Ebook Robust Engineering Designs of Partial Differential Systems and Their Applications 1St Edition Bor Sen Chen Online PDF All Chapter

Robust Engineering Designs of Partial
Differential Systems and Their

Applications 1st Edition Bor-Sen Chen
Visit to download the full and correct content document:
https://ebookmeta.com/product/robust-engineering-designs-of-partial-differential-syste
ms-and-their-applications-1st-edition-bor-sen-chen/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...
Systems Evolutionary Biology Biological Network

Evolution Theory Stochastic Evolutionary Game
Strategies and Applications to Systems Synthetic
Biology 1st Edition Bor-Sen Chen Phd
https://ebookmeta.com/product/systems-evolutionary-biology-
biological-network-evolution-theory-stochastic-evolutionary-game-
strategies-and-applications-to-systems-synthetic-biology-1st-
edition-bor-sen-chen-phd/
Systems Immunology and Infection Microbiology 1st

Edition Bor-Sem Chen
https://ebookmeta.com/product/systems-immunology-and-infection-
microbiology-1st-edition-bor-sem-chen/
Evolutionary Equations: Picard's Theorem for Partial

Differential Equations, and Applications 1st Edition
Christian Seifert
https://ebookmeta.com/product/evolutionary-equations-picards-
theorem-for-partial-differential-equations-and-applications-1st-
edition-christian-seifert/
Fundamentals of Data Engineering: Plan and Build Robust

Data Systems Joe Reis
https://ebookmeta.com/product/fundamentals-of-data-engineering-
plan-and-build-robust-data-systems-joe-reis/
Partial Differential Equations 3rd Edition Emmanuele
Dibenedetto
https://ebookmeta.com/product/partial-differential-equations-3rd-
edition-emmanuele-dibenedetto/
Fourier Analysis and Partial Differential Equations 1st

Edition Jose Garcia Cuerva
https://ebookmeta.com/product/fourier-analysis-and-partial-
differential-equations-1st-edition-jose-garcia-cuerva/
Building Integrated Photovoltaic Thermal Systems:

Fundamentals, Designs and Applications 1st Edition
Huiming Yin
https://ebookmeta.com/product/building-integrated-photovoltaic-
thermal-systems-fundamentals-designs-and-applications-1st-
edition-huiming-yin/
Building Integrated Photovoltaic Thermal Systems:

Fundamentals, Designs and Applications 1st Edition
Huiming Yin
https://ebookmeta.com/product/building-integrated-photovoltaic-
thermal-systems-fundamentals-designs-and-applications-1st-
edition-huiming-yin-2/
Practical Applications of Evolutionary Computation to

Financial Engineering Robust Techniques for Forecasting
Trading and Hedging Iba
https://ebookmeta.com/product/practical-applications-of-
evolutionary-computation-to-financial-engineering-robust-
techniques-for-forecasting-trading-and-hedging-iba/
Robust Engineering
Designs of Partial
Differential Systems and
Their Applications
Robust Engineering
Designs of Partial
Differential Systems and
Their Applications
Bor-Sen Chen
MATLAB® is a trademark of The MathWorks, Inc. and is used with permission. The
MathWorks does not warrant the accuracy of the text or exercises in this book. This book’s
use or discussion of MATLAB® software or related products does not constitute endorsement
or sponsorship by The MathWorks of a particular pedagogical approach or particular use of
the MATLAB® software.
First edition published 2022

by CRC Press
6000 Broken Sound Parkway NW, Suite 300, Boca Raton, FL 33487-2742
and by CRC Press

4 Park Square, Milton Park, Abingdon, Oxon, OX14 4RN
© 2022 Bor-Sen Chen
CRC Press is an imprint of Taylor & Francis Group, LLC
Reasonable efforts have been made to publish reliable data and information, but the author and
publisher cannot assume responsibility for the validity of all materials or the consequences of
their use. The authors and publishers have attempted to trace the copyright holders of all material
reproduced in this publication and apologize to copyright holders if permission to publish in this
form has not been obtained. If any copyright material has not been acknowledged please write
and let us know so we may rectify in any future reprint.
Except as permitted under U.S. Copyright Law, no part of this book may be reprinted,
reproduced, transmitted, or utilized in any form by any electronic, mechanical, or other means,
now known or hereafter invented, including photocopying, microfilming, and recording, or in
any information storage or retrieval system, without written permission from the publishers.
For permission to photocopy or use material electronically from this work, access www.
copyright.com or contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive,
Danvers, MA 01923, 978-750-8400. For works that are not available on CCC please contact
mpkbookspermissions@tandf.co.uk
Trademark notice: Product or corporate names may be trademarks or registered trademarks and
are used only for identification and explanation without intent to infringe.
Library of Congress Cataloging‑in‑Publication Data

Names: Chen, Bor-Sen, author.
Title: Robust engineering designs of partial differential systems and their
applications / Bor-Sen Chen.
Description: First edition. | Boca Raton, FL : CRC Press, 2022. |
Includes bibliographical references and index.
Identifiers: LCCN 2021033602 (print) | LCCN 2021033603 (ebook) |
ISBN 9781032134437 (hbk) | ISBN 9781032134451 (pbk) | ISBN 9781003229230 (ebk)
Subjects: LCSH: Reliability (Engineering)—Mathematics. |
Fuzzy systems—Design. | Differential equations, Partial.
Classification: LCC TA169 .C4685 2022 (print) | LCC TA169 (ebook) |
DDC 620/.00452—dc23
LC record available at https://lccn.loc.gov/2021033602
LC ebook record available at https://lccn.loc.gov/2021033603
ISBN: 978-1-032-13443-7 (hbk)

ISBN: 978-1-032-13445-1 (pbk)
ISBN: 978-1-003-22923-0 (ebk)
DOI: 10.1201/9781003229230
Typeset in Times
by codeMantra
Contents
Preface.................................................................................................................. xi
Author ................................................................................................................. xv
PART I Background
Chapter 1 Introduction to Partial Differential Systems ................................... 3

1.1 Partial Differential Equations and Partial
Differential Systems ............................................................. 3
1.1.1 Partial Differential Equations ................................. 3
1.1.2 Partial Differential System ...................................... 5
1.2 Numerical and Approximation Methods for PDEs
and PDSs............................................................................... 8
1.2.1 Galerkin Method ..................................................... 8
1.2.2 Finite Difference Method ...................................... 13
1.2.3 Poincaré Inequality Method .................................. 21
1.3 The Scope of the Book ....................................................... 22
1.4 Conclusion .......................................................................... 24
Chapter 2 Fuzzy Solutions to Partial Differential Equations ........................ 25

2.1 Introduction ........................................................................ 25
2.2 Problem Formulation .......................................................... 26
2.2.1 Description of PDE Problems ............................... 26
2.2.2 Description of Fuzzy Approximation.................... 28
2.3 Solving PDE Problems via Fuzzy Logic Systems .............. 30
2.3.1 Approximation Error Bound between Exact
Solution and Proposed Fuzzy Solution ................. 30
2.3.2 Adaptive Law for Adjustable Parameters θ
in Mesh Points ....................................................... 37
2.4 Simulation Results .............................................................. 42
2.5 Conclusion .......................................................................... 49
PART II Robust Signal Processing Design
Chapter 3 Robust Filter Design for Linear Stochastic Partial

Differential Systems via a Set of Sensor Measurements............... 53
3.1 Introduction ........................................................................ 53
v
vi Contents
3.2 A General H∞ Filtering Setting for Linear Stochastic

Partial Differential Systems ............................................... 54
3.3 Implementable H∞ Filter for Linear Stochastic
Partial Differential Systems ............................................... 64
3.4 Design Example of Robust H∞ Filter with Simulation ....... 80
3.5 Conclusion .......................................................................... 84
Chapter 4 Robust Filter Design for Nonlinear Stochastic Partial

Differential Systems in Sensor Signal Processing ........................ 85
4.1 Introduction ........................................................................ 85
4.2 Problem Statement and Preliminaries ................................ 87
4.3 Implementable H∞ Fuzzy Filter for Nonlinear
Stochastic Partial Differential Systems .............................. 94
4.4 Simulation Example ..........................................................110
4.5 Conclusion .........................................................................114
Chapter 5 Robust Synchronization Design of Partial Differential

Systems via Diffusion Coupling ..................................................115
5.1 Introduction .......................................................................115
5.2 Mathematical Model and Synchronization
Error Dynamic...................................................................117
5.3 Synchronization and Robust H∞ Synchronization of
Coupled PDSs with Constant Coefficients ....................... 120
5.4 Asymptotical Synchronization and H∞
Synchronization of N-Coupled PDSs with
Space-Dependent Coefficients.......................................... 129
5.5 H∞ Synchronization Criteria Based on LMI..................... 133
5.6 Numerical Examples ........................................................ 139
5.7 Conclusion ........................................................................ 144
PART III Robust Control System Design
Chapter 6 Robust Stabilization Control Design of Large Structural

Systems under Mode Truncation, Parameter Perturbations
and Actuator Saturations ............................................................. 149
6.1 Introduction ...................................................................... 149
6.2 Mathematical Notations and Preliminaries ...................... 150
6.3 System Description of LSS............................................... 151
6.4 Stabilization of LSS with Parameter Variations .............. 152
6.4.1 Extension to Structural Models with
Coupled Modes.................................................... 157
Contents vii
6.5 Stabilization of LSS with Constrained Controls .............. 158

6.6 Controller Synthesis ......................................................... 162
6.7 Simulation Examples ........................................................ 163
6.8 Conclusions....................................................................... 167
6.9 Appendix .......................................................................... 167
6.9.A Proof of Theorem 6.4.1 ....................................... 167
6.9.B Proof of Theorem 6.5.1 ....................................... 169
Chapter 7 Robust Observer-Based Output Feedback Control Design of

Large Flexible Structures: Mode State-Space Approach and
Frequency Domain Robustness Measurement Method................173
7.1 Introduction .......................................................................173
7.2 The Mathematical Model ..................................................174
7.3 Problem Formulation .........................................................176
7.4 Robust Stabilization with Respect to Control/
Observation Spillover ....................................................... 177
7.5 Robust Stabilization with Respect to the Total Spillover...... 182
7.6 Example: Modal Control of a Simply Supported Beam ... 187
7.7 Conclusion .........................................................................191
7.8 Appendix ...........................................................................191
7.8.A Proof of Theorem 7.4.1 .........................................191
7.8.B Proof of Theorem 7.4.2........................................ 193
7.8.C Proof of Theorem 7.4.3........................................ 194
Chapter 8 Robust Stabilization Design for Stochastic Linear Partial

Differential Systems under Spatiotemporal Disturbances and
Sensor Measurement Noises ....................................................... 197
8.1 Introduction ...................................................................... 197
8.2 A General H∞ Stabilization Setting for Linear
Stochastic Partial Differential Systems ............................ 199
8.3 Implementable H∞ Stabilization for Stochastic Partial
Differential Systems ......................................................... 209
8.4 Simulation Example ......................................................... 222
8.5 Conclusion ........................................................................ 226
Chapter 9 Fuzzy State-Space Modeling and Robust Observer-Based

Control Design for Nonlinear Partial Differential Systems ........ 229
9.1 Introduction ...................................................................... 229
9.2 Problem Formulation ........................................................ 231
9.3 System Representation by Fuzzy State Space Model ....... 234
9.3.1 Approximation of Nonlinear Partial
Differential Systems by Fuzzy Partial
Differential Systems ............................................ 234
viii Contents
9.3.2 Infinite-Dimensional T-S Fuzzy State-Space

Model of NPDS ................................................... 241
9.4 Robust Fuzzy Observer-Based Control Design of
Nonlinear Partial Differential Systems ............................ 250
9.4.1 Robust Stabilization of Nonlinear Partial
Differential Systems ............................................ 250
9.4.2 Fuzzy H∞ Observer-Based Control Design for
Nonlinear Partial Differential Systems ............... 259
9.6 Conclusion ........................................................................ 269
Chapter 10 Robust Tracking Control Design of Nonlinear Distributed

Parameter Time-Delayed Systems .............................................. 271
10.1 Introduction ...................................................................... 271
10.2 Reference Tracking Control Problem Formulation for
Nonlinear Distributed Parameter
Time-Delayed Systems ..................................................... 274
10.3 System Representation by Fuzzy Spatial State
Space Model ..................................................................... 277
10.4 Robust Fuzzy Observer-Based Tracking Control Design ......283
10.4.1 Robust H∞ Fuzzy Observer-Based Tracking
Control Design .................................................... 283
10.4.2 Solving Robust H∞ Tracking Control Problem
via Linear Matrix Inequality ............................... 293
10.5 An Application to Tracking Control of
Hodgkin-Huxley Nervous Systems .................................. 300
10.6 Conclusion ........................................................................ 303
Chapter 11 Robust Stabilization Control Design of Nonlinear Stochastic

Partial Differential Systems ........................................................ 305
11.1 Introduction ...................................................................... 305
11.2 System Description and Problem Formulation................. 309
11.2.1 Stochastic Stability of NSPDSs ............................310
11.2.2 Stochastic H∞ Stabilization Control for NSPDSs ..... 311
11.3 System Representation by Fuzzy Spatial State
Space Model ......................................................................312
11.4 Robust Stabilization Design for Nonlinear
Stochastic Partial Differential Systems ............................ 320
11.4.1 Robust Fuzzy Estimator-Based Stabilization
Control Design .................................................... 320
11.4.2 LMI Approach for Solving the
Robust Stochastic H∞ Stabilization
Control Problem .................................................. 329
Contents ix

11.6 Conclusion ........................................................................ 340
Chapter 12 Robust Fuzzy H∞ Estimator-Based Stabilization Design

for Nonlinear Parabolic Partial Differential Systems with
Different Boundary Conditions ................................................... 343
12.1 Introduction ...................................................................... 343
12.2 Preliminaries and Problem Formulation .......................... 347
12.3 Robust Fuzzy Estimator-Based Controller Design ........... 350
12.4 Robust Fuzzy H∞ Estimator-Based Stabilization
Design via Bilinear Matrix Inequalities ........................... 354
12.5 LMI Approach for Solutions to Robust Fuzzy H∞
Estimator-Based Stabilization Design .............................. 360
12.7 Conclusion ........................................................................ 372
Chapter 13 Low Design-Cost Fuzzy Controllers for Robust Stabilization

of Nonlinear Partial Differential Systems................................... 375
13.1 Introduction ...................................................................... 375
13.2 System Description and Problem Formulation................. 379
13.3 Robust Fuzzy H∞ Stabilization Design via Robust
Fuzzy Full-Controller ....................................................... 384
13.4 Low Design-Cost Robust Fuzzy Area-Controller for
Robust Fuzzy H∞ Stabilization Design ............................. 390
13.5 Low Design-Cost Robust Fuzzy Point-Controller for
Robust Fuzzy H∞ Stabilization Design ............................. 394
13.7 Conclusion ........................................................................ 409
References .........................................................................................................411
Index ................................................................................................................. 439
Preface
Many phenomena in science and engineering have been modeled by partial dif-
ferential systems (PDSs), for example, some phenomena related to heat flows,
fluid flows, elastic wave, flexible structure, etc., in mechanical systems, chemical
engineering systems, molecular transmission systems, biological systems, popu-
lation dynamical systems, neurophysiological systems, biodynamicical systems,
etc. In recent years, a great deal of concern has been raised regarding the study of
PDSs. In the past decades, the stabilization design problems of linear PDSs have
been widely studied by control engineers. The linear parabolic PDSs can be trans-
ferred to an equivalent infinite-dimensional ordinary differential system (ODS).
According to the separation of eigenvalues based on Galerkin’s method, the infi-
nite-dimensional ODS can be separated into a finite dimensional slow mode sys-
tem and an infinite-dimensional fast mode system. So a robust finite-dimensional
controller based on the appropriate finite-dimensional slow mode system could
be designed to stabilize the linear PDS with the robust stability to tolerate the
infinite-dimensional fast mode system. Recently, unlike the infinite-dimensional
ODS to represent the PDS, the partial differential operator on space in PDS could
be approximated by a finite-difference operator. Then, we can obtain a set of
finite-difference systems with the approximation error and truncation error to
represent the PDS. When all finite-difference grid points are represented by a
spatial vector, the set of finite-difference systems is represented by an equivalent
spatial state space system. Then a robust controller based on spatial state space
system can be developed to treat the stabilization problem of linear PDSs with
the ability to tolerate the truncation error. More recently, the divergence theorem
can be utilized to transform the diffusion matrix inequality (DMI) to bilinear
matrix inequality (BMI) and then the BMI could be transformed to linear matrix
inequality (LMI) by Poincare inequality method. Therefore the control design of
partial differential system could be significantly simplified by divergence theo-
rem and Poincare inequality. After the robust control design of linear PDSs, the
state estimation and filter design of linear stochastic PDS in signal processing
have been developed by engineers for signal processing application. Based on
fuzzy interpolation and global linearization method to interpolate several local
linear PDSs to approximate a nonlinear PDS, the robust control and filter design
methods in linear PDSs can be also extended to the nonlinear stochastic PDSs
for more practical applications. These control and filter designs of PDSs and their
applications are the main topics of the book.
The book is divided into three parts. Part I is about some backgrounds of
partial differential systems. In Part I, Galerkin and finite-difference methods are
introduced to approximated PDS by a finite-dimension ODS with a residual sys-
tem. Robust control design is also introduced to stabilize finite-dimension ODS
with the ability to tolerate the residual system. Then, for nonlinear PDSs, fuzzy
system is also introduced to interpolate several local linearized systems by fuzzy
xi
xii Preface
bases to approximate a nonlinear PDS by adaptive fuzzy algorithm. In Part II,

robust filter and estimator designs are introduced for signal processing in PDSs
under measurement noise and external disturbance in the spatiotemporal domain.
First, based on a set of sensor measurements, a robust filter design is introduced to
linear stochastic PDSs with measurement noise and external disturbance. Then,
a robust filter design is introduced to nonlinear stochastic PDSs in sensor signal
processing. In Part III, we will introduce some robust control design problems of
PDSs. We first introduce the robust stabilization control design problem of large
flexible structure systems, and then the robust stabilization design problem of
stochastic linear PDSs under spatiotemporal disturbance and sensor measurement
noise. Based on fuzzy state-space modeling, robust observer-based control design
is also introduced for the stabilization of nonlinear PDSs and nonlinear stochastic
PDSs. Finally, robust reference tracking control design is introduced for nonlinear
distributed parameter time-delay systems.
This book not only introduces recent robust filtering, robust stabilization and
robust reference tracking control design techniques in linear and nonlinear sto-
chastic partial differential systems but also applies these filter and control design
techniques to mechanical systems, aerospace systems, and control engineering
systems. Consequently, this book is useful for postgraduate students, engineers
and scientists in mechanic engineering, electrical engineering, aerospace engi-
neering, bioengineering, and ecology.
I first touched to partial differential equations in the course of electromagnet-
ics when I was a sophomore. When most students were still in confusion with the
operators like Curl and Div, I could always finish the homeworks of this course
because I had read a book of vector analysis in the previous summer vocations.
Consequently, I also took the course of electromagnetic wave in junior, which
discussed Maxwell electromagnetic wave equation in the spatiotemporal domain.
Then I had taken the graduate courses of fluid dynamics, elastic dynamics, numer-
ical forecasting of meteorology, electrodynamics, etc., in my M.S. of geophysics.
In my PhD in electrical engineering at USC, I had a roommate of architecture.
He had taken a course of finite element, which was a required but hard subject
for him and his classmates. I was always enthusiastic to help them solving home-
works of this course. At his convenience, I also learned some techniques of finite
element. Therefore, I had more background on partial differential systems than
conventional control engineers. When I became a professor, I had a co-supervised
PhD student in the department of aerospace engineering. He was interested in the
robust control design of large flexible structure and therefore I was involved in
the robust control and filter designs of partial differential systems. Consequently,
I had widely applied these knowledges to signaling processing and control engi-
neering design of partial differential systems in the last three decades.
Last, but not least, I would like to thank Dr. Gagandeep Singh, a publisher in
engineering at CRC Press, who has given me a lot of help during the course of this
book. Finally, I would like to thank Dr. Shih-Ju Ho for his help and Ms. Ci-Jun
Wang for her careful typing.
Preface xiii
MATLAB® is a registered trademark of The MathWorks, Inc. For product infor-

mation, please contact:
The MathWorks, Inc.
3 Apple Hill Drive
Natick, MA 01760-2098 USA
Tel: 508-647-7000
Fax: 508-647-7001
E-mail: info@mathworks.com
Web: www.mathworks.com
Author
Bor-Sen Chen earned a BS in electrical engineering at Tatung Institute of
Technology, Taipei, Taiwan, in 1970, and an MS in geophysics at National Central
University, Chungli, Taiwan in 1973, and a PhD at the University of Southern
California, Los Angeles, USA, in 1973. From 1973 to 1987, he was a lecturer,
associate professor, and professor at Tatung Institute of Technology. Since 1987 he
has been a professor, chair professor, and Tsing Hua Distinguished Chair Professor
with the Department of Electrical Engineering of National Tsing Hua University,
Hsinchu, Taiwan. His research interests include robust control theory and engi-
neering design, robust signal processing and communication system design, and
systems biology. He has published more than 300 journal papers, including 130
papers in control, 70 papers in signal processing and communication, and 100
papers in systems biology. He has also published 10 monographs. He was the
recipient of numerous awards for his academic accomplishments in robust control,
fuzzy control, H∞ control, stochastic control, signal processing and systems biol-
ogy, including four Outstanding Research Awards of National Science Council,
Academic Award in Engineering from the Ministry of Education, National Chair
Professor of the Ministry of Education, Best Impact Award of IEEE Taiwan Section
for his most SCI citations of IEEE members in Taiwan, etc. His current research
focuses on the H∞ team formation network tracking control of large-scale UAVs,
large-scale biped robots and their team cooperation, deep neural network-based
control design of nonlinear dynamic systems, systems medicine design based on
design specifications and deep learning schemes, etc. He is a life fellow of IEEE.
xv
Part I
Background
1 Introduction to Partial
Differential Systems
1.1 PARTIAL DIFFERENTIAL EQUATIONS AND

PARTIAL DIFFERENTIAL SYSTEMS
In this section, we introduce some partial differential equations (PDEs) at first,
especially the second-order PDEs with boundary conditions [1–3].Some funda-
mental properties such as existence, uniqueness and stability of PDEs are also
introduced. Then some corresponding partial differential systems (PDSs) are also
given to describe some nonlinear PDSs related to the robust system design prob-
lems in this book [4–16].
1.1.1 Partial Differential equations

A differential equation that contains, in addition to the dependent variable and the
independent variables, one or more partial derivatives of the dependent variable is
called a PDE. In general, it can be written as [1–3]
 ∂y ∂y ∂2 y ∂2 y 
F  x1 , x 2 , , x N , y, , , , 2 , , = 0 (1.1)
 ∂x1 ∂x2 ∂x1 ∂ x1 ∂ x 2 
where x1 , x 2 ,, x N are the independent variables with N ∈ and

y = y( x1 , x 2 ,… , x N ) ∈ is an unknown scalar function of these independent vari-
ables. The dependent variable y and its partial derivatives, which are considered
to be in a suitable bounded domain Ω of the N-dimensional space R N in the inde-
pendent variables x1 , x 2 , ..., x N , satisfy equation (1.1) identically in Ω. The order of
a PDE is the highest derivative that appears in the equation (1.1). For example, the
most general second-order 2-dimensional PDE is
∂2 y ∂2 y ∂2 y ∂y ∂y
2 + cb + cc 2 + cd + ce
ca + cf y = c (1.2)
∂x1 ∂ x1 ∂ x 2 ∂x 2 ∂x1 ∂x 2
where the coefficients ca, cb, cc, cd, ce, and cf are all allowed to be functions of
(x1,x2), and the inhomogeneity or forcing function c = c(x1,x2).
The classification of the second-order 2-dimensional PDE equation (1.2) is
based upon the possibility of reducing the equation (1.2) by a coordinate transfor-
mation to canonical or discriminant form at a point, such as [1]
cb2 ( x1p , x 2p ) − 4ca ( x1p , x 2p )cc ( x1p , x 2p ) (1.3)
DOI: 10.1201/9781003229230-2 3
4 Robust Engineering Designs of Partial Differential Systems
at a point ( x1p , x 2p ). An equation is said to be hyperbolic, parabolic, or elliptic

at a point ( x1p , x2p ) accordingly as equation (1.3) is positive, zero, or negative. If
this is true for all points ( x1 , x 2 ) ⊂ Ω, then the equation (1.2) is said to be hyper-
bolic, parabolic, or elliptic in the domain Ω; namely, second-order 2-dimensional
hyperbolic PDE, second-order 2-dimensional parabolic PDE, or second-order
2-dimensional elliptic PDE. For examples, the wave equation
∂2 y ∂2 y
− =0 (1.4)
∂x12 ∂x 22
has discriminant 4 > 0, and is hyperbolic; the heat equation
∂2 y ∂y
− =0 (1.5)
∂x12 ∂x 2
has discriminant 0 = 0, and is parabolic; the Poisson equation [3]
∂2 y ∂2 y
+ = −c (1.6)
∂x12 ∂x 22
has discriminant −4 < 0, and is elliptic.
Obviously, it is necessary to specify a boundary condition if the solution of
PDE is to be determined. The two most important kinds of boundary conditions
are [1–3]
∂y( x1 , x 2 , , x N )
= 0, ∀x1 , x 2 ,, x N ∈∂Ω (1.7)
∂nbase
∂ 
where  is the outward normal derivative, nbase is the normal vector to the
∂nbase
boundary ∂Ω. In considering time t, the initial condition y(t = 0, x1 , x2 ) is neces-
sary as well. The mathematical problem to solve such PDE is said well-posed if it
satisfies the following fundamental properties [1–3]
The first property is an obvious logical condition. The same can be said about
the uniqueness property. In order to really reflect the physical problem that has a
Introduction to Partial Differential Systems 5
unique solution, the mathematical problem must have a unique solution. For the
last property, if the solution is to have physical significance, a small change in the
initial data must produce a small change in the solution. The data in a physical
problem are normally obtained from experiment, and are approximated in order
to solve the problem by numerical or approximate methods. It is essential to know
that the process of making an approximation, such as the finite difference approx-
imation, to the data produces only a small change in the solution. These proper-
ties for the oncoming analysis discussed in the following sections are assumed
for simplicity. Surely, when simulating the PDE examples by using the numerical
approximate method, i.e., the central finite difference scheme, these properties
are guaranteed.
1.1.2 Partial Differential system

A PDS is a compound dynamic system constructed from PDE. Since most phys-
ical phenomena in reality can be described as PDE, the investigation on PDS
has become more and more popular nowadays. Among three types of the PDSs,
which are hyperbolic, parabolic, and elliptic PDSs, the most attractive one to us is
the following N-dimensional nonlinear parabolic PDS [1–2]
∂y( x ,t)
= D∆y( x , t) + f ( y( x , t)) (1.8)
∂t
on the spatio-temporal domain Ω × [0, T ], where x = [ x1 , x2 , , x N ]T ∈Ω ⊂ N is

the space variable with a smooth boundary ∂Ω and T ∈ + is the terminal time.
Here, we let Ω be a bounded convex domain for simplicity. f (⋅) is a smooth func-
T
tion with f (0) = 0, y( x, t )  y1 ( x , t), y2 ( x ,t ), , yny ( x , t )  ∈ ny and y ∈ D( Α)
is the state variable. Finally, ∆ = ∇ 2 = ∇ ⋅∇ is the Laplace operator for the
N-dimensional spatial domain, and D∆y( x , t), which represents the effect of dif-
fusion on the system, can be written as
 ∂2 y1 ( x ,t) ∂2 y1 ( x , t ) ∂2 y1 ( x ,t) 
 + ++ 
 ∂x1 2
∂x2 2
∂x N2 
 ∂2 y2 ( x ,t) ∂2 y2 ( x, t ) ∂2 y2 ( x , t ) 
 + + + 
D∆y( x , t) = [dij ]ny × ny  ∂x12 ∂x22 ∂x N2  (1.9)
 
 
 ∂2 yny ( x ,t) ∂2 yny ( x, t ) ∂2 yn y ( x , t) 
 + + + 
 ∂x12 ∂x22 ∂ x N2 
∂
where  is the outward normal derivative, n is the normal vector to the boundary
∂n
∂Ω, and the initial condition is y( x , 0)  y0 ( x ).
where y ∈ D( Α) is a known diffusion coefficient matrix. Similar to the bound-

ary conditions in equation (1.7), the boundary conditions we are interested in
are [1–3]
i. Dirichlet boundary condition (D):
y( x , t ) = 0, ∀x ∈∂Ω (1.10)
∂ y( x , t )
ii. Neumann boundary condition (N):  = 0, ∀x ∈∂Ω
∂n
Remark 1.1.1
i. Diffusion coefficient D, also called diffusivity, is an important param-

eter indicative of the diffusion mobility. It is not only encountered in
Fick’s law, but also in numerous other equations of physics and chem-
istry. Diffusion coefficient D is generally prescribed for a given pair
of species. For a multicomponent system, it is prescribed for each pair
of species in the system. The higher the diffusivity (of one substance
with respect to another), the faster they diffuse into each other. Many
physical processes have this diffusion behavior, such as fluid momen-
tum diffusion (D is kinematic viscosity, Newton’s law), thermal energy
diffusion (D is thermal diffusivity, Fourier’s law), fluid turbulent eddy
diffusion (D is eddy viscosity), porous media solute dispersion caused
by spatially varying velocity (D is aquifer dispersion coefficient), and
aquifer flow (D is hydraulic diffusivity).
ii. For broadly well-posedness of the PDS, forward diffusion, i.e., D ≥ 0,
is necessary in general. For example, the solution at t → ∞ is called the
steady-state solution, after the system reaches an equilibrium state. For
steady states, we have
∂y
=0
∂t
∂y
Therefore, for the linear diffusion equation = ∆y, we have
∂t
∆y = 0
which is the Laplace equation. The solution depends on the boundary conditions
∂2 y
in equation (1.10). In the 1-dimensional case, , which means the solution
∂x 2
is of the form ax + b, where a, b are constants. The solution can be computed
analytically, depending on the boundary conditions. However, consider the linear

inverse diffusion equation
∂y
= −∆y
∂t
This is an unstable ill-posed process. By changing the sign of the time differen-
tiation ∂t → − ∂t, one can view this also as reversing in time the forward diffu-
sion process. Any chemist knows that diffusion is a smoothing process since the
concentration of a substance tends to flatten out. Going backward, the situation
becomes more and more chaotic. Hence, one would not expect well-posedness
of the backward-in-time problem for the diffusion equation. That is, going from
a blurred signal to a sharp one (not a physical process). To avoid the possible
ill-posedness of the concerned PDS like equation (1.8), the positive semidefinite
diffusion coefficient matrix D is commonly used.
When applying a control force to the PDS equation (1.8), we have
∂y( x ,t)
= D∆y( x , t) + f ( y( x , t)) + Bu( x , t) (1.11)
∂t
where u( x ,t ) [u1 ( x , t),u2 ( x ,t ), , unu ( x , t )]T ∈ nu is the control force to be

designed with the influence matrix B ∈ ny × nu . The PDS in equation (1.11) is
so-called the N-dimensional noise-free nonlinear parabolic PDS. Generally,
systems suffer from external noise in practice. Hence, we have the following
N-dimensional nonlinear parabolic PDS
∂y( x ,t)
= D∆y( x , t) + f ( y( x , t)) + Bu( x , t) + En n( x , t ) (1.12)
∂t
where n( x , t) [n1 ( x ,t ), n2 ( x , t), , nnn ( x , t )]T ∈ L2 (Ω × + ; nn ) represents the

external noise to the PDS with the influence matrix En ∈ ny × nn . However, the
system state y may not be available in reality. Therefore, the N-dimensional non-
linear parabolic PDS equation (1.12) is reformulated as [4,5]
∂y( x ,t)
= D∆y( x , t) + f ( y( x , t)) + Bu( x , t) + En n( x , t ) (1.13)
∂t
z( x , t) = h( y( x , t)) + Em m( x , t) (1.14)
where h(⋅) is a smooth measure function with h(0) = 0 and the rest of settings are
the same as the PDS equation (1.12). z( x , t) [ z1 ( x ,t ), z2 ( x , t), , znz ( x, t )]T ∈ nz
represents the measurement output affected by the measurement noise
m( x , t) [m1 ( x ,t ), m2 ( x , t), , mnm ( x , t )]T ∈ L2 (Ω × + ; nm ) with the influence
matrix Em ∈ nz × nm . The control designs for the aforementioned PDSs in equa-
tions (1.11) and (1.12), and the state observation and control designs in equations
(1.13) and (1.14), are quite interesting topics.
1.2 NUMERICAL AND APPROXIMATION

METHODS FOR PDEs AND PDSs
In dealing with many equations arising from the modeling of physical problems,
the determination of such exact solutions in a simple domain is almost a formi-
dable task even when the boundary and/or initial data are simple. In view of the
wide-spread accessibility of today’s high speed electronic computers, numerical
and approximation methods are becoming increasingly important and useful in
real applications. This section is devoted to the analytical treatment of PDEs
and PDSs. Three main analytical methods to find the exact analytical solution
of these equations within simple domains will be discussed. In the following
sections, the Galerkin method, also called the Galerkin finite element method
for PDEs and PDSs will be presented at first. Then, the finite difference methods
(FDMs), including the methods with forward, central, and backward finite dif-
ference schemes, for PDEs and PDSs will be provided. Finally, the divergence
theorem and Poincaré inequality methods to deal with the PDEs and PDSs with
Dirichlet and Neumann boundary condition in equation (1.10) will be discussed.
These analytic methods of PDEs and PDSs will provide numerical and approxi-
mation methods for robust system designs of PDEs and PDSs in the following
chapters.
1.2.1 Galerkin methoD

Galerkin formulated a numerical approximation method that may be applied
to a problem, in which no simple variational principle exists [3]. In the area
of numerical analysis, the Galerkin method is a class of methods for convert-
ing a continuous operator problem (such as a differential equation) to a discrete
problem. In principle, it is equivalent to applying the method of variation of
parameters to a function space, by converting the equation to a weak formula-
tion. Consider the boundary value problem governed by the 2-dimensional dif-
ferential equation
Ay = f , in Ω (1.15)
with the boundary condition
B( y) = 0,on ∂Q (1.16)
where A is a self-adjoint differential operator in the Hilbert space

L2 (Ω;  ny ), f ∈ L2 (Ω;  ny ), and y ∈ D( A). In order to solve the boundary value
problem (1.15) and (1.16), we construct an approximate solution y in the form
n
yn = y0 ( x1 , x 2 ) + ∑ a φ (x , x )
i=1
i i 1 2 (1.17)
which is called the Galerkin solution of equations (1.15) and (1.16), where
φi ( x1 , x 2 ) ∈ L2 (Ω;  ny ) are known functions, y0 is introduced to satisfy the bound-
ary conditions, and the coefficients ai are to be determined. Substituting equation
(1.17) into (1.15) gives a nonzero residual Rn
n
Rn (a1 , a2 , an , x1 , x 2 ) = Ayn = Ay0 + ∑ a Aφ

i=1
i i (1.18)
In this method, the unknown coefficients ai are determined by solving the follow-
ing equations [3,4]
Rn ,φ j n
L2 ( Ω; y )
= 0, for j = 1, 2,, n (1.19)
Since A is linear, this equation (1.19) can be written as
∑a
i=1
i Aφi ,φ j
(n
L2 Ω; y )=− Ay0 ,φ j
( n
L2 Ω;  y ) (1.20)
which determines ai ’s. Substitution of the ai ’s obtained from the solution of

equation (1.20) into (1.17) gives the required approximate solution yn. However,
the general solution yn in equation (1.17) with the determination of the coefficients
ai calculated from equation (1.20) may be difficult. An interesting connection
between the Galerkin solution and the Fourier representation of y is found. We
seek a Galerkin solution in the following form:
n
yn = ∑ a φ (x , x )
i=1
i i 1 2 (1.21)
with a special restriction
 0,i ≠ j
φi , φ j
( n
L2 Ω; y ) =  1,i = j
(1.22)

Thus, the application of the Galerkin method to equation (1.15) gives
∑aφ i i Aφi ,φ j = f ,φ j
( n
L2 Ω; y ) (1.23)
i =1 ( n
L2 Ω; y )
which is, by equation (1.22),
ai = f ,φi (1.24)
( n
L2 Ω; y )
so the Galerkin solution equation (1.21) becomes

n
yn = ∑
i=1
f , φi
(n
L2 Ω; y ) φi ( x1 , x2 ) (1.25)
Evidently, the Galerkin solution (1.21) is just the finite Fourier series solution.
When n goes to infinity, we will obtain the perfect approximation
n
y = lim yn = lim
n→∞ n→∞ ∑
i=1
f , φi
(n
L2 Ω; y ) φi ( x1 , x2 ) (1.26)
Therefore, the perfect Galerkin solution like equation (1.26) results in the
infinite-dimensional (n → ∞) problem. That is, we have to use infinitely many
equations in the linear combination form to obtain the perfectly exact solution
of PDEs or PDSs. This Galerkin method makes it possible simulate solutions
of PDEs and PDSs with sufficient large number n. To guarantee the well-
posedness of the PDEs or PDSs with the Galerkin method, both the property of
boundedness
Ayn ,φ j ≤ c1 Ayn φj
( n
L2 Ω; y ) ( n
L2 Ω;  y ) ( n
L2 Ω; y ) , ∀j (1.27)
and the property of ellipticity
2
Ayn ,φ j ≥ c2 yn (1.28)
( n
L2 Ω;  y ) ( n
L2 Ω; y )
for some constants c1 > 0 and c2 > 0, are necessary. These two conditions imply
well-posedness by the Lax-Milgram theorem.
Here, we take a kind of Poisson equation for example to find the approxi-
mate solution by using the Galerkin method. Consider the following Poisson
equation:
∂2 y ∂2 y
+
∂ x12 ∂ x 22
= −1,in Ω = {( x , x ) x
1 2 1 < a, x 2 < b } (1.29)
with the boundary condition
y = 0 on ∂Ω = { x1 = ±a, x 2 = ±b} (1.30)
Now, we seek a trial solution in the for [3,4]

M M
yM ( x1 , x 2 ) = ∑ ∑
m=1,3,5 n=1,3,5
amnφmn ( x1 , x2 ) (1.31)
where
mπx1  nπx 2 
φmn ( x1 , x 2 ) = cos  cos  (1.32)
 2a   2b 
In this case A = ∆ = ∇ 2 and the residual RM is
 M M
 m2π2 n2π2  
RM = AyM + 1 = − 
 m =1
∑∑ n =1
 4 a 2 +
4 b 
2 amnφ mn  +1

(1.33)
According to the Galerkin method, we obtain

b a
kπx1  lπx
RM cos  cos  2  dx1dx 2
RM ,φkl n
L2 ( Ω;  y )
=
∫ ∫
−b −a  2a   2b 
abπ 2  k 2 l 2  16ab
=  a − 2 (−1)( k + l ) /2−1 = 0
2 + 2  kl (1.34)
4 a b  π kl
or
2 ( k + l ) /2 −1
8ab  (−1)
akl =  2  (1.35)
 π  (b 2 k 2 + a 2 l 2 )
Thus, the solution of the problem is
2 M M
(−1)( m + l ) /2−1φmn
8ab 
yM ( x1 , x 2 ) =  2 
 π  ∑ ∑
m=1,3,5 n=1,3,5
(b 2 m 2 + a 2 n 2 )
(1.36)
Moreover, in the limit M → ∞, the solution yM ( x1 , x 2 ) is in perfect agreement

with those obtained by the double Fourier series.
The Galerkin method can be used not only to simulate solutions of PDEs and
PDSs, but also to deal with the diffusion effect on the parabolic PDSs for the
control design purposes. Taking a simple 1-dimensional linear parabolic PDS for
example, we consider the PDS [3]
∂y
= y + ay + bu (1.37)
∂t
where y ∈ D ( A ) is a constant function, u ∈ is a constant control input, a ∈,
and b ∈. With the Hilbert-Schmidt theorem, the solution of equation (1.37) can
be represented as [3]
∞
y= ∑ y (t)φ (x)
l=1
1 l (1.38)
which is also called spectral expansion, where yl (t ) = y,φl ( x ) L2 (Ω;) and

φl ( x ) = λlφl ( x ) with the eigenvalues λl as well as the corresponding orthonormal
eigenfunctions φl ( x ). The orthonormal eigenfunctions φl ( x ) have the property
 0, k ≠ l
φk ( x ),φl ( x ) = (1.39)
 1, k = l
L2 ( Ω; )
and constitute a set of bases for L2 (Ω; ). λl is ordered so that Re {λl +1 } ≤ Re {λl },
where Re {λl } denotes the real part of λl . Hence, we have
∂y ∂yl (t )
,φl ( x) = , ul (t ) u,φl ( x ) (1.40)
∂t L2 ( Ω; ) ∂t L2 ( Ω;)
and
∆y,φl ( x ) L2 ( Ω; )
= λl yl (t) (1.41)
From equations (1.40) and (1.41), we can rewrite equation (1.37) as
∂yl (t )
= λl yl (t ) + ayl (t ) + bul (t ), for l = 1, 2, (1.42)
∂t
Then, we collect all yl (t ) to construct the following ordinary differential system
(ODS)
y(t ) = Ay(t ) + BU (t ) (1.43)
y(t ) [ y1 (t ), , y∞ (t ) ] ∈ ∞ , A diag ( λ1 + a, , λ∞ + a ) ∈ ∞×∞ ,

T
U(t ) =
[u1 (t ),, u∞ (t )] ∈ , and B diag ( b, , b ) ∈ ∞×∞ . To avoid infinite-
T ∞
dimensional problem of the ODS, based on the property of the eigenvalues λl and
the orthonormal eigenfunctions φl ( x ), the system state of the PDS equation (1.37)
should be partitioned into a finite-dimensional slow subsystem and an infinite-
dimensional fast residual subsystem. Hence, we define
y( x ,t ) = yc ( x , t ) + yr ( x ,t) (1.44)
where
nc ∞
yc ( x ,t )  ∑l=1
yl (t )φl ( x ) and yr ( x ,t )  ∑ y (t)φ (x)
l = nc +1
l l (1.45)
Similar to the analysis in equations (1.40)–(1.43), we can obtain [4]
yc (t ) = Ac yc (t ) + Acr yr (t ) + BcUc (t ) (1.46)

yr (t ) = Arc yc (t ) + Ar yr (t ) + BrUr (t ) (1.47)
Uc (t ) [ u1 (t ), , unc (t ) ] , Ur (t ) [ unc +1 (t ), , u∞ (t ) ] ,

T T
where yc (t ) 
[1
y (t ), , ync (t ) ]T
is the system state of the finite- dimensional slow controllable
subsystem, and yr (t ) [ ync +1 (t ), , y∞ (t ) ] is the system state of the coupled
T
infinite-dimensional fast residual subsystem. The matrices Ac , Ar , Bc and Br , with

A and B as follows:
 Ac Acr   Bc 0 
A=  and B =   (1.48)
 Arc Ar   0 Br 
Generally, the residual subsystem equation (1.47) is asymptotically stable because

one always chooses a negative definite matrix Ar , which means λnc +1 + a < 0,...,
λ∞ + a < 0 based on the property Re {λl +1 } ≤ Re {λl }, i.e., the unstable mod-
els should be put into Ac. Moreover, in this simple 1-dimensional linear para-
bolic PDS equation (1.37), we have Acr = 0 and Arc = 0; however, if the PDS is a
complex one, then the coupled matrices Acr and Arc would be nonzero matrices.
Similarly, BrUr (t ) = 0, which is so-called the control spillover, when considering
this simple PDS. Consequently, by choosing a sufficient large number nc , the con-
trol designs for the simple PDS equation (1.37) can be reformulated as ones for
the ODS equation (1.46) that can be realized via the traditional algebraic matrix
techniques. However, for the complex PDSs in the studies, the effects of the cou-
pled matrices Acr and Arc; the control spillover BrUr (t ) ; the observation spillover;
external and measurement noises need to be considered. That is, not only the
finite-dimensional slow subsystem like equation (1.46) needs to be controlled, but
also the infinite-dimensional fast residual subsystem like equation (1.47) needs
to be tolerated. Therefore, the main disadvantage of using the Galerkin method
to deal with the diffusion effect on the parabolic PDSs for the control designs is
that the coupled infinite-dimensional fast residual subsystem needs to be consid-
ered in the design procedure, which means the effect of the subsystem needs to
be tolerated to avoid unexpected control performance. Besides, the size of yc (t ),
i.e., nc , is large normally, which means dealing with the diffusion effect on the
parabolic PDSs by using the Galerkin method consumes a great deal of computa-
tional resources. In this book, the Galerkin method will be applied to the control
and state observation design of linear and nonlinear PDSs in Chapters 6, 7 and 9.
1.2.2 finite Difference methoD

In mathematics, FDMs are numerical methods for approximating the solutions
to differential equations using finite difference equations to approximate deriva-
tives. By using the FDMs, one can approximate the derivatives appearing in a
PDE or PDS by sums and differences of algebraic function values at a set of dis-
crete points, usually uniformly spaced with respect to each independent variable.
Briefly, the FDM can be characterized as follows. FDM utilizes uniformly spaced
FIGURE 1.1 An example of finite difference grids of the central finite difference scheme
on the 2-dimensional spatial domain Ω.
grids (Figure 1.1, e.g., the central finite difference scheme on the 2-dimensional
spatial domain Ω).
At each node yk ,l , each derivative is approximated by an algebraic expression
which references the adjacent nodes, i.e., yk −1,l , yk +1,l , yk ,1− l and yk ,1+ l . A system of
algebraic equations is obtained by evaluating the previous step for each node and
the system is solved for the dependent variable. For clarity, we illustrate the prin-
ciple of the FDM on the 1-dimensional spatial domain Ω, for example. The system
of algebraic equations with the sums and differences of algebraic function values
is based on the famous Taylor Series from elementary calculus [5]
∞
( x − h) ( x − h) 2
∑ (x −n!h)
n
y( x ) = y( h ) + y '(h) + y ''(h) +  = y( n ) (h) (1.49)
1! 2! n=0
where y( n ) denotes the derivative taken n times of y with respect to the space vari-
able x, n! denotes the factorial of n, and h is denoted as an increment in x and is
assumed positive. Now, we replace x and h in the previous formula (1.49) by x + h
and x, respectively. Then, we have
∞
h2
∑ hn! y
n
h
y( x + h) = y( x ) + y '( x ) + y ''( x ) +  = (n)
(x) (1.50)
1! 2! n=0
For the treatment of many problems, it is convenient to take only the first two
terms of the right hand side of the previous equation,
y( x + h) = y( x ) + hy '( x ) + O(h 2 ) (1.51)
where the expression O(h 2 ) represents the error of the approximation, which is
proportional to h 2 , and O(h 2 ) is the famous big O notation that describes the limit-
ing behavior of a function when the argument h 2 tends towards a particular value
or infinity. From this relationship, we can easily define what is known as the first
dy
order, forward finite difference scheme to , i.e.,
dx
dy y( x + h) − y( x )
= + O ( h) (1.52)
dx h
dy
Likewise, we can define the first order, backward finite difference scheme to
dx
dy y( x ) − y( x − h)
= + O (h) (1.53)
dx h
by
y( x − h) = y( x ) + hy '( x ) + O(h 2 ) (1.54)
If instead, taking up to the third order, then we have
h2
y( x + h) = y( x ) + hy '( x ) + y ''( x ) + O(h 3 ) (1.55)
2
h2
y( x − h) = y( x ) − hy '( x ) + y ''( x ) + O(h 3 ) (1.56)
2
By subtracting equation (1.56) from (1.55), we can obtain the first order, central
dy
finite difference scheme to , i.e.,
dx
dy y( x + h) − y( x − h)
= + O(h 2 ) (1.57)
dx 2h
The central finite difference scheme obviously has the best approximation since
the increment h is always chosen as a small positive number which is definitely
far less than one (0 < h  1). Therefore, the central finite difference scheme is
normally used to simulate solutions of PDEs and PDSs; thus, the illustration of
the followings focuses on the FDM with the central finite difference scheme. As
dy
to the second order, central finite difference scheme to , we take up to the forth
dx
order as follows:
h2 h3
y( x + h) = y( x ) + hy '( x ) + y ''( x ) + y '''( x ) + O(h 4 ) (1.58)
2 6
h2 h3
y( x − h) = y( x ) − hy '( x ) + y ''( x ) − y '''( x ) + O(h 4 ) (1.59)
2 6
and
y( x − h) + y( x + h) = 2 y( x) + h 2 y ''(x ) + O(h 4 ) (1.60)

d2y
Now, the second-order central finite difference scheme to 2 can be obtained
dx
as follows:
d 2 y y( x + h ) − 2 y( x ) + y( x − h )
= + O(h 2 ) (1.61)
dx 2 h2
Hence, for the numerical solutions of PDEs and PDSs on the 1-dimensional spa-
tial domain Ω, one can utilize the following approximation:
∂ y( x , t ) y( x , t + ht ) − y( x , t )
= + O(ht ) (1.62)
∂t ht
∂ y( x , t ) y( x + hx , t ) − y( x − hx , t )
= + O(hx2 ) (1.63)
∂x 2hx
∂2 y( x , t ) y( x + hx ) − 2 y( x ) + y( x − hx )
= + O(hx2 ) (1.64)
∂x 2 hx2
where hx1 and hx2 are the increment of the time and space variables x1 and x 2,
respectively. Moreover, for the numerical solutions of PDEs and PDSs on the
2-dimensional spatial domain Ω, one can use equations (1.63) and (1.64) to obtain
the following:
( ) (
∂2 y (( x1 , x 2 ) , t ) y ( x1 + hx1 , x 2 + hx2 ) , t + y ( x1 − hx1 , x 2 − hx2 ) , t
=
)
∂ x1 ∂ x 2 4 hx1 hx2
−
( ) (
y ( x1 + hx1 , x 2 − hx2 ) , t + y ( x1 − hx1 , x 2 + hx2 ) , t )
4 hx1 hx2
 ( hx + hx2 )4 
+O 1  (1.65)
 hx1 hx2 
∆y =
( ) (
y ( x1 + hx1 , x 2 ) , t − 2 y (( x1 + x 2 ) , t ) + y ( x1 − hx1 , x 2 ) , t )
2
h
x1
+
( )
2
( )
y ( x1 , x 2 + hx2 ) , t − 2 y (( x1 + x 2 ) , t ) + y y ( x1 , x 2 − hx2 ) , t t ) + O(⋅) (1.66)
h
x2
where hx1 and hx2 are the increment of the space variables x1 and x 2, respectively.
O(⋅) is proportional to the complicated function constructed from hx21 and hx22 . One
can use the results in equations (1.62)–(1.66) to obtain the numerical solutions of
PDEs and PDSs. Since the parabolic PDEs and PDSs are concerned in this book,
the followings focus on the approximation equations (1.62) and (1.66).
For the parabolic PDEs and PDSs on the 2-dimensional spatial domain Ω
(Figure 1.1), we first impose a mesh of grid points; hence, we have the interior
( )
grid points x1( k ) , x 2(l ) = ( khx1 , lhx2 ), where k = 1, 2,, N x1 , and l = 1, 2,, N x2 . With
the result of equation (1.66), but without considering O(⋅), we have
∆y =
(( ) ) (( ) ) ((
y x1( k +1) , x 2(l ) , t − 2 y x1( k ) + x 2(l ) , t ) + y x1( k −1) , x 2(l ) , t ) )
2
h x1
+
(( ) ) (( ) ) ((
y x1( k ) , x 2(l +1) , t − 2 y x1( k ) + x 2(l ) , t + y x1( k ) , x 2(l −1) , t ) ) (1.67)
2
hx2
For simplicity, one always defines hx to be the distance between mesh points, i.e.,
hx1 = hx2 = hx . This yields the simplification
1
∆yk ,l =
hx2
[ yk +1,l + yk −1,l + yk ,l +1 + yk ,l −1 − 4 yk ,l ] (1.68)
( )
where yk ,l  y x1( k ) , x 2(l ) , t = y ( khx1 ,lhx2 ). If the higher order term is considered,
then we could find out that the local truncation error for ∆yk ,l is
1 2  ∂4 y ∂4 y 
hx + (1.69)
12  ∂x14 ∂x 24  x = x ( k ) , x
1 1
(l )
2 = x2 , y = y ( x1( k ) ,x2(l ) )
( )
which is usually written as O hx2 such that the error is proportional to the space
( )
constant square O hx2 . That is, the local truncation error O hx2 , which will be ( )
discussed later, can be neglected if one chooses a sufficient small grid size hx .
Note that five points (points to the left, right, above and below of a central point,
( )
x1( k ) , x 2(l ) ) are involved in the approximation of ∆yk ,l ; hence, the FDM with the
central finite difference scheme is also called the five point formula. Obviously,
the FDMs with the forward, central, and backward finite difference schemes are
spatial-discretized methods.
Recall that whatever types of PDEs or PDSs, i.e., the hyperbolic, parabolic,
and elliptic types of PDEs or PDSs, can be numerically simulated by employ-
ing the FDM mentioned above. In order to guarantee the well-posedness of the
PDEs or PDSs with the FDM, we give the following discussion. Suppose y( x1 , x 2 )
represents the exact solution of a PDE or PDS L ( y ( x1 , x 2 )) = 0 with independent
variables x1 as well as x 2, and yk ,l is the exact solution of the corresponding spatial-
discretized algebraic equation F ( yk ,l ) = 0, which is constructed from equations
(1.62)–(1.66). Then, the FDM is said to be convergent if yk ,l tends to y( x1 , x 2 ) as
(
hx1 and hx2 tend to zero. The difference d k ,l  y x1( k ) , x 2(l ) − yk ,l  Yk ,l − yk ,l is the )
truncation error or the so-called discretization error. This error can be gener-
ally be minimized by decreasing the grid sizes hx1 and hx2 . Moreover, this error
depends on the number of terms in the truncated series equations (1.62)–(1.66),
which are used to approximate each partial derivative. Another kind of error is
introduced when a PDE or PDS is approximated by the spatial-discretized alge-
braic equation. If the solution yk ,l is replaced by the solution y( x1 , x 2 ) at the grid
points (see Figure 1.1), then the value F (Yk ,l ) is called the local truncation error
at Pk ,l . The FDM and the PDE or PDS are said to be consistent if F (Yk ,l ) tends to
zero as hx1 and hx2 tend to zero. In general, the spatial-discretized algebraic equa-
tions cannot be solved exactly because the numerical computation is carried out
only up to a finite number of decimal places. Consequently, another kind of error
is introduced in the solution yk ,l during the actual process of computation. This
error is called the round-off error, and also depends upon the type of computer
used. In practice, the actual computational solution is yk ,l *, but not yk ,l , so that the
difference rk ,l  yk ,l − y*k ,l is the round-off error at Pk ,l . In reality, the round-off
error depends mainly on the actual computational process and the FDM itself;
therefore, this error cannot be made small by allowing hx1 and hx2 to tend to zero.
Hence, the total error involved in the finite difference analysis at the point Pk ,l is
Yk ,l − y*k ,l = (Yk ,l − y k ,l ) + ( yk ,l − y*k ,l ) = d k ,l − rk ,l (1.70)
Usually, the truncation error d k ,l is bounded when yk ,l is bounded because the

value of Yk ,l is fixed for a given PDE or PDS with the prescribed boundary and
initial condition. This fact is used or assumed in order to introduce the concept of
stability. The FDM is said to be stable if the round-off errors are sufficiently small
for all k as l → ∞; that is, the growth of rk ,l can be controlled. Lax (1954) proved a
remarkable theorem which establishes the relationship between consistency, sta-
bility, and convergence for the FDM.
Theorem 1.2.1 (Lax’s Equivalence Theorem [1,2])
Given a properly posed linear initial-value problem and a FDM approximation

to it that satisfies the consistency criterion, stability is the necessary and sufficient
condition for convergence.
In other words, Lax states that for a consistent FDM for a well-posed linear
initial-value problem, the method is convergent if and only if it is stable. This
theorem is important because the convergence, which is desired, is ordinarily
difficult to establish. Nevertheless, the consistency is straightforward to verify,
which means the FDM and the PDE or PDS are consistent based on equations
(1.62)–(1.66) with hx1 → 0 and hx2 → 0. In addition, the stability is typically much
easier to show than the convergence. Thus, the convergence is usually established
via the Lax’s equivalence theorem. For showing the stability, Von Neumann sta-
bility analysis (1947) is employed. Von Neumann stability analysis (also known
as Fourier stability analysis) is a procedure used to check the stability of FDM as
applied to a PDE or PDS. Generally, the stability of numerical schemes is closely
associated with numerical error in software engineering and mathematics, which
is the combination of the truncation and round-off errors. Von Neumann stability
analysis states that a FDM is stable if the errors made at one time step ht of the
calculation do not cause the errors to increase as the computations are continued.
On the contrary, if the errors grow with time, the FDM is unstable. The stabil-
ity analysis is based on the decomposition of the errors into Fourier series. For
examples, consider 1-dimensional and 2-dimensional heat equations as
∂y( x ,t) ∂2 y ( x , t )
=α (1.71)
∂t ∂x 2
and
∂y (( x1 , x 2 ) , t ) ∂2 y (( x1 , x 2 ) , t ) ∂2 y (( x1 , x 2 ) , t )
= α1 + α2 (1.72)
∂t ∂x1 2
∂ x 22
where α ∈  + , α 1 ∈  + , and α 2 ∈ +. Then, the stability conditions of equations
(1.71) and (1.72) are
α ht 1 αh α h 1
≤ , and 12 t + 22 t ≤ (1.73)
hx2 2 hx1 hx2 2
respectively, where h 2x1 and h 2x2 have been defined in equations (1.65) and (1.66).
As to the effect of diffusion on the concerned PDS equation (1.8), we choose a
conservative stability condition as
D  ht ht ht  1
λmax  h 2 + h 2 ++ h 2  ≤ 2 (1.74)
x1 x2 xN
where λmax
D
is the maximum eigenvalue of the diffusion coefficient matrix D. With
the stability conditions (1.73) and (1.74) and the consistency based on hx1 → 0 and
hx2 → 0, the convergence can be guaranteed via the Lax’s equivalence theorem.
Surely, the FDM with the central finite difference scheme under the stability con-
dition (1.74) is utilized in our simulation results for the concerned PDSs through-
out this book.
After introducing the solutions of PDEs and PDSs via using the FDMs, we
briefly demonstrate the treatments of the diffusion effect on the parabolic PDSs
for the control design purposes. With the central finite difference scheme, the
similar methods will be used in the followings to deal with the diffusion effect.
The primary idea is to collect the system states at all grid nodes on spatial domain
as a spatial state vector. Then, the PDSs can be transformed into the ODSs with
the local truncation errors. Accordingly, the control designs for the PDSs can
be formulated as ones for the ODSs that can be realized via the traditional alge-
braic matrix techniques. Taking a simple 2-dimensional linear parabolic PDS for
example, we consider the PDS
∂y
= D y + ay + bu (1.75)
∂t
where the only difference between equations (1.75) and (1.37) is that
x = [ x1 , x 2 ] ∈Ω ⊂  2 with D ∈  +. With the typical grid mesh shown
T
in Figure 1.1, the system state y is represented by yk ,l (t ) at the grid node

x k ,l ( x1 = khx1 = khx , x 2 = lhx2 = lhx ), where k = 1, 2, , N x1 , and l = 1, 2, , N x2 , i.e.,
y( x , t ) x = xk ,l  yk ,l (t ) . At the grid points x k ,l, the central finite difference approxi-
mation for the Laplace operator ∆ can be written as follows [3]:
 ∂ 2 y( x , t )  y (t ) + yk −1,l (t ) − 2 yk ,l (t )
 ∂ x 2  = k +1,l
x = x k ,l
hx2
yk ,l +1 (t ) + yk ,l −1 (t ) − 2 yk ,l (t )
+ + O k ,l (hx2 ) (1.76)
hx2
( )
where Ok ,l hx2 is referred to as the local truncation error, which converges to
zero if the grid size hx is chosen to be as small as possible by applying the Lax’s
equivalence theorem in Theorem 1.2.1, i.e., Ok ,l hx2 → 0 as hx → 0. Therefore, ( )
the PDS equation (1.75) can be represented as follows:
yk ,l (t ) = ayk ,l (t ) + buk ,l (t )
D
+
hx2
[ yk +1,l (t ) + yk −1,l (t ) − 2 yk ,l (t ) + yk ,l+1 (t ) + yk ,l−1(t ) − 2 yk ,l (t )] (1.77)
where uk ,l (t ) is the control effort at the grid node x k ,l. To simplify the control
design, we define a spatial state vector Y(t) to collect the states yk ,l (t ) at all grid
nodes in Figure 1.1, and Y (t ) ∈  NY is given as follows:
T
Y (t ) =  y1,1 (t ), , yN x1 ,1 (t ), , yk ,l (t ), , y1, N x2 (t ), , yN x1 , N x2 (t )  (1.78)
where N y  N x1 × N x2 . In order to simplify the index of the node yk ,l (t ), we

denote the symbol y j (t ) to replace yk ,l (t ). Note that the index j is from 1 to N Y,
i.e., y1 (t )  y1,1 (t ), y2 (t )  y1,2 (t ),..., y j (t )  yk ,l (t ) ,..., yNY (t )  yN x1 , N x2 (t ), where
j = ( k − 1) N x1 + l in equation (1.78). Then, the finite difference model of two indi-
ces in equation (1.77) can be represented with only one index as follows [5,8–10]
y j (t ) = ay j (t ) + bu j (t ) + DTjY (t ) (1.79)
where Tj ∈ 1× NY expresses the interaction from the other grid nodes to the jth
node as follows:
TjY (t ) Tx1 , j ,1 + Tx2 , j ,1 , , Tx1 , j , NY + Tx2 , j , NY  Y (t )

NY
= ∑ (T
j2 =1
x1 , j , j2 + Tx2 , j , j2 ) y j2 (t ) (1.80)
in which Tx1 , j , j2 ∈ and Tx2 , j , j2 ∈ are defined as
 −2
 I , for j2 = (l − 1) N x1 + k
 hx2
 1
 I , for j2 = (l − 1) N x1 + k − 1
Tx1 , j , j2  hx2
 1
 I , for j2 = (l − 1) N x1 + k + 1
 hx2
 0, otherwise

(1.81)
 1
 I , for j2 = (l − 2) N x1 + k
 hx2
 1
 I , for j2 = lN x1 + k
Tx2 , j , j2  hx2
 −2
 I , for j2 = (l − 1) N x1 + k
 hx2
 0, otherwise

By applying some matrix multiplication techniques to (1.79), the PDS with the
state Y (t ) can be constructed; thus, the control designs for the PDS (1.75) can be
realized via the traditional algebraic matrix techniques when hx → 0. Note that if
the state y( x ,t) of the PDS (1.75) is not a constant function, i.e., y( x ,t) ∈ ny , then
Y (t ) ∈ ny NY , Ok ,l (hx2 ) ∈  ny , Tj ∈ ny × ny NY , Tx1 , j , j2 ∈ ny × ny , and Tx2 , j , j2 ∈ ny × ny .
Accordingly, the control designs for the PDSs can be realized based on control
design of DDS with state Y(t) as the previous analysis [4,5,9–11]. More details
will be given in the following chapters. The main drawback of using the cen-
tral finite difference scheme to deal with the diffusion effect on the parabolic
PDSs for the control designs is that the number of grid nodes N Y needs to be
sufficient large. That is, the tradeoff problem between the grid size hx for the
central finite difference scheme and the control performance occurs. If the grid
size h x is small enough so that Ok ,l (hx2 ) → 0, then a great deal of computational
resources is necessary because of the sufficient large N Y. Otherwise, the effect
of the local truncation error Ok ,l (hx2 ) needs to be considered to avoid unexpected
control performance. That is, a great deal of computational resources is neces-
sary as well.
1.2.3 Poincaré inequality methoD

Consider the PDS in equation (1.8) with Neumann boundary condition N in
equation (1.10). The following equations based on Green theorem and diver-
gence theorem [2,3] are useful to treat the system designs of PDSs with Neumann
boundary condition N in equation (1.10).
∫Ω
yT ( x )∆y( x ) dx =
∫ Ω
yT ( x )∇ 2 y( x ) dx
∂y( x )
=−
∫ ∇ y(x)∇y(x) dx + 2 ∫
Ω
T
∂Ω
yT ( x)  ds (by Green Theorem) (1.82)
∂n
 ∂y( x )
where ∇y( x) ⋅ n = 
∂n
By the Dirichlet and Neumann boundary conditions in equation (1.10) [2,3]
∫Ω
yT ( x )∆y( x ) dx = −
∫ ∇ y(x)∇y(x) dx
Ω
T
(1.83)
By Poincaré in equality [2,3]
cp
∫ Ω
yT ( x )y( x ) dx ≤
∫ ∇ y(x)∇y(x) dx
Ω
T
(1.84)
for some Poincaré constant cp, which can be chosen as the first nonzero eigen-
value of the Laplace operator ∆.
Then we obtain
∫ Ω
yT ( x )∆y( x ) dx ≤ −c p
∫ Ω
yT ( x ) y( x ) dx (1.85)
Based on equation (1.85), the Laplace operator ∆y( x ) in the system design of
PDSs could be treated like the conventional ordinary differential systems (PDSs).
These Poincaré inequality techniques will be applied in Chapters 5, 12, and 13 for
signal processing and control design of PDSs [12–14].
1.3 THE SCOPE OF THE BOOK

This book provides a perspective of new framework that can help bring together
the system design theory of PDSs and practical application to engineering and
bio-science. This book is divided into three parts. In part I, the background of
PDE and PDS is introduced. In part II, robust signaling and synchronization of
PDSs are given. In part III, several robust control designs of PDEs are introduced
for linear and nonlinear PDSs.
Part I is divided into two chapters. Chapter 1 gives an introduction to PDEs
and PDSs, with which this book considers robust system designs. Some funda-
mental mathematic properties like existence, uniqueness and stability are intro-
duced. Three numerical and approximation methods i.e., Galerkin method, FDM,
and Poincaré inequality method, are given to solve the partial differential Laplace
operator, which will meet in system design problems of the following chapters.
In Chapter 2, a new technique of adaptive fuzzy algorithm is introduced to solve
PDEs and PDSs encountered in science and engineering [10].
Part II is divided into three chapters. Based on finite difference scheme and
stochastic spatial state space model, Chapter 3 introduces a robust H∞ filter design
of linear stochastic PDSs with a set of sensor measurements [8]. With the help
of fuzzy spatial state space method and finite difference scheme, in Chapter 4,
a robust H∞ filter design is also introduced for nonlinear stochastic PDSs in sen-
sor signal processing [9]. In Chapter 5, based on Green theorem and Poincaré
inequality, a robust H∞ synchronization design is given to PDSs with diffusion
coupling [12].
Part III focuses on robust control system design and is divided into eight chap-
ters. Based on Galerkin method, a robust observer-based stabilization control
design is proposed in Chapter 6 for large structural systems like supported beam
problem under mode truncation, parameter perturbations and actuator saturations
[6]. Chapter 7 introduces a robust observer-based control design of large flexible
structures based on mode state space approach and frequency domain robustness
measure method. The robust observer-based controller could stabilize the con-
trolled mode state space model and tolerate the coupling spillovers from residual
mode state space model [7]. In Chapter 8, based on finite difference scheme and
spatial state space model, a robust H∞ stabilization design for stochastic linear
PDSs under spatio-temporal disturbance and sensor measurement noises. Based
on Galerkin method and fuzzy state space modeling, a robust observer-based
control design is proposed for nonlinear PDSs in Chapter 9. T-S fuzzy PDS can
interpolate several local linear partial systems to approximation any nonlinear
PDS by fuzzy bases interpolation [4]. Then the schemes in observer-based control
design in Chapter 7 could be employed for robust H∞ stabilization of nonlinear
PDSs. In Chapter 10, based on fuzzy interpolation schemes, FDM and spatial
state space model, a robust H∞ reference tracking control design of nonlinear
distributed parameter time-delayed systems is introduced for practical application
[5]. The proposed optimal robust H∞ reference tracking control design problem
of nonlinear PDSs can be transformed to a linear matrix inequalities (LMIs) –
constrained optimization problem. Based on FDM, spatial state space model
and fuzzy interpolation method, a robust H∞ stabilization design is proposed in
Chapter 11 for nonlinear stochastic PDSs. A robust estimator-based controller
is proposed for robust stochastic H∞ stabilization design to efficiently attenuate
the effect of random external disturbance and measurement noise in the spatio-
temporal domain [11]. In Chapter 12, based on divergence theorem and Poincaré
inequality to transform diffusion matrix inequality to bilinear matrix inequal-
ity (BMI), a robust fuzzy H∞ estimator-based stabilization design is proposed
for nonlinear parabolic PDS with different boundary conditions [13]. Based on
the proposed decoupling method, the robust fuzzy H∞ estimator-based stabilizing
design problem for nonlinear parabolic PDSs can be effectively solved by a set of
LMIs instead of BMI. Since design-cost of fuzzy controller for nonlinear PDSs is
very expensive, some low design-cost fuzzy controllers are proposed for nonlin-
ear parabolic PDSs [14]. Based on Green theorem and Poincaré inequality, robust
fuzzy H∞ stabilization area and point controllers are proposed as low design-cost
controller of nonlinear PDSs in Chapter 13.
1.4 CONCLUSION
In this chapter, we first describe PDSs. Then some numerical and approximation
techniques of Galerkin method, FDM and Poincaré inequality method for PDEs
and PDSs are introduced to treat of Laplace operator, which is one of the most dif-
ficult problems in the signal processing and control engineering designs of PDSs
in the following chapters. Finally, an overview of each chapter in three parts of
the book about signal processing and control designs of the partial differential
engineering systems is also outlined in this chapter.
2 Fuzzy Solutions to
Partial Differential
Equations
2.1 INTRODUCTION
It is well known that a lot of phenomena of nature or physical systems can be
modeled by partial differential equations PDEs, such as heat equations, wave
equations, and so on. Hence, studies of PDEs have become one of the main
topics of modern mathematical analysis and have attracted much attention.
However, the exact solutions to the PDEs cannot be easily obtained except for
very simple or special cases. In recent years, many methods have been devel-
oped for solving some kinds of PDEs. For example, a spreadsheet program is
used for the numerical solution of the hyperbolic equation in [17]. Moreover,
some studies produce a solution in the form of an array that contains the value
of the solution at a selected group of points [18]. Others use finite-element
methods that are famous and widely adopted in the mechanical fields to solve
some specific PDEs [19–21]. In general, a finite-element solution can only offer
one discrete solution to approximate the exact solution of PDEs through mini-
mizing the functional of the finite-element method, but this solution is often
limited differentiable. In addition, the functional of variational finite-element
method must be given first. In general, it is not easy to find the functional of the
finite-element method for complex PDEs that cannot be easily derived from the
variational calculus [22,23]. Furthermore, how close this computed solution is
to the exact solution and whether it converges to the exact solution cannot eas-
ily be checked in practice. Recently, the fuzzy spline wavelets [24] were used
to approach the solution of differential equations. On the other hand, a fuzzy
transform technology [25] was applied to approach the numerical solutions of
PDEs based on finite-difference methods in [26].
Hence, the investigation of an effective and more correct method for solving
the partial differential systems (PDSs) or the systems of PDEs is an important
task! As mentioned earlier, one cannot easily obtain a solution for complex PDEs
(high order and nonlinear types). The main reason is that exact solutions are like
unknown black boxes, and the profile of the solution cannot be easily obtained
by human works or any mathematical analysis. Since the solution of PDEs can
be regarded as an unknown system, an approximated method that can handle
any unknown system is likely to find a suitable and accurate solution for the
treated PDEs. The numerical solution via a fuzzy transform method, as in [26],
DOI: 10.1201/9781003229230-3 25
needs to solve a PDE via an algebraic equation through a fuzzy transform tech-
nique. However, for a nonlinear or high-dimensional PDE, the algebraic equation
becomes very complex. In this chapter, a fuzzy solution via adaptive algorithm is
proposed to approach the solution of PDE. Therefore, for the nonlinear or high-
dimensional cases, the solution of PDE can be easily obtained by the proposed
adaptive fuzzy algorithm.
In this chapter, we view the traditional problems of PDEs mentioned earlier
from a different perspective with the help of fuzzy logic systems. Fuzzy logic
systems have been widely used in system modeling to deal with the nonlinear
unknown systems for control or filter designs in recent years [27–31]. In most of
these fuzzy system designs [32–35], the fuzzy system was thought to be a univer-
sal approximator [36,37] for any nonlinear system. Fuzzy logic system has also
been proved to be a very good representation for a class of nonlinear dynamic
systems by the conventional schemes, and any nonlinear unknown system can be
approximated to any desired accuracy [32]. For this reason, an advanced method
that relies on the functional approximation capability of the fuzzy logic systems
via fuzzy interpolation scheme to solve the problems of PDEs will be proposed
in this chapter. This interpolation scheme employs a linear regression of fuzzy
logic system as the basic approximation element, whose parameters in conse-
quent parts are adjusted to minimize an appropriate error function via an adap-
tive algorithm. In addition, the upper bound of approximation error between the
exact solution and the proposed fuzzy solution is derived, and this result can be
easily extended to high-dimensional PDE cases. Moreover, the sufficient condi-
tion for the convergence of the proposed fuzzy solution in mesh points is also
provided.
Three attractive features of the proposed method are obtained as follows:
2.2 PROBLEM FORMULATION
2.2.1 DescriPtion of PDe Problems
Consider a class of a two-dimensional PDE formulated as follows [38]:
f ( x , y,φ ,φ x ,φ y ,φ xy ,φ xx ,φ yy ) = u( x, y) (2.1)
Fuzzy Solutions to Partial Differential Equations 27
where φ ( x , y) is a twice-differentiable function defined on a region

Ω = [a, b] × [c, d ] ⊆  2, and f is a function of variables x, y,φ ,φ x ,φ y ,φ xy ,φ xx ,φ yy,
∂φ ( x , y) ∂φ ( x , y)
where φ x = ,φ y = , and so on. Equation (2.1) can be generalized
∂x ∂y
to more than two variables x, y in an obvious way (see Remark 2.2.1). In addition
to equation (2.1), we must specify boundary conditions (BCs) that are most fre-
quently encountered in scientific and engineering problems of the form:
 ∂φ 
g  x , y, φ ,   = 0 (2.2)
 ∂n  ∂Ω
∂φ ∂φ
where  is the normal derivative on ∂Ω, i.e., some functions of φ and  evalu-
∂n ∂n
ated on boundary ∂Ω are set to zero.
Assumption 2.2.1
Suppose there exists a unique solution to (2.1) and (2.2)
Remark 2.2.1
One can straightforwardly extend the aforesaid two-variable PDE problems to

multivariable PDE problems as follows:
    
f ( x ,φ ( x ), ∇φ ( x ), ∇ 2φ ( x )) = u( x ) (2.3)
subject to certain BCs as follows:

 
φ ( x ) = B( x ), (Dirichlet) (2.4)
or
   
n∇φ ( x ) = C ( x ), (Neumann) (2.5)

where x = ( x1 , x n ) ∈ n ,φ ( x ) is the solution to be computed, and n is the out-
ward unit vector normal to the boundary.
To solve the PDE problem in equations (2.1) and (2.2), we denote two functions
β 0 and β1 as follows:
β 0 ( x , y,φ ,φ x ,φ y ,φ xy ,φ xx ,φ yy ) = f ( x, y,φ ,φ x ,φ y ,φ xy ,φ xx ,φ yy ) − u( x , y) (2.6)
 ∂φ   ∂φ 
β1  x , y,φ ,  = g  x , y,φ ,   (2.7)
 ∂n   ∂n  ∂Ω
If one exact solution φ ( x , y) for the PDEs and the BCs can be found analytically,
β 0 (*) and β1 (*) will be zero. However, it is very difficult to find one exact solution
from the PDEs with the BCs mentioned earlier by mathematical analysis or tools.
For this reason, we will develop one approximated fuzzy solution φ ( x , y θ ) in
the sequel to satisfy the PDEs equation (2.1) with BCs in equation (2.2) as pre-
cisely as possible.
Problem of the PDE: Using Assumption 2.2.1, we consider the PDE equation
(2.1) with boundary conditions (2.2). The design objective is to find one suitable
and approximated fuzzy solution φ ( x , y θ ) such that β 0 (*) and β1 (*) are as small
as possible.
2.2.2 DescriPtion of fuzzy aPProximation

Practically, the unknown solution of the PDE in equation (2.1) with the BCs in
equation (2.2) can be regarded as an unknown system. Inherently, the fuzzy logic
system is very suitable for treating this problem with its excellent approximation
ability. With the help of a fuzzy logic system, the following fuzzy solution model
is introduced to approximate the unknown solution φ ( x , y) of the PDEs with the
BCs in equations (2.1) and (2.2).
The fuzzy solution of the PDE with the BCs in equations (2.1) and (2.2) can be
described as follows:
Rα : IF x is µix and y is µ uj
(2.8)
THEN φ ( x , y) = θ ij
where Rα (α = 1, , mx × n y ) are fuzzy rules, and x and y are the input vari-
ables to the fuzzy system. φ ( x , y) = θ ij is the solution proposed by the αth rule;
µix (i = 1, ,mx ) and µ uj ( j = 1, ,n y ) are the membership functions that charac-
terize the ith and jth fuzzy sets defined in the space of the variables of x and y
coordinates, respectively. These membership functions in these cases have been
chosen as follows:
 1  x − aix  2 
µix = exp  −  x   (2.9)
 2  σ i  
 1  y − ay  2 
µ = exp  − 
u
j
j
y 
 (2.10)
 2  σ j  
where aix , a yj , σ ix , σ jy are the mean values and the standard deviations of the mem-
bership distributions, respectively. From the membership function shown in equa-
tions (2.9) and (2.10), it is easy to find out that they are nonlinear functions in C∞.
The overall fuzzy solution is given by
Another random document with
no related content on Scribd:
Indeed, the most prevailing incentive to our labor was to secure
the means of education for some male member of the family. To
make a gentleman of a brother or a son, to give him a college
education, was the dominant thought in the minds of a great many of
these provident mill-girls. I have known more than one to give every
cent of her wages, month after month, to her brother, that he might
get the education necessary to enter some profession. I have known
a mother to work years in this way for her boy. I have known women
to educate by their earnings young men who were not sons or
relatives. There are men now living who were helped to an education
by the wages of the early mill-girls.
In speaking of this subject, Mr. Thomas Wentworth Higginson
says,—
“I think it was the late President Walker who told me that in his
judgment one-quarter of the men in Harvard College were being
carried through by the special self-denial and sacrifices of women. I
cannot answer for the ratio; but I can testify to having been an
instance of this myself, and to having known a never-ending series
of such cases of self-devotion.”
Lowell, in this respect, was indeed a remarkable town, and it

might be said of it, as of Thrums in “Auld Licht Idyls,” “There are
scores and scores of houses in it that have sent their sons to college
(by what a struggle), some to make their way to the front in their
professions, and others, perhaps, despite their broadcloth, never to
be a patch upon their parents.”
The early mill-girls were religious by nature and by Puritan
inheritance, true daughters of those men and women who, as some
one has said, “were as devoted to education as they were to
religion;” for they planted the church and the schoolhouse side by
side. On entering the mill, each one was obliged to sign a “regulation
paper” which required her to attend regularly some place of public
worship. They were of many denominations. In one boarding-house
that I knew, there were girls belonging to eight different religious
sects.
In 1843, there were in Lowell fourteen regularly organized
religious societies. Ten of these constituted a “Sabbath School
Union,” which consisted of over five thousand scholars and teachers;
three-fourths of the scholars, and a large proportion of the teachers,
were mill-girls. Once a year, every Fourth of July, this “Sabbath
School Union,” each section, or division, under its own sectarian
banner, marched in procession to the grove on Chapel Hill, where a
picnic was held, with lemonade, and long speeches by the ministers
of the different churches,—speeches which the little boys and girls
did not seem to think were made to be listened to.
The mill-girls went regularly to meeting and “Sabbath-school;” and
every Sunday the streets of Lowell were alive with neatly dressed
young women, going or returning therefrom. Their fine appearance
on “the Sabbath” was often spoken of by strangers visiting Lowell.
Dr. Scoresby, in his “American Factories and their Operatives,”
(with selections from The Lowell Offering,) holds up the Lowell mill-
girls to their sister operatives of Bradford, England, as an example of
neatness and good behavior. Indeed, it was a pretty sight to see so
many wide-awake young girls in the bloom of life, clad in their
holiday dresses,—
“Whose delicate feet to the Temple of God,

Seemed to move as if wings had carried them there.”
The morals of these girls were uniformly good. The regulation

paper, before spoken of, required each one to be of good moral
character; and if any one proved to be disreputable, she was very
soon turned out of the mill. Their standard of behavior was high, and
the majority kept aloof from those who were suspected of wrong-
doing. They had, perhaps, less temptation than the working-girls of
to-day, since they were not required to dress beyond their means,
and comfortable homes were provided by their employers, where
they could board cheaply. Their surroundings were pure, and the
whole atmosphere of their boarding-houses was as refined as that of
their own homes. They expected men to treat them with courtesy;
they looked forward to becoming the wives of good men. Their
attitude was that of the German Fräulein, who said, “Treat every
maiden with respect, for you do not know whose wife she will be.”
But there were exceptions to the general rule,—just enough to
prove the doctrine of averages; there were girls who came to the mill
to work whom no one knew anything about, but they did not stay
long, the life there being “too clean for them.”
The health of the girls was good. The regularity and simplicity of
their lives, and the plain and substantial food provided for them, kept
them free from illness. From their Puritan ancestry they had inherited
sound bodies and a fair share of endurance. Fevers and similar
diseases were rare among them; they had no time to pet small
ailments; the boarding-house mother was often both nurse and
doctor, and so the physician’s fee was saved. It may be said that, at
that time, there was but one pathy and no “faith cures” nor any
“science” to be supported by the many diseases “that flesh is heir
to.”
By reading the weekly newspapers the girls became interested in
public events; they knew all about the Mexican war, and the anti-
slavery cause had its adherents among them. Lectures on the
doctrine of Fourier were read, or listened to, but none of them were
“carried away” with the idea of spending their lives in large
“phalansteries,” as they seemed too much like cotton-factories to be
models for their own future housekeeping.
The Brook Farm experiment was familiar to some of them; but the
fault of this scheme was apparent to the practical ones who foresaw
that a few would have to do all the manual labor and that an undue
share would naturally fall to those who had already contracted the
working-habit.
Mrs. Amelia Bloomer, one of the early pioneers of the dress-
reform movement, found followers in Lowell; and parlor meetings
were held at some of the boarding-houses to discuss the feasibility
of this great revolution in the style of woman’s dress. The Lowell
Journal of 1850 states that on the Fourth of July a party of
“Bloomerites” walked in the procession through the public streets,
and The London Punch embellished its pages with a neat cartoon, a
fashion-plate showing the different styles of the Bloomer costume.
This first attempt at a reform in woman’s dress was ridiculed out of
existence by “public opinion;” but from it has been evolved the
modern bicycle costume, now worn by women cyclers.
It seems to have been the fashion of the mill-girls to appear in
procession on all public occasions. Mr. Cowley, in his “History of
Lowell,” speaks of President Jackson’s visit to that city in 1833. He
says: “On the day the President came, all the lady operatives turned
out to meet him. They walked in procession, like troops of liveried
angels clothed in white [with green-fringed parasols], with cannons
booming, drums beating, banners flying, handkerchiefs waving, etc.
The old hero was not more moved by the bullets that whistled round
him in the battle of New Orleans than by the exhilarating spectacle
here presented, and remarked, ‘They are very pretty women, by the
Eternal!’”
CHAPTER V.
CHARACTERISTICS (CONTINUED).
One of the first strikes of cotton-factory operatives that ever took

place in this country was that in Lowell, in October, 1836. When it
was announced that the wages were to be cut down, great
indignation was felt, and it was decided to strike, en masse. This was
done. The mills were shut down, and the girls went in procession
from their several corporations to the “grove” on Chapel Hill, and
listened to “incendiary” speeches from early labor reformers.
One of the girls stood on a pump, and gave vent to the feelings of
her companions in a neat speech, declaring that it was their duty to
resist all attempts at cutting down the wages. This was the first time
a woman had spoken in public in Lowell, and the event caused
surprise and consternation among her audience.
Cutting down the wages was not their only grievance, nor the only
cause of this strike. Hitherto the corporations had paid twenty-five
cents a week towards the board of each operative, and now it was
their purpose to have the girls pay the sum; and this, in addition to
the cut in the wages, would make a difference of at least one dollar a
week. It was estimated that as many as twelve or fifteen hundred
girls turned out, and walked in procession through the streets. They
had neither flags nor music, but sang songs, a favorite (but rather
inappropriate) one being a parody on “I won’t be a nun.”
“Oh! isn’t it a pity, such a pretty girl as I—

Should be sent to the factory to pine away and die?
Oh! I cannot be a slave,
I will not be a slave,
For I’m so fond of liberty
That I cannot be a slave.”
My own recollection of this first strike (or “turn out” as it was

called) is very vivid. I worked in a lower room, where I had heard the
proposed strike fully, if not vehemently, discussed; I had been an
ardent listener to what was said against this attempt at “oppression”
on the part of the corporation, and naturally I took sides with the
strikers. When the day came on which the girls were to turn out,
those in the upper rooms started first, and so many of them left that
our mill was at once shut down. Then, when the girls in my room
stood irresolute, uncertain what to do, asking each other, “Would
you?” or “Shall we turn out?” and not one of them having the
courage to lead off, I, who began to think they would not go out, after
all their talk, became impatient, and started on ahead, saying, with
childish bravado, “I don’t care what you do, I am going to turn out,
whether any one else does or not;” and I marched out, and was
followed by the others.
As I looked back at the long line that followed me, I was more
proud than I have ever been since at any success I may have
achieved, and more proud than I shall ever be again until my own
beloved State gives to its women citizens the right of suffrage.
The agent of the corporation where I then worked took some
small revenges on the supposed ringleaders; on the principle of
sending the weaker to the wall, my mother was turned away from her
boarding-house, that functionary saying, “Mrs. Hanson, you could
not prevent the older girls from turning out, but your daughter is a
child, and her you could control.”
It is hardly necessary to say that so far as results were concerned
this strike did no good. The dissatisfaction of the operatives
subsided, or burned itself out, and though the authorities did not
accede to their demands, the majority returned to their work, and the
corporation went on cutting down the wages.
And after a time, as the wages became more and more reduced,
the best portion of the girls left and went to their homes, or to the
other employments that were fast opening to women, until there
were very few of the old guard left; and thus the status of the factory
population of New England gradually became what we know it to be
to-day.
Some of us took part in a political campaign, for the first time, in
1840, when William H. Harrison, the first Whig President, was
elected; we went to the political meetings, sat in the gallery, heard
speeches against Van Buren and the Democratic party, and helped
sing the great campaign song beginning:—
“Oh have you heard the news of late?”
the refrain of which was:
“Tippecanoe and Tyler too,

Oh with them we’ll beat little Van, Van,
Van is a used-up man.”
And we named our sunbonnets “log-cabins,” and set our teacups

(we drank from saucers then) in little glass tea-plates, with log-
cabins impressed on the bottom. The part the Lowell mill-girls took in
these and similar events serves to show how wide-awake and up to
date many of these middle-century working-women were.
Among the fads of those days may be mentioned those of the
“water-cure” and the “Grahamite.” The former was a theory of
doctoring by means of cold water, used as packs, daily baths, and
immoderate drinks. Quite a number of us adopted this practice, and
one at least has not even yet wholly abandoned it.
Several members of my mother’s family adopted “Professor”
Graham’s regimen, and for a few months we ate no meat, nor, as he
said, “anything that had life in it.” It was claimed that this would
regenerate the race; that by following a certain line of diet, a person
would live longer, do better work, and be able to endure any
hardship, in fact, that not what we were, but what we ate, would be
the making of us. Two young men, whom I knew, made their boasts
that they had “walked from Boston to Lowell on an apple.”
We ate fruit, vegetables, and unleavened or whole-wheat bread,
baked in little round pats (“bullets,” my mother called them), and
without butter; there were no relishes. I soon got tired of the feeling
of “goneness” this diet gave me; I found that although I might eat a
pint of mashed potato, and the same quantity of squash, it was as if I
had not dined, and I gave up the experiment. But my elder brother,
who had carried to the extremest extreme this “potato gospel,” as
Carlyle called it, induced my mother to make his Thanksgiving
squash-pie after a receipt of his own. The crust was made of Indian
meal and water, and the filling was of squash, water, and sugar! And
he ate it, and called it good. But I thought then, and still think, that his
enjoyment of the eating was in the principle rather than in the pie.
A few of the girls were interested in phrenology; and we had our
heads examined by Professor Fowler, who, if not the first, was the
chief exponent of this theory in Lowell. He went about into all the
schools, examining children’s heads. Mine, he said, “lacked
veneration;” and this I supposed was an awful thing, because my
teacher looked so reproachfully at me when the professor said it.
A few were interested in Mesmerism; and those of us who had
the power to make ourselves en rapport with others tried
experiments on “subjects,” and sometimes held meetings in the
evening for that purpose.
The life in the boarding-houses was very agreeable. These
houses belonged to the corporation, and were usually kept by
widows (mothers of mill-girls), who were often the friends and
advisers of their boarders.
Among these may be mentioned the mothers of Lucy Larcom; the
Hon. Gustavus Vasa Fox, once Assistant Secretary of the Navy;
John W. Hanson, D.D.; the Rev. W. H. Cudworth; Major General B.
F. Butler; and several others.
Each house was a village or community of itself. There fifty or
sixty young women from different parts of New England met and
lived together. When not at their work, by natural selection they sat in
groups in their chambers, or in a corner of the large dining-room,
busy at some agreeable employment; or they wrote letters, read,
studied, or sewed, for, as a rule, they were their own seamstresses
and dressmakers.
It is refreshing to remember their simplicity of dress; they wore no
ruffles and very few ornaments. It is true that some of them had gold
watches and gold pencils, but they were worn only on grand
occasions; as a rule, the early mill-girls were not of that class that is
said to be “always suffering for a breast-pin.” Though their dress was
so simple and so plain, yet it was so tasteful that they were often
accused of looking like ladies; the complaint was sometimes made
that no one could tell the difference in church between the factory-
girls and the daughters of some of the first families in the city.
Mrs. Sarah J. Hale, in The Lady’s Book, in 1842, speaking of the
impossibility of considering dress a mark of distinction, says: “Many
of the factory-girls wear gold watches and an imitation at least of all
the ornaments which grace the daughters of our most opulent
citizens.”
The boarding-houses were considered so attractive that
strangers, by invitation, often came to look in upon them, and see for
themselves how the mill-girls lived. Dickens, in his “American Notes,”
speaks with surprise of their home life. He says, “There is a piano in
a great many of the boarding-houses, and nearly all the young ladies
subscribe to circulating libraries.” There was a feeling of esprit de
corps among these households; any advantage secured to one of
the number was usually shared by others belonging to her set or
group. Books were exchanged, letters from home were read, and
“pieces,” intended for the Improvement Circle, were presented for
friendly criticism.
There was always a best room in the boarding-house, to entertain
callers in; but if any of the girls had a regular gentleman caller, a
special evening was set apart each week to receive him. This room
was furnished with a carpet, sometimes with a piano, as Dickens
says, and with the best furniture, including oftentimes the relics of
household treasures left of the old-time gentility of the house-mother.
This mutual acquaintanceship was of great advantage. They
discussed the books they read, debated religious and social
questions, compared their thoughts and experiences, and advised
and helped one another. And so their mental growth went on, and
they soon became educated far beyond what their mothers or their
grandmothers could have been. The girls also stood by one another
in the mills; when one wanted to be absent half a day, two or three
others would tend an extra loom or frame apiece, so that the absent
one might not lose her pay. At this time the mule and spinning-jenny
had not been introduced; two or three looms, or spinning-frames,
were as much as one girl was required to tend, more than that being
considered “double work.”
The inmates of what may be called these literary households
were omniverous readers of books, and were also subscribers to the
few magazines and literary newspapers; and it was their habit, after
reading their copies, to send them by mail or stage-coach to their
widely scattered homes, where they were read all over a village or a
neighborhood; and thus was current literature introduced into by and
lonely places.
From an article in The Lowell Offering, (“Our Household,” signed
H. T.,) I am able to quote a sketch of one factory boarding-house
interior. The author said, “In our house there are eleven boarders,
and in all thirteen members of the family. I will class them according
to their religious tenets as follows: Calvinist Baptist, Unitarian,
Congregational, Catholic, Episcopalian, and Mormonite, one each;
Universalist and Methodist, two each; Christian Baptist, three. Their
reading is from the following sources: They receive regularly fifteen
newspapers and periodicals; these are, the Boston Daily Times, the
Herald of Freedom, the Signs of the Times, and the Christian Herald,
two copies each; the Christian Register, Vox Populi, Literary
Souvenir, Boston Pilot, Young Catholic’s Friend, Star of Bethlehem,
and The Lowell Offering, three copies each. A magazine, one copy.
We also borrow regularly the Non-Resistant, the Liberator, the
Lady’s Book, the Ladies’ Pearl, and the Ladies’ Companion. We
have also in the house what perhaps cannot be found anywhere else
in the city of Lowell,—a Mormon Bible.”
The “magazine” mentioned may have been The Dial, that
exponent of New England Transcendentalism, of which The Offering
was the humble contemporary. The writer adds to her article:
“Nothwithstanding the divers faiths embraced among us, we live in
much harmony, and seldom is difference of opinion the cause of
dissensions among us.”
Novels were not very popular with us, as we inclined more to
historical writings and to poetry. But such books as “Charlotte
Temple,” “Eliza Wharton,” “Maria Monk,” “The Arabian Nights,” “The
Mysteries of Udolpho,” “Abellino, the Bravo of Venice,” or “The
Castle of Otranto,” were sometimes taken from the circulating library,
read with delight, and secretly lent from one young girl to another.
Our religious reading was confined to the Bible, Baxter’s “Saints’
Rest,” “The Pilgrim’s Progress,” “The Religious Courtship,” “The
Widow Directed,” and Sunday-school books.
It was fortunate for us that we were obliged to read good books,
such as histories, the English classics, and the very few American
novels that were then in existence. Cheap editions of Scott were but
just publishing; “Pickwick,” in serial numbers, soon followed;
Frederika Bremer was hardly translated; Lydia Maria Child was
beginning to write; Harriet Beecher Stowe was busy in her nursery,
and the great American novel was not written,—nor yet the small
one, which was indeed a blessing!
There were many representative women among us who did not
voice their thoughts in writing, and whose names are not on the list
of the contributers to The Offering. This was but one phase of their
development, as many of them have exerted a widespread influence
in other directions. They graduated from the cotton-factory, carrying
with them the results of their manual training; and they have done
their little part towards performing the useful labor of life. Into
whatever vocation they entered they made practical use of the habits
of industry and perseverance learned during those early years, and
they have exemplified them in their stirring and fruitful lives.
In order to show how far the influence of individual effort may
extend, it will be well to mention the after-fate of some of them. One
became an artist of note, another a poet of more than local fame, a
third an inventor, and several were among the pioneers in Florida, in
Kansas, and in other Western States. A limited number married
those who were afterwards doctors of divinity, major-generals, and
members of Congress; and these, in more than one instance, had
been their work-mates in the factory.
And in later years, when, through the death of the bread-winner,
the pecuniary support of those dependent on him fell to their lot,
some of these factory-girls carried on business, entered the trades,
or went to college and thereby were enabled to practise in some of
the professions. They thus resumed their old-time habit of supporting
the helpless ones, and educating the children of the family.
These women were all self-made in the truest sense; and it is well
to mention their success in life, that others, who now earn their living
at what is called “ungenteel” employments, may see that what one
does is not of so much importance as what one is. I do not know why
it should not be just as commendable for a woman who has risen to
have been once a factory-girl, as it is for an ex-governor or a major-
general to have been a “bobbin-boy.” A woman ought to be as proud
of being self-made as a man; not proud in a boasting way, but proud
enough to assert the fact in her life and in her works.
All these of whom I speak are widely scattered. I hear of them in
the far West, in the South, and in foreign countries, even so far away
as the Himalaya Mountains. But wherever they may be, I know that
they will join with me in saying that the discipline of their youth
helped to make them what they are; and that the cotton-factory was
to them the means of education, their preparatory school, in which
they learned the alphabet of their life-work.
Such is the brief story of the life of every-day working-girls; such
as it was then, so it might be to-day. Undoubtedly there might have
been another side to this picture, but I give the side I knew best,—
the bright side!
CHAPTER VI.
THE LOWELL OFFERING AND ITS WRITERS.
One of the most curious phases in the life of New England, and
one that must always puzzle the historian of its literature, is its
sudden intellectual blossoming half a century ago.
Emerson says, “The children of New England between 1820 and
1840 were born with knives in their brains;” and this would seem to
be true, since during or very near that time, were born the majority of
those writers and thinkers whose lives have been so recently and so
nobly rounded out,—Emerson, Bryant, Longfellow, Lowell, Whittier,
John Pierpont,—they whose influence cannot be overestimated in
bringing an ideal element into our hitherto prosaic New England life.
The seeds of this intellectual growth came suddenly, as if blown
from some far-off cultured land, and were sown broadcast. Some
found a resting-place in this little corner of New England, where were
gathered together these daughters of Puritan ancestors, and they,
too, feeling the intellectual impetus, were impelled to put in writing
their own crude thoughts. Their desire for self-improvement had
been to some extent gratified, and they now began to feel the benefit
of the educational advantages which had been opened to them. As
in “Mary Barton,” they “threw the shuttle with increasing sound,
although Newton’s ‘Principia’ lay open before them, to be snatched
at in work-hours, but revelled over at meal-time or at night.”
And the “literary” girls among us would often be seen writing on
scraps of paper which we hid “between whiles” in the waste-boxes
upon which we sat while waiting for the looms or frames to need
attention. Some of these studious ones kept note-books, with
abstracts of their reading and studies, or jotted down what they were
pleased to call their “thoughts.” It was natural that such a thoughtful
life should bear fruit, and this leads me to speak of The Lowell
Offering, a publication which was the natural outgrowth of the mental
habit of the early mill-girls, for many of the pieces that were printed
there were thought out amid the hum of the wheels, while the skilful
fingers and well-trained eyes of the writers tended the loom or the
frame.
The idea of organization for literary and educational purposes
was first proposed in 1837 by Miss Harriot F. Curtis, perhaps the
most progressive of all the mill-girls. She with her immediate
associates conceived the idea of forming a little society for mental
improvement. In The Lowell Offering of January, 1845, is the
following account of its formation written by Miss Maria Currier.
“IMPROVEMENT CIRCLE.
“In one of the corporations [the Lawrence] of this city, about eight
years ago, might have been seen, on a summer evening, a company
of four or five young females, who through the day had labored at
their several employments in some one of the factories connected
with the corporation. Perhaps they were not ambitious above others
of their sex.... But wishing to improve the talents which God had
given them, they proposed the formation of a society for mutual
improvement. An evening was appointed for the proposed purpose;
and having invited a few others to join them, they met at the time
appointed.... A president, vice-president, and secretary were chosen;
a constitution was drafted, and by-laws formed, to which each of the
members affixed her name.... At length a circle on a more extensive
scale was formed by a gentleman of this city, and a plan conceived
of bringing before the world the productions of inexperienced
females; of showing that intellect and intelligence might be found
even among factory operatives. It was then that The Offering was
published; and many of those who were present at the first meeting
of our Improvement Circle were contributors to its pages.”
At the first meeting, Miss Curtis delivered a stirring address, in

which she stated the object and scope of the organization, and the
urgent need that existed for all working-women to make an effort to
improve their minds.
The club met fortnightly, and each member contributed articles in
prose and verse, which were read at the meetings, and subjected to
the criticism of those present.
In answer to a letter of enquiry, Miss Curtis writes: “I do not
remember who composed the first circle, not even the names of the
officers; but I think Emmeline Larcom was secretary. Farther than
that I can only say, I was not anything. I never would hold any office,
—office brings trammels. I believe I wrote and read the address of
which Maria speaks. Louisa and Maria Currier, Emmeline Larcom,
Harriet Lees, and possibly Ann Carter were there.... If you want to
know whose brain conceived the idea, I suspect it was I. I was
always daring; the other five were modest and retiring.” And thus
was formed the first woman’s literary club in this country,—a remote
first cause of the hundreds which now make up the General
Federation of Women’s Clubs, since it bears the same relation to
that flourishing institution as the native crab does to the grafted tree.
Some of these early club, or improvement circle women either are,
or have been, members of similar organizations in the localities in
which they live, and have done their best to incorporate into the
constitution of the modern woman’s club the idea of “improving the
talents God has given them.” And if they have continued to live up to
this doctrine, no doubt they have attained, if not to all they may have
desired, at least to all they were capable of achieving, according to
their limitations.
It may be well to mention here that Improvement Circles
continued to be formed, and that in 1843 there were at least five in
different parts of the city. I attended one in 1845, connected with The
Lowell Offering. It met in the publication office, on Central Street, and
was well filled with factory operatives, some of whom had brought
their contributions, and waited to hear them read, with quaking
hearts and conscious faces. Harriet Farley presided, and from a pile
of manuscript on the table before her selected such contributions as
she thought the most worthy of a public reading. Among these, as I
remember, were the chapters of a novel by Miss Curtis, one of Lucy
Larcom’s prose poems, and some “pieces of poetry.” Included in
these pieces were some verses in which the wind was described as
playing havoc with nature to such an extent that—
“It took the tall trees by the hair,

And as with besoms swept the air.”
This tremendous breeze, or simile, caused a good deal of mirth

among the younger contributors, who had never heard of “The
World-Soul,” nor read Emerson’s line—
“To the green-haired forest free,”
nor Longfellow’s “The Building of the Ship,” where he speaks of the

pine-trees as—
“Shorn of their streaming hair.”
Nor yet Wordsworth’s sonnet:—
“While trees, dim seen, in frenzied numbers tear

The lingering remnant of their yellow hair.”
This was my only appearance at the Circle, as I had hitherto been

deterred from going by the knowledge that those who went were
expected to bring a written contribution to be read there. Shortly after
this, Miss Farley (one of the editors) invited me to send something to
the magazine, and I complied; but I was not an early or a constant
contributor.
In 1839, the Rev. Abel C. Thomas and the Rev. Thomas B.
Thayer, pastors of the First and Second Universalist Churches in
Lowell, established improvement circles composed of the young
people belonging to their respective parishes. These meetings were
largely made up of young men and women who worked in the mill.
They were often asked to speak; but as they persistently declined,
they were invited to write what they desired to say, and send it, to be
read anonymously at the next meeting. Many of the young women
complied with this request, and these written communications were
so numerous that they very soon became the sole entertainment of
what Mr. Thomas called “these intellectual banquets.”
A selection from the articles read at these meetings was
published by Mr. Thomas in pamphlet form, under the title, “The
Lowell Offering, a Repository of Original Articles written by Females
employed in the Mills.” Mr. Thomas’s own account of his part in
establishing the magazine will be found in chapter seven. The first
series, of four numbers, was issued from October, 1840, to March,
1841; and there was such a demand for copies, that a new series
began, The Lowell Offering proper, a monthly magazine of thirty-two
pages, which was issued regularly by its projector from that time until
October, 1842, when it passed into the hands of Miss Harriot F.
Curtis and Miss Harriet Farley, both operatives in the Lowell mills.
Under their joint editorship it was published, the first year by
William Schouler, but after that by these ladies themselves, who
were editors, publishers, and proprietors, until December, 1845,
when, with the end of Volume V. Miss Curtis retired from the
magazine, and The Lowell Offering ceased to exist.
But in September, 1847, Miss Farley resumed the publication of
the magazine and issued one copy under the title The New England
Offering; and all those who were or had been factory operatives
were invited to contribute to its pages.
This magazine was re-issued in 1848, from April to December,
continued through 1849, and until March, 1850, when it was
discontinued for want of means, and perhaps new contributors. Miss
Farley was the editor, publisher, and proprietor of The New England
Offering.
There are about seven volumes of the magazines in all,—five of
The Lowell Offering, and two of The New England Offering, including
the first four numbers in 1840, and the odd numbers of 1847 and
1850.
The prospectus of The Lowell Offering, as issued by its women-
editors in 1845, is as follows:—
THE
LOWELL OFFERING,
WRITTEN, EDITED, AND PUBLISHED
BY FEMALE OPERATIVES.
Our magazine is the only one which America has produced, of
which no other country has produced the like. The Offering is prima
facie evidence, not only of the American “factory-girls,” but of the
intelligence of the mass of our country. And it is in the intelligence of
the mass that the permanency of our republican institutions
depends.
And our last appeal is to those who should support us, if for no
other reason but their interest in “the cultivation of humanity,” and the
maintenance of true democracy. There is little but this of which we,
as a people, can be proud. Other nations can look upon the relics of
a glory come and gone—upon their magnificent ruins—upon worn-
out institutions, not only tolerated, but hallowed because they are old
—upon the splendors of costly pageant—upon the tokens of a
wealth, which has increased for ages—but we can take pride in
these. We have other and better things. Let us look upon our “free
suffrage,” our Lyceums, our Common Schools, our Mechanics’
Literary Associations, the Periodical of our Laboring Females; upon
all that is indigenous to our Republic, and say, with the spirit of the
Roman Cornelia, These, these are our jewels.
Terms: One dollar per year in advance. Postage: 100 miles and
under, 1½ cents. Over 100, 2½ cents.
Published at Lowell, Mass., monthly, by
MISSES CURTIS & FARLEY.
In order to combat the prejudice which then existed against

“female” editors and publishers, it was thought best (as Mr. Thomas
had advised) that the enterprise should be indorsed by some of the
leading men of the city; and in the original document, now before
me, these gentlemen said:—
“We wish herewith to express most cheerfully our confidence in

their talents and moral worth, and our cordial approbation of the
worthy enterprise in which they are engaged.... We wish only to
witness to all to whom this may come, that Miss Harriet Farley and
Miss Harriot Curtis are worthy of entire confidence, and are

Ebook Robust Engineering Designs of Partial Differential Systems and Their Applications 1St Edition Bor Sen Chen Online PDF All Chapter

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ebook Robust Engineering Designs of Partial Differential Systems and Their Applications 1St Edition Bor Sen Chen Online PDF All Chapter

Uploaded by

Copyright:

Available Formats

Robust Engineering Designs of Partial

Differential Systems and Their

Systems Evolutionary Biology Biological Network

Systems Immunology and Infection Microbiology 1st

Evolutionary Equations: Picard's Theorem for Partial

Fundamentals of Data Engineering: Plan and Build Robust

Fourier Analysis and Partial Differential Equations 1st

Building Integrated Photovoltaic Thermal Systems:

Building Integrated Photovoltaic Thermal Systems:

Practical Applications of Evolutionary Computation to

First edition published 2022

and by CRC Press

© 2022 Bor-Sen Chen

CRC Press is an imprint of Taylor & Francis Group, LLC

Library of Congress Cataloging‑in‑Publication Data

ISBN: 978-1-032-13443-7 (hbk)

Chapter 1 Introduction to Partial Differential Systems ................................... 3

Chapter 2 Fuzzy Solutions to Partial Differential Equations ........................ 25

PART II Robust Signal Processing Design

Chapter 3 Robust Filter Design for Linear Stochastic Partial

3.2 A General H∞ Filtering Setting for Linear Stochastic

Chapter 4 Robust Filter Design for Nonlinear Stochastic Partial

Chapter 5 Robust Synchronization Design of Partial Differential

PART III Robust Control System Design

Chapter 6 Robust Stabilization Control Design of Large Structural

6.5 Stabilization of LSS with Constrained Controls .............. 158

Chapter 7 Robust Observer-Based Output Feedback Control Design of

Chapter 8 Robust Stabilization Design for Stochastic Linear Partial

Chapter 9 Fuzzy State-Space Modeling and Robust Observer-Based

9.3.2 Infinite-Dimensional T-S Fuzzy State-Space

Chapter 10 Robust Tracking Control Design of Nonlinear Distributed

Chapter 11 Robust Stabilization Control Design of Nonlinear Stochastic

11.5 Simulation Example ......................................................... 336

Chapter 12 Robust Fuzzy H∞ Estimator-Based Stabilization Design

Chapter 13 Low Design-Cost Fuzzy Controllers for Robust Stabilization

bases to approximate a nonlinear PDS by adaptive fuzzy algorithm. In Part II,

MATLAB® is a registered trademark of The MathWorks, Inc. For product infor-

1.1 PARTIAL DIFFERENTIAL EQUATIONS AND

1.1.1 Partial Differential equations

where x1 , x 2 ,, x N are the independent variables with N ∈ and

cb2 ( x1p , x 2p ) − 4ca ( x1p , x 2p )cc ( x1p , x 2p ) (1.3)

at a point ( x1p , x 2p ). An equation is said to be hyperbolic, parabolic, or elliptic

1.1.2 Partial Differential system

on the spatio-temporal domain Ω × [0, T ], where x = [ x1 , x2 , , x N ]T ∈Ω ⊂  N is

where y ∈ D( Α) is a known diffusion coefficient matrix. Similar to the bound-

i. Dirichlet boundary condition (D):

i. Diffusion coefficient D, also called diffusivity, is an important param-

analytically, depending on the boundary conditions. However, consider the linear

where u( x ,t )  [u1 ( x , t),u2 ( x ,t ), , unu ( x , t )]T ∈ nu is the control force to be

where n( x , t)  [n1 ( x ,t ), n2 ( x , t), , nnn ( x , t )]T ∈ L2 (Ω ×  + ;  nn ) represents the

1.2 NUMERICAL AND APPROXIMATION

1.2.1 Galerkin methoD

with the boundary condition

where A is a self-adjoint differential operator in the Hilbert space

Rn (a1 , a2 , an , x1 , x 2 ) = Ayn = Ay0 + ∑ a Aφ

Since A is linear, this equation (1.19) can be written as

which determines ai ’s. Substitution of the ai ’s obtained from the solution of

with a special restriction

Thus, the application of the Galerkin method to equation (1.15) gives

so the Galerkin solution equation (1.21) becomes

and the property of ellipticity

with the boundary condition

y = 0 on ∂Ω = { x1 = ±a, x 2 = ±b} (1.30)

on the spatio-temporal domain Ω × [0, T ], where x = [ x1 , x2 , , x N ]T ∈Ω ⊂ N is

where u( x ,t ) [u1 ( x , t),u2 ( x ,t ), , unu ( x , t )]T ∈ nu is the control force to be

where n( x , t) [n1 ( x ,t ), n2 ( x , t), , nnn ( x , t )]T ∈ L2 (Ω × + ; nn ) represents the

y(t ) [ y1 (t ), , y∞ (t ) ] ∈ ∞ , A diag ( λ1 + a, , λ∞ + a ) ∈ ∞×∞ ,

Uc (t ) [ u1 (t ), , unc (t ) ] , Ur (t ) [ unc +1 (t ), , u∞ (t ) ] ,

Yk ,l − yk ,l = (Yk ,l − y k ,l ) + ( yk ,l − yk ,l ) = d k ,l − rk ,l (1.70)

TjY (t ) Tx1 , j ,1 + Tx2 , j ,1 , , Tx1 , j , NY + Tx2 , j , NY  Y (t )

Nor yet Wordsworth’s sonnet:—