10ˆ5; alp=0.75;
exec(’foi.sci’); w=[wl wh];
G1=foi(alp,’o’,N,w); G2=foi(alp,’r’,N,w);
G3=foi(alp,’m’,N,w); G4=foi(alp,’cf1’,N,w);
G5=foi(alp,’cf2’,N,w);
bode([G1;G2;G3;G4;G5],wl,wh,"rad")
legend(’Oustaloup’,’Refined Oustaloup’,’Matsuda’,...
’Curve Fitting1’,’Curve Fitting2’);
From the above commands, it can be seen that the order of approximation and
the desired frequency range are chosen as N = 5 and ω = (10−5 , 105 ). Thus, the
approximated transfer functions obtained from the above commands using various
techniques are as follows:
4.9 Simulation Study 171

80
FIR−PSE
60 IIR−PSE
Magnitude (dB)

40

20

−2 −1 0
10 10 10
Frequency (rad/s)

80

60
Phase (degree)

40

20

0
−2 −1 0
10 10 10
Frequency (rad/s)

Fig. 4.6 Bode plot of fractional-order differentiator using FIR and IIR


5623.4133 + 10101010 s + 1.796D + 08s 2 + 31945393s 3

 + 56802.154s 4 + s 5
G 1 (s) ≈ (4.60)
1 + 56802.154s + 31945393s 2 + 1.796D + 08s 3

+ 10101010s 4 + 5623.4133s 5

5.335D − 58 + 9.583D − 43s + 1.704D − 29s 2 + 3.031D − 18s 3
+ 5.390D − 09s 4 + 0.0966347s 5 + 31240.335s 6
+ 2.555D + 09s 7 + 4.490D + 12s 8 + 7.984D + 13s 9

 + 1.420D + 13s 10 + 2.528D + 10s 11 + 501246.6s 12 + s 13
G 2 (s) ≈
1.299D − 65s + 7.377D − 49s 2 + 4.149D − 34s 3 + 2.333D − 21s 4
+ 1.312D − 10s 5 + 0.0737748s 6 + 414866.19s 7
+ 2.333D + 10s 8 + 1.312D + 13s 9 + 7.377D + 13s 10

+ 4.149D + 12s 11 + 2.347D + 09s 12 + 20784.61s 13
(4.61)
172 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

13429.716 + 27013946 s + 4.858D + 08s 2

+ 85914451s 3 + 142763.38s 4 + s 5
G 3 (s) ≈  (4.62)
1 + 142763.38s + 85914451s 2 + 4.858D + 08s 3

+ 27013946s 4 + 13429.716s 5

313626.66 + 32533731 s + 52908176s 2 + 14724923s 3

+ 695318.33s 4 + s 5
G 4 (s) ≈  (4.63)
3522.6379 + 3567540.8s + 44947701s 2 + 44843529s 3

+ 8272345.1s 4 + 223169.53s 5

− 0.0000011 − 1.422D − 09s − 7.391D − 14s 2 + 6.844D − 19s 3

+ 6.229D − 24s 4
+ 1.703D − 30s 5
G 5 (s) ≈ 
− 6.124D − 09 − 0.0000003 s − 1.619D − 10s 2 − 3.550D − 15s 3

+ 5.439D − 20s 4 + 2.016D − 25s 5
(4.64)
Therefore, the resultant Bode plot of fractional-order integrator using various
techniques obtained from the above commands is shown in Fig. 4.7.
Furthermore, the approximation of the transfer functions can be done using
fotf.sci function. To illustrate this, consider the fractional-order transfer function
given as follows:
s+1
G(s) = (4.65)
10s 3.2 + 185s 2.5 + 288s 0.7 + 1

The approximated transfer function of G(s) for the order of approximation

N = 5 and the desired frequency range ω = (10−5 , 105 ) using Oustaloup, refined
Oustaloup, and both the curve fitting techniques can be obtained using the following
commands:
– –> N=5; wl=10ˆ-5; wh=10ˆ5;
exec(’fotf.sci’); n=[1 1]; np=[1 0];
d=[10 185 288 1]; dp=[3.2 2.5 0.7 0];
Go=fotf(n,np,d,dp,’o’,N,[wl wh]);
Gr=fotf(n,np,d,dp,’r’,N,[wl wh]);
Gc1=fotf(n,np,d,dp,’cf1’,N,[wl wh]);
Gc2=fotf(n,np,d,dp,’cf2’,N,[wl wh]);
bode([Go;Gr;Gc1;Gc2],wl,wh,"rad")
legend(’Oustaloup’,’Refined Oustaloup’,...
’Curve Fitting1’,’Curve Fitting2’);
In the above commands, Go, Gr , Gc1 and Gc2 represents the approximated trans-
fer functions of Oustaloup, refined Oustaloup, and both the curve fitting techniques.
Thus, the resultant Bode plot from the above commands in comparison with these
techniques is shown in Fig. 4.8.
4.9 Simulation Study 173

Oustaloup
50
Refined Oustaloup
Matsuda
Curve Fitting1
Magnitude (dB)

0
Curve Fitting2

−50

−100

−2 −1 0 1 2 3 4 5 6
10 10 10 10 10 10 10 10 10
Frequency (rad/s)

−20
Phase (degree)

−40

−60

−80

−2 −1 0 1 2 3 4 5 6
10 10 10 10 10 10 10 10 10
Frequency (rad/s)

Fig. 4.7 Bode plot of fractional-order integrator using various techniques

The coefficients of numerator and denominator of the Oustaloup based approx-

imated transfer function Go(s) can be obtained as exec(’tfdata.sci’);
[n,d]=tfdata(Go);. On the other hand, the zeros and poles of the Go(s) can be
obtained as exec( zeros_poles.sci ); [z, p] = zeros_poles(Go);. There-
fore, the resultants coefficients, zeros and poles of the Go(s) are given in Table 4.5.

4.9.1.3 Time and Frequency-Domain Plots

To illustrate all time and frequency-domain plots, consists the fractional-order trans-
fer function as follows:

5s 0.6 + 2
G(s) = (4.66)
s 3.3 + 3.1s 2.6 + 2.89s 1.9 + 2.5s 1.4 + 1.2

The pole-zero plot and the root-locus plot of the Oustaloup based approximated
transfer function G(s) for the order of approximation N = 4 and the desired fre-
quency range ω = (10−4 , 104 ) can be obtained as follows:
174 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

Oustaloup
Magnitude (dB)

−100
Refined Oustaloup
Curve Fitting1

−200 Curve Fitting2

−300
−6 −5 −4 −3 −2 −1 0 1 2 3 4 5 6
10 10 10 10 10 10 10 10 10 10 10 10 10
Frequency (rad/s)

200

100
Phase (degree)

−100

−200

−6 −5 −4 −3 −2 −1 0 1 2 3 4 5 6
10 10 10 10 10 10 10 10 10 10 10 10 10
Frequency (rad/s)

Fig. 4.8 Bode plot of fractional-order transfer function using various approximation techniques

– –> clear; clc; N=4; w=[10ˆ-4 10ˆ4]; exec(’fotf.sci’);

n=[5 2]; np=[0.6 0];
d=[1 3.1 2.89 2.5 1.2]; dp=[3.3 2.6 1.9 1.4 0];
G=fotf(n,np,d,dp,’o’,N,w);
scf(); subplot(121); plzr(G);
exec(’rlocus.sci’); subplot(122); rlocus(G,100);
Therefore, the resultant root-locus plot from the above command is shown in
Fig. 4.9. Similarly, the time domain plots both impulse and step response of approxi-
mated transfer function G(s) shown in Fig. 4.10 can be obtained using the following
commands
– –> scf(); t=0:1:200;
Gri=csim(’impulse’,t,G); Grs=csim(’step’,t,G);
subplot(121); plot(t,Gri);
xlabel("Time(sec)"); ylabel("Amplitude");
title("Impulse Response");
subplot(122); plot(t,Grs);
xlabel("Time(sec)"); ylabel("Amplitude");
title("Step Response");
4.9 Simulation Study 175

Table 4.5 Coefficients of numerator and denominator and the zeros and poles of Go(s)
Coeff. of numerator Coeff. of Zeros Poles
denominator
10000000 10910736 −50118.723 −50118.723
1.158D+11 1.706D+11 −31622.777 −31955.834
4.020D+14 8.851D+14 −15848.932 −6424.1415
4.440D+17 1.628D+18 −501.18723 −501.19184
4.812D+19 6.198D+20 −316.22777 −406.64971
1.665D+21 5.053D+22 −158.48932 −91.832328
1.929D+22 1.206D+24 −5.0118723 −20.212329
3.666D+22 5.812D+24 −3.1622777 −4.8381822
2.543D+22 6.098D+24 −1.5848932 −1.550911
7.142D+21 5.236D+24 −1 −0.1349838 + 1.2346476i
7.113D+20 3.012D+24 −0.0501187 −0.1349838 − 1.2346476i
7.584D+18 5.683D+23 −0.0316228 −0.2047152
2.565D+16 2.755D+22 −0.0158489 −0.0316423
2.799D+13 3.292D+20 −0.0005012 −0.0158489
2.977D+09 1.124D+18 −0.0001585 −0.0026804
98577.194 1.250D+15 −0.0003162 −0.0001585
1 2.220D+11 −0.0001756
8952550 −0.0003162
100

Furthermore, the step response characteristics and the stability analysis of the
approximated transfer function G(s) can be obtained as follows:
– –> exec(’stepinfo.sci’); yf=1.8;
Gsinfo=stepinfo(G,t,yf);
exec(’stability.sci’); S=stability(n,np,d,dp,’o’,N,w);
h2=h2norm(G); hi=h_norm(G); Li=linf(G);
Thus, the resultant numerical analysis obtained from the above commands are
Peak: 5.0554, PeakTime: 6, Overshoot: 180.8588, Undershoot:
0, SettlingTime: 57.4028, SettlingMin: 0.5397, SettlingMax:
5.0554 and RiseTime: 1.6006. Furthermore, the stability analysis obtained
are S=1 which denotes a Stable system. The H2 -norm, H∞ -norms and L∞ -noms as
4.293D+12i, 10.3609 and 0.0029 respectively also justifies that the approximated
transfer function is stable.
On the other hand, the frequency domain plots like Bode, Nyquist and Nichols
plots shown in Figs. 4.11, 4.12 and 4.13 of the approximated transfer function G(s)
can be obtained using the following commands
176 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

Transmission zeros and poles Evans root locus

25 000 2
Poles 0.985
0.94
0.87
0.77
0.64
0.34
0.17
0.5
20 000 1.5

15 000
1

10 000
0.5

Imaginary axis
5 000
1
Imaginary axis

0 0
0
1e+04
−0.5
−5 000 2e+04

−1 3e+04
−10 000
4e+04
−1.5
−15 000
5e+04
0.985
0.94
0.87
0.77
0.64
0.34
0.17
0.5
−20 000 −2
−60 000 −40 000 −20 000 0

−25 000 Real axis

asymptotic directions
−60 000 −40 000 −20 000 0 open loop poles
Real axis open loop zeros

Fig. 4.9 Pole-zero and root-locus plots of fractional-order transfer function G(s)

– –> scf(); bode(G,"rad");

scf(); subplot(121); nyquist(G);
subplot(122); nyquist(G); hallchart();
scf(); subplot(121); black(G);
subplot(122); black(G); nicholschart();
In the above commands, the functions hallchart() and nicholschart()
are used as grids for nyquist and nichols plots respectively. Furthermore, the peak
frequencies of the magnitude plot given in Fig. 4.11 can be obtained with the use of
Scilab function fr=freson(G) as follows:
– –> scf(); gainplot(G);
fr=freson(G); fig=gce();
p1=fig.children(1);
[phi,db]=phasemag(repfreq(G,fr));
plot(fr,db(:),’ob’);
t=datatipCreate(p1,[fr db(:)]);
datatipSetOrientation(t,"upper left");
4.9 Simulation Study 177

Impulse Response Step Response

1.4 6

1.2 5.5

1 5

0.8 4.5

0.6 4

0.4 3.5
Amplitude

Amplitude
0.2 3

0 2.5

−0.2 2

−0.4 1.5

−0.6 1

−0.8 0.5

−1 0
0 50 100 150 200 0 50 100 150 200

Time(sec) Time(sec)

Fig. 4.10 Impulse and step response of fractional-order transfer function G(s)

In the above commands it can be noted that the peak frequency fr has been
plotted on magnitude plot of G(s) which is obtained using the Scilab function
gainplot(G) as shown in Fig. 4.14. Furthermore, the datatip shown in the figure
has also been created using Scilab function datatipCreate().
On the other hand, the Asymptotic Bode plot of G(s) shown in Fig. 4.15 can be
obtained as follows:
– –> scf(); bode(G,wl,wh); bode_asymp(G,wl,wh);

4.9.2 Fractional-order PID Controllers

In this subsection, a simulation study on the performance of various fractional-order

PI and PID controllers, non-linear stable spherical tank and an unstable bioreactor
processes are considered to evaluate the performance of the proposed Scilab based
fractional-order PI and PID controllers. In all the cases, the fractional-order operators
are approximated using Oustaloup approximation with parameters N = 5, ωl = 10−3
and ωh = 103 .
178 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

0
Magnitude (dB)

−50

−100

−150

−200
−3 −2 −1 0 1 2 3 4
10 10 10 10 10 10 10 10
Frequency (rad/s)

0
Phase (degree)

−100

−200

−300
−3 −2 −1 0 1 2 3 4
10 10 10 10 10 10 10 10
Frequency (rad/s)

Fig. 4.11 Bode plot of fractional-order transfer function G(s)

4.9.2.1 Analysis of Fractional-order PID Controllers

The Bode plot of I. Podlubny’s fractional-order PID controller is given in Sect. 4.7.2
approximated using Oustaloup, refined Oustaloup, Matsuda, and curve fitting tech-
niques in comparison with conventional PID controller can be obtained using the
following commands. For this analysis the gain parameters of the controller are
chosen as K p = K i = K d = 1 and the fractional-order parameters are chosen as
λ = 0.98 and μ = 0.85.
– –> exec(’pid.sci’); exec(’fpid.sci’); C1=pid(1,1,1);
C2=fpid(1,1,1,[0.98 0.85],’o’,5,[10ˆ-5 10ˆ5]);
C3=fpid(1,1,1,[0.98 0.85],’r’,5,[10ˆ-5 10ˆ5]);
C4=fpid(1,1,1,[0.98 0.85],’m’,2*5,[10ˆ-5 10ˆ5]);
C5=fpid(1,1,1,[0.98 0.85],’cf1’,5,[10ˆ-5 10ˆ5]);
bode([C1;C2;C3;C4;C5],"rad");
legend(’$PID$’,’$PIˆ\lambda Dˆ\mu - Oustaloup$’,...
’$PIˆ\lambda Dˆ\mu - RefinedOustaloup$’,...
’$PIˆ\lambda Dˆ\mu - Matsuda$’,...
’$PIˆ\lambda Dˆ\mu - Curve Fitting$’);
4.9 Simulation Study 179

Nyquist plot Nyquist plot

10 10
−0.0706 −0.0661 −0.0706−0.0661
−0.0623 −0.0623
8 −0.0766 8 −0.0766

6 6
−0.0898 −0.0898
15°
4 4

−0.0506 30° −0.0506

2 2 45°
2dB
60°
0.0428 90° −4dB 0.0428
6dB −10dB
1e+03 0.001 20dB
1e+030.001
−20dB
10dB
0 −0.0428 0 −90°−6dB
4dB −0.0428
−60° −2dB
−45°
0.0506 −30° 0.0506
−2 −2

−15°
−4 0.0898 −4 0.0898

−6 −6
0.0766 0.0766
0.0623 0.0623
−8 −8
0.0706 0.0661 0.0706 0.0661

−10 −10
−4 −2 0 2 4 6 8 10 −6 −4 −2 0 2 4 6 8 10

Fig. 4.12 Nyquist plot of fractional-order transfer function G(s)

40 40

0.077 0.055
20 0.25dB −0.25dB
0.001 30 0dB
0.2
0 0.5dB −0.5dB

1dB −1dB
−20 20 0.077
1
2.3dB −2dB
−40
10 4dB −3dB
6dB
−60
Magnitude (dB)

Magnitude (dB)

12dB
0.2 −6dB
−80 11 0
−9dB
−100 −12dB
−10
−15dB
−120
80
−20dB
−140 −20
−25dB
−160 −30dB
−30 1
6.3e+02
−180

−200 −40
−300 −200 −100 0 100 −400 −300 −200 −100 0
Phase (deg) Phase (deg)

Fig. 4.13 Nichols plot of fractional-order transfer function G(s)

180 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

0.06165Hz
20.36dB

0
Magnitude (dB)

−50

−100

−150

−200
−3 −2 −1 0 1 2 3
10 10 10 10 10 10 10

Frequency (Hz)

Fig. 4.14 Magnitude plot of G(s) showing the peak frequencies

0
Magnitude (dB)

−100

−200

−300

−6 −5 −4 −3 −2 −1 0 1 2 3 4 5 6
10 10 10 10 10 10 10 10 10 10 10 10 10
Frequency (Hz)

200
Phase (degree)

−200

−400
−6 −5 −4 −3 −2 −1 0 1 2 3 4 5 6
10 10 10 10 10 10 10 10 10 10 10 10 10
Frequency (Hz)

Fig. 4.15 Asymptotic Bode plot of fractional-order transfer function G(s)

4.9 Simulation Study 181

From the above commands, the resultant transfer functions of the controller are
as follows:
1 + s + s2
C1 (s) = (4.67)
s

1.413D + 09 + 3.511D + 12s + 2.180D + 15s 2 + 5.484D + 16s 3
+ 3.821D + 17s 4 + 4.773D + 17s 5 + 4.014D + 17s 6
+ 5.680D + 16s 7 + 2.186D + 15s 8 + 3.515D + 12s 9

+ 1.413D + 09s 10
C2 (s) ≈  (4.68)
17782.794 + 1.741D + 09s + 4.086D + 12s 2 + 2.388D + 15s 3
+ 5.535D + 16s 4 + 3.222D + 17s 5 + 7.467D + 16s 6
+ 4.345D + 15s 7 + 1.003D + 13s 8 + 5.764D + 09s 9

+ 79432.823s 10


3.75D − 110 + 9.306D − 95s + 5.770D − 80s 2 + 1.391D − 66s 3
+ 8.537D − 54s 4 + 2.057D − 42s 5 + 1.263D − 31s 6 + 3.043D − 22s 7
+ 1.869D − 13s 8 + 0.0000045s 9 + 28.727814s 10 + 13467045s 11
+ 2.443D + 12s 12 + 1.994D + 17s 13 + 6.297D + 21s 14 + 1.475D + 25s 15
+ 8.977D + 27s 16 + 2.248D + 29s 17 + 1.547D + 30s 18 + 1.778D + 30s 19
+ 1.389D + 30s 20 + 1.942D + 29s 21 + 7.424D + 27s 22 + 1.206D + 25s 23

+ 5.009D + 21s 24 + 8.725D + 16s 25 + 3.883D + 11s 26
C3 (s) ≈  (4.69)
8.38D − 122s + 8.20D − 105s + 1.925D − 89s + 1.125D − 74s 4
2 3

+ 2.608D − 61s 5 + 1.518D − 48s 6 + 3.519D − 37s 7 + 2.048D − 26s 8

+ 4.747D − 17s 9 + 2.763D − 08s 10 + 0.6438997s 11 + 3810107.8s 12
+ 1.351D + 12s 13 + 1.632D + 17s 14 + 6.687D + 21s 15 + 1.530D + 25s 16
+ 8.866D + 27s 17 + 2.055D + 29s 18 + 1.196D + 30s 19 + 2.772D + 29s 20
+ 1.613D + 28s 21 + 3.740D + 25s 22 + 2.178D + 22s 23

+ 5.167D + 17s 24 + 3.311D + 12s 25 + 3582017.2s 26


1.267D + 11 + 3.508D + 14s + 2.420D + 17s 2 + 6.161D + 18s 3
+ 4.330D + 19s 4 + 5.405D + 19s 5 + 4.549D + 19s 6
+ 6.380D + 18s 7 + 2.426D + 17s 8 + 3.512D + 14s 9

+ 1.267D + 11s 10
C4 (s) ≈  (4.70)
65011.218 + 1.534D + 11s + 4.074D + 14s 2 + 2.652D + 17s 3
+ 6.223D + 18s 4 + 3.657D + 19s 5 + 8.410D + 18s 6
+ 4.843D + 17s 7 + 1.021D + 15s 8 + 5.273D + 11s 9

+ 1949188.2s 10
182 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

80

60
Magnitude (dB)

40

20

0
−2 −1 0 1 2 3 4
10 10 10 10 10 10 10
Frequency (rad/s)

100

50
Phase (degree)

−50

−100
−2 −1 0 1 2 3 4
10 10 10 10 10 10 10
Frequency (rad/s)

Fig. 4.16 Bode plot of PIλ Dμ controller approximated using various techniques


3.679D + 13 + 3.170D + 15s + 6.753D + 16s 2 + 2.307D + 17s 3
+ 3.792D + 17s 4 + 3.514D + 17s 5 + 1.796D + 17s 6
+ 4.594D + 16s 7 + 5.468D + 15s 8 + 2.581D + 14s 9

 + 4.080D + 12s 10
C5 (s) ≈
4.459D + 10 + 4.818D + 13s + 3.652D + 15s 2 + 6.585D + 16s 3
+ 1.583D + 17s 4 + 1.369D + 17s 5 + 4.809D + 16s 6
+ 7.224D + 15s 7 + 4.185D + 14s 8 + 8.006D + 12s 9

+ 4810788.6s 10
(4.71)
Thus, the resultant Bode plot of PIλ Dμ controller approximated using various
techniques in comparison with conventional PID controller is shown in Fig. 4.16.
Similarly, the Bode plot of M. Tenoutit’s (PI)n , (PD)n and (PID)n controllers
given in Sect. 4.7.3 with gain parameters K p = K i = K d = 1 and fractional-order
parameter n = 0.98 in comparison with conventional PID controller can be obtained
using the following commands. Here, the Oustaloup approximation technique has
been used for the implementation of controllers.
4.9 Simulation Study 183

– –> exec(’pid.sci’); exec(’fpidTen.sci’); C1=pid(1,1,1);

C2=fpidTen(1,1,0,’o’,0.98,5,[10ˆ-5 10ˆ5]);
C3=fpidTen(1,0,1,’o’,0.98,5,[10ˆ-5 10ˆ5]);
C4=fpidTen(1,1,1,’o’,0.98,5,[10ˆ-5 10ˆ5]);
bode([C1;C2;C3;C4],"rad");
legend(’$PID$’,’$(PI)ˆn$’,’$(PD)ˆn$’,’$(PID)ˆn$’);
Thus, the resultant approximated transfer functions of (PI)n , (PD)n and (PID)n
controllers are given as follows:

79432.823 + 84095975 s + 9.639D + 08s 2 + 9.720D + 08s 3

 + 92227983s + 96464.898s 3 + s
4 5 6
C2 (s) ≈ (4.72)
s + 96463.898s + 92131519s

+ 8.798D + 08s 4 + 84016543s 5 + 79432.823s 6

79432.823 + 84095975 s + 9.639D + 08s 2 + 9.720D + 08s 3

+ 92227983s 4 + 96464.898s 5 + s 6
C3 (s) ≈  (4.73)
1 + 96463.898s + 92131519s 2

+ 8.798D + 08s 3 + 84016543s 4 + 79432.823s 5

79432.823 + 84095975 s + 9.639D + 08s 2 + 1.056D + 09s 3

 + 9.721D + 08s 4
+ 92227984s 5
+ 96464.898s 6
+ s 7
C3 (s) ≈
s + 96463.898s 2 + 92131519s 3

+ 8.798D + 08s 4 + 84016543s 5 + 79432.823s 6
(4.74)
Therefore, the Bode plot of (PI)n , (PD)n and (PID)n controllers in comparison
with conventional PID controller is shown in Fig. 4.17.
In a similar way, the Bode plot of Luo’s [PI]λ and [PD]μ controllers given in
Sect. 4.7.4 with gain parameters K p = K i = K d = 1 and fractional-order parameters
λ = 0.98 and μ = 0.85 in comparison with conventional PI and PD controllers can
be obtained as follows:
– –> exec(’pid.sci’); exec(’fpiLuo.sci’); exec(’fpdLuo.sci’);
C1=pid(1,1,0); C2=fpiLuo(1,1,0.98);
C3=pid(1,0,1); C4=fpdLuo(1,1,0.85);
bode([C1;C2;C3;C4],"rad");
legend(’$PI$’,’$[PI]ˆ\lambda$’,’$PD$’,’$[PD]ˆ\mu$’);
From the above commands, the transfer functions of [PI]λ and [PD]μ controllers
are obtained as follows:

0.003332 − 0.0098s + 0.98s 2 + s 3

C2 (s) = (4.75)
s3
184 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

50
Magnitude (dB)

−50

−2 −1 0 1 2 3 4
10 10 10 10 10 10 10
Frequency (rad/s)

100
Phase (degree)

−100

−200
−2 −1 0 1 2 3 4
10 10 10 10 10 10 10
Frequency (rad/s)

Fig. 4.17 Bode plot of (PI)n , (PD)n and (PID)n controllers

1 + 0.85s − 0.06375s 2 + 0.0244375s 3

C4 (s) = (4.76)
1

Thus, the Bode plot of [PI]λ and [PD]μ controllers in comparison with conven-
tional PI and PD controllers is shown in Fig. 4.18.
On the other hand, the Bode plot of modified fractional-order PI and PID
controllers given in Sects. 4.7.6 and 4.7.7 respectively with controller parameters
K p = 1, K i = 1, α = 0.98, η = 0.1 and μ = 0.85 shown in Fig. 4.19 can be obtained
using the following commands:
– –> exec(’pid.sci’); exec(’modFPI.sci’);
exec(’modFPIss.sci’); exec(’modFPID.sci’);
C1=pid(1,1,0); C2=modFPI(1,2,0.98,’o’,5,[10ˆ-5 10ˆ5]);
C3=modFPIss(1,1,0.98,0.1,’o’,5,[10ˆ-5 10ˆ5]);
C4=pid(1,1,1); C5=modFPID(1,1,0.85,’o’,5,[10ˆ-5 10ˆ5]);
bode([C1;C2;C3;C4;C5],"rad");
legend(’PI’,’mod FPI’,’mod FPI ss’,’PID’,’mod FPID’);
4.9 Simulation Study 185

200

150
Magnitude (dB)

100

50

0
−2 −1 0 1 2 3 4
10 10 10 10 10 10 10
Frequency (rad/s)

100

50
Phase (degree)

−50

−100
−2 −1 0 1 2 3 4
10 10 10 10 10 10 10
Frequency (rad/s)

Fig. 4.18 Bode plot of [PI]λ and [PD]μ controllers

From the above commands, the approximated transfer functions of modified

fractional-order PI and PID controllers are obtained as follows:

79432.823 + 84175408 s + 1.048D + 09s 2 + 1.852D + 09s 3

 + 1.844D + 08s 4
+ 192928.8s 5
+ 2s 6
C2 (s) ≈ (4.77)
1 + 96463.898s + 92131519s 2 + 8.798D + 08s 3

+ 84016543s 4 + 79432.823s 5

7943.2823 + 8489030.4s + 1.805D + 08s 2 + 1.061D + 09s 3

 + 9.812D + 08s + 92237630s 5 + 96464.998s 6 + s 7
4
C3 (s) ≈
s + 96463.898s 2 + 92131519s 3 + 8.798D + 08s 4

+ 84016543s 5 + 79432.823s 6
(4.78)
186 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

80
PI
mod FPI
60
Magnitude (dB)

mod FPI ss
PID
40 mod FPID

20

0
−2 −1 0 1 2 3 4
10 10 10 10 10 10 10

Frequency (rad/s)

100
Phase (degree)

−100

−200
−2 −1 0 1 2 3 4
10 10 10 10 10 10 10

Frequency (rad/s)

Fig. 4.19 Bode plot of modified fractional-order PI and PID controllers


5.624D + 12 + 2.412D + 16s + 3.485D + 19s 2 + 1.749D + 22s 3
+ 7.692D + 23s 4 + 1.228D + 25s 5 + 7.702D + 25s 6
+ 1.347D + 26s 7 + 8.499D + 25s 8 + 1.703D + 25s 9
+ 1.305D + 24s 10 + 3.533D + 22s 11 + 9.329D + 19s 12

+ 8.139D + 16s 13 + 2.352D + 13s 14 + 3.163D + 08s 15
C5 (s) ≈ 
3.162D + 08 + 2.352D + 13s + 8.120D + 16s 2 + 9.277D + 19s 3
+ 3.486D + 22s 4 + 1.166D + 24s 5 + 1.317D + 25s 6
+ 4.927D + 25s 7 + 1.647D + 25s 8 + 1.861D + 24s 9
+ 6.958D + 22s 10 + 2.321D + 20s 11 + 2.611D + 17s 12

+ 9.636D + 13s 13 + 2.569D + 09s 14 + 17782.794s 15
(4.79)
As a final analysis of this study, the Bode plot of IMC based fractional-order PID
given in Sect. 4.7.5 with controller parameters K p = 1, K i = 1, K d = 1, γ = 0.98
and α = 0.98 in comparison with conventional PID can be obtained as follows:
4.9 Simulation Study 187

80
PID
IMC FPID
60
Magnitude (dB)

40

20

0
−2 −1 0 1 2 3 4
10 10 10 10 10 10 10

Frequency (rad/s)

100

50
Phase (degree)

−50

−100
−2 −1 0 1 2 3 4
10 10 10 10 10 10 10

Frequency (rad/s)

Fig. 4.20 Bode plot of IMC based fractional-order PID

– –> exec(’pid.sci’); exec(’IMCfpid.sci’); C1=pid(1,1,1);

C2=IMCfpid(1,1,1,0.98,0.98,’o’,5,[10ˆ-5 10ˆ5]);
bode([C1;C2],"rad"); legend(’PID’,’IMC FPID’);
Thus, the resultant Bode plot from the above commands is shown in Fig. 4.20.
Furthermore, the approximated transfer function obtained as follows:

5.012D + 14 + 1.592D + 18s + 1.702D + 21s 2 + 6.321D + 23s 3
+ 2.019D + 25s 4 + 2.383D + 26s 5 + 1.167D + 27s 6
+ 2.061D + 27s 7 + 1.868D + 27s 8 + 4.459D + 26s 9
+ 4.117D + 25s 10 + 1.369D + 24s 11 + 3.825D + 21s 12

+ 3.716D + 18s 13 + 1.232D + 15s 14 + 1.249D + 10s 15
C2 (s) ≈  (4.80)
6.310D + 09 + 6.220D + 14s + 1.877D + 18s 2 + 1.932D + 21s 3
+ 6.914D + 23s 4 + 2.077D + 25s 5 + 2.243D + 26s 6
+ 9.277D + 26s 7 + 9.175D + 26s 8 + 2.207D + 26s 9
+ 2.039D + 25s 10 + 6.776D + 23s 11 + 1.893D + 21s 12

+ 1.839D + 18s 13 + 6.096D + 14s 14 + 6.183D + 09s 15
188 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

4.9.2.2 Control of Non-linear Spherical Tank System

The transfer function model of the non-linear spherical tank system reported in [48]
is given as
3.62
G 1 (s) = (4.81)
330.46s + 1

The G 1 (s) can be defined in Scilab as follows:

– –> s=poly(0,’s’); s=syslin(’c’,s/((0*s)+1));
G=(3.6215/(330.46*s+1));
The tuned PID and fractional-order PID based controllers are defined as follows:

– –> Kp=6.81; Ki=1.01; Kd=1.91;

lam=0.98; mu=0.75; n=0.05;
alp=lam; gam=0.1*Kd;
Ti=Kp/Ki; Kc=Kp/Ti; Td=Kd/Kp;
The selected Oustaloup approximation parameters are specified as follows:
– –> N=5; wl=10ˆ-3; wh=10ˆ3;

Thus, the respective controllers PID, PI, PIλ Dμ , PIλ , (PID)n , IMC PID, Mod FPID
and [PI]λ as given in Sect. 4.7 are defined as

1.01
C P I D (s) = 6.81 + + 1.91s (4.82)
s
1.01
C P I (s) = 6.81 + (4.83)
s
1.01
C P I λ Dμ (s) = 6.81 + + 1.91s 0.75 (4.84)
s 0.98
1.01
C P I λ (s) = 6.81 + (4.85)
s 0.98
 0.05
1.01
C(P I D)n (s) = 6.81 + + 1.91s (4.86)
s

1.01 1.91s 0.98

C I MC P I D (s) = 6.81 + + (4.87)
s 0.98 0.19s 0.98 + 1

(1 + 6.74s 0.75 )2
C Mod F P I D (s) = 1.01 (4.88)
s 0.75
4.9 Simulation Study 189
 
1.01 0.98
C[P I ]λ (s) = 6.81 + (4.89)
s

The designed fractional-order controllers will be implemented using developed

functions in Sect. 4.7 as follows:
– –> exec(’pid.sci’); C1=pid(Kp,Ki,Kd); C11=pid(Kp,Ki,0);
exec(’fpid.sci’);
C2=fpid(Kp,Ki,Kd,[lam mu],’o’,N,[wl wh]);
exec(’fpi.sci’); C21=fpi(Kp,Ki,lam,’o’,N,[wl wh]);
exec(’fpidTen.sci’);
C3= fpidTen(Kp,Ki,0,’o’,n,N,[wl wh]);
exec(’IMCfpid.sci’);
C4=IMCfpid(Kp,Ki,Kd,gam,alp,’o’,N,[wl wh]);
exec(’modFPID.sci’); C5=modFPID(Kc,Ti,mu,’o’,N,[wl wh]);
exec(’fpiLuo.sci’); C6=fpiLuo(Kp,Ki,lam);
The closed-loop transfer functions of G 1 (s) with controllers defined above are
obtained as
– –> Gcl1=(G*C1)/(1+(G*C1)); Gcl11=(G*C11)/(1+(G*C11));
Gcl2=(G*C2)/(1+(G*C2)); Gcl21=(G*C21)/(1+(G*C21));
Gcl3=(G*C3)/(1+(G*C3)); Gcl4=(G*C4)/(1+(G*C4));
Gcl5=(G*C5)/(1+(G*C5)); Gcl6=(G*C6)/(1+(G*C6));
The step response of the system with all the designed controllers is obtained
using the following commands. Thus, the resultant plot is shown in Fig. 4.21. From
the response, it can be seen that the fractional-order controllers can be effectively
implemented in Scilab.
– –> t=0:0.01:200; scf();
plot([t’,t’,t’,t’,t’,t’,t’,t’],...
[csim(’step’,t,Gcl1)’,csim(’step’,t,Gcl11)’,...
csim(’step’,t,Gcl21)’,csim(’step’,t,Gcl2)’,...
csim(’step’,t,Gcl3)’,csim(’step’,t,Gcl4)’,...
csim(’step’,t,Gcl5)’,csim(’step’,t,Gcl6)’]);
legend(’PID’,’PI’,’$\mathrm{PIˆ\lambda}$’,...
’$\mathrm{PIˆ\lambda Dˆ\mu}$’,’$\mathrm{(PID)ˆn}$’,...
’IMC PID’,’Mod FPID’,’$\mathrm{[PI]ˆ\lambda}$’,...
"in_lower_right");
xlabel("Time (sec)"); ylabel("Amplitude");
Furthermore, the numerical assessment of the figure given in Table 4.6 can be
obtained with the use of stepinfo.sci function as follows:
– –> exec(’stepinfo.sci’); yf=1;
Gcl1s=stepinfo(Gcl1,t,yf); Gcl11s=stepinfo(Gcl11,t,yf);
Gcl2s=stepinfo(Gcl2,t,yf); Gcl21s=stepinfo(Gcl21,t,yf);
Gcl3s=stepinfo(Gcl3,t,yf); Gcl4s=stepinfo(Gcl4,t,yf);
Gcl5s=stepinfo(Gcl5,t,yf); Gcl6s=stepinfo(Gcl6,t,yf);
190 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

1.6

1.4

1.2

1
Amplitude

0.8
PID
PI
0.6

0.4

IMC PID
0.2 Mod FPID

0
0 20 40 60 80 100 120 140 160 180 200
Time (sec)

Fig. 4.21 Step response of G 1 (s) with various fractional-order controllers

Table 4.6 Step info of G 1 (s) with various controllers

Controller tr (s) ts (s) %OS
PID 9.8296 99.6926 37.4939
PI 9.7107 98.3596 37.9709
PIλ 9.8950 96.5135 35.4722
PIλ Dμ 10.2393 97.4630 34.1414
(PID)n 9.1032 94.8763 40.2484
IMC PID 10.2300 97.5639 34.9400
Mod FPID 12.0992 71.0458 4.4632
[PI]λ 10.0441 105.1809 37.7394

From the table, it can be seen that the Mod FPID controller has the least overshoot
of 4.46% and the fastest settling time of 71.04 s. However, for rise time (PID)n
controller has the fastest rise time of 9.10 s. Furthermore, the open-loop Bode plots
of the system with various controllers can be obtained using the following commands.
Thus, the resulting plot is shown in Fig. 4.22.
– –> scf();
bode([G*C1;G*C11;G*C21;G*C2;G*C3;G*C4;G*C5;G*C6],wl,wh);
legend(’PID’,’PI’,’$\mathrm{PIˆ\lambda}$’,...
’$\mathrm{PIˆ\lambda Dˆ\mu}$’,’$\mathrm{(PID)ˆn}$’,...
’IMC PID’,’Mod FPID’,’$\mathrm{[PI]ˆ\lambda}$’,...
"in_lower_left");
4.9 Simulation Study 191

40

20
PID
0
PI
Magnitude (dB)

−20

−40

−60 IMC PID

Mod FPID
−80

−100
−3 −2 −1 0 1 2 3
10 10 10 10 10 10 10
Frequency (Hz)

−50
Phase (degree)

−100

−150

−3 −2 −1 0 1 2 3
10 10 10 10 10 10 10
Frequency (Hz)

Fig. 4.22 Bode plot of G 1 (s) with various fractional-order controllers

Similarly, the other frequency-domain plots like Nyquist and Nichols are obtained
using the following commands. Thus, the Nyquist and Nichols plot of G 1 (s) with
various controllers are shown in Figs. 4.23 and 4.24 respectively.
– –> scf();
nyquist([G*C1;G*C11;G*C21;G*C2;G*C3;G*C4;G*C5;G*C6]);
legend(’PID’,’PI’,’$\mathrm{PIˆ\lambda}$’,...
’$\mathrm{PIˆ\lambda Dˆ\mu}$’,’$\mathrm{(PID)ˆn}$’,...
’IMC PID’,’Mod FPID’,’$\mathrm{[PI]ˆ\lambda}$’,...
"in_lower_left");

scf();
black([G*C1;G*C11;G*C21;G*C2;G*C3;G*C4;G*C5;G*C6]);
legend(’PID’,’PI’,’$\mathrm{PIˆ\lambda}$’,...
’$\mathrm{PIˆ\lambda Dˆ\mu}$’,’$\mathrm{(PID)ˆn}$’,...
’IMC PID’,’Mod FPID’,’$\mathrm{[PI]ˆ\lambda}$’,...
"in_lower_left");
192 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

Nyquist plot
20
−0.00259

−0.00259
15
−0.00259
−0.00259 −0.00259
−0.00381
10
PID −0.00381
−0.00381
−0.00381
−0.00381
5 PI
−0.0092
−0.0092
−0.0092
−0.0092
1e+03
0 0.0092
0.0092
0.0092
0.0092
IMC PID 0.00381
0.00381
0.00381
−5 Mod FPID 0.00381

−10 0.00259 0.00381

0.00259
0.00259
0.00259
−15
0.00259

−20
−60 −55 −50 −45 −40 −35 −30 −25 −20 −15 −10 −5 0 5 10

Fig. 4.23 Nyquist plot of G 1 (s) with various fractional-order controllers

40
0.0026
0dB 0.0026
0.0026
0.25dB −0.25dB
0.0026
30
0.5dB 0.0026
−0.5dB

1dB −1dB
20

2.3dB −2dB
10 4dB 0.013 −3dB
6dB 0.013
0.013
0.013
0.013
Magnitude (dB)

PID 12dB
−6dB
0 PI 0.071
−9dB
−12dB
−10 −15dB
0.071
0.071
0.071
0.071
25
IMC PID −20dB
−20
Mod FPID −25dB

−30dB
−30 2.3dB
−39dB
−40
−400 −350 −300 −250 −200 −150 −100 −50 0
Phase (deg)

Fig. 4.24 Nichols plot of G 1 (s) with various fractional-order controllers

4.9 Simulation Study 193

4.9.2.3 Control of Unstable Bioreactor Process

To further evaluate the performance of the proposed Scilab based fractional-order

controllers, an unstable Bioreactor Process reported in [48] is selected. Thus, the
transfer function model of the reactor process is given by

−0.9951s − 0.2985
G 2 (s) = (4.90)
s2 + 0.1302s − 0.0509

From Eq. (4.90), it can be seen that the process is unstable with a pole at 0.1697.
For the control of this process, the tuned PID and fractional-order PID parameters
are K p = −0.54, K i = −0.07, K d = −0.05, λ = α = 0.90, μ = 0.85 and n = 0.1.
Thus, the controller transfer functions of PID, PIλ Dμ , (PID)n , IMC PID and Mod
FPID are given as 
0.07
C P I D (s) = − 0.54 + + 0.05s (4.91)
s

0.07
C P I λ Dμ (s) = − 0.54 + 0.90 + 0.05s 0.85 (4.92)
s
 0.1
0.07
C(P I D)n (s) = − 0.54 + + 0.05s (4.93)
s

0.07 0.05s 0.90
C I MC P I D (s) = − 0.54 + 0.90 + (4.94)
s 0.0005s 0.90 + 1

(1 − 0.07s 0.85 )2
C Mod F P I D (s) = 7.71 (4.95)
s 0.85
The designed controllers will be implemented for the control of unstable biore-
actor process in Scilab as follows:
– –> s=poly(0,’s’); s=syslin(’c’,s/((0*s)+1));
G=((-0.9951*s-0.2985)/(sˆ2+0.1302*s-0.0509));
Kp=-0.5374; Ki=-0.0702; Kd=-0.0532;
lam=0.9; mu=0.85; n=0.1; alp=lam; gam=0.01*Kd;
Ti=Kp/Ki; Kc=Kp/Ti; Td=Kd/Kp; N=5; wl=10ˆ-3; wh=10ˆ3;
exec(’pid.sci’); C1=pid(Kp,Ki,Kd); exec(’fpid.sci’);
C2=fpid(Kp,Ki,Kd,[lam mu],’o’,N,[wl wh]);
exec(’fpidTen.sci’); exec(’IMCfpid.sci’);
C3=fpidTen(Kp,Ki,Kd,’o’,n,N,[wl wh]);
C4=IMCfpid(Kp,Ki,Kd,gam,alp,’o’,N,[wl wh]);
exec(’modFPID.sci’);
C5=modFPID(Kc,Ti,mu,’o’,N,[wl wh]);
Gcl1=(G*C1)/(1+(G*C1)); Gcl2=(G*C2)/(1+(G*C2));
Gcl3=(G*C3)/(1+(G*C3)); Gcl4=(G*C4)/(1+(G*C4));
Gcl5=(G*C5)/(1+(G*C5));
194 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

1.6

1.4

1.2

1
Amplitude

0.8

0.6 PID

0.4
IMC PID
Mod FPID
0.2

0
0 10 20 30 40 50 60 70 80 90 100
Time (sec)

Fig. 4.25 Step response of G 2 (s) with various fractional-order PID controllers

Therefore, the step response of bioreactor with the above-designed fractional-

order PID controllers can be obtained in Scilab as follows:
– –> t=0:0.01:100; scf();
plot([t’,t’,t’,t’,t’],[csim(’step’,t,Gcl1)’,...
csim(’step’,t,Gcl2)’,csim(’step’,t,Gcl3)’,...
csim(’step’,t,Gcl4)’,csim(’step’,t,Gcl5)’])
legend(’PID’,’$\mathrm{PIˆ\lambda Dˆ\mu}$’,...
’$\mathrm{(PID)ˆn}$’,’IMC PID’,’Mod FPID’,...
"in_lower_right"); xgrid;
xlabel("Time (sec)"); ylabel("Amplitude");
Thus, the resultant step response plot is shown in Fig. 4.25. Furthermore, the
assessment of the figure numerically in terms of tr , ts and %OS is given in Table 4.7
is obtained as follows:

– –> yf=1; Gcl1s=stepinfo(Gcl1,t,yf);

Gcl2s=stepinfo(Gcl2,t,yf); Gcl3s=stepinfo(Gcl3,t,yf);
Gcl4s=stepinfo(Gcl4,t,yf); Gcl5s=stepinfo(Gcl5,t,yf);
From the assessment, it can be seen that the designed fractional-order PID con-
trollers can be successfully implemented for this unstable process using Scilab. It
can also be seen that the Mod FPID controller has outperformed all the other con-
trollers with least overshoot of 12.01% and the fastest rise time of 1.22 s. However,
for settling time the (PID)n controller has the fastest settling time of 24.53 s.
4.9 Simulation Study 195

Table 4.7 Step info of G 2 (s) with various controllers

Controller tr (s) ts (s) %OS
PID 2.0840 26.6758 42.7678
PIλ Dμ 2.1322 31.0234 39.9924
(PID)n 1.9186 24.5391 43.7835
IMC PID 2.1436 30.8774 40.1460
Mod FPID 1.2216 78.6227 12.0132

Similarly, the frequency response plots like Bode, Nyquist, and Nichols for the
process with the designed fractional-order PID controllers can be obtained as follows:

– –> scf(); bode([G*C1;G*C2;G*C3;G*C4;G*C5],wl,wh);

legend(’PID’,’$\mathrm{PIˆ\lambda Dˆ\mu}$’,...
’$\mathrm{(PID)ˆn}$’,’IMC PID’,’Mod FPID’,...
"in_lower_right");
scf(); nyquist([G*C1;G*C2;G*C3;G*C4;G*C5]);
legend(’PID’,’$\mathrm{PIˆ\lambda Dˆ\mu}$’,...
’$\mathrm{(PID)ˆn}$’,’IMC PID’,’Mod FPID’);
Thus, the resultant Bode and Nyquist are shown in Figs. 4.26 and 4.27 respectively.
Thus, in all the cases it can be concluded that the fractional-order controllers can be
effectively designed, implemented and analyzed using Scilab.

4.9.3 Fractional-order Filters

This section analyses the designed fractional-order filters for various values of param-
eters. In all the cases, the pole frequency (ωc ) is chosen as 5 rad/s. Furthermore,
the approximation is done using Oustaloup’s technique with parameters N = 5,
ωl = 10−3 and ωh = 103 .

4.9.3.1 Analysis of Fractional-order Low-Pass Filter

The Bode responses of the fractional-order low-pass for various values of α between
(0, 1) shown in Fig. 4.28 can be obtained using the commands
196 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

40

20
Magnitude (dB)

−20

−40

−60
−3 −2 −1 0 1 2 3
10 10 10 10 10 10 10
Frequency (Hz)
PID

400 IMC PID

Mod FPID
300
Phase (degree)

200

100

−3 −2 −1 0 1 2 3
10 10 10 10 10 10 10
Frequency (Hz)

Fig. 4.26 Bode plot of G 2 (s) with various fractional-order PID controllers

– –> omegaC=5; N=5; wl=10ˆ-3; wh=10ˆ3;

exec(’foLPF.sci’); i=1;
for alpha=0.1:0.1:0.9
H(i)=foLPF(alpha,omegaC,’o’,N,[wl wh]);
i=i+1;
end
scf();
bode([H(1);H(2);H(3);H(4);H(5);H(6);H(7);H(8);H(9)],...
wl,wh);
legend(’$\alpha=0.1$’,’$\alpha=0.2$’,’$\alpha=0.3$’,...
’$\alpha=0.4$’,’$\alpha=0.5$’,’$\alpha=0.6$’,...
’$\alpha=0.7$’,’$\alpha=0.8$’,’$\alpha=0.9$’,...
"in_lower_left");

The impulse and step responses of the fractional-order low-pass filter for 0 < α <
1 shown in Fig. 4.29 can be obtained as follows:
4.9 Simulation Study 197

Nyquist plot
50
PID
40 0.00259

30
0.00259 IMC PID
20
Mod FPID
0.00259
0.00259 0.00673
10 0.00673
0.00673
0.00673
1e+03
1e+03
0 −0.00673
−0.00673
−0.00673
−0.00259 −0.00673
−10
−0.00259

−20
−0.00259

−30
−0.00259
−40

−50
−14 −12 −10 −8 −6 −4 −2 0 2 4 6

Fig. 4.27 Nyquist plot of G 2 (s) with various fractional-order PID controllers

−10
Magnitude (dB)

−20

−30

−40

−3 −2 −1 0 1 2 3
10 10 10 10 10 10 10

Frequency (rad/sec)

0
Phase (degree)

−20

−40

−60

−80
−3 −2 −1 0 1 2 3
10 10 10 10 10 10 10

Frequency (rad/sec)

Fig. 4.28 Bode response of fractional-order low-pass filter for 0 < α < 1
198 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

Impulse Response
5

3
Amplitude

0
0 1 2 3 4 5 6 7 8 9 10
Time (sec)

Step Response
1

0.8
Amplitude

0.6

0.4

0.2

0
0 1 2 3 4 5 6 7 8 9 10

Time (sec)

Fig. 4.29 Impulse and step responses of foLPF for 0 < α < 1

– –> t=0:0.01:10; scf(); subplot(211);

plot([t’,t’,t’,t’,t’,t’,t’,t’,t’],...
[csim(’impulse’,t,H(1))’,csim(’impulse’,t,H(2))’,...
csim(’impulse’,t,H(3))’,csim(’impulse’,t,H(4))’,...
csim(’impulse’,t,H(5))’,csim(’impulse’,t,H(6))’,...
csim(’impulse’,t,H(7))’,csim(’impulse’,t,H(8))’,...
csim(’impulse’,t,H(9))’]);
legend(’$\alpha=0.1$’,’$\alpha=0.2$’,’$\alpha=0.3$’,...
’$\alpha=0.4$’,’$\alpha=0.5$’,’$\alpha=0.6$’,...
’$\alpha=0.7$’,’$\alpha=0.8$’,’$\alpha=0.9$’,...
"in_lower_left");
a=gca(); a.data_bounds=[0 0;10 5];
title("Impulse Response");
xlabel("Time (sec)"); ylabel("Amplitude");
4.9 Simulation Study 199

subplot(212);
plot([t’,t’,t’,t’,t’,t’,t’,t’,t’],...
[csim(’step’,t,H(1))’,csim(’step’,t,H(2))’,...
csim(’step’,t,H(3))’,csim(’step’,t,H(4))’,...
csim(’step’,t,H(5))’,csim(’step’,t,H(6))’,...
csim(’step’,t,H(7))’,csim(’step’,t,H(8))’,...
csim(’step’,t,H(9))’]); title("Step Response");
xlabel("Time (sec)"); ylabel("Amplitude");
The Bode responses of the fractional-order low-pass for various values of α
between (1, 2) shown in Fig. 4.30 can be obtained using the commands
– –> exec(’foLPF.sci’); i=1;
for alpha=1.1:0.1:1.8
H(i)=foLPF(alpha,omegaC,’o’,N,[wl wh]);
i=i+1;
end
clf();
bode([H(1);H(2);H(3);H(4);H(5);H(6);H(7);H(8);])
legend(’$\alpha=1.1$’,’$\alpha=1.2$’,...
’$\alpha=1.3$’,’$\alpha=1.4$’,’$\alpha=1.5$’,...
’$\alpha=1.6$’,’$\alpha=1.7$’,’$\alpha=1.8$’,...
"in_lower_left");

0
Magnitude (dB)

−50

−100

−3 −2 −1 0 1 2 3
10 10 10 10 10 10 10

Frequency (rad/sec)

0
Phase (degree)

−50

−100

−150

−3 −2 −1 0 1 2 3
10 10 10 10 10 10 10

Frequency (rad/sec)

Fig. 4.30 Bode response of fractional-order low-pass filter for 1 < α < 2
200 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

Similarly, the impulse and step responses of the filter for 1 < α < 2 shown in
Fig. 4.31 can be obtained as follows:
– –> t=0:0.01:10; scf(); subplot(211);
plot([t’,t’,t’,t’,t’,t’,t’,t’],...
[csim(’impulse’,t,H(1))’,csim(’impulse’,t,H(2))’,...
csim(’impulse’,t,H(3))’,csim(’impulse’,t,H(4))’,...
csim(’impulse’,t,H(5))’,csim(’impulse’,t,H(6))’,...
csim(’impulse’,t,H(7))’,csim(’impulse’,t,H(8))’]);
legend(’$\alpha=1.1$’,’$\alpha=1.2$’,’$\alpha=1.3$’,...
’$\alpha=1.4$’,’$\alpha=1.5$’,’$\alpha=1.6$’,...
’$\alpha=1.7$’,’$\alpha=1.8$’,"in_lower_left");
title("Impulse Response");
xlabel("Time (sec)"); ylabel("Amplitude");

subplot(212);
plot([t’,t’,t’,t’,t’,t’,t’,t’],...
[csim(’step’,t,H(1))’,csim(’step’,t,H(2))’,...
csim(’step’,t,H(3))’,csim(’step’,t,H(4))’,...
csim(’step’,t,H(5))’,csim(’step’,t,H(6))’,...
csim(’step’,t,H(7))’,csim(’step’,t,H(8))’]);
title("Step Response");
xlabel("Time (sec)"); ylabel("Amplitude");

4.9.3.2 Analysis of Fractional-order High-Pass Filter

Similarly, the Bode responses of the fractional-order high-pass for various values of
α between (0, 1) shown in Fig. 4.32 can be obtained using the commands
– –> omegaC=5; N=5; wl=10ˆ-3; wh=10ˆ3;
exec(’foHPF.sci’); i=1;
for alpha=0.1:0.1:0.9
H(i)=foHPF(alpha,omegaC,’o’,N,[wl wh]);
i=i+1;
end
scf();
bode([H(1);H(2);H(3);H(4);H(5);H(6);H(7);H(8);H(9)],...
wl,wh)
legend(’$\alpha=0.1$’,’$\alpha=0.2$’,’$\alpha=0.3$’,...
’$\alpha=0.4$’,’$\alpha=0.5$’,’$\alpha=0.6$’,...
’$\alpha=0.7$’,’$\alpha=0.8$’,’$\alpha=0.9$’,...
"in_lower_left");
Furthermore, the impulse and step response the filter for order 0 < α < 1 shown
in Fig. 4.33 can be obtained using the following commands
4.9 Simulation Study 201

Impulse Response
2

1
Amplitude

−1

0 1 2 3 4 5 6 7 8 9 10

Time (sec)
Step Response

1.5
Amplitude

0.5

0
0 1 2 3 4 5 6 7 8 9 10

Time (sec)

Fig. 4.31 Impulse and step responses of foLPF for 1 < α < 2

– –> t=0:0.01:10; scf(); subplot(211);

plot([t’,t’,t’,t’,t’,t’,t’,t’,t’],...
[csim(’impulse’,t,H(1))’,csim(’impulse’,t,H(2))’,...
csim(’impulse’,t,H(3))’,csim(’impulse’,t,H(4))’,...
csim(’impulse’,t,H(5))’,csim(’impulse’,t,H(6))’,...
csim(’impulse’,t,H(7))’,csim(’impulse’,t,H(8))’,...
csim(’impulse’,t,H(9))’]);
legend(’$\alpha=0.1$’,’$\alpha=0.2$’,’$\alpha=0.3$’,...
’$\alpha=0.4$’,’$\alpha=0.5$’,’$\alpha=0.6$’,...
’$\alpha=0.7$’,’$\alpha=0.8$’,’$\alpha=0.9$’,...
"in_lower_left");
a=gca(); a.data_bounds=[0 0;10 5];
title("Impulse Response");
xlabel("Time (sec)"); ylabel("Amplitude");
202 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

Magnitude (dB) 20

−20

−40
−3 −2 −1 0 1 2 3
10 10 10 10 10 10 10

Frequency (rad/sec)

80
Phase (degree)

60

40

20

0
−3 −2 −1 0 1 2 3
10 10 10 10 10 10 10

Frequency (rad/sec)

Fig. 4.32 Bode response of fractional-order high-pass filter for 0 < α < 1

subplot(212); plot([t’,t’,t’,t’,t’,t’,t’,t’,t’],...
[csim(’step’,t,H(1))’,csim(’step’,t,H(2))’,...
csim(’step’,t,H(3))’,csim(’step’,t,H(4))’,...
csim(’step’,t,H(5))’,csim(’step’,t,H(6))’,...
csim(’step’,t,H(7))’,csim(’step’,t,H(8))’,...
csim(’step’,t,H(9))’]); title("Step Response");
xlabel("Time (sec)"); ylabel("Amplitude");
The Bode responses of the fractional-order high-pass filter for various values of
α between (1, 2) shown in Fig. 4.34 can be obtained using the commands
– –> exec(’foHPF.sci’); i=1;
for alpha=1.1:0.1:1.8
H(i)=foHPF(alpha,omegaC,’o’,N,[wl wh]);
i=i+1;
end
scf();
bode([H(1);H(2);H(3);H(4);H(5);H(6);H(7);H(8);])
legend(’$\alpha=1.1$’,’$\alpha=1.2$’,...
’$\alpha=1.3$’,’$\alpha=1.4$’,’$\alpha=1.5$’,...
’$\alpha=1.6$’,’$\alpha=1.7$’,’$\alpha=1.8$’,...
"in_lower_left");
4.9 Simulation Study 203

Impulse Response
5

4
Amplitude

0
0 1 2 3 4 5 6 7 8 9 10

Time (sec)

Step Response

3
Amplitude

0
0 1 2 3 4 5 6 7 8 9 10

Time (sec)

Fig. 4.33 Impulse and step responses of foHPF for 0 < α < 1

Furthermore, for this order of α ∈ (1, 2), the impulse and step responses of the
filter shown in Fig. 4.35 can be obtained using the commands
– –> t=0:0.01:10; scf(); subplot(211);
plot([t’,t’,t’,t’,t’,t’,t’,t’],...
[csim(’impulse’,t,H(1))’,csim(’impulse’,t,H(2))’,...
csim(’impulse’,t,H(3))’,csim(’impulse’,t,H(4))’,...
csim(’impulse’,t,H(5))’,csim(’impulse’,t,H(6))’,...
csim(’impulse’,t,H(7))’,csim(’impulse’,t,H(8))’])
title("Impulse Response");
xlabel("Time (sec)"); ylabel("Amplitude");
204 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

Magnitude (dB)

−50

−3 −2 −1 0 1 2 3
10 10 10 10 10 10 10

Frequency (rad/sec)

150
Phase (degree)

100

50

0
−3 −2 −1 0 1 2 3
10 10 10 10 10 10 10

Frequency (rad/sec)

Fig. 4.34 Bode response of fractional-order high-pass filter for 1 < α < 2

subplot(212) plot([t’,t’,t’,t’,t’,t’,t’,t’],...
[csim(’step’,t,H(1))’,csim(’step’,t,H(2))’,...
csim(’step’,t,H(3))’,csim(’step’,t,H(4))’,...
csim(’step’,t,H(5))’,csim(’step’,t,H(6))’,...
csim(’step’,t,H(7))’,csim(’step’,t,H(8))’])
legend(’$\alpha=1.1$’,’$\alpha=1.2$’,’$\alpha=1.3$’,...
’$\alpha=1.4$’,’$\alpha=1.5$’,’$\alpha=1.6$’,...
’$\alpha=1.7$’,’$\alpha=1.8$’,"in_lower_left");
title("Step Response");
xlabel("Time (sec)"); ylabel("Amplitude");
From the Bode responses of both Fractional-order Low-pass and High-pass filters
given in Sects. 4.9.3.1 and 4.9.3.2 respectively, it can be observed that for α > 1,
the magnitude plot shows a resonant peak, which increases as the order approaches
to 2. On the other hand, the variation in phase response is linear for 0 < α < 1. As
the value α of increases from 1, the response exhibits saturation behavior towards
the asymptotic values. Furthermore, from the impulse and step responses, it can be
seen that for 1 < α < 2, the responses exhibits oscillations and the magnitude of
oscillations increases as the value of α approaches to 2.
4.9 Simulation Study 205

Impulse Response
8

4
Amplitude

−2

0 1 2 3 4 5 6 7 8 9 10

Time (sec)
Step Response
4

2
Amplitude

−2

0 1 2 3 4 5 6 7 8 9 10

Time (sec)

Fig. 4.35 Impulse and step responses of foHPF for 1 < α < 2

4.9.3.3 Analysis of Fractional-order Band-Pass and All-Pass Filter

On the other hand, the Bode response of fractional-order band-pass and all-pass
filters for 0 < α < 1 shown in Figs. 4.36 and 4.37 respectively are obtained using
the following commands. From the responses, it can be seen that the characteristics
of the fractional-order band and all-pass filters can be achieved effectively using the
developed Scilab commands.
– –> // Fractional-order Band-pass Filter
exec(’foBPF.sci’); N=5; wl=10ˆ-6; wh=10ˆ6; i=1;
for bita=0.1:0.1:0.8
H(i)=foBPF(0.9,bita,omegaC,’o’,N,[wl wh]); i=i+1;
end
bode([H(1);H(2);H(3);H(4);H(5);H(6);H(7);H(8)],wl,wh)
legend(’$\beta=0.1$’,’$\beta=0.2$’,’$\beta=0.3$’,...
’$\beta=0.4$’,’$\beta=0.5$’,’$\beta=0.6$’,...
’$\beta=0.7$’,’$\beta=0.8$’,"in_lower_left");
206 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

0
Magnitude (dB)

−50

−6 −5 −4 −3 −2 −1 0 1 2 3 4 5 6
10 10 10 10 10 10 10 10 10 10 10 10 10

Frequency (rad/sec)

100
Phase (degree)

50

−50

−6 −5 −4 −3 −2 −1 0 1 2 3 4 5 6
10 10 10 10 10 10 10 10 10 10 10 10 10

Frequency (rad/sec)

Fig. 4.36 Bode response of fractional-order band-pass filter

– –> // Fractional-order All-pass Filter

exec(’foAPF.sci’); N=5; wl=10ˆ-6; wh=10ˆ6; i=1;
for alpha=0.1:0.1:0.9
H(i)=foAPF(alpha,omegaC,’o’,N,[wl wh]); i=i+1;
end
bode([H(1);H(2);H(3);H(4);H(5);H(6);H(7);H(8);H(9)],...
wl,wh);
legend(’$\alpha=0.1$’,’$\alpha=0.2$’,’$\alpha=0.3$’,...
’$\alpha=0.4$’,’$\alpha=0.5$’,’$\alpha=0.6$’,...
’$\alpha=0.7$’,’$\alpha=0.8$’,’$\alpha=0.9$’,...
"in_lower_left");

4.9.3.4 Performance Analysis on Process Plant

To further analyze the performance of the designed filters, a second order stable
process plant given in Eq. (4.96) is considered.

1
G(s) = (4.96)
(s + 1)2
4.9 Simulation Study 207

−5
Magnitude (dB)

−10

−15

−20

−6 −5 −4 −3 −2 −1 0 1 2 3 4 5 6
10 10 10 10 10 10 10 10 10 10 10 10 10

Frequency (rad/sec)

150
Phase (degree)

100

50

0
−6 −5 −4 −3 −2 −1 0 1 2 3 4 5 6
10 10 10 10 10 10 10 10 10 10 10 10 10

Frequency (rad/sec)

Fig. 4.37 Bode response of fractional-order all-pass filter

For this process, the designed integer-order and fractional-order low-pass filters
with parameters ωc = 0.5 rad/s and α = 0.85 are:

20
F(s) = (4.97)
s + 20

(20)0.85
F(s) = (4.98)
s 0.85 + (20)0.85

The approximated transfer function of the filter using Oustaloup is given as fol-
lows: 
0.5547s 5 + 481.31s 4 + 24783.02s 3

+ 80196.30s 2 + 16309.05s + 196.84
F(s) ≈  (4.99)
355.36s 5 + 298336.95s 4 + 169336.95s 3

+ 124867s 2 + 17176.62s + 197.84
208 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

Fig. 4.38 Xcos diagram of the process with and without filters

Noise Signal
0.1

0.09

0.08

0.07

0.06
Amplitude

0.05

0.04

0.03

0.02

0.01

0
0 5 10 15 20 25 30 35 40 45 50
Time (sec)

Fig. 4.39 Generated noise signal used for the simulation

Furthermore, the designed PID controller for the control of this process is given
as follows:
1.42
C(s) = 2.03 + + 0.65s (4.100)
s
Therefore, the Xcos; a graphical editor in Scilab for the design of dynamical
systems diagram with these filters is shown in Fig. 4.38. The noise signal used for
the simulation is shown in Fig. 4.39. Therefore, the step response of the system
4.9 Simulation Study 209

Step Response

1.5
Amplitude

0.5
without filter
with integer−order filter
with Fractional−order filter
0
0 5 10 15 20 25 30 35 40 45 50

Time (sec)

1.5
Control Signal

0.5

0 5 10 15 20 25 30 35 40 45 50

Time (sec)

Fig. 4.40 Step responses of stable process with low-pass filters

with these filters is shown in Fig. 4.40. From the responses, it can be seen that the
fractional-order filters can be effectively implemented and analyzed using Scilab-
Xcos.
From the analysis, it can be seen that the developed fractional-order filters can
be effectively used for real-time processes. Furthermore, the characteristics of the
filter are also analyzed for various values of α in terms of Bode, impulse and step
responses. The simulation study shows that the developed fractional-order filters can
be effectively implemented and analyzed using Scilab.

4.10 Summary

In this work, a Scilab based toolbox for fractional-order systems and PID controllers
is developed. The proposed toolbox is the first open-source toolbox in open-source
software. The toolbox includes the definition of fractional-order parameters, approx-
imation techniques, fractional-order differentiator/integrator, fractional-order based
210 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

systems, and PI/PID controllers. Furthermore, the toolbox allows for time and fre-
quency domain as well as stability analysis of the fractional-order systems. The sim-
ulation study on various fractional-order based systems and controllers shows that
the implementation of fractional-order systems can be done easily on the developed
Scilab based toolbox.

References

1. Machado, J.T., Kiryakova, V., Mainardi, F.: Recent history of fractional calculus. Commun.
Nonlinear Sci. Numer. Simul. 16(3), 1140–1153 (2011)
2. Matušu, R.: Application of fractional order calculus to control theory. Int. J. Math. Models
Methods Appl. Sci. 5(7), 1162–1169 (2011)
3. Caponetto, R.: Fractional Order Systems: Modeling and Control Applications. World Scientific
(2010)
4. Petráš, I.: Fractional-Order Nonlinear Systems: Modeling, Analysis and Simulation. Springer
Science & Business Media (2011)
5. Shah, P., Agashe, S.: Review of fractional PID controller. Mechatronics 38, 29–41 (2016)
6. Xue, D., Chen, Y., Atherton, D.P.: Linear Feedback Control: Analysis and Design with MAT-
LAB. Siam (2007)
7. Monje, C.A., Chen, Y., Vinagre, B.M., Xue, D., Feliu-Batlle, V.: Fractional-Order Systems and
Controls: Fundamentals and Applications. Springer Science & Business Media (2010)
8. Tavazoei, M.S.: Time response analysis of fractional-order control systems: a survey on recent
results. Fract. Calc. Appl. Anal. 17(2), 440–461 (2014)
9. Bingi, K., Ibrahim, R., Karsiti, M.N., Hassan, S.M., Harindran, V.R.: Real-time control of
pressure plant using 2DOF fractional-order PID controller. Arab. J. Sci. Eng. 44(3), 2091–
2102 (2019)
10. Bingi, K., Ibrahim, R., Karsiti, M.N., Hassan, S.M.: Fractional order set-point weighted PID
controller for pH neutralization process using accelerated PSO algorithm. Arab. J. Sci. Eng.
43(6), 2687–2701 (2018)
11. Oustaloup, A., Melchior, P., Lanusse, P., Cois, O., Dancla, F.: The CRONE toolbox for Matlab.
In: CACSD. Conference Proceedings. IEEE International Symposium on Computer-Aided
Control System Design, Anchorage, USA, 25–27 Sept 2000
12. Malti, R., Melchior, P., Lanusse, P., Oustaloup, A.: Object-oriented CRONE toolbox for frac-
tional differential signal processing. Signal Image Video Process. 6(3), 393–400 (2012)
13. Malti, R., Victor, S.: Crone toolbox for system identification using fractional differentiation
models. IFAC-PapersOnLine 48(28), 769–774 (2015)
14. Malti, R., Melchior, P., Lanusse, P., Oustaloup, A.: Towards an object oriented CRONE toolbox
for fractional differential systems. IFAC Proc. Vol. 44(1), 10830–10835 (2011)
15. Valerio, D., Da Costa, J.S.: Ninteger: a non-integer control toolbox for MatLab. Proceedings
of Fractional Differentiation and Its Applications, Bordeaux (2004)
16. de Oliveira Valério, D.P.M.: Ninteger v. 2.3 Fractional Control Toolbox for MATLAB. Lisboa,
Universidade Technical (2005)
17. Tepljakov, A., Petlenkov, E., Belikov, J.: FOMCON: fractional-order modeling and control
toolbox for MATLAB. In: Proceedings of the 18th International Conference Mixed Design of
Integrated Circuits and Systems-MIXDES 2011, Gliwice, Poland, 16–18 June 2011
18. Tepljakov, A.: Fractional-Order Modeling and Control of Dynamic Systems. Springer (2017)
19. Tepljakov, A., Petlenkov, E., Belikov, J., Finajev, J.: Fractional-order controller design and
digital implementation using FOMCON toolbox for MATLAB. In: 2013 IEEE Conference on
Computer Aided Control System Design (CACSD), Hyderabad, India, 28–30 Aug 2013
References 211

20. Lachhab, N., Svaricek, F., Wobbe, F., Rabba, H.: Fractional order PID controller (FOPID)-
toolbox. In: 2013 European Control Conference (ECC), Zurich, Switzerland, 17–19 July 2013
21. Chen, Y., Petras, I., Xue, D.: Fractional order control-a tutorial. In: 2009 American Control
Conference, St. Louis, MO, USA, 10–12 June 2009
22. Dzieliński, A., Sierociuk, D.: Simulation and experimental tools for fractional order control
education. IFAC Proc. Vol. 41(2), 11654–11659 (2008)
23. Sierociuk, D.: Fractional order discrete state-space system simulink toolkit user guide. http://
www.ee.pw.edu.pl/~dsieroci/fsst/fsst.htm (2005)
24. Pisoni, E., Visioli, A., Dormido, S.: An interactive tool for fractional order PID controllers. In:
2009 35th Annual Conference of IEEE Industrial Electronics, Porto, Portugal, 3–5 Nov 2009
25. Dormido, S., Pisoni, E., Visioli, A.: An interactive tool for loop-shaping design of fractional-
order PID controllers. In: Proceedings of the 4th IFAC Workshop on Fractional Differentiation
and Its Applications, Badajoz, Spain, 18–20 Oct 2010
26. Dormido, S., Pisoni, E., Visioli, A.: Interactive tools for designing fractional-order PID con-
trollers. Int. J. Innov. Comp. Inf. Control 8(7), 4579–4590 (2012)
27. Petráš, I.: Fractional derivatives, fractional integrals, and fractional differential equations in
Matlab. In: Assi , A. (ed.) Engineering Education and Research Using MATLAB, pp. 239–
264, IntechOpen (2011)
28. Marinov, T.M., Ramirez, N., Santamaria, F.: Fractional integration toolbox. Fract. Calc. Appl.
Anal. 16(3), 670–681 (2013)
29. Lanusse, P., Malti, R., Melchior, P.: Crone control system design toolbox for the control engi-
neering community: tutorial and case study. Phil. Trans. R. Soc. A 371(1990), 20120149 (2013)
30. Yousfi, N., Melchior, P., Rekik, C., Derbel, N., Oustaloup, A.: Design of centralized CRONE
controller combined with MIMO-QFT approach applied to non-square multivariable systems.
Int. J. Comput. Appl. 45(16) (2012)
31. Tepljakov, A.: Fractional-Order Calculus Based Identification and Control of Linear Dynamic
Systems. Tallinn University of Technology (2011)
32. Sohal, J.S.: Improvement of artificial neural network based character recognition system, using
SciLab. Optik 127(22), 10510–10518 (2016)
33. Campbell, S.L., Chancelier, J.P., Nikoukhah, R.: Modeling and Simulation in SCILAB.
Springer, New York (2006)
34. Bunks, C., Chancelier, J.P., Delebecque, F., Goursat, M., Nikoukhah, R., Steer, S.: Engineering
and Scientific Computing with Scilab. Springer Science & Business Media (2012)
35. Wouwer, A.V., Saucez, P., Vilas, C.: Simulation of ode/pde Models With Matlab, Octave and
Scilab. Springer, Cham (2014)
36. Rohit, M.T., Ashish, M.K.: Digital Image Processing Using SCILAB. Springer, Cham (2018)
37. Ma, L., Xia, F., Peng, Z.: Integrated design and implementation of embedded control systems
with scilab. Sensors 8(9), 5501–5515 (2008)
38. Pendharkar, I.: Rltool for Scilab: a public domain tool for SISO system design. IEEE Control
Syst. Mag. 25(1), 23–25 (2005)
39. Magyar, Z., Žáková, K.: Scilab based remote control of experiments. IFAC Proc. Vol. 45(11),
206–211 (2012)
40. Landau, I.D., Zito, G.: Digital Control Systems: Design, Identification and Implementation.
Springer Science & Business Media (2007)
41. Bingi, K., Ibrahim, R., Karsiti, M.N., Hassan, S.M.: Fractional-order filter design for set-point
weighted PID controlled unstable systems. Int. J. Mech. Mechatron. Eng. 17(5), 173–179
(2017)
42. Bingi, K., Ibrahim, R., Karsiti, M.N., Hassan, S.M., Harindran, V.R.: Fractional order PI con-
trollers for real-time control of pressure plant. In: 2018 5th International Conference on Control,
Decision and Information Technologies, Thessaloniki, Greece, 10–13 Apr 2018
43. Luo, Y., Chen, Y.: Fractional order [proportional derivative] controller for a class of fractional
order systems. Automatica 45(10), 2446–2450 (2009)
44. Tavakoli-Kakhki, M., Haeri, M.: Fractional order model reduction approach based on retention
of the dominant dynamics: application in IMC based tuning of FOPI and FOPID controllers.
ISA Trans. 50(3), 432–442 (2011)
212 4 Scilab Based Toolbox for Fractional-order Systems and PID Controllers

45. Feliu-Batlle, V., Perez, R.R., Rodriguez, L.S.: Fractional robust control of main irrigation canals
with variable dynamic parameters. Control Eng. Pract. 15(6), 673–686 (2007)
46. Feliu-Batlle, V., Rivas-Perez, R., Castillo-Garcia, F.J.: Fractional order controller robust to
time delay variations for water distribution in an irrigation main canal pool. Comput. Electron.
Agric. 69(2), 185–197 (2009)
47. El-Khazali, R.: Fractional-order PIλ Dμ controller design. Comput. Math. Appl. 66(5), 639–646
(2013)
48. Latha, K., Rajinikanth, V., Surekha, P.M.: PSO-based PID controller design for a class of stable
and unstable systems. ISRN Artif. Intell. 2013 (2013)
Chapter 5
Scilab Based Toolbox for
Fractional-order Chaotic Systems

5.1 Introduction

Over the years the chaotic behavior of fractional-order nonlinear systems has been
utilized in many real-world applications such as engineering, finance, microbiology,
biology, physics, robotics, mathematics, economics, philosophy, meteorology, com-
puter science and civil engineering [1, 2]. From the investigation of researchers, it
was found that fractional-order chaotic systems possess less memory and display
more sophisticated dynamics compared to its integer-order systems. Recently the
attention of researcher’s shifts towards creating a chaotic system and its fractional-
order form with a more complicated topological structure becomes a desirable task
and many of the researchers have made a great contribution. For example, Duffing
introduced a duffing equation which can be extended to the complex domain to study
strange attractors and chaotic behavior of forced vibrations of industrial machinery
[3–5]. Similarly, Van der Pol introduced a model to study oscillations in vacuum tube
circuits. The Van der Pol oscillator represents a nonlinear system with an interesting
behavior that exhibits naturally in several applications, such as heartbeat, neurons,
and acoustic models [6–8].
On the other hand, Lorenz proposed the simplified equations of convection rolls
arising in the equations of the atmosphere. From the numerical solution of the equa-
tions, it can be observed that the chaotic behavior is known as the butterfly effect
which is very sensitive to initial conditions [9–11]. Similarly, Chen proposed a sim-
ple autonomous system which is similar but nonequivalent to the Lorenz’s attractor.
Furthermore, it should be noted that the attractors of Lorenz and Chen systems are
double scroll attractor [12]. An example of a single manifold chaotic system was pro-
posed by Rössler. The attractor has only one manifold which is useful in modeling the
equilibrium of chemical reactions [13, 14]. The dynamical behavior of these chaotic
systems together with many other chaotic systems has been analyzed in Matlab by
Petráš in [15–18]. The authors developed a Matlab toolbox named Fractional Order
Chaotic Systems which is very useful in analyzing the chaotic behavior of various

© Springer Nature Switzerland AG 2020 213

K. Bingi et al., Fractional-order Systems and PID Controllers,
Studies in Systems, Decision and Control 264,
https://doi.org/10.1007/978-3-030-33934-0_5
214 5 Scilab Based Toolbox for Fractional-order Chaotic Systems

fractional-order nonlinear systems. Therefore, in this work, an attempt is made to

develop and analyze the dynamic behavior of fractional-order chaotic systems using
the open-source software called Scilab. This is because Scilab is extremely effective
and available as a free open-source software package for scientific computations. It
includes hundreds of general and specialized functions for numerical computations,
2D and 3D visualization and data analysis [19, 20].

5.2 Fractional-order Chaotic Oscillators

This section presents the design and analysis of fractional-order Van der Pol and
duffing oscillator for various commensurate and non-commensurate orders using
Scilab.

5.2.1 Fractional-order Van der Pol Oscillator

The model of Van der Pol oscillator describing the oscillations of triode in electrical
circuit is defined as follows:

ẍ(t) + (x(t)2 − 1)ẋ(t) + x(t) = 0 (5.1)

where  > 0 is the control parameter and represents a relaxation oscillator. On the
other hand, if  = 0, the oscillator represents a simple linear oscillator.
The state-space representation of oscillator in Eq. (5.1) is defined as follows:

ẋ1 (t) = x2 (t)

(5.2)
ẋ2 (t) = −x1 (t) − (x1 (t)2 − 1)x2 (t)

The fractional-order form of the Van der Pol oscillator can be obtained by
fractional-ordering the classical state-space equation in Eq. (5.2) with parameters
α1 and α2 as follows:

Dαt 1 x1 (t) = x2 (t)

(5.3)
Dαt 2 x2 (t) = −x1 (t) − (x1 (t)2 − 1)x2 (t)

where α1 and α2 are the fractional-order parameters in (0, 2).

The solution for the fractional-order form of the Van der Pol oscillator can be
obtained using the Grünwald-Letnikov’s definition in Eqs. (4.2) and (4.3) as follows:
5.2 Fractional-order Chaotic Oscillators 215


k
x1 (tk ) = x2 (tk−1 )h α1 − c(α1)
j x 1 (tk− j )
j=1
(5.4)
  
k
x2 (tk ) = − x1 (tk ) − (x12 (tk ) − 1)x2 (tk−1 ) h α2
− c(α2)
j x 2 (tk− j )
j=1

Therefore, the function representing the fractional-order Van der Pol oscillator
using Scilab with syntax [x1,x2]=fracVdp(alpha,epsi,t,h,X) can be
written as follows:
function [x1,x2]=fracVdp(alpha,epsi,t,h,X)
n=round(t/h); x1(1)=X(1); x2(1)=X(2);
alpha1=alpha(1); alpha2=alpha(2);
c01=1;c02=1;
for j=1:n
c1(j)=(1-((1+alpha1)/j))*c01;
c2(j)=(1-((1+alpha2)/j))*c02;
c01=c1(j); c02=c2(j);
end
for j=2:n
x1(j)=x2(j-1)*hˆalpha1 - sumGL(x1,c1,j);
x2(j)=(-x1(j)-epsi*((x1(j)ˆ2)-1)...
*x2(j-1))*hˆalpha2 - sumGL(x2,c2,j);
end
endfunction

function [f]=sumGL(r,c,k)
temp=0;
for j=1:k-1
temp=temp+c(j)*r(k-j);
end
f=temp;
endfunction

In the above function, alpha is the fractional-order parameters, epsi is the con-
trol parameter, t is the simulation time, h is the step size and X=(X1(0),X2(0)) is
the initial condition. Furthermore, the coefficients c1 and c2 are calculated accord-
ing to the relations given in Eq. (4.3).
Thus, the phase portraits of fractional-order Van der Pol oscillator will be
obtained and analyzed for various values of fractional-order parameters and control
parameters using the developed Scilab function [x1,x2]=fracVdp(alpha,
epsi,t,h,X). Furthermore, in all the cases, the parameters of simulation time and
initial conditions used are t = 50, h = 0.005, X 1(0) = 1 and X 2(0) = 0. Therefore,
the phase portrait of fractional-order Van der Pol oscillator for various commen-
216 5 Scilab Based Toolbox for Fractional-order Chaotic Systems

−1

−2

−3

−4
−2 −1.5 −1 −0.5 0 0.5 1 1.5 2

Fig. 5.1 Phase portrait of fractional-order Van der Pol oscillator for various values of fractional-
order parameters

surate orders of fractional-order parameters α1 = α2 = 0.75, α1 = α2 = 0.85, and

α1 = α2 = 0.95 is obtained as follows:

– –> scf();
[x11,x21]=fracVanderpol([0.75,0.75],1,50,0.005,[1,0]);
[x12,x22]=fracVanderpol([0.85,0.85],1,50,0.005,[1,0]);
[x13,x23]=fracVanderpol([0.95,0.95],2,50,0.005,[1,0]);
plot(x11,x21,’b’,x12,x22,’-.r’,x13,x23,’--k’);
xlabel(’$x_1$’); ylabel(’$x_2$’);
legend([’$\alpha_1=\alpha_2=0.75$’,...
’$\alpha_1=\alpha_2=0.85$’,...
’$\alpha_1=\alpha_2=0.95$’],"in_upper_left")

Thus, the resultant phase portrait plot is shown in Fig. 5.1. Furthermore, the time
response of states x1 and x2 for the above condition is obtained using the following
code and the resultant plot is shown in Fig. 5.2.
5.2 Fractional-order Chaotic Oscillators 217

−1

−2
0 5 10 15 20 25 30 35 40 45 50 55

−2

−4
0 5 10 15 20 25 30 35 40 45 50 55

Fig. 5.2 Time response of fractional-order Van der Pol oscillator for various values of fractional-
order parameters

– –> scf(); t=0.005:0.005:50;

subplot(211)
plot(t,x11’,’b’,t,x12’,’-.r’,t,x13’,’--k’);
xlabel(’$t$’); ylabel(’$x_1$’);
subplot(212)
plot(t,x21’,’b’,t,x22’,’-.r’,t,x23’,’--k’);
xlabel(’$t$’); ylabel(’$x_2$’);
legend([’$\alpha_1=\alpha_2=0.75$’,...
’$\alpha_1=\alpha_2=0.85$’,...
’$\alpha_1=\alpha_2=0.95$’],"in_lower_left")
From the figures, it can be observed that the oscillator generates a stable limit
cycle when α1 , α2 increases from 0 to 1. On the other hand, it can also be observed
that the oscillator approaches towards the equilibrium point (0, 0) when α1 , α2 varies
from 1 to 0.
In a similar way, the phase portrait of fractional-order Van der Pol oscillator for
α1 = α2 = 0.5 and various values of control parameters  = 1, 2, 3 and 4 is obtained
as follows:
218 5 Scilab Based Toolbox for Fractional-order Chaotic Systems

– –> exec(’fracVdp.sci’); scf();

[x11,x21]=fracVdp([0.5,0.5],1,50,0.005,[1,0]);
[x12,x22]=fracVdp([0.5,0.5],2,50,0.005,[1,0]);
[x13,x23]=fracVdp([0.5,0.5],3,50,0.005,[1,0]);
[x14,x24]=fracVdp([0.5,0.5],4,50,0.005,[1,0]);
plot(x11,x21,’b’,x12,x22,’:r’,x13,x23,’-.m’,...
x14,x24,’--k’);
xlabel(’$x_1$’); ylabel(’$x_2$’);
legend([’$\epsilon=1$’,’$\epsilon=2$’,...
’$\epsilon=3$’,’$\epsilon=4$’],"in_upper_left")
On the other hand, the time response of the states x1 and x2 for this condition
of α1 = α2 = 0.5 and various values of control parameters  = 1, 2, 3 and 4 is
obtained using as follows:

– –> scf(); t=0.005:0.005:50;

subplot(211)
plot(t,x11’,’b’,t,x12’,’:r’,t,x13’,’-.m’,t,x14’,’--k’);
xlabel(’$t$’); ylabel(’$x_1$’);
subplot(212)
plot(t,x21’,’b’,t,x22’,’:r’,t,x23’,’-.m’,t,x24’,’--k’);
xlabel(’$t$’); ylabel(’$x_2$’)
legend([’$\epsilon=1$’,’$\epsilon=2$’,...
’$\epsilon=3$’,’$\epsilon=4$’],"in_lower_left")
Therefore, the resultant phase portrait and the time response plots of the states from
the above commands are given in Figs. 5.3 and 5.4 respectively. From the figures, a
similar observation of stable limit cycle has been observed when  increases from 0
to 1. Also, in this case, the oscillator approaches towards the equilibrium point (0, 0)
when  varies from 1 to 0.
Similarly, to analyze the phase portraits of the oscillator for commensurate and
non-commensurate orders, the oscillator has been simulated for commensurate and
non-commensurate fractional-order parameters α1 = α2 = 0.9 and α1 = 1.2, α2 =
0.8 using the following commands:

– –> [x11,x21]=fracVdp([0.9,0.9],1,50,0.005,[1,0]);
subplot(121); plot(x11,x21);
xlabel(’$x_1$’); ylabel(’$x_2$’); title(’A’);
legend([’$\alpha_1=\alpha_2=0.9$’])
[x12,x22]=fracVdp([1.2,0.8],1,50,0.005,[1,0]);
subplot(122); plot(x12,x22);
xlabel(’$x_1$’); ylabel(’$x_2$’); title(’B’);
legend([’$\alpha_1=1.2,\alpha_2=0.8$’])
Thus, the resultant phase portrait plots of commensurate and non-commensurate
orders are shown in Figs. 5.5 and 5.6 respectively. From the figures, it can be observed
that for orders less than 1, a stable limit cycle has been observed whereas, for orders
5.2 Fractional-order Chaotic Oscillators 219

−1

−2

−3

−4
−1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1

Fig. 5.3 Phase portrait of fractional-order Van der Pol oscillator for various values of control
parameter

greater than 1, some strange nature has been found in the outer areas which may be
described as bad bands.

5.2.2 Fractional-order Duffing Oscillator

The duffing oscillator is a model of a periodically forced steel beam which is deflected
toward the two magnets as shown in Fig. 5.7. From the figure, it can be seen that the
model is an example of a periodically forced oscillator with a nonlinear elasticity
which can be written as

d2x dx  
m +δ + βx(t) + αx 3 (t) = γ cos(ωt) (5.5)
dt 2 dt
where
• m is the mass of the ball,
• δ ≥ 0 is the damping constant,
• γ ≥ 0 is the amplitude of external periodic force,
220 5 Scilab Based Toolbox for Fractional-order Chaotic Systems

0.5

−0.5

−1
0 5 10 15 20 25 30 35 40 45 50 55

−2

−4
0 5 10 15 20 25 30 35 40 45 50 55

Fig. 5.4 Time response of fractional-order Van der Pol oscillator for various values of control
parameter

• ω is thefrequency of the periodic force,

• F = − βx(t) + αx 3 (t) is the restoring force of the rod.
From the restoring force (F), it should be noted that if α > 0, then the spring is
hardening spring. On the other hand, for α < 0 then the spring is softening spring.
Furthermore, for β < 0, then the Duffing oscillator can be regarded as a model of
periodically forced steel rod which is deflected towards the two magnets.
From Eq. (5.5), the state-space formulation of duffing oscillator is given as follows:

ẋ1 (t) = x2 (t)

(5.6)
ẋ2 (t) = −δx2 (t) − βx1 (t) − αx13 (t) + γ cos(ωt)

In a similar way to fractional-order Van der Pol oscillator, the fractional-order

form of the duffing oscillator is obtained by fractional-ordering the classical states
with parameters α1 and α2 as follows:
5.2 Fractional-order Chaotic Oscillators 221

2.5

1.5

0.5

−0.5

−1

−1.5

−2

−2.5
−2 −1.5 −1 −0.5 0 0.5 1 1.5 2

Fig. 5.5 Phase portrait of fractional-order Van der Pol oscillator for α1 = α2 = 0.9

Dαt 1 x1 (t) = x2 (t)

(5.7)
Dαt 2 x2 (t) = −δx2 (t) − βx1 (t) − αx13 (t) + γ cos(ωt)

The solution for the fractional-order duffing oscillator can be obtained using the
Grünwald-Letnikov’s definition in Eqs. (4.2) and (4.3) as follows:


k
x1 (tk ) = x2 (tk−1 )h α1 − c(α1)
j x 1 (tk− j )
j=1
  
k
x2 (tk ) = − δx2 (tk−1 ) − βx1 (tk ) − αx13 (tk ) + γ cos(ωt) h α2 − c(α2)
j x 2 (tk− j )
j=1
(5.8)
Therefore, the function representing the fractional-order duffing oscillator using
Scilab with syntax [x1,x2]=fracDuff(alpha,par,t,h,X) can be written
as follows:
function [x1,x2]=fracDuff(alpha,par,t,h,X)
n=round(t/h); x1(1)=X(1); x2(1)=X(2);
d=par(1); b=par(2); a=par(3); g=par(4);
w=par(5); alpha1=alpha(1); alpha2=alpha(2);
222 5 Scilab Based Toolbox for Fractional-order Chaotic Systems

2.5

1.5

0.5

−0.5

−1

−1.5

−2

−2.5

−3
−3 −2.5 −2 −1.5 −1 −0.5 0 0.5 1 1.5 2 2.5 3

Fig. 5.6 Phase portrait of fractional-order Van der Pol oscillator for α1 = 1.2, α2 = 0.8

External periodic Force

Light Flexible Rod

Steel Ball
M
X

N N

Magnet Magnet

S S

Fig. 5.7 Duffing oscillator model

5.2 Fractional-order Chaotic Oscillators 223

c01=1;c02=1;
for j=1:n
c1(j)=(1-((1+alpha1)/j))*c01;
c2(j)=(1-((1+alpha2)/j))*c02;
c01=c1(j); c02=c2(j);
end
for j=2:n
x1(j)=x2(j-1)*hˆalpha1 - sumGL(x1,c1,j);
x2(j)=(-(d*x2(j-1))-(b*x1(j))...
-(a*x1(j)ˆ3)+(g*cos(w*h*j)))*hˆalpha2...
-sumGL(x2,c2,j);
end
endfunction

function [f]=sumGL(r,c,k)
temp=0;
for j=1:k-1
temp=temp+c(j)*r(k-j);
end
f=temp;
endfunction

In the above function, alpha is the fractional-order parameters, par denotes the
system parameters (i.e., δ, β, α, γ and ω), t is the simulation time, h is the step size
and X=(X1(0),X2(0)) is the initial condition. Furthermore, the coefficients c1
and c2 are calculated according to the relations given in Eq. (4.3).
Thus, the phase portraits of fractional-order duffing oscillator will be obtained
and analyzed for commensurate and non-commensurate orders of fractional-order
parameters using the developed Scilab code presented in Sect. 5.2.2. Furthermore, in
all the cases, the parameters of oscillator, simulation time and initial conditions used
are δ = 0.15, β = −1, α = 1, γ = 0.3, ω = 1, t = 200, h = 0.05, X 1(0) = 0.21 and
X 2(0) = 0.13. Therefore, the phase portrait of fractional-order duffing oscillator for
commensurate orders of fractional-order parameters α1 = α2 = 0.95 is obtained as
follows:
– –> exec(’fracDuff.sci’); scf();
alpha=[0.95 0.95]; par=[0.15 -1 1 0.3 1];
[x1,x2]=fracDuff(alpha,par,200,0.05,[0.21 0.13]);
plot(x1,x2); xlabel(’$x_1$’); ylabel(’$x_2$’);
Similarly, the time response of the states x1 and x2 for commensurate orders of
fractional-order parameters α1 = α2 = 0.95 is obtained using the following com-
mands:
– –> t=0.05:0.05:200; subplot(211); plot(t,x1);
xlabel(’$t$’); ylabel(’$x_1$’); subplot(212);
plot(t,x2); xlabel(’$t$’); ylabel(’$x_2$’);
224 5 Scilab Based Toolbox for Fractional-order Chaotic Systems

0.8

0.6

0.4

0.2

−0.2

−0.4

−0.6

−0.8

−1
−1.5 −1 −0.5 0 0.5 1 1.5

Fig. 5.8 Phase portraits of fractional-order duffing oscillator for commensurate order α1 = α2 =
0.95

Therefore, the resultant phase portrait and time response of fractional-order duff-
ing oscillator for the commensurate orders is given in Figs. 5.8 and 5.9 respectively.
From the figures, it can be seen very clearly that the oscillator produced a stable limit
cycle around the three equilibrium points of the system. This can be seen more clearly
when the system will be simulated for the integer-orders α1 = α2 = 1. Therefore,
the phase portrait for α1 = α2 = 1 shown in Fig. 5.10 is obtained using the following
commands:

– –> exec(’fracDuff.sci’);
scf(); alpha=[1 1]; par=[0.15 -1 1 0.3 1];
[x1,x2]=fracDuff(alpha,par,200,0.05,[0.21 0.13]);
plot(x1,x2); xlabel(’$x_1$’); ylabel(’$x_2$’);
Similarly, the phase portrait and time response of fractional-order duffing oscil-
lator for non-commensurate orders of fractional-order parameters α1 = 0.9, α2 = 1
are obtained using the following commands.
5.2 Fractional-order Chaotic Oscillators 225

1.5

0.5

−0.5

−1

−1.5
0 20 40 60 80 100 120 140 160 180 200 220

0.5

−0.5

−1
0 20 40 60 80 100 120 140 160 180 200 220

Fig. 5.9 Time response of fractional-order duffing oscillator for commensurate order α1 = α2 =
0.95

– –> exec(’fracDuff.sci’); scf();

alpha=[0.9 1.0]; pa=[0.15 -1 1 0.3 1];
[x1,x2]=fracDuff(alpha,pa,200,0.05,[0.21 0.13]);
plot(x1,x2); xlabel(’$x_1$’); ylabel(’$x_2$’);
scf(); subplot(211); plot(t,x1);
xlabel(’$t$’); ylabel(’$x_1$’);
subplot(212); plot(t,x2);
xlabel(’$t$’); ylabel(’$x_2$’);
Therefore, the resultants phase portrait and time response plots of fractional-
order duffing oscillator for non-commensurate orders α1 = 0.9, α2 = 1 are given in
Figs. 5.11 and 5.12 respectively. From the figures, a similar observation of stable
limit cycle has been observed. However, the oscillator approaches towards the two
equilibrium point only. This is because the system having one unstable equilibrium
which satisfies the chaotic behavior of the system. For a dynamical system to be
classified as chaotic, it must have the following properties:
• must be period infinity orbit or a periodic orbit,
• must be bounded and
• must be sensitive to initial conditions.
226 5 Scilab Based Toolbox for Fractional-order Chaotic Systems

1.2

0.8

0.6

0.4

0.2

−0.2

−0.4

−0.6

−0.8

−1

−1.2
−2 −1.5 −1 −0.5 0 0.5 1 1.5 2

Fig. 5.10 Phase portraits of fractional-order duffing oscillator for order α1 = α2 = 1

From Figs. 5.11 and 5.12, it can be observed that the phase portrait is periodic and
has been bounded. Furthermore, to test the oscillator for sensitive to initial condi-
tions, the oscillator is simulated for two sets of initial conditions with very minimal
difference i.e., (0.21, 0.13) and (0.211, 0.13). Therefore, the resultant phase portrait
shown in Figs. 5.13 and 5.14 respectively can be obtained as follows:
– –> exec(’fracDuff.sci’); alp=[0.9 1.0]; t=200;
h=0.05; init=[0.211 0.13]; par=[0.15 -1 1 0.3 1];
[x1,x2]=fracDuff(alp,par,t,h,[0.21 0.13]);
[x11,x21]=fracDuff(alp,par,t,h,init);
plot(x1,x2,’b’,x11,x21,’r’);
xlabel(’$x_1$’); ylabel(’$x_2$’);
legend([’[0.21 0.13]’,’[0.211 0.13]’])

scf(); t=0.05:0.05:200;
subplot(211); plot(t,x1,’b’,t,x11,’r’);
xlabel(’$t$’); ylabel(’$x_1$’);
subplot(212); plot(t,x2,’b’,t,x21,’r’);
xlabel(’$t$’); ylabel(’$x_2$’);
legend([’[0.21 0.13]’,’[0.211 0.13]’])
5.2 Fractional-order Chaotic Oscillators 227

0.8

0.6

0.4

0.2

−0.2

−0.4

−0.6

−0.8

−1
−1.5 −1 −0.5 0 0.5 1 1.5

Fig. 5.11 Phase portraits of fractional-order duffing oscillator for non-commensurate order α1 =
0.9, α2 = 1.0

From the figure, it can be observed that the oscillator is very sensitive to initial
conditions. Here, a small change in the initial value of the current trajectory has been
lead to a significant change in future behavior. Therefore, the system can be classified
under a chaotic system.

5.3 Fractional-order Lorenz’s Chaotic System

The fractional-order form of the Lorenz’s chaotic system is defined as follows:

 
Dαt 1 x(t) = p y(t) − x(t)
 
Dαt 2 y(t) = x(t) q − z(t) − y(t) (5.9)
Dαt 3 z(t) = x(t)y(t) − r z(t)

where
• α1 , α2 and α3 are the fractional derivative orders,
• p is the Prandtl number,
228 5 Scilab Based Toolbox for Fractional-order Chaotic Systems

1.5

0.5

−0.5

−1

−1.5
0 20 40 60 80 100 120 140 160 180 200 220

0.5

−0.5

−1
0 20 40 60 80 100 120 140 160 180 200 220

Fig. 5.12 Time response of fractional-order duffing oscillator for non-commensurate order α1 =
0.9, α2 = 1.0

• q is the Rayleigh number and

• r is the physical dimension of the layer.
Therefore, the numerical solution for the fractional-order Lorenz’s chaotic system
can be obtained using the Grünwald-Letnikov’s definition in Eqs. (4.2) and (4.3) as
follows:

   
k
x(tk ) = p y(tk−1 ) − x(tk−1 ) h α1 − c(α1)
j x(tk− j )
j=1

    
k
y(tk ) = x(tk ) q − z(tk−1 ) − y(tk−1 ) h α2 − c(α2)
j y(tk− j ) (5.10)
j=1

  
k
z(tk ) = x(tk )y(tk ) − r z(tk−1 ) h α3 − c(α3)
j z(tk− j )
j=1
5.3 Fractional-order Lorenz’s Chaotic System 229

1
[0.21 0.13]
0.8 [0.211 0.13]

0.6

0.4

0.2

−0.2

−0.4

−0.6

−0.8

−1
−1.5 −1 −0.5 0 0.5 1 1.5

Fig. 5.13 Phase portrait of fractional-order duffing oscillator for sensitivity to initial conditions

Thus, the function representing the fractional-order Lorenz’s chaotic system using
Scilab with syntax [x1,x2,x3]=fracLorenz(alpha,param,t,h,X) can
be written as follows:
function [x1,x2,x3]=fracLorenz(alpha,param,t,h,X)
n=round(t/h); x1(1)=X(1); x2(1)=X(2); x3(1)=X(3);
alpha1=alpha(1);alpha2=alpha(2); alpha3=alpha(3);
p=param(1);q=param(2);r=param(3);
c01=1; c02=1; c03=1;
for j=1:n
c1(j)=(1-((1+alpha1)/j))*c01;
c2(j)=(1-((1+alpha2)/j))*c02;
c3(j)=(1-((1+alpha3)/j))*c03;
c01=c1(j); c02=c2(j); c03=c3(j);
end
for j=2:n
x1(j)=(p*(x2(j-1)-x1(j-1)))*hˆalpha1-sumGL(x1,c1,j);
x2(j)=(x1(j)*(q-x3(j-1))-x2(j-1))*hˆalpha2-sumGL(x2,c2,j);
x3(j)=((x1(j)*x2(j))-(r*x3(j-1)))*hˆalpha3-sumGL(x3,c3,j);
end
endfunction
230 5 Scilab Based Toolbox for Fractional-order Chaotic Systems

1.5

0.5

−0.5

−1

−1.5
0 20 40 60 80 100 120 140 160 180 200 220

1
[0.21 0.13]
[0.211 0.13]
0.5

−0.5

−1
0 20 40 60 80 100 120 140 160 180 200 220

Fig. 5.14 Time response of fractional-order duffing oscillator for sensitivity to initial conditions

function [f]=sumGL(r,c,k)
temp=0;
for j=1:k-1
temp=temp+c(j)*r(k-j);
end
f=temp;
endfunction

In the above function, alpha is the fractional-order parameters (i.e., α1 , α2 and

α3 ), par denotes the oscillators parameters (i.e., p, q and r ), t is the simulation
time, h is the step size and X=(X1(0),X2(0),X3(0)) is the initial condition.
Furthermore, the coefficients c1, c2 and c3 are calculated according to the relations
given in Eq. (4.3).
5.3 Fractional-order Lorenz’s Chaotic System 231

50

45

40

35

30

25

20

15

10

0 −30
−20
−10
20 10 0
0 10
−10 20
−20
30

Fig. 5.15 Chaotic attractor of Lorenz’s system for commensurate orders of α1 = α2 = α3 = 0.995

5.3.1 Commensurate Orders

In this simulation study, the chaotic attractor of Lorenz’s system has been obtained for
the commensurate orders α1 = α2 = α3 = 0.995. Furthermore, the attractor has been
generated with system parameters p = 10, q = 28, r = 8/3 for time t = 100 and
step size h = 0.005 with initial conditions as x1 (0) = x2 (0) = x3 (0) = 0.1. Thus,
the Scilab commands for generating the chaotic attractor of Lorenz’s system are as
follows:
– –> exec(’fracLorenz.sci’);
[x1,x2,x3]=fracLorenz([0.995,0.995,0.995],...
[10,28,8/3],100,0.005,[0.1,0.1,0.1]);
param3d(x1,x2,x3);
xlabel(’$x_1$’);
ylabel(’$x_2$’);
zlabel(’$x_2$’);
Thus, the resultant chaotic attractor of Lorenz’s system is shown in Fig. 5.15.
Furthermore, the projections of the attractor on x − y, y − z and x − z planes shown
in Fig. 5.16a, b and c respectively can be obtained as follows:
232 5 Scilab Based Toolbox for Fractional-order Chaotic Systems

30

20

10

−10

−20

−30
−20 −15 −10 −5 0 5 10 15 20 25

50

40

30

20

10

0
−30 −25 −20 −15 −10 −5 0 5 10 15 20 25 30

Fig. 5.16 Chaotic attractor of Lorenz’s system projected on x − y, y − z and x − z planes

– –> scf();
subplot(311);
plot(x1,x2); xlabel(’$(a) x$’); ylabel(’$y$’);
subplot(312);
plot(x2,x3); xlabel(’$(b) y$’); ylabel(’$z$’);
subplot(313);
plot(x1,x3); xlabel(’$(c) x$’); ylabel(’$z$’);
On the other hand, the time response of the system for the states x, y and z shown
in Fig. 5.17 can be obtained as follows:
5.3 Fractional-order Lorenz’s Chaotic System 233

50

40

30

20

10

0
−20 −15 −10 −5 0 5 10 15 20 25

Fig. 5.16 (continued)

– –> scf(); t=0.005:0.005:100;

subplot(311);
plot(t,x1); xlabel(’$(a) t$’); ylabel(’$x$’);
subplot(312);
plot(t,x2); xlabel(’$(b) t$’); ylabel(’$y$’);
subplot(313);
plot(t,x3); xlabel(’$(c) t$’); ylabel(’$z$’);

5.3.2 Non-commensurate Orders

Similarly, the chaotic attractor of Lorenz’s system for the non-commensurate orders
of α1 = α2 = 1 and α3 = 0.995 with parameters p = 10, q = 28, r = 8/3, t = 100,
h = 0.005, x1 (0) = 0.1, x2 (0) = 0.1 and x3 (0) = 0.1 can be obtained as follows:
– –> scf(); exec(’fracLorenz.sci’);
[x1,x2,x3]=fracLorenz([1,1,0.99],...
[10,28,8/3],100,0.005,[0.1,0.1,0.1]);
subplot(221); param3d(x1,x2,x3);
xlabel(’$x$’); ylabel(’$y$’); zlabel(’$z$’);
subplot(223);
plot(x1,x2); xlabel(’$x$’); ylabel(’$y$’);
subplot(223);
plot(x2,x3); xlabel(’$y$’); ylabel(’$z$’);
subplot(224);
plot(x1,x3); xlabel(’$x$’); ylabel(’$z$’);
234 5 Scilab Based Toolbox for Fractional-order Chaotic Systems

20

10

−10

−20
0 10 20 30 40 50 60 70 80 90 100

30

20

10

−10

−20

−30
0 10 20 30 40 50 60 70 80 90 100

Fig. 5.17 Time responses of Lorenz’s system for the states x, y and z

Therefore, the resultant chaotic attractor together with the projections on x − y,

y − z and x − z planes is shown in Fig. 5.18.
Furthermore, the time response of the system shown in Fig. 5.19 for the states x,
y and z can be obtained as follows:
– –> scf(); t=0.005:0.005:100;
subplot(311);
plot(t,x1); xlabel(’$(a) t$’); ylabel(’$x$’);
subplot(312);
plot(t,x2); xlabel(’$(b) t$’); ylabel(’$y$’);
subplot(313);
plot(t,x3); xlabel(’$(c) t$’); ylabel(’$z$’);
5.3 Fractional-order Lorenz’s Chaotic System 235

50

40

30

20

10

0
0 10 20 30 40 50 60 70 80 90 100

Fig. 5.17 (continued)

30

50
20

40 10

30 0

20 −10

10 −20

0 −30
−20
20 10 0 −20 −10 0 10 20
0 −10 −20 20

50 50

40 40

30 30

20 20

10 10

0 0
−30 −20 −10 0 10 20 30 −20 −15 −10 −5 0 5 10 15 20

Fig. 5.18 Chaotic attractor of Lorenz’s system for non-commensurate orders of α1 = α2 = 1 and
α3 = 0.995
236 5 Scilab Based Toolbox for Fractional-order Chaotic Systems

20

15

10

−5

−10

−15

−20
0 10 20 30 40 50 60 70 80 90 100

30

20

10

−10

−20

−30
0 10 20 30 40 50 60 70 80 90 100

Fig. 5.19 Time responses of Lorenz’s system for the states x, y and z

From the chaotic attractors shown in Figs. 5.15, 5.16 and 5.18, it can be observed
that the chaotic behavior is known as butterfly effect which is very sensitive to initial
conditions.
5.4 Fractional-order Chen’s Chaotic System 237

50

40

30

20

10

0
0 10 20 30 40 50 60 70 80 90 100

Fig. 5.19 (continued)

5.4 Fractional-order Chen’s Chaotic System

The fractional-order form of the chaotic system proposed by the Chen is described
as follows:  
Dαt 1 x(t) = p y(t) − x(t)
Dαt 2 y(t) = −(r − p)x(t) − x(t)z(t) + r y(t) (5.11)
Dαt 3 z(t) = x(t)y(t) − qz(t)

where
• α1 , α2 and α3 are the fractional derivative orders and
• p, q and r are the system parameters
Therefore, the numerical solution for the fractional-order Chen’s chaotic system
can be obtained using the Grünwald-Letnikov’s definition in Eqs. (4.2) and (4.3) as
follows:

  
k
x(tk ) = p y(tk ) − x(tk ) h α1 − c(α1)
j x(tk− j )
j=1

  
k
y(tk ) = − (r − p)x(tk ) − x(tk )z(tk−1 ) + r y(tk−1 ) h α2 − c(α
j
2)
y(tk− j )
j=1

  
k
z(tk ) = x(tk )y(tk ) − qz(tk−1 ) h α3 − c(α3)
j z(tk− j )
j=1
(5.12)
238 5 Scilab Based Toolbox for Fractional-order Chaotic Systems

Thus, the function representing the fractional-order Chen’s chaotic system using
Scilab with syntax [x1,x2,x3]=fracChen(alpha,param,t,h,X) can be
written as follows:
function [x1,x2,x3]=fracChen(alpha,param,t,h,X)
n=round(t/h); x1(1)=X(1); x2(1)=X(2); x3(1)=X(3);
alpha1=alpha(1);alpha2=alpha(2); alpha3=alpha(3);
p=param(1);q=param(2);r=param(3);
c01=1; c02=1; c03=1;
for j=1:n
c1(j)=(1-((1+alpha1)/j))*c01;
c2(j)=(1-((1+alpha2)/j))*c02;
c3(j)=(1-((1+alpha3)/j))*c03;
c01=c1(j); c02=c2(j); c03=c3(j);
end
for j=2:n
x1(j)=(p*(x2(j-1)-x1(j-1)))*hˆalpha1-sumGL(x1,c1,j);
x2(j)=(-(r-p)*x1(j)-x1(j)*x3(j-1)+r*x2(j-1))*hˆalpha2...
-sumGL(x2,c2,j);
x3(j)=(x1(j)*x2(j)-q*x3(j-1))*hˆalpha3-sumGL(x3,c3,j);
end
endfunction

function [f]=sumGL(r,c,k)
temp=0;
for j=1:k-1
temp=temp+c(j)*r(k-j);
end
f=temp;
endfunction

Similar to fractional-order Lorenz’s function, here also, alpha denotes the

fractional-order parameters (i.e., α1 , α2 and α3 ), par denotes the oscillators param-
eters (i.e., p, q and r ), t is the simulation time, h is the step size and X=(X1(0),
X2(0),X3(0)) is the initial condition. Furthermore, the coefficients c1, c2 and
c3 are calculated according to the relations given in Eq. (4.3).

5.4.1 Commensurate Orders

This simulation study provides the chaotic attractor of fractional-order Chen’s system
for the commensurate orders of α1 = α2 = α3 = 0.9. Here, the parameters for the
simulation are used as p = 32, q = 3 and r = 28. Thus, the chaotic attractor for the
simulation time t = 100 and step size h = 0.005 can be obtained as follows:
5.4 Fractional-order Chen’s Chaotic System 239

40

60 20

0
40

−20
20

−40
−20 −40
20 0 0 −40 −30 −20 −10 0 10 20 30
−20 −40
40 20

70 70

60 60

50 50

40 40

30 30

20 20

10 10
−40 −30 −20 −10 0 10 20 30 40 −40 −30 −20 −10 0 10 20 30

Fig. 5.20 Chaotic attractor of Chen’s system for commensurate orders of α1 = α2 = α3 = 0.9

– –> scf(); exec(’fracChen.sci’);

[x1,x2,x3]=fracChen([0.9,0.9,0.9],[35,3,28],...
100,0.005,[-9,-5,14]);
subplot(221);
param3d(x1,x2,x3); xlabel(’$x$’);
ylabel(’$y$’); zlabel(’$z$’);
subplot(222);
plot(x1,x2); xlabel(’$x$’); ylabel(’$y$’);
subplot(223);
plot(x2,x3); xlabel(’$y$’); ylabel(’$z$’);
subplot(224);
plot(x1,x3); xlabel(’$x$’); ylabel(’$z$’);
Thus, the resultant chaotic attractor of fractional-order Chen’s system for the
commensurate orders from the above commands is shown in Fig. 5.20. Furthermore,
the figure also gives the projections of the attractor on x − y, y − z, and x − z planes.
240 5 Scilab Based Toolbox for Fractional-order Chaotic Systems

80 40

60 20

0
40

−20
20

−40
40 20 0 0 −40 −20 0 20 40
−20 −40
50

80 80

60 60

40 40

20 20

−40 −20 0 20 40 −40 −30 −20 −10 0 10 20 30 40

Fig. 5.21 Chaotic attractor of Chen’s system for non-commensurate orders of α1 = 0.995, α2 = 1.0
and α3 = 1.005

5.4.2 Non-commensurate Orders

Similarly, the chaotic attractor of fractional-order Chen’s system for the non-
commensurate orders of α1 = 0.995, α2 = 1.0 and α3 = 1.005 with system param-
eters p = 32, q = 3 and r = 28 with simulation parameters t = 100 and h = 0.005
shown in Fig. 5.21 can be obtained as follows:
– –> scf(); exec(’fracChen.sci’);
[x1,x2,x3]=fracChen([0.995,1.0,1.005],[35,3,28],...
100,0.005,[-9,-5,14]);
subplot(221);
param3d(x1,x2,x3); xlabel(’$x$’);
ylabel(’$y$’); zlabel(’$z$’);
subplot(222);
plot(x1,x2); xlabel(’$x$’); ylabel(’$y$’);
subplot(223);
plot(x2,x3); xlabel(’$y$’); ylabel(’$z$’);
subplot(224);
plot(x1,x3); xlabel(’$x$’); ylabel(’$z$’);
5.4 Fractional-order Chen’s Chaotic System 241

From the chaotic attractors given in Figs. 5.20 and 5.21, it can be observed that
the behavior of Chen’s system is similar but nonequivalent to the Lorenz’s attractor.
Furthermore, it should be noted that the attractors of Lorenz and Chen systems are
double scroll attractor. An example of a single manifold chaotic system will be
discussed in the next section.

5.5 Fractional-order Rössler’s Chaotic System

The fractional-order form of the Rössler’s chaotic system which is useful in modeling
the chemical equilibrium reactions that has only one manifold is described as follows:
 
Dαt 1 x(t) = − y(t) − z(t)
Dαt 2 y(t) = x(t) + py(t) (5.13)
 
Dαt 3 z(t) = q + z(t) x(t) − r

where
• α1 , α2 and α3 are the fractional derivative orders and
• a, b and b are the system parameters
Therefore, the numerical solution for the fractional-order Lorenz’s chaotic oscil-
lator can be obtained using the Grünwald-Letnikov’s definition in Eqs. (4.2) and (4.3)
as follows:

  
k
x(tk ) = − y(tk−1 ) − z(tk−1 ) h α1 − c(α1)
j x(tk− j )
j=1

  
k
y(tk ) = x(tk ) + py(tk−1 ) h α2 − c(α
j
2)
y(tk− j ) (5.14)
j=1

   α3 
k
z(tk ) = q + z(tk−1 ) x(tk ) − r h − c(α3)
j z(tk− j )
j=1

Thus, the function representing the fractional-order Rössler’s chaotic system using
Scilab with syntax [x1,x2,x3]=fracRossler(alpha,param,t,h,X) can
be written as follows:
function [x1,x2,x3]=fracRossler(alpha,param,t,h,X)
n=round(t/h); x1(1)=X(1); x2(1)=X(2); x3(1)=X(3);
alpha1=alpha(1); alpha2=alpha(2); alpha3=alpha(3);
p=param(1); q=param(2); r=param(3);
c01=1; c02=1; c03=1;
for j=1:n
242 5 Scilab Based Toolbox for Fractional-order Chaotic Systems

c1(j)=(1-((1+alpha1)/j))*c01;
c2(j)=(1-((1+alpha2)/j))*c02;
c3(j)=(1-((1+alpha3)/j))*c03;
c01=c1(j); c02=c2(j); c03=c3(j);
end
for j=2:n
x1(j)=(-(x2(j-1)+x3(j-1)))*hˆalpha1-sumGL(x1,c1,j);
x2(j)=(x1(j)+(p*x2(j-1)))*hˆalpha2-sumGL(x2,c2,j);
x3(j)=(q+x3(j-1)*(x1(j)-r))*hˆalpha3-sumGL(x3,c3,j);
end
endfunction

function [f]=sumGL(r,c,k)
temp=0;
for j=1:k-1
temp=temp+c(j)*r(k-j);
end
f=temp;
endfunction

5.5.1 Commensurate Orders

This simulation study provided the chaotic attractor of fractional-order Rössler’s

system for commensurate orders α1 = α2 = α3 = 0.9 with system parameters p =
0.5, q = 0.2 and r = 10. Thus, the chaotic attractor of the system shown in Fig. 5.22
for the time t = 120 and h = 0.005 with initial conditions x(0) = 0.5, y(0) = 1.5
and z(0) = 0.1 can be obtained as follows

– –> scf(); exec(’fracRossler.sci’);

[x1,x2,x3]=fracRossler([0.9,0.9,0.9],[0.5,0.2,10],...
120,0.005,[0.5,1.5,0.1]);
param3d(x1,x2,x3);
xlabel(’$x(t)$’); ylabel(’$y(t)$’); zlabel(’$z(t)$’);
title(’Fractional-order Rossler’s system’);

5.5.2 Non-commensurate Orders

In a similar way, the chaotic attractor of fractional-order Rössler’s system for non-
commensurate orders α1 = 0.9, α2 = 0.85 and α3 = 0.95 with parameters p = 0.5,
5.5 Fractional-order Rössler’s Chaotic System 243

Fractional−order Rossler System

45

40

35

30

25
z(t)

20

15

10

0
20
15 15
10 10
5 5
0 0
−5
−10 −5
−15 −10 x(t)
y(t) −15
−20

Fig. 5.22 Chaotic attractor of Rössler’s system for commensurate orders of α1 = α2 = α3 = 0.9

q = 0.2, r = 10, t = 120, h = 0.005, x(0) = 0.5, y(0) = 1.5 and z(0) = 0.1 can
be obtained as follows:
– –> scf(); exec(’fracRossler.sci’);
[x1,x2,x3]=fracRossler([0.9,0.85,0.95],[0.5,0.2,10],...
120,[0.5,1.5,0.1]);
param3d(x1,x2,x3);
xlabel(’$x(t)$’); ylabel(’$y(t)$’); zlabel(’$z(t)$’);
title(’Fractional-order Rossler’s system’);
Thus, the resultant chaotic attractor of the fractional-order Rössler’s system is
shown in Fig. 5.23. From the figure, it can be noted that the attractor is similar to the
Lorenz’s attractor, but is simpler and has only one manifold.
The summary of all the developed functions for the fractional-order chaotic sys-
tems in this toolbox including their Scilab syntaxes is given in Table 5.1.
244 5 Scilab Based Toolbox for Fractional-order Chaotic Systems

Fractional−order Rossler System

20

15
z(t)

10

0
15 20
10 15
5 10
0 5
−5 0
−10 −5
y(t) −15 −10 x(t)
−20
−15

Fig. 5.23 Chaotic attractor of Rössler’s system for non-commensurate orders of α1 = 0.9, α2 =
0.85 and α3 = 0.95

Table 5.1 List of functions developed in Scilab based toolbox for fractional-order chaotic systems

Function Syntax
Fractional-order Van der Pol Oscillator [x1,x2]=fracVanderpol(alpha,epsi,t,h,X)
Fractional-order Duffing Oscillator [x1,x2]=fracDuff(alpha,param,t,h,X)
Fractional-order Lorenz’s System [x1,x2,x3]=fracLorenz(alpha,param,t,h,X)
Fractional-order Chen’s System [x1,x2,x3]=fracChen(alpha,param,t,h,X)
Fractional-order Rössler’s system [x1,x2,x3]=fracRossler(alpha,param,t,h,X)

5.6 Summary

In this work, a Scilab based toolbox for fractional-order chaotic systems is devel-
oped. The proposed toolbox is the first open-source toolbox in open-source soft-
ware. The toolbox includes fractional-order Van der Pol and duffing oscillators and
fractional-order Lorenz, Chen and Rössler’s systems. Using the developed func-
tions, the dynamic behavior of these chaotic systems has been analyzed effectively
5.6 Summary 245

for various commensurate and non-commensurate orders using Scilab. However, it

should be noted that the execution time of Scilab is slower than Matlab which can
be observed while running these complex fractional-order chaotic systems [21, 22].

References

1. Zhang, W., Liao, S.K., Shimizu, N.: Dynamic behaviors of nonlinear fractional-order differ-
ential oscillator. J. Mech. Sci. Technol. 23(4), 1058–1064 (2009)
2. Elwakil, A.S.: Fractional-order circuits and systems: an emerging interdisciplinary research
area. IEEE Circuits Syst. Mag. 10(4), 40–50 (2010)
3. Ge, Z.M., Ou, C.Y.: Chaos in a fractional order modified Duffing system. Chaos Solitons
Fractals 34(2), 262–291 (2007)
4. Shen, Y., Yang, S., Xing, H., Gao, G.: Primary resonance of Duffing oscillator with fractional-
order derivative. Commun. Nonlinear Sci. Numer. Simul. 17(7), 3092–3100 (2012)
5. Baleanu, D., Machado, J.A.T., Luo, A.C.: Fractional Dynamics and Control. Springer Science
& Business Media (2011)
6. Shen, Y., Yang, S., Sui, C.: Analysis on limit cycle of fractional-order van der Pol oscillator.
Chaos Solitons Fractals 67, 94–102 (2014)
7. Matouk, A.E.: Chaos, feedback control and synchronization of a fractional-order modified
autonomous Van der Pol-Duffing circuit. Commun. Nonlinear Sci. Numer. Simul. 16(2), 975–
986 (2011)
8. Tavazoei, M.S., Haeri, M., Attari, M., Bolouki, S., Siami, M.: More details on analysis of
fractional-order van der Pol oscillator. J. Vib. Control 15(6), 803–819 (2009)
9. Grigorenko, I., Grigorenko, E.: Chaotic dynamics of the fractional Lorenz system. Phys. Rev.
Lett. 91(3), 034101 (2003)
10. Wu, X.J., Shen, S.L.: Chaos in the fractional-order Lorenz system. Int. J. Comput. Math. 86(7),
1274–1282 (2009)
11. Munmuangsaen, B., Srisuchinwong, B.: A hidden chaotic attractor in the classical Lorenz
system. Chaos Solitons Fractals 107, 61–66 (2018)
12. Li, C., Peng, G.: Chaos in Chen’s system with a fractional order. Chaos Solitons Fractals 22(2),
443–450 (2004)
13. Zhang, W., Zhou, S., Li, H., Zhu, H.: Chaos in a fractional-order Rössler system. Chaos Solitons
Fractals 42(3), 1684–1691 (2009)
14. Li, C., Chen, G.: Chaos and hyperchaos in the fractional-order Rössler equations. Physica A
341, 55–61 (2004)
15. Petráš, I.: Fractional-Order Nonlinear Systems: Modeling, Analysis and Simulation. Springer
Science & Business Media (2011)
16. Petráš, I.: A note on the fractional-order Volta’s system. Commun. Nonlinear Sci. Numer.
Simul. 15(2), 384–393 (2010)
17. Petráš, I.: A note on the fractional-order Chua’s system. Chaos Solitons Fractals 38(1), 140–147
(2008)
18. Petráš, I.: Stability of fractional order systems with rational orders: a survey. Fract. Calc. Appl.
Anal. 12(3), 269–298 (2009)
19. Campbell, S.L., Chancelier, J.P., Nikoukhah, R.: Modeling and Simulation in SCILAB.
Springer, New York (2006)
20. Bunks, C., Chancelier, J.P., Delebecque, F., Goursat, M., Nikoukhah, R., Steer, S.: Engineering
and Scientific Computing with Scilab. Springer Science & Business Media (2012)
21. Sharma, N., Gobbert, M.K.: A Comparative Evaluation of Matlab, FreeMat, and Scilab for
Research and Teaching. UMBC Faculty Collection, Octave (2010)
22. Bordeianu, C.C., Besliu, C., Jipa, A., Felea, D., Grossu, I.V.: Scilab software package for the
study of dynamical systems. Comput. Phys. Commun. 178(10), 788–793 (2008)
Appendix
Tuning of Controller Parameters Using
Accelerated Particle Swarm Optimization

A.1 Introduction

This chapter focus on tuning of PID parameters (K p , Ti & Td ), set-point weighting

parameter (b) and fractional-order parameters (λ & μ) of the designed SWPIλ Dμ and
PIλ -PDμ controllers presented in Sects. 2.2.3 and 2.3. First, the standard SWPIλ Dμ
controller parameters θcp = {K p , Ti , Td , b, λ & μ} presented in Sect. 2.2.1 will be
tuned using accelerated particle swarm optimization (APSO) algorithm. Then for
effective comparison, the controller parameters of industrial and parallel configu-
rations of SWPIλ Dμ will be derived using the conversion relations developed in
Sect. 2.2.7.
Furthermore, the PIλ -PDμ controller parameters θcp = {k p , ki , kd , λ & μ} are
derived from parallel SWPIλ Dμ presented in Sect. 2.3.4. The summary of all these
conversion relations developed in this research for the proposed fractional-order
controllers is given in Fig. A.1. Therefore, the tuning of standard SWPIλ Dμ con-
troller using APSO algorithm and the development of APSO from particle swarm
optimization (PSO) algorithm is explained in the succeeding part of the section.

A.2 Accelerated Particle Swarm Optimization

The standard particle swarm optimization (PSO) is a metaheuristic optimization algo-

rithm inspired by birds’ flocking behavior [1, 2]. At the beginning of the algorithm,
velocities and positions of a group of N birds (or particles) are initialized randomly
within the dimensional search space D. In the next stage, each particle is attracted
towards the current global best position G ∗ and its own best position St∗ based on
cost function J .
Therefore, the update of velocity and position equations for the PSO algorithm
are given as follows:
© Springer Nature Switzerland AG 2020 247
K. Bingi et al., Fractional-order Systems and PID Controllers,
Studies in Systems, Decision and Control 264,
https://doi.org/10.1007/978-3-030-33934-0
 248

Fig. A.1 Conversion of SWPIλ Dμ and PIλ -PDμ controller parameters

Appendix: Tuning of Controller Parameters Using Accelerated …
Appendix: Tuning of Controller Parameters Using Accelerated … 249

Vit+1 = Vit + αε1 (G ∗ − Sit ) + βε2 (Si∗ − Sit ) (A.1)

Sit+1 = Sit + Vit+1 (A.2)

where
• Si is the position of ith particle,
• Vi is the velocity of ith particle,
• ε1 and ε2 are random numbers N (0, 1) and
• α and β are acceleration constants.
The use of local best position S ∗ in (A.1) is to increase the diversity of the solutions.
However, using some randomness, this diversity can be simulated in the simplified
APSO algorithm. This simplified version accelerates the convergence of PSO using
global best only. Therefore, the velocity and position updates are achieved using the
following equations:

Vit+1 = Vit + αεn + β(G ∗ − Sit ) (A.3)

Sit+1 = (1 − β)Sit + βG it + αεn (A.4)

A further advancement to the APSO is to reduce the randomness. This can be

done by using a monotonically decreasing function as:

α = αo γ t (A.5)

where
• εn is the random number drawn from N (0, 1),
• α = 0.1L ∼ 0.5L is the acceleration constant of Si ,
• L is the scale of each variable,
• β = 0.1 ∼ 0.7 is the acceleration constant of search domain,
• αo = 0.5 ∼ 1 is the starting value of the level of randomness and
• 0 < γ < 1 is the control variable.
The pseudo-code for the accelerated particle swarm optimization (APSO) algo-
rithm is given as follows:
Require: N , bir d Step, D, αo , γ , β; Output: θcp ;
1: Initialization:
2: for i = 1 to N do
3: for d = 1 to D do
4: Sid = rand(d, i)
5: Vid = rand(d, i)
6: end for
7: εi = rand(i)
8: Pi = Si
9: if J (Pi ) < J (G) then
10: G ∗ = Pi
250 Appendix: Tuning of Controller Parameters Using Accelerated …

11: end if
12: end for
13: Iteration Steps:
14: repeat
15: for i = 1 to N do
16: if J (Si ) < J (Pi ) then
17: Pi = Si
18: end if
19: if J (Pi ) < J (G) then
20: G ∗ = Pi
21: end if
22: end for
23: for i = 1 to N do
24: for d = 1 to D do
25: Vid = Vid + αi + β(G ∗ − Sid )
26: Sid = (1 − β)Sid + βG ∗ + αεi
27: end for
28: end for
29: iter = iter + 1
30: α = αo γ iter
31: until iter > bir d Step
The block diagram of the SWPIλ Dμ controller based on the APSO algorithm for
tuning the controller parameters of standard SWPIλ Dμ controller with an objective
of minimizing error function (J ) is given in Fig. A.2. In the figure, θcp denotes the
controller parameters.
On the other hand, it is reported from the literature based on the previous works in
[3–5], the commonly used objective functions for optimization are error-based time
domain performance criteria. These are in the form of integral squared error (ISE),
integral time squared error (ITSE), mean squared error (MSE), integral absolute error
(IAE), integral time absolute error (ITAE). Furthermore, it should be noted that the
integral criteria such as ISE, IAE, ITAE and ITSE often result in a better closed-loop
response of a control system.
The ISE criteria should be used where there is a large deviation of error from
the set-point. In the case of small deviations, IAE criterion is used. On the other
hand, when the error persists for a long time, ITAE criterion helps. This is because
multiplying the error with time to augment the error function (J ) term at high values.
Therefore, in this work for the purpose of optimization, the performance indices used
are the integral absolute error (IAE), integral squared error (ISE) and integral time
absolute error (ITAE) as given in the following equations:
 T
JI AE = |e(τ )|dτ (A.6)
0
Appendix: Tuning of Controller Parameters Using Accelerated … 251

Fig. A.2 Block diagram for tuning of SWPIλ Dμ controller

 T
JI S E = e(τ )2 dτ (A.7)
0

 T
JI T AE = τ |e(τ )|dτ (A.8)
0

References

1. Yang, X.S.: Engineering Optimization: An Introduction with Metaheuristic Applications. Wiley

(2010)
2. Yang, X.S., Deb, S., Fong, S.: Accelerated particle swarm optimization and support vector
machine for business optimization and applications. In: International Conference on Networked
Digital Technologies, Macau, China, 11–13 July 2011
3. Vilanova, R., Visioli, A.: PID Control in the Third Millennium. Springer (2012)
4. Xue, D., Chen, Y., Atherton, D.P.: Linear Feedback Control: Analysis and Design with MAT-
LAB. Siam (2007)
5. Sahib, M.A., Ahmed, B.S.: A new multiobjective performance criterion used in PID tuning
optimization algorithms. J. Adv. Res. 7(1), 125–134 (2016)
Index

A Continuous stirred-tank reactor, 13, 39

Accelerated particle swarm optimization, CSTR, see continuous stirred-tank reactor
247, 249 Curve fitting approximation, 102, 143
Acetic acid, 39, 42, 43
Agitator, 40
AI, see analog input D
Analog input, 44 Derivative kick effect, 1, 11, 13, 14, 19, 22,
Analog output, 44 26, 47, 48, 51, 55, 58, 61, 62, 73, 77,
AO, see analog output 80, 82, 85, 94, 95
Approximation table, 114 Deterioration rate, 85, 94, 95
APSO, see accelerated particle swarm opti- DFOC, 138
mization Differential calculus, 1
Asymptotic Bode plot, 164 DISO, see dual-input-single-output
Automatic voltage regulator, 19 Disturbance rejection, 9, 11, 14, 20, 22, 23,
30, 34–37, 44, 45, 47, 49–55, 57–61,
63, 65, 67, 69, 72, 75, 76, 78, 80, 81,
B 83, 84, 86, 87, 89, 90, 92, 94, 95
Binomial expansion, 155 Dual-input-single-output, 13, 95
Bode, 11, 107, 111, 120, 137, 138, 162, 164,
182–184, 190, 195, 200
E
Electroneutrality principle, 43
C EnPID, 9
Caputo, 17, 139 Equivalent circuit implementation, 2, 17,
Carlson, 3, 101, 102 101
CFE, see continued fractional expansion
Chaotic attractor, 231, 233–236, 238–243
Chaotic behavior, 225 F
Charef, 102 Feedback, 14, 15, 17, 18, 27, 28, 95
Chemical mixing tank, 39 Feedforward, 14, 27
Commande Robuste d’Ordre Non Entier Field-programmable gate array, 19
(CRONE), 17, 137, 138 Finite impulse response, 101
Commensurate orders, 231, 238, 242 FIR, see finite impulse response
Component Separated type, 29 FIT, see fractional integration tool
Conductivity transmitters, 40 Fixed-pole approximation, 101
Continued fractional expansion, 3, 101, 102 Flow control valve, 44

© Springer Nature Switzerland AG 2020 253

K. Bingi et al., Fractional-order Systems and PID Controllers,
Studies in Systems, Decision and Control 264,
https://doi.org/10.1007/978-3-030-33934-0
254 Index

Flow transmitters, 39 Industrial structure, 25

FOTF, see Fractional-order Transfer Func- Infinite impulse response, 101
tion Integral absolute error, 250
FPGA, see field-programmable gate array Integral squared error, 250
Fractional calculus, 1, 2 Integral time absolute error, 45, 49, 250
Fractional integration tool, 137 Integral time squared error, 250
Fractional State Space Tool (FSST), 17, 137, Ionic product of water, 43
138 ISE, see integral squared error
Fractional-order all-pass filter, 152 Iso-damping, 2, 11, 16, 17, 137
Fractional-order band-pass filter, 151 ITAE, see integral time absolute error
Fractional-order capacitor, 19 ITSE, see integral time squared error
Fractional-order chaotic oscillators, 214
Fractional-order chaotic systems, 213
Fractional-order Chen, 237 K
Fractional-order controllers, 1, 17 Kalman filter, 138
Fractional-Order Differentiator (FOD), 145
Fractional-order duffing, 219
L
Fractional-order filters, 148
Level control, 19
Fractional-order high-pass filter, 150
Level sensors, 40
Fractional-order impedance, 19
Levy’s cost function, 103, 133
Fractional-Order Integrator (FOI), 146
L∞ -norm , 164
Fractional-order Lorenz, 227
Load frequency, 19
Fractional-order low-pass filter, 149
Load regulation, 2, 10, 20, 32, 47, 60, 94, 95
Fractional-Order Modeling and Control
(FOMCON), 17, 137, 138, 210
Fractional-order Rössler, 241 M
Fractional-Order Transfer Function (FOTF), Maclaurin Series, 101
147 MAE, see mean absolute error
Fractional-Order Van der Pol, 214 Magnitude plot, 164
Frequency distribution mode, 102 Mass balance, 43
Frequency response data, 102, 104, 105, 143 Matrix Laboratory (MATLAB), 17, 137, 138
Fundamental modeling, 39 Matsuda, 101, 142
Fuzzy, 14 Mean absolute error, 105, 107, 111, 121, 124,
127, 131
Metaheuristic optimization algorithm, 247
G Model predictive control, 17
Grünwald-Letnikov, 17, 139, 163, 166, 214, Model reference adaptive control, 17
221, 228, 237, 241 Mod FPI, 157, 158
GUI, 138 Mod FPID, 158
Modified PID controllers, 11, 13
H
Hand valve, 63 N
Heuristic optimization, 13 Nichols, 138, 162, 164, 175, 195
H∞ -norm, 106, 164 Noise filter, 17
H2 -norm, 106, 164 Non-commensurate orders, 233, 240, 242
Non-Integer (NINTEGER), 17, 137, 138
Non-linear PID (NPID), 9, 11
I Non-linear stable spherical tank, 177, 188
IAE, see integral absolute error Nyquist, 138, 162, 164, 175, 195
Ideal structure, 23
IIR, see infinite impulse response
IMC-FOPID, 156, 158, 163 O
Improper, 101 Optimal control, 17
Index 255

Oscillatory, 1, 9, 11, 14, 15, 20, 49, 54, 58, PPI, 9, 11

61, 72, 76, 80, 85, 94, 148 Practical realization, 2, 17, 101
Oustaloup, 3, 102, 140, 163 Pre-filter, 14, 27, 29
Overshoot, 11, 14, 22, 44, 45, 49, 52, 54, 58, Pressure indicating controller, 63
60, 67, 72, 73, 76, 78, 82, 83, 85, 94, Pressure process plant, 62
95, 190, 194 Pressure transmitter, 63
Process and Instrumentation Diagram
(P&ID), 39, 63
P Process control valve, 63
Parallel structure, 24 PSO, see particle swarm optimization
Particle swarm optimization, 247
PCI, see Peripheral Component Interconnect
PCV, see process control valve R
(PD)n , 155 Refined Oustaloup, 141
PDμ , 158 Relay-based auto-tuning approach, 14
[PD]μ , 155, 158, 183, 184 Riemann-Liouville (RL), 17, 138, 139
Peak frequencies, 164 Rise time, 44, 46, 50, 52, 54, 56, 58, 60, 65,
Peripheral Component Interconnect (PCI), 72, 78, 82, 83, 85, 159, 190, 194
44, 64 Robotic manipulator, 19
PH, 39, 40, 43, 45, 64 Robustness, 11, 14, 18
PH Neutralization, 39
PH transmitter, 40
Phase plot, 164 S
Phase portraits, 215, 218, 223, 226 Sanathanan-Koerner (SK), 103, 105
PIλ , 158 Scilab, 139, 141, 145, 147, 150, 158, 164,
PIλ -PDμ , 35–39, 51, 53–57, 59, 61–65, 67, 176, 193–195, 205, 208, 209, 214,
69, 78, 79, 81–83, 85–95, 247, 248 215, 221, 229, 238, 243, 244
PIλ -PDμ 1, 35, 55–57, 59–61, 64, 66, 68, 81– Second order stable process plant, 206
83, 87, 90, 94 Set-point tracking, 2, 10, 11, 13, 18, 20, 22,
PIλ -PDμ 2, 37, 55–61, 64, 66, 68, 81–83, 85, 30, 32, 34–37, 44, 45, 47, 49, 51–55,
87, 90, 94 57–59, 61, 63, 65, 67, 69, 73, 75, 76,
PIλ Dμ , 2, 9–11, 15–18, 22, 37, 44, 45, 47– 78, 80–82, 84–95
50, 52, 55–58, 60–62, 64–66, 68, 69, Set-point weighting parameters, 13, 31
72–77, 79, 81, 82, 87, 90, 94, 101, Settling time, 44, 47, 50, 52, 54, 56, 58, 60,
126, 129, 130, 133, 158, 163, 182, 65, 72, 75, 76, 78, 82, 83, 85, 94, 159,
188, 193, 195 161, 190, 194
PI-PD, 2, 4, 9–11, 13–15, 20, 32–35, 37, 38, Simulink, 17, 44, 64, 138
52–58, 62–65, 78–80, 82–85, 89–91, Single-input-single-output, 13, 27, 95
94, 95 Singular-value sigma-plot, 164
PI-PD1, 33, 53–57, 61, 64, 66, 68, 80–83, SISO, see single-input-single-output
85, 90, 94 Sliding mode control, 17
PI-PD2, 34, 53–57, 61, 64, 66, 68, 80, 82, Sodium hydroxide, 39, 42
83, 85, 90, 94 Stability boundary locus, 3, 102
PIC, see pressure indicating controller Standard structure, 20
(PI)n , 17, 155, 158 State-space approach, 102
(PID)n , 154, 158 Steady-state, 31, 157
P&ID, see Process and Instrumentation Dia- Step info, 159, 164
gram SWPIλ -Dμ , 32, 33, 39, 47, 48, 50, 52, 59–62,
PIDτd , 10, 11 66, 68, 75, 77, 79, 81, 90, 94
[PI]λ , 155, 183, 184, 188 SWPIλ Dμ , 20, 22–24, 26, 27, 37–39, 44, 45,
[PI]μ , 158 47, 49, 51, 53, 59, 62, 65, 67, 69, 72–
Pole zero plot, 163 80, 82, 87, 92–94, 247, 248, 250
Power series expansion, 3, 17, 101, 102, 144, SWPID, 9–11, 13
145, 169 Sysquake, 17, 137, 138
256 Index

T V
Taylor series, 101 Vector fitting method, 102
TID, see tilt-integral-derivative
Tilt-integral-derivative, 19
Time moments approach, 18, 102 W
Titration, 39, 43 Wolfram Mathworld, 17
Transfer function estimation, 105, 143
Transient response, 31
X
Twin-rotor system, 19 Xcos, 139, 208, 209

Z
U Ziegler-Nichols (ZN), 67
Unstable bioreactor processes, 177, 193 ZN, see Ziegler-Nichols