Introduction to Optimization

for Environmental and Chemical

Engineers
 Introduction to Optimization
for Environmental and Chemical
Engineers

By
Louis Theodore and Kelly Behan
CRC Press
Taylor & Francis Group
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 33487-2742

© 2018 by Taylor & Francis Group, LLC

CRC Press is an imprint of Taylor & Francis Group, an Informa business

No claim to original U.S. Government works

Printed on acid-free paper

International Standard Book Number-13: 978-1-138-48912-7 (Hardback)

This book contains information obtained from authentic and highly regarded sources. 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 copy-
right 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 pho-
tocopying, 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, please access www.copyright.com (http://
www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA
01923, 978-750-8400. CCC is a not-for-profit organization that provides licenses and registration for a variety of users.
For organizations that have been granted a photocopy license by the CCC, a separate system of payment has been
arranged.

Trademark Notice:  Product or corporate names may be trademarks or registered trademarks, and are used only for
identification and explanation without intent to infringe.

Visit the Taylor & Francis Web site at 

http://www.taylorandfrancis.com

and the CRC Press Web site at 

http://www.crcpress.com
 TO

Donald J. Trump

who may yet save our great nation from

enemies within and abroad (L.T.)

My mother and father

who inspire me to be extraordinary (K.B.)

Contents

Preface............................................................................................................................................ xiii
About the Book............................................................................................................................ xvii
Authors.......................................................................................................................................... xix

Part I  Optimization Fundamentals and Principles

1 Optimization Overview.........................................................................................................3
1.1 History of Optimization...............................................................................................4
1.2 The Computer Age.........................................................................................................6
1.3 The Scope of Optimization...........................................................................................7
1.4 Conventional/Established Optimization Procedures..............................................7
1.5 Contemporary Optimization: Linear Programming................................................ 8
References................................................................................................................................ 11

2 Mathematical Operations.................................................................................................... 13
2.1 The Quadratic Equation.............................................................................................. 15
2.2 Interpolation and Extrapolation................................................................................ 17
2.3 Significant Figures and Approximate Numbers..................................................... 20
2.4 Errors.............................................................................................................................22
2.5 Differentiation.............................................................................................................. 23
2.6 Numerical Integration................................................................................................. 26
2.6.1 Trapezoidal Rule............................................................................................. 27
2.6.2 Simpson’s Rule................................................................................................ 28
2.7 Simultaneous Linear Algebraic Equations............................................................... 29
2.7.1 Gauss–Jordan Reduction...............................................................................30
2.7.2 Gauss Elimination..........................................................................................30
2.7.3 Gauss–Seidel Approach................................................................................. 31
2.8 Nonlinear Algebraic Equations................................................................................. 31
2.9 Ordinary Differential Equations............................................................................... 32
2.10 Partial Differential Equations.................................................................................... 37
2.10.1 Parabolic PDE.................................................................................................. 37
2.10.2 Elliptical PDE................................................................................................... 39
References................................................................................................................................ 40

3 Perturbation Techniques...................................................................................................... 41
3.1 One Independent Variable..........................................................................................42
3.2 Two Independent Variables........................................................................................ 45
3.3 Three Independent Variables..................................................................................... 48
References................................................................................................................................ 49

4 Search Methods..................................................................................................................... 51
4.1 Interval Halving Method............................................................................................ 53
4.2 The Bisection Method.................................................................................................. 56
vii
viii Contents

4.3 The Golden Section Method....................................................................................... 57

4.4 Method of Steepest Ascent/Descent......................................................................... 60
References................................................................................................................................63

5 Graphical Approaches..........................................................................................................65
5.1 Rectangular Coordinates............................................................................................ 66
5.1.1 Quadrants........................................................................................................ 67
5.1.2 Linear versus Non-Linear.............................................................................. 67
5.1.3 Intercepts.......................................................................................................... 68
5.2 Logarithmic-Logarithmic (Log-Log) Coordinates.................................................. 68
5.3 Semi-Logarithmic (Semi-Log) Coordinates............................................................. 73
5.3.1 Plotting Straight Lines on Semi-Logarithmic Paper.................................. 73
5.3.2 Other Graphical Coordinates........................................................................ 76
5.4 Methods of Plotting Data............................................................................................ 76
5.5 Applications..................................................................................................................77
References................................................................................................................................ 81

6 Analytical Methods...............................................................................................................83
6.1 Breakeven Considerations..........................................................................................83
6.2 One Independent Variable..........................................................................................85
6.3 General Analytical Formulation of the Optimum.................................................. 88
6.4 Two Independent Variables........................................................................................ 89
6.5 Three Independent Variables..................................................................................... 93
References................................................................................................................................ 94

7 Linear Programming............................................................................................................. 95
7.1 Definitions..................................................................................................................... 96
7.2 Basic Concepts of Optimization................................................................................. 97
7.3 Applied Mathematical Concepts in Linear Programming.................................... 99
7.4 Applied Engineering Concepts in Linear Programming.................................... 102
7.5 Other Real-World Applied Concepts in Linear Programming........................... 105
References.............................................................................................................................. 107

Part II  Environmental Engineering Applications

8 Air Pollution Management................................................................................................ 111

8.1 Outdated Air Pollution Control Device.................................................................. 111
8.2 Coal-Fired Power Plant Options.............................................................................. 113
8.3 Particulate Control Equipment Options................................................................. 114
8.4 Recovering Dust......................................................................................................... 117
8.5 Incinerating Mercury-Contaminated Waste.......................................................... 119
8.6 Optimizing Incinerator Performance...................................................................... 121
References.............................................................................................................................. 123

9 Water Pollution Management........................................................................................... 125

9.1 Break-Even Point Operation..................................................................................... 125
9.2 Wastewater Treatment Equipment Options........................................................... 126
Contents ix

9.3 Water Disinfection Options...................................................................................... 127

9.4 Water Plant Operation............................................................................................... 131
9.5 Generating Waste Revenues from Waste Streams................................................ 132
9.6 Membrane Construction Costs................................................................................ 133
References.............................................................................................................................. 135

10 Solid Waste Management.................................................................................................. 137

10.1 Solid Waste Treatment............................................................................................... 137
10.2 Cement Kiln versus Rotary Kiln Incinerator......................................................... 139
10.3 Sludge Waste Receiving Tank Size.......................................................................... 142
10.4 Discounted Cash Flow Application......................................................................... 144
10.5 Maximum Mercury Waste Emission Constraints................................................. 147
10.6 Structural Considerations of Concrete Mix........................................................... 149
References.............................................................................................................................. 151

11 Health Risk Assessment..................................................................................................... 153

11.1 Minimum Purging Time for a Vessel to Permit Entry......................................... 153
11.2 Minimizing the Cost of Cancer-Fighting Drugs................................................... 155
11.3 Reducing Insurance Costs........................................................................................ 157
11.4 Reducing Health Concerns Associated with Flue Gas Emissions...................... 158
11.5 Vitamin Caloric and Protein Requirements........................................................... 160
11.6 Pharmaceutical Cancer Drug................................................................................... 161
References.............................................................................................................................. 163

12 Hazard Risk Assessment................................................................................................... 165

12.1 Safety Expenditures versus Return of Safety Investments................................. 165
12.2 Plant Safety Proposal Comparisons........................................................................ 167
12.3 Effect of Failure Distribution Rate........................................................................... 169
12.4 Estimating the Effect of Worst-Case Explosions...................................................172
12.5 Economics of Fire Protection Equipment............................................................... 174
12.6 Optimizing the Operation of a Cement Incinerator............................................. 177
References.............................................................................................................................. 180

Part III  Chemical Engineering Applications

13 Fluid Flow Applications..................................................................................................... 183

13.1 Fan Selection............................................................................................................... 183
13.2 Pump Selection........................................................................................................... 185
13.3 Pipe Diameter Selection............................................................................................ 188
13.4 Two-Stage Compressor.............................................................................................. 189
13.5 Ventilation Models..................................................................................................... 191
13.6 Three-Stage Compressor........................................................................................... 195
References.............................................................................................................................. 197

14 Chemical Reactor Applications........................................................................................ 199

14.1 Comparing a Continuous Stirred Tank Reactor to a Tubular Flow Reactor
Performance��������������������������������������������������������������������������������������������������������������� 199
x Contents

14.2 Optimizing Fluidized-Bed Reactor Performance.................................................. 204

14.3 Two Reactors in Series: Achieving Maximum Conversion.................................. 208
14.4 Maximizing Selectivity............................................................................................. 210
14.5 Optimizing Batch Reactor Performance................................................................. 212
14.6 Optimizing Operating Schedules............................................................................ 214
References.............................................................................................................................. 215

15 Mass Transfer Applications............................................................................................... 217

15.1 Optimum Reflux Ratio.............................................................................................. 217
15.2 Optimizing a Filter Press Operation....................................................................... 219
15.3 Optimizing Production Rates.................................................................................. 221
15.4 Process and Equipment Modifications................................................................... 223
15.5 Economic Analysis for an Absorber versus an Extractor..................................... 227
15.6 Processing Crude Oil................................................................................................. 232
References.............................................................................................................................. 233

16 Heat Transfer Applications................................................................................................ 235

16.1 Steam Options............................................................................................................ 235
16.2 Optimum Insulation Thickness............................................................................... 238
16.3 Selecting the Most Profitable Exchanger................................................................ 239
16.4 Critical Insulation Thickness................................................................................... 242
16.5 Recovering Quality Energy...................................................................................... 246
16.6 Maximizing Profit through Energy Recovery....................................................... 247
References.............................................................................................................................. 248

17 Plant Design Applications................................................................................................. 249

17.1 Shipping Facilities...................................................................................................... 249
17.2 Tank Farms................................................................................................................. 252
17.3 Cyclone Selection and Design..................................................................................254
17.4 Minimizing the Cost of a Batch Plant Operation..................................................258
17.5 Ventilation Models with System Variables............................................................. 259
17.6 Plant Structure Design..............................................................................................261
References.............................................................................................................................. 262

Part IV  Select Optimization Applications

18 Select Environmental Engineering Applications......................................................... 265

18.1 Air Management........................................................................................................ 265
18.2 Water Management.................................................................................................... 268
18.3 Solid Waste Management......................................................................................... 271
18.4 Health Risk Assessment............................................................................................ 273
18.5 Hazard Risk Assessment.......................................................................................... 279
18.5.1 DRaT II Model............................................................................................... 281
18.5.2 DRaT IV Model.............................................................................................. 282
18.5.3 DRaT V Model............................................................................................... 283
References.............................................................................................................................. 283
Contents xi

19 Select Chemical Engineering Applications................................................................... 285

19.1 Fluid Flow................................................................................................................... 285
19.2 Chemical Reactors...................................................................................................... 287
19.3 Mass Transfer Operations......................................................................................... 293
19.4 Heat Exchangers......................................................................................................... 295
19.5 Plant Design................................................................................................................ 297
References..............................................................................................................................300

20 Advanced Optimization Term Projects.......................................................................... 303

20.1 Sulfuric Acid Process................................................................................................. 303
20.2 Optimum Indoor Ventilation Flow Rates...............................................................304
20.3 Fluid Transportation System....................................................................................306
20.4 Mass Transfer Application........................................................................................ 311
References.............................................................................................................................. 315
Index.............................................................................................................................................. 317
Preface

It is no secret that in recent years, the number of people entering the engineering field has
increased. Most are beginning college students and some had earlier chosen a non-tech-
nical major/career path. A large number of these individuals are today seeking degrees in
environmental and/or chemical engineering. These prospective students will require an
understanding and appreciation of the numerous mathematical and optimization meth-
ods that are routinely employed in practice. This technical stepping stone to a successful
career is rarely provided at institutions that award engineering degrees. This introductory
text on optimization attempts to supplement existing curricula with a sorely needed tool
to eliminate this void.
More on the need for this book. The question often arises as to the educational back-
ground required for engineering students to possess meaningful analysis capabilities since
technology has changed the emphasis that is placed on certain technical subjects. Before
computer usage became popular, instruction in engineering analysis was (and still is in
many places) restricted to simple systems, and most of the effort was devoted to solving a
few elementary equations that were derived. These cases were mostly of academic inter-
est, and because of their simplicity, were of little practical value. To this end, a considerable
amount of time is now required to acquire skills in mathematics, especially in numerical
methods, statistics, and optimization. In fact, most engineers are given courses in classi-
cal mathematics, but experience shows that very little of this knowledge is retained after
graduation for the simple reason that these mathematical methods are not adequate for
solving most systems of equations encountered in industry. In addition, advanced math-
ematic skills are either not provided in courses or are forgotten through sheer disuse since
they are not related to engineering practice.
The material in this book was prepared primarily for environmental and chemical
engineering students and, to a lesser extent, for engineering professionals who wish to
obtain a better understanding of the various optimization methods that can be employed
in solving technical problems. In presenting the text material, the authors have stressed
the pragmatic approach in the application of mathematical tools to later assist the reader in
grasping the role of optimization in engineering problem-solving situations.
As noted previously, this book serves two purposes. It may be used as a textbook for
beginning environmental and chemical students or as a “reference” book for practicing
engineers involved with optimization applications. For both audiences, it is assumed that
the reader has already taken basic courses in physics and chemistry, and should have a
minimum background in mathematics through elementary calculus. The authors’ ulti-
mate aim is to offer the reader the fundamentals of several optimization methods with
accompanying practical engineering applications. The reader is encouraged through the
references to continue his or her own development beyond the scope of the presented
material.
As is usually the case in preparing any text, questions of what to include and what to
omit have been particularly difficult. As noted above, the material in this book attempts
to address optimization calculations common to both the environmental and chemi-
cal engineering disciplines. The book provides the reader with numerous solved illus-
trative examples so that the interrelationship between both theory and application
is emphasized in nearly all of the chapters. One key feature of this book is that the

xiii
xiv Preface

solutions to the problems are presented in a stand-alone manner. Throughout the book,
the illustrative examples are laid out in such a way as to develop the reader’s technical
understanding of optimization, with more difficult examples located at or near the end
of each chapter.
The book is divided into four (IV) parts (see also Table of Contents):

Part I: Optimization Fundamentals and Principles

Part II: Environmental Engineering Applications
Part III: Chemical Engineering Applications
Part IV: Select Industrial Applications

It should be noted that Part IV was an afterthought. After completing the manuscript,
the authors felt that it lacked some real-world applications drawn from both the envi-
ronmental and chemical engineering industries. As such, three chapters were added to
this work—one concerned with environmental engineering, one concerned with chemical
engineering, and one concerned with term projects. Chapter 18, titled Select Environmental
Engineering Applications, is an extension of Part II and contains five sections—air man-
agement, water management, solid waste management, health risk assessment, and haz-
ard risk assessment. Chapter  19, titled Select Chemical Engineering Applications, is on
the other hand an extension of Part III and also contains five sections—fluid flow, chemi-
cal reactors, mass transfer applications, heat transfer applications, and plant design.
Chapter 20 contains four sections that address more advanced optimization material.
Most chapters contain a short introduction to the topic in question, which is followed by
developmental material, which, in turn, is followed by several illustrative examples. Since
this book offers material not only to individuals with limited technical background but
also to those with extensive engineering industrial experience, this book can be used as a
text in either a general introductory optimization course or (perhaps) as a training tool in
industry for challenged engineering professionals.
Hopefully, the text is simple, clear, to the point, and imparts a basic understanding of the
theory and application of some of the important optimization methods employed in prac-
tice. It should also assist the reader in helping master the difficult task of explaining what
was once a very complicated subject matter in a way that is easily understood. The authors
feel that this distinguishes this text from the numerous others in this field.
Is should also be noted that the authors have long advocated that basic science
courses—particularly those concerned with mathematics—should be taught to engi-
neers by an engineer; and the books adopted for use in these courses should also be
written by an engineer. For example, a mathematician may lecture on differentiation—
say dy /dx —not realizing that in a real-world application involving an estuary, y  could
refer to concentration while x  could refer to time. The readers of this book will not
encounter this problem.
The reader should also note that parts of the material in the book were drawn from one
of the author’s notes of yesteryear. In a few instances, the original source was not available
for referencing purposes. Any oversight will be corrected in a later printing/edition.
The authors wish to express appreciation to those who have contributed suggestions for
material covered in this book. Their comments have been very helpful in the selection and
presentation of the subject matter. Special appreciation is extended to Drs. Walter Matystik
and Francesco Ricci.
Preface xv

Finally, the authors are especially interested in learning the opinions of those who read
this book concerning its utility and serviceability in meeting the needs for which it was
written. Corrections, improvements, and suggestions will be considered for inclusion in
any later editions.

Lou Theodore 
East Williston, NY 

Kelly Behan 
Mineola, NY 
About the Book

As noted in both the Preface and Table of Contents, this optimization book is divided into
four parts:

Part I: Optimization Fundamentals and Principles

Part II: Environmental Engineering Applications
Part III: Chemical Engineering Applications
Part IV: Select Industrial Applications

Following an optimization overview (Chapter 1) and a review of mathematical opera-

tions (Chapter  2), Part  I then examines the five optimization approaches covered in
the book: perturbation techniques (Chapter  3), search methods (Chapter  4), graphical
techniques (Chapter 5), analytical methods (Chapter 6), and the general subject of linear
programming (Chapter  7). Chapters 3–6 primarily examine traditional optimization
from an unconstrained  perspective, while Chapter 7 approaches optimization from a
constrained  perspective. The above approach provides optimization topics ranging from
the optimization of a single variable function to more detailed nonlinear constrained
problems. Note also that although the optimization methods are provided with com-
puters in mind, programming and coding are not addressed. The applications that fol-
low in the next three Parts (II, III, IV) discuss how to improve existing systems and/or
designing improved and more efficient new systems.
Part II keys on optimization applications in environmental engineering. The five topics
addressed are air management (Chapter 8), water management (Chapter 9), solid waste
management (Chapter 10), health risk assessment (Chapter 11), and hazard risk assessment
(Chapter 12). Each chapter contains a brief introduction to the subject in question, which
is then followed by a host of examples dealing with industrial applications that employ
some—if not all—of the optimization methods introduced in Part I. Also note that the
performance criteria employed can vary from classical economic choices to technological
factors, such as energy conservation, degree of purity, etc.
Part III keys on optimization applications in chemical engineering. The five types
addressed are fluid flow (Chapter 13), chemical reactors (Chapter 14), mass transfer (Chapter
15), heat transfer (Chapter 16), and plant design (Chapter 17). Each chapter contains a brief
introduction to the subject in question, which is then followed by a host of examples that
key on industrial applications that employ some—if not all—of the optimization methods
introduced in Part I.
Part IV, contains three chapters that are concerned with what the authors have termed,
“Select Industrial Applications”. The first chapter covers environmental engineering while
the second chapter keys on chemical engineering. Each of these two chapters contains five
applications—one for each of the subject areas addressed earlier in Parts II and III. The
third chapter provides more advanced optimization material that some might refer to as
term projects, and it contains four illustrative examples.
Regarding the subject of references, the authors recommend the following six texts for
what is described as “suggested reading” for topics associated with environmental and
chemical engineering:

xvii
xviii About the Book

1. R. Perry and D. Green, Perry’s Chemical Engineers’ Handbook , 8th edition, McGraw-
Hill, New York City, NY, 2008.
2. L. Theodore, Chemical Engineering: The Essential Reference , McGraw-Hill, New York
City, NY, 2014.
3. J. Reynolds, J. Jeris, and L. Theodore, Handbook of Chemical and Environmental
Engineering Calculations , John Wiley & Sons, Hoboken, NJ, 2004.
4. L. Theodore, R. Dupont, and K. Ganesan, Unit Operations in Environmental
Engineering , Wiley‑Scrivener, Salem, MA, 2018.
5. M. K. Theodore and L. Theodore, Introduction to Environmental Management , CRC
Press/Taylor & Francis, Boca Raton, FL, 2000.
6. G. Burke, B. Singh, and L. Theodore, Handbook of Environmental Management and
Technology , 2nd edition, John Wiley & Sons, Hoboken, NJ, 2000.

These six books provide extensive background information, developmental material,

illustrative examples and problems, and other references for many of the chapters in each
Part of this book.
The four recommended texts concerned with mathematical methods are

1. R. Ketter and S. Prawel, Modern Methods in Engineering Computation , McGraw-Hill,

New York City, NY, 1969.
2. H. Mickley, T. Sherwood, and C. Reed, Applied Mathematics in Chemical Engineering ,
McGraw-Hill, New York City, NY, 1957.
3. M. Moyle, Introduction to Computers for Engineers , John Wiley & Sons, Hoboken, NJ, 1967.
4. J. Happel, Chemical Process Economics , John Wiley & Sons, Hoboken, NJ, 1958.

The beginning of each chapter (not including Part IV) contains additional references that
are topic specific. The reader, particularly the beginning student, should consider review-
ing (and perhaps reading) these references before attempting to grasp the fundamentals,
principals, concepts, and so on, associated with each chapter.
Finally, an important point needs to be made. There have been numerous occasions dur-
ing one of the author’s 50-year tenure as an educator when students solved a problem
using a packaged program such as Excel, MathCAD, and so on. For example, the problem
could have involved the solution to a differential equation or the regression of some data.
On being questioned how the packaged program performed the calculation, the student
almost always responded with something to the effect of, “I don’t know, and I don’t care.”
For this reason, the reader should note that no detailed attempt was made to introduce and
explain the packaged computer programs presently available that provide easy access to
trouble-free solutions of (even) complex mathematical problems, including optimization.
The emphasis in this introductory text on optimization was to provide the reader with
an understanding of fundamental principles in order to learn how these methods can be
used to obtain answers to questions within one’s own subject matter specialty; becoming
“computer literate” in this field was not the objective. Thus, anyone privy to applicable
software can obtain detailed solutions to large and complex optimization problems by
following relatively simple instructions associated with the software/computer. It should
also be noted that significant material of an example/problem nature was drawn from
lecture notes prepared for several courses offered by one of the authors during his career
as an educator; the source of some of this material is no longer known.
Authors

Over the past 50  years, Dr. Louis Theodore was a successful educator at Manhattan
College (holding the rank of full professor of chemical engineering), graduate program
director, researcher, professional innovator, and communicator in the engineering field.
During this period, he was primarily responsible for his program achieving a number two
ranking by US News & World Report. He has authored nearly 100 text/reference books
and over 100 technical papers. He currently serves as a part-time consultant to the US
Environmental Protection Agency (EPA) and runs Theodore Tutorials. He is a member of
Phi Lambda Upsilon, Sigma Xi, Tau Beta Pi, American Chemical Society, American Society
of Engineering Education, Royal Hellenic Society, and a fellow of the International Air &
Waste Management Association (AWMA). Dr. Theodore is the recipient of the AWMA’s
prestigious Ripperton award that is “presented to an outstanding educator, who, through
example, dedication, and innovation has so inspired students to achieve excellence in their
professional endeavors.” He was also the recipient of the American Society for Engineering
Education’s (ASEE) AT&T Foundation award for “excellence in the instruction of engineer-
ing students.”

Ms. Behan was employed in the utility industry and worked in the area of develop-
ing growth jobs and analyzing site plans. She also worked for the New York City Fire
Department and was primarily involved in physiological research on World Trade Center
victims. Ms. Behan previously served as Editorial Manager for a book on basketball coach-
ing. She is currently involved with structural and civil engineering activities. In addition,
her recent activities have centered on structural engineering with Turner Construction.

xix
 Part I

Optimization Fundamentals
and Principles

Sophisticated optimization mathematical techniques are developing that enable envi-

ronmental and chemical engineers to consider, in more quantitative ways, problems con-
cerned with technology, society, and the environment. The traditional approach, classical
calculus, to determine the best values for optimal variables is still employed, but linear
programming is being used to an increasing extent, although this method is used only
for limited application by most practicing engineers. The presentation here is to sim-
ply acquaint engineers with the fact that other simpler optimization methods also exist.
Further study is required for the reader to attain proficiency in advanced optimization
theory and practice. Finally, the age of computers has, to a certain extent, allowed many
professionals to become comfortable working in an area that was once considered to be
beyond their abilities.
Seven introductory optimization related chapters complement the presentation for
Part I. The titles are as follows.

Chapter 1: Optimization Overview

Chapter 2: Mathematical Operations
Chapter 3: Perturbation Techniques
Chapter 4: Search Methods
Chapter 5: Graphical Approaches
Chapter 6: Analytical Methods
Chapter 7: Linear Programming
1
Optimization Overview

As one might suspect, the term optimization has come to mean different things to differ-
ent people. It has also come to mean different things for different applications; for example,
it could involve a simple two-step calculation or one that requires the use of a detailed
numerical method. To take this a step further, the authors were initially undecided on how
to include optimization subject matter in this text. After much deliberation and medita-
tion, it was decided to present introductory material in the first Part of the book. This deci-
sion was primarily influenced by the desire for this book to be a book that both introduces
optimization mathematical methods and addresses applications that appear in Parts II
and III for environmental and chemical engineers, respectively, as well as in Part IV.
The optimization problem has been described succinctly by Aris1 as “getting the best
you can out of a given situation.” Problems amenable to solution by mathematical optimi-
zation techniques generally have one or more independent variables whose values must be
chosen to yield a viable solution to distinguish between the many viable solutions gener-
ated by the different choices of these variables. Mathematical optimization techniques are
generally used for guiding the problem solver to the choice of variables that maximizes the
“goodness” measure (e.g., profit) or that minimizes some “badness” measure (e.g., cost).
In addition to the optimization definition by Aris presented previously, one might offer the
following generic definition for many engineers: “Optimization is concerned with determin-
ing the ‘best’ solution to a given problem.”2 Alternatively, a dictionary would offer something
to this effect: “to make the most of … develop or realize to the utmost extent … often the most
efficient or optimum use of.” The process of optimization in practice is required in the solu-
tion of many problems and can involve the maximization or minimization of a mathematical
function. As one might suppose, many of the applications involve economic considerations.
One of the most important areas for the application of mathematical optimization tech-
niques is in chemical engineering design. Topics can include the following:

1. Generation of best functional representations (e.g., curve fitting)

2. Design of optimal control systems
3. Determining the optimal height (or length) of pollutant mass transfer control units
4. Determining the optimal diameter of units
5. Finding the best equipment materials of construction
6. Generating operating schedules
7. Selecting operating conditions

This first chapter in Part I serves to provide a broad overview of the general subject of
optimization. Section headings and titles are listed here:

1.1 History of Optimization

1.2 The Computer Age

3
4 Introduction to Optimization for Chemical and Environmental Engineers

1.3 The Scope of Optimization

1.4 Conventional/Established Optimization Procedures
1.5 Contemporary Optimization: Linear Programming

The first three sections need no introduction. The fourth section addresses standard
(some might say traditional) optimization procedures employed by engineers today and in
the past. These include such topics as brute force methods, perturbation schemes, elemen-
tary search methods, graphical approaches, and analytical methods. A separate chapter is
devoted to each of these topics later in this Part. The last section is called Contemporary
Optimization (some might replace contemporary with complex) and primarily keys on
linear programming. These approaches will be emphasized in the solution to the environ-
mental and chemical engineering illustrative examples in Parts II and III, respectively, as
well as in Part IV.

Illustrative Example 1.1

Qualitatively discuss optimization

SOLUTION:
Optimization is viewed by many as a tool in decision-making. It often aids in the selec-
tion of values that allow the practicing environmental and chemical engineer to better
solve a problem. This brief answer provides a qualitative look at optimization.
As noted previously, in its most elementary and basic form, one may say that optimi-
zation is concerned with the determination of the “best” solution to a given problem.
This process is required in the solution of many general problems in engineering and
applied science – in the maximization (or minimization) of a given function(s), in the
selection of a control variable to facilitate the realization of a desired condition, in the
scheduling of a series of operations or events to control completion dates of a given proj-
ect, in the development of optimal layouts of organizational units within a given design
space, and so on. In engineering design, once a particular subject or process scheme
has been selected for study, it is common practice to optimize the process from both a
capital cost and a O&M (operation and maintenance) standpoint.
There are many optimization procedures available, most of them too detailed for
meaningful application in a text of this nature. These sophisticated optimization tech-
niques, some of which are routinely used in the design of conventional chemical and
petrochemical plants, invariably involve computer calculations. Although the use of
these techniques in the majority of environmental and chemical engineering applica-
tions was not warranted in the past, more and more real-world problems are requiring
the use of optimization techniques.

1.1 History of Optimization
The subject of mathematics encompasses the study of relationships among quantities,
magnitudes, and properties and of logical operations by which unknown quantities,
magnitudes, and properties may be deduced. In the past, mathematics was essentially
regarded as a science of magnitudes and numbers. Toward the middle of the 19th century,
however, mathematics came to be regarded increasingly as the science of relations, or as
Optimization Overview 5

the science that draws necessary conclusions. This latter view encompasses mathemati-
cal or symbolic logic, the science of using symbols to provide an exact theory of logical
deduction and references based on definitions, axioms, postulates, and rules for com-
bining and transforming primitive elements into more complex relations and theorems.
Enter optimization.
This brief survey of the history of optimization traces the evolution of mathematical
ideas and concepts, beginning in prehistory. Indeed, mathematics is nearly as old as
humanity itself; evidence of a sense of geometry and interest in geometrical patterns
has been found in the designs of prehistoric pottery and textiles, and in cave paintings.
Primitive counting systems were almost certainly based on using the fingers of both
hands, as evidenced by the predominance of the number 10 as the base for many number
systems employed today.
Interestingly, mathematics of the late 19th century and the 20th century is character-
ized by an interest in unifying elements across numerous fields of mathematical endeavor,
especially in logic. For example, group theory has proved a particularly effective unifier.
The amount of new math and the particular topics arising at that time were numerous and
varied: unified set theory, intuitive geometry, the development of the number systems –
including methods of numeration, binary and other bases of notation, and modular arith-
metic and measurement, with attention to accuracy, precision, and error. It also included
studies of algebraic systems, linear algebra, modern algebra, vectors, matrices, logic, truth
tables, the nature of proofs, Venn and Euler diagrams, relations, functions, probability and
statistics, linear programming, and (of course) computer programming.
As to the origin of optimization, it depends on who provides the response, because
there are so many aspects of optimization of interest to the practitioner. For example, some
claim it was Thomas Edison when he developed a long-lasting, high-quality light bulb in
the 1870s. His success was primarily the result of an extensive trial-and-error search for
the optimum filament material. A few now refer to it as the Edisonian approach. One of the
authors refers to it as the perturbation approach; on occasion, he has modestly termed it
the Theodore approach.
Advances in computing have not only reduced the cost and time required to perform any
iteration calculation but also provide a better understanding of how this method works.
However, the success of this approach still relies heavily on the limitations of the user’s
intuition and experience since one often cannot predict or even comprehend the effects of
changing numerous variables in a complex system. Despite this barrier, the computer has
expanded and improved the employment of the classical methods of optimization.
With the promise of reducing design time and cost while improving product quality,
automated design optimization held tremendous potential. Starting with a sub-optimal
design, a numerical optimization algorithm could be used to iteratively adjust a set of pre-
selected design parameters in an attempt to achieve a set of design targets. This new class
of optimization technology enables broader, more comprehensive, and faster searches for
innovative designs than was possible using previous generations of tools. Moreover, it
requires no expertise in optimization theory, so it is easier to use for non-experts and
experts alike. By leveraging an engineer’s potential to discover new design concepts,
this new class of optimization technology overcomes the limits of human intuition and
extends the designer’s professional capability to achieve breakthrough designs and accel-
erated innovation.
As noted earlier, today’s computers now allow an optimization procedure (or math-
ematics, computer science, operations research, mathematical optimization, or math-
ematical programming) to select the best element (with regard to some criteria) from
6 Introduction to Optimization for Chemical and Environmental Engineers

some set of available alternatives. In the simplest case, an optimization problem con-
sists of optimizing (maximizing or minimizing) a real function by systematically
choosing input values from within an allowed set of variables in order to compute
the value of the function. Mathematically speaking, optimization allows one to find
the best available value(s) of some (often referred to as the objective) function given a
defined domain (or a set of constraints), including a variety of different types of objec-
tive functions and different types of domains. These various options are introduced
in the last section of this chapter and treated in more detail in Chapter 7. A feasible
solution to the objective function that minimizes (or maximizes), if that is the goal, is
called an optimal solution.

1.2 The Computer Age

Although digital computers are often viewed as fast calculators or special slide rul-
ers, they are, in fact, rather general devices for manipulating symbolic information. As
noted, the symbols being manipulated are numbers or digits (hence the name digital
computer), and the operations being performed on the symbols are the standard arith-
metical operations, such as addition and subtraction. Famularo offered the following
thoughts in the late 1960s: “The digital computer can be viewed as a high-speed calcu-
lator that, with the availability of subroutines and a compiling language, is able to per-
form many mathematical operations such as add, multiply, generate analytic functions
and logic decisions, etc.”3 Although the details of coding are not in the scope of this
text, there is an intermediate step between the equations and the coding program, and
that is to arrange the computing procedure in block diagram or information flow form.
The operation of each block is described inside the block, and the computer performs
the instructions around the loop until the condition in the decision block is satisfied, at
which time the most recently computed values were once punched out on tape or card.
For a large computer, this entire sequence, including a half-dozen cycles around the
loop, would be accomplished within a millisecond or less. Of course, as the number of
equations to be solved increases, the time required to obtain a solution increases. For
cases where the solution time becomes excessive, mathematics or other more sophisti-
cated programming is used to accelerate the convergence and decrease the total com-
putation time required.
High-speed computing equipment (today’s computers) has had a tremendous impact
on engineering design, scientific computation, and data processing. The ability of
computers to handle large quantities of data and to perform the mathematical opera-
tions described previously at tremendous speeds permits the analysis of many more
applications and more engineering variables than could possibly be handled on the
aforementioned slide ruler – the trademark of engineers of yesteryear. Scientific cal-
culations previously estimated in lifetimes of computation time are currently gener-
ated in seconds and, on many occasions, microseconds, and in some rare instances,
nanoseconds.4
Increased growth is expected in the microcomputer field as inexpensive information-
processing devices continue to be developed. The cost will continue to decrease, while
the cost of the associated software will probably tend to increase; computerized processes
will, therefore, be quite inexpensive unless customized programs must be written.
Optimization Overview 7

1.3 The Scope of Optimization

One can conclude from the aforementioned that the theory and application of optimiza-
tion is mathematical in nature, and it typically involves the maximization or minimization
of a function (usually known), which represents the “performance” of some “system.” This
is carried out by the finding of values for those variables, which cause the function to yield
an optimal value.
There are a large number of traditional mathematical optimization methods available in
practice. For example, problems might simply involve selecting the “best” choice between
two alternatives – in effect a yes/no option. Or, it could involve three or more alternatives.
A number of others are available in the mathematics literature. Some of these methods are
reviewed in the chapters to follow in this Part I.
There are a large number of “contemporary” optimization approaches available to the
practicing environmental and chemical engineer. Most of the these methods involve find-
ing the optimizer of a function (or functions) by a trial-and-error method until a satisfac-
tory optimizer is achieved. The various methods involve a different procedure for making
educated guesses. The overall process institutes the first guess, and changes this guess
until a satisfactory solution is obtained.

1.4 Conventional/Established Optimization Procedures

On to traditional optimization. One of the authors4 has often employed the phrase “…has
come to mean different things to different people.” No technical phrase better fits the
description of optimization. For many, including the authors, standard (or conventional,
traditional, established, typical, etc.) optimization keys on four topics:

1. Perturbation methods
2. Search methods
3. Graphical approaches
4. Analytical methods

Introductory details on these topics follow, noting that linear programming briefly receives
an introduction in Section 1.5.
Topic (1) basically involves a guessing game when attempting to solve an optimiza-
tion problem. High-powered mathematics is usually not involved. The same can be said
for any trial-and-error method whether it be systematic or not. Numerous elementary
search methods (2) are available but only two will be discussed in Chapter 4. Graphical
approaches (3), as one might suppose, involve graphing available data and calculations in
an attempt to arrive at an “optimization” solution. Analytical methods (4) often employ
ordinary and partial derivatives in obtaining a solution.
Topics (1)–(4) have been defined by most mathematics as direct approaches, while linear
programming is referred to as an indirect approach. As one might assume, indirect meth-
ods are generally preferred. There are, however, numerous other methods – in addition
to those mentioned previously – available for solving optimization problems. These can
8 Introduction to Optimization for Chemical and Environmental Engineers

involve the aforementioned procedures that simply require a yes or no answer, select-
ing the best option of the two alternatives, selecting the best of more than two options,
and so on.
The selection of the independent variable(s) is often set in academic problems involv-
ing standard optimization calculations. The choice of these variables is usually based
on past experience or sound engineering judgment. For example, the performance of an
absorber – for control of a pollutant in an environmental engineering application or for
recovery in a chemical engineering application – is usually assumed to be a function of
temperature and pressure. Other effects, such as the time of day, or lunch details, or … are
usually and understandably neglected. This often reduces the problem to one that does not
involve the need to resort to any sophisticated optimization mathematical methods.
One should also note that these optimization problems, like many engineering prob-
lems, usually involve a simple two-step procedure:

1. Developing a describing equation (or equations) for the system to be optimized

2. Solving the describing equation or equations

Step (1) falls in the domain of the engineer. Step (2) traditionally fell in the domain of
both the engineer and mathematician. However, the second step is essentially no longer a
concern to many engineers with the advent of computers, particularly as it applies to con-
temporary optimization – a topic that is addressed in the next section.

Illustrative Example 1.2

Which of the four traditional optimization methods discussed previously, that is,

1. Perturbation methods
2. search methods
3. graphical approaches
4. analytical methods

are most often employed by the environmental and chemical engineer?

SOLUTION:
This is a difficult question to answer since it is most probably a function of the experi-
ence, interest, and capabilities of the individual being questioned. The answer for one of
the authors (2) earlier in his career would probably be (1) and (3). In more recent times,
both authors would definitely select (4).

1.5 Contemporary Optimization: Linear Programming

Perhaps the most important tool employed in detailed optimization studies by prac-
ticing engineers is linear programming. Linear programming consists of a linear, mul-
tivariable function, which is to be optimized (maximized or minimized), subject to a
particular number of constraints. The constraints are normally expressed in linear form.
Integer linear programming refers to optimization problems in which at least some of the
Optimization Overview 9

variables must assume integer values. If the assumption of linearity cannot be made, as
in the case of linear programming, there exist some general procedures for nonlinear
problems. The collection of techniques developed for these problems is called nonlinear
programming.
The application of linear programming usually involves a constrained optimization
problem. The objective is to locate the values of the decision variable(s) that either maxi-
mize or minimize a linear objective function, where the decision variables are subject to
linear constraints. These mathematical exercises involve locating a value (the feasibility
value) that minimizes the objective function while satisfying the constraints. Thus, linear
programming provides a method of obtaining the optimum feasible value from a near
infinite number of values.
Some optimization problems can be divided into parts, for which each part is then opti-
mized. In some instances, it is possible to attain the optimum for the original problem by
simply realizing how to optimize these constituent parts. This process is very powerful, as
it allows one to solve a series of smaller, easier problems rather than one large one. One of
the best-known techniques to attack such problems is dynamic programming. This approach
is characterized by a process that is performed in stages, such as manufacturing processes.
The possible states of any stage of the process must include enough information regarding
conclusions and/or decisions made at the preceding stage (or stages) to allow an informed
decision at the present stage. Rather than solving the problem as a whole, dynamic pro-
gramming thus optimizes one stage at a time to produce an optimal set of decisions for
the whole process–in effect reformatting the optimization problem as a multistage one.
Although dynamic programming has applicability in some environmental and chemical
engineering systems/processes, it is not reviewed in this book.
One of the major responsibilities in optimization is to construct the correct objective
function to be maximized (or minimized). For example, in a simple environmental prob-
lem, the objective function might be the profit associated with the operation of a landfill
that accepts two categories of waste: “inert” and “non-inert.” One optimization problem
might be to maximize the profit (from handling and treating the waste) by employing an
operating treatment schedule subject to the landfill’s capacity and environmental regula-
tory requirements. This problem might also be rephrased in the following manner: What
combination of the quantities of the two wastes will produce the maximum profit subject
to the constraints connected with the landfill’s processing capabilities and the different
regulatory requirements imposed on each waste?
The objective function may be quite simple and easy to calculate in some environmental
and chemical engineering applications, or it may be complicated and difficult to not only
calculate but also specify and/or describe. The objective function may also be very elusive
due to the presence of conflicting or dimensionally incompatible objectives; for example,
one might be asked to optimize the profit for the aforementioned landfill that not only
minimizes air and water contamination and/or emissions but also is aesthetically appeal-
ing. Thus, it may not be always possible to quantify an objective function and hence be able
to use any of the mathematical optimization procedures available.

Illustrative Example 1.3

The dependent variable (function) y is equal to four times x1 minus three times x2, where
x1 and x2 are positive independent variables. The sum of x1 and x2 are both constrained
by zero and 10, and their cross product is limited to the range of zero and 36. Describe
this constrained mathematical system.
10 Introduction to Optimization for Chemical and Environmental Engineers

SOLUTION:
The (objective) function is

y = 4 x1 − 3 x2

The constraints are

0 ≤ x1 + x2 ≤ 10

and

0 ≤ x1x2 ≤ 36

plus

0 ≤ x1 and 0 ≤ x2

The remainder of Part I addresses six additional topics related to optimization. Chapter
numbers and subject matter are again presented here:

Chapter 2: Mathematical Operations

Chapter 3: Perturbation Techniques
Chapter 4: Search Methods
Chapter 5: Graphical Approaches
Chapter 6: Analytical Methods
Chapter 7: Linear Programming

This material provides optimization fundamentals and principles ranging from the
optimization (or minimization) of a single variable function to more detailed nonlinear
constrained problems. Although the optimization methods are provided with comput-
ers in mind, the reader is again reminded that programming and coding details are not
addressed. Also note that the performance criteria employed can vary from classical eco-
nomic choices to technological factors, such as energy conservation, the degree of purity,
and so on.
Finally, most engineering optimization applications involve economics: maximizing
profit, minimizing cost, or both. In addition to the traditional engineering applications
discussed earlier, environmental applications can involve minimizing toxic emissions,
maximizing energy usage and conservation, minimizing health and safety risks, standard
economic concerns, and so on. The reader should not lose sight of the fact that most real-
world industry applications involving optimization usually require simple solutions. Here
is a comment from a retired engineer:5 “Generally, the consumer goods industry where
I worked does not rely on sophisticated models. Basically, optimization was conducted
to determine overall consumer preference or liking, to determine how to use minimum
resources, and lastly, to determine how to minimize costs subject to constraints. In the
determination of consumer preference, we typically used design of experiments6 to exam-
ine the broad range of attributes that led to maximum product liking or decision to pur-
chase. Occasionally, we modeled processes and used regression models.6 In that case, we
could search or compute the desired maximum. The last set of techniques used is what the
Optimization Overview 11

business world calls ‘Operations Research.’ Typically, linear or dynamic programming was
used to evaluate an ‘objective function subject to constraints.’ For example, we used linear
programming in hot dog manufacturing. The goal was to come up with a mixture of meat
cuts that minimized costs while meeting governmental requirements for protein, fat, and
water content.”

References
1. R. Aris, Discrete Dynamic Programming, Blaisdell, Boston, MA, 1964.
2. L. Theodore, personal notes, East Williston, NY, 2011.
3. J. Famularo, personal communication to L. Theodore (with permission), Englewood Cliffs, NJ,
1973.
4. L. Theodore, personal notes, East Williston, NY, 2005.
5. R. Altomare, communication to L. Theodore, Bronx, NY, 2016.
6. S. Shaefer and L. Theodore, Probability and Statistics Applications for Environmental Science, CRC
Press/Taylor & Francis Group, Boca Raton, FL, 2007.
2
Mathematical Operations

The natural numbers, or so-called counting numbers, are the positive integers: 1, 2, 3, …
and the negative integers: −1, −2, −3, … The following applies to real numbers:

a > b means that a – b is a positive real number (2.1)

If a < b and b < c, then a < c (2.2)

If a < b and c > 0, then ac < bc (2.3)

If a < b and c < d then a + c < b + d (2.4)

1 1
If a < b and ab > 0, then > (2.5)
a b
a + b = b + a; a ⋅ b = b ⋅ a = ( a )(b ) (2.6)

a + (b + c ) = ( a + b ) + c; a ⋅ (bc ) = ( ab ) ⋅ c (2.7)

a (b + c ) = a ⋅ b + a ⋅ c; with a + 0 = a; a ⋅ 1 = a (2.8)

If x = a, where a > 0, then x = a, or x = − a (2.9)

If x < c, then − c < x < c, where c > 0 (2.10)

1
a− n = , a ≠ 0 (2.11)
an

( ab )n = anb n (2.12)

(a )
n m
= a nm , a n a m = a n+ m (2.13)

a0 = 1, a ≠ 0 (2.14)

log ab = log a + log b , a > 0, b > 0 (2.15)

log a n = n log a (2.16)

 a
log   = log a − log b (2.17)
 b

13
14 Introduction to Optimization for Chemical and Environmental Engineers

 1
log n a =   log a (2.18)
 n
ln a = (log e 10 ) common log a = ( 2.3026 ) common log a = 2.3026 log a (2.19)

Based on the previous, one may write

( 4)(9) = (3.60)(10)1
(6)3 = (2.16)(10)2
3
3375 = (1.50 )(10 ) =
1
( 3
3.375 )( 3
1000 )
log ( 245) = 2.3892

log ( 24.5) = 1.3892

log ( 2.45) = 0.3892

log (0.245) = 9.3892 − 10 = −0.6108

ln ( 245) = ( 2.3892)( 2.3026 ) = 5.501

ln ( 24.5) = 3.199

ln ( 2.45) = (0.3892)( 2.3026 ) = 0.896

ln (0.245) = −11.4065
The authors have assumed that the reader is not only familiar with that described previ-
ously but also other more advanced topics that can include simultaneous linear algebraic
equations, nonlinear equations, differentiation, integration, the solution of both ordinary
differential equations and partial differential equations, and the Monte Carlo method.
Only the numerical solution approaches for the previously mentioned are reviewed in
this chapter. Elementary statistics and least squares approximation are not reviewed, but
details are available in the literature.1
Ten sections complement this mathematical operations chapter:

2.1 The Quadratic Equation

2.2 Interpolation and Extrapolation
2.3 Significant Figures and Approximate Numbers
2.4 Errors
2.5 Differentiation
2.6 Numerical Integration
2.7 Simultaneous Linear Algebraic Equations
2.8 Nonlinear Algebraic Equations
2.9 Ordinary Differential Equations
2.10 Partial Differential Equations
Mathematical Operations 15

The first four sections may be viewed as elementary material. The last six sections are
more quantitative and can be viewed as fitting in the domain of numerical methods. Since
a detailed treatment of each of these topics is beyond the scope of this text, the reader is
referred to the literature2–4 for more extensive analysis and additional information. The
remainder of this chapter briefly examines the topics listed previously.

2.1 The Quadratic Equation

Given any quadratic equation of the general form

ax 2 + bx + c = 0 (2.20)

a number of methods of solution are possible depending on the specific nature of the equa-
tion in question. If the equation can be factored, then the solution is straightforward. For
instance, consider

x 2 − 3 x = 10 (2.21)

First, the equation is put into the standard form,

x 2 − 3 x − 10 = 0 (2.22)

which can be factored as follows:

( x − 5) ( x + 2) = 0 (2.23)
This condition can be met, however, only when the individual factors in parentheses are
zero, i.e., when x = 5 and/or x = −2. That these are indeed the solutions to the equation may
be verified by substitution.
If, upon inspection, no obvious means of factoring the above equation can be found, an
alternative approach may exist. For example, in the equation

4 x 2 + 12x = 7 (2.24)

the expression

4 x 2 + 12x (2.25)

could be factored as a perfect square if it were

4 x 2 + 12x + 9 (2.26)

which equals

(2x + 3)2 (2.27)

16 Introduction to Optimization for Chemical and Environmental Engineers

This can easily be achieved by adding 9 to the left side of the equation. The same amount
must then, of course, be added to the right side as well, resulting in:

4 x 2 + 12x + 9 = 7 + 9 (2.28)

(2x + 3)2 = 16 (2.29)

This can be reduced to

(2x + 3 ) = 16 (2.30)

or

2x + 3 = +4 (2.31)

and

2x + 3 = −4 (2.32)

Since 16 has two solutions, the first equation leads to the solution x = 1/2 or x = 0.5, while
the second equation leads to the solution x = −7/2, or x = −3.5.
If the methods of factoring or completing the square are not possible, any quadratic
equation can always be solved by the quadratic formula. This provides a method for deter-
mining the solution of the equation if it is once again in the form

ax 2 + bx + c = 0 (2.33)

In all cases, the two solutions of x are given by the formula

−b ± b 2 − 4 ac
x= (2.34)
2a
For example, to find the roots of

x 2 − 4 x = −3 (2.35)

the equation is first put into the standard form

x 2 − 4 x + 3 = 0 (2.36)

As a result, a = 1, b = −4 and c = 3. These terms are then substituted into the quadratic for-
mula presented in Equation 2.34.

− ( −4 ) ± ( −4)2 − 4 (1)(3) 4 ± 16 − 12
x= = (2.37)
2 (1) 2

4± 4 4±2
= = = 3, and 1 (2.38)
2 2
Mathematical Operations 17

The practicing environmental and chemical engineer occasionally has to solve not just a
single equation but several at the same time. The problem is to find the set of all solutions
satisfying all equations. These are called simultaneous equations, and specific algebraic
techniques may be used to solve them. For example, a simple solution exists for the follow-
ing two linear equations with two unknowns:

3 x + 4 y = 10 (2.39)

2x + y = 5 (2.40)

The variable y in Equation 2.40 is isolated (y = 5 − 2x), and then this value of y is substituted
into Equation 2.39. The result is Equation 2.41.

3 x + 4 ( 5 − 2x ) = 10 (2.41)

This reduces the problem to one involving the single unknown x, and it follows that

−5x = −10 (2.42)

so that

x = 2 (2.43)

When this value is substituted into either previous equation, it follows that

y = 1 (2.44)

A faster method of solving simultaneous equations, however, is obtained by observing

that if both sides of Equation 2.40 are multiplied by 4, then

8 x + 4 y = 20 (2.45)

3 x + 4 y = 10 (2.39)

If Equation 2.39 is subtracted from Equation 2.45, 5x = 10 results, or x = 2. This procedure
leads to another development in mathematics, matrices, which can help to produce solu-
tions for any set of linear equations with a corresponding number of unknowns (see also
Section 2.7).

2.2 Interpolation and Extrapolation

Experimental data (and data in general) in environmental and chemical engineering may
be presented using a table, a graph, or an equation. Tabular presentation permits retention
of all significant figures of the original numerical data. Therefore, it is the most numeri-
cally accurate way of reporting data. However, it is often difficult to interpolate between
data points within tables.
18 Introduction to Optimization for Chemical and Environmental Engineers

Graphical or tabular presentation of data is usually used if no theoretical or empir-

ical equations can be developed to fit the data. This type of presentation of data is
one method of reporting experimental results. For example, heat capacities of benzene
might be tabulated at various temperatures. Data may also be presented graphically.
One should note that graphs are inherently less accurate than numerical tabulations.
However, they are useful for visualizing variations in data and for interpolation and
extrapolation.
Interpolation is of practical importance to the practitioner because of the occasional
necessity of referring to sources of information expressed in the form of a table. Logarithms,
trigonometric functions, water properties of steam, liquid water and ice, vapor pressures,
and other physical and chemical data are commonly given in the form of tables in the
literature. Although these tables are sometimes given in sufficient detail so that interpola-
tion may not be necessary, it is important to be able to interpolate properly when the need
arises.
Assuming that a series of values of the dependent variable y for corresponding tabu-
lated values of the independent variable x is provided, the problem of interpolation is
to obtain the correct value of yet other values of x. (Extrapolation refers to a value of x
lying outside the range of tabulated values of x.) Clearly, interpolation or extrapolation
may be accomplished by using data for x and y to develop a linear relationship between
the two variables. The general method would be to fit two points (y1, x1) and (y2, x 2) by
means of

y = a + bx (2.46)

and then employ this equation to calculate y for some value of x lying between x1 and
x 2. Most practitioners do this mentally when reading values from the (steam) tables.
If a number of points are used, a polynomial of a correspondingly higher degree may
also be employed.1 Interpolation is necessary to find y when x is some value not given
in the table. Thus, interpolation may be viewed as the process of finding the value of
a function at some arbitrary point where the function is not known but is represented
over a given range as a table of discrete points. (See also Table 2.1 where y represents a
reservoir’s height as a function of time in days during a rainy season.) For instance, one
may be interested in finding y when x = 7. (The process of finding x when y is known
is referred to as inverse interpolation). Given a table with (x, y) data (Table 2.1), one can
draw a picture (Figure 2.1) and write the equation of the straight line through the points
(x1, y1) and (x 2, y2) for y.

TABLE 2.1
Reservoir Height
x, Time y, Height
0 30
3 31
6 33
9 35
12 39
15 46
18 52
Mathematical Operations 19

35

30

25

20
y2
y

15
y1
10

5 x1 x2
0
1 2 3 4 5 6 7 8 9 10
x

FIGURE 2.1
Interpolation procedure.

 y − y1 
y − y1 =  2
 x2 − x1 
( x − x1 ) (2.47)
Equation 2.47 can be solved for y in terms of x

( y 2 − y1 ) ( x − x1 )
y = y1 + (2.48)
x2 − x1
or (for points 0 and 1)

y 0 ( x − x 0 ) + ( y1 − y 0 ) ( x − x 0 )
y= (2.49)
x1 − x0

Illustrative Example 2.1

Refer to Table 2.1. Find y at x = 11.

SOLUTION:
Proceed as follows, keying (obviously) on the values for x = 9 and x = 12.

Data Point i xi yi x i − x

1 9 35 2
2 12 39 1

Apply Equation 2.48 with x = 11. Therefore,

y = 35 +
( 4)( 2) = 35 + 8 = 37.67
3 3
Inverse interpolation involves estimating x which corresponds to a given value of y, and
(as noted previously) extrapolation involves estimating values of y outside the interval in
which the data x0, … , xn fall. It is generally unwise to extrapolate any empirical relation
significantly beyond the last data point or before the first point. If, however, a certain form
of equation is predicted by theory and substantiated by (other) available data, reasonable
extrapolation is ordinarily justified.
20 Introduction to Optimization for Chemical and Environmental Engineers

2.3 Significant Figures and Approximate Numbers1

Significant figures provide an indication of the precision with which a quantity is mea-
sured or known. The last digit represents in a qualitative sense, some degree of doubt.
For example, a measurement of 8.32 nm (nanometers) implies that the actual quantity is
somewhere between 8.315 and 8.325 nm. This applies to calculated and measured quan-
tities; quantities that are known exactly (e.g., pure integers) have an infinite number of
significant figures. Note, however, that there is an upper limit to the accuracy with which
physical measurements can be made.
The method for counting the significant digits of a number follows one of two rules
depending on whether or not a decimal point is present. The significant digits of a number
always start from the first nonzero digit on the left to either

1. the last digit (whether it is nonzero or zero) on the right, if there is a decimal point
present, or
2. the last nonzero digit on the right of the number, if there is no decimal point
present.

For example,

370 Has 2 significant figures

370. Has 3 significant figures
370.0 Has 4 significant figures
28,070 Has 4 significant figures
0.037 Has 2 significant figures
0.0370 Has 3 significant figures
0.02807 Has 4 significant figures

Whenever quantities are combined by multiplication and/or division, the number of

significant figures in the result should equal the lowest number of significant figures of
any of the quantities. In long calculations, the final result should be rounded off to the
correct number of significant figures. When quantities are combined by addition and/or
subtraction, the final result cannot be more precise than any of the quantities added or
subtracted. Therefore, the position (relative to the decimal point) of the last significant digit
in the number that has the lowest degree of precision is the position of the last permissible
significant digit in the result. For example, the sum of 3702, 370, 0.037, 4, and 37 should be
reported as 4113 (without a decimal). The least precise of the five numbers is 370, which has
its last significant digit in the tens position. Therefore, the answer should also have its last
significant digit in its tens position.
Unfortunately, environmental and chemical engineers rarely concern themselves with
significant figures in their calculations. However, it is recommended that the reader
attempt to follow the calculation procedure set forth in this section.
In the process of performing engineering scientific calculations, very large and very
small numbers are often encountered. A convenient way to represent these numbers is to
use scientific notation (see also the introduction to this chapter). Generally, a number repre-
sented in scientific notation is the product of a number and 10 raised to an integer power.
For example,
Mathematical Operations 21

28, 070, 000, 000 = 2.807 × 1010 = ( 2.807 )(10 )

10

0.000002807 = 2.807 × 10 −6 = ( 2.807 )(10 )
−6

A positive feature of using scientific notation is that only the significant figures need to
appear in the number.
When approximate numbers are added, or subtracted, the results are also in terms of the
least precise number. Since this is a relatively simple rule to master, note that the answer
follows the aforementioned rule of precision. For example,

6.04 L + 2.8 L − 4.173 L = 4.7 L (2.50)

(The complete result is 4.667 L.) The terms in Equation 2.50 have two, one, and three deci-
mal places, respectively. The least precise previous number (least decimal places) is 2.8, a
value carried only to the tenths position. Therefore, the answer must be reported to the
tenths position only. Thus, the correct answer is 4.7 L, with the last 6 and the 7 dropped
from the 4.667 L, and the first 6 is rounded up to provide 4.7 L.
In multiplication and division of approximate numbers, finding the number of signifi-
cant digits is used to determine how many digits to keep (i.e., where to truncate). One must
first understand significant digits in order to determine the correct number of digits to keep
or remove in multiplication and division problems. As noted earlier in this section, the dig-
its 1 through 9 are considered to be significant. Thus, the numbers 123, 53, 7492, and 5 contain
three, two, four and one significant digits, respectively. The digit zero must be considered
separately.
Zeroes are significant when they occur between significant digits. In the following examples,
all zeroes are significant: 10001, 402, 1.1001, 500.09 has five, three, five, and four signifi-
cant figures, respectively. Zeroes are not significant when they are used as placeholders. When
used as a placeholder, a zero simply identifies where a decimal is located. For example,
each of the following numbers has only one significant digit: 1000, 500, 60, 0.09, 0.0002.
In the number 1200, 540, and 0.0032 there are two significant digits, and the zeroes are
not significant. When zeroes follow a decimal and are preceded by a significant digit, the zeroes
are significant. In the following examples, all zeroes are significant: 1.00, 15.0, 4.100, 1.90,
10.002, 10.0400. For 10.002, the zeroes are significant because they fall between two sig-
nificant digits. For 10.0400, the first two zeroes are also significant because they fall
between two significant digits; the last two zeroes are significant because they follow a
decimal and are preceded by a significant digit. Thus, when approximate numbers are
multiplied or divided in a problem, the result is expressed as a number having the same
number of significant digits as the expression in the problem having the least number of
significant digits.
When truncating (removing final, unwanted digits), rounding is normally applied to
the last digit to be kept. Thus, if the value of the first digit to be discarded is less than 5,
one should retain the last kept digit with no change. If the value of the first digit to be
discarded is 5 or greater, one should increase the last kept digit’s value by one. Assume,
for example, only the first two decimal places are to be kept for 25.0847 (the 4 and 7 are to
be dropped). The number is then 25.08. Since the first digit to be discarded (4) is less than
5, the 8 is not rounded up. If only the first two decimal places are to be kept for 25.0867
(the 6 and the 7 are to be dropped), it should be rounded to 25.09. Since the first digit to be
discarded (6) is 5 or more, the 8 is rounded up to 9.
22 Introduction to Optimization for Chemical and Environmental Engineers

When adding or subtracting approximate numbers, a rule based upon precision deter-
mines how many digits are kept. In general, precision relates to the decimal significance
of a number. When a measurement is reported as 1.005 cm, one can say that the number is
precise to the thousandth of a centimeter. If the decimal is removed (1005 cm), the number
is precise to thousands of centimeters.
For example, in some water pollution studies, a measurement in gallons or liters may
be required. Although a gallon or liter may represent an exact quantity, the measuring
instruments that are used are only capable of producing approximations. Using a stan-
dard graduated flask in liters as an example, can one determine whether there is exactly
one liter? Not likely. In fact, one would be pressed to verify that there was a liter to within
±1/10 of a liter. Therefore, depending upon the instruments used, the precision of a given
measurement may vary.
If a measurement is given as 16.0 L, the zero after the decimal indicates that the measure-
ment is precise to within 1/10 L, that is, 0.1 L. Given a measurement of 16.00 L, one has
precision to the 1/100 L. As noted, the digits following the decimal indicate how precise
the measurement is. Thus, precision is used to determine where to truncate when approxi-
mate numbers are added or subtracted.

2.4 Errors
This section introduces the various classes of errors that arise in technical calculations
while the next section demonstrates the propagation of some of these errors. As one might
suppose, numerous books have been written on the general subject of “errors.” Different
definitions for errors appear in the literature, but what follows is the authors’ attempt to
clarify the problem.1
Any discussion of errors would be incomplete without providing a clear and concise
definition of two terms: the aforementioned precision and accuracy. The term precision is
used to describe a state or system or measurement for which the word precise implies
little to no variation; some refer to this as reliability. Alternatively, accuracy is used to
describe something free from the matter of errors. The accuracy of a value, which may be
represented in either absolute or relative terms, is the degree of agreement between the
measured value and the true value.
All measurements and calculations are subject to two broad classes of errors: determi-
nate and indeterminate. The error is known as a “determinate error” if an error’s mag-
nitude and sign are discovered and accounted for in the form of a correction. All errors
that either cannot be or are not properly allowed for in magnitude and sign are known as
“indeterminate errors.”
A particularly important class of indeterminate errors is that of accidental errors. To
illustrate the nature of these, consider the very simple and direct measurement of tem-
perature. Suppose that several independent readings are made and that temperatures are
read to 0.1°F. When the results of the different readings are compared, it may be found that
even though they have been performed very carefully, they may differ from each other by
several tenths of a degree. Experience has shown that such deviations are inevitable in all
measurements and that these result from small unavoidable errors of observation due to
the sensitivity of measuring instruments and the keenness of the sense of perception. Such
errors are due to the combined effect of a large number of undetermined causes, and they
can be defined as “accidental errors.”
Mathematical Operations 23

Regarding the words precision and accuracy, it is also important to note that a result
may be extremely precise and at the same time inaccurate. For instance, the temperature
readings just mentioned might all agree within 1°F. From this, it would not be permis-
sible to conclude that the temperature is accurate to 1°F until it can be definitively shown
that the combined effects of uncorrected constant errors and known errors are negligible
compared with 1°F. It is quite conceivable that the calibration of the thermometer might be
grossly incorrect. Errors such as these are almost always present and can never be detected
individually. Such errors can be detected only by obtaining the readings with several dif-
ferent thermometers and, if possible, several independent methods and observers.
It should also be understood once again that most numerical calculations are by their
very nature inexact. The errors are primarily due to one of three sources: inaccuracies in
the original data, lack of precision in carrying out calculations, or inaccuracies introduced
by approximate or incorrect methods of solution. Of particular significance are the afore-
mentioned errors due to “round-off” and the inability to carry more than a certain number
of significant figures.
The errors associated with the method of solution are usually the area of greatest con-
cern.3 These usually arise as a result of approximations and assumptions made in the
development of an equation used to calculate a desired result and should not be neglected
in any error analysis.
Finally, many list the following three errors associated with a computer (calculator):

1.
Truncation error: With the truncation of a series after only a few terms, one is com-
mitting a generally known error. This error is not machine-caused but is due to
the method.
2.
Round-off error: This is machine-made and is caused by the limitations of the par-
ticular computer. If one computes using eight digits in multiplication, one would
expect 16 digits by hand calculation. The computer retains only the most signifi-
cant eight of these. If the number was rounded, this error would be minimized,
but rounding is generally not done.
3.
Propagation or inherited error: This is caused by sequential calculations that include
points previously calculated by the computer that already are erroneous owing
to the two previously mentioned errors. Since the result is already off the solu-
tion curve, one cannot expect any new points computed to be on the correct solu-
tion curve. Adding the round-off errors and truncation errors into the calculation
causes further errors to propagate adding more error at each step.

2.5 Differentiation
Several differentiation methods are available for generating expressions for a derivative.
Consider the problem of determining the benzene concentration – time gradient dC/dt5 at
t = 4.0 s; refer to Table 2.2. Six differentiating methods are presented subsequently.5
Method 1: This method involves the selection of any three data points and calculating
the slope m of the two extreme points. This calculated slope is approximately equal to the
slope at the point lying in the middle. The value obtained is the “equivalent” of the deriva-
tive at that point 4. Using data points from 3.0 to 5.0, one obtains
24 Introduction to Optimization for Chemical and Environmental Engineers

TABLE 2.2
Time-concentration Data
Time (s) Concentration of Benzene (mg/L), C
0.0 7.46
1.0 5.41
2.0 3.80
3.0 2.70
4.0 2.01
5.0 1.63
6.0 1.34
7.0 1.17

C5 − C3
Slope = m =
t5 − t3

1.63 − 2.70
= = −0.535
5.0 − 3.0
Method 2: This method involves determining the average of two slopes. Using the same
points chosen previously, two slopes can be calculated, one for points 3 and 4 and the other
for points 4 and 5. Adding the two results and dividing them by 2 will provide an approxi-
mation of the derivative at point 4. For the points used in this method, the results are:

C 4 − C3
m1 = slope1 =
t 4 − t3

2.01 − 2.70
= = −0.69
4.0 − 3.0
C − C4
m2 = slope2 = 5
t5 − t 4

1.63 − 2.01
= = −0.38
5.0 − 4.0
−0.69 + ( −0.38 )
mavg = slopeavg = = −0.535
2
Method 3: This method consists of selecting any three data points (in this case the same
points chosen before) and fitting a curve to it. The equation for the curve is obtained by
employing a second-order equation and solving it with the three data points. The follow-
ing equation results:

C = 0.155t 2 − 1.775t + 6.63

The analytical derivative of the equation is calculated and may then be used to evaluate
the slope at any point. Here, point 4 is used:

dC
= 0.31t − 1.775
dt
Mathematical Operations 25

Evaluated at t = 4.0 s

dC
= 0.31 ( 4.0 ) − 1.775 = −0.535
dt
Method 4: This method uses the method of least squares.1 In this case, all the data points
are used to generate a second-order polynomial equation. This equation is then differenti-
ated and evaluated at the point where the value of the derivative is required. For example,
Microsoft Excel can be employed to generate the regression equation. Once all the coef-
ficients are known, the equation is once again analytically differentiated.

C = 0.1626t 2 − 1.9905t + 7.3108

dC
= 0.3252t − 1.9905
dt
Evaluated at t = 4.0 s:

dC
= 0.3252 ( 4.0 ) − 1.9905 = −0.6897
dt
Methods 5 and 6: These two methods are somewhat similar. They are based on five data
points used to generate coefficients. For this development, represent C and t by f and x (as
it appeared in the literature6), respectively. Method 5 employs five data points to generate a
five coefficient (fourth-order) model using an equation of the form f = A + Bx+ Cx2+ Dx3+ Ex4.
This method is known as interpolating. A set of equations is used to evaluate numerical
derivatives from the interpolating polynomial. The describing equations are listed here:

−25 f 0 + 48 f1 − 36 f 2 + 16 f 3 − 3 f 4
mo = f ′ ( x0 ) = (2.51)
12h

−3 f 0 − 10 f1 + 18 f 2 − 6 f 3 + 3 f 4
m1 = f ′ ( x1 ) = (2.52)
12h

f i − 2 − 8 f i −1 + 8 f i + 1 − f i − 2
mi = f ′ ( xi ) = (2.53)
12h

− f n− 4 + 6 f n−3 − 18 f n−2 + f n−1 + f 3

mn−1 = f ′ ( xn−1 ) = (2.54)
12h

−3 f n− 4 − 16 f n−3 + 36 f n−2 − 48 f n−1 + 25 f 3

mn = f ′ ( xn ) = (2.55)
12h
where:
h = xi + 1 − xi
fi is the function evaluated at i

For example, the equation obtained for “the five-data set” from 1.0 to 5.0 s, that is, t = 1.0,
2.0, 3.0, 4.0, and 5.0 s, using the equations given previously is

f ( x ) = −0.0012x 4 + 0.002x 3 + 0.2616 x 2 − 2.34 x + 7.467

26 Introduction to Optimization for Chemical and Environmental Engineers

All these equations are evaluated for each value of x and f(x). The value obtained for point
4.0 is −0.5448. Method 6 also employs five data points, but only three coefficients are gen-
erated for a second-order polynomial equation of the form f = A + Bx+ Cx2. Another set of
equations is used to evaluate the derivative at each point using this method. The describ-
ing equations are provided here:6

f ′ ( x0 ) =
( −54 f0 + 13 f1 − 40 f2 + 27 f3 − 26 f 4 ) (2.56)
70 h

f ′ ( x1 ) =
( −34 f0 + 3 f1 + 20 f2 + 17 f3 − 6 f 4 ) (2.57)
70 h

f ′ ( xi ) =
( −2 fi−2 − fi−1 + fi+1 + 2 fi−2 ) (2.58)
70 h

f ′ ( x n −1 ) =
(6 fn−4 − 17 fn−3 − 20 fn−2 − 3 fn−1 + 34 f3 ) (2.59)
70 h

f ′ ( xn ) =
(26 fn−4 − 27 fn−3 − 40 fn−2 − 13 fn−1 + 54 f3 ) (2.60)
70 h

At point 4.0, the solution for the derivative using the method is −0.6897.
Comparing all six values obtained for the derivative at t = 4.0 s, one can conclude that the
answers are in close proximity to each other. It is important to note that these are approxi-
mate values and that they vary depending on the approach and the number of data points
used to generate the equations.
Some useful analytical derivatives in engineering calculations are provided in the
literature.4

2.6 Numerical Integration
Numerous engineering and science problems require the solution of integral equations.
In a general sense, the problem is to evaluate the function on the right-hand side (RHS) of
Equation 2.61.

∫
I = f ( x ) dx (2.61)
a

where I is the value of the integral. There are two key methods employed in their solution:
analytical and numerical. If f(x) is a simple function, it may be integrated analytically. For
example, if f(x) = x2, then

∫
I = x 2dx = (
1 3
3
)
b − a 3 (2.62)
a
Mathematical Operations 27

If, however, f(x) is a function too complex to integrate analytically, for example,

I = log  tanh e x
 ( ) (2.63)
2
−2

one may resort to any of the many numerical methods available. Two simple numerical inte-
gration methods that are commonly employed in environmental and chemical engineer-
ing practice are the trapezoidal rule and Simpson’s rule. These are described subsequently.

2.6.1 Trapezoidal Rule
In order to use the trapezoidal rule to evaluate the integral I given by Equation 2.61 as
b

∫
I = f ( x ) dx (2.61)
a

one may use the equation

h
I=  y0 + 2 y1 + 2 y 2 +  + 2 y n−1 + y n  (2.64)
2
where:
h is the incremental change in x; that is, ∆x
yi are the values of f(x) at xi, that is, f(xi)
In addition,

y0 = f ( x0 ) = f ( x = a ) (2.65)

y n = f ( xn ) = f ( x = b ) (2.66)

b a
h=
(2.67)
n
This method is known as the trapezoidal rule because it approximates the area under
the function f(x) – which is generally curved – with a two-point trapezoidal rule calcula-
tion. The error associated with this rule is illustrated in Figure 2.2.
f(x)

f(x)

f(a) f(b)

x
= Error = Calculated value

FIGURE 2.2
Trapezoidal rule error.
28 Introduction to Optimization for Chemical and Environmental Engineers

There is an alternative for improving the accuracy of this calculation – the interval (a, b)
can be subdivided into smaller intervals. The trapezoidal rule can then be applied repeat-
edly in turn over each subdivision.

2.6.2 Simpson’s Rule
A higher-degree interpolating polynomial scheme can be employed for more accurate
results. One of the more popular integration approaches is Simpson’s rule. For Simpson’s
3-point (or one-third) rule, one may use the equation

h
I=  y a + 4 yb+ a/2 + yb  (2.68)
3

The equation for the general form of Simpson’s rule (where n is an even integer) is

h
I=
3
( y0 + 4 y1 + 4 y2 +  + 4 yn −1 + yn ) (2.69)

This method also generates an error, although it is usually smaller than that associated
with the trapezoidal rule. A diagrammatic representation of the error for a 3-point calcula-
tion is provided in Figure 2.3.
The reader should note that the trapezoid rule is often the quickest but least accurate
way to perform a numerical integration by hand. However, if the step size is decreased,
the answer should converge – subject to round-off error – to the analytical solution. The
results of each numerical integration must be added together to obtain the final answer for
smaller step sizes.
Some useful analytical integrals in engineering calculations are provided in the literature.4

f(x)

f(x)
Error
Simpson

f(x0) f(x1) f(x2)

x
x0 x1 x2

= Error

FIGURE 2.3
Simpson’s rule error.
Mathematical Operations 29

2.7 Simultaneous Linear Algebraic Equations

The environmental and chemical engineer often encounters problems that contain not
only more than two or three simultaneous algebraic equations but also those that can be
nonlinear. There is, therefore, an obvious need for systematic methods of solving simulta-
neous linear and simultaneous nonlinear equations.7 This section addresses the linear sets
of equations; information on nonlinear sets is available in the literature.8
Consider the following set of n equations:

a11x1 + a12 x2 + … a1n xn = y1

a21x1 + a22 x2 + … a2 n xn = y 2
(2.70)


an1x1 + an 2 x2 + … ann xn = y n
where:
a is the coefficient of the variable x
y is a constant

This set is considered to be linear as long as none of the x terms are nonlinear, for exam-
2
ple, x2 or ln x1; thus, a linear system requires that all terms in x be linear. The previously
mentioned system of linear algebraic equations may be set in matrix forms:

 a11 a12 … a1n   x1   y1 

 a a … a  x  y 
 21 22 2n
  2  =  2  (2.71)
… … … …  …  … 
    
 an1 an 2 … ann   xn   y n 
However, it is often more convenient to represent Equation 2.71 in the augmented matrix
form provided in the following equation:

 a11 a12 … a1n y1 

a a …a y 
 21 22 2n 2
 (2.72)
… … … … …
 
 an1 an 2 … ann y n 
Methods of solution available for solving these linear sets of equations include

1. Gauss–Jordan reduction
2. Gauss elimination
3. Gauss–Seidel approach
4. Cramer’s rule
5. Cholesky’s methods

Only methods 1 to 3 are discussed in this section.

30 Introduction to Optimization for Chemical and Environmental Engineers

2.7.1 Gauss–Jordan Reduction
Carnahan and Wilkes2 provide an example that solved the following two simultaneous
equations using the Gauss–Jordan reduction method:

3 x1 + 4 x2 = 29 (2.73)

6 x1 + 10 x2 = 68 (2.74)

The four-step procedure is provided here:

1. Divide Equation 2.73 through by the coefficient of x1:

4 29
x1 + x2 = (2.75)
3 3

2. Subtract a suitable multiple, in this case 6, of Equation 2.74 from Equation 2.73, so
that x1 is eliminated. Equation 2.75 remains untouched, leaving

34 5 4 29
− x2 + x2 = (2.76)
3 3 3 3
with the corresponding solution

2x2 = 10 (2.77)

3. Divide Equation 2.77 by the coefficient of x2, that is, solve Equation 2.77.

x2 = 5 (2.78)

4. Subtract a suitable factor of Equation 2.78 from Equation 2.75 so that x2 is elimi-
nated. When (4/3)x2 = 20/3 is subtracted from Equation 2.75, one obtains

x1 = 3 (2.79)

2.7.2 Gauss Elimination
Gauss elimination is another method used to solve linear sets of equations. This
method utilizes the augmented matrix as presented in Equation 2.72. The goal with
Gauss elimination is to rearrange the augmented matrix into a triangle form, where all
the elements below the diagonal are zero. This is accomplished in much the same way
as in Gauss–Jordan reduction. The procedure employed follows. Start with the first
equation in the set. This is known as the pivot equation and will not change through-
out the procedure. Once the matrix is in triangle form, back substitution can be used
to solve for variables.7
Gauss elimination is useful for systems that contain fewer than 30 equations. Systems
larger than 30 equations become subject to round-off error where numbers are truncated by
computers performing the calculations. This method is also effective to calculate the rank,
determinant, and inverse of an invertible matrix.
Mathematical Operations 31

2.7.3 Gauss–Seidel Approach
Another approach to solving an equation or series/sets of equations is to make an informed
or educated guess. If the first assumed value(s) does not work, the value is updated. By
carefully noting the influence of these guesses on each variable, one can approach these
answers or correct set of values for a system of equations.7 The reader should note that
when this type of iterative procedure is employed, a poor initial guess does not prevent the
correct solution from ultimately being obtained.
Ketter and Prawler3 provide several excellent illustrative examples.

2.8 Nonlinear Algebraic Equations

The subject of the solution to a nonlinear algebraic equation is considered in this section.
Although several algorithms are available in the literature, the presentation will focus
on the Newton–Raphson (NR) method of evaluating the root(s) of a nonlinear algebraic
equation.
The solution to the equation

f ( x ) = 0 (2.80)

is obtained by guessing a value for each x, for example (xold), that will satisfy this equa-
tion. This value is continuously updated (xnew) using the equation (the prime represents a
derivative)

f ( xold )
xnew = xold − (2.81)
f ′ ( xnew )

until either little or no change in (xnew − xold) is obtained. One can also express this opera-
tion graphically (see Figure 2.4). Noting that

f(x)

df (x)
Slope – – f ¢ (xold)
dx x = xold

f(xold)

xnew xold
Exact solution
since f(x) = 0

FIGURE 2.4
NR method.
32 Introduction to Optimization for Chemical and Environmental Engineers

df ( x ) ∆f ( x ) f ( xold ) − 0
f ′ ( xold ) = ≈ = (2.82)
dx ∆x xold − xnew

one may rearrange Equation 2.82 to yield Equation 2.83 below. The xnew then becomes xold
in the next calculation.
This method is also referred to as Newton’s method of tangents (NMT) and is a widely used
method for improving a first approximation to a root to the aforementioned equation of
the form f(x) = 0. This development can be rewritten in subscripted form to (perhaps) better
accommodate a computer calculation. Thus

f ( xn )
f ′ ( xn ) = (2.83)
x n − x n +1
from which

f ( xn )
x n +1 = x n − (2.84)
f ′ ( xn )

The term xn + 1is again the improved estimate of xn – the previous guess for a solution
to the equation f(x) = 0. The value of the function and the value of the derivative of the
function are determined at x = x n, and using the proposed procedure, and the new
approximation to the root xn + 1 is obtained. The same procedure is repeated, with the
new value, to obtain a still better approximation of the root. This continues until suc-
cessive values for the approximate root differ by less than a prescribed small value
ε, which controls the allowable error (or tolerance) in the root. Relative to the previous
estimate, ε is given by

x n +1 − x n
ε= (2.85)
xn
Despite its popularity, the method suffers for two reasons. First, an analytical expres-
sion for the derivative, specifically, f′(xn), is required. In addition to the problem of having
to compute an analytical derivative value at each iteration, one would expect Newton’s
method to converge fairly rapidly to a root in all cases. However, as is common with some
numerical methods, it may fail occasionally in certain instances. A possible initial oscil-
lation followed by a displacement away from a root can occur. Note, however, that the
method would have converged if the initial guess had been somewhat closer to the exact
root. Thus, the first guess may be critical to the success of the calculation.8,9

2.9 Ordinary Differential Equations

The Runge–Kutta (RK) method is one of the most widely used techniques in engineering
practice for solving first-order differential equations. For the equation

dy
= f ( x , y ) (2.86)
dx
Mathematical Operations 33

the solution takes the form

h
y n +1 = y n +
6
(D1 + 2D2 + 2D3 + D4 ) (2.87)
where

D1 = hf ( x , y )

 h D 
D2 = hf  xn + , y n + 1 
 2 2 (2.88)
 h D 
D3 = hf  xn + , y n + 2 
 2 2 
D4 = hf ( xn + h, y n + D3 )

The term h represents the increment in x. The term yn is the solution to the equa-
tion at yn + 1, and yn + 1 is the solution to the equation at xn + 1 where xn + 1 = xn + h. Thus,
the RK method provides a straightforward means for developing expressions for ∆y,
namely, xn + 1 −yn, in terms of the function f(x,y) at various “locations” along the interval
in question.
Consider a simple equation of the form

dC
= a + bC (2.89)
dt
where:
t = 0
C = C0

The RK algorithm given previously becomes (for t = h)

h
C1 = C0 +
6
(D1 + 2D2 + 2D3 + D4 ) (2.90)
where

D1 = hf ( x , y ) = h ( a + bC0 )

 h D    D 
D2 = hf  xn + , y n + 1  = h  a + b  C0 + 1  
 2 2   2 

 h D    C + D2   (2.91)
D3 = hf  xn + , y n + 2  = h  a + b  0 
 2 2    2  

D4 = hf ( xn + h, y n + D3 ) = h  a + b (C0 + D3 )

The same procedure is repeated to obtain values for C2 at t = 2h, C3 at t = 3h, and so on.
34 Introduction to Optimization for Chemical and Environmental Engineers

The RK method can also be used if the function in question also contains the indepen-
dent variable.8,9 Consider the following equation:

dC
= f (C , t ) (2.92)
dt
For this situation, one obtains

h
C1 = C0 +
6
(D1 + 2D2 + 2D3 + D4 ) (2.93)
with

D1 = hf (C , t )

 D h
D2 = hf  C0 + 1 , t0 +  (2.94)
 2 2

 D h
D3 = hf  C0 + 2 , t0 + 
 2 2

D4 = hf (C0 + D3 , t0 + h )

For example, if

dC
= 5C − e − Ct (2.95)
dt
then

   D1   h  
 −  C0 + 2   t0 + 2   
dC  D1 
= h  5  C0 +  − e  
 (2.96)
dt   2 
Situations may arise when there is a need to simultaneously solve more than one ordi-
nary differential (ODE). In a more general case, one could have n dependent variables
y1, y2, … yn with each related to a single-independent variable x by the following system
of n simultaneous first-order ODEs:

dy1
= f 1 ( x , y1 , y 2 , … , y n )
dx
dy 2
= f 2 ( x , y1 , y 2 , … , y n )
dx (2.97)

dy n
= f n ( x , y1 , y 2 , … , y n )
dx
Mathematical Operations 35

Note that the equations in Equation 2.97 are interrelated, that is, they are dependent on
each other. This is illustrated in the following two equations:8

dC
= Ae − E/RT C = f (C , t ) (2.98)
dt

dT ∆H
= − kC = g (C , t ) (2.99)
dt ρCp
or in a more general sense

dy
= f (x, y , z) ; (e.g., xyz ) (2.100)
dx

dz
dt
= g (x, y , z) ; (e.g., x y e ) (2.101)
2 2 −z

The RK algorithm for Equations 2.100 and 2.101 is

1
y1 = y 0 +
6
(RY1 + 2RY2 + 3RY3 + RY4 ) (2.102)

1
z1 = z0 +
6
(RZ1 + 2RZ2 + 3RZ3 + RZ4 ) (2.103)

where y1 − y0 = ∆y , z1 − z0 = ∆z , h = ∆x and

RY1 = h × f ( x0 , y0 , z0 )

RZ1 = h × f ( x0 , y0 , z0 )

 h RY1 RZ1 
RY2 = h × f  x0 + , y0 + , z0 + 
 2 2 2 
 h RY1 RZ1 
RZ2 = h × f  x0 + , y0 + , z0 + 
 2 2 2 
(2.104)
 h RY2 RZ2 
RY3 = h × f  x0 + , y0 + , z0 + 
 2 2 2 
 h RY2 RZ2 
RZ3 = h × f  x0 + , y0 + , z0 + 
 2 2 2 

RY4 = h × f ( x0 + h, y0 + RY3 , z0 + RZ3 )

RZ4 = h × f ( x0 + h, y0 + RY3 , z0 + RZ3 )

36 Introduction to Optimization for Chemical and Environmental Engineers

Although the RK approach (and other similar methods) has traditionally been employed
to solve first-order ODEs, it can also treat higher ODEs. The procedure requires reducing
an nth-order ODE to n first-order ODE. For example, if the equation is of the form8

d2 y
= f ( y , x ) (2.105)
dx 2
set
dy
z= (2.106)
dx
so that
dz d 2 y
= (2.107)
dx dx 2
The second-order equation in Equation 2.107 has now been reduced to the two first-order
ODEs provided in Equation 2.108:

d 2 y dz
= = f ( y , x ) (2.108)
dx 2 dx

dy
= z
dx
The procedure expressed in Equation 2.102 and 2.103 can be applied to generate a solu-
tion to Equation 2.105. Note, however, that the first derivative (i.e., dy/dx or its estimate) is
required at the start of the calculation. Extending the procedure to higher-order equations
is left as an exercise for the reader.
The selection of increment size remains a variable to the practicing engineer. Few numer-
ical analysis methods provided in the literature are concerned with error analysis. In gen-
eral, round-off and numerical errors appear as demonstrated earlier in Figure 2.5. In the limit,
when the increment nears 0, one approaches an analytical solution. However, the number
of calculations correspondingly increases the error Ɛ, which increases exponentially as the
increment nears 0. Note that selecting the increment size that will minimize the error is
rarely a problem in practice; in addition, computing it is also rarely a concern.

Roundoff
ε

Numerical

Increment

FIGURE 2.5
Roundoff and numerical error.
Mathematical Operations 37

2.10 Partial Differential Equations

Many practical problems in engineering practice involve at least two independent vari-
ables; thus, the dependent variable is defined in terms of (or is a function of) more than one
independent variable. The derivatives describing these independent variables are defined
as partial derivatives. Differential equations containing partial derivatives are referred to
as partial differential equations (PDEs).
It has been said that “the solution of a partial differential equation is essentially a guess-
ing game.” In other words, one cannot expect to be provided with a formal method that
will yield exact solutions for all PDEs.9 Fortunately, numerical methods for solving these
equations were developed during the mid-to late 20th century, and they are routinely used
today.
The three main PDEs encountered in practice are briefly introduced here, employing T
(e.g., the temperature is the dependent variable) with t (time) and x, y, z (position) as the
independent variables (note that any dependent variable, such as pressure or concentra-
tion, could have been selected).
The parabolic equation

∂T ∂ 2T
= α 2 (2.109)
∂t ∂z
The elliptical equation:

∂ 2T ∂ 2T
+ = 0 (2.110)
∂x 2 ∂y 2
The hyperbolic equation:

∂ 2T ∂ 2T
= α (2.111)
∂t 2 ∂x 2
The preferred numerical method of solution involves finite differencing. Only the para-
bolic and elliptical equations are considered subsequently.

2.10.1 Parabolic PDE
Examples of parabolic PDEs include

∂T ∂ 2T
= α 2 (2.112)
∂t ∂x
and (the two-dimensional)

∂T  ∂ 2T ∂ 2T 
= α  2 + 2  (2.113)
∂t  ∂x ∂y 

Ketter and Prawler,3 as well as many others, have reviewed the finite difference approach
to solving Equation 2.112. This is detailed subsequently.
38 Introduction to Optimization for Chemical and Environmental Engineers

Consider the (t, x) grid provided in Figure 2.6. The partial derivatives may be replaced by

∂T ∆T −T4 + T2 −T4 + T2
≅ = = ; ∆t = k (2.114)
∂t ∆t 2 ( ∆t ) 2k
and

∂T 2 ∆  ∆T  T3 − 2T0 − T1
2 ≅ = ; ∆x = h (2.115)

∂x ∆x  ∆x  h2

Substituting Equations 2.114 and 2.115 into Equation 2.112 leads to

−T4 + T2 T3 − 2T0 − T1
= (2.116)
2k h2
Solving for T2:

T2 = T4 + 2r (T3 − 2T0 − T1 ) (2.117)

where r = k/h2.
Thus, one may calculate T2 if T0, T1, T3, and T4 are known. Unfortunately, stability and error
problems arise in employing the previously mentioned approach. These can be removed
by replacing the central difference term in Equation 2.114 by a forward difference term:

∂T ∆T −T0 + T2 −T0 + T2
≅ = = (2.118)
∂t ∆t ( ∆t ) k

With this substitution, Equation 2.117 becomes

T2 = T0 + r (T3 − 2T0 + T1 ) (2.119)

It can be shown that the problem associated with the central difference derivative is
removed if r ≤ 0.5.

T2

t
∆t = k

T3 T0 T1 x

T4

∆x = h

FIGURE 2.6
Two variable grid.
Mathematical Operations 39

2.10.2 Elliptical PDE
For this equation, examine the grid in Figure 2.7. Using finite differences to replace the
derivatives in Equation 2.110 ultimately leads to
1

4
(T1 + T2 + T3 + T4 ) ; ∆x = ∆y (2.120)
T0 =

In effect, each T value calculated reduces to the average of its four nearest neighbors in
the square grid. This difference equation may then be written at each interior grid point,
resulting in a linear system of N equations, where N is the number of grid points. The sys-
tem can then be solved by one of several methods provided in the literature.9,10
Another solution method involves applying the Monte Carlo approach, requiring the use
of random numbers.9,10 Consider the squares shown in Figure 2.8. If the describing equa-
tion for the variation of T within the grid structure is
∂ 2T ∂ 2T
+ = 0 (2.110)
∂x 2 ∂y 2

T2

y
∆y

T3 T0 T1 x

T4

∆x

FIGURE 2.7
Elliptical grid.

y= a

7 8 9

6 5 4

1 2 3

y= 0 x
x= 0 x =a

FIGURE 2.8
Monte Carlo approach.
40 Introduction to Optimization for Chemical and Environmental Engineers

with specified boundary conditions (BCs) for T(x, y) of T(0, y), T(a, y), T(x, 0), and T(x, a), one
may employ the following approach:

1. Proceed to calculate T at point 1 (i.e., T1).

2. Generate a random number between 00 and 99.
3. If the number is between 00 and 24, move to the left. For 25 to 49, 50 to 74, 75 to 99,
move upward, to the right, and downward, respectively.
4. If the move in step 3 results in a new position that is at the outer surface (bound-
ary), terminate the first calculation for point 1 and record the T value of the bound-
ary at that new position. However, if the move results in a new position that is not
at a boundary and is still at one of the other eight interval grid points, repeat steps
2 and 3. This process is continued until an outer surface boundary is reached.
5. Repeat steps 2 to 4 numerous times, for example, 1000 times.
6. After completing step 5, sum all the T values obtained and divide this value by the
number of times steps (2 to 4) have been repeated. The resulting value provides a
reasonable estimate of T1.
7. Return to step 1 and repeat the calculation for the remaining eight grid points.

This method of solution is not limited to square systems. In addition, one of the authors11
has applied this method of solution to numerous real-world applications.

References
1. S. Schaefer and L. Theodore, Probability and Statistics Applications for Environmental Science, CRC
press/Taylor & Francis Group, Boca Raton, FL, 2007.
2. B. Carnahan and J. Wilkes, Digital Computing and Numerical Methods, John Wiley & Sons,
Hoboken, NJ, 1973.
3. R. Ketter and S. Prawler, Modern Methods of Engineering Computations, McGraw‑Hill, New York
City, NY, 1973.
4. J. Reynolds, J. Jeris, and L. Theodore, Handbook of Chemical and Environmental Engineering
Calculations, John Wiley & Sons, Hoboken, NJ, 2004.
5. L. Perez, homework assignment submitted to L. Theodore, Manhattan College, Bronx, NY,
2003.
6. F. Lavery, The Perils of Differentiating Engineering Data Numerically, Chem Eng, New York
City, NY, 1979
7. L. Theodore and C. Prochaska, Introduction to Mathematical Methods for Environmental Engineers
and Scientists, Wiley‑Scrivener, Salem, MA, 2018.
8. L. Theodore, class notes, Manhattan College, Bronx, NY, 1971.
9. L. Theodore, personal notes, East Williston, NY, 1965.
10. L. Theodore, Heat Transfer for the Practicing Engineer, John Wiley & Sons, Hoboken, NJ, 2011.
11. L. Theodore, personal notes, East Williston, NY, 1985.
3
Perturbation Techniques

A significant number of optimization problems face the environmental and chemical engi-
neer. The optimal design of industrial processes, as well as process equipment, has long
been of concern to the practicing engineer, and indeed, for some, might be regarded as a
definition of the function and goal of applied engineering. The attainment of an optimum
design is generally a result of factors that include mathematical analysis, empirical infor-
mation, and both the subjective and objective experience of the environmental and/or
chemical engineer.
Regarding design, formal optimization techniques have as their goal the development of
procedures for the attainment of an optimum in a system that can be characterized math-
ematically. The mathematical characterization may be (1) partial or complete, (2) approxi-
mate or exact, and/or (3) empirical or theoretical. The resulting optimum may be a final
implementable design or a guide to practical design and a criterion for judging practical
designs. In either case, the optimization techniques should serve as an important part of
the total effort in the design of the units, structures, and control of not only equipment but
also industrial system processes.
One simple procedure for solving optimization problems that is recommended by the
authors is a perturbation scheme. This involves a systematic change of variables, one by one,
in an attempt to locate an optimum. To be practical, this can mean that one must limit
the number of variables by assuming values to those (process) variables that are known
beforehand to play an insignificant role. Reasonable guesses or simple or shortcut math-
ematical methods can further simplify the procedure. Much information can be gathered
from this type of study since it usually identifies those variables that significantly impact
the solution, such as on the overall performance of equipment or process, and it helps iden-
tify the major contributors affecting the optimization calculations.1
What does the previous paragraph mean? If the relationship between the indepen-
dent variables and the dependent variable cannot be solved graphically or analytically,
some form of trial-and-error calculation may be employed to arrive at the optimum. For
example, a sequence of values over a range of the independent variables may be calcu-
lated to completely map out the region of interest. Another scheme is to vary one factor
at a time, holding the other(s) constant. These two schemes are outlined in this chapter
in the three sections that follow. Finally, some in the technical community might malign
this approach of solving optimization problems. Certainly, linear programming, which
is addressed in the last two chapters in this Part, has its place in mathematical methods,
but so does the perturbation approach. The readers should keep this in mind when
there is a need to solve an engineering problem that involves optimization. In fact,
one of the illustrative example applications in Part III involves the application of the
Monte Carlo method,1 which some might argue requires a trial-and-error perturbation
approach.

41
42 Introduction to Optimization for Chemical and Environmental Engineers

This chapter contains three sections that address an optimization problem using the
perturbation approach when there is

3.1 One independent variable

3.2 Two independent variables
3.3 Three independent variables

3.1 One Independent Variable

This first section introduces the reader to the perturbation method and how it applies to an
equation involving one dependent variable, for example, y, and one independent variable,
for example, x. One could therefore write

y = f ( x ) (3.1)

Three simple examples follow.

Illustrative Example 3.1

Suppose one is interested in obtaining the maximum and/or minimum of the formula

y = 100 (1 − x ) (3.2)
2

over the range (constraint)

0 ≤ x ≤ 1 (3.3)

SOLUTION:
To apply the perturbation method of solution, one could calculate y at 0.1 intervals of x
between 0 and 1. The results are provided in Table 3.1.

TABLE 3.1
Perturbation Method for
y = 100(1 −x)2
y x
100 0.0
81 0.1
64 0.2
49 0.3
36 0.4
25 0.5
16 0.6
9 0.7
4 0.8
1 0.9
0.0 1.0
Perturbation Techniques 43

There is both a maximum and a minimum over the 0.0–0.1 range of x. As expected,
the maximum occurs at

x = 0 with y = 100

and the minimum occurs at

x = 1.0 with y = 0.0

Illustrative Example 3.2

Consider the following function

f ( x ) = x 3 − 2x 2 − 5x + 6

Calculate the minimum of the function over the range 0 ≤ x ≤ 4 using a perturbation
technique.

SOLUTION:
Calculate f(x) in the range (0, 4) in increments of 0.4 for x. Results are presented in Table 3.2.
Table 3.2 suggests the minimum is located approximately at x = 2.0 with a value of 4.0.

TABLE 3.2
Calculations for Illustrative Example 3.2.
f(x) x
6.0 0.0
3.74 0.4
1.23 0.8
−1.15 1.2
−3.02 1.6
−4.0 2.0
−3.7 2.4
−1.73 2.8
2.29 3.2
8.74 3.6
18.0 4.0

Illustrative Example 3.3

Consider the following function

11, 900
y = 162 − 2.33 x −
x

Locate the value of x that corresponds to the minimum value of y. Once again employ
the perturbation technique.
44 Introduction to Optimization for Chemical and Environmental Engineers

TABLE 3.3
Calculations for Illustrative Example 3.3
y x
60 −176.1
62 −174.4
64 −173.1
66 −172.1
68 −171.4
70 −171.1
72 −171.1
74 −171.2
76 −171.7
78 −172.3
80 −173.2

SOLUTION:
The following results if one were to neglect the last term in y:

y = 162 − 2.33 x

Arbitrarily, set y = 0 and solve for x.

162
x= ≈ 70
2.33
At this value of x,

11, 900
y = 162 − 2.33 x −
70

y = 171.1

Generate values for x that straddle the value of 70 over the range 60–80 in increments
of 2. Results are provided in Table 3.3.
It appears the minimum is indeed around 70. Additional calculations could further
refine these results. For example, for x = 71, y = −171.0. This example will be revisited in
Chapter 6.

Illustrative Example 3.4

Obtain a value for x2 that will minimize the function

  k −1 
 
 k −1 
  
  x2   k   x3   k 
y = f ( x2 ) =   +  − 2 ;
 x1   x2  
 

if , x1 = 1, x3 = 10, and k = 1.4.

Perturbation Techniques 45

TABLE 3.4
Results of Illustrative Example 3.4
y x2
2 0.800
3 0.777
4 0.780
5 0.800
6 0.814
7 0.814
8 0.875
9 0.901

SOLUTION:
Set x1 = 1, x3 = 10, and k = 1.4 in the previous equation.

 x 0.285   10 0.285  
y ( x2 ) =  2  +  − 2
 1   x2  

Since 1 ≤ x ≤ 10, select value of x2 in the interval (2, 9) in increments of 1.0 and calculate y.
The results are provided in Table 3.4.
The results of Table 3.4 suggest that the minimum of y is approximately 0.777 and
occurs at x2 ≅ 3.0. The reader may check this for x2 = 3.5 where y = 0.777.

3.2 Two Independent Variables

This section presents four examples that illustrate the method of solution when there
are two independent variables. These examples introduce the reader to the concept of
constraints.

Illustrative Example 3.5

Obtain the maximum and/or minimum for the following function

y = 3 x1 + 5x2

0 ≤ x1 ≤ 3, 0 ≤ x2 ≤ 4

SOLUTION:
Proceed as illustrated below. The function can be calculated at the 20 points shown
in parenthesis in Table 3.5. The value of the Table 3.5 point locations for Illustrative
Example 3.5 are shown immediately to the right of the point in parenthesis in
Table 3.6.
As expected, the optimum y is 29 and occurs at x1 = 3 and x2 = 4, and the minimum y is
located at x1 = 0 and x2 = 0 with a minimum value of 0.
46 Introduction to Optimization for Chemical and Environmental Engineers

TABLE 3.5
Optimization Calculations for Illustrative Example 3.5
(0, 0) (1, 0) (2, 0) (3, 0)
(0, 1) (1, 1) (2, 1) (3, 1)
(0, 2) (1, 2) (2, 2) (3, 2)
(0, 3) (1, 3) (2, 3) (3, 3)
(0, 4) (1, 4) (2, 4) (3, 4)

TABLE 3.6
Function Values for Illustrative Example 3.5
(0, 0) 0 (1, 0) 3 (2, 0) 6 (3, 0) 9
(0, 1) 5 (1, 1) 8 (2, 1) 11 (3, 1) 14
(0, 2) 10 (1, 2) 13 (2, 2) 16 (3, 2) 19
(0, 3) 15 (1, 3) 18 (2, 3) 21 (3, 3) 24
(0, 4) 20 (1, 4) 23 (2, 4) 26 (3, 4) 29

A modified form of the previous equation will be revisited in Chapter 7 (Illustrative

Example 7.2) when a graphical solution is provided that includes a plot of x1 vs.x2 for
various values of y.

Illustrative Example 3.6

Obtain the maximum and minimum for the following objective function with the same
constraints employed in the previous example.

E = 3 x1 + 5x2 + x12 + 2x2 2 − 5x1x2

SOLUTION:
The values for 20 points appear in Table 3.7.
The above function exhibits both a maximum and a minimum over the specified
range for x1 and x2. For this example, the minimum is located at (0, 0) with a value of 0,
and the maximum is located at the point (0, 4) with a value of 52.
This problem could also have been solved analytically; the solution is provided in
Chapter 6.

Illustrative Example 3.72

Employing the perturbation approach, obtain the maximum value of the function

11, 900
y = 162 − 2.33 x1 − − 1.86 x1
x1x2

TABLE 3.7
Calculations for Illustrative Example 3.6
(0, 0) 0 (1, 0) 4 (2, 0) 10 (3, 0) 18
(0, 1) 7 (1, 1) 6 (2, 1) 7 (3, 1) 10
(0, 2) 18 (1, 2) 12 (2, 2) 8 (3, 2) 6
(0, 3) 33 (1, 3) 22 (2, 3) 13 (3, 3) 6
(0, 4) 52 (1, 4) 36 (2, 4) 22 (3, 4) 10
Perturbation Techniques 47

Earlier calculations suggest that

0 ≤ x1 ≤ 30, 0 ≤ x2 ≤ 30

SOLUTION:
The calculations for this example are presented in Table 3.8.
It appears that the maximum value of y is approximately 80.4 and is located at x1 = 10
and x2 = 30. This problem will also be revisited in Chapter 6.

Illustrative Example 3.8

Obtain values for x1 and x2 that minimize the subsequent function below employing a
perturbation approach. Both x1 and x2 are constrained in the range (1, 10) with x2 > x1.

 x  0.285  x2  0.285  10  0.285 

y ( x1 , x2 ) =  1  +  −  − 3
 1   x1   x2  

SOLUTION:
Since x2 > x1, initially assume increments of 1 for x1 and x2 in the range of (2, 5) and (4, 9),
respectively. Calculated values for y for various combinations of x1 and x2 are provided
in Table 3.9.
The results in Table 3.9 suggest the minimum occurs in the neighborhood of x1 = 2.0
and x2 = 4.0. What further calculation would the reader suggest? One of the authors fol-
lowed the previous calculations with two additional calculations, the results of which
are presented here:

x1 = 1.5, x2 = 4.0; y = 0.743

x1 = 2.0, x2 = 3.5; y = 0.739

These two latter values for y further implies that the minimum is indeed in the range of
x1 = 2.0 and x2 = 4.0. The reader is referred to both Chapters 4, 5, 6, and 7 and Part III – Fluid
Flow, Chapters 13. The reader will soon discover that the analytical answer is

x1 = 2.154, x2 = 4.642; y = 0.644

TABLE 3.8
Calculation Chart for Illustrative Example 3.7
x2 x1
0 5 10 15 20 25 30
0 −∞ −∞ −∞ −∞ −∞ −∞ −∞
5 −∞ −335.0 −117.9 −59.5 −40.8 −38.0 −43.0
10 −∞ −97.0 +1.1 +19.8 +18.7 +9.7 −3.4
15 −∞ −17.6 +40.8 +46.3 +38.5 +25.5 +9.9
20 −∞ +22.1 +60.6 +59.5 +48.5 +33.5 +16.5
25 −∞ +45.9 +72.5 +67.4 +54.4 +38.2 +20.4
30 −∞ +61.7 +80.4 +72.7 +58.4 +41.4 +23.1
48 Introduction to Optimization for Chemical and Environmental Engineers

TABLE 3.9
Calculation Chart for y(x1, x2) for Illustrative Example 3.8
x2 x1
4 5 6 7 8 9
2 0.654 0.734 0.743 0.754 0.810 0.784
3 0.671 0.742 0.742 0.748 0.798 0.766
4 − 0.768 0.762 0.765 0.810 0.775
5 − − 0.792 0.790 0.791 0.794

3.3 Three Independent Variables

This section presents one illustrative example involving three independent variables.

Illustrative Example 3.9

Consider the following function:

y = 3 x1 − 5x2 − 8 x3

Assume the same constraints as the previous two examples for x1 and x2; the constraint
for x3 is

1 ≤ x3 ≤ 2

Obtain the maximum and minimum of y.

SOLUTION:
Apply the perturbation method again as shown below. See Table 3.10 for
x3 = 1:

TABLE 3.10
Calculations for x3 = 1 for Illustrative Example 3.9
(0, 0) −8 (1, 0) −5 (2, 0) −2 (3, 0) 1
(0, 1) −13 (1, 1) −10 (2, 1) −7 (3, 1) −4
(0, 2) −18 (1, 2) −15 (2, 2) −12 (3, 2) −9
(0, 3) −23 (1, 3) −20 (2, 3) −17 (3, 3) −14
(0, 4) −28 (1, 4) −25 (2, 4) −22 (3, 4) −19

For x3 = 2, refer to Table 3.11:

TABLE 3.11
Calculations for x3 = 2 for Illustrative Example 3.9
(0, 0) −16 (1, 0) −13 (2, 0) −10 (3, 0) −1
(0, 1) −21 (1, 1) −18 (2, 1) −15 (3, 1) −6
(0, 2) −26 (1, 2) −23 (2, 2) −20 (3, 2) −11
(0, 3) −31 (1, 3) −28 (2, 3) −25 (3, 3) −16
(0, 4) −36 (1, 4) −33 (2, 4) −30 (3, 4) −21
Perturbation Techniques 49

Based on the these two trials for x3, the maximum is located at (3, 0) and the minimum
value is located at (0, 4) with corresponding y values of 1 and −36, respectively. The
previous solution could be further refined by performing calculations at x3 = 1.25, 1.5,
and 1.75.

References
1. L. Theodore, Chemical Engineering: The Essential Reference, McGraw-Hill, New York City, NY,
2014.
2. L. Theodore, personal notes, Theodore Tutorials, East Williston, NY, 2015.
4
Search Methods

The mathematicians’ literature abounds with search methods, that is, methods that allow
one to search for either a solution to an equation or the optimum value for an equation.
There are both indirect and direct approaches. However, direct methods – those that pro-
duce an exact answer – are generally preferred by environmental and chemical engineers.
These methods have also been defined in the literature as seek methods, and usually apply
to single-independent variable unimodal equations; they provide results in intervals (or
limits) of uncertainty for the variable in a (minimum) number of trials during the search
process. The unimodal function referred to for single-independent variable equations con-
tains a single optimization (maximum or minimum) point over a specified interval/range.
Two of the procedures that are reviewed in this chapter are the interval halving (or bisec-
tion) and golden section search methods. These two methods can also be employed to
obtain the solution to some equations. For example, the interval halving method depends
on finding an approximation to a solution of the form f(x) = 0. Initial guesses x0 and x1 are
sequenced provided that f(x) is continuous for x0 ≤ x ≤ x1 and the product [(  f(x0))(  f(x1))] < 0.
This guarantees that the curve y = f(x) crosses the x-axis between x0 and x1. A new approxi-
mation of f(x) is then calculated at

x2 =
( x0 + x1 ) (4.1)
2
that is, f(x2). If [(f(x1))(f(x2))] < 0 then f(x) is calculated at

x3 =
( x1 + x2 ) (4.2)
2
If, however, [(f(x0))(f(x2))] < 0 then f(x) is calculated at

x3 =
( x0 + x2 ) (4.3)
2

Illustrative Example 4.1

Solve

x 3 − 9x + 1 = 0

for the root x* between x = 2 and x = 4 by interval halving.

SOLUTION:
First, calculate

f ( 2 ) = −9

51
52 Introduction to Optimization for Chemical and Environmental Engineers

f ( 4 ) = 29

The results assure that a root exists between x = 2 and x = 4. Now, set

x2 =
( x1 + x0 ) = 2 + 4 = 3
2 2
Since f(x = 3) = 1, the next approximation is determined by noting that [f(2) · f(3)] = (−9)
(1) = −9 < 0. Therefore, set

2+3
x3 = = 2.5
2

The reader is left the exercise of showing that x = x* ≈ 2.95.

Perhaps the simplest of optimization methods that have been employed in the past by
practitioners are the interval halving and bisection methods. Unfortunately, both have
come to mean different things to different people, and to add to the confusion, both terms
have been used interchangeably. The calculations can be performed with or without the
need to evaluate the derivative of the function in question. For this text, the method involv-
ing derivatives will be referred to as the bisection method.
Some complications can arise for search methods involving multi-independent variable
equations. The most popular of these methods is the steepest ascent/descent method. This
topic is also reviewed in this chapter although there are numerous other methods that can
be employed.
Finally, it should be noted that nonlinear algebraic equations are often solved using the
Newton–Rhapson method, as described earlier and in the literature.1 However, the method
requires that the derivative of the function be available.
Several of the search methods do not require the derivative of the function in generating
a solution. It should be noted that an expression for the derivative f′(x) from f(x) can nor-
mally be derived for many problems involving simple functions. When f(x) is such that its
derivative cannot be obtained or is such that it is very complex, an approximation for f(x)
may be used. By definition,

f ( x + ∆x ) − f ( x )
f ( x ) = lim (4.4)
∆x → 0 ∆x
or

f ( x + ∆x ) − f ( x − ∆x )
f (x) ≈ (4.5)
2∆x
Four sections complement the presentation in this chapter. Search procedures are pro-
vided for

4.1 Interval halving method

4.2 Bisection method
4.3 Golden section method
4.4 Method of steepest ascent/descent
Search Methods 53

4.1 Interval Halving Method

A search procedure can be employed that is based on the interval halving method
described below for optimizing (obtaining the maximum or minimum of) a function. This
method is based on the fact that f(x) is unimodal and it is only necessary to compare the
function f(x) at two different points (e.g., f(x1) and f(x2)) in order to predict which of the
subintervals within these two points the optimum does not lie. Thus, one can institute a
search in which the optimum is found by repeatedly eliminating sections of the original
bounded interval in question. The search is terminated when the remaining subinterval is
reduced to an acceptable one.
The main advantage of this search method is that it only requires an algebraic cal-
culation of the function. In particular, the function need not be differentiable or be in
a mathematical or analytical form. All that is required is that the function f(x) can be
determined at a specific x. It should be noted that there is also a search procedure that
requires the calculation of the derivative at the points in question, as described in the
next section.
This optimization search method reduces the interval of the location of the indepen-
dent variable that contains the solution by one-half the previous interval. It has also been
defined as the three-point equal-interval search method since it requires three equally spaced
trial points in the search interval. The search method for finding the minimum of a function
f(x) over the interval a ≤ x ≤ b is now detailed.

1. Compute f(x) at the midpoint x = xm of the interval, L, that is, f(xm).

L = b − a (4.6)

xm =
( a + b ) (4.7)
2
2. Set

L
x1 = a + (4.8)
4

L
x2 = b + (4.9)
4
3. Compute f(x1) and f(x2).
4. Compare f(x1) and f(xm). If (x1) ≤ f(xm), eliminate the interval xm to b from consider-
ation so that the remaining interval is a to xm. The midpoint of the new interval is
previously specified x1. Update xm in this new interval by setting

x1 = xm (4.10)

5. Calculate an updated L,

L = xm − a = x1 − a (4.11)
54 Introduction to Optimization for Chemical and Environmental Engineers

The solution for the minimum is in the new interval for L in Example 4.11. Unless
L has achieved an acceptable small value, return to step (2) and repeat the calcula-
tion for further refinement.
6. If (x2) ≤ f(xm), eliminate the interval for a to xm from consideration. The remaining
interval is xm to b. The midpoint of the new interval is x2. Update xm in this new
interval by setting

x2 = xm (4.12)

7. Calculate an updated L,

L = b − xm = b − x2 (4.13)

The solution for the minimum is in the new interval of L. Unless L has achieved
an acceptable small value, return to step (2) and repeat the calculation for further
refinement.
8. If by chance both f(x1) ≥ f(xm) and f(x2) ≤ f(xm), eliminate the interval (a, x1) and (b, x2).
The remaining (new) interval has therefore changed from (a, b) to (x1, x2) with the
same midpoint xm. The calculation then proceeds as outlined previously.
9. If by chance both f(x1) ≤ f(xm) and f(x2) ≤ f(xm), there is a fault in the mathematical
description of the minimization system since only a maximum is possible.

Illustrative Example 4.2

Obtain the value of x that provides the minimum value of the function

f ( x ) = x 3 − 3x 2 + 2

using the interval halving method provided in steps (1–9) above. Assume that 2 ≤ x ≤ 6.

SOLUTION:
Apply the algorithm provided in steps (1)–(9).
Step 1:

L =6 − 2 = 4

xm =
(2 + 6) = 4
2
Since:

f ( x ) = x 3 − 3x 2 + 2

f ( xm ) = f ( 4 ) = 64 − 48 + 2 = 18

Step 2:

2
x1 = 2 + =3
2
Search Methods 55

2
x2 = 6 − =5
2
Step 3:

f (3) = 2

f ( 5 ) = 52

Step 4: Since f(3) ≤ f(4), the remaining (surviving) interval is (2, 4). The midpoint of the
new interval is 3. The process is continued as per the suggested algorithm. The solution
will ultimately approach a value of x = 2.0 with f(x) = −2.0.

Illustrative Example 4.3

Consider the following function:

f ( x ) = x 3 − 2x 2 − 5x + 6

Obtain the minimum value of this function over the range (1, 3) by employing the inter-
val halving approach.

SOLUTION:
For this example, a = 1, b = 3, and

xm = 2.0

Set

x1 = 1.5

and

x2 = 2.5

Calculate f(1.5), f(2.0), and f(2.5).

f (1.5 ) = −2.63

f ( 2.0 ) = −4.0

f ( 2.5 ) = −3.38

Since f(1.5) ≥ f(2.0) and f(2.5) ≥ f(2.0), eliminate the interval (1, 1.5) and (2.5, 3). The
remaining interval is then (1.5, 2.5) with the same midpoint xm = 2.0. Proceed to cal-
culate f(1.75) and f(2.25). The remainder of the problem is left as an exercise for the
reader. The reduced interval will ultimately center around x = 2.1 with f(x) approxi-
mately 4.0.
56 Introduction to Optimization for Chemical and Environmental Engineers

4.2 The Bisection Method

This method assumes the function is not only unimodal but also differentiable. Since the
function can be differentiated, one begins the calculation for a “minimum” by evaluating
the derivative at the midpoint of the initial “uncertainty” interval. Thus, if the interval
range is (x1, x2) set x3 = (x1 + x2)/2 and calculate f′(x3). If f′(x3) ≥ 0, then the minimum lies to
the left of x3 and the uncertainty interval has been reduced to (x1, x3); that is, the minimum
cannot be to the right of x3. On the other hand, if f′(x3) ≤ 0 the minimum lies to the right
of x3 and the uncertainty interval has reduced to (x3, x2). However, if f′(x3) = 0, x3 is the
minimum and the search can be terminated. The process is repeated iteratively until the
uncertainty interval is acceptable. A similar procedure is employed for the calculation of
a maximum.
For both this and the interval halving method, the length of the uncertainty interval is
reduced by a factor of 1/2.

Illustrative Example 4.4

Resolve the previous example employing the bisection approach.

SOLUTION:
Note once again that this method requires the calculation of the derivative of the func-
tion. For this example,

f ′ ( x ) = 3x 2 − 4x − 5

Set

1+ 3
xm = x3 = = 2.0
2
and

f ′ ( x ) = f ( x = 2 ) = −1

Since f′(x = 2.0) ≤ 0, the minimum lies to the right of x3 and the uncertainty interval is
now (2, 3). Continuing, set

2+3
x3 = = 2.5
2
and

f ′ ( 2.5 ) = 3.75

The reader is left the exercise of continuing the calculation – employing the new interval
(2, 2.5) – and showing that the minimum approaches an interval centered about x = 2.1
with f(2.1) = −4.06.
Search Methods 57

4.3 The Golden Section Method

The reader is referred to the literature for details and derivation of this method.1–3 Only
the algorithm is presented in this section. The principle of the golden section method
can be stated in essay form as follows: In the division of a line of unit length into two
unequal parts, the ratio of the larger of the two segments to the total length of the line should
be the same as the ratio of the smaller to the larger segment. These two segments are pro-
vided in Figure 4.1. Based on the previously mentioned premise, one may write

α α
= (4.14)
1 1− α

The solution to Equation 4.14 is α = 0.618.

The calculation for this method proceeds as follows. If f(x3) > f(x4), the minimum must lie
in the interval (x1, x3). If f(x3) < f(x4), the minimum is located in the interval (x4, x2). For the
former case, the calculation is continued employing the new (and reduced) interval with
x5 located between x1 and x4 and (1 − α)α or 0.236L (if x2 − x1 = L) displaced from x1 (See also
Figure 4.2). Thus, when f(0.618) >  f(0.382) the (0.618, 1.0) interval is discarded, leaving the
interval (0.0, 0.618) or 0.618L. The next calculation is performed at x5 = 0.236 where once
again the interval range ratio (IRR) is

0.236 − 0
IRR = = 0.618
0.618 − 0.236

α 1−α

1 − α = α2 2α − 1 = α3 1−α

x1 = 0 x4 x3 x2 = 1.0

FIGURE 4.1
Golden section diagram – first trial.
58 Introduction to Optimization for Chemical and Environmental Engineers

α 1−α

1 − α = α2 2α − 1 = α3 1−α

α
4=
α
2α − 1 = α3 α2 − α3 2α − 1
= 2 − 3α
1 − α = α2 1 − α = α2
First interval
reduction = 1− ∝

x1 = 0 x5 x4 x3 x2 = 1.0

FIGURE 4.2
Golden section diagram – second trial.

The calculation then continues as outlined previously. If f(0.236) > f(0.362), discard the

range (0.236, 0.362), leaving only the interval (0.0, 0.236). The next value to be calculated is
at x6 = 0.1465 (See Figure 4.3). The interval ratio range remains

0.236 − 0.146
IRR = = 0.618
0.146 − 0

Finally, it should be clear that the golden section search method outperforms the previous
two methods by requiring the least number of calculations for the same reduction of the
interval range.

Illustrative Example 4.5

Resolve the previous introductory example (Illustrative Example 4.2) employing the
golden section method. Here

f ( x ) = x 3 − 3x 2 + 2

Assume the initial range is again 1 ≤ x* ≤ 3.

SOLUTION:
First note for this example the interval is

L = 3 −1= 2
Search Methods 59

First set x4.

( )
x4 = 1 + α 2 L

( )
x 4 = 1 + 0.618 2 L = 1 + ( 0.382 )( 2 )

x 4 = 1.764

Calculate f(x) at x = x4 = 1.764.

f ( x 4 ) = f (1.764 )

f ( x 4 ) = −1.846

Set x3.

x3 = 3 − ( 0.382 )( 2 ) = 2.236

Calculate f(x) at x = x3 = 2.236.

f ( x3 ) = f ( 2.236 )

= −1.819

1
α 1−α
1−α= α2 2α − 1 = α3 1−α

α
α4 =
2α − 1 = α3 2α − 1 (First interval
α2 − α3
reduction)
= 2 − 3α
1 − α = α2 1 − α = α2

1 − α = α2 (Second interval
reduction)

α4 α5 α4 = (α2)2

x1 = 0 x6 x5 x4 x3 x2 = 1.0

FIGURE 4.3
Golden selection; subsequent interval – third trial.
60 Introduction to Optimization for Chemical and Environmental Engineers

Since f(1.764) < f(2.236), the interval of (2.236, 3) is eliminated from consideration, and the
new limits are 1 ≤ x* ≤ 2.236. Set x5 where

x5 = 1 + ( 0.618 )

( ) ( )
x5 = 1 + α 3 L = 1 + 0.236 3 ( 2) = 1.472

and proceed as noted previously. As one would suppose, the minimum will ultimately
converge at x = 2.0.

4.4 Method of Steepest Ascent/Descent

Perhaps the one search method that has received the most attention in the literature is
that described by the title of this last section. All the one-independent variable search
methods developed earlier presume a unimodal function. However, this assumption is
less likely to be the case when the function in question is dependent on more than one
independent variable. For multidimensional problems, regions rather than lines within
the total space, need to be examined. A logical extension of the one-independent variable
case discussed before would be to select a starting point for the search and, while holding
all but one of the independent variables constant, alter that one until an optimum is real-
ized. Then, allow a second variable to seek its relative optimum value. In turn, by altering
only one variable at a time, holding all others constant, all (say n) independent quantities
are allowed to vary. This is defined as the sectional search method and is the essence of
the aforementioned method of steepest ascents/descents. This will be illustrated later for
a two-independent variable example. Happel4 described the “steepest” method in rela-
tively simple terms in his classic book titled Chemical Process Economics. What follows is his
edited write-up, adapted with permission.
“There are many situations, especially where experimentation is involved, in which nei-
ther an analytical nor graphical representation is available for the function to be optimized
for the entire range of the variables involved. The question then arises as to what course
the experimenter should follow, be it in the research laboratory or full-scale process. One
may employ any of the several trial-and-error procedures previously discussed earlier.
However, if the results are known over only a small range of the variable, some efficient
method of extrapolation is necessary to minimize the amount of trial-and-error. Such a
scheme is defined by some as the method of steepest ascents (or descents). The concept
behind the idea of predicting a path of steepest ascent is simply that a plane does a good
job of approximating a curved surface over a limited area. Thus, for a situation involv-
ing the maximizing of a function of two independent variables, a contour map with the
(independent) variables as horizontal coordinates and the objective (dependent) function
as the vertical coordinate will better illustrate the problem in question. The best-fitting
plane is obtained in a small area under investigation. Then, it is established as the direc-
tion of steepest ascent. Experiments (or calculations) are performed along this path until
a decline in response is noted. Additional observations are taken around this point to
confirm whether a maximum has indeed been reached or whether a new path of steepest
ascent should be predicted. The method, of course, does not eliminate the possibility that
more than one optimum may exist in the area of interest.”
Search Methods 61

In practice, the method of steepest ascents is generally applied to studies involving

more than two variables. As one may suppose, it is difficult to illustrate these cases
graphically, but the mechanics are entirely similar to any two-dimensional approach.
More specifically, the partial derivatives of the function5 will indicate its rate of change
with respect to a series of variables x1, x 2, x3…. From these derivatives the gradient
∇[f(x)] is computed:

∂f ( x ) ∂f ( x ) ∂f ( x )
∇  f ( x )  = i1 + i2 + i3 +… (4.15)
∂x1 ∂x2 ∂x3

where i1, i2, i3… are unit vectors in the direction of coordinate axes representing the vari-
ables, each variable being associated with one dimension in a multidimensional space.5
The gradient ∇[f(x)] is a vector that has the property of being directed along the path
on which the function increases most rapidly (i.e., the path of “steepest ascent”). The
direction is along a path proportional to the rate of increase for a unit change of each
variable.
Summarizing, the method of steepest descent is a gradient operation where a step
size is chosen to achieve the maximum amount of decrease of the objective function
for each individual step. The method of steepest descent is an example of an iterative
algorithm in which the operation generates a sequence of points, each calculated on
the basis of the points preceding it. The operation is a decent method because the cor-
responding value of the objective value decreases in value as each new point is gen-
erated by the algorithm. The steepest descent algorithm proceeds as follows: at each
step, starting from the point. x1, a line search until a minimizer, x2, is located. A typi-
cal sequence is provided later in Figure 4.4. Note that the method of steepest descent
moves in orthogonal steps.
The following example, similar to Illustrative Examples 3.4 and 3.8, illustrates the way
the procedure works. By applying it in a case where the result is known, it is possible to
readily note its limitations. Technically, it is correct only at the base chosen for each ascent.
Additional details on this search method are presented later in the text.

Illustrative Example 4.6 4

Obtain values for x1 and, x2 that will minimize the following function.

  k −1 
 
 k −1 
 
 k −1 
  
  x1  k   x2  k   x3  k 
y = f ( x1 , x2 ) =   +  +  − 3  ; x0 = 1, x3 = 10, k = 1.4
x0  x1   x2 
  

Refer to Illustrative Example 3.8 for additional details.

SOLUTION:
Happel’s graphical/analytical solution4 follows where f(x1, x2) has been multiplied by 4.0.
Contours representing other values of y for values of x2 and x3 are shown in Figure 4.4.
The optimum values of x1 and x2 that minimize y will later be shown to be

x1 = 2.154

x2 = 4.642
62 Introduction to Optimization for Chemical and Environmental Engineers

10
x2 = 2.70

8 2.65

2.60

6
x2

2.59

2.58
5

2
0 1 2 3 4 5
x1

FIGURE 4.4
Contours of y; Note: The path of steepest descent is shown by a dotted line.

Assume at the start of the calculation for the steepest ascend method that x1 = 4 and
x2 = 7. The direction of steepest ascent at x1 = 4 and x2 = 7 is obtained in the following
manner:4

 1− k  1  1− k   1− 2 k 
∂y   −     
= x0 k 
x
1
k
− x2 k   k 
x1 = 0..079
∂x1
Similarly,

∂y
= 0.009
∂x2
To obtain the path of steepest descent, increments of x must be taken proportionally to
these rates of change; thus,

∆x1 0.079
= = 8.8
∆x2 0.009
This represents the values that will produce the maximum change in the objec-
tive function. If increments of x2 are taken equal to 0.05, then the increments of
∆x1 = (0.05) ⋅ (8.8) = 0.44. In this way, the first descent is established starting at x1 = 4 and
x2 = 7, where x1 = 4.00 − 0.44 = 3.56 and x2 = 7.00 − 0.05 = 6.95 but also (Table 4.1).
Search Methods 63

TABLE 4.1
Results of Illustrative Example 4.4
x1 x2 x
4.00 7.00 2.681
3.56 6.95 2.653
3.12 6.90 2.629
2.68 6.85 2.615 (min)
2.24 6.80 2.622

The same procedure is repeated until convergence is achieved. This procedure is

demonstrated in more detail in Part III, Chapter 13, Illustrative Example 13.6.

References
1. R. Ketter and S. Prawel, Modern Methods of Engineering Computation, McGraw-Hill, New York
City, NY, 1969.
2. G. Reklaitis, A. Ravindran, and K. Ragsdell, Engineering Optimization, Wiley Interscience,
Hoboken, NJ, 1983.
3. H. Mickley, T. Sherwood, and C. Reed, Applied Mathematics in Chemical Engineering, McGraw-
Hill, New York City, NY, 1957.
4. J. Happel, Chemical Process Economics, John Wiley & Sons, Hoboken, NJ, 1958.
5. L. Theodore, Transport Phenomena for Engineers, Theodore Tutorials, East Williston, NY, origi-
nally published by the International Textbook Company, Scranton, PA, 1971.
5
Graphical Approaches

Engineering data is often best understood when presented in the form of graphs or math-
ematical equations. The earlier preferred method of obtaining the graphical relations
between variables was to plot the data as straight lines and use the slope-intercept method
to obtain coefficients and exponents. Therefore, it behooves the environmental and chemi-
cal engineer to be aware of the methods of obtaining straight lines on the various types of
graph paper, and of determining the equations of such lines.1 One always strives to plot
data as straight lines because of the simplicity of the curve, ease of interpolation, and ease
of extrapolation. Graphical methods proved invaluable in the past in the analysis of the
relatively complex processes of that era. Much of the basic physical and chemical data are
still best represented graphically.
One or more of the many types of graphical representations may be employed for the
following purposes:

1. As an aid in visualizing a process for the representation of quantitative data

2. For the representation of quantitative data
3. For the representation of a theoretical equation
4. For the representation of an empirical equation
5. For the comparison of experimental data with a theoretical expression
6. For the comparison of experimental data with an empirical expression
7. For solving some optimization problems

The relation between two quantities, y – the dependent variable and x – the independent
variable, is commonly obtained as a tabulation of values of y for a number of different val-
ues of x. The relation between y and x may not be easy to visualize by studying tabulated
results and is often best seen by plotting y vs. x. If the conditions are such that y is known
to be a function of x only, the functional relation will be indicated by the fact that the
points may be represented graphically by a smooth curve. Deviations of the points from a
smooth curve can provide a measure of the reliability of the data. If y is a function of two
variables, x1 and x2, a series of results of y in terms of x1 may be obtained for a (definite)
constant x2. When plotted, the data will be represented by a family of curves, each curve
representing the relation between x1 and y for a constant value of x2. If another variable x3
is involved, one may have separate graphs for constant values of x3, each showing another
family of curves of y vs. x1. One may expand this method of representing data to more than
three independent variables, although this rarely occurs in practice. In addition, graphs
may be employed to perform common mathematical manipulations, such as integration
and differentiation, operations that usually arise in analytical analyses.
As noted previously, graphs also can be used for interpolation and extrapolation, and
to perform common mathematical manipulations, such as integration and differentiation.
The data may be of almost any type encountered in engineering and science practice.

65
66 Introduction to Optimization for Chemical and Environmental Engineers

It may be physical property data, such as the variation in density and viscosity with tem-
perature along a tube in a heat exchanger. One of the things that should be kept in mind
is the loss of accuracy arising due to the use of graphs whenever one works with plotted
data. In general, the number of significant figures may be only as great as the size of the
divisions on the graph.
Five sections complement this presentation. Section numbers and subject titles follow:

5.1 Rectangular coordinates

5.2 Logarithmic-logarithmic (log-log) coordinates
5.3 Semi-logarithmic (semi-log) coordinates
5.4 Methods of plotting data
5.5 Applications

5.1 Rectangular Coordinates
Rectangular (sometimes referred to as Cartesian) coordinate graph paper is most generally
used to represent equations of the form

y = mx + b (5.1)

where:
y = variable represented on the ordinate
x = variable represented on the abscissa
m = slope of the line
b = y-intercept at x = 0

For engineering use, the most common form of graph paper is the 8 1/2 × 11-inch sheet,
having 20 lines per inch, with every fifth line accented, and every tenth line heavily
accented. This type of coordinate graph may be obtained on drawing or tracing paper,
with lines in black, orange, or green, and with or without accented and heavy lines; it is
also available in other sizes. Also note that once again the bulk of the development in the
last section of this chapter will key on rectangular coordinates.
All graphs are created by drawing a line or lines, about one or more axes. For the pur-
pose of this chapter, the presentation will be primarily concerned with graphs built around
two axes, the x-axis and the y-axis. The graph data may be presented as coordinates of the
x and y-axis in the form (x, y), or they may be presented as the abscissa (distance from the
y-axis or the x coordinate) and the ordinate (distance from the x-axis or the y coordinate).
Referring to Figure 5.1, point L5 has an abscissa (x coordinate) of +60 and an ordinate of
+40 (y coordinate). Using coordinate notation, the point can be described as (60, 40) where
the points are listed as (x, y). Point L2 can be described as (−6.7, 0).
All coordinates are produced as the result of solving an equation by assigning different
values to x or y, and solving for the value of the other. By convention, equations are usually
stated in the following form: y equals some value(s) of x and other constant(s), for example,
y = mx + b.
Graphical analysis 67

y-axis
60

50
Quadrant II Quadrant I
L5
40
(60, 40)
30
x +4
0.6
y=
20
L4
10 L3 (10, 10)
L2

x-axis
(0, 4)
0
(–6.7, 0)
–10
L1
–20 (–40, –20)

–30
Quadrant III Quadrant IV
–40
–40 –30 –20 –10 0 10 20 30 40 50 60

FIGURE 5.1
Linear coordinates.

5.1.1 Quadrants
The x and y-axes divide a graph into four quadrants. The quadrants are usually num-
bered in a counter-clockwise sequence, beginning with the upper right quadrant (see
Figure 5.1). Note that in quadrants I and IV, the x values are positive while in quadrants
II and III, they are negative. The y values are positive in quadrants I and II and negative
in quadrants III and IV. The quadrant identifications are only of value when information
is provided that a point or result lies in a certain quadrant. Without knowledge of the
precise values of x or y one can still determine whether the x and y values are positive or
negative.

5.1.2 Linear versus Non-Linear

Equations are identified by the type of line (or curve) they produce when plotted on a
graph. A linear equation, as the name implies, will result in a plot of a straight line on a
graph. With the linear equation, each incremental change of x (or y) will result in an incre-
mental change of a fixed ratio in y (or x). The ratio of change between y and x is known as
the slope of the line. A linear equation is also referred to as a first-degree equation. The line
plotted between L1 and L5 in Figure 5.1 is linear.
A nonlinear equation will result in a line having one or more curves, when plotted, and
the ratio of change between x and y is not constant. The line between points N0 and N5
in Figure 5.2 represents the plot of a nonlinear equation. Nonlinear equations come in
a variety of forms and may be quickly identified since they may have at least one term
containing an exponent. The equation y = πx2 (see Figure 5.2) is representative of a non-
linear equation.
68 Introduction to Optimization for Chemical and Environmental Engineers

y-axis
90

80 N5 (5, 78.5)

70

60
N4
50 y = π x2
(4, 50.3)

40

30 N3
(3, 28.3)
20
N2
(2, 12.6)
10
N1
N0

x-axis
(1, 3.1)
0
(0, –0)

–10
–3 –2 –1 0 1 2 3 4 5 6 7

FIGURE 5.2
Nonlinear equation.

5.1.3 Intercepts
Reference to the x and y intercepts are often made when working with graphs. An inter-
cept is the point at which a line crosses the x- or y-axis. The x-intercept is the point where a
line crosses the x-axis (y = 0), and the y-intercept is the point where a line crosses the y-axis
(x = 0). In Figure 5.1, the x-intercept occurs are (−6.7, 0) and the y-intercept occurs at (0, 4).
There can be only one x- and one y-intercept in a linear equation. With nonlinear equa-
tions, there may be no x- or y-intercept, or there may be multiple intercepts. Note that in
Figure 5.3 there are two x-intercepts [(−2, 0) and (0, 0)] and one y-intercept (0, 0).

5.2 Logarithmic-Logarithmic (Log-Log) Coordinates

A logarithmic scale is easily constructed by plotting (on rectangular paper) numbers from
one to ten on the ordinate versus the logarithm of the number on the abscissa. If the points
representing the numbers on the ordinate are projected to the resultant curve and then
upward, the scale formed by the vertical lines will be a logarithmic scale. The construction
is illustrated in Figure 5.4. The utility of the logarithmic scale lies in the fact that one can
use actual numbers on the scale instead of the logarithms. When two logarithmic scales
are placed perpendicular to each other and lines are drawn vertically and horizontally
to represent major divisions, full logarithmic graph paper results, that is, a log-log graph.
The term log-log or full logarithmic is used to distinguish between another kind of graph
or relationship yet to be discussed, that is, semi-logarithmic (see next section). The full
logarithmic graph paper may be obtained in many types depending on the scale length,
sheet size, and number of cycles desired. The paper is specified by the number of cycles
Graphical analysis 69

y-axis
16
(–5, 15) (3, 15)
14

12

x + x2
10

y=2
(–4, 8) (2, 8)
8

4
(–3, 3)
(1, 3)
2
(–2, 0)

x-axis
0
(0, 0)
(–1, –1)
–2

–4
–10 –8 –6 –4 –2 0 2 4 6 8 10
x

FIGURE 5.3
Intercept analysis.

1 2 3 4 5 6 7 8 9 10

10

7
Dependent variable, y

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Log of a dependent variable, x

FIGURE 5.4
The construction of a logarithmic scale.
70 Introduction to Optimization for Chemical and Environmental Engineers

of the ordinate and abscissa; for example, logarithmic 2 × 3 cycles means two cycles on the
ordinate and three on the abscissa. Cycle lengths are generally the same on both axes of a
particular graph. In addition, the distance between numbers differing by a factor of 10 is
constant on a logarithmic scale.
Equations of the general form

y = bx m (5.2)

where:
y = a variable
x = a variable
b = a constant
m = a constant

will plot as straight lines on (full) logarithmic paper. A form of the equation analogous
to the slope-intercept equation for a straight line is obtained by writing and then subse-
quently plotting the equation in logarithmic form, as represented in Equation 5.3.

log y = log b + m log x (5.3)

If log y versus log x were plotted on rectangular coordinates, a straight line of slope m and
y-intercept log b would be obtained. An expression for the slope is obtained by differenti-
ating Equation 5.3.

d (log y )
= m (5.4)
d (log x)

When x = 1, m log x = 0, Equation 5.3 reduces to

log y = log b (5.5)

Hence, log b is the y-intercept.

To illustrate the previous, a linear plot of the equation

y = 2x1.3 (5.6)

on rectangular coordinate paper is given in Figure 5.5. It should be noted that although
the slope was obtained from a ratio of logarithmic differences, the same result could have
been obtained using measured differences if the scale of ordinate and abscissa were the
same (in this case they were).
Inspection of Figure 5.5 reveals that equal measured distances on the logarithmic scale
are equivalent to equal logarithmic differences on the abscissa of that plot. It therefore fol-
lows that since the slope of Equation 5.3 is a ratio of coordinate differences, the slope on
full logarithmic paper could be obtained using measured distances. The plot of Equation
5.6 on 2 × 2 cycle logarithmic paper is given in Figure 5.6. The slope using logarithmic dif-
ferences is

1.08 − 0.43
m= = 1.3
0.60 − 0.10
Graphical analysis 71

1.08
1.0

log10 y 0.43

0.65
m= = 1.3
0.5
0.43
0.5
0.302
0.1 0.6

0
log10 x

FIGURE 5.5
A linear plot of y = 2x1.3 on rectilinear graph paper.

100
80
60

40

20
a

10
y

8
6

4
b

1
1 2 4 6 8 10 20 40 60 80 100
x

FIGURE 5.6
The equation y = 2x1.3 plotted on logarithmic graph paper.

To further illustrate the use of full logarithmic paper, consider the determination of a
mathematical equation to represent data that are known to plot as a straight line on log-log
paper. If data, as provided in Table 5.1, are to be plotted, it is evident that more than one
cycle will be needed in both directions. For x, the data covers the range from

Range ( x ) = 2 × 100 to 101 to 6.2 × 101 (5.7)

72 Introduction to Optimization for Chemical and Environmental Engineers

TABLE 5.1
Log-Log x-y Data
x y
2.0 3.6
3.9 6.0
7.0 9.4
14.0 16.0
26.5 26.0
62.0 50.0

and for the y the range is

Range ( y ) = 3.6 × 100 to 101 to 5.0 × 101 (5.8)

In both cases, two cycles are needed to cover the ranges from 100 to 101 and from 101 to 102.
A general rule would be that the number of cycles is equal to the number of exponents of
ten involved in the range of data, in this case two for each coordinate. The data is plotted
in Figure 5.7. The slope may once again be calculated by:

log 50 − log 3.6

m = c/d =
log 62 − log 2.0
The y-intercept is 2.13; hence the equation of the line is

log y = 2.13 + 0.765 log x (5.9)

100
80
60

40

20
c

10
y

8
6

d
2

1
1 2 4 6 8 10 20 40 60 80 100
x

FIGURE 5.7
Determination of a mathematical equation from raw data.
Graphical analysis 73

In cases where the line x = 1 is several cycles removed from the data range, it may be incon-
venient to extrapolate; the intercept can then be determined by substituting the value of
the slope and one data point and solving for b.

5.3 Semi-Logarithmic (Semi-Log) Coordinates

Graph paper made with one logarithmic scale and one arithmetic scale is termed semi-
logarithmic. The logarithmic scale normally is the ordinate and the arithmetic scale is the
abscissa on semi-logarithmic paper. The paper is available in many styles depending on
the sheet size, type of paper, color of lines, number of cycles, length of cycles, and the
divisions on the uniform scale. The designation semi-logarithmic, 7 × 5 to the 1/2 inch refers
to semi-logarithmic paper whose logarithmic scale contains seven cycles whose uniform
scale contains five divisions to the half inch. The designation 10 division per inch (70 divi-
sions) by two 5-inch cycles refers to paper whose arithmetic scale is seven inches long and
contains ten divisions per inch, and whose logarithmic scale contains two cycles for each
five inches. Thus, a semi-logarithmic graph uses a standard scale for one axis and a loga-
rithmic scale for the other axis, that is, the graph uses scales of different mathematical
proportions. The reason for this use of scales is that a logarithmic (exponential) plot would
quickly exceed the physical boundaries of the graph if one were to plot values of some base
having exponents of 2, 3, 4, 5, 6, 7 and 8 against an ordinary numeric scale.

5.3.1 Plotting Straight Lines on Semi-Logarithmic Paper

Semi-logarithmic paper can be used to graph equations as straight lines. These equations
take the form

y = ne mx (5.10)

or

y = n10 mx (5.11)

where:
y = dependent variable
x = independent variable
n = a constant
m = a constant
e = the natural logarithm base

Here again, a form of the equation analogous to the slope-intercept equation can be
obtained by placing the previous equations in logarithmic form. For example, Equation
5.10 may be written as

log y = b + mx (5.12)

where b = log n
74 Introduction to Optimization for Chemical and Environmental Engineers

An expression for slope is obtained from the differential form of Equation 5.12:

d (log y )
= m (5.13)
dx

so that the y-intercept is log n or b.

One can illustrate the use of semi-logarithmic paper by plotting a segment of the curve
representing the equation

x
log y = 2.2 + (5.14)
46.3

The line may be constructed (see also Figure 5.8) by choosing values of x, solving
for y, and plotting the results. Note that values of y and not log y are plotted on the
ordinate. The slope is a logarithmic difference divided by an arithmetic difference
and in this case may be calculated from the two indicated points (see Figure 5.8) as
follows:

log ( 1587 ) − log ( 158.7 ) 1

m= =
43.6 − 0 46.3

The slope of a line on semi-logarithmic coordinates may be calculated from any two
points, but the simplest method is the one used previously in which the slope is the

5000
y

1000

500

100
0 10 20 30 40 50 60 70 80
x

FIGURE 5.8
x
The equation log y = 2.2 + plotted on semi-logarithmic paper.
46.3
Graphical analysis 75

logarithmic difference on one cycle (equal to log 10 or 1.0) divided by the arithmetic dif-
ference of one cycle. Note that if the cycle had been taken between y = 300 and y = 3000,
the slope would be

log ( 3000 ) − log(300) 1

m= =
59 − 12.7 46.3

It should also be noted that the y-intercept is 2.2, that is, the of log 159.
The method of obtaining a mathematical equation from data that are known to plot a
straight line on semi-logarithmic paper may be illustrated by the use of the data presented
in Table 5.2. The data is plotted in Figure 5.9. These cycles on the logarithmic scale were
necessary because values of y ranged from

4.32 × 10 2 to 2.7 × 10 4 (5.15)

TABLE 5.2
Semi-Log-Log x-y Data
x y
0.93 432
1.52 886
2.56 2,830
3.20 6,000
3.80 12,000
4.50 27,000

Use of semilogarithmic graph paper

50,000

10,000
y

1,000

100
0 1 2 3 4 5
x

FIGURE 5.9
The determination of a mathematical equation from raw data.
76 Introduction to Optimization for Chemical and Environmental Engineers

and three exponents of ten are involved, for example, 2, 3, and 4. The slope may be calcu-
lated from the middle cycle as follows:

1 1
m= =
3.65 − 1.65 2

The line is extrapolated to x = 0 to give a y-intercept of 150; hence, the equation represent-
ing the data is

x
log y = log 150 + (5.16)
2
or

y = 150 ( 10 )
x /2
(5.17)

5.3.2 Other Graphical Coordinates

Triangular (or trilinear) coordinates are used to represent systems defined by three vari-
ables. The graph could be in the form of an equilateral triangle, each side of which is
usually divided into 100 equal parts. This has found some application in mass transfer
applications.1 In addition to the more common types of paper previously discussed, the
practicing environmental and chemical engineer may have need for probability versus
arithmetic, probability versus logarithmic, logarithmic versus reciprocals and other special
type of graphs. Non-standard scales, if desired, may be constructed in a similar manner
to the logarithmic scale shown earlier. For example, logarithmic-probability (log-normal)
graphics find applications in air pollution studies2 to describe particle size distribution.

5.4 Methods of Plotting Data

The simplest procedure to employ in plotting equations of various forms is provided in
Table 5.3. Various additional forms are available in the literature. Details on statistical meth-
ods for calculating the coefficients in the seven equations are provided in the literature.3
TABLE 5.3
Plotting Procedure
y = a + bx Plot y vs. x
y = axn Plot log y vs. log x or y vs. x on logarithmic coordinates

y = c + ax n First, obtain c as the intercept on a plot of y vs. x; then

plot log (y − c) vs. x on logarithmic coordinates
y = abex Plot log y vs. x or y vs. x on semi-logarithmic coordinates
y = abx Plot log y vs. x or y vs. x on semi-logarithmic coordinates
b 1
y = a+ Plot y vs.
x x
x x 1 1
y= Plot vs. x or vs.
a + bx y y x
Graphical analysis 77

5.5 Applications
This last section contains four illustrative examples. Some are extensions of problems dis-
cussed earlier with the material keying on solutions employing graphical approaches.

Illustrative Example 5.1

Consider the following relationship

y = f (x) = A(x) − B(x)

where:

0.429x
A(x) =
x + 15
and

B ( x ) = 0.006 x

Determine the value of x for which f(x) = 0. Use a graphical approach.

SOLUTION:
The describing equation for f(x) is

0.429x
y (x) = f (x) = − 0.006 x
x + 15
Refer to Figure 5.10. It can be seen that f(x) = 0 at approximately x = 56.6.

Illustrative Example 5.2

Refer to the previous example. Determine the value of x that will provide the maximum
value of f(x).

0.5

0.4
CP

0.3
RV
0.2
y

Profit
0.1
Break-even
0

–0.1

–0.2
0 10 20 30 40 50 60 70 80
x

FIGURE 5.10
x-y plot from Illustrative Example 5.1.
78 Introduction to Optimization for Chemical and Environmental Engineers

SOLUTION:
Refer to Figure 5.10 once again. The maximum value of f(x) is 0.75 and occurs at x
approximately equal to x = 15.0. This problem will be revised in next chapter and an
exact analytical solution to the problem will be obtained.

Illustrative Example 5.3

The relationship

11, 900
y = f ( x ) = 162 − 2.33 x −
x
was solved earlier in Chapter 3 (see also Illustrative Example 3.3) using a perturbation
approach. Recalculate the minimum value of f(x) using a graphical approach.

SOLUTION:
The following y vs. x values were provided in Illustrative Example 3.3. These calcula-
tions in Table 5.4 can be plotted. That is left as an exercise for the reader. Hint: The
solution appears here:

x = 71

f ( 71) = −171.0

As expected, the results compare favorably with the answer provided earlier. The reader
should also note that this problem will be solved analytically in the next chapter.

Illustrative Example 5.4

Refer to Illustrative Example 3.7 and consider a slightly modified form of that equation.

11, 900
y = 2.33 x1 + + 1.86 x2 + 10
x1x2
Determine the values of x1 and x2, which will produce the minimum value of y in the
above equation.

TABLE 5.4
y vs. x value for Illustrative Example 5.3
y x
60 −176.1
62 −174.4
64 −173.1
66 −172.1
68 −171.4
70 −171.1
72 −171.1
74 −171.2
76 −171.7
78 −172.3
80 −173.3
Graphical analysis 79

SOLUTION:
When two or more independent variables are to be optimized, the procedure for deter-
mining the optimum conditions may be more difficult than when one independent
variable is involved. For this case, the relationship among y, x1, and x2, could be shown
as a curved surface in a three-dimensional plot, with a minimum value of y occurring
at the optimum values of x1 and x2 However, the use of a three-dimensional plot may
not be practical for some engineering calculations. The optimum values of x1 and x2
can be found graphically on a two-dimensional plot by the method that is provided in
Figure 5.11.
For the equation in question, f(x) is plotted against one of the independent variables x1,
with the second variable x2 held at a constant value. A series of such plots can be made
with each dashed curve representing a different constant value of the second variables.
As shown in Figure 5.11, each of the curves (a, b, c, d, and e) gives one value of the vari-
able x1 at the point where the f(x) or f(x1, x2) is a minimum. The curve AB represents the
locus of all these minimum points and the optimum value of x1 and x2 occurs at the
minimum point on the dashed curve AB – noted as (x1* , x2* ) at y*.
The following constant values of x2 are chosen arbitrarily for the example at hand:

= =
xI 32 =
xII 26 =
xIII 20 xIV 15 xV = 12

Plots of y vs. x1 at each constant value of x2, that is, xI, xII, xIII, xIV, and xV are generated.
These plots are presented as solid lines in Figure 5.12 as curves a, b, c, d, and e. A sum-
mary of these results is presented in Table 5.5 with x2*, and x1* representing the mini-
mum values of x2 for the solid curves a, b, c, d, and e, respectively. The dashed curve
AB represents the relationship for the minimum five x1*’s as a function of y. The point
x1* represents the minimum y, which occurs at approximately x2 III and x1*III ; the corre-
sponding value of y is 121.6.

A B

(a) x2I x2V (e)

y = f(x)

(b) x2II x2IV (d)

x2* = x2III (c)

y*
Optimum (minimum) y*= f (x1*, x2*)
x1* x1

FIGURE 5.11
Optimum value of the function of (y) in terms of two independent variables x1 and x2.
80 Introduction to Optimization for Chemical and Environmental Engineers

A B

(a) x2I = 32 x2V = 12 (e)

y = f(x)

x2II = 26 x2IV = 15
(b) (d)

x2* = x2III = 20 (c)

y*
Optimum (minimum) y*= f (x1*, x2*)
x1* = 16 x1

FIGURE 5.12
Optimum value function of (y) in terms of two independent variables (x1, x2).

TABLE 5.5
Optimal x2, and x1* Results for Illustrative Example 5.4

x3 Optimum x1* Minimum y

128.3
x2I = 32 x1* I = 12.7
123.6
x2II = 26 x1* II = 14.1
121.6
x2III = 20 x1* III = 16.0
123.9
x2IV = 15 x1* IV = 18.5
128.5
xV2 = 12 x1*V = 20.7

The reader should also note that similar graphical procedures can be used when
there are more than two independent variables, for example, x1, x 2, and, x3, where
y = f(x1, x 2, x3). For this case, and as noted earlier, the first step would be to make
a plot similar to Figure 5.12 at one constant value of x3. Similar plots would then
be made at other constant values of x3. Each plot would give an optimum value of
x1, x 2, and y for a particular x3. Finally, the overall optimum value of x1, x 2, and y
could be obtained by plotting of x3 versus the individual optimum value of y, where
y = f(x) = f(x1, x 2, x3).
Graphical analysis 81

References
1. L. Theodore and F. Ricci, Mass Transfer Operations for the Practicing Engineer, John Wiley & Sons,
Hoboken, NJ, 2010.
2. L. Theodore, Air Pollution Control Equipment Calculations, John Wiley & Sons, Hoboken, NJ,
2008.
3. S. Shaefer and L. Theodore, Probability and Statistics Applications for Environmental Science, CRC
Press/Taylor & Francis Group, Boca Raton, FL, 2007.
6
Analytical Methods

The subject of analytical methods serves as an excellent starting point before preceding to
the last subject in this first Part—linear programming. In terms of introduction, once again
consider the following. Since the derivative dx/dy or f′(x), represents the rate of change of
y with respect to the change in x, it is evident that the derivative will be zero if the func-
tion passes through a maximum or minimum. If the occurrence of such a maximum or
minimum is to be determined and located, the derivative must be equated to zero, thus
providing the condition for which the maximum or minimum exists. This analytical pro-
cedure is of considerable value in environmental and chemical engineering calculations
since the location of a maximum or minimum is frequently of practical importance. As one
might suppose, many of the applications are related to economics, for example, maximiz-
ing profit and/or minimizing cost.
Five sections complement this presentation. Section numbers and titles follow:

6.1 Breakeven Considerations

6.2 One Independent Variable
6.3 General Analytical Formulation of the Optimum
6.4 Two Independent Variables
6.5 Three Independent Variables

6.1 Breakeven Considerations
In solving many environmental and chemical engineering optimization problems, one
often selects optimum conditions from the standpoint of economics. Consider, a simple
case in which there is only one independent variable affecting profit P, that is, income I
minus cost C. For example, refer to Figure 6.1. The term y is the dependent variable that
will later be referred to as the objective function and x is the independent variable. The
former term could represent the aforementioned profit and x could be the operating tem-
perature. As can be seen from Figure 6.1, the “system” operates at a loss when operating
below x* and operates at a profit when above x*. The term x* is located at the breakeven point,
that is where the profit is zero.
The above breakeven problem can be solved by three methods presented below:1

Method 1: Solve the problem analytically as follows. Assume I and C can be repre-
sented in equation form, that is,

I = I ( x ) (6.1)

83
84 Introduction to Optimization for Chemical and Environmental Engineers

C = C ( x ) (6.2)

Since

P = I – C (6.3)

the solution for x, when P = 0, provides the breakeven point (of operation).
Method 2: Select various values of x and calculate both the income and cost, and then
the profit. The breakeven point arises when the profit is zero.
Method 3: Graph the results.2 A typical solution is presented in Figure 6.1. See also the
previous chapter.

Economic studies often reveal that certain cost factors increase while others decrease
relative to an independent variable. Now relate this concept to an LNG (liquefied nitro-
gen gas) plant. Liquefied nitrogen gas is used by gas utility companies around the coun-
try to heat residential and commercial establishments. This gas has an extremely low
boiling point and must be kept at a temperature of −260°F to be maintained in a liqui-
dated state. LNG is most easily and safely transported in its liquid form. It is odorless
and colorless in its gaseous state, making it extremely difficult to spot a leak if stored as
a gas. The gas companies transport this liquefied LNG gas from trucks and other sources
and store it in large tanks until it is converted to a gas for customer use. Consider the
following LNG example:
The safety engineer employed at Theohan Gas Company was inspecting the LNG
transport pipes when she noticed the insulation on one of the pipes looked damaged.
After removing a section of the insulation, it was clear it needed to be replaced and the
engineer proposed replacing it with a thicker insulating cover to prevent/reduce leakage.
Looking at this from an economic standpoint, as the thickness of the insulating cover
(x) on an LNG pipe increases, the cost increases, but correspondingly the LNG lost in
travel (the LNG transportation cost) decreases. Here the thickness of the cover serves as
the independent variable, and if the installed cost of the insulation and the LNG transpor-
tation leakage losses could both be determined as functions of the independent variable,

(I, x)

Income, I

(P, x)
Profit, P x
x*

Cost, C

(C, x)

FIGURE 6.1
Breakeven figure.
Analytical Methods 85

a condition would exist for which the total cost is a minimum. This solution is pictured in
Figure 6.2 and the income optimum is located at the square point, x*. Here, the total cost
is plotted against the insulating cover thickness. This problem will be revisited in Part III,
Section 16.2, Illustrative Example 16.2.

6.2 One Independent Variable

The case of a dependent variable and a single (one) independent variable was reviewed in
the previous chapters. Here

y = f ( x ) (6.4)

As noted above, the optimization of y can be generated by setting

dy
= 0 (6.5)
dx

Fixed charges
Total cost

Economic
optimum

LNG transportation cost/profit analysis

Leakage cost loss

x*
Insulation thickness, x

FIGURE 6.2
Total cost vs. insulating cover thickness.
86 Introduction to Optimization for Chemical and Environmental Engineers

The optimum y is obtained by substituting the value of x from Equation 6.5 with Equation
6.6. And, y is a maximum if

d2 y
< 0 (6.6)
dx 2

When x from Equation 6.5 is substituted with Equation 6.4 the calculated y is a minimum if

d2 y
> 0 (6.7)
dx 2

To summarize, a maximum occurs when dy/dx = 0 and d2y/dx2 < 0; a minimum occurs

when dy/dx = 0 and d2y/dx > 0.

Illustrative Example 6.1

Refer to Illustrative Example 3.3.
Consider the following function

11, 900
y = 162 – 2.33x +
x
Determine the maximum (or minimum) of the above function by analytical means.

SOLUTION:
Apply Equation 6.5.

dy 11, 900
= −2.33 − =0
dx x2
Solving for x,

11, 900
x2 = = 5, 107
2.33
x = 71.5
The second derivative d2y/dx2 is negative since

dy 2 23, 800
2
=−
dx x3
The calculated value of y at x = 71.5 is therefore a maximum with a value of

11, 900
y = 162 – ( 2.33 )( 71.5 ) +
71.5
= 162 – 166.6 + 166.4

= 161.8

Illustrative Example 6.2

Refer to Illustrative Example 5.1. Solve the problem analytically.
Analytical Methods 87

SOLUTION:
Once again, the describing equation is

0.429x
y= − 0.006 x
x + 15.0
To proceed analytically, calculate dy/dx and set it equal to zero.

dy 0.429 0.429x
= − − 0.006 = 0
dx x + 15.0 ( x + 15.0)2
Solve for x by any suitable method.

0.429x
0.429 − − 0.006 ( x + 15.0 ) = 0
x + 15.0
x = 17.5 psf
Since dy2/dx2 < 0, the corresponding maximum profit is

( 0.429 )(17.5 ) −
y= ( 0.006 )(17.5 )
17.5 + 15.0
= 0.231 − 0.105

= 0.126
The above analytical result compares favorably with the results provided earlier in
Illustrative Example 5.1.

Illustrative Example 6.3

Refer to Illustrative Examples 3.4 and 4.4. Solve the following problem employing an
analytical approach if x1 = 1, x2 = 10, and k = 1.4.

SOLUTION:
As noted in Illustrative Example 3.2 the total work is given by

  k −1
 
 k −1
  
  x2   k   x3   k 
y = f ( x2 )   +  − 2
 x1   x2 
 
If this quantity is to be a minimum, then the derivative must be zero, that is,

dy  k − 1   k  −  k   k − 1   k   k  
 1− k   1  1− k   1− 2 k 

=  x
 1 x 2 −  x
 3 x2 =0
dx2  k  k  

Solving for x2 yields

 2k − 2   1− k   1− k 
     
x2 2k 
= x1 k 
x3 k 

or

 k −1  1− k   1− k 
2     
x2  k 
= x1 k 
x3 k 

88 Introduction to Optimization for Chemical and Environmental Engineers

Thus,

x22 = x1x3

x2 = ( x1x3 )
1/2

For x1 = 1, x3 = 10, one obtains

1/2
x2 = (1)(10 )  = 3.16

dy 2
Since < 0 the corresponding minimum value of y is
dx22
y = f ( x2 = 3.162 ) = 2.726

This result compares favorably with the answers provided earlier in Illustrative
Examples 3.4.
Interestingly, Happel2 addressed the problem of maximizing the work requirement
for a three-stage compressor operating between 1 and 10 atm. This problem was solved
using an analytical approach along with the method of steepest ascent2 (see also the
previous development) in Chapter 4, Illustrative Example 4.6, and in Part III, Chapter
13, Illustrative Example 13.6. The analytical method for the two intermediate pressures
P2 and P3 yielded the equations (see Illustrative Example 6.7 for details):

( )
1/3
P2 = P12 P4

(
P3 = P42 P1 )

With P1 = 1 and P2 = 10 atm,

P2 = 101/3 = 2.154 atm

P3 = 1001/3 = 4.642 atm

and

E = 2.575 ft ⋅ lbf

6.3 General Analytical Formulation of the Optimum

When more than one independent variable is involved in determining a function, for exam-
ple, profitability, a more elaborate treatment is necessary than was employed above. Thus, for
a system of x1,x2,x3, … xn independent variables, the function y will depend on these variables.

y ( profit ) = f ( x1 , x2 , x3 ,…, xn ) (6.8)

Analytical Methods 89

Instead of profit, y may naturally be expressed in other convenient or equivalent forms.

The problem is to specify x1,x2,x3, … xn so that y will be a maximum (or minimum). In gen-
eral, this will involve some type of trial-and-error.
It is possible in some cases to set up an explicit function that will relate the objective func-
tion y to the independent variables. Differentiation will then result in a series of partial deri-
vations dy/dx1, dy/dx2 etc. Setting these partial derivatives equal to zero will result in the same
number of simultaneous equations as the variables involved. These equations, subject to limi-
tations, can be solved for the corresponding optimum values of these variables. The necessary
conditions for an optimum at a point where the first derivatives become zero include restric-
tions on the second derivatives, continuity of the function involved throughout the range of
independent variables, and absence of optimum conditions at limiting values of one or more
of the variables. These conditions can on occasion restrict the applicability of this system.2
Thus, for the case of a two-variable function, y = f(x1, x2), a point x1 = a, x2 = b, will have a
maximum value y = f(x1, x2) if the following conditions hold:

∂y ∂y
= = 0 (6.9)
∂x1 ∂x2
with

∂2y ∂2y
2
< 0; 2 < 0 (6.10)
∂x1 ∂x2
and

2
∂ 2 y d2 y  ∂ 2 y 
>  (6.11)
∂x12 dx22  ∂x1∂x2 
Similarly, if ∂ 2 y / ∂x12 and ∂ 2 y / ∂x22 are positive and Equation 6.11 holds, there will be a
minimum. But, if

2
∂2y ∂2y  ∂2y 
>  (6.12)
∂x12 ∂x22  ∂x1∂x2 
then the point concerned will be a saddle point. The case where

2
∂2y ∂2y  ∂2y 
=  (6.13)
∂x12 ∂x22  ∂x1∂x2 
is open and the value may be a maximum, a minimum, or neither.

6.4 Two Independent Variables

Illustrative Example 6.4
An objective function is specified as:

f ( x1 , x2 ) = f = (x12 − 2x1 + x22 + x1x2 + 1)

90 Introduction to Optimization for Chemical and Environmental Engineers

Determine if the function has a maximum or a minimum. Also, calculate that value.

SOLUTION:
For the above equation

∂f
= 2x1 − 2 + x2 = 0
∂x1

∂f
= 2x2 + x1 = 0
∂x2
The following solution results, when the above two equations are set equal to zero.

4
x1 = −
3

2
x2 =
3
Since

∂2 f
=2
∂x12

∂2 f
=2
∂x22
the solution provided for x1 and x2 represent a minimum with

y = f ( x1 , x2 ) = 5

Illustrative Example 6.5

A profit equation is given by:

11, 900
P = 162 − 2.33 x1 − 1.86 x2 − =0
x12 x2
Calculate values of x1 and x2 that would either maximize or minimize P.

SOLUTION:
Begin by generating the first derivative and setting them equal to zero.

∂P 11, 900
= −2.33 + 2 =0
∂x1 x1 x2
∂P 11, 900
= −1.86 + =0
∂x2 x1x2 2
Solving simultaneously yields

x1 = 16
x2 = 20
Analytical Methods 91

The corresponding P (16,20) is a maximum since both second derivatives are negative.
Thus,

P (16, 20 ) = 50.3

Illustrative Example 6.6

Obtain values for T and P, which will yield the minimum emissions (E) in the equation
below by employing an analytical method of analysis.

11, 900
E = 2.33T + + 1.86P + 10
TP

SOLUTION:
Begin once again by presenting the first derivatives and setting them equal to zero.

∂E 11, 900
= 2.33 − =0
∂T T 2P

∂E 11, 900
= 1.86 − =0
∂P TP 2
For this example, the simultaneous solution to the above three equations yields

T = 16

T = 20

Furthermore,

E = E (16, 20 ) = 121.6

Since

∂ 2E ( 2 )(11, 900 )
= =+
∂T 2 (16)2 ( 20 )

∂ 2E ( 2 )(11, 900 )
= =+
∂P 2 (16 ) (20)2
the optimum conditions represent a minimum emission. The interested reader should
compare the results of this illustrative example with that in the previous illustrative
example.

Illustrative Example 6.7

Refer to Illustrative Example 3.8, 4.6, and 6.3. Resolve the Illustrative Example 3.8
“search” problem employing a purely analytical approach.
92 Introduction to Optimization for Chemical and Environmental Engineers

SOLUTION:
The describing equation is once again

  k −1
 
 k −1
 
 k −1
  
  x2   k   x3   k   x 4   k 
y =   +  +  − 3  ; x1 = 1, x 4 = 10, k = 1.4
 x1   x2   x3 
 

If this quantity is to be a minimum, ∂y/∂x2 and ∂y/∂x3 must be zero.

 1− k   1  1− k   1− 2 k 
∂y   −     
= x1 k 
x2  k  − x3 k 
x2 k 

∂x 2

 1− k   1  1− k   1− 2 k 
∂y   −     
= x2 k 
x3  k  − x 4 k 
x3 k 

∂x 3
Setting ∂y/∂x2 and ∂y/∂x3 = 0, ultimately leads to

x22 = x1x3 and x32 = x2 x 4

so that

( ) ( )
1/2 1/2
x2 = x12 x 4 x2 = x 4 2 x1

Setting x1 = 1 and x4 = 10 yields the solution

x2 = 2.154

x3 = 4.642

and

y = y ( 2.154, 4.642 ) = 2.572

One can further show that (see also Part III Chapter 13)

∂2y
=+
∂x2 2

∂2y
=+
∂x3 2

∂2y
=+
∂x2∂x3
and

 ∂2y   ∂2y  ∂2y

    >
 ∂x2   ∂x3  ∂x2∂x3
2 2

Therefore, the previously calculated y is a minimum.

Analytical Methods 93

6.5 Three Independent Variables

This section presents one illustrative example involving three independent variables.

Illustrative Example 6.8

Consider now a traditional mathematician’s exercise to determine the maximum or
minimum of a function dependent on three variables.

P = 8 x + 6 y + 8 z + xy + xz + yz − 2x 2 − 3 y 2 − 4 z 2

SOLUTION:
∂P
= 8 + y + z − 4x = 0
∂x

∂P
= 6 + x + z − 6y = 0
∂y

∂P
= 8 + y + x − 8z = 0
∂z

The solution to the above three linear simultaneous equations is

x = −2.900, y = 1.794, z = 1.806

for which

P = 78.07

To determine whether the above P is a maximum or minimum, calculate the following

three second-order partial derivatives.

∂ 2P
= −4
∂x 2

∂ 2P
= −6
∂y 2

∂P
= −8
∂z 2

Since the three derivatives are negative, the value of P of 78.07 is a maximum; that is,
any other calculated P for a different x, y, and z would produce a value of P less than
78.07. The “proof” is left as an exercise for the reader. However, if one were to select
values of x, y, and z straddling the maximum set of values, P should be below 78.07.
For example, the value of P for x = 2, y = 2, and z = 2 is −20. For x = y = z = 10, the value of
P is −380.
94 Introduction to Optimization for Chemical and Environmental Engineers

References
1. L. Theodore, Chemical Engineering: The Essential Reference, McGraw-Hill, New York City, NY,
2014.
2. J. Happel, Chemical Process Economics, John Wiley & Sons, Hoboken, NJ, 1958.
7
Linear Programming

Linear programming is one of several techniques for optimization that has increased
in usefulness because of the availability of digital computers. The title results from the
assumptions in the method that linear relationships describe the problem/system/rela-
tionship under consideration. For example, each production element increase per day in a
process could produce a corresponding linear profit increase per day. However, there are
many instances where a linear model does not accurately describe the totality of the prob-
lem/system/relationship. It may then be necessary to resort to other procedures such as
nonlinear programming; some brief details will be presented for handling the nonlinear
situation. The definitions associated with some of these terms appear earlier in this Part.
The reader should note that the terms linear programming and nonlinear programming
are essentially similar from an applications perspective. There were problems around the
middle of the last century because of some difficulties that arose in attempting to solve
nonlinear programming problems. However, the arrival of the modern-day computer (see
Chapter 1) and sophisticated software (e.g., Excel) have removed these problems. Unless a
solution is presented graphically or algebraically, it will be obtained directly from Excel so
that the reader need not be concerned with whether a system’s pertinent describing equa-
tions and constraints are linear or nonlinear. Thus, details regarding the solution method-
ology for both linear programming and nonlinear programming problems as well as for
some of the illustrative examples will not be presented. They are simply not necessary and
beyond the scope of this introductory optimization text. However, illustrative examples
involving traditional optimization will address both the development of the describing
equations, that is, the model, and the solution to the problem.
Other than linear programming, dynamic programming is another often used optimiza-
tion method by the practicing environmental and chemical engineer. As described earlier,
it is a process involving decisions made at multiple stages. Just as an optimization model is
defined by its variables, constraints, and objective function, a multistage decision process
is defined by its stages and decisions. At any stage, there is often enough information about
decisions made at preceding stages to make informed decisions at the present stage. Thus,
some optimization problems can be reformulated as multistage decision process problems.
One then proceeds through subsequent stages until the final stage of the process is reached.
Some additional details are provided in the next section.
As one might suppose, there are numerous other optimization methods available to the prac-
ticing environmental engineer and scientist. Several of these are detailed in Perry’s Handbook:1

1. Mixed integer
2. Mixed integer linear programming
3. Mixed integer nonlinear programming
4. Quadrative programming
5. Differential/nondifferentiable systems
6. Convex/nonconvex

95
96 Introduction to Optimization for Chemical and Environmental Engineers

The previously mentioned topics are not reviewed in this text. The reader is referred to
the optimization literature2–6 for details.
Five sections compliment the presentation for this chapter:

7.1 Definitions
7.2 Basic concepts of optimization
7.3 Applied mathematical concepts in linear programming
7.4 Applied engineering concepts in linear programming
7.5 Other real-world applied concepts in linear programming

7.1 Definitions
Seven key definitions – as they apply to optimization and linear programming are pro-
vided below (in alphabetical order).

1.
Algebra: A branch of mathematics in which letters are used to represent basic
arithmetic relations. As in arithmetic, the basic operations of algebra are addition,
subtraction, multiplication, and division. Arithmetic, however, cannot generalize
such mathematical relations as the Pythagorean theorem (which states that the
sum of the squares of the sides of any right triangle is also a square). Arithmetic
can produce specific instances of this relations (e.g., 3, 4, and 5, where 32 + 42 = 52).
Algebra can make a purely general statement that fulfills the conditions of the
theorem, that is, a2 + b2 = c2. Any number multiplied by itself is termed squared
and is indicated by a superscript number 2. For example, 3 × 3 is noted as 32 and is
termed three squared; similarly, a × a is equivalent to a2. Classical algebra, which is
concerned with solving equations, uses symbols instead of specific numbers and
employs arithmetic operations to establish procedures for manipulating symbols,
while modern algebra has evolved from classical algebra by placing more atten-
tion to the structures within mathematics.
2.
Constraint: “A constraining or being constrained…confinement or restrictions…
forced…something that constrains.” The term constraint is also defined as “to bond
together, draw together…to force into or hold in, close bounds, confine…restrain.”
3.
Dynamic programming: The dimensionality of a particular operation is reduced
from (m)(n) variables to m repeated n times. Thus, if it is not practical or possible
to solve a large complex system of equations, the system is reduced in complexity
in a manner that requires the solution of a smaller subset of the original problem,
noting that for each n (e.g., each stage in a mass transfer operation)7, 8 only m vari-
ables are involved in the optimization process. One can then work backward and
combine the solutions generated at each n.
4.
Linear algebra: A branch of mathematics that is concerned with systems of linear
equations, linear transformations, vectors, vector spaces, and related topics. The
major interest is with linear equations.
5.
Linear programming: A mathematical and operations research technique used
in engineering, science, administrative engineering, and economic planning to
Linear Programming 97

maximize the linear function(s) of a large number of variable(s) subject to certain

constraint(s). It is basically used to find a set of values, chosen from a prescribed set
of numbers, that will maximize or minimize, that is, optimize, a given equation.
6.
Operations research: A confusing term that most often has been used to describe
linear programming.
7.
Optimization: As noted earlier in this Part, optimization has come to mean differ-
ent things to different people. Its definition has also varied in the literature. In its
most elementary and basic form, optimization is concerned with the determination
of the optimum solution to a given problem. This process is required in the solu-
tion of many problems in environmental and chemical engineering that involve
the maximization or minimization of a given function(s). Some specific examples
include
a. Environmental engineering research
b. Applied science
c. In the selection of control variables to accomplish a desired operation
d. In the scheduling of a series of operations or events to control dates of a given
project
e. The development of optimal layouts of organizational units within a given
design space, and so on.7

7.2 Basic Concepts of Optimization

The term linear programming defines a particular class of optimization problems in which
the constraints of the system can be expressed as linear equations or inequalities and the
function is linear with respect to the independent design variable(s). These techniques are
widely used to solve a number of engineering, environmental, chemical, structural, eco-
nomic, and societal problems. The primary reasons for its wide use today are the availabil-
ity of a host of commercial software to solve complex problems and the ease with which
any data variation that could involve sensitivity analysis can be handled.
As noted in the earlier chapters in this Part, the optimization problem in which the
performance measure is a function of one variable is the most elementary type of opti-
mization problem since it is a type of problem that the environmental and chemical engi-
neer commonly encounters in practice. It also allows single-variable optimization to be
employed as a sub-problem within iterative procedures for solving multivariable optimi-
zation problems.
Constraints, constraints, constraints. Constraints continue to be imposed on individuals,
organizations, groups by governments, religions, relatives, friends, and so on. It seems to
be getting worse each year. And, one of the hallmarks of linear programming is the inclu-
sion of constraints in the problem, that is, the objective function. Linear programming
involves linear constraints. For example, the function

f ( x ) = x 4 + 3 x 2 − 2x + 5; all values of x (7.1)

is an unconstrained function. However, the function

98 Introduction to Optimization for Chemical and Environmental Engineers

f ( x ) = x 6 − 4 x + 6; − 3 ≤ x ≤ 6 (7.2)

is a constrained function. When constraints are present, the function being optimized is
usually referred to as the objective function to distinguish it from any other functions that
may be used to define the constraints or the problem. Although the emphasis to follow is
on linear constraints, many real-world applications also involve nonlinear constraints and
nonlinear objective functions. To further complicate the problem, the constraints can be of
an equality or monoequality/inequality nature.
From a basic perspective, the problem could be to optimize the objective function

y = f ( x1 , x2 ) (7.3)

The constraints can include

g ( x1 , x2 ) < 0 (7.4)

x1 ≥ 0 (7.5)

x2 ≥ 0 (7.6)
This is a simple yet typical problem in optimization – the maximization or minimization of
a real-value function of a number of real variables (occasionally just a single variable) subject
to a number of constraints (occasionally the number is zero). However, the mathematical
model normally contains several (n, e.g.) variables known as the independent design variables
denoted as x1, x2, x3, … xn. This set of variables has been described by some as a vector, x. One
may view each x as one dimension in the set (space) of variables, and any particular set of
variables, and any particular set of values for these variables in this system as a solution.
One of the more important recent areas for the application of mathematical optimiza-
tion techniques is in engineering and science. Many problems require an iterative and/or
trial-and-error solution and the problem of concern is often stated in an ambiguous and
open-ended way. The first task is to decide what the problem is and to identify the require-
ments or specifications for any proposed and/or feasible solution. One then generates a
concept, perhaps in the form of a rough configuration, for the object, system, or process to
be optimized. This phase of the problem is usually the most difficult and can require the
greatest ingenuity and engineering judgment.
The next step attempts to model mathematically the object, system, or process. This model
may assume a wide variety of forms, but the usually preferred ones are mathematical in
nature. One may then use many mathematical, scientific, and engineering tools. These can
involve not only the most empirical of data correlation but also the most rigorous math-
ematical procedures to the least formal cut-and-try methods for assisting in the analysis of
this aspect of the solution.
As noted above, it may happen that certain functional equality constraints also apply, for
example,

f1 ( x1 , x2 , x3 ,… , xn ) = 0

 (7.7)

f k ( x1 , x2 , x3 ,… , xn ) = 0
There may also be certain inequality constraints such as
Linear Programming 99

g1 ( x1 , x2 , x3 ,… , xn ) < 0

 (7.8)

g m ( x1 , x2 , x3 ,… , xn ) < 0
that are specified on the space so that only points in a portion of the total space may be
considered as acceptable values for the solution of the problem. This region is defined as
the feasible space and points in the region are defined as feasible solutions and/or values.
The previously mentioned equality constraints – Equation 7.7 – come from functional
relationships that must be satisfied among the aforementioned n variables. If there is to
be a choice in the values for which at least some of the variables may be assumed, then k
must be smaller than n. The inequality constraints in Equation 7.8 usually arise from some
specified limitations (maximum permissible stress, minimum emissions, minimum allow-
able temperatures, etc.). Note that there is no upper limit on the value of m. Additional
information is provided in Section 7.4.
Summarizing, the optimization problem to be considered is that of minimizing (mini-
mizing should be italicized) f(x) subject to certain constraints. The term f(x), is a real-valued
function called the objective function (or cost function). A vector x is an n-vector of indepen-
dent variables x1, x2, …, xn. These variables are often referred to as decision variables. The
problems require finding the “best” values of the vector x over all the possible values of x.
The “best” value refers to the one that results in the smallest value of the aforementioned
objective function. Optimization problems can also involve maximization of the objective
function. This can be represented similarly to the minimization presentation above since
maximizing f(x) is equivalent to minimizing f(x). Therefore, one can apply the presented to
the minimization problem without concern.
Finally, the following comments are provided on the objective function. It may be quite
simple and easy to determine, or complicated and difficult to calculate. The objective func-
tion may also be very elusive due to the presence of conflicting or dimensionally incompat-
ible sub-objectives; for example, one might be asked to operate an absorber7 at minimum
cost and that it is aesthetically appealing. In some cases, it may not be possible to find a
quantitative objective function. In this case, one can assume that an alternative suitable
objective function can be formulated.

7.3 Applied Mathematical Concepts in Linear Programming

As noted earlier, linear programming deals with the determination of an optimum
solution of a problem expressed in linear relationships where there may be a large
number of possible solutions. In general, the methods of solution employed are of a
trial-and-error nature, where the procedure systemically follows a mathematical treatment
that minimizes one’s labor involved and ensures that a correct result has been obtained.
The following graphical approach to a linear programming problem illustrates some of the
mathematics employed in the solution.
Consider now an example that involves minimizing an objective variable E (e.g., expen-
diture or emission)

E = a1x1 + a2 x2 (7.9)
100 Introduction to Optimization for Chemical and Environmental Engineers

subject to two constraints

B ≤ b1x1 + b2 x2 (7.10)

C ≤ c1x1 + c2 x2 (7.11)

In addition,

x1 ≥ 0 (7.12)

x2 ≥ 0 (7.13)

In terms of a solution, first examine the previous system of equations analytically. The
problem can be viewed in Figures (7.1) and (7.2). First, note that only the first quadrant need
be considered since Equations 7.12 and 7.13 must hold. Equations 7.10 and 7.11 are rewritten
in the following form:

B  b1 
x2 ≤ − x1 (7.14)
b2  b2 
and

C  c1 
x2 ≤ − x1 (7.15)
c2  c2 

x2 x2

Equation Equation
(7.14) (7.15)
x2

x1 x3
(a) (b)

x1

x2
(c)

FIGURE  7.1
Two-variable minimization problems.
Linear Programming 101

x2

E 1 =a 1
E*

=
x2

E2

a 1 x 1+
=
E

x1
=

+
3
a

a2 x2
1x

x2
a2
1 +
a
2x
2
x2
x1
(a)

x2
E1
E 2 E*
E3
=

x2
x1
(b)

FIGURE  7.2
Solution to a two-variable minimization problem.

It is apparent that the only values of x1 and x 2 that will satisfy Equations 7.14 and 7.15
lie above the solid lines (the cross-hashed area) in Figure 7.1. This shaded area is termed
the constraint set. Now place in Figure 7.2 a series of straight lines (or contours) corre-
sponding to Equation 7.9 with a different value of E used for each line (see Figure 7.2a).
Of all these lines, the smallest E for which x1 and x 2 remain in the specified region will
be that minimum E (see Figure 7.2b), i.e., E3. This will yield that proportion of x1 and
x 2 that minimizes the objective function. Stated in other words, one desires the lowest-
value contour having some point in common with the constraint set. For the particular
value of slope chosen for Equation 7.9, the minimum point is seen to correspond to E
where

E3 = Es* (7.16)

The point that represents the minimum is therefore determined by the constraints a1
and a2 of the problem. Further, if the constants c1 and c2 in Equation 7.11 are changed, it
is possible that the minimum point would occur elsewhere. The situation is essentially
reserved for a maximization problem, that is, for a profit P. This is demonstrated in
Figure (7.3).
It should be pointed out that if the objective function – Equation 7.9 – was nonlinear,
the contour lines would be curved, and if the constraint equations were nonlinear, the
constraint lines would not be straight. It should be intuitively obvious that locating the
minimum point graphically is more complicated in this case.
102 Introduction to Optimization for Chemical and Environmental Engineers

x2

P3
=P
*
x2

P3
*
P2
P1
x1
x1

FIGURE  7.3
Possible solution to a two-variable maximization problem.

The previously mentioned methodology can be extended to more than two variables
and two constraints. The following would apply for three variables and four constraints:

y = f ( x1 , x2 , x3 ) = a1x1 + a2 x2 + a3 x3 (7.17)

and

B = b1x1 + b2 x2 + b3 x3 (7.18)

C = c1x1 + c2 x2 + c3 x3 (7.19)

D = d1x1 + d2 x2 + d3 x3 (7.20)

E = e1x1 + e2 x2 + e3 x3 (7.21)

and

x1 ≥ 0 (7.22)

x2 ≥ 0 (7.23)

x3 ≥ 0 (7.24)

Naturally, the previous can be rewritten for n objective functions and m constraints.

7.4 Applied Engineering Concepts in Linear Programming

An engineer’s first task in problem-solving is to decide what the problem is and to identify
the requirements or specifications for any proposed and/or feasible solution. One then
generates a concept, perhaps in the form of a rough configuration, for the object, or process
Linear Programming 103

to be optimized. As noted previously, this phase of the problem is often the most difficult
and requires the greatest ingenuity and engineering judgment.
The next step attempts to model the object, system, or process. This model may assume
a wide variety of forms, but the usually preferred ones are mathematical in nature. One
may then use many mathematical, scientific, and engineering tools. As noted earlier, these
can involve not only the most empirical of data correlations but also the most rigorous
mathematical procedures or formal cut-and-try methods for assisting in the analysis of
this aspect of the solution.
Regarding constraints, Figure (7.4) illustrates a simple situation involving the operation
of a control equipment absorber7–10 with temperature and pressure limits for which each
depends on the magnitude of the other. The former might come from limitations due to
equilibrium considerations, since “the higher the temperature in the absorber the lower
the recovery absorptive capacity of the absorbing liquid while the higher the pressure in
the absorber increases the absorption.”8–10 The region of feasible solutions for the absorber
problem is the region enclosed by the constraints. The optimization problem is to find the
one set of variables or point in the feasible solution space that corresponds to the best solu-
tion. In order to locate the best feasible solution of the potentially infinite number of feasi-
ble solutions using a mathematical optimization technique, it is essential that a meaningful
objective function f(x) of the independent design variables be specified for this operation
concerned with the control and/or recovery of a particular component of gaseous emis-
sion. For simplicity, assume that one must choose for the aforementioned two independent
variables – temperature T and pressure P. The space is, therefore, two-dimensional, with
temperature and pressure as the coordinates. In practice, there usually are certain inequal-
ity constraints placed upon the range on both T and P. As shown in Figure 7.4, there could
be lower limits for the temperature and pressure, and an upper limit for each that depends
on the magnitude of the other.
Once again, the optimization problem is to find the one set of variables or point, x, in
the feasible solution space that corresponds to the best solution. In order to locate the best

Upper constraint on
Lower both T and P
Pressure, P

constraint Feasible
on T solution

Lower constraint on P

Temperature, T

FIGURE  7.4
Feasible solution for operation of an absorber.
104 Introduction to Optimization for Chemical and Environmental Engineers

of the potentially infinite number of feasible solutions using a mathematical optimization

technique, it is essential that the aforementioned objective function f(x) – see Equation 7.25,
for example – of the design variables be specified to provide for a comparison of alterna-
tive solutions.
Perhaps Happel11 best described linear programming from a practicing engineer’s per-
spective. His edited writing follows. The general problem of finding a maximum usually
assumes that some partial derivative of the function y = f (x1, x2, x3) vanishes within the
range of interest of the variables. In practical problems, it often happens, however, that the
optimum solutions will be obtained at the boundary or limiting values of the variables.
This is, of course, always true where the function to be optimized is linear. In such cases,
explicit analytical solution is not possible, and the optimum must be obtained by some
method or iteration.
In general, even the simplest type of formulations may require a great deal of computa-
tion in practical applications, which is facilitated by the availability of high-speed com-
puters. Linear programming, discussed in greater detail subsequently, is beginning to be
used to a substantial extent by practicing engineers. As noted earlier, linear programming
deals with the determination of an optimum solution of a problem expressed in linear
relationships where there are a large number of possible solutions. In general, the methods
employed are trial-and-error, where the procedure systematically follows a mathematical
treatment that minimizes the labor involved and ensures that a correct result has been
obtained.
In order to apply linear programming to a problem, the relationships between vari-
ables must be expressed as a set of linear equations or inequalities. The variables must
be otherwise independent and must exceed the number of equations. Mathematically
expressed,

∑ a x = b ( i = 1, 2,…, m; j = 1,…, n; n > m) (7.25)

i
ij i i

where the sum of the coefficient aij times the value of each variable xi equals a requirement
bi in the ith equation. The number of variables must exceed the number of equations, so
that an infinite number of solutions is possible. The best one is selected. The relation-
ships in Equation 7.25 are often expressed as inequalities. “Dummy” additional variables
can then be introduced in order to convert these inequalities into appropriate equations.
Another requirement is that no quantity can be negative. Thus, mathematically

xi ≥ 0 and bi ≥ 0 (7.26)

This requirement will ensure that only useful answers are obtained.
The solution of a problem by linear programming also requires a profit function or
objective to be maximized (or minimized). This requirement is stated as

P= ∑ c x = maximum ( or minimum ) (7.27)

i
i i

where ci is the profit (or cost) per unit of xi used. The solution is carried out by an iterative
process, which at each stage of calculation assigns either zero or positive values to all xi
variables. As the calculation proceeds, the values selected, while meeting the constraints
imposed, will tend to increase the summation to be maximized at each stage. If a solution
exists, a maximum profit value will finally be realized.”
Linear Programming 105

7.5 Other Real-World Applied Concepts in Linear Programming

This last section attempts to clarify the ideas presented previously via three illustrative
examples. An attempt will also be made to relate the variables to real-world quantitative
factors. That will occur in the next part of the text. In fact, the first illustrative example will
be revisited as it applies to pollutant emission concerns from a coal-fired boiler in Part II.

Illustrative Example 7.1

You have been requested to obtain the maximum value of P subject to the following
conditions:
Objective function:

P = 0.3 x1 + 5x2 (1)

Constraints:

x2 ≤ −0.5x1 + 3 (2)

x2 ≤ −3 x1 + 6 (3)

x1 ≤ 3 (4)

x2 ≤ 4 (5)

Once again, the solution is first provided in graphical form. Refer to Figure 7.5.
Conditions (4) and (5) require that the solution, that is, the values x1 and x2 that will max-
imize P, lie within the rectangle provided in Figure 7.5a. Conditions (2) and (3) require
that the solution be within the area ABCD in Figure 7.5b. When the objective function
equation is superimposed on Figure 7.5b, Figure 7.5c results. The maximum P is located
at x1 = 1.2 and x2 = 2.4 for which P = 15.6. Excel provides the same solution.

Illustrative Example 7.2

A company manufactures two products, A and B, with an accompanying unit profit of
PA and PB, respectively. A minimum number of N products can be produced on a daily
basis. The daily production ratio of A and B is xA and xB, respectively. In addition, the
fractional defective rate of producing A and B are a and b, respectively. Each defective
product reduces the profit associated with product A and B by c and d, respectively.
Determine the describing equations (the model for the production rate of A and B) that
will maximize daily profits P.

SOLUTION:
As noted, xA and xB are the daily production rates of A and B respectively. Therefore,

P = x A PA + xB PB − x A ac − xBbd

x A + xB ≥ N

x A , xB ≥ 0

This problem will be evaluated later in the text.

106 Introduction to Optimization for Chemical and Environmental Engineers

(5)
4.0
(3)
B
3.0
x2 x2 C

2.0 (4)

(2)
1.0

1.0 x1 3.0 A D x1
(a) (b)

5.00
4.80
3.12
Objective function
P*
x2 = 2.4

x1 = 1.2 5.2 6.7 8.2

(c)

FIGURE  7.5
Graphical solution for Illustrative Example 7.1.

Illustrative Example 7.3

A batch chemical reactor12 can produce either product A, B, or C. The daily production
levels of each are represented by xA, xB, and xC with an accompanying unit profit of
PA, PB, and PC, respectively. The two-unit expressions of concern for each product are
manpower (M) and energy costs (E), neither of which can exceed Y and Z, respectively.
Develop the linear programming model to optimize the profit P for this system.

SOLUTION:
Based on the problem statement one concludes:

P = PA x A + PB xB + PC xC

M A x A + M B x B + MC x C ≤ Y

EA x A + EB xB + EC xC ≤ Z

x A , x B , xC ≥ 0

Numerical problems of this nature will also be revisited later.

Linear Programming 107

References
1. D. Green and R. Perry (editors), Perry’s Chemical Engineers’ Handbook, 8th edn, McGraw-Hill,
New York City, NY, 2007.
2. R. Ketter and S. Prawel, Modern Methods of Engineering Computation, adapted from, McGraw-
Hill, New York City, NY, 1969.
3. L. Lapidus, Digital Computations for Chemical Engineers, adapted from, McGraw-Hill, New York
City, NY, 1962.
4. E. Chong and S. Zak, An Introduction to Optimization, 4th edn, John Wiley & Sons, Hoboken, NJ,
2013.
5. G. Reklaitis, A. Ravindran, and K. Ragsdell, Engineering Optimization: Methods and Applications,
John Wiley & Sons, Hoboken, NJ, 1983.
6. A. Levy, The Basics of Practical Optimization, Society for Industrial and Applied Mathematics,
Philadelphia PA, 2009.
7. H. Mickley, T. Sherwood, and C. Reed, Applied Mathematics in Chemical Engineering, McGraw-
Hill, New York City, NY, 1957.
8. L. Theodore and F. Ricci, Mass Transfer Operations for the Practicing Engineer, John Wiley & Sons,
Hoboken, NJ, 2010.
9. L. Theodore, Chemical Engineering: The Essential Reference, McGraw-Hill, New York City, NY,
2014.
10. L. Theodore, F. Ricci, and T. VanVilet, Thermodynamics for the Practicing Engineer, John Wiley &
Sons, Hoboken, NJ, 2009.
11. J. Happel, Chemical Process Economics, John Wiley & Sons, Hoboken, NJ, 1958.
12. L. Theodore, Chemical Rector Analysis and Application for the Practicing Engineer, John Wiley &
Sons, Hoboken, NJ, 2008.
 Part II

Environmental Engineering
Applications

Five chapters compliment the presentation in Part II. The titles follow:

Chapter 8: Air Pollution Management

Chapter 9: Water Pollution Management
Chapter 10: Solid Waste Management
Chapter 11: Health Risk Assessment
Chapter 12: Hazard Risk Assessment

Optimization provides important methods for decision-making in environmental

engineering. It has evolved from a methodology of academic interest into a technology
that continues to have a significant impact on both engineering research and practice.
Optimization algorithms include tools for experimental design, model development, and
statistical analysis; process synthesis analysis, design, and retrofit; model predictive con-
trol and real-time optimization; and, planning, scheduling, and the integration of process
operations.
Part II keys on optimization applications in environmental engineering. The five top-
ics addressed are: air pollution management (Chapter 8), water pollution management
(Chapter 9), solid waste management (Chapter 10), health risk assessment (Chapter 11),
and hazard risk assessment (Chapter 12). Each chapter contains a brief introduction to the
subject in question which is followed by a host of illustrative examples keying on indus-
trial applications that employ some—if not all—of the optimization methods introduced
in Part I.
The reader should note that illustrative examples involving traditional optimization
will address both the development of the describing equation(s); that is, the model and
the solution to the problem. Linear programming illustrative examples will key solely on
the modeling aspect of the problem and not the solution, which more often than not will
be “obtained” from Excel. This statement particularly applies to some of the illustrative
examples at the end of each chapter.
8
Air Pollution Management

The first chapter in Part II contains six illustrative examples. Example titles are listed here:

8.1 Outdated air pollution control device

8.2 Coal-fired power plant options
8.3 Particulate control equipment options
8.4 Recovering dust
8.5 Incinerating mercury-contaminated waste
8.6 Optimizing incinerator performance

Seven select references for suggested reading for Chapter 8 are provided below:

1. L. Theodore, Chemical Engineering: The Essential Reference, McGraw-Hill, New York

City, NY, 2014.
2. J. Reynolds, J. Jeris, and L. Theodore, Handbook of Chemical and Environmental
Engineering Calculations, John Wiley & Sons, Hoboken, NJ, 2004.
3. L. Theodore, R. Dupont, and K. Ganesan, Unit Operations in Environmental
Engineering, Wiley‑Scrivener, Salem, MA, 2018.
4. M.K. Theodore and L. Theodore, Introduction to Environmental Management, CRC
Press/ Taylor & Francis Group, Boca Raton, FL, 2009.
5. G. Burke, B. Singh, and L. Theodore, Handbook of Environmental Management and
Technology, 2nd edn, John Wiley & Sons, Hoboken, NJ, 2000.
6. L. Stander and L. Theodore, Environmental Regulatory Calculations Handbook, John
Wiley & Sons, Hoboken, NJ, 2007.
7. L. Theodore, Air Pollution Control Equipment Calculations, John Wiley & Sons,
Hoboken, NJ, 2008.

8.1 Outdated Air Pollution Control Device

Air pollution control technology is self-defeating if it creates undesirable side effects in
meeting (limited) air pollution control objectives. Air pollution control should be consid-
ered in terms of both the total technological system and ecological consequences. The
former considers the technology that can be brought to bear on not only individual pieces
of equipment but also the entire technological system. Consideration of ecological side
effects must also be taken into account; for example, the problem of disposal of possibly
unmanageable accumulations of contaminants. These may be concentrated in the collec-
tion process, such as groundwater pollution resulting from landfill practices or pollution
of streams from the discharges of air pollution control systems.
111
112 Introduction to Optimization for Chemical and Environmental Engineers

Gaseous and particulate pollutants discharged into the atmosphere can be controlled.
The five generic devices available for particulate control include gravity settlers, cyclones
(centrifugal separators), electrostatic precipitators, wet scrubbers, and baghouses (fabric
filtration). The four generic devices for gases include condensers, absorbers, adsorbers, and
incinerators. Detailed information on these control devices are available in the literature.1
There are a number of factors to be considered prior to selecting a particular piece of air
pollution control hardware. In general, they can be grouped into three categories: environ-
mental, engineering, and economic. The latter category is discussed later since economics
are considered in the applications to follow. Key economic factors include1

1. Capital cost (equipment, installation, engineering, etc.)

2. Operating cost (utilities, maintenance, etc.)
3. Expected equipment lifetime and salvage value

Illustrative Example 8.1

The annual total operation cost of an outdated air pollution control device is $75,000.
Under a proposed air toxin emission reduction plan, the installation of a new process-
ing system will require an initial cost of $150,000, but it will reduce annual air toxin
pollution control costs to $15,000 in each of the first five years and $25,000 in the latter
years. Determine the annualized cost for the new processing system by assuming the
system only has five years operational life. The interest rate is 7%. Compare the costs for
both operations and select the optimum option.
Note: The capital recovery factor (CRF) or annual payment of a capital investment can
be calculated as follows:

i (1 + i )
n

CRF = (8.1)
(1 + i) n
−1
where:
i = interest, fractional value
n = equipment lifetime, years

SOLUTION:
The annualized cost for a new process is determined based on the following input data:

Capital cost = $150, 000

Interest , i = 7%

Term , n = 5 yr

With i = 7% and n = 5 yr, the CRF from Equation 8.1 is:

i (1 + i ) 0.07 (1 + 0.07 )
n 5

CRF = = = 0.2439
(1 + i)n − 1 (1 + 0.07 )5 − 1
The total annualized cost for the new process is then calculated by Equation 8.2:
Annualized cost = (CRF)(AIC) + AOC (8.2)
= Annualized installation + Annualized operation cost

Annualized cost = (0.2439)($150, 000 ) + $15, 000 = $36, 585 + $15, 000 = $51, 585
Air Pollution Management 113

Since this annual cost is lower than the original annual cost of $75,000 for the old pro-
cess, select the new process.

8.2 Coal-Fired Power Plant Options

An environmental standard of performance is a standard for emissions of air pollutants
that reflects the degree of emission limitation achievable through the application of, for
example, best available control technology (BACT) or the best system of emission reduc-
tion that (taking into account the cost of achieving such a reduction and any non-air-qual-
ity health and environmental impact plus energy requirements) the EPA (Environmental
Protection Agency) determines has been adequately demonstrated. Four source categories
(as they relate to coal) for which performance standards have been established are

1. Fossil fuel-fired steam generators

2. Electric utility steam-generating units
3. Industrial/commercial/institutional steam-generating units
4. Small industrial/commercial steam-generating units

Coal combustion products (CCPs) can play an important role in its usage. CCPs are the
by-products generated from burning coal in coal-fired power plants. These by-products
include2

1. Fly ash
2. Bottom ash
3. Boiler slag
4. Flue gas desulfurization gypsum
5. Other types of material, such as fluidized-bed combustion ash, cenospheres, and
scrubber residues

The key air pollution control option for coal-fired boilers today is the baghouse, details
of which will be provided in the next section (Section 8.3).

Illustrative Example 8.2

Consider the following application. A baghouse is needed at a small coal-fired power
plant for a design operating period of 20 years. If the unit fails at any time (bag
meltdown), a 45% (of the initial cost) reinvestment cost will result. Two companies
submit bids for this particulate control device with the following cost and operating
characteristics data (see Table 8.1):

TABLE 8.1
Data for Coal-Fired Power Plant Options
Time to Annual Salvage Value
Company Initial Cost Failure (yr) Operation Cost in Year 20
A $15,000 10 $2,300 0
B $22,000 45 $1,100 0
114 Introduction to Optimization for Chemical and Environmental Engineers

If each alternative has an annual probability of failure that is inversely proportional

to its time to failure, choose the optimum option for particulate control for this facility
assuming an annual interest rate of 4%.

SOLUTION:
The annual probability of failure of the baghouse provided by Company A is 1/10 = 0.10,
while that of Company B is 1/45 = 0.022. With these results, the estimated annual cost of
each of the baghouse options can be determined as follows:

Annual cost = Annualized capital cost + Reinvestment cost + Annual O/M cost (8.3)

The value of the CRF for 20 years at 4% is 0.0736 (see Section 8.1). For Company A, the
estimated annual cost is:

Annual cost = ($15, 000 )(0.0736 ) + $15, 000 (0.45)(0.10 ) + $2, 300

= $1,, 104 + $675 + $2, 300

= $4, 079 / yr
For Company B, the estimated annual cost is

Annual cost = ($22, 000 )(0.0736 ) + $22, 000 (0.45)(0.022) + $1, 100

= $1, 620 + $218 + $1, 100

= $2, 938 / yr
From this present worth evaluation, Company B’s bid for the baghouse installation
should be selected since the annual cost is minimized.

8.3 Particulate Control Equipment Options

Electrostatic precipitators (ESPs) are unique among gas cleaning equipment in that the
forces separating the particulates from the gas stream are applied directly to the particu-
lates themselves, and hence the energy required to affect the separation is considerably
less than for other types of gas cleaning apparatuses. Gas pressure drops through the pre-
cipitator may be of the order of 1 inch of water or less as compared with pressures of up to
100 inches of water for scrubbers and 10 inches of water for baghouses. The fundamental
advantage of electrostatic precipitation has resulted in its widespread use in applications
where large gas volumes are to be handled and high efficiencies are required for collection
of small particles.1
As the name scrubber implies, wet collectors or wet scrubbers are devices that use a liq-
uid for removing particles or polluted gases from an exhaust gas stream. Water sprays can
be injected into the gas stream; gas can be forced to pass through sheets or films of liquid;
or, the gas can move through beds of plastic spheres covered with liquid. Each of these
techniques can effectively remove particulate matter from process exhaust gases. They can
also effectively remove gases such as HCl or SO2, but removal conditions must be right. In
addition, gas-liquid contact can bring about gas conditioning, and to a lesser extent, liq-
uid conditioning. In effect, these air pollution control devices can remove both particulate
Air Pollution Management 115

matter and gases from effluent gas streams. The evaluation and selection of a scrubber
system is somewhat simplified by the observation that collection efficiency is a function
of the amount of energy required to operate the scrubber. This means, that independent
of design and generally speaking, the more power put into the system, the greater the col-
lection efficiency. Therefore, for systems of nearly equivalent power inputs, the collection
efficiencies should be nearly equivalent. The selection or evaluation procedure between
two systems could then concentrate on ease of operation, potential maintenance problems,
and comparative costs.1
One of the oldest, simplest, and most efficient methods for removing solid particulate
containments from gas streams is by filtration through fabric media. The fabric filter is
capable of providing high collection efficiencies for particles as small as 0.1 µm and will
remove a substantial quantity of those particles as small as 0.01 µm. In its simplest form,
the industrial fabric filter consists of a woven or felted fabric through which dust-laden
gases are forced. A combination of factors results in the collection of particles on the fab-
ric filters. When woven fabrics are used, a dust cake eventually forms; this, in turn, acts
predominantly as a sieving mechanism. When felted fabrics are used, this dust cake is
minimal or almost nonexistent, and the primary filtering mechanisms are a combination
of inertial forces, impingement, and so on, on the fabric. These are essentially the same
mechanisms that are applied to particle collection on wet scrubbers where the collection
media is in the form of liquid droplets rather than solid fibers.1
Proper selection of a particular system for a specific application can be extremely diffi-
cult and complicated. In view of the multitude of complex and often ambiguous pollution
control regulations, it is in the best interest of the prospective user (as noted above) to work
closely with regulatory officials as early as possible in the process.
The final choice in equipment selection is usually dictated by that piece of equipment
capable of achieving compliance with regulatory codes at the lowest uniform annual cost
(amortized capital investment plus operation and maintenance costs). More recently, there
have been attempts to include liability problems, neighbor/consumer goodwill, employee
concerns, and so on, in the economic analysis, but these effects – although important – are
often extremely difficult to quantify.1
In order to compare specific control equipment alternatives, knowledge of the particular
application and site is also essential. A preliminary screening, however, may be performed
by reviewing the advantages and disadvantages of each type of air pollution control equip-
ment. For example, if water or a waste treatment system is not available at the site, this may
preclude the use of a wet scrubber system and instead focus particulate removal on dry
systems, such as cyclones, baghouses, and/or electrostatic precipitators. If auxiliary fuel is
unavailable on a continuous basis, it may not be possible to combust the organic pollutant
vapors in an incineration system. If the particulate-size distribution in the gas stream is
relatively fine, gravity settlers and cyclone collectors most probably would not be consid-
ered. If the pollutant vapors can be reused in the process, control efforts may be directed to
adsorption systems. There are many other situations where knowledge of the capabilities
of the various control options, combined with common sense, will simplify the selection
process.1

Illustrative Example 8.3

A stream of 100,000 acfm of particulate-contaminated gas from a waste facility is to
be cleaned according to environmental regulations. You have been requested to install
the optimum pollution control device at the site. A reputable vendor – Behan Doyle
116 Introduction to Optimization for Chemical and Environmental Engineers

Environmental Inc. – has provided information on the cost of the equipment, as well as
installation, operating and maintenance costs, according to the conditions specified for
it to be treated. Given in Table 8.2 is all the collected data. Determine what equipment
you would select for this work in order to minimize costs on an annualized basis.
Costs are based on comparable overall collection efficiencies. The interest rate is 10%
and there is zero salvage value.

SOLUTION:
The first step is to convert the equipment, installation, and operating costs to total costs
by multiplying each by the total gas flow, 100,000 acfm. Hence, for the electrostatic pre-
cipitator, the total costs are

 $3.1 
Equipment cost = 100, 000 acfm  = $310, 000
 acfm 

 $0.80 
Installation cost = 100, 000 acfm  = $80, 000
 acfm 

 $0.06 
Operating cost = 100, 000 acfm   = $6, 000/yr
 acfm − yr 
Note that the operating costs are on an annualized basis. The equipment cost and the
installation cost must then be converted to an annual basis using Equation 8.1 with
n = 20 and i − 0.10.3

CRF =
(0.1)(1 + .10)20 = 0.1175
(1 + .10 )20 − 1
 
The annual costs for the equipment and the installation is the product of the CRF and
the total costs of each.

Equipment Annual Cost = $310, 000 (0.1175)

= $36, 412/yr

Installation Annual Cost = $80, 000 (0.1175)

= $9, 400/yr
The calculations for the scrubber and the baghouse are performed in the same man-
ner. The three devices can be compared after all the annual costs are added. The tabu-
lated results are provided in Table 8.3.

TABLE 8.2
Precipitator, Scrubber, and Baghouse Cost; Illustrative Example 8.3
Electrostatic Precipitator Venturi Scrubber Baghouse
Equipment cost $3.1/acfm $1.9/acfm $2.5/acfm
Installation cost $0.80/acfm $1.4/acfm $1.0/acfm
Operating cost $0.06/acfm-yr $0.06/acfm-yr $0.095/acfm-yr
Maintenance cost $14,000/yr $28,000/yr $9,500/yr
Lifetime of equipment 20 yr 15 yr 20 yr
Air Pollution Management 117

TABLE 8.3
Calculation for Air Pollution Equipment; Illustrative Example 8.3
Electrostatic Precipitator Venturi Scrubber Baghouse
Equipment cost $310,000 $190,000 $250,000
Installation cost $80,000 $140,000 $100,000
CRF 0.1175 0.1315 0.1175
Annual equipment $36,412 $24,985 $29,375
Annual installation $9,400 $18,410 $11,750
Annual operating $6,000 $6,000 $9,500
Annual maintenance $14,000 $28,000 $9,500
Total annual cost $65,812 $77,395 $60,125

According to the analysis, the baghouse is the most economically attractive, that is, the
optimum device, since the annual cost is the lowest.

8.4 Recovering Dust
From an economic point of view, the break-even point of a process operation is defined as
that condition when the costs (C) exactly balance the income (I). The Profit (P) is, therefore,

P = I − C (8.4)

At break-even, the profit is zero.

Illustrative Example 8.4

A process emits 50,000 acfm of gas containing a dust (it may be considered ash and/
or metal) at a loading of 2.0 gr/ft3. A particulate control device is employed for particle
capture and the dust captured from the unit is worth $0.03/lb of dust. Experimental
data have shown that the collection efficiency, E, is related to the system pressure drop,
∆P, by the formula:

∆P
E= (8.5)
∆P + 15.0
where:
E = fractional collection efficiency
∆P = pressure drop, lbf/ft2

If the overall fan is 55% efficient (overall) and electric power costs $0.18/kWh, at what
collection efficiency is the cost of power equal to the value of the recovered material?
What is the pressure drop in inches of water at this condition?

SOLUTION:
The value of the recovered material (RV) may be expressed in terms of the collection effi-
ciency E, the volumetric flow rate q, the inlet dust loading w, and the value of the dust (DV):

RV = ( q) ( w )( DV )(E) (8.6)
118 Introduction to Optimization for Chemical and Environmental Engineers

Substituting in Equation 8.6 yields:

 50, 000 ft 3   2.0 gr   1 lb   0.03$ 

RV =   3    (E) = 0.429E$/ min
 min   ft   7000 gr   lb 

The recovered value can be expressed in terms of pressure drop, that is, replace E by
Equation 8.5:

RV =
(0.429)( ∆P) $/min (8.7)
∆P + 15.0

The cost of power (CP) in terms of ∆P, q, the cost of electricity (CE) and the fan efficiency,
Ef, is

( )
CP = ( q) ( ∆P )( CE) / E f (8.8)

Substitution yields

 50, 000ft 3   ∆P lbf   0.18$   1 min ⋅ kW   1   1h 

CP =       = 0.006 ∆P$/min
 min   ft 2   kWh   44, 200 ft ⋅ lbf   0.55   60 min 

The pressure drop at which the cost of power is equal to the value of the recovered
material is found by equating RV with CP:

RV = CP (8.9)

Substituting Equations 8.7 and 8.8 for RV and CP, respectively ultimately leads to

∆P = 56.5 lbf /ft 2

= 10.9 H 2 O

Figure 8.1 shows the variation of RV, CP, and profit with pressure drop. The collection
efficiency corresponding to the previously calculated ∆P from Equation 8.5 is:

∆P (8.5)
E=
∆P + 15.0
56.5
=
56.5 + 15.0
= 0.79

= 79%

The reader should note that operating below this efficiency (or the corresponding pres-
sure drop) will produce a profit; operating above the value leads to a loss.
The operating condition for maximum profit can also be estimated from the graph.
Calculating this value is left as an exercise for the reader. [Hint: Set the first derivative
of the profit (i.e., RV-CP) with respect to ∆P equal to zero. The answer is 13.9 lbf/ft2.]
However, this application will be revisited later in the book.
The reader should realize that this problem was addressed earlier in the text. One is
left the option of revisiting Illustrative Example 5.1 and 6.2.
Air Pollution Management 119

0.5

0.4
CP

0.3
RV

Value, $/min
0.2
Profit
0.1
Break-even
0

–0.1

–0.2
0 10 20 30 40 50 60 70 80
Pressure drop, lbf /ft2

FIGURE 8.1
Profit as a function of pressure drop; illustrative example 8.4.

8.5 Incinerating Mercury-Contaminated Waste

Combustion and/or incineration is often used to control the emissions of organic compounds
from process industries. At a sufficiently high temperature and adequate residence time, any
hydrocarbon can be oxidized to carbon dioxide and water by the combustion process.
Combustion systems are often relatively simple devices capable of achieving very high
removal destruction efficiencies. They consist of burners, which ignite the fuel and organic
material, and a chamber, which provides appropriate residence (or detention/retention)
time for the oxidation process. Because of the high cost and decreasing supply of fuels,
combustion systems may be designed to include some type of heat recovery. Combustion
is also used for the more serious emission problems that require high destruction efficien-
cies, such as emission of toxic or hazardous gases. There are, however, some problems that
may occur when using combustion. Incomplete combustion of many organic compounds
results in the formation of aldehydes and organic acids as well as metals, such as mercury,
lead, calcium, and so on, which may create additional problems. Oxidizing organic com-
pounds containing sulfur or halogens produce unwanted pollutants such as sulfur diox-
ide, hydrochloric acid, hydrofluoric acid, or phosgene. If present, these pollutants would
require a scrubber to remove them prior to release into the atmosphere.
Several basic combustion systems are in use. Although these devices are physically simi-
lar, the conditions under which they operate may be different. Choosing the proper device
depends on many factors, including the type of hazardous contaminants in the waste
stream, concentration of combustibles in the steam, process flow rate, control require-
ments, and an economic evaluation.4
Combustion is a chemical process occurring from the rapid combination of oxygen with
various elements or chemical compounds resulting in the release of heat. The process of
combustion is also referred to as oxidation or incineration. Most fuels used for combustion
along with the waste are composed essentially of carbon and hydrogen, but can include
other elements, such as sulfur, nitrogen, and chlorine. Although combustion seems to be a
very simple process that is well understood, in reality it is not. The exact manner in which
120 Introduction to Optimization for Chemical and Environmental Engineers

a fuel or waste is oxidized occurs in a series of complex, free radical chain reactions. The
precise set of reactions by which combustion occurs is termed the mechanism of combus-
tion. By analyzing the mechanism of combustion, the rate at which the reaction proceeds
and the variables affecting the rate can be estimated. For most combustion devices, the rate
of reaction proceeds extremely fast compared to the mechanical operation of the device.
Maintaining efficient and complete combustion is somewhat of an art rather than a sci-
ence. Therefore, this section focuses on the factors that influence the completeness of com-
bustion rather than analyzing the mechanism involved.4,5
To achieve complete combustion once the air (oxygen), waste, and fuel have been brought
into contact, the following conditions must be provided: a temperature high enough to ignite
the waste/fuel mixture, turbulent mixing of the air and waste/fuel; and, sufficient residence
time for the reaction to occur. These three conditions are referred to as the “three T’s of com-
bustion.” Time, temperature, and turbulence govern the speed and completeness of reaction.
They are not independent variables since changing one can affect the other two.4
For combustion processes, ignition is accomplished by adding heat to speed up the oxi-
dation process. Heat is needed to combust any mixture of air and fuel until the ignition
temperature of the mixture is reached. By gradually heating a mixture of fuel and air, the
rate of reaction and energy released will gradually increase until the reaction no longer
depends on the outside heat source. In effect, more heat is being generated than is lost to
surroundings. The ignition temperature must be reached or exceeded to ensure complete
combustion. To maintain combustion of the waste, the amount of energy released by the
combusted waste must be sufficient to heat the incoming waste (and air) up to its ignition
temperature; otherwise, a fuel must be added. The ignition temperature of various fuels
and compounds can be found in the literature.4,5
The dangers of human exposure to many metallic chemical elements have long been
recognized. Metals such as mercury, lead, cadmium, and arsenic are toxic. Health effects
range from retardation and brain damage, especially in children from lead poisoning,
to the impairment of the central nervous system as a result of mercury exposure. Since
these toxic metals are chemical elements, they cannot be broken down by any chemical
or biological process. As a result of this, it is not uncommon to have metallic buildup or
bioaccumulation.6
Mercury and most of its compounds are toxic substances. Nineteenth-century hat mak-
ers, for example, developed a characteristic shaking and slurring of speech from occu-
pational exposure to large quantities of inorganic mercury during the manufacturing
process – symptoms that gave rise to the phrase “mad as a hatter.” Such problems result
from impairment of the central nervous system. In addition, high levels of mercury can
cause kidney damage, birth defects, and in extreme cases, death. The most common source
of the silvery liquid can be found in the bulbs of thermometers and thermostats. Large
quantities of mercury can also be found in one type of barometer or another. Small quanti-
ties are also found in fluorescent lights, mercury switches, batteries, hearing aids, smoke
detectors, watches, and cameras.6 Consider the following application.

Illustrative Example 8.5

Two hospital wastes are received and stored in separate tanks at an incineration facility.
The first, a sludge waste with a net heating value (NHV) of 6000 Btu/lb contains 2% Hg
by weight. The second, a mercury-contaminated plating waste with an NHV of 8000 Btu/
lb, contains 8% Hg by weight. A minimum of 1000 lb/h of each waste is to be incinerated.
Because of pump limitations, no more than 5000 lb/h of each waste can be utilized.
Air Pollution Management 121

The incinerator can operate with a heat rate between 25 and 40 million Btu/h. These
specifications provide two additional constraints on the operation of the incinerator.

3890
msludge = 3890 − mplate ; < 40 MMBtu/h (8.10)
5000

2431
msludge = 2431 − mplate ; > 25 MMBtu/h (8.11)
3125
Determine the maximum amount of mercury-contaminated plating waste that can be
incinerated employing a graphical approach involving the plotting of the two principal
variables, msludge and mplate (plating waste).

SOLUTION:
Let msludge and mplate represent the flowrates of the two wastes to the waste incinerator
in lb/h. The restrictions on the available blending options are provided in Table 8.4 and
shown graphically in Figure 8.2 using numbered lines for the constraints. The permissi-
ble area in Figure 8.2 is the quadrilateral area (ABCD) bounded by the lines 2, 4, 5, and 6.
The point farthest to the right (point A), represents the maximum rate at which plating
waste can be incinerated. This occurs where constraints 2 and 5 intersect. The describ-
ing equation for a minimum sludge rate of 1000 lb/h is then (employing Equation 8.12)

lb lb  3890 lb/h 
1000 = 3890 −  mplate
h h  5000 lb/h 
so that

mplate = 3715 lb/h

Can the reader determine the maximum rate at which the sludge can be incinerated?

8.6 Optimizing Incinerator Performance

The performance of an incinerator is strongly influenced by temperature T. The effect pres-
sure, P, is normally assumed to be negligible. However, some innovative and/or specially
designed incinerators can be adversely affected by pressure.

TABLE 8.4
Flowrate Restrictions Related to Illustrative Example 8.5
Line 1 Maximum of 5000 lb/h of sludge msludge < 5000
Line 2 Minimum of 1000 lb/h of sludge msludge > 1000
Line 3 Maximum of 5000 lb/h of plating waste mplate < 5000
Line 4 Minimum of 1000 lb/h of plating waste mplate > 1000
Line 5 Maximum heat rate of 40 MMBtu/h5 3890
msludge = 3890 − mplate
5000
Line 6 Minimum heat rate of 25 MMBtu/h5 2431
msludge = 2431 − mplate
3125
122 Introduction to Optimization for Chemical and Environmental Engineers

6000
Line 4 Line 3
5000 Line 1

msludge lb/h 4000

3000 B
Line 5

2000

C D
1000 Line 2
A
Line 6
0
0 1000 2000 3000 4000 5000 6000
mplate lb/h

FIGURE 8.2
Diagram showing six constraints for mercury incinerator application.

Illustrative Example 8.6

The emissions rate, E, of a toxic pollutant as a function of both temperature and pressure
from one such unit has been described by the following equation:

12, 000
E (lb/h ) = 0.23T + 0.19P + + 4.31; T = °C, P = psi (8.12)
PT
Concerned about the worst-case scenario of emissions, you have been hired to determine
the maximum emission rate from the previously mentioned incinerator. Earlier tests sug-
gest that the allowable operating range of the unit is in the 25°C–50°C temperature and
14.7–50.0 psi pressure ranges. Obtain the solution to this application employing a brute
force perturbation approaches (The reader may choose to revisit Illustrative Example 6.6).

SOLUTION:
Generate emission rates from Equation 8.12 for a number of combinations of pressure
and temperature. Some of the calculated results are provided in Table 8.5.

TABLE 8.5
Incineration Emission Rates as a Function of Temperature and
Pressure; Illustrative Example 8.6
P, psi T,°C
25.0 30.0 35.0 40.0 45.0 50.0
14.7 45.51 41.21 38.27 36.71 33.59 34.93
20.0 37.86 35.01 33.30 32.31 31.79 31.61
25.0 34.01 31.96 30.82 30.26 30.08 30.16
30.0 31.76 30.24 29.49 29.21 29.25 29.51
40.0 29.66 28.81 28.53 28.61 28.93 29.41
50.0 29.16 28.71 28.72 29.01 29.49 30.11
Air Pollution Management 123

The reader is left the exercise of refining (expanding) the table. The results suggest that
the minimum emission rate occurs with T ≈ 35–40°C and P ≈ 40 psi with E ≈ 28.5 lb/h.
The reader should also realize that the application could have been solved analyti-
cally by employing an approach by requiring partial derivatives.7,8

∂E 12, 000
= 0.23 − (8.13)
∂T T 2P

∂E 12, 000
= 0.19 − (8.14)
∂P TP 2
Setting both Equations 8.13 and 8.14 equal to zero and solving simultaneously leads to

T = 35.06°C

P = 42.44 psi

The corresponding value of E is

12, 000
E = 0.23 ( 35.06 ) + 0.19 ( 42.44 ) + + 4.31
( 42.44)( 35.06)
= 8.06 + 8.06 + 8.06 + 4.31

= 28.49 lb/h

Evaluating the second partial derivatives is necessary to determine if E is a maximum

or minimum.

∂ 2E
=+
(12, 000)( 2)
∂T 2 T 3P

∂ 2E
=+
(12, 000)( 2)
∂P 2
TP 3
Since both second partial derivatives are positive, the calculated value for the previously
mentioned E is indeed a minimum.

References
1. L. Theodore, Air Pollution Control Equipment Calculations, John Wiley & Sons, Hoboken, NJ,
2008.
2. K. Skipka and L. Theodore, Energy Resources: Availability, Management, and Environmental
Impacts, CRC Press/Taylor & Francis Group, Boca Raton, FL, 2016.
3. S. Shaefer and L. Theodore, Probability and Statistics for Environmental Science, CRC Press/Taylor
& Francis Group, Boca Raton, FL, 2007.
4. J. Santoleri, J. Reynolds, and L. Theodore, Introduction to Hazardous Waste Incineration, 2nd edn,
John Wiley & Sons, Hoboken, NJ, 2004.
5. L. Theodore, “Chemical Reactor Analysis and Application for the Practicing Engineer”, John
Wiley & Son, Hoboken, NJ, 2012.
124 Introduction to Optimization for Chemical and Environmental Engineers

6. L. Theodore and R. Dupont, “Environmental Health and Hazard Risk Assessment: Principles
and Calculations”, CRC Press/Taylor & Francis Group, Boca Raton, FL, 2012.
7. L. Theodore, personal notes, East Williston, NY, 2000.
8. L. Theodore and C. Prochaska, Introduction to Mathematical Methods for Environmental Engineers
and Scientists, Wiley‑Scrivener, Salem, MA, 2018.
9
Water Pollution Management

The second chapter in Part II contains six illustrative examples. Example titles are listed
here:

9.1 Break-even point operation

9.2 Wastewater treatment equipment options
9.3 Water disinfection options
9.4 Water plant operation
9.5 Generating waste revenues from waste streams
9.6 Membrane construction costs

Eight select references for Chapter 9 are provided here:

1. L. Theodore, Chemical Engineering: The Essential Reference, McGraw-Hill, New York

City, NY, 2014.
2. J. Reynolds, J. Jeris, and L. Theodore, Handbook of Chemical and Environmental
Engineering Calculations, John Wiley & Sons, Hoboken, NJ, 2004.
3. L. Theodore, R. Dupont, and K. Ganesan, Unit Operations in Environmental
Engineering, Wiley‑Scrivener, Salem, MA, 2018.
4. M. K. Theodore and L. Theodore, Introduction to Environmental Management, CRC
Press/Taylor & Francis, Boca Raton, FL.
5. G. Burke, B. Singh, and L. Theodore, Handbook of Environmental Management and
Technology, 2nd edn, John Wiley & Sons, Hoboken, NJ, 2000.
6. L. Stander and L. Theodore, Environmental Regulatory Calculations Handbook, John
Wiley & Sons, Hoboken, NJ, 2007.
7. R. Thomann and J. Muller, Principles of Surface Water Quality Modeling and Control,
Harper & Row, New York City, NY, 1987.
8. R. Dupont, K. Ganesan, and L. Theodore, Pollution Prevention, Sustainability,
Industrial Ecology, and Green Science and Engineering, CRC Press/Taylor & Francis,
Boca Raton, FL, 2017.

9.1 Break-Even Point Operation

From an economic point of view, the break-even point of a process operation is defined as
that condition when the costs (C) exactly balances the income (I). As noted in the previous
chapter, the profit (P) is, therefore, P = I − C. At break-even, the profit is zero.

125
126 Introduction to Optimization for Chemical and Environmental Engineers

Illustrative Example 9.1

The cost and income (in dollars) for a particular operation are given by the following
equations:

I = $60, 000 + 0.021N

C = $78, 000 + 0.008 N

where N is the yearly production of the item being manufactured.

Calculate the break-even point for this operation.

SOLUTION:
Write the equation relating C to I. Note that at break-even operation, P = 0.

I =C

Substitute for C and I in terms of N:

$60, 000 + 0.021N = $78, 000 + 0.008 N

Solving for N at the break-even point

N ≈ 1, 384, 600

Calculating the cost at the break-even point:

C = $78, 000 + 0.008 N

= $78, 000 + 0.008 (1, 384, 600 )

≈ $89, 077

The reader should note that as N decreases below 1,384,600 items, P is negative (there is
a cost). Higher values of N lead to profits.

9.2 Wastewater Treatment Equipment Options

Numerous technologies exist for treating industrial wastewater. These technologies range
from simple clarification in a settling pond to a complex system of advanced technologies
requiring sophisticated equipment and skilled operators. Finding the proper technology or
combination of technologies adequate to treat a particular wastewater to meet federal and
local requirements and yet still be cost-effective can be a challenging task.1
Treatment technologies can be divided into three broad categories: physical, chemical,
and biological. Many treatment processes combine two or all three categories to provide
the most economical treatment. There are a multitude of technologies for each of these
categories. Therefore, although the technologies selected for subsequent discussion are
among the most widespread in the environmental field, they represent only a fraction of
the available technologies.
Water Pollution Management 127

Physical treatment processes include clarification (sedimentation), floatation, and oil-

water separation. Chemical treatment processes include coagulation, precipitation, and
neutralization. Biological processes include actuated sludge, aerobic attached growth, aer-
obic lagoons (stabilization ponds), and anaerobic lagoons.1

Illustrative Example 9.2

Consider the following application. Three different water pollution control devices are
available for the removal of a toxic contaminant from a wastewater stream. The service
life is 10 years for each device. The capital and annual operating costs are as follows (see
Table 9.1):
Which is the most economical unit (that will minimize cost)? Employ a straight-line
depreciation method of analysis.

SOLUTION:
To select the most economical control device, a comparison can be performed among
the three units based on the total annualized cost (TAC). Table 9.2 can be used to sim-
plify these calculations:
A comparison among units A, B, and C indicates that unit C has the lowest TAC and
should be selected as the most economical unit of the three being evaluated. A similar result
would be obtained if a return on investment (ROI) method of analysis were employed.

9.3 Water Disinfection Options

The exploration of an individual water quality problem begins with an investigation of
the impact of bacteria and the organisms that may cause communicable diseases. This is
appropriate because:

1. The kinetics of bacterial die-away are usually considered to be first order.

2. It is the oldest of the water pollution problems in the discovery of the links between
contaminated water and communicable disease.

TABLE 9.1
Economic Cost Data for Waste Pollution Control Options
Device Initial Cost Annual Operating Cost Salvage Value in Year 10
A $300,000 $50,000 0
B $400,000 $35,000 0
C $450,000 $25,000 0

TABLE 9.2
Total Annual Cost Calculations
Equipment A B C
Capital investment $300,000 $400,000 $450,000
Depreciation $30,000 $40,000 $45,000
Operating costs $50,000 $35,000 $25,000
Total annual costs $80,000 $75,000 $70,000
128 Introduction to Optimization for Chemical and Environmental Engineers

3. The problem is still quite relevant today and is manifested in the continuing high
levels of waterborne diseases in some countries, closing of bathing beaches, and
restrictions on water consumption.

The transmission of waterborne diseases (e.g., gastroenteritis, amoebic dysentery, chol-

era, and typhoid fever) has been a matter of concern for many years. The impact of high
concentrations of disease-producing organisms on waste uses can be significant. Bathing
beaches may be closed permanently or intermittently during rainfall conditions when
high concentrations of pathogenic bacteria and discharge from urban runoff and com-
bined sewer overflows. Diseases associated with water used for drinking purposes con-
tinue to occur.
The modes of transmission of pathogens are (1) ingestion of contaminated water and
food, and (2) exposure to infected persons or animals. Infections of the skin, eyes, ears,
nose, and throat may result from immersion in the water while bathing. The water uses
impacted by pathogenic bacteria, viruses, and parasites are2

1. Drinking water: municipal, domestic, industrial, and individual supplies

2. Primary contact recreation, such as bathing and water skiing
3. Secondary contact recreation, such as boating and floating
4. Shellfishing, such as harvesting for clams and oysters

There are three broad categories for control of pathogenic bacteria, viruses, and para-
sites: control at the input source of the microorganism, control at the area of water use, and
control of the product that is affected by contamination. Point sources of municipal wastes
are the usual principal inputs of communicable disease organisms, and such inputs can
be reduced by treatment of wastes without disinfection, and disinfection by chlorination,
ultraviolet (UV) radiation, ozonation, and chlorine dioxide. Controls of the area of water
use would include bathing restrictions on a transient basis (i.e., during and after storms,
resulting in combined sewer overflows), construction of dikes, and diversion structures to
protect a given area. Controls of the product would include chlorination or other forms of
disinfection of water used for municipal water supply, treatment plants for contaminated
shellfish to allow depuration of bacteria prior to marketing, and distribution of high-qual-
ity bottled water during emergencies.1

Illustrative Example 9.3

Consider the following application. A large wastewater treatment facility currently uses
chlorine for disinfection. The average chlorine dosage is 6.0 mg/L at an average flow
rate of 70 MGD. Chlorine (Cl2) costs $1.00/lb. Sulfur dioxide (SO2) is used to dechlorinate
the effluent before it is discharged (a requirement placed on the facility to protect the
fish inhabiting the receiving water) and is consumed at an average dose of 2.0 mg/L. Its
cost is $1.20/lb.
Concerned about operating costs and the risks associated with Cl2 usage, the utility
has completed a pilot study showing that UV light disinfection could be used to replace
Cl2 and SO2. The UV system has a capital cost of $12,000,000 and will cost $500,000 a
year to operate with a 20-year service life. The utility will draw from bank savings
accounts currently earning 3.0% simple interest per year to pay for the UV system’s capi-
tal cost. Does the savings in operating costs justify the capital cost? Assume straight-line
depreciation of the UV system, with $0 salvage value at the end of its service life.
Water Pollution Management 129

ADDITIONAL INFORMATION:
Conversion factor: A dose of 1 mg/L = 8.34 lb/106 gal (MG)
Straight-line depreciation equation:

Initial Value − Salvage Value

Depreciation Rate in $ yr = (9.1)
Service Life
Rate of ROI equation in %yr:

Scenario A operating costs − Scenario B operating costs

ROI = (9.2)
Investmeent
where:
Scenario A = estimated cost for continuing the status quo
Scenario B = estimated cost for proposed alternatives

SOLUTION:
Calculate the daily Cl2 and SO2 dosage rates as follows:

Cl 2 dosage rate = ( 70 MGD )( 6.0 mg L ) ( 8.34 L ⋅ lb MG ⋅ mg )

= 3503 lb day

SO 2 dosage rate = (70 MGD )( 2.0 mg L ) ( 8.34 L ⋅ lb MG ⋅ mg )

= 1168l l b MG ⋅ mg
These daily rates are then converted to annual rates:

Annual Cl 2 dosage rate = ( daily Cl 2 rate )( day yr )

= ( 3503 lb day ) ( 365 day yr )

= 1, 278, 595 lb yr

Annual SO 2 dosage rate = ( daily SO 2 rate )( day yr )

= ( 1168 lb day ) ( 365 day yr )

= 426, 320 lb yr
The annual costs for Cl2 and SO2 addition are calculated as follows:

Annual Cl 2 cost = ( annual Cl 2 dosage rate)( Cl 2 cost in $ lb) (9.3)

= (1, 278, 595 lb yr )( $1.00 lb)

= $1, 278, 595 yr

Annual SO 2 cost = ( annual SO 2 dosage rate) ( SO 2 cost in $ lb) (9.4)

= ( 426, 320 lb yr )( $1.00 lb)

= $511, 584 yr
130 Introduction to Optimization for Chemical and Environmental Engineers

Total operating cost = Annual Cl 2 cost + Annual SO 2 cost (9.5)

= $1, 278, 595 yr + $511, 584 yr

= $1, 790, 179 yr

The straight-line depreciation of the system is then calculated as follows:

initial value − salvage value (9.6)

Depreciation =
service life

= ( $12, 000, 000 − $0 ) ( 20 yr )

= $600, 000 yr

The total annual cost of the UV system is then calculated as:

Total annual cost = Straight line depreciation + Operating cost (9.7)

= $6000, 000 yr + $500, 000 yr

= $1, 100, 000 yr

The annual cost savings for switching to the UV system is determined as follows:

Annual cost savings = Scenario A operating cost − Scenario B operating cost (9.8)

where scenario A is the total annual Cl2 cost and scenario B is the total annual UV
cost.

Annual cost savings = Annual Cl 2 cost − Annual UV cost

= $1, 790, 000 yr − $1, 100, 000 yr

= $690, 000 yr

The expected ROI on the investment in the UV system is determined as follows:

ROI = 100 ( Annual cost savings ) ( investment ) (9.2)

= 100 ( $690, 000 yr ) ( 2, 000, 000 )

= 100 ( 0.0575 )

= 5.8%

This is more than the 3.0% simple interest the utility could make from cash in its bank
account (at the time of preparation of this problem). Thus, based on operating cost con-
siderations alone, the utility should invest in the UV system. Other considerations may
make UV even more attractive. There may be hidden costs to continued Cl2 and SO2
usage, such as maintaining safety equipment and training, hazardous materials plan-
ning, worker and public concerns, stockholder support, and, perhaps, capital costs to
maintain or upgrade the Cl2 and SO2 storage and dosing systems. Additional calcula-
tions should be made that include these as well as potentially other hidden costs before
a final decision is made.
Water Pollution Management 131

9.4 Water Plant Operation

Illustrative Example 9.4
The Canadian Rocky Mountain Water Plant is located in the foothills of the Canadian
Rockies near Calgary. The plant “produces” bottled water and is the hub of their
operation. The bottled water is delivered to two local municipalities – Edmonton
(1) and Winnipeg (2). Locality (1) and (2) are approximately 200 and 650 miles from
the plant.
Henry Hydroxic was recently assigned the task of maximizing profits for the plant’s
operation. The following information is available. The plant can only produce a maxi-
mum of 17,300 bottles/day. Income derived from the sale of water to (1) and (2) is
27 cents and 38 cents per bottle respectably. However, Harry is confronted with two
constraints – one associated with energy costs and one associated with environmental
regulations. The energy constraint is based on company policy that energy operating
costs – 0.015 and 0.035 $/bottle for (1) and (2) – be limited to $1700/day. The environ-
mental constraints arise because the water contains 0.3 ppm dissolved SO2; local policy
at (1) and (2) require a “tax” of 1.1 cents and 1.9 cents/bottle, respectively. Company
policy limits this charge to $1000/day.
Unable to arrive at a solution to the company’s request, Harry has requested that you
provide a weekly delivery schedule that will maximize profits, P.

SOLUTION:
Set x1 and x2 equal to the daily delivery schedule of bottled water for Edmonton1 and
Winnipeg2, respectively. The profit objective function is then

P = 0.27 x1 + 0.38 x2 $ day (9.9)

The two constraints are

Energy: 0.015x1 + 0.035x2 ≤ 1700 $ day (9.10)

Environmental: 0.011x1 + 0.019x2 ≤ 1000 $ day (9.11)

In addition, based on physical grounds,

x1 ≥ 0 (9.12)

x2 ≥ 0 (9.13)

Employing the previously mentioned three equations, Excel provides the following
solution:

x1 = 27 , 000 bottles day

x2 = 37 , 000 bottles day

P = 21, 350 $ day

132 Introduction to Optimization for Chemical and Environmental Engineers

9.5 Generating Waste Revenues from Waste Streams

The waste from one industry is often an acceptable raw material for another. In response
to this common interest, material and waste exchanges were established. Material and waste
exchanges were organized to enable industrial process wastes, by-products, surpluses, or
materials that do not meet specifications to be transferred from one company to another
company where they can be used as a process input.3 Because many of these materials are
typically of low or negative value, it usually does not pay to transport them long distances.
As a result, these exchanges are generally regional ventures servicing a wide variety of
industries. Many of these exchanges have been operating in the United States since the
early 1970s. Initially, a major stumbling block to widespread adoption of this method of
waste recycling in this country has been the lack of an intermediary or an information
bank to help match supply and demand. Although more waste exchanges have been set
up in this country in recent years, they have not been fully accepted by industry because
of liability concerns, particularly for wastes that can be considered hazardous under the
Resource Conservation and Recovery Act (RCRA).4
Waste exchanges are usually nonprofit operations; however, some waste exchanges
charge for access to their database services or for copies of their catalogs. A partial list of
waste exchanges is available in the literature.5

Illustrative Example 9.5

As noted previously, a waste stream may be one company’s problem, but another com-
pany’s source of revenue. A chemical company has found an environmental waste
exchange that is prepared to pay $50/ton for Waste A that contains 40% chromium (Cr),
15% nickel (Ni), and 45% water; $20/ton for Waste B that contains 10% chromium, 80%
nickel, and 10% water; and, $35/ton for Waste C that contains 5% chromium, 20% nickel,
and 75% water. How many tons of waste A, B, and C should be purchased that will min-
imize the cost ($) of the waste and yet ensure that the mixture contains at least 20 tons
of chromium, 12 tons of nickel, and 15 tons of water? Develop the describing equations
for this operation.

SOLUTION:
Set x1, x2, and x3 as the tons of waste A, B, and C, respectively. The objective function for
the cost is given by

$ = 50 x1 + 20 x2 + 35x3 (9.14)

The constraints are:

0.40 x1 + 0.15x2 + 0.45x3 ≥ 20 (9.15)

0.10 x1 + 0.80 x2 + 0.10 x3 ≥ 12 (9.16)

0.05x1 + 0.20 x2 + 0.75x3 ≥ 15 (9.17)

and

x1 ≥ 0 (9.18)
Water Pollution Management 133

x2 ≥ 0 (9.19)

x3 ≥ 0 (9.20)

To minimize the waste cost, 29.0 tons of Waste A, 9.4 tons of Waste B and 15.6 tons of
Waste C should be purchased. This will produce a minimum cost of $2,181. The calcula-
tions are left as an exercise for the reader. Hint: Employ Excel.

9.6 Membrane Construction Costs

Membrane processes today are state-of-the-art separation technologies that show continued
promise for technical growth and wide-scale commercialization. They are used in many
industries for process stream and product concentration, purification, and fractionation. The
need for membrane research and development (R&D) is important because of the increasing
use of membrane technology in traditional and emerging engineering fields. Membrane pro-
cesses are increasingly finding their way into the growing engineering areas of biotechnology,
green engineering, specialty chemical manufacturing, biomedical engineering, as well as the
traditional chemical process industry. Membrane technology is also being looked at as either
a replacement for or supplement to traditional separations, such as distillation or extraction.
Membrane processes are generally more efficient and effective since they can simultaneously
concentrate and purify, and they can perform separations at ambient conditions.6
There are four major membrane processes of interest to the practicing engineer:

1. Reverse osmosis (hyperfiltration)

2. Ultrafiltration
3. Microfiltration
4. Gas permeation

The four processes have their differences. The main difference between reverse osmosis
(RO) and ultrafiltration (UF) is that the size/diameter of the particles or molecules in solu-
tion to be separated/ concentrated are generally solids or colloids rather than molecules
in solution. In microfiltration (MF), the particles to be separated/concentrated are gener-
ally solids or colloids rather than molecules in solution. Gas permeation (GP) is another
membrane process that employs a nonporous semipermeable membrane to “fractionate”
a gaseous stream.6
The heart of the membrane process is the membrane itself. A membrane is an ultra-thin
semipermeable barrier separating two fluids that permits the transport of certain species
through the barrier from one fluid to the other. The membrane is typically made from
various polymers, such as cellulose acetate or polysulfone, but ceramic and metallic mem-
branes are also used in some applications. The membrane is selective since it permits the
transport of certain species while rejecting others. The term semipermeable is frequently
used to describe the selective action.6
134 Introduction to Optimization for Chemical and Environmental Engineers

Illustrative Example 9.6

Consider the following application. Theodore International manufactures four different
membranes (A, B, C, and D) for water purification systems. Four input variables affect
the production process labor requirements: individual days, raw membrane material, lb,
processed membrane material, lb; and labor costs. Each membrane has different input
requirements. Current information on the production process for the four membranes is

TABLE 9.3
Membrane Data; Illustrative Example 9.6
Membrane
A B C D Maximum Input
Individual days 3 3 1 2 40
Lbs of raw membrane 5 5 7 6 80
Lbs of preferred membrane 3 3 6 10 65
Labor costs 2 5 4 3 50

provided in Table 9.3. If the A, B, C, and D membranes produced sell for $5, $5, $8, and
$20, respectively, determine the maximum production revenue subject to the constraints
listed in Table 9.3.

SOLUTION:
Set x1, x2, x3, and x4 as the production numbers for membranes A, B, C, and D. The objec-
tive function describing the production membrane number for maximizing the total
revenue (R) is presented here:

R = f ( x1 , x2 , x3 , x 4 ) = 5x1 + 5x2 + 8 x3 + 20 x 4 (9.21)

The constraints are

3 x1 + 3 x2 + x3 + 2x 4 ≤ 40 (9.22)

5x1 + 5x2 + 7 x3 + 6 x 4 ≤ 175 (9.23)

3 x1 + 3 x2 + 6 x3 + 10 x 4 ≤ 65 (9.24)

2x1 + 5x2 + 4 x3 + 3 x 4 ≤ 50 (9.25)

with

x1 ≥ 0 (9.26)

x2 ≥ 0 (9.27)

x3 ≥ 0 (9.28)

x 4 ≥ 0 (9.29)

The solution to this problem, subject to the information provided is from the previous
equation,
Water Pollution Management 135

x1 = 6.71

x2 = 3.71

x3 = 1.47

x 4 = 1.82

P = $100

The calculations are left as an exercise for the reader. Hint: employ Excel.

References
1. G. Burke, B. Singh, and L. Theodore, Handbook of Environmental Management and Technology,
2nd edn, John Wiley & Sons, Hoboken, NJ, 2000.
2. M. K. Theodore and L. Theodore, Introduction to Environmental Management, CRC Press/Taylor
& Francis Group, Boca Raton, FL.
3. L. Theodore and F. Ricci, Mass Transfer Operations for the Practicing Engineer, John Wiley & Sons,
Hoboken, NJ, 2010.
4. R. Dupont, K. Ganesan, and L. Theodore, Pollution Prevention: The Waste Management Approach
for the 21st Century, CRC Press/Taylor & Francis Group, Boca Raton, FL, 2000.
5. R. Dupont, K. Ganesan, and L. Theodore, Pollution Prevention, Sustainability, Industrial
Engineering, CRC Press/Taylor & Francis Group, Boca Raton, FL, 2017.
6. S. Slater, Membrane Technology, NSF Workshop Notes, Manhattan College, Bronx, NY, 1991
(adapted with permission).
10
Solid Waste Management

The third chapter in Part II contains six illustrative examples. Example titles are listed
below:

10.1 Solid waste treatment

10.2 Cement kiln versus rotary kiln incinerator
10.3 Sludge waste receiving tank size
10.4 Discounted cash flow application
10.5 Maximum mercury waste emission constraint
10.6 Structural considerations of concrete mix

Eight select references for Chapter 10 are provided here:

1. L. Theodore, Chemical Engineering: The Essential Reference, McGraw-Hill, New York

City, NY, 2014.
2. J. Reynolds, J. Jeris, and L. Theodore, Handbook of Chemical and Environmental
Engineering Calculations, John Wiley & Sons, Hoboken, NJ, 2004.
3. L. Theodore, R. Dupont, and K. Ganesan, Unit Operations in Environmental
Engineering, Wiley‑Scrivener, Salem, MA, 2018.
4. M. K. Theodore and L. Theodore, Introduction to Environmental Management, CRC
Press/Taylor & Francis Group, Boca Raton, FL, 2008.
5. G. Burke, B. Singh, and L. Theodore, Handbook of Environmental Management and
Technology, 2nd edn, John Wiley & Sons, Hoboken, NJ, 2000.
6. L. Stander and L. Theodore, Environmental Regulatory Calculations Handbook, John
Wiley & Sons, Hoboken, NJ, 2007.
7. J. Santoleri, J. Reynolds, and L. Theodore, Introduction to Hazardous Waste
Incineration, 2nd edn, John Wiley & Sons, Hoboken, NJ, 2000.
8. R. Dupont, K. Ganesan, and L. Theodore, Pollution Prevention, Sustainability,
Industrial Ecology, and Green Science and Engineering, CRC Press/Taylor & Francis
Group, Boca Raton, FL, 2017.

10.1 Solid Waste Treatment

Technical considerations involving solid waste treatment include waste characteristics,
technical sustainability of the available treatments, and treatment objectives. The waste
must be analyzed so that the proper form of treatment may be selected. Should the waste

137
138 Introduction to Optimization for Chemical and Environmental Engineers

contain more than one type (e.g., a mixture of heavy metals and organics), more than one
method of treatment may be needed.
Waste treatment may be divided into five categories: physical, chemical, biological, ther-
mal (incineration), and solidification or encapsulation (a subcategory of physical treatment).
Physical treatments may be used alone or with the other types of treatments; their function
is to concentrate wastes, reduce waste volume, and separate different waste components for
continued treatment or disposal. Chemical treatments are used to convert hazardous wastes
into other less hazardous forms; they are generally used on one substance or similar sub-
stances because the reactions involved are specific. Biological treatments are more specific
than the other treatment types and are used mainly on organic wastes. Thermal treatment is
usually concerned with incineration (see later paragraph). Solidification and encapsulation are
also treatments for specific types of waste; these methods work well on inorganic wastes, but
physical and/or chemical pretreatment is usually necessary. Incineration is one of the most
popular forms of thermal waste treatment, especially when organic wastes are involved. It
destroys most organic components in the waste, and the resulting sludge and ash are almost
always less hazardous and easier to dispose of than the original waste. It is difficult to obtain
a permit for incineration, however, and many stringent regulations govern its use.1,2
Processes to perform various waste treatment tasks include:

1. Separation of heavy metals from liquid waste streams

2. Destruction of organics
3. Separation of toxic anions from any liquid waste stream
4. Processes that can handle slurries or sludges
5. Processes the can handle tars or solids

Not all of these processes are currently feasible for hazardous wastes.
The objective of the treatment is another important factor that influences treatment selec-
tion. It may be desired, for example, to destroy the hazardous components in the waste,
reduce the hazardous components, and then isolate them, or merely to separate the haz-
ardous components. The end product of the waste treatment must be compatible with the
type of ultimate disposal to be used for the waste: sewage discharge, landfilling, deep-well
injection, and so on. For example, if the waste is to be disposed of by landfarming, the end
product of the waste treatment should consist of biodegradable organic components.
Economic feasibility and environmental factors are obviously also important factors to
be considered in treatment selection. Cost is a strong function of the amount of waste to
be treated and the simplicity and degree of commercial usage of the treatment process.
The various costs include capital investment, operating costs, utility costs, and expenses
associated with the final disposal of the waste product. Environmental factors, in the form
of state and local environmental regulations, also influence waste treatment selection.
The Environmental Protection Agency (EPA) Resource Conservation and Recovery Act
(RCRA) regulations especially control the type of waste treatment that may be used and
the characteristics of the waste that is to finally be disposed.1

Illustrative Example 10.1

A copper solid waste treatment facility can produce three grades of copper – pure grade
(99.99%), medium grade (99.95%), and poor grade (99.90%). Because of industry’s cop-
per demand, the production of the three classes of copper is limited. The profit after
Solid Waste Management 139

producing the three copper wastes, A, B, and C, is 2.20, 1.94, and 3.10, respectively,
$/lb. Copper Waste A generates 43%, 14%, and 43% of pure, medium, and poor grade
copper, respectively. Copper Waste B generates 35%, 35%, and 30% of pure, medium,
and poor grade copper, respectively. Copper Waste C generates 75%, 7%, and 18% of
pure, medium, and poor grade copper, respectively. However, the maximum produc-
tion of pure, medium, and poor grade copper is limited to 340,000, 66,000, and 145,000
lb/mo. Calculate the daily processing rate of the three wastes to maximize the profits.

SOLUTION:
Set x1, x2, and x3, as the daily processing rate of wastes A, B, and C, respectively. The
objective function for the income I is then

I = 2.20 x A + 1.94 xB + 3.10 xC (10.1)

There are three constraints:

0.43 x1 + 0.35x2 + 0.75x3 ≤ 340, 000 (10.2)

0.14 x1 + 0.35x2 + 0.07 x3 ≤ 66, 000 (10.3)

0.43 x1 + 0.30 x2 + 0.18 x3 ≤ 145, 000 (10.4)

with

x1 ≥ 0 (10.5)

x2 ≥ 0 (10.6)

x3 ≥ 0 (10.7)

Excel provides the following solution:

x1 = 154, 838

x2 = 59, 255

x3 = 336, 907

I = $1, 263, 052

10.2 Cement Kiln versus Rotary Kiln Incinerator

In the field of waste incineration, there is more experience with liquid injection (LI) incin-
erators than all other types combined. However, various other types of incinerators are
used to handle solid wastes as well as wastes in other forms, that is, liquid sludges, slur-
ries, and fumes. The subsequent discussion is confined to the LI type of incinerator, which,
as the name implies, is applicable almost exclusively to pumpable liquid waste. The waste
is burned directly in a burner (combustor) or injected into the flame zone or combustion
zone of the incinerator chamber (furnace) via nozzles. The heating value of the waste is the
primary determining factor for nozzle location.
140 Introduction to Optimization for Chemical and Environmental Engineers

Waste liquids in the process industries are numerous in kind and therefore defy easy
definition. Disposal of these waste liquids has become a serious problem for plant opera-
tors. Regulations today cover wastes once considered wastewater streams, as well as the
organic streams known to be made up of hazardous compounds. LI incinerators are avail-
able today to handle the various liquid streams that are being generated by the process
industries. These units may be on-site (at the location of the generator) or off-site (at a sister
facility or at a commercial disposal operation).
LI incinerators are usually refractory-lined chambers (horizontal, vertical with orienta-
tion up or down), generally cylindrical, and equipped with a primary combustor (waste
and auxiliary fuel-fired) and often secondary combustors, or injection nozzle for low-cal-
orific-value materials (aqueous wastes containing either organic or inorganic compounds,
or both). These units operate at temperature levels from 950°C (1742°F) to 1700°C (3092°F).
Residence time in the chamber may vary from milliseconds to 2.5 s. The viscosity deter-
mines whether the material being incinerated is considered a liquid, slurry, or sludge.
Liquid units today are capable of burning high viscosity materials of 4500 Saybolt seconds
universal (SSU) or less.1,2 Critical to the operation of the unit would be the atomizing noz-
zle used to convert the liquid stream into finely atomized droplets.2 When considering the
LI incinerator, one must be concerned about the ability of the system, that is, the design,
components, controls, and air pollution control system, to meet the requirements of the
federal and state regulations.
The physical, chemical, and thermodynamic properties of the waste must be consid-
ered in the basic design requirements of the total incinerator system. These include stor-
age tanks, mixers, pumps, control valves, piping, atomizers, combustors, refractory, heat
recovery, quench system(s), and the air pollution equipment. The types of data needed
by the designer to properly engineer the total system and the individual components are
available in the literature.1
Rotary kiln (RK) process incinerators were originally designed for lime processing.
Recently the RK has been utilized to process hazardous waste because it has the unique
capability to stabilize the combustion process for many different materials that may be
fed simultaneously. These kilns have demonstrated the ability to handle a wide variety of
materials with minimal impact on their performance. The result is minimal preprocessing
of the wastes and is one of the main reasons for the RK popularity.
The processing of raw materials has been a historic application of the RK. The basic
building materials of a community – cement, lime, coal, iron ore, aggregates, and other
bulk materials – require heating to high temperatures. These materials historically have
been produced in an RK. This high-temperature treatment and the continual mixing and
blending of the raw materials for cement are constantly exposed to the heat of the kiln.
Lime, lightweight aggregate, cement, and other process kilns may operate at temperatures
up to 1925°C (1350°F). Hazardous waste incinerators operate in the range of 700−1315°C
(1300−2400°F) depending on the waste types and shapes fed.
The RK is a cylindrical refractory-lined shell that is mounted at a slight incline from the
horizontal plane to facilitate mixing the waste materials and exposing the surface to the
auxiliary burner flame plus the waste fuel flames and flames generated over the burning
surface of solids. The kiln accepts all types of solid waste materials with heating values
between 555 and 8333 kcal/kg (1000 and 15,000 Btu/lb), and even higher. Solid wastes and
drummed wastes are usually fed by a pack-and-drum feed system, which may consist of
a bucket elevator for loose solids and a conveyor system for drummed wastes. Pumpable
sludges and slurries are injected into the kiln through nozzles. Temperatures for burning
vary from 700°C to 1315°C (1300°F to 2400°F) depending upon the waste types and shape.
Solid Waste Management 141

RKs are classed as slagging [>1000°C (1832°F)] or non-slagging [<1000°C (1832°F)]. The
non-slagging rotary kiln for hazardous waste destruction is not a total incineration sys-
tem. The kiln does not attain high destruction and removal efficiencies (DRE) required
of an incinerator due to the basic manner that it operates. It will volatilize the organic
content of the materials placed into it. The volatilized organics are carried in the com-
bustion products exit gas stream. A secondary combustion chamber (SCC) is located
immediately downstream of the kiln to perform the final combustion (destruction) pro-
cess on the kiln exit combustion products. The non-volatile part of the material fed into
the kiln, such as metals, stones, sand, and other solid inorganics exit as ash. The ash and
residue exiting the kiln should contain minimum to zero organics and may be treated
and disposed of in a hazardous waste landfill. Some metals and other elements, due to
the operating temperature will vaporize within the kiln and be carried in the exhaust
gas stream.
The RK combines three very important aspects of waste material processing into one
piece of equipment: mixing, temperature, and time. The mixing process is well known
in the production of cement where lime, iron, and sand or clay are mixed while being
heated. Good mixing is required to obtain a quality product. Therefore, the processing
equipment must allow for the wide variations in the material while achieving a uniform
bottom ash that meets regulatory requirements. The waste material that is placed into
the kiln must have sufficient time for its organic components to volatilize. With proper
time, the heat source within the kiln can raise the waste material to a desired tempera-
ture. The RK provides excellent mixing through the tumbling action to distribute the
heat evenly to all of the material being processed and to expose the material to the heat
source. Interestingly, the original kilns used for waste incineration were called “tumble
burners.”

Illustrative Example 10.2

Consider the following application. Plans are underway to construct and operate a com-
mercial solid hazardous waste facility in Trashtown in the state of Anatnom. The com-
pany is still undecided as for whether to install a cement or RK incinerator at the waste
site. The cement unit is less expensive to purchase and operate than a comparable RK
system. However, projected waste treatment income from the RK unit is higher since
it will handle a larger variety and quantity of solid wastes. In addition, the RK will be
treating solids as well as liquid wastes.
Based on the economic and financial data provided in Table 10.1, select the incinerator
that will yield the higher annual profit. Calculations should be based on an interest rate
of 12% and a process lifetime of 12 years for both incinerators.

SOLUTION:
For both units:

CRF = (0.12) (1 + 0.12) (1 + 0.12)12 − 1

12

 
= 0.1614
Annual capital and installation costs for the cement kiln (CK) are

CK costs = ( 2, 625, 000 + 1, 575, 000 )( 0.1614 )

= $677 , 780/yr
142 Introduction to Optimization for Chemical and Environmental Engineers

TABLE 10.1
Economic and Financial Data Based on Illustrative Example 10.2
Costs/Credits Cement Kiln Rotary Kiln
Capital ($) 2,625,000 2,975,000
Installation ($) 1,575,000 1,700,000
Operation ($/yr) 400,000 550,000
Maintenance ($/yr) 650,000 775,000
Income ($/yr) 2,000,000 2,500,000

Annual capital and installation costs for the RK unit are

RK costs = ( 2, 975, 000 + 1, 700, 000 )( 0.1614 )

= $754, 545/yr
A comparison of costs and credits for both incinerators is given in Table 10.2.
An RK incinerator is recommended.

10.3 Sludge Waste Receiving Tank Size

The normal distribution is used to obtain probabilities concerning the mean X of a sample
of n observations on a random variable X. If X is normally distributed with mean µ and
standard deviation σ, then X, the sample mean, is normally distributed with mean µ and
standard deviation σ n. For example, suppose X is normally distributed with mean 100
and standard deviation 2. Then X, the mean of a sample of 16 observations on X, is nor-
mally distributed with mean 100 and standard deviation 0.5. To calculate the probability
that X is greater than 101, one would write3,4

 X − 100 101 − 100 

P ( X > 101) = P  > (10.8)
 0.5 0.5 

 X − 100 101 − 100  X − 100

P ( X > 101) = P  >  ; Z=
 0.5 0.5  0.5
(10.9)
P ( X > 101) = P(Z > 2)

TABLE 10.2
Incinerator Cost Comparisons for Illustrative Example 10.2
Cement Kiln Rotary Kiln
Total installed ($/yr) 678,000 755,000
Operation ($/yr) 400,000 550,000
Maintenance ($/yr) 650,000 775,000
Total annual cost ($/yr) 1,728,000 2,080,000
Income credit ($/yr) 2,000,000 2,500,000
Profit ($/yr) 272,000 420,000
Solid Waste Management 143

From a standard normal distribution table,3,4

P = 0.023

However, if X is not normally distributed, then X, the mean of a sample of n observations

of X, is still approximately normally distributed with mean µ and standard deviation σ n,
provided the sample size n is large (>30). This result is based on an important theorem in
probability called central limit theorem.3,4

Illustrative Example 10.3

Consider the following application. Tests indicate that a sludge waste arriving in
55-gal drums to a treatment facility has a mean lead content of 11 ppm with a stan-
dard deviation of 10 ppm. The drums are unloaded into a 250-gal receiving tank.
The facility is required to keep the lead concentration entering the incinerator at
or below 15 ppm in order to meet the required regulatory particulate emission
levels. Assume that the lead contents from one drum to the next are not correlated
and that the tank is nearly full. What size should the receiving tank be to ensure,
with 98% confidence, that the facility treats a waste with a mean lead concentration
below 15 ppm?

SOLUTION:
For this condition, the probability that the lead (Pb) concentration in the receiving tanks
exceeds 15 ppm is

1.0 − 0.98 = 0.02 or 2.0%

The value of z0 from the standard normal table3 is then 2.05, which corresponds to the
number of standard deviations above the mean tank concentration. According to the
aforementioned central limit theorem, the standard deviation of the mean lead concen-
tration is given by

σ 10 ppm
= (10.10)
n n
where n is the number of drums. Therefore, for a 2% probability that the mean concen-
tration in the tank exceeds 15 ppm, the number of drums can be found by solving

15 − 11
2.05 =
 10 
 
 n
or

n = 26.3

The tank volume V must then be (approximately)

V = 26.3 drums ( 55 gal/drum )

= 1446 gal
144 Introduction to Optimization for Chemical and Environmental Engineers

10.4 Discounted Cash Flow Application

Illustrative Example 10.4
Two small commercial solid waste facility designs are under consideration. The first
design involves fixation and the second involves encapsulation. (In fixation, a solidify-
ing agent is used to solidify the waste. Encapsulation is a process in which the waste is
surrounded by a bundle; it has often been solidified by a chemical agent.) For the fixa-
tion system, the total capital cost (TCC) is $2.5 million, the annual operating costs (AOC)
are $1.2 million, and the annual revenue generated from the facility (R) is $3.6 million.
For the encapsulation system, TOC, AOC, and R are $3.5, 1.4, and 5.3 million, respec-
tively. Using straight-line depreciation and the discounted cash flow method, which
design is more attractive? Assume a 10-year facility lifetime and a two-year construc-
tion period. Note that the solution involves the calculation of the rate of return for each
of the two proposals.2,5
An after-tax rate of return on an initial investment of at least 30% is usually desirable.
The method used to arrive at a rate of return is discussed below. An annual after-tax
cash flow can be computed as the annual revenues (R) less the annual operating costs
(AOC) and less total income taxes (IT). Total IT can be estimated at 50% (this number
could be significantly lower subject to the passage of the new tax laws at the time of the
preparation of this manuscript) of taxable income (TI).

IT = 0.5 ( TI ) (10.11)

The TI is obtained by subtracting the AOC and the depreciation of the plant (D) from
the revenues generated (R) or

TI = R − AOC − D (10.12)

If straight-line depreciation is assumed, the plant will depreciate uniformly over the
life of the plant. Thus, for a 10-year lifetime, the facility will depreciate 10% each year.

D = 0.1( TCC ) (10.13)

The annual after-cash flow (A) is then

A = R − AOC − IT (10.14)

This procedure involves a trial-and-error solution. There are both positive and negative
cash flows. The positive cash flows consist of A and the recoverable working capital in
year 10. Both should be discounted backward to time = 0, the year the facility begins
operation. The negative cash flows consist of the TCC and the initial working capital
(WC). In actuality, the TCC is assumed to be spent evenly over the two-year construc-
tion period. Therefore, one-half of this flow is adjusted forward from after the first con-
struction year, (time = −1 yr) to the year the facility begins operating (time = 0). The other
half, plus the WC, is assumed to be expended at time = 0. Forward adjustment of the 50%
TCC is accomplished by multiplying by an economic parameter known as the single-
payment compound amount factor F/P, given by

F
= (1 + i ) (10.15)
m

P
Solid Waste Management 145

where:
i = rate of return (fraction)
n = the number of years (in this case, 1 yr)

For the positive cash flows, the annual after-tax cash flow (A) is discounted backward
by using a parameter known as capital recovering factor (CRF) or the uniform series
present worth factor (P/A). This factor is dependent on both interest rate (rate of return)
and the lifetime of the facility and is defined by

(1 + i ) − 1 (10.16)
n

P/A =
i (1 + i )
n

where n is the lifetime of the facility (in this case, 10 years).

The recoverable working capital at year 10 is discounted backward by multiplying
WC by the single payment present worth factor (P/F)

1
P/F = (10.17)
(1 + i )
n

where n is the lifetime of the facility (in this case, 10 years).

The positive and negative cash flows are now equated and the value of i, the rate of
return, may be determined by trial-and-error from Equation 10.18.

Term 1 = Term 2 − Term 3 + Term 4 (10.18)

where:

 (1 + i )10 − 1 
Term 1 =   A; worth at year 0 of annual after − tax cash flows (10.19)
 i (1 + i ) 
10

 1 
Term 2 =  
10 WC ; worth at year 0 of recoverable WC after 10 year (10.20)
 (1 + i ) 

 1 
Term 3 =  WC + TCC  ; assumed exoenditures at year = 0 (10.21)
 2 

1
( TCC ) (1 + i ) ; worth at year = 0 of assumed expenditures at year = −1 (10.22)
1
Term 4 =
2

SOLUTION:
For the fixation system, calculate, D, WC, TI, IT, and A. The depreciation is

D = 0.1( TCC )

= ( 0.1)( $2, 500, 000 ) (10.23)

= $250, 000
The WC is set at 10% of the TCC.
146 Introduction to Optimization for Chemical and Environmental Engineers

WC = 0.1( TCC )

= ( 0.1)( $2, 500, 000 ) (10.24)

= $250, 000
In addition,

TI = R − AOC − D

= $3, 600, 000 − $1, 200, 000 − $250, 000 (10.25)

= $2, 150, 000

and

IT = ( 0.5 ) TI

= ( 0.5 )( $2, 150, 000 ) (10.26)

= $1, 075, 000

The after-tax cash flow is calculated using

A = R − AOC − IT
= $3, 600, 000 − $1, 200, 000 − $1, 075, 000 (10.27)

= $1, 325, 000

The rate of return, i, for the fixation unit is also calculated. The rate of return can be
computed by solving the following equation:

 (1 + i )10 − 1   1 
10 WC = WC + ( 0.5 ) TCC + ( 0.5 ) TCC ( 1 + i ) (10.28)
1
 A+ 
 I (1 + i )   (1 + i ) 
10

or (on substituting)

 (1 + i )10 − 1   1 

 I (1 + i ) 
10 ( )
 1.325 × 10 6 + 
 (1 + i ) 
10 ( )
 0.250 × 10 6

( ) ( 
) ( (
= 0.250 × 10 6 + ( 0.5 ) 1.250 × 10 6 +  ( 0.5 ) 1.250 × 10 6 )) (1 + i ) 
1

By trial-and-error,

i = 39.6%

For the encapsulation system,

WC = D (10.29)

= (0.1)($3, 500, 000 )

= $350, 000
Solid Waste Management 147

TI = $5, 300, 000 − $1, 400, 000 − $350, 000

= $3, 550, 000

IT = (0.5) ($3, 550,, 000 )

= $1, 775, 000

The annual after-tax cash flow is

A = $5, 300, 000 − $1, 400, 000 − $1, 775, 000

= $2, 125, 000
The rate of return equation for the encapsulation unit becomes

 (1 + i )10 − 1   1 

 I (1 + i ) 
10 ( )
 2.125 × 10 6 +  (
 (1 + i ) 
10
 0.3650 × 10 6 )

( ( ) )
= 0.350 × 10 6 + ( 0.5 ) 1.750 × 10 6 + ( 0.5 ) 1.750 × 10 6 (1 + i ) 

1

By trial-and-error,

i = 44.8%

Hence, by the discounted cash flow method, the rate of return on the initial capital
investment is approximately 5% greater for the encapsulation system than the fixa-
tion system. From a purely financial standpoint, the encapsulation system is more
attractive.

10.5 Maximum Mercury Waste Emission Constraints

Are mercury emissions important? This question was briefly discussed in Chapter 8,
Section 8.5. This application involves an extension of that problem.
One of the most important issues is what environmental factors influence the formation
of methyl mercury, the element’s most toxic chemical compound and the one most easily
incorporated into an aquatic food chain. More than 95% of the mercury in fish is in the
form of methyl mercury. Most atmospheric mercury occurs simply in its elemental form,
which must be oxidized before being carried to the earth in rain. In a lake, the transfor-
mation of the oxidized form to methyl mercury, a process called methylation, takes place
mainly through the action of bacteria. The amount of methyl mercury available in the
water at any time depends on a balance between methylation and its opposite reaction,
demethylation. Researchers are particularly interested in determining what factors can tip
this balance one way or the other.
Fish can absorb the methyl mercury directly through their gills or ingest it by eating
smaller organisms. Microscopic plants (phytoplankton) absorb considerable amounts of
methyl mercury from the water, so the small fish that eat them receive an already concen-
trated dose. Large predatory fish then eat the smaller ones, and methyl mercury builds
up in their bodies because it is difficult to excrete. Eventually, the magnitude of such
148 Introduction to Optimization for Chemical and Environmental Engineers

bioconcentration becomes staggering. Fish at the top of a food chain may have levels of
methyl mercury a million times the level in the surrounding water.6,7

Illustrative Example 10.5

Consider the following application. Refer to Chapter 8, Section 8.5. If no more than
150 lb/h of mercury can be incinerated, what is the maximum amount of mercury-
contaminated plating waste that can be incinerated in order to operate at maximum
profit?

SOLUTION:
A seventh constraint is added to the earlier application (See also Table 8.4 and Figure 8.1).
It is required to keep the Hg < 150 lb/h. This may be represented by an equation in the
form:

0.02msludge + 0.08mplate < 150 lb/h (10.30)

The constraint is shown as line 7 in Figure 10.1. The equation of this line (providing the
upper limit of msludge) is

150  0.08 
msludge = −  mplate
0.02  0.02  (10.31)
= 7500 − 4.0mplate
Given all seven constraints, the permissible area is then the quadrilateral ABCD
bounded by lines 2, 5, 6, and 7. The farthest point to the right in this area (point C) on

8000

7000 Line 7

6000 Line 4 Line 3

5000 Line 1
msludge lb/h

4000

3000 A
Line 5
B
2000

D C
1000 Line 2
Line 6

0
0 1000 2000 3000 4000 5000 6000
mplate lb/h

FIGURE 10.1
Diagram showing seven constraints for mercury application.
Solid Waste Management 149

the figure is the maximum incineration rate of the plating waste. This occurs where
constraints 6 and 7 interest, that is,

 2431 lb/h 
7500 lb/h − 4.0mplate = 2431 lb/h −   mplate
 3125 lb/h 
mplate = 1573 lb/h

Interestingly, the shaded area on the figure represents the permissible operating range
for this unit given all seven constraints. The operating range does not allow much flex-
ibility. The solution is deterministic, and variation in waste, operator error, waste analy-
sis uncertainty, and so on, are unaccounted for. This graphical technique is adequate for
the two variables investigated in this problem.
This problem can also be solved using the optimization techniques such as linear
programming, a technique that can handle numerous constraints and variables and
can be used, for example, to minimize supplemental fuel costs while operating within
given system constraints. This approach will be employed to solve numerous applica-
tions later in this book.

10.6 Structural Considerations of Concrete Mix

While the content of this book is directed toward environmental and chemical engi-
neers, Section 10.6 is of great relevance to civil engineers, in particular structural
engineers. Structural engineers deal with various types of construction–ranging from
bridges, to buildings, and even roads. There are also some who specialize in the evalua-
tion of structures. These practicing engineers are responsible for evaluating both exist-
ing and future structures. They determine whether existing structures should be built
upon, left-alone, or demolished, and what the foundations for future projects should
be. One of the main materials they work with is concrete. Concrete can be viewed as
an artificially engineered material made from a mixture of Portland cement, water, fine
and coarse aggregates, and a small amount of air. It is the most widely used construc-
tion material in the world. Concrete is used more than all other building materials
combined in the United States.
Concrete is the only major building material that can be delivered to the job site in a
pliable state. This unique quality makes concrete most desirable as a building material
because it can be molded to virtually any form or shape and provide for continuity of
structural action. Concrete provides a wide latitude in surface textures and colors and can
be adapted to many uses, including construction of highways and streets, bridges, dams,
large buildings, airport runways, irrigation structures, breakwaters, piers and docks, side-
walks, silos and farm buildings, homes, and even barges and ships.7
Other desirable qualities of concrete as a building material are its strength, economy,
and durability. Depending on mixture of materials used, concrete will support, in com-
pression, approximately 700,000 g/cm2 (10,000 lbf/in2). The tensile strength of concrete is
much lower, but by using properly designed steel reinforcing, structural members can be
made that are as strong in tension as they are in compression.7
It is convenient to think of concrete in terms of two major components: paste and inert
materials. The paste consists of Portland cement, water, and some air – either in the form
of naturally entrapped air voids or minute, intentionally entrained air bubbles. The inert
150 Introduction to Optimization for Chemical and Environmental Engineers

materials are usually composed of sand (fine aggregate) and gravel, crushed stone, and
slag (coarse aggregate).7
Under normal conditions, concrete grows stronger as it grows older. The chemical reac-
tions between cement and water that cause the paste to harden and bind the aggregates
together require time. The reactions take place very rapidly at first and then more slowly
over a long period of time. In the presence of moisture, concrete continues to gain strength
for years.7
Concrete mixtures are usually specified in terms of the dry-volume ratios of cement,
sand, and coarse aggregates used. A 1:2:3 mixture, for instance, consists of one part by vol-
ume of cement, two parts of sand, and three parts of coarse aggregate. Depending on the
applications, the proportions of the ingredients in the concrete can be altered to produce
specific changes in its properties, particularly strength and durability. The ratios can vary
from 1:2:3 to 1:2:4 and 1:3:5; these mixtures are about 1–1.5 times the volume of the cement.
For high-strength concrete, the water content is kept low, with just enough water added to
preserve the ductility of the mix.7

Illustrative Example 10.6

Consider the following application. As a recently hired structural engineer, you are
given two different types of concrete to be used in the construction of a new high rise
in Snehta, Eceerg. The first type (1) contains 30% glass cement, 40% metal gravel, and
30% sand (all percentages by weight). The second type contains (by weight) 10% glass
cement, 20% metal gravel, 70% sand. The first type of concrete costs $5 per pound and
the second type costs $1 per pound. How many pounds of each type of concrete should
one buy and mix together so that the cost, C, is minimized but one can obtain a concrete
mixture that has at least a proportional total of five pounds of cement, three pounds of
gravel and four pounds of sand for optimal strength?

SOLUTION:
The objective function is:

C = f ( x1 , x2 ) = 5x1 + x2 (10.32)

where x1 and x2 represent the mass of concrete (1) and (2), respectively.
The constraints are:

0.3 x1 + 0.1x2 ≥ 5 (10.33)

0.4 x1 + 0.2x2 ≥ 3 (10.34)

0.3 x1 + 0.7 x2 ≥ 4 (10.35)

with

x1 ≥ 0 (10.36)

x2 ≥ 0 (10.37)

The calculation is left as an exercise for the reader. (Hint: x1 = 0, and x2 = 50.)
Solid Waste Management 151

References
1. T. Shen, Y. Choi, and L. Theodore, EPA Manual: Hazardous Waste Incineration, USEPA/APTI,
RTP, NC, 1985.
2. J. Santoleri, J. Reynolds, and L. Theodore, Introduction to Hazardous Waste Incineration, 2nd edn,
John Wiley & Sons, Hoboken, NJ, 2000.
3. F. Taylor and L. Theodore, Probability and Statistics, A Theodore Tutorial, Theodore Tutorials,
East Williston, NY, originally published by USEPA/APTI, RTP, NC, 1999.
4. S. Shaefer and L. Theodore, Probability and Statistics for Environmental Science, CRC Press/Taylor
& Francis Group, Boca Raton, FL, 2007.
5. L. Theodore and R. Dupont, "Environmental Health and Hazard Risk Assessment: Principles
and Calculations", CRC Press/Taylor & Francis Group, Boca Raton, FL, 2014.
6. R. Dupont, K. Ganesan, and L. Theodore, Pollution Prevention, Sustainability, Industrial Ecology,
and Green Science and Engineering, CRC Press/Taylor & Francis Group, Boca Raton, FL, 2017.
7. K. Behan, personal notes, Buffalo, NY, 2017.
11
Health Risk Assessment

The next to last chapter in Part II contains six illustrative examples. Example titles are
listed here:

11.1 Minimum purging time for a vessel to permit entry

11.2 Minimizing the cost of cancer-fighting drugs
11.3 Reducing insurance costs
11.4 Reducing health concerns associated with flue gas emissions
11.5 Vitamin caloric and protein requirement
11.6 Pharmaceutical cancer drug

Seven select references concerned with health risk assessment (HRA) are provided here:

1. L. Theodore, Chemical Engineering: The Essential Reference, McGraw-Hill, New York

City, NY, 2014.
2. J. Reynolds, J. Jeris, and L. Theodore, Handbook of Chemical and Environmental
Engineering Calculations, John Wiley & Sons, Hoboken, NJ, 2004.
3. L. Theodore, R. Dupont, and K. Ganesan, Unit Operations in Environmental
Engineering, Wiley‑Scrivener, Salem, MA, 2018.
4. M. K. Theodore and L. Theodore, Introduction to Environmental Management, CRC
Press/Taylor & Francis Group, Boca Raton, FL, 2008.
5. G. Burke, B. Singh, and L. Theodore, Handbook of Environmental Management and
Technology, 2nd edn, John Wiley & Sons, Hoboken, NJ, 2000.
6. L. Theodore and R. Dupont, Environmental Health and Hazard Risk Assessment:
Principles and Calculations, CRC Press/Taylor & Francis Group, Boca Raton, FL,
2012.
7. R. Dupont, K. Ganesan, and L. Theodore, Pollution Prevention: Sustainability,
Industrial Ecology, and Green Science and Engineering, CRC Press/Taylor & Francis
Group, Boca Raton, FL, 2017.

11.1 Minimum Purging Time for a Vessel to Permit Entry

The atmosphere in many process vessels and tanks in chemical and petrochemical plants
is maintained free of oxygen to avoid fire, explosion, and/or product deterioration. It is
sometimes necessary, however, for a worker to enter such a vessel for maintenance or
inspection tasks. When this is done, the vessel is normally isolated from the rest of the

153
154 Introduction to Optimization for Chemical and Environmental Engineers

plant and purged with air to provide a breathable atmosphere. Alternatively, the worker
may wear an airpack or other breathing apparatus.1

Illustrative Example 11.1

A 1200 ft3 vessel at a chemical plant normally containing hydrocarbons and blanketed
with nitrogen is to be inspected intermittently. After the vessel is drained and all poten-
tially explosive chemicals removed, the vessel is isolated from the rest of the plant using
blinds on incoming and exiting flow lines. A 100 ft3/min (cfm) flow of air is then started
in the vessel. The purging time is presently eight hours.
The plant has initiated a study to maximize process performance by optimizing
(hopefully minimizing) the purging time of the operation. As part of the study, the
plant engineer has requested information on what is the minimum period of purging
time required to ensure safe entry into the vessel? In effect, how long should this air
flow go on before a worker attempts to enter the vessel?1
The oxygen concentration in the vessel should be at least 19.5% before the worker
enters it. The oxygen content in the purge air is 21%. Also, assume the contents of the
vessel are well mixed.

SOLUTION:
Set up an unsteady-state material balance for oxygen in the vessel. Let
C = oxygen concentration, mole fraction
V = vessel volume, ft3
F = air flow rate, ft3/min
t = time, minutes
CA = oxygen concentration in purge air
C0 = oxygen concentration in vessel at time zero

Apply the conservation law2,3

Accumulation = ( flow in ) − ( flow out ) (11.1)

Substituting appropriate terms:

(V ) 
dC 
 = ( F )(C A ) − ( F )(C ) (11.2)
dt 
Solve the resulting differential equation. This is a simple, first-order equation. The
result is

  t 
C = C A − ( C A − C0 ) 1 − exp  −   (11.3)
  V F  
Rearranging to solve for t:

 V   (C A − C0 )  (11.4)
t =   ln  
 F   (C A − C ) 
Calculate the time required for the oxygen concentration to reach 18% for C0 = 0:

 V   (C A )  (11.5)
t =   ln  
 F   (C A − C ) 
Health Risk Assessment 155

Inserting the values for C, CA, F, and V:

 1200   0.21 
t= ln = (12)(2.64)
 100   0.21 − 0.195 

= 31.6 min

= 0.53 h

In practice, some extra time would be allowed as a safety factor before entering and
an assistant should be standing by outside the vessel.

11.2 Minimizing the Cost of Cancer-Fighting Drugs

Here are two definitions the reader should be familiar with:4

1.
Cancer: A tumor formed by mutated cells
2.
Carcinogen: A cancer-causing chemical

For carcinogens, risks are estimated as the incremental probability of an individual

developing cancer over a lifetime as a result of exposure to the potential carcinogen (i.e.,
incremental or excess individual lifetime cancer risk). Guidelines are provided in the
Environmental Protection Agency (EPA)’s Guidelines for Carcinogen Risk Assessment.5 For
some carcinogens, there may be sufficient information on the mechanism of action that a
modification of this approach is possible.
Perhaps the key technical term in the cancer field is the slope factor. The slope factor
(SF) converts estimated daily intakes, averaged over a lifetime of exposure, directly to the
incremental risk of an individual developing cancer. Because relatively low intakes (com-
pared to those experienced by test animals) are most likely from environmental exposures
at most sites, it is generally assumed that the dose-relationship will be linear in the low-
dose portion of the multistage model dose-response curve. Under this assumption, the SF
is a constant, and risk will be directly related to intake. This linear low-dose risk equation
is described in the following manner:

Risk LD = ( CDI )( CSF ) (11.6)

where:
RiskLD is a unitless probability (e.g., 2 × 10−5) of an individual developing cancer at low-
dose conditions
CDI is the chronic daily intake averaged over 70 years, mg/kg‑day
CSF is the carcinogenic SF, expressed in (mg/kg‑day)−1

The risk obtained from Equation 11.6 may be viewed as the product of an exposure term
(CDI) and a cancer “potency” factor, SF. However, this linear equation is valid only at low-
risk levels (i.e., below estimated risks of 0.01). For situations where chemical intakes might
be high (i.e., risks above 0.01), an alternative equation should be used. The one-hit equation,
156 Introduction to Optimization for Chemical and Environmental Engineers

which is consistent with the linear low-dose model given previously and described in the
following equation, should be used instead:

Risk HD = 1 − e (
− CDI×SF )
(11.7)

where RiskLD is a unitless probability (e.g., 2 × 10−5) of an individual developing cancer at

high dose levels.
Because the SF is often an upper 95% confidence limit of probability of response based
on experimental animal data used in a multistage model, the carcinogenic risk estimate
will generally be an upper-bound estimate. This means that the EPA is reasonably confi-
dent that the “true risk” will not exceed the risk derived through use of this model and is
likely to be less than that predicted.

Illustrative Example 11.2

Consider the following application and note that this is a health risk problem concerned
with cancer-fighting drugs. A cancer institute has determined that there are n drugs
available for cancer “treatment.” Each drug costs $c and each drug contains a maximum
number of m anti-cancer ingredients. The institute’s researchers have also determined
that a minimum of a units of nutrients per day is required to successfully contain the
spread of cancer. If each drug (of a total of n) contains aij units of the ith nutrient, obtain
the objective function and constraints to minimize the total cost of the drugs necessary
for a patient to remain in a state of remission.

SOLUTION:
Denote xj as the number of a specific drug required for remission. The total cost TC
objective function to be minimized is

TC = f ( x1 , x2 ,… xn ) = c1x1 + c2 x2 +…+ cn xn (11.8)

The nutritional constraints associated with each of the m ingredients are given by

a11x1 + a2 x2 ,… + a1n xn > f1

a21x1 + a22 x2 ,… + a2 n xn > f 2

(11.9)


am1x1 + am 2 x2 ,… + amn xn > f m

with

xn ≥ 0
xn ≥ 0
(11.10)


xn ≥ 0
Health Risk Assessment 157

11.3 Reducing Insurance Costs

Illustrative Example 11.3

A structural engineering summer intern at NAHEB Construction Associates was
assigned a project concerned with reducing a company’s insurance costs by installing
induced draft fans on every floor of the company’s eight-story building. The purpose
of the fans is to provide the ventilation needed to reduce the indoor air pollutants that
reportedly have affected the health and performance of several of the company’s per-
sonnel. After several months of reviewing the literature and numerous discussions with
Lou Theodore, the nation’s foremost authority on air pollution, the intern developed the
following economic model:

Bx
IR = A + − Dx ; $ yr (11.11)
C+x
( 1) ( 2 ) ( 3 )

where:
IR is the annual insurance reduction
Term1 is the savings factor concerned with making the employees aware of the
problem
Term2 is the savings factor concerned with the installation of x fans
Term3 is the savings factor (a cost) concerned with the purchase of x fans
Determine the number of fans that should be installed to maximize the annual insur-
ance reduction.

SOLUTION:
In order to reduce insurance costs (a maximization operation), the intern decided to
solve the insurance reduction (IR) model analytically. The calculation involved generat-
ing d(IR)/dx and setting the result equal to zero, that is,

d ( IR )
= 0 (11.12)
dx
For the equation provided in the problem statement,

d ( IR ) B Bx
= − − D = 0 (11.13)
dx C + x ( C + x )2
Equation 11.13 is a form similar to that for a quadratic equation, that is,

( )
B ( C + x ) − Bx − D C 2 + 2Cx + x 2 = 0 (11.14)

or

Dx 2 + 2CDx + C ( CD − B ) = 0 (11.15)

The solution to Equation 11.15 is given by

158 Introduction to Optimization for Chemical and Environmental Engineers

−b ± b 2 − 4 ac
x= (11.16)
2a
where:
a = D
b = 2CD
c = C(CD−B)

Substitution yields

−2CD ± 4C 2D2 − 4CD2 − 4BCD

x= (11.17)
2D
The intern was notified that she would be hired next summer to generate numerical
values for coefficient A, B, C, and D.

11.4 Reducing Health Concerns Associated with Flue Gas Emissions

As noted earlier, there are many definitions of the word “risk.” Some of the definitions
provided are as follows: It is a combination of uncertainty and damage; a ratio of hazards
to safeguards; a triplet combination of event, probability, and consequences; or even a mea-
sure of economic loss or human injury in terms of both the incident likelihood and the
magnitude of the loss or injury.
People face all kinds of risks every day, some voluntarily and others involuntarily.
Therefore, risk plays a very important role in today’s world. Studies on cancer caused a
turning point in the world of risk because it opened the eyes of risk scientists and health
professionals to the world of health risk assessments (HRAs).
The usual objective of HRA and the accompanying calculations are to evaluate the
potential for adverse health effects from the release of chemicals into the environment.
Unfortunately, the environment is very complex since there is a vast array of potential
receptors present. The task of testing and evaluating all of the enormous number of
chemicals on the market for their impact on human populations and ecosystems becomes
extremely difficult. To further complicate this problem, health is a concept that has come
to mean different things to different people. Some have defined it as follows: “…a state of
complete physical, mental and social well-being and not merely the absence of disease or
infirmary.” Many other definitions and concepts have been proposed and appear in the
literature.
Since 1970, the field of HRA has received widespread attention within both the scientific
and the regulatory communities. It has also attracted the attention of the public. Properly
conducted risk assessments and risk assessment calculations have received fairly broad
acceptance, in part because they put into perspective the terms toxic, health, risk, and
hazard (see also Chapter 13). Toxicity is an inherent property of all substances. It states
that all chemical and physical agents can produce adverse health effects at some specific
exposure conditions. In contrast, exposure to a chemical that has the capacity to produce
a particular type of adverse effect represents a health problem or “hazard.” Health risk
(in a general sense), however, is the probability or likelihood that an adverse outcome will
occur in a person or a group that is exposed to a particular concentration or dose of the
Health Risk Assessment 159

hazardous agent. Health risk is therefore a function of exposure and dose. Consequently,
HRA is defined as the process or procedure used to estimate the likelihood that humans
or ecological systems will be adversely affected by a chemical or physical agent under a
specific set of conditions.
Corporations that manufacture, sell, and purchase chemicals now realize that the chem-
icals they handle present health risks to their employees, their customers, and/or the pub-
lic. The assessment process is often an enormous task since the health risks of scores of
chemicals may have to be assessed from a risk perspective. Each individual HRA is a
multistep process consisting of

1. Health problem identification

2. Toxicological concerns
3. Exposure characterization
4. Risk characterization

Illustrative Example 11.4

In an attempt to minimize the reduction of a host of gaseous pollutants, a state-run util-
ity has proposed to convert some of its coal-fired boilers to oil-fired in order to reduce
health concerns associated with flue gas emissions. DuPont Associates was hired to
generate a model (objective function) that would minimize the cost of the replacement
relative to the quantity of oil to be combusted. Two oils – (1) and (2) – are available to
supply the necessary energy for the new boilers, x1 and x2. The proposal also requested
that an attempt be made to salvage as many of the old boilers (x3) currently burning
coal.6 The cost associated with each oil varies with its content, regulatory requirements,
heating value, and so on. The engineers at Dupont Associates proposed the following
cost model, DA:

DA = 3 x1 + 4 x2 − x3 ; 10 −4 $ (11.18)

(Salvaging the existing boilers resulted in a credit.) The constraints, primarily due to
emission rules for NOX, SOX, and ROX were:

x2 + x3 ≤ 100 ( nitrous oxide, NOX ) (11.19)

x1 + x3 ≤ 50 ( sulfur dioxide, SOX ) (11.20)

x1 + x2 x3 + x32 ≤ 1000 ( particulates, ROX ) (11.21)

with
x1 < x2 < x3 ≤ 50 (11.22)

The state-run facility has been informed that capital cost expenditures would have to
be below $500,000 for this cost minimization project.

SOLUTION:
Excel provides the following solution:

x1 = 4
160 Introduction to Optimization for Chemical and Environmental Engineers

x2 = 10

x3 = 26

with

DA = $26 × 10 4 = $260, 000

Thus, this minimization project should go forward. The solution requires the pur-
chase of four new boilers that burn oil (1), 10 new boilers that will burn oil (2), and the
reassignment of 26 of the old coal-fired boilers.

11.5 Vitamin Caloric and Protein Requirements

A protein is any of a large number of organic compounds that makes up living organ-
isms and is essential to their functioning. Protein molecules range from the long, insol-
uble fibers that make up connective tissue and hair to the compact, soluble globules
that can pass through cell membranes and set off metabolic reactions. They are all large
molecules, ranging in molecular weight from a few thousand to more than a million,
and they are specific for each species and for each organ of each species. Humans have
an estimated 30,000 different proteins, of which only about 2% have been adequately
described. Proteins in the diet primarily serve to build and maintain cells, but their
chemical breakdown also provides energy, yielding close to the same four calories per
gram as do carbohydrates.

Illustrative Example 11.5

Vitamin (x1) costs $0.20 per mg and contains 50 cal/vitamin and 20 mg. protein/
vitamin. Vitamin  (x 2) costs $0.15 per mg and contains 40 cal/vitamin and 10 mg.
protein/vitamin. What amounts of vitamins x1 and x 2, in mg would meet the mini-
mum regulatory requirement while minimizing the cost, C. The US Department of
Agriculture (USDA) provided the following objective function for C:

C($) = 0.20 x1 + 0.15x2 (11.23)

along with the constraints

50 x1 + 40 x2 ≥ 300 (11.24)

20 x1 + 10 x2 ≥ 120 (11.25)

and

x1 ≤ 6 (11.26)

x2 ≤ 6 (11.27)
Health Risk Assessment 161

SOLUTION:
Excel provides the following solution for the number of vitamins:

x1 = 6

x2 = 0

and

C = $1.20

11.6 Pharmaceutical Cancer Drug

Cancer is defined as new growth of tissue resulting from a continuous proliferation of
abnormal cells that have the ability to invade and destroy other tissues. Cancer is known
throughout the world as a fearsome cause of illness and death. Not all cancers are the
same, however, nor are all persons equally at risk. The disease can occur in any tissue
where cells are growing and dividing, including such widely different tissues as those of
the lungs, skin, brain, stomach, and bones. Each kind of cell gives rise to its own kinds of
cancer, with its own causes, frequency, symptoms, treatment, and outcome.
Some of the most common kinds of cancer are those of the lung, prostate, and large
bowel (for older men) and of the breast, large bowel, and uterus (for older women). Cancer
is less common in children and young adults; when it occurs, it is likely to arise in the bone
marrow (the forms of cancer called leukemia), lymphatic tissue (including Hodgkin’s dis-
ease), brain, kidneys, and bones. These generalizations exclude the common forms of skin
cancer, which – unique among cancers – are easily diagnosed and treated in their early
stages and that rarely cause serious disability unless long neglected.
Cancer is often discovered because it causes symptoms that lead the patient to seek medi-
cal help, or it may be found by special screening examinations beforehand. Good reasons
exist to believe that the earlier a cancer is detected and treated, the higher the chance of cure.
Cancer is a disease of individual cells, and even one cell can grow and divide many
times to produce a new cancer mass. Thus, the cure can be accomplished only by removing
or killing every single malignant cell. The three accepted ways of doing this are surgery,
radiation, and drugs, which are often used in combination. Several dozen drugs are used
at present to treat cancer, and research has shown just how the drugs should be used for
best effect. The size and timing of each dose may have profound effects on drug action, as
may the drug’s integration with surgical or radiation treatment. Moreover, because some
of the drugs attack cancer cells in different ways, the use of as many as four or five such
drugs in a precisely determined combination or sequence is commonly more effective
than the use of just one.

Illustrative Example 11.6

ABC Drug Laboratory is working to produce four variations of a new cancer drug.
Initial estimates of the predicted hourly profit from the production of drugs A, B, C,
and D were $100, $150, $200, and $250 respectively. Table 11.1 shows how much time
162 Introduction to Optimization for Chemical and Environmental Engineers

TABLE 11.1
Drug Laboratory Production Time Expenditure (Hours)
for Illustrative Example 11.6
Lab Drug A Drug B Drug C Drug D
1 10 4 8 9
2 6 5 3 9
3 3 9 7 12

is required in each of the three research labs to make one unit of each of the four
products.
Laboratory 1 has no more than 300 man-hours available per day. Laboratory 2 is lim-
ited to no more than 400 man-hours per day. Similarly, Laboratory 3 is limited to 425
man-hours per day.
As a recent biomedical engineering graduate, you have been hired to determine the
number of A, B, C, and D units that should be produced to maximize daily profit, P. The
following information is provided

P = 100 x1 + 150 x2 + 200 x3 + 250 x 4 ; $ h (11.28)

where x1, x2, x3, and x4 represent the number of drugs A, B, C, and D, respectively, pro-
duced per hour. The man-hour constraints are

10 x1 + 4 x2 + 8 x3 + 9 x 4 ≤ 300 (11.29)

6 x1 + 5x2 + 3 x3 + 9x 4 ≤ 400 (11.30)

3 x1 + 9x2 + 7 x3 + 12x 4 ≤ 425 (11.31)

and

x1 ≥ 0 (11.32)

x2 ≥ 0 (11.33)

x3 ≥ 0 (11.34)

x 4 ≥ 0 (11.35)

SOLUTION:
Excel provides the following solution:

x1 = 0

x2 = 29

x3 = 23

x4 = 0
Health Risk Assessment 163

with a profit of

P = $8, 950 h

= $214, 900 day ( round the clock operation )

References
1. L. Theodore and R. Dupont: Environmental Health and Hazard Risk Assessment: Principles and
Calculations, CRC Press/Taylor & Francis Group, Boca Raton, FL, 2012.
2. J. Reynolds, J. Jeris, and L. Theodore, Handbook of Chemical and Environmental Engineering
Calculations, John Wiley & Sons, Hoboken, NJ, 2004.
3. L. Theodore, R. Dupont, and K. Ganesan, Unit Operations in Environmental Engineering,
Wiley‑Scrivener, Salem, MA, 2018.
4. L. Theodore, J. Reynolds, and K. Morris, Concise Dictionary of Environmental Terms, Gorden and
Breach Science Publishers/CRC Press/Taylor & Francis Group, Boca Raton, FL, 1997.
5. USEPA, Guidelines for Carcinogen Risk Assessment, EPA/630/R-98/002, Risk Assessment Forum,
Washington, DC, 2005.
6. K. Skipka and L. Theodore, “Energy Resources: Availability Management, and Environmental
Impacts”, CRC Press/Taylor & Francis Group, Boca Raton, FL, 2014.
12
Hazard Risk Assessment

The last chapter in Part II contains six illustrative examples. Example titles are listed here:

12.1 Safety expenditures versus return of safety investments

12.2 Plant safety proposal comparisons
12.3 Effect of failure distribution rate
12.4 Estimating the effect of worst-case explosions
12.5 Economics of fire protection equipment
12.6 Optimizing the operation of a cement incinerator

Seven references are recommended for suggested reading:

1. L. Theodore, Chemical Engineering: The Essential Reference, McGraw-Hill, New York

City, NY, 2014.
2. J. Reynolds, J. Jeris, and L. Theodore, Handbook of Chemical and Environmental
Engineering Calculations, John Wiley & Sons, Hoboken, NJ, 2004.
3. L. Theodore, R. Dupont, and K. Ganesan, Unit Operations in Environmental
Engineering, Wiley‑Scrivener, Salem, MA, 2018.
4. M. K. Theodore and L. Theodore, Introduction to Environmental Management, CRC
Press/Taylor & Francis Group, Boca Raton, FL, 2008.
5. G. Burke, B. Singh, and L. Theodore, Handbook of Environmental Management and
Technology, 2nd edn, John Wiley & Sons, Hoboken, NJ, 2000.
6. L. Theodore and R. Dupont, Environmental Health and Hazard Risk Assessment:
Principles and Calculations, CRC Press/Taylor & Francis Group, Boca Raton, FL,
2012.
7. L. Theodore, Environmental Risk Analysis: Probability Distribution Calculations, CRC
Press/Taylor & Francis Group, Boca Raton, FL, 2016.

12.1 Safety Expenditures versus Return of Safety Investments

Protective systems are employed to reduce/eliminate hazards and risks and may be
viewed as a special category of process equipment. The management of the protec-
tive system’s operation and maintenance and testing information should occur within
the overall management approach. Provisions for protection and safety equipment
should be incorporated in the original plant design. The size of the plant, nature of the

165
166 Introduction to Optimization for Chemical and Environmental Engineers

hazards, and potential exposures will determine the amount, kind, and location of this
equipment.
Regarding fires, water is the primary extinguishing agent, and it should be available
in adequate supply and at adequate pressure at all locations in the plant. The layout for
various types of fire protection installations and the appropriate recommendations for
their design are found in the standards of the National Fire Protection Association. Fire
hydrants, hose lines, and automatic sprinklers and water spray systems should all be a
part of the permanent equipment facilities of the plant.
Fire extinguishing systems can include foam, carbon dioxide, and dry chemicals. Wetting
agents and high-expansion foam have been used in some plant protection systems. All fire
extinguishing systems should be evaluated for their potential health risks as well as their
overall effectiveness before incorporation into a plant.
Since many chemical plants have severe potential health and hazard problems, it is
essential to provide medical facilities and first aid stations in case rapid medical response
is required. In addition, showers and eyewash stations are necessary in certain hazard-
ous areas. Also, guards and covers should be provided for all moving equipment. Ten of
the key protective equipment elements in an environmental or chemical process plant or
manufacturing facility are listed here:1

1. Pressure relief/vent collection

2. Release devices (flares, scrubbers, etc.)
3. Plant equipment isolation
4. Critical alarms/interlocks
5. Fire detection/protection
6. Hazardous/toxic gas detection
7. Flame arrestors
8. Emergency system services
9. Appropriate grounding and bonding
10. Personal protection equipment

Illustrative Example 12.1

BETH Construction Company has decided to implement a new proposal from Doyle
Engineers in the construction of a 57-story high-rise building in San Francisco that
will house offices on floors 1–6 and apartments on the floors above. The new proposal
from the structural engineer at Doyle recommends that their recent patented quake-
buster columns be inserted at strategic locations in the building in order to allow the
structure to survive earthquakes up to 7.5 on the Richter scale. Billy B engineers have
estimated that the value (worth) of the augmented high rise, including the column
costs, is given by

x ( 2.56 )
V = 25.6 + − 0.015x ; MM$ (12.1)
x + 3.5

where x = number of quake-buster columns

Determine the optimum number of columns that BETH should install in the high rise.
Hazard Risk Assessment 167

SOLUTION:
This illustrative example requires the optimization of V. To employ an analytical solu-
tion, proceed as follows:

dV 2.56 x ( 2.56 )
= − − 0.015 (12.2)
dx x + 3.5 x + 3.52

Setting this equation equal to zero, that is, dV/dx = 0 and solving for x (positive/negative)
yield x ≅ 20. Therefore, 20 quake-buster columns should be employed in the new high-
rise to maximize the value of the structure, increasing it from 25.6 to 27.5 MM$.

12.2 Plant Safety Proposal Comparisons

The academic training of engineers with respect to plant equipment has traditionally
focused on design and predicting performance. Optimizing operation from a safety
perspective is often an overlooked consideration. Little to no effort has historically been
expended in attempting to answer the question

WHAT CAN GO WRONG?

This question requires answers in today’s high-tech environment. Failure to properly and
realistically address and answer this question can result/lead to an

ACCIDENT !

Once an industrial accident has occurred, the responsible individual (usually the oper-
ator or engineer) should react as quickly as possible in a responsible manner (exercise
good judgment) to detect and correct the situation, and hopefully reduce and/or eliminate
potential danger(s).
Industry now knows that equipment sometimes partially fails and sometimes cata-
strophically fails. In any event, it behooves the engineer to carefully and thoroughly exam-
ine the various pieces of plant equipment that can fail. A partial list of some of the more
common equipment that can be expected to fail in a plant with some regularity is pro-
vided subsequently. Details are available in the literature.1,2

1. Foundations
2. Structural steel
3. Vessels and tanks
4. Pumps
5. Compressors
6. Fans
7. Heat exchangers
8. Turbines
9. Electrical systems
10. Instrumentation and controls
168 Introduction to Optimization for Chemical and Environmental Engineers

11. Piping systems

12. Valves and joints
13. Mass transfer operations
14. Heaters and furnaces
15. Chemical reactors
16. Fire protection equipment
17. Safety equipment

Deviations, that is, abnormal conditions that can lead to a failure and/or an accident gener-
ally occur because of

1. Abnormal temperatures
2. Abnormal pressures
3. Material flow stoppage
4. Equipment leaks
5. Equipment spills
6. Materials failure due to wear
7. Materials failure due to imperfections
8. Materials failure due to poor maintenance
9. Materials failure due to corrosion

Some specific operational failures include

1. Blocked outlets
2. Opening/closing valves
3. Cooling water failure
4. Power failure
5. Instrument air failure
6. Thermal expansion
7. Vacuum problems

Illustrative Example 12.2

Consider the following application. Four different plant safety proposals are under con-
sideration at the KBLT Power Plant. The life of each project is 20 years. The individual
capital and annual operating costs are provided in Table 12.1. The salvage value at the
termination of the project is zero and there are no demolition costs.
Which is the least expensive project? Assume the straight-line depreciation method
applies.

SOLUTION:
The total annualized cost calculations (TAC) are provided in Table 12.2.
Comparing projects 1, 2, 3, and 4 indicates that unit 3 has the lowest TAC and should
be selected as the most economical project of the four being evaluated.
Hazard Risk Assessment 169

TABLE 12.1
Cost Data for Four Safety Purposes; Illustrative Example 12.1
Project Initial Cost, $ Operating Cost, $/yr
1 2,600,000 200,000
2 2,400,000 250,000
3 2,450,000 275,000
4 1,850,000 350,000

TABLE 12.2
Cost Analysis for Four Safety Purposes; Illustrative Example 12.1
Project 1 2 3 4
Capital cost ($) 2,600,000 2,400,000 2,450,000 1,850,000
Depreciation ($) 130,000 120,000 125,000 92,250
Operating costs ($) 200,000 150,000 275,000 350,000
TAC ($) 330,000 270,000 400,000 442,250

12.3 Effect of Failure Distribution Rate

The probability distribution of a random variable concerns the distribution of probabil-
ity over the range of the random variable. The distribution probabilities are specified by
the probability distribution function (pdf). This introduction is devoted to providing general
properties of the pdf in the case of discrete and continuous random variables, as well as
an introduction to the cumulative distribution function (cdf). Special pdfs finding extensive
application in hazard and risk analysis are available in the literature.3
The pdf of a discrete random variable X is specified by f(x), where f(x) has the following
essential properties:

1. f ( x ) = P ( X = x ) = probability assigned to the outcome

corresponding to the number x in the range of X ; (12.3)
i.e., X is a specifically designated value of x

2. f ( x ) ≥ 0 (12.4)

3. ∑ f (x) = 1 (12.5)
x

Property 1 indicates that the pdf of a discrete random variable generates probability by
substitution. Property 2 and Property 3 restrict values of f(x) to non-negative real numbers
and numbers whose sum is 1, respectively.
The pdf of a continuous random variable x has the following properties:

∫ f ( x ) dx = p(a < x < b) (12.6)
a
170 Introduction to Optimization for Chemical and Environmental Engineers

f ( x ) ≥ 0 (12.7)

∫ f ( x ) dx = 1 (12.8)
−∞

Equation 12.6 indicates that the pdf of a continuous random variable generates probability
by integration of the pdf over the interval whose probability is required. When this inter-
val contracts to a single value, the integral over the interval becomes zero. Therefore, the
probability associated with any particular value of a continuous random variable is zero.
Consequently, if x is continuous

P ( a ≤ x ≤ b ) = P( a < x ≤ b)

= P( a < x < b) (12.9)

= P( a ≤ x < b)

Equation 12.7 restricts the values of f(x) to non-negative numbers. Equation 12.8 follows
from the fact that

P ( −∞ < x < ∞ ) = 1 (12.10)

The following properties of the cdf of a random variable, x, can be deduced directly from
the definition of F(x).

F ( b ) − F ( a ) = P( a < x ≤ b) (12.11)

F ( +∞ ) = 1 (12.12)

F ( −∞ ) = 1 (12.13)

F ( x ) is a nondecreasing function of x (12.14)

These four properties apply to the cases of both discrete and continuous random variables.

Illustrative Example 12.3

Consider the following application. You have been assigned to run ten 80-day corrosion
tests. Each testing device can be used for only one test, after which it must be discarded.
Device A costs $400 each. Data from past tests indicate these devices have a failure rate
distribution curve shown in Figure 12.1.

1. How many devices would you expect to use?

2. Competing device, Device B, with the uniform failure rate distribution pro-
vided in Figure 12.2, is available for $500 each. To minimize costs, is it eco-
nomically advantageous to purchase the more expensive device?
Hazard Risk Assessment 171

0.015
0.0125
0.010
f(t) 0.075
0.005
0.025
0.0
8 16 32 64 128
Time (d)

FIGURE 12.1
Failure rate, f(t), as a function of time for Device A; Illustrative Example 12.3.

0.015

0.010
f(t)= 0.008
f(t)

0.005

0.0
8 16 32 64 128
Time (d)

FIGURE 12.2
Failure rate, f(t) as a function of time for Device B; Illustrative Example 12.3.

SOLUTION:

1. Calculate the probability of the device lasting at least 80 days using the follow-
ing equation:

80 8 80

∫
F ( t ) = f ( t ) dt =
0
∫ (0.0125) dt + ∫ (0.005) dt
0 8

= 0.10 + 0.36 = 0.46

number of tests required 10

Number of A devices most likely required = = = 21.7
probability of success 0.46

Therefore, 22 are needed at a total cost of $8,800.

2. The probability of success for the B device devices is

80

F (t ) =
∫ (0.008 ) dt = 0.64
0
172 Introduction to Optimization for Chemical and Environmental Engineers

Therefore, the number of B devices needed = 10/0.64 = 15.6 = 16 for a total cost = $8,000.

The more expensive devices appear to be justified.
Since this question relates to “most likely” outcomes, it would also be acceptable to
base the calculations on factional devices needed. Doing so yields

 $ 400 
Device A: 21.7  = $8, 680
 device 

 $500 
Device B: 15.6  = $7 , 800
 device 
So, the conclusion remains the same.

12.4  Estimating the Effect of Worst-Case Explosions1,4

One of the most hazardous situations encountered in a process plant is the atmospheric
detonation of hydrocarbon vapors that have been accidentally released. Such explosions,
referred to as “unconfined vapor cloud explosions,” have been responsible for some of the
industry’s worst accidents, such as the one at Flixborough, England, in 1974.1
An explosion at the Nypro Ltd. caprolactam factory in Flixborough, England, on June 1,
1974, was one of the most serious in the history of the chemical industry and the most seri-
ous in the history of the United Kingdom. Of those working on the site, 28 were killed and
36 others injured. Outside the plant, 53 people were reported injured, while 1821 houses
and 167 shops suffered damage. The estimated cost of the damage was well over $100
million.
The cyclohexane oxidation plant in Flixborough, which stored this solvent at about 120
psig and 145°C, contained six reactors in series. One reactor had been removed for repairs,
and the resulting gap was bridged by a temporary 20-in pipe, connected by a bellows at
each end and supported on temporary scaffolding. This pipe collapsed, and in the minute
or so that elapsed before ignition, about 50 tons of cyclohexane escaped. The extensive
damage that occurred could have resulted from the deflagration of as little as 10 to 20 tons
of the chemicals.
It was later determined that no calculations had been done to ascertain whether the
bellows or pipe could withstand the strain. Only the capacity of the assembly to carry the
required flow was calculated. No reference was made to any accepted standard, nor to
the designer’s guide issued by the manufacturer of the bellows. Neither the pipe nor the
complete assembly unit was pressure tested before being fitted. Apparently, no one real-
ized that the pressurized assembly would be subject to a turning moment, imposing shear
forces on the bellows for which they were not designed. Nor did anyone observe that the
hydraulic thrust on the bellows would tend to make the pipe buckle at the joints. The tem-
porary pipe connection functioned satisfactorily after the initial installation. It was never
closely inspected, however, even though when at operating temperature and pressure, it
was observed to lift off the scaffolding that had been put in to support it.
What happened on the final shift will never be known because all those in the control
room were killed, and all instrumentation and records were destroyed. The equivalent
force of the explosion was estimated at 15 tons of TNT, minimum.
Hazard Risk Assessment 173

Explosions are frequently calibrated in terms of the aforementioned equivalent mass of

TNT. One lb of TNT releases 1983 Btu (500 Kcal) on detonation; therefore, any explosion
that releases 1983 Btu is said to be a “1 lb of TNT” explosion.
The overpressure at any distance from an explosion can be estimated using the follow-
ing method. The overpressure from one kg of TNT, as a function of distance, is given by

1737 1.875
P ( psi ) = + − 0.01556 (12.15)
D2 D
with D in feet, for the range from 3 to 400 ft. The distance for equal overpressures then
scales as the 1/3 power of the energy of the explosion (E):

1
 D1   E1  3
 D  =  E  (12.16)
2 2

Illustrative Example 12.4

Consider the following hypothesis. The hot hydrocarbon vapor from a catcracker at
Petro Unlimited suddenly releases vapors to the atmosphere, mixes with the air, and
explodes.

1. Estimate the maximum possible energy release (Emax) and the TNT equivalency.
2. Estimate the radius from the center of the explosion at which overpressures of
1 psi (sufficient to injure many people and cause damage to buildings), and 5
psi (sufficient to break wooden telephone poles) would occur.
3. Based on the results of (1) and (2), discuss how the hazard risks from the cat-
cracker can be minimized.

The cracker has a volume of 800 ft3. The vapors at the time of release consist of light
hydrocarbons with an average molecular weight of 55, at 25 psi and 350°F. The average
heat of combustion in air of the hydrocarbons is 19,000 Btu/lb.

SOLUTION:

1. Calculate the moles of hydrocarbon released and convert to lbs:

PV
N hc =
RT

=
( 25)(8000) = 2.3 lbmol
(10.73)(810)
Mhc = ( 2.30 )( 55 ) = 127 lbs    

Determine the maximum possible combustion energy release and convert it to

lb TNT equivalent:

Emax = (127 )(19, 000 ) = 2, 400, 000 Btu

2, 400, 000
= = 1210 lb TNT
1983
174 Introduction to Optimization for Chemical and Environmental Engineers

2. Calculate the distance for an overpressure of 5 psi. Employ Equation 12.15. For
5 psi overpressure with 1 kg TNT:

1737 1.875
5= + − 0.01156
D2 D
Solving for D:

D = 18.8 ft

  Convert the energy of the blast to kg TNT:

1210 lb TNT
E= = 550 kg TNT
2.2 lb / kg
Scale the distance for 550 kg TNT. Employ equation 12.16:

1 1
 E 3  550  3
D1 = D2  1  = (18.8 )   = 154 ft
E
 2  1 
  Finally, calculate the distance for an overpressure of 1 psi. The distance for 1
psi overpressure with 1 kg TNT is given by:

1737 1.875
1= + − 0.01156
D2 D
Solving for D:

D = 42.1 ft

Scale the distance for 550 kg TNT employing Equation 12.16.

1 1
 E 3  550  3
D1 = ( D2 )  1  = ( 42.1)   = 345 ft
E
 1  1 
3. It has been reported that in most hydrocarbon vapor explosions, only 10 to 25%
of the maximum possible combustion energy is released. This results from the
vapor/air mixture being ignited when the mixture is still “rich”, that is, before
the amount of air for complete combustion has had a chance to mix with the
hydrocarbon vapors. Thus, efforts should be directed to reducing the amount
of energy released. In addition, consideration should be given to surrounding
the catcracker with barriers.

12.5 Economics of Fire Protection Equipment

The most common cause of fire accidents in environmental and chemical process plants is
equipment failure. This is primarily a result of poor equipment maintenance or poor equip-
ment layout and design. Maintenance performed according to a detailed and well-structured
schedule will significantly reduce the occurrence of fire accidents. The second largest cause
Hazard Risk Assessment 175

of fire accidents is ignorance of the properties of a specific chemical or chemical process.

Proper training of employees will increase their knowledge of the properties of a specific
chemical and chemical process, and it can prevent many of these chemical fire accidents.
Leaks and spills resulting from an equipment failure will frequently cause fires. Large
leaks can occur when a vessel, pipe, or pump fails. Failure of a pump seal, pipe flange, or
bore connection will cause a somewhat smaller leak. Since fires at pumps and/or at flanges
are highly possible, it is essential to assess and minimize the effect of such fires on the
equipment above or near pumps or flanges. Incidence of pump fires can be minimized by
using double mechanical seals. Pipe flanges also can be protected against fires by placing
an insulating pad between the joint ring and pipe bore and by specifying more massive
flanges to minimize the temperature gradient.
Thermal insulation (or lagging) on plant equipment may become soaked or impregnated
with oils and other flammable liquids. When the lagging gets hot, spontaneous combus-
tion can occur. Lagging fires are affected by oil leaks, insulation material, and temperature.
Spontaneous combustion occurs only when the oil is non-volatile, since volatile oil evapo-
rates more easily, thus delaying the accumulation of oil that reaches its auto-ignition temper-
ature (AIT).1 A continuous leak may actually inhibit lagging fires since the lagging fire can
be over-saturated. A good insulating material with a low thermal conductivity is favored by
the lagging fire. A low density, porous structure provides surface area for the air to diffuse
in. A thick lagging gives good insulation; therefore, spontaneous combustion is more likely
on a thicker lagging. To prevent lagging fires, the following precautions are recommended:

1. Prevent leaks.
2. Use appropriate sealings (cement finish or aluminum foil).
3. Use special insulation materials such as foam glass or crimped aluminum sheeting.

The potential for fire hazards is rather high in the chemical industry. However, this
potential is generally judged to be less than that of an explosion or toxic release. The scale
of a fire hazard can be determined by assessing the following factors: 1,2

1.
Inventory. The larger the inventory of material, the greater the loss potential.
2.
Energy. The more energy available for a release, such as the stored energy in a
material state or chemical reaction, the greater the potential.
3.
Time. A higher rate of release of a hazard and the distance over which it may cause
injury or damage also directly affect the potential.
4.
Exposure. The intensity of the hazard and the distance over which it may cause
injury or damage also directly affect the potential.

Most fire accidents involve a large loss of life and extensive property damage. Frequently,
a fire is preceded by an explosion. Explosions are usually more lethal than fires.1 Some
major fires in industrial plants have been described in detail in the literature.1,2 The ther-
mal radiation intensity and the time duration of fires often are used to estimate injury and
damage due to a fire.
The best way to fight a fire is to remove any one of three essential conditions required
to sustain the fire:

1. The fuel source (e.g., by eliminating the leaks in a process plant)

2. Heat (e.g., by dousing the fire with water)
176 Introduction to Optimization for Chemical and Environmental Engineers

3. The supply of oxygen (e.g., by applying foams or inert gases). Various fire extin-
guishing systems can be used. These include
a. Water systems
i. Automatic sprinklers
ii. Fire hoses
b. Foam systems
i. Chemical foams: formed by the chemical reaction in which bubbles of CO2
gas and a foaming agent combine to produce and expand froth.
ii. Mechanical foams: bubbles of air, which are produced when air and water
are mechanically agitated with a foam-making agent.
iii. High-expansion foams: tiny foam bubbles filled with air, created by a fan
blowing the air through a mesh over which a detergent base solution is
flowing. High-expansion foam is used where water damage is a problem
or access to an area is not feasible. For example, such a foam would be
suitable if a flammable liquid held in an atmospheric storage tank were to
spill as a result of overfilling, causing the formation of an ignitable vapor
cloud.
c. Carbon dioxide systems.
d. Dry chemical systems (e.g., sodium bicarbonate). These are not toxic and do
not conduct electricity or freeze.
e. Water spray systems. These are used for explosion protection of buildings,
tanks, and control of flammable liquid fires.
f. Steam jet systems. These are used to smother some fires in closed containers or
in confined spaces.

Illustrative Example 12.5

Consider the following application. Ricci Consultants have provided Behan and
Theodore (BAT) Associates with the following model to determine a company’s profits
(P) gained from the installation of two different fire protection devices. Assume the
number of devices installed are x1 and x2. The model is given by

P = 2000 x1 + 2750 x2 (1)

2x1 + 3 x2 ≤ 30 (2)

x1 + 10 x2 ≤ 25 (3)

with

x1 ≥ 0 (4)

x2 ≥ 0 (5)

Qualitatively explain the previously mentioned profit mathematical model; that is, what
number of the two devices will maximize the profits.
Hazard Risk Assessment 177

SOLUTION:
Equation (1) is the objective function and describes the profit that will be produced from
the installation of the safety devices. Equations (2) and (3) are part of the constraints
on the limitation of the number of devices that can be installed because of space and
regulatory issues. Equation (4) and (5) require that the number of devices (naturally) be
a positive number.
The calculations are left as an exercise for the reader. (Hint x1 = 14 devices and x2 = 2
devices)

12.6 Optimizing the Operation of a Cement Incinerator

Monte Carlo simulation is a procedure for mimicking observations on a random variable
that permits verification of results that would ordinarily require difficult mathematical
calculations or extensive experimentation. The method normally uses computer programs
called random number generators. A random number is a number selected from the interval
(0,1) in such a way that the possibilities that the number comes from any two sub inter-
vals of equal “length” are equal. For example, the probability that the number is in the
sub-interval (0.1,0.3) is the same as the probability that the number is in the subinterval
(0.5,0.7). Thus, random numbers are observations on a random variable x having a uni-
form distribution on the interval (0,1). This means that the probability distribution func-
tion (pdf) of x is specified by

f ( x ) = 1; 0 < x < 1 (12.17)

= 0; elsewhere (12.18)

The previously mentioned pdf assigns equal probability to subintervals of equal

length in the interval (0,1). Using random number generators, Monte Carlo simulation
can generate observed values of a random variable having any specified pfd. For exam-
ple, to generate observed values of T, the time to failure, when T is assumed to have a
pdf specified by f(t), one first uses the random number generator to generate the value of
T between 0 and 1. The solution is an observed value of the random variable T having a
pdf specified by f(t).
Monte Carlo methods find application in a host of applied mathematical and engineer-
ing areas. Most of these involve the solution of problems that cannot be solved by analyti-
cal or ordinary numerical methods. Solving these types of problems using Monte Carlo
methods has recently been employed in optimization.
In terms of introduction, consider the system pictured in Figure 12.3. The system is a
solid and the variable T is the temperature; the base temperature is maintained at 100°C
and the other three surfaces at 0°C. For this system, the temperature at each location can
be described by the average temperature of its four neighboring points. The Monte Carlo
procedure provided previously can be used to generate the temperature profile in the solid
in the following manner.
178 Introduction to Optimization for Chemical and Environmental Engineers

y
0°C
0°C 0°C 0°C 0°C
y = b 0°C

1 2 3
0°C 0°C

4 5 6
0°C 0°C

7 8 9
0°C 0°C

y= 0 x
100°C 100°C 100°C 100°C 100°C
x= 0 100°C x =a
a =b

FIGURE 12.3
Monte Carlo gird approach.

With specified boundary conditions (BCs) for T(x, y) of T(0, y), T(a, y), T(x, 0), and T(x, a),
one may employ the following approach:

1. Proceed to calculate T at point 1 (i.e., T1).

2. Generate a random number between 00 and 99.
3. If the number is between 00 and 24, move to the left. For 25 to 49, 50 to 74, 75 to 99,
move upward, to the right, and downward, respectively.
4. If the move in step 3 results in a new position that is at the outer surface (bound-
ary), terminate the first calculation for point 1 and record the T value of the bound-
ary at that new position. However, if the move results in a new position that is not
at a boundary and is still at one of the other eight interval grid points, repeat steps
2 and 3. This process is continued until an outer surface boundary is reached.
5. Repeat steps 2 to 4 numerous times, for example, 1000 times.
6. After completing step 5, sum all the T values obtained and divide this value by the
number of times steps (2 to 4) have been repeated. The resulting value provides a
reasonable estimate of T1.
7. Return to step 1 and repeat the calculation for the remaining eight grid points.

The results?

T1 = 7.07°C

T2 = 9.80°C
Hazard Risk Assessment 179

T3 = 7.16°C

T4 = 18.67°C

T5 = 25.12°C

T6 = 18.77°C

T7 = 42.93°C

T8 = 52.57°C

T9 = 42.80°C

This procedure can be extended to rectangles; it is not limited to squares.

Illustrative Example 12.6

Maximize the following profit objective function, MC<, for a cement incinerator utiliz-
ing a Monte Carlo approach:5,6

 x −x 
− x6 tanh  3 6 
( )( )
MC = x13 + x23 − x33 + x12 x22 x32 − x 42 x52 x62 + ln ( x1 ) ln ( x2 x3 x 4 ) + e  x1 
; $ (12.19)

The independent variables, x1, x2, and x3 relate to operating conditions while x4, x5, and
x6 relate to health and safety concerns.
The constraints are

4 ≤ x1 ≤ 99

4 ≤ x2 ≤ 99

4 ≤ x3 ≤ 99

4 ≤ x 4 ≤ 99

4 ≤ x5 ≤ 99

4 ≤ x6 ≤ 99

Note that there are (96)6 MC integer values to be potentially checked. Outline a method
of solution.

SOLUTION:
Apply the Monte Carlo method in the solution to the problem. Here is one of the author’s
approaches.7 It involves the reading/sampling of 10,000 (it could be more or less) pos-
sible solutions and selecting the maximum value derived from the objective function
calculations. First, set MC = 0.
180 Introduction to Optimization for Chemical and Environmental Engineers

1. Generate a random 2-digit number from 4 to 99 and assign it to x1.

2. Generate a random 2-digit number from 4 to 99 and assign it to x2.
3. Generate a random 2-digit number from 4 to 99 and assign it to x3.
4. Generate a random 2-digit number from 4 to 99 and assign it to x4.
5. Generate a random 2-digit number from 4 to 99 and assign it to x5.
6. Generate a random 2-digit number from 4 to 99 and assign it to x6.
7. Substitute the previous x1, x2, x3, x4, x5, and x6 values into the objective function
MC and calculate MC.
8. Store (retain) this MC value.
9. Compare this new value of MC with the previous value of MC. Store or retain
the larger of the two.
10. Repeat steps (1–9) 9999 times.
11. The last man (or is it person) standing, that is, the last retained MC is the maxi-
mum value of MC.

The answer in (11) represents the maximum value of the sample of the 10,000 solu-
tions. The “actual” or “true” maximum of the entire population of (96)6 MC’s is prob-
ably slightly higher than the value in (11). Naturally, the more solutions attempted will
improve the estimate provided in (11).
The reader should consider what other approaches could be employed to improve on
the accuracy of the answer provided by the Monte Carlo method.

References
1. L. Theodore and R. Dupont, Environmental Health and Hazard Risk Assessment: Principles and
Calculations, CRC Press/Taylor & Francis Group, Boca Raton, FL, 2012.
2. L. Theodore, Chemical Engineering: The Essential Reference, McGraw-Hill, New York City, NY,
2014.
3. J. Reynolds, J. Jeris and L. Theodore, Handbook of Chemical and Environmental Engineering
Calculations, John Wiley & Sons, Hoboken, NJ, 2004.
4. D. Kaufmann, Process Synthesis and Design, A Theodore Tutorial, Theodore Tutorials, East
Williston, NY, originally published in the USEPA/APTI, RTP, NC, 1996.
5. L. Theodore and C. Porchaska, Introduction to Mathematical Methods for Environmental Engineers
and Scientists, Wiley‑Scrivener, Salem, MA, 2018.
6. S. Shaefer and L. Theodore, Probability and Statistics Applications for Environmental Science, CRC
Press/Taylor & Francis Group, Boca Raton, FL, 2007.
7. L. Theodore, Personal Notes, East Williston, NY, 2017.
 Part III

Chemical Engineering
Applications

Five chapters compliment the presentation in Part III. The titles follow:

Chapter 13: Fluid Flow Applications

Chapter 14: Chemical Reactor Applications
Chapter 15: Mass Transfer Applications
Chapter 16: Heat Transfer Applications
Chapter 17: Plant Design Applications

Note that the examples in Chapter   17 apply to both environmental and chemical
engineering.
Optimization provides important methods for decision-making in chemical engineer-
ing. It has evolved from a methodology of academic interest into a technology that contin-
ues to have a significant impact on both engineering research and practice. Optimization
algorithms include tools for experimental design, model development, and statistical anal-
ysis; process synthesis analysis, design, and retrofit; model predictive control and real-
time optimization; and planning, scheduling, and the integration of process operations.
Part III keys on optimization applications in chemical engineering. The five top-
ics addressed are: fluid flow (Chapter  13), chemical reactors (Chapter  14), mass transfer
(Chapter  15), heat transfer (Chapter  16), and plant design (Chapter  17). Each chapter con-
tains a brief introduction to the subject in question, which is followed by a host of illus-
trative examples keying on industrial applications that employ some— if not all— of the
optimization methods introduced in Part I. As noted in the previous paragraph, the per-
formance criteria can take many forms. For chemical engineering applications, empha-
sis has been placed on process variables, for example, heat exchange area, not equipment
details, for example, baffle spacing, and so on.
182 Introduction to Optimization for Chemical and Environmental Engineers

The reader should note that illustrative examples involving traditional optimization
will address both the development of the describing equation(s), that is, the model, and
the solution to the problem. Linear programming illustrative examples will key solely on
the modeling aspect of the problem and not the solution, which more often than not will
be “ obtained”  from Excel. This statement particularly applies to some of the illustrative
examples at the end of the chapters.
13
Fluid Flow Applications

The first chapter in Part III contains six illustrative examples concerned with the general
subject of fluid flow. The title and one-line descriptions of each example are provided here:

13.1 Fan selection

13.2 Pump section
13.3 Pipe diameter
13.4 Two-stage compressor
13.5 Ventilation models
13.6 Three-stage compressor

The five references that follow provide suggested reading material:

1. R. Perry and D. Green, Perry’s Chemical Engineers’ Handbook, 8th edn, McGraw‑Hill,
New York City, NY, 2008.
2. L. Theodore, Chemical Engineering: The Essential Reference, McGraw‑Hill, New York
City, NY, 2014.
3. J. Reynolds, J. Jeris, and L. Theodore, Handbook of Chemical and Environmental
Engineering Calculations, John Wiley & Sons, Hoboken, NJ, 2004.
4. L. Theodore, R. Dupont, and K. Ganesan, Unit Operations in Environmental
Engineering, Wiley‑Scrivener, Salem, MA, 2018.
5. P. Abuleucia and L. Theodore, Fluid Flow for the Practicing Chemical Engineer, John
Wiley & Sons, Hoboken, NJ, 2009.

13.1 Fan Selection
If a pressure difference is required between two points in a system, a prime mover such
as a fan, pump, or compressor is usually used to provide the necessary pressure and/or
flow impetus. Environmental and chemical engineers are often called on to specify prime
movers more frequently than any other piece of processing equipment, particularly in the
chemical industry. In a general sense, these prime movers are to a process plant what the
engine is to one’s automobile. Whether one is processing petrochemicals, caustic soda,
acids, and so on, the fluid must usually be transferred from one point to another some-
where in the process. At a chemical plant, chemicals must be loaded or unloaded, sent to
heat exchangers or cooling coils, transferred from one processing unit to another, or pack-
aged for shipment.1

183
184 Introduction to Optimization for Chemical and Environmental Engineers

To move material (either a fluid or slurry) through the various pieces of equipment
at a facility (including piping and duct work) requires mechanical energy, not only to
impart an initial velocity to the material, but more importantly, to overcome pressure
losses that occur throughout the flow path. This energy may be imparted to the mov-
ing stream in one or all of three modes: an increase in the stream’s velocity, an increase
in the stream pressure, or an increase in stream height. In the first case, the additional
energy takes the form of increased external kinetic energy as the bulk stream veloc-
ity increases. In the second, the internal energy (mainly potential energy, but usually
some kinetic as well) of the stream increases. This pressure increase may also cause a
stream temperature rise. The term fans and blowers are often used interchangeably, and
no distinction will be made between the two in the following discussions. Whatever
is stated about fans applies equally to blowers. Generally speaking, however, fans are
used for low-pressure (drop) operation, usually below 2 psi. Blowers are employed
when generating pressure heads in the 2.0–14.7 psi range. High-pressure operations
require compressors.
Fans are usually classified as centrifugal or axial-flow types. In centrifugal fans, the gas
is introduced into the center of the revolving wheel (the eye) and discharges at right angles
to the rotating blades. In axial-flow fans, the gas moves directly (forward) through the axis
of rotation of the fan blades. Both types are used in industry, but it is the centrifugal fan
that is employed at most facilities.
The gas in a centrifugal fan is subjected to centrifugal forces. These forces compress the
gas giving it additional static pressure. Centrifugal fans are enclosed in a scroll-shaped
housing that helps convert kinetic energy to static pressure. Gas rotating between the fan
blades is compressed in the fan scroll, which increases the static pressure. Centrifugal fans
are classified by blade configuration as not only radial, forward curved, backward curved,
but also as air foil and radial tip-forward curved heel.
Radial or straight blade fans physically resemble a paddle wheel with long radial blades
attached to the rotor and are the simplest design of all centrifugal fans. This enables most
radial blade fans to be built with great mechanical strength and to be easily repaired.
These fans can be used in a variety of situations, especially heavy-duty applications. This
type of fan can handle erosive and corrosive gases as well as very viscous gases. It is
particularly well-suited for high static pressure operations and can generate pressures in
excess of 50 in H2O. When operated properly, the horsepower efficiency range is 55–69%,
with 65% as a typical value. Forward curved fans are the most popular for general ventila-
tion purposes (high flow rates and low static pressures). These fans have both the heel
and the tip of the blade curved forward in the direction of rotation. Blades are smaller
and spaced much closer together than in other blade designs. They are generally not used
with dirty gases when dust or sticky materials are present because contaminants easily
accumulate on the blades and cause imbalance. Efficiencies range from 52% to 71%, with
65% being typical. Backward curved or backward inclined fans have blades inclined in a
direction opposite to that of the direction of rotation. This feature causes the gas to leave
the tip of the blade at a lower velocity than the wheel-tip speed, a factor that improves the
mechanical efficiency. These types of fans are not suitable for a heavily particulate-laden
gas, sticky material, or abrasive dust. Centrifugal forces tend to build up particulate mat-
ter on the backside of the fan blades. The airfoil is similar to the backward curved fan,
except that the blade has been contoured to increase stability and operating efficiency.
These fans are more expensive to construct than backward curved fans, but they have
lower power requirements. They are rarely used in air pollution control where the gas
must be clean and noncorrosive.1
Fluid Flow Applications 185

Illustrative Example 13.1

The monthly operating cost of four outdated (and in some cases corroded) fans at the
Belmont Detergent plant is $5600. Under a proposed plan from chief engineer Mike Ice,
the installation of the new fans will cost $94,000 with a reduced monthly operating cost,
$2200 due to the increased efficiency. Determine the annualized cost for the four new
fans if their operating life is estimated to be six years. The interest rate may be assumed
to be 4.9%. The capital recovery factor (CRF) or annual payment of a capital investment
can be calculated as follows:

i (1 + i )
n
A
CRF =   = (13.1)
 P i ,n ( 1 + i ) − 1
n

Should Mike Ice replace the fans?

SOLUTION:
The annualized cost for the new fans is determined based on the following input data:
Capital cost = $94,000
Interest, i = 4.9%
Life, n = 6 yrs
For i = 0.049 and n = 6, the CRF is:

0.049 (1 + 0.049 )
6

CRF =
(1 + 0.049 )
6
−1

= 0.1958
The total annualized cost for the fans is therefore

Annualized cost = Installation cost + Operation cost

= ( 0.1958 ) ( $94, 000 ) + ( $2, 200 )(12 ) = $44, 804
Since the cost is lower than the annual cost of $67,200 for the old process, the proposed
plan should be instituted.

13.2 Pump Selection
Pumps are required to transport liquids, liquid-solid mixtures such as slurries and slud-
ges, auxiliary fuel, and so on. Pumps are also needed to transport water to and/or from
such peripheral devices as boilers, quenchers, scrubbers, heat exchangers, distillation col-
umns, and so on.
Pumps may be classified as reciprocating, rotary, or centrifugal. The reciprocating and
rotary types are referred to as positive displacement pumps because unlike the centrifugal
type, the liquid or semiliquid flow is broken up into small portions as it passes through the
pump. These three classes of units are described below.1,2
Reciprocating pumps operate by the direct action of a piston on the liquid contained in a
cylinder. As the liquid is compressed by the piston, the higher pressure forces it through
discharge valves to the pump outlet. As the piston retracts, the next batch of low-pres-
sure liquid is drawn into the cylinder and the cycle is repeated. The piston may be either
186 Introduction to Optimization for Chemical and Environmental Engineers

directly steam driven or moved by a rotating crankshaft through a crosshead. The rate
of liquid delivery is a function of the volume swept out by the piston and the number of
strokes per unit time. A fixed volume is delivered for each stroke, but the actual delivery
may be less because of both leakage past the piston and failure to fill the cylinder when
the piston retracts. The volumetric efficiency of the pump is defined as the ratio of the
actual volumetric discharge to the pump displacement. For well-maintained pumps, the
volumetric efficiency is at least 95%. Reciprocating pumps are used for some applications.
Reciprocating pumps can deliver the highest pressure of any type of pump (20,000
psig); however, their capacities are relatively small compared to the centrifugal pump.
Also, because of the nature of the operation of the reciprocating pump, the discharge flow
rate tends to be somewhat pulsating. Liquids containing abrasive solids can damage the
machine surfaces of the piston and cylinder. Because of its positive displacement opera-
tion, reciprocating pumps can also be used to measure liquid volumetric flow rates.
The rotary pump combines rotation of the liquid with positive displacement. The rotat-
ing elements mesh with elements of the stationary casing in much the same way that two
gears mesh. As the rotating elements come together, a pocket is created that first enlarges,
drawing in liquid from the inlet or suction line. As rotation continues, the pocket of liquid
is trapped, reduced in volume, and then forced into the discharge line at a higher pressure.
Centrifugal pumps are the most widely used in the process industry because of simplic-
ity of design, low initial cost, low maintenance, and flexibility of application. Centrifugal
pumps have been built to move as little as a few gallons per minute against a small head,
and as much as several thousand gallons per minute against a pressure of several hun-
dred pounds force per square inch (psi). In its simplest form, this type of pump consists
of an impeller rotating within a casing. Fluid enters the pump near the center of the rotat-
ing impeller and is thrown outward by centrifugal force. The kinetic energy of the fluid
increases from the center of the impeller to the tips of the impeller vanes. This high veloc-
ity is converted to a high pressure as the fast-moving fluid leaves the impeller and is driven
into slower moving fluid in the volute or diffuser.

Illustrative Example 13.2

Consider the following application. Two centrifugal pumps are being considered for
purchase to handle a flow of 140 gpm with a required pressure increase of 120 psi. Both
are nominal 20-HP pumps, which should be adequate for this service. Pump A costs
less, but pump B operates more efficiently. Piping, installation, and foundation costs
will be the same for either pump. Which pump should be selected?
Data: The pump is expected to operate with a 90% stream factor. (The “stream factor”
is the fraction of time that the plant operates over the course of a year.) The cost of elec-
tricity at the plant is $0.085/kWh. Additional data is provided in Table 13.1.
Plant economic criterion: 3-year payout.

SOLUTION:
The hydraulic horsepower (HHP) required is1,2

TABLE 13.1
Data for Illustrative Example 13.2
Pump Specifications Pump A Pump B
Cost delivered to plant $4,500 $6,000
Efficiency at desired flow conditions 55% 60%
Fluid Flow Applications 187

HHP = ( 0.000583 ) ( q ) ( ∆P ) q [ = ] gpm ∆P [ = ] psi

= ( 0.000583 )(140 )(120 ) (13.2)

= 9.79 HP

The brake horsepower (BHP) for each pump is1,2

BHP = HHP/E f ; E f = pump fractional efficiency (13.3)

9.79
BHPA = = 17.8 HP
0.55

9.79
BHPB = = 16.3 HP
0.60
Converting to kW,

kWA = (17.8 ) ( 0.745 kW/HP ) = 13.3

kWB = (16.3 ) ( 0.745 kW/HP ) = 12.2

The total electrical energy in kWh required for three years becomes

kWA = (13.3 ) ( 3yr )( 8760h/yr ) ( 0.90 ) = 314, 000 kWh

kWB = (12.2 ) ( 3yr )( 8760h/yr ) ( 0.90 ) = 288, 000 kWh

The cost of electricity for three years (the payout period) is

Cost A = ( 314, 000 kWh ) ( $0.085/kWh ) = $26, 700

Cost B = ( 288, 000 kWh ) ( $0.085/kWh ) = $24, 400

The total three-year cost for each pump can now be calculated (see also Table).
Pump B is the more economical choice. Note, however, that the difference is only
about 2%. Small changes in the cost of electricity, the actual pump efficiencies, or the
stream factor could shift the choice to pump A.
In the final design and purchase of equipment for a new plant, hundreds of decisions
such as this are made after the overall process flow design is completed.

TABLE 13.2
Pump Selection Criteria
Cost Pump A Pump B
Purchase cost $4,500 $6,000
Electricity cost $26,700 $24,400
Total $31,200 $30,400
188 Introduction to Optimization for Chemical and Environmental Engineers

13.3 Pipe Diameter Selection

When plants are designed and developed, designers try to minimize the cost of opera-
tions and maximize efficiency. One way this is commonly done in plants is by strategically
picking the piping and pipe fittings to maximize flow and minimize cost. For any given
set of flow conditions, as pipe diameter increases, capital cost correspondingly increases.
However, with increasing pipe diameter, comes decreasing pumping or fan charges.
Conversely, an increase in pressure drop will generally result from a decreased pipe size.
Therefore, an optimum economic pipe diameter must exist. This optimum pipe diameter is
found at the point when the sum of pumping or fan costs and fixed costs is at a minimum.

Illustrative Example 13.3

Consider the following application. Outline how to determine the optimum economic
pipe diameter for a flow system.

SOLUTION:
The usual calculational procedure is as follows:

1. Select a pipe diameter.

2. Obtain the annual operating cost.
3. Obtain the capital equipment cost.
4. Convert the capital cost to an annual basis.
5. Sum the two annual costs in steps 2 and 4.
6. Return to step 1, selecting a different pipe diameter.

The only variable that will appear in the resulting total-cost expressions is the pipe
diameter. The optimum economic pipe diameter can be generated analytically by taking
the derivative of the total annual cost with respect to pipe diameter, setting the result
equal to zero, and solving for the diameter. The derivative operation can be replaced by
a trial-and-error procedure that involves calculating the total cost for various diameters
and simply selecting the minimum. These calculations can be employed to generate
a graph of the form provided in Figure 13.1. In any event, computerized optimization

Total cost
Cost $/yr

Operating cost

Capital cost

Pipe diameter

FIGURE 13.1
Optimum Pipe Diameter.
Fluid Flow Applications 189

procedures and generalized charts are available that enable the practicing engineer to
optimize pipe diameter for numerous applications.
The reader should note that this topic will be revisited in Part IV, Chapter 19.
The reader should also note that tube/pipe size comes into play with heat e­ xchangers –
a topic that is reviewed in Chapter 15. For heat exchangers, as the tube/pipe diameter is
decreased, the heat transferal increases (since the coefficient of heat transfer increases)
somewhat in many cases; hence, the required area decreases. However, the pressure
drop increases, and consequently there is an optimum diameter corresponding to the
minimum sum of fixed/capital charges and power/operating costs. Note that the dif-
ficulty of cleaning the tubes increases with a decrease in diameter, which may result in
higher maintenance costs.

13.4 Two-Stage Compressor
Compressors operate in a similar fashion to pumps and have the same classifications:
rotary, reciprocating, and centrifugal. An obvious difference between the two opera-
tions is the large decrease in volume resulting from the compression of a gaseous stream
compared to the negligible change in volume caused by the pumping of a liquid stream.
Details on the three compressor classifications follow.
Centrifugal compressors are employed when large volumes of gases are to be handled at
low to moderate pressure increases (0.5–50 psi). Rotary compressors have smaller capaci-
ties and can achieve discharge pressures up to 100 psi. Reciprocating compressors are the
most common type used in industry and are capable of compressing small gas flows to
as much as 3500 psig. With specially designed compressors, discharge pressures as high
as 25,000 psig can be reached, but these devices are capable of handling only very small
volumes and do not work well for all gases. For the environmental applications mentioned
earlier – atomizing of liquids for combustion or of venturi scrubber water for gas c­ leaning –
­reciprocating compressors are normally used.
The following equation may be used to calculate compressor power requirements when
the compressor operation is adiabatic and the gas (usually air) follows ideal gas behavior:3,4

( γ − 1)/γ 
 γRT   P2 
Ws =   − 1 (13.4)
 γ − 1   P1  

where:
Ws = compressor work required per lbmol of air
R = 1.987 Btu/(lbmol · °R)
T = air temperature at compressor inlet conditions (°R)
P1, P2 = air inlet and discharge pressures
γ = ratio of the heat capacity at constant pressure to that at constant volume – typi-
cally 1.3 for air

The temperature ratio across a compression stage is3,4

( k − 1)/k
T2  P2 
= ; adiabatic operation (13.5)
T1  P1 
190 Introduction to Optimization for Chemical and Environmental Engineers

( N − 1)/N
T2  P2 
= ; polytropic operation (13.6)
T1  P1 
where:
k = adiabatic exponent, Cp/Cv
N −1 K −1
N = polytropic exponent, =
N KEp
P1, P2 = suction, discharge pressures, respectively psia
T1, T2 = suction, discharge temperatures, respectively °R
Ep = polytropic efficiency, fraction

It is normally assumed that the usual centrifugal compressor is uncooled internally

and thus follows a polytropic path.4 The temperature of the gas is often limited to protect
against temperature excursions. Intercooling can be employed to retain temperatures at
reasonable levels during high overall compression ratio applications.

Illustrative Example 13.4

Consider the two-stage reversible adiabatic compression of a gas from an initial pres-
sure, P1, to a final pressure, P3. If all fixed charges on the compressors are assumed to be
essentially independent of the inter-stage pressure employed, that is, they are constant,
then the optimum operation involves the determination of the inter-stage pressure, P2,
for which the total power requirement is a minimum. For a fixed gas flow, this cor-
responds to the minimum work for the two stages. (Note: For optimum results, inter-
cooling back to the inlet temperature is desired.)4 Obtain the equation describing P2, in
terms of P1 and P3. If the gas enters at T1 and is cooled to T1 between stages, the total
work, E, is given by Equation 13.72,4:

  k −1
 
 k −1
  
 k   P2   k   P3   k   (13.7)
E = NRT   + − 2
 k − 1   P1   P 
1

 

where:
P = pressure, psia
N = lb mol of gas compressed
R = ideal gas constant, (consistent units)
k = ratio of heat capacity at constant pressure to the heat capacity at constant vol-
ume for gas compressed
T = inert gas temperature, °R
E = total work, ft · lbf

Calculate the inter-stage pressure P2 that will minimize the work if P1 = 14.6 psia and
P2 = 79.2 psia.

SOLUTION:
If the dependent variable E is to be a minimum, then the derivative must be zero, that is,

d ( E)  k   k   k  −  k   k − 1   k   k  
 1− k   1  k −1  1− 2 k 

= NRT1   P P − P P  = 0 (13.8)
 k − 1   k − 1   
1 2 3 2
dP2 k  

Fluid Flow Applications 191

Solving for the values of P2 that satisfies this condition gives

 2 k −2   k −1   k −1 
     
P2 k 
= P1 k 
P3  k 
(13.9)

or

 2( k −1)   k −1   k −1 
 
k     
P2 = P1 k 
P3  k 
(13.10)

Solution for P2 yields

P2 = P1P3 (13.11)

If P1 = 14.6 psia and P3 = 79.2 psia,

P2 = (14.6 )( 79.2 ) = 1156 = 34 psia

This approach to the solution is found in most handbooks and thermodynamics

texts.3,4
The location of a maximum or minimum becomes obvious if the function is plotted,
and this graphical procedure may be followed in economic-based calculations. The pre-
viously mentioned analytical method has the advantage of directness and simplicity
in many cases. The formal method involves calculation of the function at a number of
points and requires more time to obtain the result. When the curve is obtained, how-
ever, its shape indicates the importance of operating at the exact optimum. If the curve
is flat, a considerable variation in operating conditions will not appreciably affect the
costs and/or profits. If the maximum or minimum is not flat, it may be quite important
to operate at the optimum point. In some situations, the dependent variable may be a
function of several independent variables. The determination of a maximum or mini-
mum value under such circumstances is reviewed later in this chapter in Illustrative
Example 13.6, where the problem of a three-stage compressor is discussed.
Note: The reader may choose to review Illustrative Examples 4.6 and 6.3 in Part I.

13.5 Ventilation Models
Indoor air pollution is rapidly becoming a major health issue in the United States. Indoor
pollutant levels are quite often higher than outdoors, particularly where buildings are
tightly constructed to save energy. Since most people spend nearly 90% of their time
indoors, exposure to unhealthy concentrations of indoor air pollutants is often inevitable.
The degree of risk associated with exposure to indoor pollutants depends on how well
buildings are ventilated and the type, mixture, and amounts of pollutants in the building.1,2,5
Alternatively, industrial ventilation is the field of applied science concerned with con-
trolling airborne contaminants to produce healthy conditions for workers and a clean
environment for the manufacture of products. To claim that industrial ventilation will
prevent contaminants from entering the workplace is naïve and unachievable. More to
the point, and within the realm of achievement, is the goal of controlling contaminant
exposure within prescribed limits. To accomplish this goal, one must be able to describe
192 Introduction to Optimization for Chemical and Environmental Engineers

the movement of gaseous and particle contaminants in quantitative terms that take into
account

1. The spatial and temporal rate at which contaminants are generated

2. The velocity field of the air in the workplace
3. The spatial relationship between source, workers, and openings through which
air is withdrawn or added
4. Exposure limits (time-concentration relationships) that define unhealthy
conditions1,2

In general, most gaseous and particulate control/recovery equipment is more economi-

cally efficient when handling higher concentrations of contaminants, all else being equal.
Therefore, the gas handling system should be designed to concentrate contaminants in
the smallest possible volume of air. This is important since, exclusive of the fan, the cost of
the control equipment is based principally upon the volume of gas to be handled and not
on the quantity of particulate to be removed. The reduction of emissions by process and
system control is an important adjunct to any cleaning technology.
Although ventilation does not remove the contaminants (particularly nanoagents) from
the workplace, it does provide an opportunity to either recover/control or dilute (into
the atmosphere) any problem emissions. Regarding manufacture and production, some
European nations and unions either require or recommend that closed operations be
employed.
It is important to note that because of the size of nanoparticles, their behavior in a ven-
tilation system will approach that of a gas. Thus, it is safe to conclude that the ventilation
procedures and calculations that are currently employed in practice may be applied to
nanoparticles.
The reader is referred to the work of Heinsohn5 for an excellent treatment of ventila-
tion. In addition, it should be noted that several of the problems to follow were excerpted
and/or adapted from publications6,7 resulting from National Science Foundation (NSF)-
sponsored faculty workshops and a subsequent publication.

Illustrative Example 13.5

Your consulting firm has received a contract to develop, as part of an emergency prepa-
ration plan, a mathematical model describing the concentration of a nanoparticle in a
medium-sized ventilated laboratory room.8 The following information/data (SI units)
is provided:
V = volume of the room, m3
v0 = volumetric flow rate of ventilation air, m3/air
c0 = concentration of the nanochemical in ventilation air, gmol/m3
c = concentration of the nanochemical leaving ventilated room, gmol/m3
c1 = concentration of the nanochemical initially present in ventilated room, gmol/m3
r = rate of disappearance of the nanochemical in the room due to reaction and/or
other effects, gmol/m3‑min

As an authority in the field (having taken the Theodore course on Health, Safety, and
Accident Management), you have been requested to

1. Obtain the equation describing the concentration in the room as a function of

time if there are no “reaction” effects, that is, r = 0.
Fluid Flow Applications 193

2. Obtain the equation describing the concentration in the room as a function of

time if r = −k. Note that the minus sign is carried since the agent is disappearing.

SOLUTION:
Use the laboratory room as the control volume. Apply the concentration law for mass to
the nanochemical:

rate of mass  rate of mass  rate of mass   rate of mass 

 − + =  (13.12)
 in   out   generated  accumulated 
Employing the notation specified in the problem statement gives

rate of mass 
  = v0 c0
 in 

rate of mass 
  = v0 c
 out 

rate of mass 
  = rV
 generated 

 rate of mass  d(cV )

 =
accumulated dt

Substituting the previous gives

d(cV ) (13.13)
v0 c0 − v0 c + rV =
dt

Since the laboratory room is constant, V may be taken out of the derivative term. This
leads to

v0 dc
( c0 − c ) + r = (13.14)
V dt
The term V/v0 represents the average residence time the nanochemicals reside in the
room and is usually designated as τ. The previous equation may then be rewritten as

dc ( c0 − c )
= + r (13.15)
dt τ

1. If r = 0,

dc ( c0 − c )
= (13.16)
dt τ
Separating variables

dc dt
=
( c0 − c ) τ
194 Introduction to Optimization for Chemical and Environmental Engineers

c t
dc dt

∫
ci
( c0 − c )
=
∫τ
0

 c −c  t
ln  0 = τ
 c0 − ci 

t
 c0 − c  − τ
 c −c  = e
 0 i

which ultimately leads to

t t
− −
c = c0 − ( c0 − ci ) e τ = c0 + ( ci − c0 ) e τ (13.17)

2. If r = −k,

dc ( c0 − c ) c c  c − kτ  c
= −k = 0 −k− = 0 − (13.18)
dt τ τ τ  τ  τ

Separating variables,

dc
= dt
 c0 − k τ  c 
 τ  τ  −
  

dc dt
=
( c0 − k τ ) − c  τ

c t
dc dt

∫
ci
( c0 − k τ ) − c 
=
∫τ
0

 ( c0 − k τ ) − c  dt
− ln  =
 ( c0 − k τ ) − ci  τ

( c0 − kτ ) − c =e
t
− 
τ

( c0 − kτ ) − ci
which ultimately leads to

−
t  −  
t
c = ci e τ + ( c0 − k τ ) 1 − e  τ   (13.19)
 
Fluid Flow Applications 195

13.6 Three-Stage Compressor

Illustrative Example 13.6

Compressors, unlike fans and pumps, find only limited and specialized application.
They are primarily employed to increase the pressure of a fluid in some types of sys-
tems, for example, when liquids are broken into tiny droplets (atomized) before enter-
ing a unit. This can be accomplished through the use of a high-pressure stream of air
or steam that impinges on the liquid stream and atomizes it. The pressurizing of the
air or steam is accomplished through the use of compressors. Compressors are also
used for atomization on certain types of air pollution control devices such as venturi
scrubbers, which depend on fine water droplets to remove particulates from the flue
gas stream. These compressors often operate with several stages. The following is an
example involving three stages.
Consider the case of a three-stage compressor. It is desired to calculate the two inter-
mediate pressures P2 and P3 in atm. It can be shown that the energy expended by the
compressor, E, in ft · lbf is given by

  k −1
 
 k −1
 
 k −1
  
 k   P2   k   P3   k   P4   k 
E ( P2 , P3 ) = NRT  +  +  − 3  (13.20)
 k − 1   P1   P2   P3 
 
Discuss the approach that was employed in Part I, to solve this problem if NRT = 1,
k = 1.4, P1 = 1 atm and P4 = 10 atm.

SOLUTION:
The reader is first referred to Illustrative Example 4.6 in Section 4.4. The method of
steepest ascent/descent was employed. The results provided were presented as 3

P2 = 2.154 atm

P3 = 4.642 atm

with

E ( min ) = 2.575 ft ⋅ lbf

The problem was also solved in Illustrative Example 6.7 using analytical means. The
results were essentially identical to those provided in Illustrative Example 4.6. Another
approach that could have been employed to solve the problem was to use any brute
force trial-and-error methods that could have included graphical analysis.
Regarding the analytical solution, the reader should note that for the variable E to be
a minimum, first ∂E/∂P2 and ∂E/∂P3 must be zero. The derivatives are given by

∂E   1−k   −1   k −1   1− 2 k  
   
= NRT1  P1 k  P2 k  − P3  k  P2 k   (13.21)
∂P2  

∂E   1−k   −1   k −1   1− 2 k  
   
= NRT1  P2 k  P3  k  − P4  k  P3  k   (13.22)
∂P3  
196 Introduction to Optimization for Chemical and Environmental Engineers

Setting ∂E/∂P2 and ∂E/∂P3 = 0, leads to

P2 2 = P1P3 (13.23)

and

P3 2 = P2 P4 (13.24)

Therefore

( )
1/3
P2 = P12 P4 (13.25)

1
(
P3 = P4 2 P1 ) 3 (13.26)

If one now differentiates ∂E/∂P2 with respect to P2 and substitutes the values for P2
and P3 from the previous, the following is obtained:

 k − 1    3 k  −  3 k  
 1− 5 k   1+ k 
 ∂ 2E 
 ∂P  = 2 NRT1    P1 P4  (13.27)
2
2
 k   
0

The subscript 0 indicates that the value of ∂ 2E / ∂P2 2 is at the possible minimum that one

is attempting to verify. Since the powers of P1 and P4 must be positive, and k > 1, it is clear
 ∂ 2E 
that,  is greater than zero, a positive quantity. Similarly, by differentiation of
 ∂P2 2  0
Equation 13.21 with respect to P3 and substitution of values from Equation 13.24 and 13.25

 k − 1    3 k  −  3 k  
 1− 3 k   1+ 3 k 
 ∂ 2E 
 ∂P 2  = 2 NRT1    P1 P4  (13.28)
3  k   
0

This, by similar reasoning, is positive. The product of the second derivatives lead to

 NRT1 ( k − 1)  
2 2−8k   2+ 4k 
 ∂ 2E   ∂ 2E  3k 
 −
 3 k 
 ∂P 2   ∂P 2  = 4   P1 P4 (13.29)
2 0 3 0  k 
Similarly, the cross derivative is given by

 NRT1 ( k − 1)  
2 2−8k   2+ 4k 
 ∂ 2E  3k 
 −
 3 k 
 ∂P ∂P  =   P1 P4 (13.30)
2 3 0  k 
Thus, it has also been demonstrated that
2
 ∂ 2E   ∂ 2E   ∂ 2E 
 ∂P 2   ∂P 2  >  ∂P ∂P  (13.31)
2 3  2 3
Finally, it should be noted that considerable accuracy in computation is often required
to establish the optimum. This frequently occurs in applying other methods so that it is
necessary to use an analytical expression to establish the optimum accurately. Another
limitation of the procedure, using the method of steepest descent/ascent previously
discussed and the general analytical differentiation procedure, is that they do not take
into account external constraints imposed on the system. For example, the gradient may
lead into an area in which operations are not possible.
Fluid Flow Applications 197

References
1. I. Farag, Fluid Flow, A Theodore Tutorial, Theodore Tutorials, East Williston, NY, originally
published by the USEPA/APTI, RTP, NC, 1995.
2. P. Abulencia and L. Theodore, Fluid Flow for the Practicing Chemical Engineer, John Wiley & Sons,
Hoboken, NJ, 2009.
3. L. Theodore, Heat Transfer Applications for the Practicing Engineer, John Wiley & Sons, Hoboken,
NJ, 2011.
4. L. Theodore, F. Ricci, and T. VanViliet, Thermodynamics for the Practicing Engineer, John Wiley &
Sons, Hoboken, NJ, 2009.
5. R. Heinsohn, Industrial Ventilation: Engineering Principle, John Wiley & Sons, Hoboken, NJ, 1991.
6. R. Dupont, T. Baxter, and L. Theodore, Environmental Management: Problems and Solutions, CRC-
Lewis Publishers, Boca Raton, 1998.
7. K. Ganesan, L. Theodore, and R. Dupont, Air Toxins: Problems and Solutions, Gordon and Breach,
New York City, NY, 1996.
8. L. Theodore, “Nanotechnology: Basic Calculations for Engineers and Scientists”, John Wiley &
Sons, Hoboken, NJ, 2006.
14
Chemical Reactor Applications

The second chapter in Part III contains six illustrative examples concerned with the gen-
eral subject of chemical reactors (CR). The title and one-line descriptions of each example
are provided here:

14.1 Comparing continuous stirred tank reactor (CSTR) to tubular flow (TF) reactor
performance
14.2 Optimizing fluidized-bed reactor performance
14.3 Two reactors in series: achieving maximum conversion
14.4 Maximizing selectivity
14.5 Optimizing batch reactor performance
14.6 Optimizing operating schedules

The six references that follow are suggested reading for chemical reactors:

1. R. Perry and D. Green, Perry’s Chemical Engineers’ Handbook, 8th edn, McGraw‑Hill,
New York City, NY, 2008.
2. L. Theodore, Chemical Engineering: The Essential Reference, McGraw‑Hill, New York
City, NY, 2014.
3. J. Reynolds, J. Jeris, and L. Theodore, Handbook of Chemical and Environmental
Engineering Calculations, John Wiley & Sons, Hoboken, NJ, 2004.
4. L. Theodore, R. Dupont, and K. Ganesan, Unit Operations in Environmental
Engineering, Wiley‑Scrivener, Salem, MA, 2018.
5. L. Theodore, Chemical Reactor Analysis and Applications for the Practicing Engineer,
John Wiley & Sons, Hoboken, NJ, 2012.
6. L. Theodore, Chemical Reaction Kinetics, A Theodore Tutorial, Theodore Tutorials,
East Williston, NY, originally published by USEPA/APTI, RTP, NC, 1992.

14.1 Comparing a Continuous Stirred Tank Reactor

to a Tubular Flow Reactor Performance
There are certain inherent advantages and disadvantages to using each of the three cat-
egories of reactors: batch reactors (B), CSTRs (C), and (TF) reactors. The advantages and
disadvantages of these three classes of reactor are provided in the literature.1,2 These reac-
tors are sometimes used in combination for any one of a variety of reasons.

199
200 Introduction to Optimization for Chemical and Environmental Engineers

A reaction where mixing is important is the tank flow or (CSTR); it is also referred to as a
“back-mix” reactor. This type of reactor, like the batch reactor discussed in a later section,
essentially consists of a tank or kettle equipped with an agitator. As with the batch reac-
tor, perfect mixing is normally assumed so that both the concentration and temperature
within the reactor is the same, that is, there is no spatial variation so that both terms are
not a function of position within the reactor. This perfect mixing assumption also requires
that both the concentration and temperature at the discharge point (the outlet pipe) is
identical to that in the reactor. Fogler3 notes that for perfect mixing: “The idea that the com-
position is identical everywhere in the reactor and in the exit pipe requires some thought.
It might seem that, since the concentration changes instantly at the entrance where mixing
occurs, reaction occurs there and nothing else happens in the reactor because nothing is
changing. However, reaction occurs throughout the reactor, but mixing is so rapid that
nothing appears to change with time and position.” One can extend this reasoning to also
conclude the same within the reactor and the discharge port.
This class of reactor finds application in industry when

1. The production rate (or quantity) is high.

2. The reaction time is relatively low.

The bottom line is that in addition to producing greater quantities with smaller equip-
ment, most industrial reactors are operated in a continuous rather than batch mode. They
also require less labor and maintenance; however, flow processes are more difficult to start
and stop than their batch counterparts.
As noted previously, the CSTR essentially consists of a tank or kettle equipped with an
agitator. It may be operated under steady or transient conditions. Reactants are fed contin-
uously, and the product(s) are withdrawn continuously. The reactants and products may
be liquid, gas, or solid, or a combination of them all. If the contents are perfectly mixed,
the reactor design problem is greatly simplified for steady conditions because the mixing
results in the aforementioned uniform concentration, temperature, and so on, throughout
the reactor. This means that the rate of reaction is also constant and the describing equa-
tions are not differential and, therefore, do not require integration. In general, CSTRs are
used for liquid phase reactions. High reactant concentrations can be employed with low
flow rates so that conversions approaching 100% can be achieved. However, the overall
economics of the system may be reduced because of low throughput rates.
CSTRs (as well as TF reactors to be addressed shortly) are often connected in series, in
such a manner that the exit stream of one reactor is the feed stream for another reactor.
Under these conditions, it is convenient to define the conversion at any point downstream
in the battery of CSTR reactors in terms of inlet conditions, rather than with respect to any
one of the reactors in the series. The conversion X (or X A) is then the total moles of A that
have reacted up to that point per mole of A fed to the first reactor. However, this definition
should only be employed if there are no side stream withdrawals and the only feed stream
enters the first reactor in the series. The conversion from reactors 1, 2, 3,… in the series are
usually defined as X1, X2, X3, …, respectively, and effectively represent the overall conver-
sion for that reactor relative to the feed stream to the first reactor.
As noted, the CSTR has certain advantages because of the near uniform temperature,
concentration, and so on, that results because of mixing. However, for perfect mixing,
the reaction occurs at a rate determined by the concentration of the discharge (or exit)
stream from the reactor. Since the rate usually decreases with the extent of reaction (con-
version), the CSTR operates in the range between the high reaction rate corresponding to
Chemical Reactor Applications 201

the concentration in the feed stream and the (normally) lower rate corresponding to the
concentration leaving the reactor. The CSTR therefore requires a larger volume than a
tubular reactor to accomplish a given degree of conversion. This statement applies to all
positive reaction orders. The reaction/residence time ratio for a zero order reaction is one;
that is, both perform identically. However, the previous analysis occasionally does not
apply, particularly if the reaction mechanism is complex and/or non-elementary.
The second reactor to be examined is the TF reactor. The most common type involves
the use of the single-pass cylindrical tube; another type is one that consists of a number of
tubes in parallel. Both will be reviewed in more detail subsequently.
The TF reactor can be conceptualized by two different views. The first is to consider a
CSTR (see previous development) of given volume V, and feed rate q (or F) and divide the
volume into a number of much smaller CSTRs in series. In the limit, an infinite number
of these CSTRs with a total volume V reduces to a TF reactor. In other words, the time of
reaction in a flow system may be visualized as the time required for an infinitesimal parcel
of mass to pass through the reactor in an imaginary compartment (a CSTR of differential
volume) that moves with the flowing stream and reacts, producing changes to the tem-
perature, pressure, and composition of the stream. The second approach is to envision a
pipe (or tube) of uniform diameter with reactants and products flowing axially along the
pipe. This simpler perspective is also a TF reactor.
In terms of kinetic behavior, the TF reactor may be viewed as “identical” to a batch reac-
tor. The time t or reaction that can be defined for a batch reactor is now expressed in terms
of τ (the residence time) and represents the length of time necessary for any given incre-
ment of feed to travel the entire length of the reactor.
TF reactors are characterized by a continuous and decreasing concentration of reactants in
the direction of flow. Either horizontal or vertical orientation is common. The reactants are
charged continuously at one end, and the products are removed continuously at the other
end. The unit almost always operates in a steady-state mode. This greatly simplifies design
and predictive calculations. It is a unit that is amendable to automatic control and to experi-
mental work. As noted, the most common type of TF reactor is the single-pass cylindrical
tube. Another type is one that consists of a number of tubes in parallel. These reactor(s)
may be vertical or horizontal. The feed is charged continuously at the inlet of the tube, and
the products are continuously removed at the outlet. If heat exchange with surroundings is
required, the reactor setup includes a jacketed tube and effectively acts as either a double
pipe or a single pass heat exchanger.1 If the reactor is “empty,” a homogeneous reaction –
one phase present – usually occurs. If the reactor contains catalyst particles, the reaction is
said to be heterogeneous. This type will be considered later in this chapter.
TF reactors are usually operated under steady conditions so that physical and chemical
properties do not vary with time. Unlike the batch and tank flow reactors, there is no mechan-
ical mixing. Thus, and as indicated earlier, the state of the reacting fluid will vary from point
to point in the system, and this variation may be in both the radial r and axial z-direction. The
describing equations are then differential, with position as the independent variable.
In the describing equations to follow, the reacting system is assumed to move through
the reactor in plug flow. It is further assumed that there is no mixing in the axial direc-
tion and complete mixing in the radial direction so that the concentration, temperature,
and so on, do not vary through the cross section of the tube. Thus, the reacting fluid flows
through the reactor in an undisturbed plug of mass. Note once again that time for this
hypothetical plug to flow through this type of reactor is the same as the contact or reac-
tion time in a batch reactor. Under these conditions, the form of the describing equations
for batch reactors will also apply to TF reactors.1,2 In a plug flow reactor, the entire feed
202 Introduction to Optimization for Chemical and Environmental Engineers

stream moves with the same radially uniform axial velocity along parallel streamlines.
The entire feed stream, therefore, has the same residence time since there is no mixing
in the axial direction but (effectively) complete mixing radially. From a qualitative point
of view, as the length of the reactor approaches infinity, the concentration of a (single)
reactant approaches zero for irreversible reactions (except zero order) and the equilibrium
concentration for reversible reactions.
In actual practice, TF reactors deviate from the plug flow model because of the veloc-
ity variations in the radial direction. For any of these conditions, the residence time for
annular elements of fluid within the reactor will vary from some minimal value at a point
where the velocity is a maximum, to a maximum value near the wall where the velocity
approaches zero. The concentration and temperature profiles, as well as the velocity pro-
file, are therefore also not constant across the reactor. The describing equations based on
the plug flow assumption are then not applicable.
The design equation for tubular reactors operating in a steady-state mode will be dif-
ferential in form, unlike the algebraic design equation for CSTRs. The describing equation
is of an ordinary differential form for variations along the axial (longitudinal) length of the
reactor. A partial differential equation arises if there are variations along both the axial and
radial directions.
The key assumption with a CTSR is that it is completely mixed. The key assumption
with a TF unit is that plug flow is present, that is, the fluid is completely unmixed along the
reactor length and flows down the tube as a plug since it is perfectly mixed radially. Even
though the plug-flow approximation may not apply rigorously in some applications, the
simplicity of solutions in the limit of perfect plug flow makes it a very useful model from
both an academic and industrial perspective.
Ultimately, the performance of the three different classes of reactor is based on a host of
considerations. They include

1. Conversion
2. Production rate
3. Volume requirement(s)
4. Selectivity (and the effect of unwanted by-products)

Energy considerations, space, power requirements, and so on, also play a role but super-
imposed on all of the previously mentioned is the economics associated with the choice.
Walas4, in an earlier article, provides some general “rules of thumb” regarding reactor
selection. Some of his (unedited) suggestions are noted here:

1. The rate of reaction in every instance must be established in the laboratory, and
the residence time or space velocity and product distribution must be found in a
pilot plant.
2. The optimum proportions of stirred tank reactors are a liquid level equal to tank
diameter; at high pressures, slimmer proportions are more economical.
3. Power input to a homogenous-reaction stirred tank is 0.5–1.5 hp/1000 gal, but
input is three times this amount when heat is to be transferred.
4. CSTR behavior is approached when the mean residence time is 5–10 times the time
needed to achieve homogeneity, which is accomplished by 500–2000 revolutions of
a properly designed stirrer.
Chemical Reactor Applications 203

5. Batch reactions are conducted in stirred tanks for small, daily production rates,
or when the reaction times are long, or when some condition, such as feed rate or
temperature, must be programmed in some way.
6. Relatively slow reactions of liquids and slurries are conducted in continuous
stirred tanks. A battery of four or five units in series is most economical.
7. TF reactors are suited to high production rates at short residence times (seconds
or minutes) and when substantial heat transfer is needed. Jacketed tubes or shell-
and-tube construction are used.
8. For conversions under about 95% of equilibrium, the performance of a five-stage
CSTR unit approaches plug flow.

Illustrative Example 14.1

Consider the following applications. The liquid reaction

A → products

follows the rate law

k1C1A/2
−rA = (14.1)
1 + k 2CA2
where:
k1 = 5 (gmol/L)1/2/h
k2 = 10(L/gmol)2

Initial concentration of A = 0.5 gmol/L

Feed rate of A = 200 gmol/h

1. Find the volume necessary to achieve 60% conversion in a CSTR.

2. Would a larger or smaller TF reactor be required to achieve the same degree of
conversion?
3. Explain the difference in the results of (2).

SOLUTION:
1. The rate of reaction, −rA, in terms of the conversion variable X is

C A = C A 0 (1 − X ) ; X = X A (14.2)

so that
k1C A1/02 (1 − X )
1/2

−rA = (14.3)
1 + k 2C A2 0 (1 − X )
2

The design equation if the reaction is conducted in a CSTR is1,2

FA 0 X
V= ; X = 0.6
− rA
(14.4)
 1 + (10 )(0.5)2 (1 − 0.6)2 
= FA 0 X  
 ( 5)(0.5) (1 − 0.6)1/2 
1/2
204 Introduction to Optimization for Chemical and Environmental Engineers

Further substitution yields

 
 1 + (10 )( 0.5 ) (1 − 0.6) 
2 2
V = ( 200 )( 0.6 )  1 1  ; X = 0.6
 ( 5 )( 0.5 ) 2 (1 − 0.6) 2 
 
= 75.1 liters

2. The design equation if the reaction is conducted in a TF reactor is1,2

X
dX
V = FA 0 −
∫ 0
rA
(14.5)
1 + k 2C A2 0 (1 − X )
X 2
F
= A10/2
k1C A 0 ∫0
(1 − X )
1/2 dX

The integral I may be defined as

1 + k 2C A2 0 (1 − X )
0.6 2

I=
∫
0
(1 − X )1/2
dX

(14.6)
0.6

=
∫ f (X ) dX
0

Simpson’s three-point rule is used to evaluate the integral:1,2

 h
I =    f (0.0 ) + ( 4 ) f (0.3 ) + f (0.6 ) h = 0.3
 3

 0.3 
= ( 3.5) + ( 4 )( 2.659) + 2.214 
 3  

= 1.635
The volume is then

FA 0 I
V=
k1C A1/02 (14.7)
= 92.5 liters
Note that the integral could have been evaluated by any one of several methods.
3. Surprisingly, the CSTR, is the optimum choice since it requires a smaller vol-
ume. This result arises because of the unique nature of the rate expression.

14.2 Optimizing Fluidized-Bed Reactor Performance

Fluidization is the process in which fine solid particles are transformed into a fluid-like
state through contact with either a gas or liquid, or both. Fluidization is normally carried
out in a vessel filled with catalyst solids. The fluid is normally introduced through the
Chemical Reactor Applications 205

bottom of the vessel and forced through the bed. At a low flow rate, the fluid (liquid or
gas) moves through the void spaces between the stationary and solid catalyst particles
and the bed is referred to as fixed. As the flow rate increases, the catalysts begin to vibrate
and move about slightly, resulting in the onset of an expanded bed. When the flow of fluid
reaches a certain velocity, the solid catalysts become suspended because the upward fric-
tional force between the catalyst and the fluid balances the gravity force associated with
the weight of the catalyst. This point is termed minimum fluidization or incipient fluidization
and the velocity at this point is defined as the minimum or incipient fluidization velocity.
Beyond this stage, the bed enters the fluidization state where bubbles of fluid rise through
the solid catalysts, thereby producing a circulatory and/or mixing pattern.5
From a force balance perspective, as the flow rate upward through a packed bed is
increased, a point is reached at which the frictional drag and buoyant force is enough to
overcome the downward force exerted on the bed by gravity. Although the bed is usually
supported at the bottom by a screen, it is free to expand upward, as it will if the velocity is
increased above the aforementioned minimum fluidization velocity. At this point, the cata-
lysts are no longer supported by the screen, but rather are suspended in the fluid and act
and behave as the fluid. The bed is then said to be fluidized. From a momentum or force bal-
ance perspective, the sum of the drag, buoyancy, and gravity forces must be equal to zero.
The terminal settling velocity can be evaluated for the case of flow past one catalyst par-
ticle in the bed. By superimposition, this case is equivalent to that of the terminal velocity
that a catalyst particle would attain flowing through a fluid. Once again, a force balance
can be applied, and empirical data can be used to evaluate a friction (drag) coefficient.1,2
At intermediate velocities between the minimum fluidization velocity and the terminal
velocity, the bed is expanded above the volume that it would occupy at the minimum
value. Note also that above the minimum fluidization velocity, the pressure drop stays
essentially constant.
One of the novel characteristics of fluidized beds is the uniformity of temperature found
throughout the system. Essentially constant conditions are known to exist in both the hori-
zontal and vertical directions in both short and long beds. This homogeneity is due to the
turbulent motion and rapid circulation rate of the solid catalyst particles within the fluid
stream described previously. In effect, excellent fluid-particle contact results. Temperature
variations can occur in some beds in regions where quantities of relatively hot or cold
catalyst particles are present but these effects can generally be neglected. Consequently,
fluidized beds find wide application in industry, that is, oil cracking, zinc coating, coal
combustion, gas desulfurization, heat exchangers, plastics cooling, and fine powder
granulation.

Illustrative Example 14.2

Consider the following application. A fluidized-bed reactor is to be designed to destroy
99.99% of a unique liquid hazardous waste. Based on laboratory and pilot plant studies,
researchers have described the waste reaction by a first-order reversible mechanism.
Their preliminary findings are given here:

k
A↔R A = waste
k′

k = 1.0e −10 ,000/T T = °R

206 Introduction to Optimization for Chemical and Environmental Engineers

k′ = 9.89e −35 ,000/T T = °R

Calculate the bed reaction operating temperature that will minimize the volume of
the reactor and achieve the optimum desired degree of waste conversion.

SOLUTION:
A fluid-bed reactor is best described by a CSTR model.1,2 The rate of reaction, −rA, in
terms of the concentration of A, CA, is1,2

−rA = kC A1 − k′CR1 (14.8)

Since a fluidized-bed incinerator is best described by a CSTR,1,2

V C A1 − C A1
= ; q = volumetric flow rate (14.9)
q −rA

C A 0 − C A1
= (14.10)
kC A1 − k′CR1
where:
V = reactor volume
q = volumetric flowrate of the waste

CA1 and CR1, the outlet concentrations of A and R, respectively, can now be expressed in
terms of CA0. Note that for 99.99% destruction of the waste, A, the conversion variable
X A becomes

X A = 0.9999

Thus,

C A1 = 0.0001C A 0

CR1 = 0.9999C A 0

The design equation for V may now be rewritten in terms of CA0 and the k’s

V C A 0 − 0.0001(C A 0 )
=
q ( k )( 0.0001) C A 0 − ( k′ )( 0.9999 ) CA 0

0.9999(CA 0 ) 0.9999
= =
( 0.0001k − 0.9999k )( CA0 ) 0.0002k − 0.9999k
′
The reaction velocity constants, k and k′, are described by the Arrhenius equation given
in the problem statement. Thus,

k = Ae − E/RT (14.11)

k′ = A′e − E′/RT (14.12)

A = 1.0
Chemical Reactor Applications 207

A’ = 9.89

E
= −10, 000
R

E
= −35, 000
R
To calculate the operating temperature that will require the minimum volume to accom-
plish a conversion of 99.99%, minimize the volume by setting dV/dT = 0 and solving for
the temperature:

  dk   dk′  
d (V / q )   dT  − ( 9999 )  dT  
= ( 0.9999 )       (14.13)
( k − 9999k )
2
dT  ′ 
 
 
Setting

d (V / q )
= 0 (14.14)
dT

dk dk′
= ( 9999 )
dT dT

dk′
( )
≈ 10 4
dT
Since

dk AE − E/RT
= e (14.15)
dT RT 2

dk′ A′E′ − E′/RT

= e (14.16)
dT RT 2
rearranging the previous equation yields

E′ − E
 A E
10 4 =     e RT
 A ′   E′ 
 −35 , 000 ′ − −10 , 000 
( ) ( )
 1.0   −10, 000   
4
10 =  e T
 9.89   −35, 000 
25 , 000
−
346, 150 = e T
T = 1960 °R

= 1500°F
This represents the optimum operating temperature to minimize the volume of the
reactor.
208 Introduction to Optimization for Chemical and Environmental Engineers

14.3 Two Reactors in Series: Achieving Maximum Conversion

The sequence in which reactors are placed can have an impact on the design and/or the
final conversion.

Illustrative Example 14.3

Consider the following application. Your company has two reactors of equal volume,
which it would like to use in the production of a specified product formed by a first-
order irreversible liquid reaction. One reactor is a CSTR and the other is a TF reac-
tor. How should these reactors be hooked up to achieve maximum conversion (see
Figure  14.1)? Justify your answer using concentration (not conversion) terms in the
design equations.

SOLUTION:
Write the design equation for a CSTR in terms of the concentration (not the conversion)
and the residence time τ.1,2

V = FA0 ( X − 0 ) / ( −rA ) (14.17)

= qC A0 ( X − 0 ) / kC A1 (14.18)

Since C A0 X = C A0 − C A1 and V/q = τ.

τ = (C A0 − C A1 )/ kC A1 (14.19)

C A1 = C A0 / (1 + k τ ) (14.20)

Write the design equation for a TF reactor in terms of the concentration (not the conver-
sion) and the residence time τ.

X
 1 

∫
V = FA0  −
0
 rA
 dX (14.21)


X
 1

0
∫
V/q = C A0  −  dX (14.22)
 rA 

(a) q

or

(b) q

FIGURE 14.1
Physical arrangements of CSTR and TF.
Chemical Reactor Applications 209

Since C A0 dX = −dC A

C A1
1  1 
τ=−
k ∫  C
CA0
 dC A
A 
(14.23)

C 
−k τ = ln  A1 
 CA0 
CA1
= exp ( −k τ )
CA0
C A1 = C A0 exp ( −k τ ) (14.24)

Consider the first case (a) – a CSTR followed by a TF reactor – and determine the outlet
concentration C A1 from the CSTR in terms of τ1.

C A1 = C A0 / (1 + k τ1 ) (14.25)

Determine the outlet concentration, C A2 , for the TF reactor. Set τ = τ2 for the TF reactor

C A2 = C A1 exp ( − k τ 2 )
(14.26)
= C A0 exp ( − k τ 2 ) / (1 + τ 2 )
Since τ = τ1 = τ2,

C A2 = C A0 exp ( −k τ ) / (1 + k τ ) (14.27)

Consider the second case (b) – a TF reactor followed by a CSTR – and determine C A1
and C A2.

C A1 = C A0 exp ( −k τ )

C A2 = C A0 / (1 + k τ ) (14.28)

= C A0 exp ( −k τ ) / (1 + k τ )
The previous results are identical; that is, both cases provided the same outlet
conversion.
How would the analysis in the previous example be affected if the volumes of the two
reactors are not equal? One would repeat the previous calculation if the two volumes
are not equal by employing τT and τc, rather than τ. For the first case,

C A2 = C A0 exp ( −k τT ) / (1 + k τC ) (14.29)

For the second case,

C A2 = C A0 exp ( −k τT ) / (1 + k τC ) (14.29)

The results are once again identical.

How would the results of the previous two examples be affected if the reaction was
not first order? The reader should not interpret the previous two results too loosely. For
any order reaction other than first order, the sequence in which the reactors are placed
210 Introduction to Optimization for Chemical and Environmental Engineers

may affect the performance. The reader is left the exercise of verifying this statement;
however, for a zero order or ∞ order reaction, that is, for n equal to either zero or ∞, the
“performance” or degree of conversion would still be identical.

14.4 Maximizing Selectivity
When multiple reactions (either parallel or consecutive) occur, the side or undesired reac-
tions compete with the main or desired reaction; the less predominant the main reaction
is, the smaller the amount of desired product is for a given amount of reactant. Suppose
the following hypothetical parallel reactions occur:

A→B

A → C (14.30)

A→D

Suppose also that, with a starting amount of 10 mol of A, four form the desired product B,
two form the undesired product C, one forms the undesired product D, and three remain
unreacted. While this process produces 4 mol of valuable product, it could have produced
10 if everything went the way a practicing engineer would want it to; that is, if all 10 mol
of A reacted to form 10 mol of B. The ratio of the 4 mol of B actually produced, to the 10 it
potentially could have produced, is called the yield (0.40 or 40%). By definition, the yield of
a reaction is a measure of how much of the desired product is produced relative to how
much would have been produced if only the desired reaction occurred and if that reaction
went to completion. Obviously, the fractional yield must be a number between zero and
infinity. Another definition that has been used by industry is the “ratio of product formed
to the initial (inlet) reactant.” An additional term used in conjunction with multiple reac-
tions is selectivity. Selectivity is a measure of how predominant the desired reaction is rela-
tive to one of the undesired products produced by a side reaction. Obviously, selectivity
becomes more important than conversion if “undesired” side reaction(s) take precedence
over “desired” reaction(s). Other definitions of selectivity that have appeared in the litera-
ture include the popular “ratio of the amount of desired products formed to the amount
of all products formed.” Definitions have also included “rate of production of one product
relative to another product (point selectivity)” and “amount of one product formed relative
to another product (overall selectivity).” In the previously mentioned example, the selec-
tivity of B over C is 2.0 and that of B over D is 4.0. These definitions are employed in the
illustrative example to follow.

Illustrative Example 14.4

Consider the following reactions3

kD
A → D ( desirable ) (14.31)

kU
A → U (undesirable) (14.32)
Chemical Reactor Applications 211

with

rD = kDC Aα1 (14.33)

rU = kU C Aα2 (14.34)

and

−rA = rD + rU (14.35)

= kDC Aα1 + kU C Aα 2 (14.36)

Suggest procedures that can be employed to maximize the point selected, S, where
S = rD/rU.

SOLUTION:
If one desires rD ≫ rU, then the point (or local) selectivity is

rD kD α1 −α2
S= = CA (14.37)
rU kU
If α1 − α2 = a, then

k  a
S= D  CA (14.38)
 kU 
To maximize S, the need to maintain the product of both terms in Equation 14.38 is
high. If a is positive, one needs to maximize CA. This generally means no inert(s), no
product(s), and for gas phase reactions to operate at low temperature and high pressure.
If a is negative, the reverse applies.
The effect of temperature is handled through the Arrhenius equation:

1
kD AD − RT ( ED −EU )
= e (14.39)
kU AU
If ED > EU, the reaction should be run at the highest temperature economically possible.
If ED < EU, it should be run at the lowest temperature possible (while maintaining a rea-
sonable rate).
The reader should also attempt to analyze the following application:

k1
A + B → D (14.40)

k2
A + B → D (14.41)

where:

rD = k1C Aα1CBβ1 (14.42)

rU = k 2C Aα 2CBβ 2 (14.43)
212 Introduction to Optimization for Chemical and Environmental Engineers

For this case

rD  k1  α1 −α2
S= = (
rU  k 2 
CA)( )
CBβ1 −β2 (14.44)

Hint: Consider attempting to maximize the product of CA and CB.

Consider also the following application. In some cases, it may be possible to reduce
the generation of undesirable by-products from chemical reactions by altering the oper-
ating conditions. For example, consider the reaction scheme
k1
A → U (14.45)

k2
A + D ↔ U (14.46)
k3

where D is the desired product and U is an undesired by-product. This problem is also
left as an exercise for the reader.

14.5 Optimizing Batch Reactor Performance

A batch reactor is a solid vessel or container. Batch reactors were unquestionably the first
type of reactor to be employed by industry. A batch reactor may be open or closed. Reactants
are usually added to the reactor simultaneously. The contents are then mixed (if necessary)
to ensure no spatial variations in the concentration of the species present. The reaction then
proceeds. There is no transfer of mass into or out of the reactor during this period. This con-
centration of reactants and products change with time; thus, this is a transient or unsteady-state
operation. The reaction is terminated when the desired chemical change has been achieved.
The contents are then discharged and sent elsewhere, usually for further processing.
The batch reactor is one of the most common types of reactors used in industry. However,
it is almost always the choice of the chemist working in the laboratory where processes of a
chemical nature originate. It is simple, versatile, and easy to operate. Examples at the labo-
ratory level include a test tube or a multi-neck flask. The simplest example at the indus-
trial level is the autoclave. These units are primarily used for small-scale or complicated
processes that manufacture expensive products, that is, pharmaceuticals, dyestuff, and so
on. These units almost always involve low production rates, often a major disadvantage
with the conversion solely a function of reaction time. Production rates can be increased by
either increasing the conversion or scaling up (increasing the size of) the reactor. Shipping
often limits the size of this class of reactor due to transportation problems. Larger units are
therefore generally fabricated on-site.
The batch reactor may operate

1. Isothermally
2. Non-isothermally
3. Adiabatically
4. Any combination of the above
Chemical Reactor Applications 213

In (1), the temperature is maintained constant during the course of the reaction. This con-
dition can be approached by providing sufficient heat exchanger equipment to account
for enthalpy effects arising during reaction. This mode of operation finds its major appli-
cation in laboratory kinetic studies. Most industrial reactors are described by (2). Some
energy in the form of heat is added to or removed from the reactor, but isothermal condi-
tions are not satisfied. For (3), the reactor is insulated to minimize heat transfer between
the reactor contents and the surroundings. Many industrial reactors attempt to operate
in this manner.
As noted previously, batch reactors are commonly used in experimental studies. Their
industrial applications are somewhat limited. They are seldom used for gas phase reac-
tions, for example, combustion, since small quantities (mass) of product are produced with
even a very large-sized reactor. It is primarily used for liquid phase reactions when small
quantities of reactants are to be processed and thus finds limited application in industry.
As a rule, batch reactors are less expensive to purchase but more expensive to operate than
either CSTRs or TF reactors.

Illustrative Example 14.5

Consider the following application. When batch reactors are employed in industry the
times required to charge and empty the reactor can become significant relative to the
reaction time. The time required to complete one sequence of events is called the cycle
time tc and can consist of several terms when applied to chemical reactors:

tc = treact + tfill + tempty + tstandby + theat, cool + …

= treact + tdowntiime (14.47)

= t + td ; t = treact
Expand the previously mentioned development in terms of maximizing the production
rate in a batch reactor.6.

SOLUTION:
For the reactor

aA → pP

the “production” rate, PR, for reactant A is given by

N A0 X A
PR ( A ) = (14.48)
t + td
and is a function of both X and t. Normally the production rate refers to one of the
products formed. This production rate, using the same conversion variable, is then (for
product P)

pN X
PR ( P ) = PR =   A0 A (14.49)
 a  t + td
where p and a are the stoichiometric coefficients in the reaction equation for species P
and A, respectively.
To maximize production, one needs to express PR in terms of X (replace t by the
design equation that contains X) and solve the equation
214 Introduction to Optimization for Chemical and Environmental Engineers

d ( PR )
= 0 (14.50)
dX A
or alternatively

d ( PR )
= 0 (14.51)
dt
For example, if the stoichiometric reaction equation is given by

kA
A → products

then

X A = 1 − e − k At

or

ln (1 − X A )
t=− (14.52)
kA
To maximize P, one can either set

d ( PR )
; X A = 1 − e − k At (14.53)
dt
or

d ( PR ) ln (1 − X A )
; t=− (14.54)
dX A kA
equal to zero and then solve for either X A or t.
One should check to see that the second-order derivative is negative, indicating that
the solution is a maximum and not a minimum.

14.6 Optimizing Operating Schedules

Illustrative Example 14.6
As a recently hired engineer at Rezifp Pharmaceuticals and Chemicals Corp., you have
been assigned the task of developing a weekly operating schedule of the company’s
three chemical producing batch reactors. Reactors A, B, and C, produce x1, x2, and x3
tanks of product respectively, on a weekly basis. Each reactor essentially requires the
same feed materials. Additional information for this system is provided here:

1. There are three major costs associated with each reactor: labor (engineers),
labor (operators), and raw materials.
2. Reactors A, B, and C require 3, 1, 1 engineers per tank of product per week
respectively, but are limited to a maximum of 30 engineers.
Chemical Reactor Applications 215

3. Weekly labor requirements for operators for reactors A, B, and C are 3, 2, and 4
operators per tank of product respectively, but are limited to a maximum of 82
operators.
4. Only 110 tons of raw material is available on a weekly basis that serves as feed
to the reactors. Each tank of product A, B, and C produced require 2, 4, and 5
tons of raw material, respectively.
5. The income, I, of this chemical operation for products A, B, C are 400, 500, and
800 $/tank, respectively. Develop a model that will maximize income.

SOLUTION:
The objective function I is given by:

I ($/week) = 400 x1 + 500 x2 + 800 x3 (14.55)

The constraints are

3 x1 + x2 + x3 ≤ 30 ( engineers ) (14.56)

3 x1 + 2x2 + 4 x3 ≤ 82 ( operators ) (14.57)

2x1 + 4 x2 + 5x3 ≤ 110 ( materials ) (14.58)

with

x1 ≥ 0 (14.59)

x2 ≥ 0 (14.60)

x3 ≥ 0 (14.61)

This problem can be solved using Excel, which gives the weekly outcome of x1 = 3, x2 = 8,
x3 = 15, and I = $17,200.

References
1. L. Theodore, Chemical Reaction Kinetics, A Theodore Tutorial, Theodore Tutorials, East Williston,
NY, originally published by the USEPA/APTI, RTP, NC, 1992.
2. L. Theodore, Chemical Reactor Analysis and Applications for the Practicing Engineer, John Wiley &
Sons, Hoboken, NJ, 2012.
3. S. Fogler, Elements of Chemical Reaction Engineering, 4th edn, Prentice-Hall, Upper Saddle River,
NJ, 2006.
4. S. Walas, Rules of Thumb: Selecting and Diagnosing Equipment, Chem. Eng., New York City, NY,
1987.
5. D. Kunii and O. Levenspieal, Fluidization Engineering, John Wiley & Sons, Hoboken, NJ, 1969.
6. L. Theodore, personal notes, Manhattan College, Bronx, NY, 1967.
15
Mass Transfer Applications

The third chapter in Part III contains six illustrative examples concerned with the general
subject of mass transfer. The titled and one-line description of each example is provided
here:

15.1 Optimum reflux ratio

15.2 Optimizing filter press operation
15.3 Optimizing production rates
15.4 Process and equipment modifications
15.5 Economic analysis for an absorber versus an extractor
15.6 Processing crude oil

Seven references are provided for suggested reading for this chapter:

1. R. Perry and D. Green, Perry’s Chemical Engineers’ Handbook, 8th edn, McGraw‑Hill,
New York City, NY, 2008.
2. L. Theodore, Chemical Engineering: The Essential Reference, McGraw-Hill, New York
City, NY, 2014.
3. J. Reynolds, J. Jeris, and L. Theodore, Handbook of Chemical and Environmental
Engineering Calculations, John Wiley & Sons, Hoboken, NJ, 2004.
4. L. Theodore, R. Dupont, and K. Ganesan, Unit Operations in Environmental
Engineering, Wiley‑Scrivener, Salem, MA, 2018.
5. L. Theodore, F. Ricci, and T. VanVliet, Thermodynamics for the Practicing Engineer,
John Wiley & Sons, Hoboken, NJ, 2009.
6. L. Theodore and F. Ricci, Mass Transfer Operation for the Practicing Engineer, John
Wiley & Sons, Hoboken, NJ, 2020.
7. R. Treybal, Mass Transfer Operations, McGraw-Hill, New York City, NY, 1955.

15.1 Optimum Reflux Ratio

It is no secret that the subject matter of distillation is solely located in the domain of the
chemical engineer. In many ways, it is a topic that separates (no pun intended) chemical
engineering from other engineering and applied science disciplines. As such, the authors
made a conscious decision to provide preferential treatment to distillation and to include
development material that may be lacking in other sections of this chapter.

217
218 Introduction to Optimization for Chemical and Environmental Engineers

Distillation may be defined as the separation of the components of a liquid feed mixture
by a process involving partial vaporization through the application of heat. In general,
the vapor evolved is recovered in liquid form by condensation. The more volatile (lighter)
components of the liquid mixture are obtained in the vapor discharge at a higher concen-
tration. The extent of the separation is governed by two important factors: the properties of
the components involved and the physical arrangement of the unit for distillation.
In continuous distillation, a feed mixture is introduced to a column where vapor rising
up the column is contacted with liquid flowing downward (which is provided by condens-
ing the vapor at the top of the column). This process removes or absorbs the less volatile
(heavier) components from the vapor, thus effectively enriching the vapor with the more
volatile (lighter) components. This occurs in the section above the feed stream, which is
referred to as the enriching or rectification section of the column. The product (liquid or
vapor) removed from the top of the column is rich in the more volatile components and
is defined as the distillate. The section below the feed stream is referred to as the stripping
section of the column. In this section, the liquid is stripped of the lighter components by
the vapor produced in a reboiler at the bottom of the column, called the bottoms, which is
richer in the heavier components.
Distillation columns are used throughout industry when mixtures (primarily in liquid
form) must be separated. One such example is the petroleum industry. In such an applica-
tion, crude oil is fed into a large distillation column and different fractions (oil mixtures of
varying composition and volatility) are withdrawn at different heights in the column. Each
fraction, such as jet fuel, home heating oil, gasoline, and so on, is used by both industry and
the consumer in a variety of ways. The separation achieved in a distillation column depends
primarily on the relative volatilities1,2 of the components to be separated, the number of con-
tacting trays (plates) or packing height, and the ratio of liquid and vapor flow rates.
Distillation columns are rarely designed with packing in large-scale production because
of the liquid distribution problems that arise with large diameter units and the enormity
of the height of many columns. However, where applicable, towers filled with packing are
competitive in cost, and are particularly useful in cases where the pressure drop must
be low, and/or the liquid holdup must be small. Packed towers are occasionally used for
bench-scale or pilot plant work. In contrast, use of trayed towers extends to many areas of
the chemical industry.

Illustrative Example 15.1

The design of a distillation unit is ordinarily based on specifications giving the degree
of separation required for a feed supplied to the unit at a known composition, tempera-
ture, and flow rate. The design engineer must determine the size of column and reflux
ratio necessary to meet the specifications. As the reflux ratio is increased, the number of
theoretical stages required for the given separation decreases.1–3 An increase in reflux
ratio, therefore, may result in lower fixed charges for the distillation column and greater
costs for the reboiler heat supply and condenser coolant. Outline, in graphical form, an
economic analysis of this system.

SOLUTION:
As indicated in Figure 15.1, the optimum reflux ratio occurs at the point where the sum
of fixed charges and operating costs is a minimum. As a rough approximation, the opti-
mum reflux ratio usually falls in the range of 1.1–1.5 times the minimum reflux ratio.
This figure illustrates the general method for determining the optimum reflux ratio in
distillation operations.
Mass Transfer Applications 219

100,000

Total cost = (1) + (2)

80,000

60,000
Annual cost

Steam and cooling - water

cost (1)
40,000

Fixed charges for entire distillation column (2)

20,000
Mini-
mum
reflux
ratio Optimum reflux ratio
0
1.0 1.2 1.4 1.6 1.8 2.0
Reflux ratio

FIGURE  15.1
Optimum reflux ratio in distillation operation.

15.2 Optimizing a Filter Press Operation

Filtration is one of the most common applications of the flow of fluids through packed
beds. As carried out industrially it is similar to the filtration carried out in a chemical labo-
ratory using a filter paper in a funnel. The object is still the separation of a solid from the
fluid in which it is carried, and the separation is accomplished by forcing fluid through a
porous filter. The solids are trapped within the pores of the filter and (primarily) build up
as a layer on the surface of this filter. This fluid, which may be either gas or liquid, passes
through the bed of solids and through the retaining filter.
As noted above, solid particles are removed from a slurry in a filtration process by pass-
ing it through a filtering medium. (According to Webster, a slurry is defined as a “watery
mixture of insoluble matter such as mud, lime or plaster of Paris.”) The solids are depos-
ited on the filtering of a solid nature, which is normally referred to as the filter mixture of a
fluid, and particles of a solid are separated by the filter medium that permits the flow of
the fluid but retains the particles of the solid. Therefore, it primarily involves the flow of
fluids through porous media.
In all filtration processes, the mixture of slurry flows as a result of some driving force, for
example, gravity, pressure (or vacuum), or centrifugal force. In each case, the filter medium
supports the particles as a porous cake. This cake, supported by the filter medium, retains
the solid particles in the slurry with successive layers added to the cake as the filtrate
passes through the cake and medium.
The several procedures for creating a driving force on the fluid, the different methods
of cake deposition and removal, and the different means for removal of the filtrate from
the cake subsequent to its formation, result in a great variety of filter equipment. In
general, filters may be classified according to the nature of the driving force supporting
filtration.
220 Introduction to Optimization for Chemical and Environmental Engineers

There are various types of filtration equipment employed by industry. Included in

this list are gravity filters, plate-and-frame filters, leaf filters, and rotary vacuums filters.
Gravity filters are the oldest and simplest type. These filters consist of tanks with perfo-
rated bottoms filled with porous sand through which the slurry passes. The most com-
mon type, however, is the plate-and-frame filter where plates are held rigidly together in
a frame. Leaf filters are similar to the plate-and-frame filters in that a cake is deposited
on each side of the leaf and the filtrate flows to the outlet in the channels provided by
the coarse drainage screen in the leaf between the cakes. The leaves are immersed in the
slurry. Rotary vacuum filters are used where a continuous operation is desirable, particu-
larly for large-scale operations. The filter drum is immersed in the slurry where a vacuum
is applied to the filter medium that causes the cake to deposit on the outer surface of the
drum as it passes through the slurry.
The plate-and-frame filter press is perhaps the most widely used type of filtering devices
in the chemical industry. Plate-and-frame filter presses are used in a wide variety of indus-
tries. The chemical industry has used the filter press in order to separate the solid portion
of slurry from the liquid. A chemical, for example, zinc, builds up on the frames. The filter
press is then opened and the wet cake, containing solid zinc, can be collected, removed and
dried. The pharmaceutical industry also uses the filter press in similar applications. The
solids collected on the inside of the frames can be dried and later sold as medication. The
sugar industry also employs the filter press to separate solid sugar from solution. A slurry
is sent through a filter press and the solid cake is collected on the frames. This solid is later
dried, crystallized, and sold as sugar. In the pottery industry, the solid cake is collected in
the filter press. This cake is then used for the production of various pottery products. The
filter press is also used in the w ­ astewater treatment industry. A waste stream containing
sludge is passed through a filter press. Once the solids are removed, a smaller volume of
“liquid” can be disposed of.

Illustrative Example 15.2

Consider the following application. Tests with a plate-and-frame filter press, operated at
constant pressure, have shown that the relation between the volume of filtrate delivered
and the time in operation can be represented as follows:

( )
V f2 = 2.25 × 10 4 ( t f + 0.11) (15.1)

where Vf = cubic feet of filtrate delivered in filtering time t hours.

The cake formed in each cycle must be washed with an amount of water equal to
one-sixteenth times the volume of filtrate delivered per cycle. The washing rate remains
constant and is equal to one-fourth of the filtrate delivery rate at the end of the filtra-
tion. The time required per cycle for dismantling, dumping, and reassembling is six
hours. Under the conditions where the preceding information applies, determine the
total cycle time necessary to permit the maximum output of filtrate during each 24-hour
cycle.

SOLUTION:
Let tf = hours of filtering time per cycle and filtrate delivered per cycle = Vf,cycle.

( )
1/2
V f , cycle = 150 t f + 0.11 ; ft 3 (15.2)
Mass Transfer Applications 221

The rate of filtrate delivery at the end of each cycle is

1
dV f 150
Vw × 4 =
dt f
=
2
( t f + 0.11) 2 ; ft 3 (15.3)

where Vw = washing rate; ft3/h

volume of wash water

Time for washing =
washing rate
1 (15.4)
( 4)( 2)(150)(t f + 0.11) 2
= 1
(16)(150)(t f + 0.11)
−
2

t f + 0.11
= ; h (15.5)
2

Total time per cycle = t f +

( t f + 0.11) + 6 = 1.5t + 6.06; h (15.6)
f
2

24
Cycles per 24 h = (15.7)
1.5t f + 6.06

The volume of filtrate delivered every 24 hours is

1  
24
( )
V f , cycle ( cycles per 24 h ) = 150 t f + 0.11 2   (15.8)
 1.5t f + 6.06 

At the optimum cycle time giving the maximum output of filtrate per 24-hour,

d ( volume of filtrate delivered 24 h )

(15.9)
dt f

Performing the differentiation4 (left as an exercise) and solving for tf ultimately yields

t f ,opt = 3.8 h

15.3 Optimizing Production Rates

Basic maximization principles can be applied when determining the most favorable con-
ditions in the operation of a process. One of the most important variables in any pro-
cess operation is the amount of product produced per unit of time. The production rate
depends on many factors, such as the number of hours in operation per day, per week, or
per month; the stress placed on the equipment; and, the sales market available. From an
analysis of the costs involved under different situations and consideration of other factors
affecting the particular process, it is possible to determine an optimum rate of production or
a so-called economic lot size.
222 Introduction to Optimization for Chemical and Environmental Engineers

Design engineers create process designs based on the production capacity for
the plants. However, the optimum rate of production may vary from the “designed
capacity” once the plant is put in operation. For instance, suppose a process has been
designed with a designed capacity of one batch of production every six hours. After
the plant has been put into operation, it has been determined that the total produc-
tion per month can be increased if the time of production of each batch is reduced.
However, if the production time per batch is decreased, more labor hours are required,
per cent conversion of raw materials is reduced, and steam and power costs increase.
While the engineer may have maximized the total production per month at the time
of design, this obvious example demonstrates how price and market conditions con-
stantly change and the process operation can often be improved over time. The follow-
ing analysis indicates the general method for determining economic production rates
and lot sizes.
The total product cost per unit of time is generally divided into the two classifications of
operating costs and organization costs. Operating costs depend on the rate of production and
include expenses for direct labor, raw materials, power, heat, supplies, and similar items,
which are a function of the amount of material produced. Organization costs are due
to expenses for directive personnel, physical equipment, and other services or facilities,
which must be maintained irrespective of the amount of material produced. Organization
costs are independent of the rate of production.4
It is convenient to consider operating costs on the basis of one unit of production. When
this is done, the operating costs can be divided into two types of expenses as follows:

1. Minimum expenses for raw materials, labor, power, and so on that remains con-
stant and must be paid for each unit of production as long as any amount of mate-
rial is produced
2. Extra expenses due to increasing the rate of production

These extra expenses4 are known by some as superproduction costs. They can become par-
ticularly important at high rates of production. Examples of superproduction costs are
extra expenses caused by overload on power facilities, additional labor requirements, or
decreased efficiency of conversion.

Illustrative Example 15.3

Consider the following application. A plant produces membranes1,2 at the rate of P units
per day generating income of $173/membrane. The variable costs per membrane have
been found to be $47.73 + 0.1P1.2. The total daily operation fixed charges are $1750, and all
other expenses are constant at $7325 per day. Determine the daily profit at

1. A production schedule giving the minimum cost per membrane

2. A production schedule giving the maximum daily profit

SOLUTION:
1. The total cost per membrane is given by

1750 + 7325
cTm = 47.73 + 0.1P1.2 + ; $ membrane (15.10)
P
Mass Transfer Applications 223

The production schedule for minimum cost per membrane is

dcTm 9075
= 0 = 0.12P 0.2 − 2 ; number of membranes day (15.11)
dP P
Solution for P gives

9075
P 2.2 =
0.12
P = 165 membranes day
The daily profit at a production schedule for minimum cost per membrane is DP ($/day),
where DP is

 1750 + 7325  (15.12)

DP =  173 − 47.73 − 0.1P1.2 −  P
 P

Therefore, the daily profit for the minimum cost per member is (with P = 165)

 9075 
DP (min ) = 173 − 47.73 − 0.1(165) −
1.2
165
 165 
= $4035 $ day

2. The production schedule for maximum profit per day

d ( DP )
= 0 = 125.27 − 0.22P1.2 (15.13)
dP
Solving Equation 15.13 yields

P = 198membranes per day for maximum daily profit

The daily profit at production schedule for maximum daily profit is therefore

 9075 
DP (max ) = 173 − 47.73 − 0.1(198 ) −
1.2
198
 198 
= $4439 $ day

15.4  Process and Equipment Modifications 5

Source/waste reduction at the process-bath level can be established by material substitu-
tion, bath life extension, and/or drag-out reduction. Material substitution options include
modifying the chemistry of the process baths or replacing the chemicals used for a given
process. Deionized water can be used to replace tap water for process bath makeup and
rinsing operations. Natural contaminants found in tap water (e.g., carbonates, phosphates,
etc.) can reduce the rinse water efficiency, minimize the potential for drag-out recovery,
and increase the frequency of changing the process bath.
224 Introduction to Optimization for Chemical and Environmental Engineers

The life of a cleaning bath can be extended by proper and regular bath maintenance (i.e.,
regular replenishing, pH measurement, metal content measurement, etc.). The life of a plating
bath can also be extended through bath treatment that removes metal contaminants. Copper
is a common metal contaminant that builds up in plating baths. Copper can be removed from
a bath by “dummying.” The dummying process is based on the principle that the copper can
be plated at a low electrical current. When the copper content of a process bath becomes too
high, an electrolyte panel is placed in the bath and a low current (1–2 amperes per square foot)
is then run through the system. At this low current, only the copper in the bath will be plated
out on the panel and other bath additives will remain unaffected by the current.
Process chemical loss as a result of drag-out is the most significant source of chemicals pres-
ent in the wastewater. The factors that contribute to drag-out are the workplace size and shape,
bath viscosity, bath concentration, surface tension, and temperature. The available techniques
of reducing process chemical drag-out are summarized in the following guidelines:

1. Maintain the minimum bath chemical concentration acceptable within operating

range.
2. Maximize bath operating temperature to lower solution viscosity.
3. Reduce surface tension by using wetting agents in the process bath.
4. Maintain proper racking orientations to achieve the best possible drainage.
5. Withdraw workpieces at slow rates and allow for sufficient drainage time.
6. Use a spray rinse above the process tank.
7. Use drainage boards between process baths and rinse tanks to collect and return
drippage to the process baths.
8. Use drag-out tanks to recover the process chemicals for reuse in the process baths.

Illustrative Example 15.4

A nickel electroplating line uses a dip-rinse tank to remove excess plating metals from
the parts. Currently, a single tank is used that requires R gal/h of fresh rinse water to
clean F parts/h (see Figure 15.2). Assume the cleaning is governed by the following
equilibrium relation:

f1 ounces of metal residue part

λ= = (15.14)
r1 ounces of metalresidue galbath

1. Calculate the reduction in rinse water flowrate (a pollution prevention mea-

sure) if a two-stage countercurrent rinse tank is used (as compared to the sin-
gle-stage unit), and 99% of the residue must be removed. Assume the drag-out
volume is negligible.
2. Discuss optimization implications for further wastewater treatment/reuse.

finF f1F
1
r1R rinR

FIGURE  15.2
Flow diagram for one-stage operation.
Mass Transfer Applications 225

SOLUTION:
1. Using the flow diagram provided in Figure 15.2, a material balance for the
residue may be written

f in F + rin R = f1F + r1R (15.15)

This equation may be rearranged in terms of λ. For the single-stage operation, set i = 1:

R ( f in − f1 ) λ ( f in − f1 )
= = (15.16)
F r1 f1

The fraction of residue removed, x, is

x=
( fin − f1 ) (15.17)
f in

Now R/F may be expressed in terms of λ and x:

R  x 
=λ  (15.18)
F  1 − x 

The flow diagram for a two-stage (stages 1 and 2) countercurrent operation is provided
in Figure 15.3. Material balances on the residue for each stage are

stage 1: f in F + rin R = f1F + r1R (15.19)

stage 2: f1F + r1R = f 2 F + r2 R (15.20)

Each of these equations may be solved for R/F/ in terms of λ and f1 using the defining
equation for λ:

R λ ( f in − f1 )
stage 1: = (15.21)
F f1 − f 2

R λ ( f1 − f 2 )
stage 2: = (15.22)
F f2
If one substitutes x into the mass balance expressions,

R f in − f1
stage 1: = (15.23)
F f1 − fin (1 − x )

finF f1F f 2F

r2R r1R rinR

FIGURE  15.3
Flow diagram for two-stage countercurrent operation.
226 Introduction to Optimization for Chemical and Environmental Engineers

R f1 − f in (1 − x )
stage 2 : = (15.24)
F f in (1 − x )
The right-hand sides (RHS) of stages 1 and 2 equations may be set equal to each other:

f in − f1 f1 − f in (1 − x )
= (15.25)
f1 − f in (1 − x ) f in (1 − x )
This equation may be rearranged to obtain a quadratic equation in f1:

f12 − (1 − x ) f in  f1 −  x (1 − x ) f in2  = 0 (15.26)

Solving for f1 yields


{ } 
0.5
2
 f in (1 − x ) ±  − f in (1 − x )  − 4 (1) x (1 − x ) f in
2

f1 =   (15.27)
2
This can be rewritten as

 f 
f1 =  in  1 − x ± ( 3 x + 1) (1 − x )   (15.28)
0.5

 2     
The following expression results if (1 − x) is factored out:

 f in (1 − x )    3 x + 1  0.5 
f1 =   1 ±  1 − x   (15.29)
 2   

Only the “+” term is physically reasonable; therefore,

 f in (1 − x )    3 x + 1  0.5 
f1 =   1 +  1 − x   (15.30)
 2   
The term f1 may be substituted into the stage 2 mass balance equation. Equation 15.31
results:

 f in (1 − x )    3 x + 1  0.5  
λ   1 +    − f2 
R  2    1 − x   
=  (15.31)
F f2

However, f2 can be expressed in terms of x and fin:

f 2 = f in (1 − x ) (15.32)

This can be substituted into the previous equation to yield

 f in (1 − x )    3 x + 1  0.5  
λ   1 + 
   − f in (1 − x )
R  2   1− x  
=  (15.33)
F f in (1 − x )
Mass Transfer Applications 227

which reduces to

R  λ   3 x + 1  
0.5
=     − 1 (15.34)
F  2   1 − x  
The rinse water requirements for both a single and two-stage countercurrent unit
becomes

 0.99 
Single − stage : R = λ  F = 99.0lambaF (15.35)
 1 − 0.99 

  
 λ    3 (0.99) + 1 
0.5

Two-stage : R =     − 1  F = 9.46lambaF (15.36)

 2    1 − 0.99  
  
The rinse water flowrate reduction is therefore

99.0λF − 9.46λF
= 0.904 = 90.4% reduction (15.37)
( 99.0λF )
2. For fixed residue removal, the total mass of metals in the rinse water will be the
same; with reduced water duty, the metals concentration increases. A smaller
water volume now needs to be processed (or sewered). Alternatively, if the
metals concentration is sufficiently high, it can be returned to the plating bath
(for reuse or can be recovered) to further optimize performance.

15.5 Economic Analysis for an Absorber versus an Extractor

Two mass transfer operations of importance to both the environmental and chemical engi-
neers are absorption and extraction. Brief descriptions of both operations follow.
The removal of one or more selected components from a gas mixture by absorption is an
important operation in environmental and chemical engineering. The process of absorp-
tion conventionally refers to the intimate contacting of a mixture of gases with a liquid
so that part of one or more of the constituents of the gas will dissolve with a liquid. The
contact usually takes place in some type of packed or plate column.
Since gas absorption is concerned with the removal of one or more species from a gas
stream by treatment with a liquid, necessary information includes the solubility of these
constituents in the absorbing liquid. In gas absorption operations, the equilibrium of inter-
est is that between a non-volatile absorbing liquid (solvent) and a solute gas. The solute is
ordinarily removed from its mixture in a relatively large amount of a carrier gas that does
not dissolve in the absorbing liquid.1–3
The engineering design of gas absorption equipment must be based on a sound applica-
tion of the principles of diffusion equilibrium and mass transfer. The main requirement
in equipment design is to bring the gas into intimate contact with the liquid, that is, to
provide a large interfacial area and a high intensity of interface renewal, and to minimize
resistance and maximize the mass transfer driving force. The contacting of the phases can
be achieved in many different types of equipment, the most important of which are either
packed or plate columns. The final choice often rests with the various criteria that may
228 Introduction to Optimization for Chemical and Environmental Engineers

have to be met. For example, if the pressure drop through the column is large enough such
that horsepower costs become significant, a packed column may be preferable to a plate-
type column because of the lower pressure drop. Again, primary emphasis in this section
is placed on packed and plate columns.1–4
In most processes involving the absorption of gaseous constituents from a gas stream,
the gas stream is the process fluid; hence, its inlet conditions (flow rate, composition, and
temperature) are usually known. The objectives, then, in the design of an absorption col-
umn, are the determination of the solvent (liquid) flow rate and the calculation of the prin-
cipal dimensions of the equipment (column diameter and height).
The general design procedure consists of a number of steps that have to be taken into
consideration (details of which follow shortly). These include1,2

1. Solvent selection.
2. Equilibrium data evaluation.
3. Estimation of operating data (usually obtained from a mass and energy balance,
where the energy balance determines whether the absorption process can be con-
sidered isothermal or adiabatic).
4. Column selection (should the column selection not be obvious or specified, calcu-
lations must be carried out for the different types of columns, and the final selec-
tion based on economic considerations).
5. Calculation of column diameter (for packed columns that is usually based on
flooding conditions, and for plate columns it is based on the optimum gas velocity
or the liquid handling capacity of the plate).
6. Estimation of the column height or the number of plates (for a packed column, the
column height is obtained by multiplying the number of transfer units, obtained
from a knowledge of equilibrium and operating data by the height of a transfer
unit; for plate columns, the number of theoretical plates, often determined from
the plot of equilibrium and operating lines, is divided by the estimated overall
efficiency to give the number of actual plates, which in turn allows the column
height to be estimated from the plate spacing).
7. Determination of pressure drop through the column (for packed columns, correla-
tions dependent on packing type, column operating data, and physical properties
of the constituents involved need to be available to estimate the pressure drop
through the packing; for plate columns, the pressure drop per plate is obtained
(often estimated) and multiplied by the number of plates).

In gas absorption operations, the choice of a particular solvent is also important.

Frequently, water is used as it is inexpensive and plentiful, but the following properties
must also be considered:

1.
Gas solubility: A high gas solubility is desired since that increases the absorption
rate and minimizes the quantity of solvent necessary; generally, a solvent of a
chemical nature similar to that of the solute to be absorbed will provide good
solubility.
2.
Volatility: Low solvent vapor pressure is desired since the gas leaving an absorp-
tion unit is ordinarily saturated with the solvent and much will therefore be lost.
3.
Corrosiveness.
Mass Transfer Applications 229

4.
Cost (particularly for solvents other than water).
5.
Viscosity: Low viscosity is preferred for reasons of rapid absorption rates,
improved flooding characteristics, lower pressure drops, and good heat transfer
characteristics.1,2
6.
Chemical stability: The solvent should be chemically stable and, if possible,
non-flammable.
7.
Toxicity.
8.
Low freezing point: If possible, a low freezing point is favored since any solidifica-
tion of the solvent in the column could prove disastrous.

Once the solvent is specified, the choice (and design) of the absorption system may be
determined.
Extraction is a term that is used for any operation in which a constituent of a liquid or a
solid is transferred to another liquid (the solvent). The term liquid–liquid extraction describes
the processes in which both phases in the mass transfer process are liquids. The terms
solid–liquid extraction is restricted to those situations in which a solid phase is present and
includes those operations frequently referred to as leaching, lixiviation, and washing. These
terms are subsequently used interchangeably.
Extraction involves the following two steps: contact of the solvent with the liquid or solid
to be treated so as to transfer the soluble component (solute) to the solvent, and separation
or washing of the resulting solution. The complete process may also include a separate
recovery procedure involving the solute and solvent; this is normally accomplished by
another operation, such as evaporation, distillation, or stripping. Thus, the streams leav-
ing the extraction system usually undergo a series of further operations before the fin-
ished product is obtained; either one or both solutions may contain the desired material.
In addition to the recovery of the desired product or products, recovery of the solvent for
recycling is also often an important consideration.
In practice, the manner and the equipment in which these operations are carried out are
based on the difference in physical states. Because solids are more difficult to handle and
do not readily lend themselves to continuous processing, leaching is commonly accom-
plished in a batch-wise fashion by agitating the crude mixture with the leaching agent and
then separating the residual insoluble from the resultant solution. Liquid–liquid extraction
may also be carried out in a batch operation. The ease of moving liquids, however, makes
liquid extraction more amenable to continuous flow in various types of columns and/or
stages.
Liquid–liquid extraction is used for the removal and recovery of primarily organic sol-
utes from aqueous and non-aqueous streams. Concentrations of solute in these streams
range from either a few hundred parts per million to several mole mass percent. Most
organic solutes may be removed by this process. Extraction has been specifically used
in removal and recovery of phenols, oils, and acetic acid from aqueous streams, and in
removing and recovering freons and chlorinated hydrocarbons from organic streams.
Treybal6 has described the liquid–liquid extraction process in the following manner.
“If an aqueous solution of acetic acid is agitated with a liquid such as ethyl acetate, some
of the acid but relatively little water will enter the ester phase. Since the densities of the
aqueous and ester layers are different at equilibrium, they will settle on cessation of
agitation and may be decanted from each other. Since the ratio of acid to water in the
ester layer is now different from that in the original solution and also different from that
in the residual water solution, a certain degree of separation has occurred. This is an
230 Introduction to Optimization for Chemical and Environmental Engineers

example of stage-wise contact and it may be carried out either in a batch or continuous
fashion. The residual water may be repeatedly extracted with more ester to additionally
reduce the acid content.” As will be discussed shortly, one may arrange a countercurrent
cascade of stages to accomplish the separation. Another possibility is to use some sort
of countercurrent or crosscurrent continuous-contact device where discrete stages are
involved.
The solution whose components are to be separated is the feed to the process. The
feed is composed of a dilutant and solute. The liquid contacting the feed for purposes of
extraction is referred to as the solvent. If the solvent consists primarily of one substance
(aside from small amounts of residual feed material that may be present in a recycled
or recovered solvent), it is called a single solvent. A solvent consisting of a solution of one
or more substances chosen to provide special properties is a mixed solvent. The solvent-
lean, residual feed solution, with one or more constituents removed by extraction, is
referred to as the raffinate. The solvent-rich solution containing the extracted solute(s) is
the extract.
The degree of separation that arises because of the aforementioned solubility difference
of the solute in the two phases may be obtained by providing multiple-stage countercur-
rent contacting and subsequent separation of the phases, similar to a distillation operation.
In distillation, large density differences between the gas–liquid phases are sufficient to
permit adequate dispersion of one fluid in the other as each phase moves through the col-
umn. However, in liquid extraction, the density differences are significantly smaller and
mechanical agitation of the liquids is frequently employed at each stage to increase contact
and to increase the mass transfer rates.
The minimum requirement of a liquid extraction unit is to provide intimate contact
between two relatively immiscible liquids for the purposes of mass transfer of con-
stituents from one liquid phase to the other, followed by the aforementioned physi-
cal separation of the two immiscible liquids. Any device or combination of devices
that accomplishes this is defined in this text as a stage. If the effluent liquids are in
equilibrium, so that no further change in concentration would have occurred within
them after longer contact time, the stage is considered a theoretical or ideal stage. The
approach to equilibrium stage provides a mechanism by which two immiscible phases
intimately mix until equilibrium concentrations are reached and then physically sep-
arated into clear layers. A multi-stage cascade is a group of stages usually arranged
in a countercurrent flow between stages for the purpose of enhancing the extent of
separation.

Illustrative Example 15.5

Consider the following application. Doyle Financial Consultation has been hired to con-
duct an optimization study involving two mass transfer devices. Plans are underway
to construct and operate some type of mass transfer unit that will serve to purify a
product stream from a chemical reactor. The company is still undecided as to whether
to install an absorber or an extraction unit. The extraction unit is less expensive to pur-
chase and operate than a comparable absorber system, primarily because of manpower
costs. However, projected income from the absorber unit is higher since it will handle
a larger quantity of process liquid and provide a purer product. Based on the economic
and financial data given in Table 15.1, select the mass transfer unit that will yield the
higher annual profit. Calculations should be based on an interest rate of 12% and a pro-
cess lifetime of 12 years for both units.
Mass Transfer Applications 231

TABLE  15.1
Economic Data Extraction vs. Absorber Comparison
Costs/Credits Extraction Absorber
Mass transfer unit $750,000 $800,000
Peripherals $1,875,000 $2,175,000
Total capital $2,625,000 $2,975,000
Installation $1,575,000 $1,700,000
Operation $400,000/yr $550,000/yr
Maintenance $650,000/yr $775,000/yr
Income $2,000,000/yr $2,500,000/yr

SOLUTION:
Calculate the capital recovery factor (CRF):

( 0.12 ) (1 + 0.12 )
12

CRF =
(1 + 0.12 ) − 1
12
(15.38)
= 0.1614
Determine the annual capital and installation costs for the extraction (EXT) unit:

COST ( EXT ) = ( 2, 625, 000 + 1, 575, 000 )( 0.1614 )

= $677 , 880 yr
Determine the annual capital and installation costs for the absorber (ABS) unit:

COST ( ABS ) = ( 2, 975, 000 + 1, 700, 000 )( 0.1614 )

= $754, 545 yr
See Table 15.2 for a comparison of costs and credits for both devices.

Finally, calculate the profit for each unit on an annualized basis:

PROFIT (EXT ) = 2, 000, 000 − 1, 728, 000 = + 272, 000 $ yr

PROFIT ( ABS ) = 2, 500, 000 − 2, 080, 000 = + 420, 000 $ yr

An absorber unit should be selected based on the previous economic analysis.

TABLE  15.2
Cost Analysis for Illustrative Example 15.5
Extraction Absorber
Total installed ($/yr) 678,000 755,000
Operation ($/yr) 400,000 550,000
Maintenance ($/yr) 650,000 775,000
Total annual cost ($/yr) 1,728,000 2,080,000
Income credit ($/yr) 2,000,000 2,500,000
232 Introduction to Optimization for Chemical and Environmental Engineers

The reader should note that detailed cost estimates are beyond the scope of this text.
Such procedures are capable of producing accuracies in the neighborhood of ±5%.
However, such estimates generally require many months of engineering work. This type
of analysis is designed to give the reader a basis for a preliminary cost analysis only.

15.6 Processing Crude Oil

No discussion on oil processing would be complete without another brief introduction
to distillation. Distillation may be defined as the separation of the components of a liq-
uid feed mixture by a process involving partial vaporization through the application of
heat. In general, the vapor evolved is recovered in liquid form by condensation. The more
volatile (lighter) components of the liquid mixture are obtained in the vapor discharge at
a higher concentration. The extent of separation is governed by two important factors: the
properties of the components involved and by the physical arrangement of the unit used
for distillation.
In order to provide fuels, crude oil has to be refined. It is the aforementioned distilla-
tion column that is the workhorse of this process. The liquid is first boiled to turn it into a
light vapor, which is passed into a tower where it rises and cools. Each of the components
of crude oil has a different boiling point, so at each stage of the cooling a different liquid,
or fraction, can be drawn off at different heights in the column. Parts of the mixture that
are not fully vaporized at first will be collected as a liquid at the bottom of the tower and
recirculated.
At the end of this stage, known as fractional distillation, the crude oil mixture will have
been separated into several distinct substances. The components that boiled at the lowest
temperature will include propane and butane. Liquids that are boiled at higher tempera-
tures will include fuels used in cars and aircraft, while heavier oils needed to power diesel
locomotives or to fire industrial boilers would need even more heat. But none of the prod-
ucts will be ready for the consumer until they have been refined even further.
Products are vaporized and returned to liquid form as they pass through a series of
processes designed to purify them and improve their performance as fuels. For example,
most modern refineries reduce the sulfur content in oil products. Besides minimizing
pollution, this process yields pure sulfur, which can be used in the manufacture of fertil-
izers. If the first refining stage produces too much heavy fuel oil and not enough of the
lighter products, such as gasoline for automobiles, heavier products are passed through
another unit known as a hydrocracker, which will break them down (react) into lighter
oils. Finally, the output from various parts of the refinery may be blended and combined
with other chemicals to produce a complete range of the fuels most suited to the needs of
consumers.7

Illustrative Example 15.6

A distillation column processes two types of crude oil streams: North Texas and
Venezuelan. Because of consumer demand, the production of diesel fuel, home heating
oil, and gasoline must be limited. Based on the information in Table 15.3, estimate the
optimum daily usage rate of each crude oil to maximize profit, P.
The profit generated on processing North Texas (A) and Venezuelan crude (B) is $2.00/
gal and $1.60/gal, respectively.
Mass Transfer Applications 233

TABLE  15.3
Crude Oil Usage and Production Information
North Texas (%) Venezuelan (%) Maximum Production Rate (gal/day)
Diesel fuel 8 11 1,500
Home heating oil 29 54 5,500
Gasoline 63 35 11,000

SOLUTION:
Set NA and NB equal to the daily usage rate of North Texas and Venezuelan crude. One
may then write the following based on the problem statement:

P = 2.00 N A + 1.60 N B ; N = gal day (15.39)

Constraints

0.08 N A + 0.11N B ≤ 1500 (15.40)

0.29N A + 0.54 N B ≤ 5500 (15.41)

0.63 N A + 0.35N B ≤ 11, 000 (15.42)

N A ≥ 0 (15.43)

N B ≥ 0 (15.44)

By trial-and-error, or from an optimization program, one obtains (approximately)

N A = 16, 800 gal day

N B = 1150 gal day

P = 35, 500 $ day

References
1. L. Theodore and J. Barden, Mass Transfer Operations, A Theodore Tutorial, Theodore Tutorials,
East Williston, NY, originally published by USEPA/APTI, RTP, NC, 1995.
2. L. Theodore and F. Ricci, Mass Transfer Operations for the Practicing Engineer, John Wiley & Sons,
Hoboken, NJ, 2010.
3. R. Treybal, Mass Transfer Operations, McGraw-Hill, New York City, NY, 1955.
4. M. Peters, Plant Design and Economics for Chemical Engineers, McGraw-Hill, New York City, NY,
1958.
5. L. Theodore and R. Allen, Pollution Prevention, A Theodore Tutorial, Theodore Tutorials, East
Williston, NY, originally published by USEPA/APTI, RTP, NC, 1956.
6. R. Treybal, Liquid Extraction, McGraw-Hill, New York City, NY, 1956.
7. K. Skipka and L. Theodore, Energy Resources: Availability, Management, and Environmental
Impacts, CRC Press/Taylor & Francis Group, Boca Raton, FL, 2014.
16
Heat Transfer Applications

The fourth chapter in Part III contains six illustrative examples concerned with the gen-
eral subject of heat transfer (HT). The title and one-line description of each example are
provided here:

16.1 Steam options

16.2 Optimum insulation thickness
16.3 Selecting the most profitable exchanger
16.4 Critical insulation thickness
16.5 Recovering quality energy
16.6 Maximizing profit through energy recovery

Six references are provided below for suggest reading for this chapter:

1. R. Perry and D. Green, Perry’s Chemical Engineers’ Handbook, 8th edn, McGraw‑Hill,
New York City, NY, 2008.
2. L. Theodore, Chemical Engineering: The Essential Reference, McGraw-Hill, New York
City, NY, 2014.
3. J. Reynolds, J. Jeris, and L. Theodore, Handbook of Chemical and Environmental
Engineering Calculations, John Wiley & Sons, Hoboken, NJ, 2004.
4. L. Theodore, R. Dupont, and K. Ganesan, Unit Operations in Environmental
Engineering, Wiley‑Scrivener, Salem, MA, 2018.
5. L. Theodore, Heat Transfer Applications for the Practicing Engineer, John Wiley &
Sons, Hoboken, NJ, 2011.
6. A. M. Flynn, T. Akashige, and L. Theodore, Kern’s Process Heat Transfer, 2nd edn,
Wiley‑Scrivener, Salem, MA, 2018.

16.1 Steam Options
The term phase for a pure substance refers to a state of matter that is gas, liquid, or solid.
Latent enthalpy effects are associated with phase changes. These phase changes involve
no change in temperature, but there is a transfer of energy to/from the substance. There
are three possible latent effects, as detailed here:

1. Vapor–liquid
2. Liquid–solid
3. Vapor–solid

235
236 Introduction to Optimization for Chemical and Environmental Engineers

Vapor–liquid changes are referred to as condensation when the vapor is condensing and
vaporization when liquid is vaporizing. Liquid–solid changes are referred to as melting
when the solid melts to liquid and freezing when a liquid solidifies. Vapor–solid changes
are referred to as sublimation. One should also note that there are enthalpy effects associ-
ated with a phase change of a solid to another solid form; however, this enthalpy effect is
small compared to the other effects mentioned previously. Finally, vapor–liquid changes
are of primary interest to the practicing environmental and chemical engineer. This leads
to the subject of steam tables.
Water is one of the most abundant substances in the earth’s crust and its properties
have been determined with considerable accuracy. Since water is used in most chemical
operations as a coolant in the liquid form or as a heating medium in the form of steam, the
environmental and chemical engineer should be familiar with the compilation of data on
water known as the Steam Tables. These tables are available in the literature.1–3

Illustrative Example 16.1

Consider the following 2017 application. Based on the results of a 2017 energy audit study
for a new process, it is necessary to heat 50,000 lb/h of an organic liquid from 150 to 330°F.
The liquid is at a pressure of 135 psia. A simple steam-heated shell-and-tube floating-head
carbon steel exchanger is the preferred equipment choice.1–3 Steam is available at 150 psia
(135 psig) and 300 psia (285 psig). The higher pressure steam should result in a smaller
heat exchanger, but the steam will cost more. Which steam choice would be better?
Data:

• The heat capacity of the organic liquid is 0.6 Btu/lb · °F.

• The plant on-stream operation factor is expected to be 90%.

Steam properties are provided in Table 16.1.

Heat exchanger cost correlation (1998 basis):

• Base cost (BC) = 117 A0.65

• Installation factor (IF) = 3.29

Pressure factors (PF):

• 100 to 200 psig = 1.15

• 200 to 300 psig = 1.20

Cost indexes (CI):

• 1998 = 230
• 2016 = 360
• Capital cost (C) = (BC)(IF)(PF)(CI).

SOLUTION:
Calculate the overall heat duty:

 p (T2 − T1 ) =  50, 000   0.6

lb Btu 
 ( 330 − 150°F )
 = mc
Q
 h lb ⋅ °F  (16.1)
= 5, 400, 000 Btu / h
Heat Transfer Applications 237

Calculated log-mean temperature differences for each case are provided in Table 16.2.

Calculate the required heat transfer area for each case:

5, 400, 000
A150 = = 436 ft 2
(138 )( 89.8 ) 

5, 400, 000
A 300 = = 244 ft 2
(138 )(160.5 ) 
Determine the capital cost for each case:

 360 
Cost 150 = (117 )( 436 )  ( 3.29)(1.15) = $36, 000
0.65

230 

 360 
Cost 300 = (117 )( 244 )  ( 3.29)(1.20 ) = $25, 800
0.65

230 
Obtain the steam requirement for each case in lb/yr:

St150 =
( 5, 400, 000 Btu/h )(8760 × 0.9 h/yr ) = 49.3 million lb/yr
863.6 Btu/lb

St 300 =
( 5, 400, 000 Btu / h )(8760 × 0.9 hr / yr) = 52.6 million lb / yr
809.0 Btu / lb
Use the previous calculations to determine the annual steam cost for each case:

( )
StCost 150 = 49.3 × 10 6 lb / yr ( 0.00520 $ / lb ) = $256, 000 / yr

( )
StCost 300 = 52.6 × 10 6 lb / yr ( 0.00575 $ / lb ) = $303, 000 / yr

The 300-psia exchanger costs $10,200 less to purchase and install, but it costs $47,000
per year more to operate. Choosing the more expensive 150-psia exchanger is the obvi-
ous choice.

TABLE 16.1
Steam Data
150 psia 300 psia
Saturation temperature, °F 358.0 417.0
Latent heat (enthalpy), Btu/lb 863.6 809.0
Cost, $/1000 lb 5.20 5.75

TABLE 16.2
Log-Mean Temperature Differences Calculations
150 psia case 300 psia case
ΔT1 358 − 150 = 208 417 − 150 = 267
ΔT2 358 − 330 = 28 417 − 330 = 87
ΔtLM = LMTD 89.8°F 160.5°F
238 Introduction to Optimization for Chemical and Environmental Engineers

16.2 Optimum Insulation Thickness

Industrial thermal insulation usually consists of materials of low thermal conductivity
combined in a way to achieve a higher overall resistance to heat flow. Webster4 defines
insulation as, “to separate or cover with a non-conducting material in order to prevent the
passage or leakage of…heat…etc.” Insulation is defined in Perry’s1 in the following man-
ner: “Materials or combinations of materials which have air or gas-filled pockets or void
spaces that retard the transfer of heat with reasonable effectiveness are thermal insula-
tors. Such materials may be particulate and/or fibrous, with or without binders, or may
be assembled, such as multiple heat-reflecting surfaces that incorporate air or gas-filled
void spaces.” Refractory materials also serve the chemical process industries. In addition
to withstanding heat, refractory also provides resistance to corrosion, erosion, abrasion,
and/or deformation.
Insulation must often be considered an integral part of any exchanger. In addition, the
extent of the problem of keeping heat out of a storage vessel containing a liquid refrigerant
or a cryogenic liquid varies widely. Generally, one must decide on the permissible and/
or allowable heat losses (leaks) since insulation costs money, and an economic analysis
must be performed. Thus, the main purpose of insulation is to minimize radiative and
convective heat transfer and to use as little material as possible in providing the optimal
insulation. When choosing appropriate insulation, the following factors are usually taken
into consideration:

1. Ruggedness
2. Convenience
3. Volume and weight
4. Ease of handling
5. Thermal effectiveness
6. Cost

The thermal conductivity (k) of a material is a major consideration in determin-

ing the thermal effectiveness of the insulation material. Different types of insulation
obviously have different k values and there are five categories of insulation. These
include

1. Vacuum insulation, which employs an evacuated space that reduces radiant heat
transfer
2. Multilayer insulation, referred to by some as superinsulation, which consists of
alternating layers of highly reflective material and low conductivity insulation in
a high vacuum
3. Powder insulation, which utilizes finely divided particulate material packed
between surfaces
4. Foam insulation, which employs non-homogeneous foam whose thermal conduc-
tivity depends on the amount of insulation
5. Special insulation, which includes composite insulation that incorporates many of
the advantageous qualities of the other types of insulation
Heat Transfer Applications 239

The optimum thickness of insulation to be used in a particular application can be deter-

mined by a cost calculation for a number of standard thicknesses. The procedure is to
calculate the heat loss for various thicknesses of insulation. The annual cost of heat loss
can be determined for each thickness and the annual cost of the installed insulation can be
determined from the initial cost and any applicable depreciation rate.
As one would suppose, as the thickness of the insulation is increased, the cost of heat
loss decreases (a credit), but the fixed charges on the insulation increase. The optimum
thickness is determined by the minimum of the total variable costs.

Illustrative Example 16.2

Consider the following application. It is desired to choose the most economical thick-
ness of insulation for a steam line that is 1000 ft long. The annual fixed costs for the
insulation are 20% of the initial cost. The installed cost in dollars per foot in pipe is
$6000(t) where t is the thickness in inches. The heat losses for the heat losses for varying
insulation thickness are provided in Table 16.3.
If saturated steam at 125 psig is flowing in the line and the value of steam is $0.80 per
1,000 lbs, what is the optimum thickness of insulation that should be employed? Note:
For saturated steam at 125 psig, the latent heat of vaporization is 868 Btu/lb.

Hint: The reader may choose to revisit Part I, Section 6.2.

SOLUTION:
The cost of heat loss is given as

Cost  1, 000, 000 Btu   $0.80 

=
Million Btu Lost  868 Btu / lbsteam   1000 lb steam  (16.2)

= $0.922
The insulation that should be chosen, based on Table 16.4, is in the two-inches range.
A plot of these values may also be performed; it would result in a graph similar to that
seen in Figure 16.1.

16.3 Selecting the Most Profitable Exchanger

Of the various types of heat exchangers that are employed in industry, perhaps the two
most popular are the double pipe and the shell-and-tube. Despite the fact that shell-and-
tube heat exchangers generally provide greater surface area for heat transfer with a more
TABLE 16.3
Insulation or Heat Loss Data
Insulation
Thickness (inches) Heat Losses (Btu/h)
0 600,000
1 300,000
2 120,000
3 60,000
4 20,000
6 7,000
240 Introduction to Optimization for Chemical and Environmental Engineers

TABLE 16.4
Cost Data; Illustrative Example 16.3
Insulation Initial Annual Cost of Total
Thickness Installed Fixed Annual Heat Loss Annual Heat Annual
(inches) Cost ($) Cost ($) (Million Btu)* Loss ($) Cost ($)
0 0 0 5,260 4,860 4,860
1 6,000 1,200 2,620 2,420 3,620
2 12,000 2,400 1,050 970 3,370
3 18,000 3,600 524 484 4,084
4 24,000 4,800 175 162 4,962
6 36,000 7,200 61 56 7,256
* There are 8760 h per yr.

Total cost
Total annual cost $

Value of heat loss

Fixed costs

Insulation thickness

FIGURE 16.1
Optimum thickness of insulation; consistent units.

compact design, greater ease of cleaning, and less possibility of leakage, the double pipe
heat exchanger still finds use in practice.
The double pipe unit consists of two concentric pipes. Each of the two fluids – hot and
cold – flow either through the inside of the inner pipe or through the annulus formed
between the outside of the inner pipe and the inside of the outer pipe. Generally, it is more
economical (from a heat efficiency perspective) for the hot fluid to flow through the inner
pipe and the cold fluid through the annulus, thereby reducing heat losses to the surround-
ings. In order to ensure sufficient contacting time, pipes longer than approximately 20 ft
are extended by connecting them to return bends. The length of pipe is generally kept to a
maximum of 20 ft because the weight of the piping may cause the pipe(s) to sag. Sagging
may allow the inner pipe to touch the outer pipe, distorting the annulus flow region and
disturbing proper operation. When two pipes are connected in a “U” configuration by a
return bend, the bend is referred to as a hairpin. In some instances, several hairpins may
be connected in series.
Double pipe heat exchangers have been used in the chemical process industry for over
90 years. The first patent on this unit appeared in 1923. The unique (at that time) design
provided “the liquid to be heated or cooled and flow longitudinally and transversely
around a tube containing the cooling or heating liquid, etc.” The original design has not
changed significantly since that time. Although this unit is not extensively employed in
Heat Transfer Applications 241

industry (the heat transfer area is small relative to other heat exchangers), it serves as an
excellent starting point from an academic and/or training perspective in the treatment of
all the various heat exchangers.
Shell-and-tube (also referred to as tube and bundle) heat exchangers provide a large heat
transfer area economically and practically. The tubes are placed in a bundle and the ends
of the tubes are mounted in tube sheets. The tube bundle is enclosed in a cylindrical shell,
through which the second fluid flows. Most shell-and-tube exchangers used in practice are
of welded construction. The shells are built as a piece of pipe with flanged ends and neces-
sary branch connections. The shells are made of seamless pipe up to 24 inches in diameter;
they are made of bent and welded steel plates if above 24 inches. Channel sections are
usually of built-up construction, with welding-neck forged-steel flanges, rolled-steel bar-
rels, and welded-in pass partitions. Shell covers are either welded directly to the shell, or
they are built-up constructions of flanged and dished heads and welding-neck forged-steel
flanges. The tube sheets are usually non-ferrous castings in which the holes for inserting
the tubes have been drilled and reamed before assembly. Baffles can be employed to both
control the flow of the fluid outside the tubes and provide turbulence.
There are vast industrial uses of shell-and-tube heat exchangers. These units are used
to heat or cool process fluids, either through a single-phase heat exchanger or a two-phase
heat exchanger. In single-phase exchangers, both the tube side and shell side fluids remain
in the same phase that they enter. In two-phase exchangers (examples include condensers
and boilers), the shell side fluid is usually condensed to a liquid or heated to a gas, while
the tube-side fluid remains in the same phase.
Generally, shell-and-tube exchangers are employed when double pipe exchangers do not
provide sufficient area for heat transfer. Shell-and-tube exchangers usually require fewer
materials of construction and are usually more economical when compared to double pipe
and/or multiple double pipe heat exchangers in parallel.

Illustrative Example 16.3

Consider the following application. Plans are underway to construct and operate a com-
mercial hazardous waste facility in Dumpsville in the state of Egabrag. The company
is still undecided as to whether to install a double pipe or shell-and-tube or extended
surface (finned) heat exchanger at the plant site to recover energy. The double pipe (DP)
unit is less expensive to purchase and operate than either a comparable shell-and-tube
(ST) or extended surface (ES) system. However, projected energy recover income from
the ST and ES units are higher since they will handle a larger quantity and different
temperature steam.
Based on economic and financial data provided in Table 16.5, select the heat exchanger
that will yield the higher annual profit. Calculations should be based on an interest rate
of 5% and a process lifetime of 20 years for the three exchangers.

TABLE 16.5
Costs/Credits Data; Illustrative Example 11.3
Costs/Credits Double Pipe (DP) Shell-and-tube (ST) Extended Surface (ES)
Capital ($) 26,250 29,750 34,350
Installation ($) 15,750 17,000 17,550
Operation ($/yr) 4,000 5,500 5,900
Maintenance ($/yr) 6,500 7,750 8,450
Income ($/yr) 20,000 25,000 25,100
242 Introduction to Optimization for Chemical and Environmental Engineers

SOLUTION:
Calculate the capital recovery factor CRF:

i (16.3)
CRF =
1 − (1 + i )− n 
 
0.05
=
1 − (1 + 0.05)−20 
 
= 0.080
Determine the annual capital and installation costs for the DP unit:

DP costs = ( Capital + Installation ) ( CRF ) (16.4)

= ( 26, 250 + 15, 750 )( 0.080 )

= $3, 360 / yr
Determine the annual capital and installation costs for the ST unit:

ST costs = ( Capital + Installation ) ( CRF ) (16.5)

= ( 29, 750 + 17 , 000 )( 0.080 )

= $3, 740 / yr
Determine the annual capital and installation costs for the ES unit:

ES costs = ( Capital + Installation ) ( CRF ) (16.6)

= ( 34, 350 + 17 , 550 )( 0.080 )

= $4, 150 / yr
The profit is the difference between the income credit and the total annual cost:

Profit = income credit − total cost (16.7)

DP ( profit ) = 20, 000 − 13, 860 = 6, 140 $ / yr

ST ( profit ) = 25, 000 − 16, 990 = 8, 010 $ / yr

ES ( profit ) = 25, 100 − 18, 500 = 6, 600 $ / yr

The information in Table 16.6 provides a comparison of the costs and credits for both
exchangers.
A shell-and-tube heat exchanger should therefore be selected based on the previously
mentioned economic analysis.

16.4 Critical Insulation Thickness

It would normally seem that the thicker the insulation, the less the heat loss, that is, increas-
ing the insulation should reduce the heat loss to the surroundings. But, this is not always
the case. There is a “critical insulation thickness” below which the system will experience a
Heat Transfer Applications 243

TABLE 16.6
Comparison of Economic Exchanger Results; Illustrative Example 16.3
Double pipe Shell-and-tube Extended Surface
Total installed ($/yr) 3,360 3,740 4,152
Operation ($/yr) 4,000 5,500 5,900
Maintenance ($/yr) 6,500 7,750 8,450
Total annual cost ($/yr) 13,860 16,990 18,500
Income credit ($/yr) 2,000 25,000 25,100
Profit ($/yr) 6,140 8,010 6,600

greater heat loss due to an increase in insulation. This situation arises from “small” diam-
eter pipes when the increase in area increases more rapidly than the resistance arising
from the thicker insulation; in effect, the heat loss can initially increase with increasing
insulation.
Consider the system shown in Figure 16.2. The area terms for the heat transfer equation
in rectangular coordinates are no longer valid in cylindrical coordinates equal in rectan-
gular coordinates (for the inside surface, the heat transfer area is given by 2πrL).3,4 Applying
the development to a pipe/cylinder system provided by Theodore2,3 leads to

T1 − T0
Q. =
1  1 xw 1  1
+ +
2πri L  hi  k w 2πLrim , i 2πrL  h0 
 
(16.8)
2πL (T1 − T0 ) 2πL (T1 − T0 )
= =
1 ln r0
+
r
+
1 ( ) f (r )

ri hi kw rh0
where:

f (r) =
1 ln r0
+
r
+
( )
1
(16.9)
ri hi kw rh0

h0
ki
T0
kw ∆Xi
∆Xw 1
h1 T1
r ri r0

FIGURE 16.2
Critical insulation thickness for a pipe.
244 Introduction to Optimization for Chemical and Environmental Engineers

Assuming that goes through a maximum or minimum as r is varied, L’Hopital’s rule can
be applied to Equation (16.9):
 df ( r ) 
− 
2πL ( T1 − T0 )  dr  = 0 (16.10)
 f (r) 
2

 
with
df ( r ) 1 1
− =− + (16.11)
dr rki r 2 h0
One may therefore write,
df ( r ) 1 1
− =− + 2 = 0 (16.12)
dr rki r h0
For this maximum/minimum condition, set r = rc and solve for rc
ki
rc = (16.13)
h0
The second derivate of Equation (16.9) provides information as to whether there is a maxi-
mum or minimum at rc.

d2 f ( r ) 1 2 h2 h2
2
= 2 − 3 = 03 = 03 ( 1 − 2 ) (16.14)
dr r ki r h0 k k
Clearly, the second derivative is a negative number; therefore, a maximum is located at r = rc. It
then decreases monotonously as r is increased beyond rc. However, one should exercise care in
interpreting the implications of the previously mentioned development. This result applies only
if r0 is less than rc; that is, it generally applies to “small” diameter pipes/tubes. Thus, rc repre-
sents the outer radius (not the thickness) of the insulation that will maximize the heat loss, and
at which point any further increase in insulation thickness will result in a decrease in heat loss.
A graphical plot of the resistance R2,3 versus r is provided in Figure 16.3. (The curve is
inverted for the plot of Q vs. r.) In other words, the maximum heat loss from a pipe occurs
when the critical radius equals the ratio of the thermal conductivity of the insulation to the
surface coefficient of heat transfer (see also Equation 16.13). This ratio has the dimension
of length (e.g., ft). Note that once again the equation for ri can be rewritten in terms of a
dimensionless number defined as the Biot number, Bi:
ki
= Bi−1 (16.15)
h0 rc
To apply the following equation for a base wall (r = rc), r must be greater than rc, that is,
r > rc. The radius at which this occurs is denoted at r*. The term r* may be obtained by solv-
ing the equation

2πL (T1 − T0 ) 2πL (T1 − T0 )

=
( ) ( ) ( )
(16.16)
ln r0 ln r0 ln r*
1 r1 1 1 r1 r0 1
+ + + + +
ri hi kw r0 h0 ri hi ki ki r* h 0
One may solve for r* using a suitable trial-and-error procedure.
Heat Transfer Applications 245

Resistance, R

ri r0 rc r*
Radius, r

FIGURE 16.3
Resistance associated with the critical insulation thickness for a bare surface.

The previously mentioned development applies where r is less than rc. If r0 is larger
than rc, the previous analysis again applies, but only to the results presented for r > rc; that
is, it will decrease indefinitely as r increases. Note that there is no maximum/minimum
(inflection) for this case since values of ln(r/r0) are indeterminate for r > r0. Once again, it
approaches zero in the limit as r approaches infinity.
As noted earlier, as the thickness of the insulation is increased, the cost associated with
heat loss decreases, but the insulation cost increases. The optimum thickness is determined
by the minimum of the total costs. Thus, as the thickness of the insulation is increased, the
heat loss reaches a maximum value and then decreases with further increases in insula-
tion. Reducing this effect can be accomplished by using an insulation of low conductivity.

Illustrative Example 16.4

With reference to the previously mentioned development, calculate the outer critical
radius of insulation (k = 0.44 Btu/h · ft · °F) on a 2.0-inch outside diameter (OD) pipe.
Assume the air flow heat transfer coefficient to be 1.32 Btu/h · ft2 · °F. Comment on the
effect of insulation on the heat rate lost from the pipe.

SOLUTION:
Assume the resistances are approximately equal, i.e., ki = horc and employ Equation
(16.13). Since BL ≈ 1.0, set

ki
rc =
h0
0.44
= = 0.333 ft
1.32
= 4.0 inches
The critical insulation thickness is therefore

 2.0 
L1 = ∆xi = 4.0 −   = 3.0 inches
 2 
Since r0 = 1.0 inches, r0 < rc, so that the heat loss will increase as insulation is added, but it
will start to decrease when the radius of the insulation increases above rc, that is, when r > rc.
246 Introduction to Optimization for Chemical and Environmental Engineers

16.5 Recovering Quality Energy

The law of conservation of energy is defined by many as the first law of thermodynamics.
Its application allows calculations of energy relationships associated with various pro-
cesses. The second law of thermodynamics is referred to as the “limiting law”. Historically,
the study of the second law was developed by individuals, such as Carnot, Clausius, and
Kelvin in the middle of the 19th century. This development was made purely on a macro-
scopic scale and is referred to as the “classical approach” to the second law.
Concerns involving conservation of energy issues gained increasing prominence dur-
ing and immediately after the OPEC oil embargo of 1973. In addition, global popula-
tion growth has led to an increasing demand for energy. Although the use of energy has
resulted in great benefits, the environmental and human health impact of this energy use
has become a concern. One of the keys to reducing and/or eliminating this problem will
be achieved through what has come to be referred to as meaningful energy conservation.

Illustrative Example 16.5

Discuss optimizing the recovery of quality energy from an environmental perspective.

SOLUTION:
The first law of thermodynamics is a conservation law concerned with energy transfor-
mation. Regardless of the types of energy involved in processes – thermal, mechanical,
electrical, elastic, magnetic, and so on – the change in the energy of a system also allows
free convertibility from one form of energy to another, as long as the overall quantity is
conserved. Thus, this law places no restriction on the conversion of work into heat, or on its
counterpart – the conversion of heat into work. However, the second law is another matter.
One of the areas where the meaningful energy conservation measures can be realized
is in the design and specification of process (operating) conditions for heat exchangers.
This can be best accomplished by the inclusion of second law principles in the analysis.
The quantity of heat recovered in an exchanger is not alone in influencing size and cost.
As the temperature difference driving force (LMTD) in the exchanger approaches zero,
the “quality” heat recovered increases.5,6
Any discussion of energy conservation regarding heat exchangers leads to an impor-
tant second law consideration – energy has quality as well as quantity. Because work is
100% convertible to heat, whereas the reverse situation is not true, work is a more valu-
able form of energy than heat. Although it is not obvious, it can also be shown through
second law arguments that heat also has “quality” in terms of the temperature at which
it is discharged from a system. The higher the temperature, the greater the possible
energy transformation into work. Thus, thermal energy stored at high temperatures is
generally more useful to society than that available at lower temperatures. While there
is an immense quantity of energy stored in rivers and oceans for example, its present
availability to society for performing useful tasks is its “quality” since it is degraded
when it is transferred by means of heat transfer from one temperature to a lower one.
Other forms of energy degradation include energy transformation due to frictional
effects and electrical resistance. Such effects are highly undesirable if the use of energy
for practical purposes is to be maximized.5,6
The second law provides the means of measuring this energy degradation through a
thermodynamic term referred to as entropy, and it is the second law (of thermodynamics)
that serves to define this important property. It is normally designated as S with units of
energy per absolute temperature (e.g., Btu/R or cal/K). Furthermore, entropy calculations
can provide quantitative information on the “quality” of energy and energy degradation.5,6
Heat Transfer Applications 247

There are a number of other phenomena that cannot be explained by the first law of
conservation of energy. It is the second law of thermodynamics that provides an under-
standing and analysis of these adverse effects. However, among these considerations,
it is the second law that can allow the measuring of the aforementioned “quality” of
energy, including its effect on the design and performance of heat exchange.

16.6 Maximizing Profit through Energy Recovery

Illustrative Example 16.6
Two hot liquid (essentially water) process streams are discharged from sources A and B
in a plant. A young engineer has proposed to recover energy from the two streams by
employing two heat exchangers that are currently available on-site: a double pipe (DP)
exchanger and a shell-and-tube (ST) exchanger. Due to diameter differences in the two
lines that discharge the hot liquid from source A, there is a 60/40 split (mass basis) of the
flowrate. There are also two lines discharging liquid from source B with a 75/25 split to
the ST/TB exchangers. Due to the location of the exchangers relative to the two sources,
the 40% flow from A and 75% flow from B can be diverted (fed) to the ST exchanger. The
remaining discharge can only be sent to the DP exchanger.
The young engineer has obtained additional information. The maximum flowrate
that exchanger ST and DP can accommodate is 12,000 ft3/day and 6000 ft3/day respec-
tively. In addition, the maximum flow that be drawn from sources A and B is 8000 ft3/
day and 6000 ft3/day respectively.
The profit value of the recovered energy from sources A and B are $1.70/ft3 and $2.00/
ft , respectively. Develop the describing equations that will provide values of the (volu-
3

metric) flowrates from A and B, that is, qA and qB, and that will maximize the profit P for
the proposed processes.

SOLUTION:
The line diagram of the system is provided in Figure 16.4.
The describing equations are as follows. The objective function for the profit is

P = (1.70 ) qA + ( 2.00 ) qB (16.17)

0.75 qB
qST ≤ 12,000
qA ≤ 8000 Shell and tube (ST)
A
0.40 qA

0.60 qA
qDP ≤ 6000
Double pipe (DP)
0.25 qB
qB ≤ 6000

FIGURE 16.4
Shell-and-tube double pipe exchangers; Illustrative Example 16.6.
248 Introduction to Optimization for Chemical and Environmental Engineers

The constraints are

qA ≤ 8, 000 (16.18)
qB ≤ 6, 000 (16.19)
0.75qA + 0.40qB = qST ≤ 12, 000 (16.20)
0.60qA + 0.25qB = qDP ≤ 6, 000 (16.21)
qA ≥ 0; qB ≥ 0 (16.22)

SOLUTION:
One may also calculate the annual (365-day basis) profit based on this condition.
Employing a suitable optimization program gives

qA = 7 , 500 ft 3 /day

qB = 6, 000 ft 3 /day
P = 24, 750 $/day

= $8, 910, 000 $/year
Note that operating, maintenance, depreciation, and so on, costs have not been included
in the analysis.

References
1. R. Perry and D. Green, Perry’s Chemical Engineers’ Handbook, 8th edn, McGraw-Hill, New York
City, NY, 2008.
2. L. Theodore, Chemical Engineering: The Essential Reference, McGraw-Hill, New York City, NY,
2014.
3. L. Theodore, Heat Transfer Applications for the Practicing Engineer, John Wiley & Sons, Hoboken,
NJ, 2011.
4. Webster’s New World Dictionary, Second College Edition, Prentice-Hill, Upper Saddle River, NJ,
1971.
5. L. Theodore, F. Ricci, and T. VanVliet, Thermodynamics for the Practicing Engineer, John Wiley &
Sons, Hoboken, NJ, 2009.
6. L. Theodore and J. Reynolds, Thermodynamics, A Theodore Tutorial, Theodore Tutorials, East
Williston, NY, originally published by the USEPA/APTI, RTP, NC, 1991.
17
Plant Design Applications

The last chapter in Part III contains six illustrative examples concerned with the general
subject of plant design (PD). The title and one-line description of each example are pro-
vided here:

17.1 Shipping facilities

17.2 Tank farms
17.3 Cyclone selection and design
17.4 Minimizing the cost of a batch plant operation
17.5 Ventilation models with system variables
17.6 Plant structure design

The seven references that follow are for suggested reading for plant design applications:

1. R. Perry and D. Green, Perry’s Chemical Engineers’ Handbook, 8th edn, McGraw‑Hill,
New York City, NY, 2008.
2. L. Theodore, Chemical Engineering: The Essential Reference, McGraw‑Hill, New York
City, NY, 2014.
3. J. Reynolds, J. Jeris, and L. Theodore, Handbook of Chemical and Environmental
Engineering Calculations, John Wiley & Sons, Hoboken, NJ, 2004.
4. L. Theodore, R. Dupont, and K. Ganesan, Unit Operations in Environmental
Engineering, Wiley‑Scrivener, Salem, MA, 2018.
5. J. Happel, Chemical Process Economics, John Wiley & Sons, Hoboken, NJ, 1958.
6. F. Vilbrandt and C. Dryden, Chemical Engineering Plant Design, McGraw-Hill, New
York City, NY, 1959.
7. M. Peters, Plant Design and Economics for Chemical Engineers, McGraw-Hill, New
York City, NY, 1958.

17.1 Shipping Facilities
A large chemical plant complex is likely to have facilities for shipping by pipeline, sea,
inland waterway, rail, and road. Smaller plants may have facilities for only one or two of
these modes. In any case, they must be designed for safe loading and unloading. In addi-
tion, there is often a major economic incentive to load and unload ships, barges, rail cars,
and trucks as quickly as possible.

249
250 Introduction to Optimization for Chemical and Environmental Engineers

The cost of shipping materials, both raw materials and products, is often a very significant
factor in the chemical industry. The energy cost associated with shipping can range from
0.001 gallons of fuel per ton-mile of cargo moved for supertankers to 0.2 for cargo jet aircraft.

Illustrative Example 17.11

A new process plant on the Ohio River expects to receive 500,000 gallons per day of a
single raw material and ship out 100,000 gallons per day of each of two liquid products
and 500,000 lbs each of four solid products. Estimate the total shipping cost per day
for this plant’s operations if all shipments are by barge and if all shipments are by rail.
Select the optimum mode of shipping. Data is provided in Tables 17.1 and 17.2.

SOLUTION:
The number of tons per day in transit is given by the volume per day times the density. The
density of water is 8.33 (specific gravity = 1.00) in lb/gal, and 2000 converts from lbs to tons.
To calculate the shipping cost, the tons are multiplied by the number of miles shipped and
the ton-mile shipping cost. Liquid products are calculated in the same manner. Solid prod-
ucts are calculated similarly, except that the number of pounds per day is already given.
Calculate the shipping cost of the feed by barge and rail:
Barge

( 500, 000)  2000  (800)(0.015) = 25, 000 $ day

8.33

Rail

( 500, 000)  2000  (800)(0.08) = 133, 300 $ day

8.33

TABLE 17.1
Shipping Distance Data; Illustrative Example 17.1
Distance
Shipped
Product (miles)
Feed 800
Liquid product A 250
Liquid product B 200
Solid product C 40
Solid product D 300
Solid product E 45
Solid product F 120

TABLE 17.2
Shipping Cost Data; Illustrative Example 17.1
Transportation Type Distance (miles) Cost ($/ton-mile)
Barge 20–80 0.03
80–300 0.02
300–1000 0.015
Rail 10–100 0.08
100–1000 0.04
Plant Design Applications 251

Calculate the shipping cost of liquid products by barge and rail.

Barge, Product A:

(100, 000)  2000  (800)( 250)(0.02) = 2, 080 $ day

8.33

Barge, Product B:

(100, 000)  2000  (800)( 200)(0.02) = 1, 670 $ day

8.33

Rail, Product A:

(100, 000)  2000  (800)( 250)(0.04) = 4, 170 $ day

8.33

Rail, Product B:

(100, 000)  2000  (800)( 200)(0.04) = 3, 330 $ day

8.33

Calculate the shipping cost of solid products by barge and rail.

Barge, Product C:

 500, 000 
  ( 40 )(0.03 ) = 3, 000 $ day
2000 
Barge, Product D:

 500, 000 
  ( 300 )(0.02) = 1, 500 $ day
2000 
Barge, Product E:

 500, 000 
  ( 45)(0.03 ) = 340 $ day
2000 
Barge, Product F:

 500, 000 
  (120 )(0.02) = 600 $ day
2000 
Rail, Product C:

 500, 000 
  ( 40 )( 0.03 ) = 800 $ day
 2000 
Rail, Product D:

 500, 000 
  ( 300 )(0.04 ) = 3, 000 $ day
2000 
Rail, Product E:

 500, 000 
  ( 45)(0.08 ) = 900 $ day
2000 
252 Introduction to Optimization for Chemical and Environmental Engineers

Rail, Product F:

 500, 000 
  (120 )(0.04 ) = 1, 200 $ day
2000 
The sum of the costs for barge transport are shown in Table 17.3. The sum of the costs for
rail transport are shown in Table 17.4.

Conclusion:
Transport by ship and/or barge is usually much cheaper than transport by rail or truck.
It often requires longer travel times, however, and thus more inventory can be tied up
in shipping.

17.2 Tank Farms
Liquid feeds, products, intermediates, and fuels at chemical and petrochemical plants are
stored in tanks, which are usually located in a “tank farm” adjacent to the process plant
area. The tank capacities are most often expressed in gallons or barrels (one barrel = 42
gallons). The individual tanks may range in size from a thousand gallons or less to several
million gallons.

TABLE 17.3
Costs for Barge Transport
Product Cost ($)
Feed 25,000
Product A 2,080
Product B 1,670
Product C 300
Product D 1,500
Product E 340
Product F 1,200
Total: 31,490

TABLE 17.4
Costs for Rail Transport
Product Cost ($)
Feed 133,300
Product A 4,170
Product B 3,330
Product C 800
Product D 3,000
Product E 900
Product F 600
Total: 146,100
Plant Design Applications 253

Whenever possible, liquids are stored at ambient temperature and pressure. Volatile liq-
uids may have to be stored under pressure. In addition, they often require vapor recovery
systems2 or other devices to prevent releases to the atmosphere as the tanks are filled and
emptied. Liquids that are very viscous at ambient temperature or that would solidify at
ambient temperature are kept in heated tanks.
Standard practice calls for each tank to be surrounded by a dike or berm sufficiently
high to contain all the liquid stored in a tank in case it should rupture. Firefighting equip-
ment are normally located near tanks containing flammable liquids.

Illustrative Example 17.2

Consider the same process plant described in Illustrative Example 17.1. The new process
plant expects to receive 500,000 gallons per day of a single raw material and ship out 100,000
gallons per day of each of two liquid products and 500,000 lbs each of four solid products.
Estimate the storage tank requirements for the liquid feed and products. If shipments are all
made by barge, the plant will require tankage for a 15-day supply of feed and a 10-day supply
of each product. If shipments are by rail, which can be scheduled more reliably, the plant will
require a seven-day supply of feed and a five-day supply of each product. For ease of main-
tenance, there should be at least three tanks for each liquid, with one tank being “off-line”
at any given time. Which mode of shipment should be employed for optimum operation?
Available standard tank sizes are provided in Table 17.5.

SOLUTION:
Calculate the storage requirement for feed and select an appropriate number of tanks:
Barge shipping:

Storage needed = (15)( 500, 000 ) = 7 , 500, 000 gallons

Use : three 120 ft × 48 ft tanks or five 100 ft × 36 ft tanks

Rail shipping:

Storage needed = (7 )( 500, 000 ) = 3, 500, 000 gallons

Use : three 100 ft × 36 ft tanks or five 70 ft × 36 ft tanks

Calculate the storage requirement for one of the liquid products.

Barge shipping:

Storage needed = (10 )(100, 000 ) = 1, 000, 000 gallons

Use : four 45 ft × 36 ft tanks or six 35 ft × 30 ft tanks

TABLE 17.5
Available Standard Tank Sizes; Illustrative Example 17.2
Capacity (gallons) Height (ft) Diameter (ft)
216,000 30 35
429,000 36 45
1,040,000 36 70
2,110,000 36 100
4,060,000 48 120
254 Introduction to Optimization for Chemical and Environmental Engineers

Rail shipping

Storage needed = ( 5)(100, 000 ) = 500, 000 gallons

Use : four 35 ft × 30 ft tanks

Rail shipping 45 ft × 36 ft tanks would also be satisfactory. Using the 35 ft × 30 ft tanks
allows a little more flexibility since there would be a total of eight tanks the of the same
size.
Specify a single set:
Barge shipping

Feed: five 100 ft × 36 ft tanks

Product A: four 35 ft × 36 ft tanks
Product B: four 45 ft × 36 ft tanks

Rail shipping

Feed: three 100 ft × 36 ft tanks

Product A: four 35 ft × 36 ft tanks
Product B: four 35 ft × 36 ft tanks

It should be mentioned that many different combinations would be acceptable. It

appears either modes of shipment would be acceptable.
Storage tanks can present major risks both in terms of fire and environmental pollu-
tion. In both cases, this is primarily due to the large inventories of materials that could
be involved in any accident. There has been a major effort in industry in the last several
decades to reduce the amount of storage at process plants, especially for flammable and/
or toxic materials.

17.3 Cyclone Selection and Design

Objects moving in circular paths tend to move away from the center of their motion. The
object moves outward as if a force is pushing it out. This force is known as centrifugal
force. The whirling motion of the gas in a cyclone, that is, centrifugal separator, causes
particulate matter in the gas to sense this force and move out to the walls and ultimately
be collected.2
Cyclones provide a relatively low-cost method of recovering or removing particulate
matter from exhaust gas streams. Cyclones are somewhat more complicated in design than
simple gravity settling systems, and their removal efficiency is accordingly much better
than that of settling chambers. However, cyclones are not as efficient as electrostatic pre-
cipitators, baghouses, and venturi scrubbers, but they are often installed as pre-cleaners or
pre-recovery units before these more effective devices.2
Cyclones come in many sizes and shapes and have no moving parts. From the small 1
and 2 cm diameter source sampling cyclones used for particle size analysis, to the large
5 m diameter cyclone separators used after wet scrubbers, the basic separation principle
remains the same. As noted previously, particles enter the device with the flowing gas; the
Plant Design Applications 255

gas stream is forced to turn, but the larger particles have more momentum and cannot turn
with the gas. These larger particles impact and, under the influence of gravity, fall down
the cyclone wall and are collected in a hopper. The gas stream actually turns a number of
times in a helical pattern, much like the funnel of a tornado. The repeated turnings provide
many opportunities for particles to pass through the streamlines, thus hitting the cyclone
wall.
The removal efficiency of a cyclone for a given size particle is very dependent on the
cyclone dimensions. The efficiency at a given volumetric flow rate is most affected by the
diameter. The overall length determines the number of turns of the vortex. The greater
the number of turns, the greater the efficiency. The length and width of the inlet are also
important, since the smaller the inlet, the greater the inlet velocity becomes. A greater
inlet velocity gives greater efficiency but also increases the pressure drop. A number of
equations have been developed for determining the fractional cyclone efficiency, Ei, for
a given size particle. As noted earlier, fractional efficiency is defined as the fraction of
particles of a given size collected in the cyclone, compared to those of that size going into
the cyclone.
Multiple-cyclone collectors (multiclones) are high-efficiency devices that consist of a
number of small-diameter cyclones operating in parallel with a common gas inlet and out-
let. The flow pattern differs from a conventional cyclone in that instead of bringing the gas
in at the side to initiate the swirling action, it is imparted by a stationary vane positioned
in the path of the incoming gas. The diameters of the collecting tubes usually range from
6 to 24 in. with pressure drops in the 2–6 in range.2
Properly designed units can be constructed and operated with collection efficiency as
high as 90% for particulates in the 5–10 µm range. The most serious problems encountered
with these systems involve plugging and flow equalization. Since the gas flow to a multi-
cone is axial (usually from the top), the cross-sectional area available for flow inlet condi-
tions is given by the annual area between the outlet tubes and cyclone body. The outlet
tube diameter is usually one-half the body diameter.2
The pressure drop across a cyclone is an important parameter to the purchaser of such
equipment. Increased pressure drop means greater costs for power to move an exhaust
gas through the recovery/control device. With cyclones, an increase in pressure drop usu-
ally means that there will be an improvement in collection efficiency (one exception to
this is the use of pressure recovery devices attached to the exit tube; these reduce the
pressure drop but do not adversely affect collection efficiency). For these reasons, there
have been many attempts to predict pressure drops from design variables. The idea is
that with having such an equation, one could work back and optimize the design of new
cyclones.2

Illustrative Example 17.3

Consider the following application. A recently hired plant engineer has been assigned the
job of selecting and specifying a cyclone unit to be used to reduce an inlet fly ash loading
(with the particle size distribution given in Table 17.6) from 3.1 gr/ft3 to an outlet value
of 0.06 gr/ft3. The flowrate from the coal-fired plant’s boiler is 100,000 acfm. Fractional
efficiency data provided by a vendor is presented subsequently (see Figure  17.1) for
three different types of cyclones (multiclones).
Which type and how many cyclones are required to meet the previously mentioned
specifications? The optimum operating pressure drop is 3.0 in. H2O; at this condition,
the average inlet velocity may be assumed to be 60 ft/s.
256 Introduction to Optimization for Chemical and Environmental Engineers

TABLE 17.6
Particle Size Distribution; Illustrative Example 17.3
Particle Diameter
Range (µm) Weight Fraction (w i)
5–35 0.05
35–50 0.05
50–70 0.10
70–110 0.20
110–150 0.20
150–200 0.20
200–400 0.10
400–700 0.10

100
6" diameter
24" diameter
80
Collection efficiency, %

12" diameter
60

40

20

0
0 20 40 60 80 100
Particle diameter, microns

FIGURE 17.1
Fractional efficiency data; Illustrative Example 17.3.

SOLUTION:
Calculate the required collection efficiency, ER:

 3.1 − 0.06 
ER = 
3.1 
(100)


= 98%
= 0.98
Calculate the average particle size associated with each size range (see Table 17.7).
Table 17.8 provides the overall efficiency, E6, for the 6-in tubes. Since E6 > ER, the 6-in
tubes will do the job. Table 17.9 is generated for the 12-in tubes. Since the overall effi-
ciency, E12 < ER, the 12-in tubes will not do the job. Thus, it will be necessary to use the
6-in tubes for a conservative design.2
Since the outlet tube diameter is one-half the body diameter, the inlet cross-sectional
area (for axial flow) for each 6-in (0.5 ft) tube will be

(
A = 0.785 0.52 − 0.252 )

= 0.147 ft 2
Plant Design Applications 257

TABLE 17.7
Average Particle Size Calculations
Particle Diameter Average Particle
Range (µm) Diameter (µm)
5–35 20
35–50 42.5
50–70 60
70–110 90
110–150 130
150–200 175
200–400 300
400–700 550

TABLE 17.8
Overall Efficiency Calculations for 6-in Tubes
Average Particle Weight Efficiency for Ei (w i) for 6-in
Diameter (µm) Fraction (w i) 6-in Tubes (%) Tubes (%)
20 0.05 89 4.45
42.5 0.05 97 4.85
60 0.10 98.5 9.85
90 0.20 99 19.8
130 0.20 100 20
175 0.20 100 20
300 0.10 100 10
550 0.10 100 10
E6 = 98.95

TABLE 17.9
Overall Efficiency Calculations for 12-in Tubes
Average Particle Weight Efficiency for Ei (w i) for 12-in
Diameter (µm) Fraction (w i) 12-in Tubes (%) Tubes (%)
20 0.05 82 4.1
42.5 0.05 93.5 4.67
60 0.10 96 9.6
90 0.20 98 19.6
130 0.20 100 20
175 0.20 100 20
300 0.10 100 10
550 0.10 100 10
E12 = 98.95

Since the velocity in each tube is 60 ft/s, the number of tubes, n, is given by

(60)(60)(0.147 )( n) = 100, 000

Solving for n,

n = 190 tubes
258 Introduction to Optimization for Chemical and Environmental Engineers

Thus, 190 tubes are required in this multiple-cyclone unit. A 15 × 15, 14 × 14, or 12 × 16
design is recommended.

17.4  Minimizing the Cost of a Batch Plant Operation 3,4

Analyses of some batch cycle operations can be carried out conveniently by using the
time for one cycle as a basis. Equations similar to those provided below can be developed,
including the total annual cost or annual rate of production.

Total annual cost (TAC ) = (operating and shutdowncosts cycle) (cycles year )
(17.1)
+ annualized fixed costs

Annual Production ( AP ) = ( production cycle)(cycles year ) (17.2)

operating + shutdown time year

Cycles Year ( CY ) = (17.3)
operating + shutdown time cycle

Illustrative Example 17.4

Consider the following application. An explosive chemical is being produced via a batch
operation. Each cycle consists of the operating time necessary to complete the reaction
plus a total time of 1.4 hours for charging this feed and discharging the product. The
operating time per cycle is equal to 1.4(PPB)0.33 h, where PPB is the pounds per product
produced per batch. The operating costs during the operating period are $20 per batch,
and the costs during the discharge-charge period are $15 per hour. The annual fixed costs
(AFC) are a function of the size of the batch where

AFC = 300 (PPB )

0.75
; $ (17.4)

In addition, raw-material and miscellaneous costs amount to $300,000 per year.

Determine the cycle time for the condition of minimum total cost per year. The plant
operates 24 hours per day for 300 days per year. The annual production is 1.2 million lb
of product.

SOLUTION:

annual production 1, 200, 000

Cycles Year = = ; cycles/year (17.5)
production batch PPB

Cycle time = operating + shutdown time = 1.4 (PPB )

0.33
+ 1.4 h r (17.6)

Operating + shutdown costs cycle = ( 20 ) 1.4 (PPB ) + (15)(1.4 ) ; $ /cycle (17.7)

0.33

Annual fixed costs = 300 (PPB )

0.75
+ 300, 000; $/year (17.8)
Plant Design Applications 259

(
Total annual costs (TAC ) = 28 (PPB )
0.33
) 1, 200, 000 
+ 21 
 PPB 
(17.9)
+300 (PPB )
0.75
+ 300, 000; $

The total annual cost is a minimum where d(TAC)/d(PPB) = 0. By performing the dif-
ferentiation, one obtains

PPB = 2432 lb per batch

This result could also have been obtained by plotting total annual cost versus PPB and
determining the value of PPB at the point of minimum annual cost. For conditions of
minimum annual cost and 1 million lb/year production,

Cycle time = (1.5)( 2432)

0.33
+ 1.4 = 11.9 h r (17.10)

In addition,

 1, 200, 000 
Total time used per year = (11.6 )  = 5872 hr
 2432 

Total time available per year = ( 300 )( 24 ) = 7200 hr

Thus, for conditions of minimum annual cost and a production of 1.2 million lb per
year, the available operating and shutdown time would be sufficient.

17.5 Ventilation Models with System Variables

Illustrative Example 17.5

Refer to Illustrative Example 13.5.

1. Obtain the equation describing the concentration in the room as a function of

time if r = −kc. Note once again that the minus sign is carried since the chemical
is disappearing.
2. If r = −k, discuss the effect on the resultant equation if k is extremely small, that
is, k ⟶ 0.
3. For part 1, discuss the effect on the resultant equation if k is extremely small,
that is, k ⟶ 0.
4. If r = 0, qualitatively discuss the effect on the final equation if the volumetric
flow rate, v0 varies sinusoidally.
5. If r = 0, qualitatively discuss the effect on the final equation if the inlet concen-
tration, c0 varies sinusoidally.

1. If r = −kc

dc c0 − c c c
= − kc = 0 − − kc (17.11)
dt τ τ τ
260 Introduction to Optimization for Chemical and Environmental Engineers

c0  kτ + 1 
= − c
τ  τ 

dc
= dt
c0 − c ( k τ + 1)

c t
dc dt

∫ci
c0 − c ( k τ + 1)
=
∫τ
0

 1   c0 − (1 + k τ ) c  t
− ln  =
 1 + k τ   c0 (1 + k τ ) ci  τ

 c0 − (1 + k τ ) c   t
−   (1+ k τ )
 τ
 c 1 + kτ c  = e
 0( )i
Rearranging yields

 c  −   (1+ k τ ) 
 t  t
−   (1+ k τ )
 τ
c = ci e +  0  1 − e  τ   (17.12)
 1 + kτ  
 

2. If r = −k but k = 0,

 t
−   −  
 t

c = ci e  τ
+ (c0 − k τ ) 1 − e  τ  
 

 t  t
−  − 
 τ  τ
= ci e + c0 − c0 e

Rearranging yields

 t
− 
= c0 + (ci − c0 ) e  τ
(17.13)

3. If k ≈ 0 in (1),

 c  −   (1+ k τ ) 
 t  t
−   (1+ k τ )
 τ
c = ci e +  0  1 − e  τ  
 1 + kτ  
 
 t  t
−  − 
 τ  τ
= ci e + c0 − c0 e (17.14)

 t
− 
= c0 + (ci − c0 ) e  τ

As expected, the results of (2) and (3) are identical.

Plant Design Applications 261

4. If v0 varies and τ varies, solving the equation becomes more complex. Variations
need to be included in the describing equation

dc c0 − c
= + r (17.15)
dt τ
This may require a numerical solution.
5. If both c0 = c0(t) and v0 = vd(t), the solution in part (4) again applies.

17.6  Plant Structure Design5

When a structure is in the process of being designed and developed, there are multiple
considerations that come into play. One of the authors(5) would argue that the most impor-
tant consideration is safety. The safety of a structure is directly related to its ability to
withstand loads. These loads include weight of materials used, weather loads (i.e., snow,
wind, rain, etc.), weight of furniture and materials within the structure, and many more.
However, both authors would also argue that the second most important factor is cost.
Structures can be designed to last indefinitely if budget was not a consideration, but in
today’s economy the cost of a design often determines if the design will be implemented.
With this in mind, the up-and-coming engineer should seek to optimize the safety and cost
of any project with which they are involved.

Illustrative Example 17.6

Behan Structural Engineers (BSE) has been hired to provide the design for a new nano-
technology plant. The proposal requests a closed rectangular structure to enclose a plant
producing toxic nanoparticles. The structure is to have a volume of 16,000 ft3, but the
perimeter of the base should be no more than 220 ft. The depth is to be no more than 60
ft, and the width no more than 80 ft. Moreover, the width should not exceed three times
the depth, and the height is to be no more than two-thirds of the width. The cost of the
structural material of which the three sides and roof are to be constructed is $40/ft2.
Provide a design of the structure that minimizes the cost of the structural material.

SOLUTION:
As shown in Figure 17.2, x1 = depth (ft), x2 = width (ft), and x3 = height (ft). The material
cost is

Roof = 40 x1x2 ; $ (17.16)

Front and back wall = 40 x2 x3 ; $ (17.17)

Sidewall = 80 x1x3 ; $ (17.18)

The objective function for the cost C is given as

C = 40 x1x2 + 40 x2 x3 + 80 x1x3 (17.19)

262 Introduction to Optimization for Chemical and Environmental Engineers

x3 = height

x1 = depth

x2 = width

FIGURE 17.2
Rectangular structure; Illustrative Example 17.6.

There are also six constraints.

Volume = x1x2 x3 = 16, 000 ft 3 (17.20)

The inequality constraints will include the perimeter limitation,

Perimeter: 2 ( x1 + x2 ) ≤ 220 ft (17.21)

Width

x2 ≤ 3 x1 (17.22)

Height

2
x3 ≤ x2 (17.23)
3
Variable limits

0 ≤ x1 ≤ 60 ft (17.24)

0 ≤ x2 ≤ 80 ft (17.25)

The optimum design solution is left as an exercise for the reader. (Hint: the values gen-
erated from Excel were [in ft] x1 = 13.9, x2 = 41.6, and x3 = 27.9).

References
1. D. Kauffman, Process Synthesis and Design, A Theodore Tutorial, Theodore Tutorials, East
Williston, NY, originally published by USEPA/APTI, RTP, NC 1992.
2. L. Theodore, Air Pollution Control Equipment Calculation, John Wiley & Sons, Hoboken, NJ, 2008.
3. M. Peters, Plant Design and Economics for Chemical Engineers, McGraw-Hill, New York City, NY,
1958.
4. L. Theodore, Chemical Engineering: The Essential Reference, McGraw-Hill, New York City, NY,
2014.
5. K. Behan, personal notes, University of Buffalo, Buffalo, NY, 2017.
 Part IV

Select Optimization Applications

As one would conclude from the title of this Part, it is concerned with applications. Two
chapters complementing the presentation for this Part contain five sections that are loosely
based on the contents of Parts II and III; chapter and section titles are listed below. The
third (and last) chapter is concerned with advanced optimization term projects; four sec-
tions compliment this presentation.
Chapter 18: Select Environmental Engineering Applications

18.1 Air Management

18.2 Water Management
18.3 Solid Waste Management
18.4 Health Risk Assessment
18.5 Hazard Risk Assessment

Chapter 19: Select Chemical Engineering Applications

19.1 Fluid Flow

19.2 Chemical Reactors
19.3 Heat Exchangers
19.4 Mass Transfer Operations
19.5 Plant Design

Chapter 20: Advanced Optimization Term Projects

20.1 Sulfuric Acid Process

20.2 Optimum Indoor Ventilation Flowrate
20.3 Fluid Transportation System
20.4 Mass Transfer Application
264 Introduction to Optimization for Chemical and Environmental Engineers

It should be noted that last two sections of Chapter  20 were drawn (with permission)
from the classic work of Happel’ s 1959 John Wiley & Sons text, Chemical Process Economics .
Happel’ s original units and notation were retained in these presentations.
18
Select Environmental Engineering Applications

Five sections complement the presentation in this chapter:

18.1 Air management

18.2 Water management
18.3 Solid waste management
18.4 Health risk assessment
18.5 Hazard risk assessment

18.1 Air Management
Gaseous (and particulate) pollutants discharged into the atmosphere can be controlled.
The four generic devices for gases include absorbers, adsorbers, scrubbers, and incin-
erators. Gas absorption, as applied to the control of air pollution, is concerned with the
removal of one or more pollutants from a contaminated gas stream by treatment with a
liquid. The necessary condition is the solubility of these pollutants in the absorbing liquid.
Design and performance characteristics of any particular control system are important
considerations. Optimizing the performance of an air pollution control device is often a
standard operating practice.

Illustrative Example 18.1

Consider the absorber system shown in Figure 18.1. It was designed to operate with a
maximum pollutant/solute discharge concentration of 50 ppm. However, once the unit
was installed and running, the unit operated with a discharge of 60 ppm. Rather than
purchase a new unit, what options are available to bring the unit into compliance with
the specified design concentration?

SOLUTION:
Theodore1 proposed a host of possible solutions. Many of them are listed here:

1. Use a smaller packing size. A different size may produce a better packing
factor.
2. Use a different type of packing. A different packing type may also produce a
better packing factor.
3. Make sure there is no channeling inside the column. The packing should be
randomly distributed and there should be no open spots where water will
accumulate.

265
266 Introduction to Optimization for Chemical and Environmental Engineers

50 ppm (design)

90°F water

1 atm

rasching rings)
5 ft (one inch
1000 acfm gas
1000 ppm (solute)
90°F

FIGURE 18.1
Faulty design?

4. Increase the water flow entering the column. In effect, the slope of the operat-
ing line will increase1 and provide a greater driving force for mass transfer.
Thus, there will be a greater mass transfer between the liquid and gas phase
and the efficiency of the column will increase.
5. Use a liquid other than water to scrub the gas. Another liquid may cause better
mass transfer between the gas and liquid phase.
6. Check to see if flooding is occurring. If the column is near the flooding point,2
there will not be much mass transfer. The water will be pushed up the column
by the gas rather than the water coming down the column.
7. Lower the initial concentration in the carrier gas (if possible).
8. Increase the height of the bed. This will provide more surface area for mixing
and more time in the absorber.
9. Increase the pressure. This will reduce the slope of the operating line1 and lead
to a greater driving force.
10. Decrease the temperature. This will reduce the slope of the operating line and
lead to a greater driving force.
11. Increase the pressure and decrease the temperature.
12. Additional packing height could be added at the top of the column. This would
increase the amount of the solute removed from the gaseous phase.
13. Add sprays at the bottom of the column.
14. Modify the process producing the problem. Since the details of this process
are not included in the problem statement, no specific recommendations can
be mentioned in this discussion.
15. Finally, before considering changes to the system, one should undertake a thor-
ough inspection of the unit and surrounding components. The emission moni-
toring system should be recalibrated. All valves, fittings, and pipes should be
checked for plugging or leaks. The liquid distributor should be checked to
make sure it is functioning properly. The distribution of the packing should be
inspected to make sure it is as uniform as possible. Any problems encountered
during this inspection should be corrected immediately. Following the main-
tenance check, the performance of the unit should be reevaluated.
Select Environmental Engineering Applications 267

The reader should note that some of the previously mentioned recommendations could
lead to higher pressure drops and potential problems with the flow. An example of this
problem would be the implementation of suggestion (1). The reader is left the exercise of
determining what other steps could lead to flow/pressure drop problems. Hint: There
are at least six suggestions that fall into the proposed solution, including firing the envi-
ronmental engineer who designed the absorber.
A similar situation could exist with a gaseous adsorption unit. Once again, there are
options available to bring the unit into compliance with the specified design concen-
tration. Some of Theodore’s1 recommendations are listed below. Assume the adsorber
packing is activated carbon and there is no liquid flow.

1. Check/increase the depth of the adsorption bed. Increasing the depth of the
bed will increase the removal capacity of the column.
2. Change the adsorbent type of size. Use an adsorbent with a higher saturation
capacity.
3. Reducing the flow will increase the residence time and should allow for more
adsorption.
4. Decrease the inlet temperature. This will favor the adsorption equilibrium
process.
5. Increase the system pressure. This will also favor the adsorption equilibrium
process.
6. Regenerate the adsorbent for a longer period.
7. Replace the adsorbent.
8. Make up any adsorbent lost to carryover (as necessary).
9. Consider changing the flow direction.

Again, pressure drop considerations should be included in the analysis.

What if the pollution control device is a 24-year-old perforated plate distillation col-
umn2 that is no longer delivering the degree of separation required for a wastewater
process. In effect, the issue with the current operation is its inability to perform the
separation at its optimum. Rather than replace the unit, you have been asked to recom-
mend what other possible steps can be taken to the existing unit to get it back “on line.”
Here are some recommendations:1

1. The first thing that needs to be investigated is what is the current state of the
column? How was it originally designed? Were safety margins included in the
design that allow for some “tweaking” of the actual structure of the column?
How much change in pressure (drop) is allowed? What will the column mate-
rials stand up to?
2. A possibility for the column is changing the flow rate of gas through the col-
umn. If the gas is made to pass through the column at a higher velocity, it
should result in more foaming of the liquid on the trays. This will increase the
mass transfer area. This will also help to separate the products. Unfortunately,
this could create new problems with flooding and entrainment. This needs to
be examined very carefully before it is implemented.
3. It also might be possible to change the reboiler and condenser2 conditions. It is
possible after 24 years of service that these two devices are not working at their
optimum settings. An examination should be made of their contribution to the
failure to perform the desired separation. Is the column failing because these
devices are no longer functioning? If so, then they need to be either retrofitted
or replaced with newer models.
4. The composition and condition of the feed stream also need to be examined.
Is it the same as what the column was originally designed to separate? Has the
process been altered upstream in such a way that the separation is no longer
268 Introduction to Optimization for Chemical and Environmental Engineers

feasible? Perhaps it has changed through the years in such a manner that the
column is no longer designed to deal with the stream; if so, the feed stream
could be altered to adhere to the original conditions to make the column oper-
ate correctly again. Also, the feed tray (location) could be changed. Finally, the
inlet feed quality might also be changed.2
5. Adding trays is the most difficult thing to implement but would probably have
the largest impact. The trays could be added near the top of the column if a
safety margin of headspace had been designed into the column. A possibility
would be the complete removal of all the trays and reinstalling them with a
smaller tray spacing between them. If that is done, the environmental engineer
should consider replacing the trays with a better tray design. Significant work
has gone into tray design in the past half-century; improvements have been
made to optimize their efficiency and capabilities.
6. Finally, it is possible that the column needs cleaning and that too could opti-
mize/improve the column’s performance.

18.2 Water Management
Shipping and storage facilities often require a majority of the total land area at chemical
and petrochemical plants and also represent a very large total capital investment. Though
their design is not usually so sophisticated as that of the actual process units, they often
represent the largest potential source of safety and environmental problems because of the
large inventories involved.
Gaseous feeds, products, and fuels may be shipped to and from a plant by pipeline, pres-
surized truck, or tank car, or in individual cylinders. Pipelines are by far the least expen-
sive means of shipment of gases, though they have very high initial capital costs. Gases
are the least practical materials to store at a plant site. They may be stored in large storage
tanks or in trucks, tank cars, and individual cylinders. Underground, hollowed-out salt
domes are sometimes used where very large amounts of storage are needed, such as for
natural gas, ethane, and ethylene.
Liquid feeds, products, and fuels may be shipped by pipeline, tanker ship barge, tank
car, truck, or, occasionally, individual small containers. Pipelines and very large ocean-
going tankers are the least expensive shipping means. Barges are the next most expensive,
followed by rail cars and trucks. Liquid feeds, products, intermediates, and by-products
are commonly stored in large tanks in a “tank farm” adjacent to the process units. Tanks
must be surrounded by dikes or other barriers to prevent loss of the liquid if a tank should
rupture. Vent condensers, floating roofs, and other features are used with volatile liquids
to prevent release of their vapors to the atmosphere.
Solid feeds, products, and fuels are commonly shipped by ocean freighter, barge, rail car,
and truck. Storage may be in silos or covered hoppers, in sacks, or special containers, or
simply in piles out in the open.
As noted in the previous chapter, feeds, products, intermediates, and fuels at chemical and
petrochemical plants are stored in tanks, which are usually located in a “tank farm” adjacent to
the process plant area. The tank capacities are most often expressed in gallons or barrels (one
barrel equaling 42 gallons). The individual tanks may range in size from a thousand gallons or
less to several million gallons. Whenever possible, liquids are stored at ambient temperature
and pressure. Volatile, hazardous, and toxic liquids may have to be stored under pressure. In
addition, they often require vapor recovery systems or other devices to prevent releases to the
Select Environmental Engineering Applications 269

atmosphere as the tanks are filled and emptied (defined as working losses). Liquids that are
very viscous at ambient temperature or that would solidify at ambient temperature are kept
in heated tanks. Standard practice calls for each tank to be surrounded by a dike or berm suf-
ficiently high to contain all the liquid stored in a tank in case it should rupture. Firefighting
equipment is usually permanently located near tanks containing flammable liquids.3
Storage tanks can present major risks both in terms of fire and environmental pollution. In
both cases, this is primarily due to the large inventories of materials that could be involved
in any accident. There has been a major effort in industry in the last century to reduce the
amount of storage at process plants, especially for flammable and/or toxic materials.3
From a water pollution perspective, the major problem with these tank farms falls under
the heading of operation, storage, and transportation. The objective of any water quality
management program is to control the discharge of any liquid pollutants so that water
quality is not degraded to an unacceptable extent below the natural background level.
The impact of the pollutant on water quality must be predicted and compared to the back-
ground water quality that would be present without human intervention.

Illustrative Example 18.2

Consider a chemical plant that uses two liquid feeds of different densities as provided
in Table 18.1. It produces four different liquid chemical products of varying densities (ρ)
following chemical reaction and separation, as presented in Table 18.2. The plant stor-
age requirements call for maintaining 4–5 weeks supply of each feed, 4–6 weeks supply
of products A, B, and C and a maximum of 1–2 weeks supply of product D. The plant
operates year-round, but each tank must be emptied once a year for a week of mainte-
nance. Tanks are normally dedicated to one feed or product and one or two could be
used as “swing” tanks; however, one day of cleaning is required between uses with
different liquids.
You have been assigned the task of specifying a set of tanks from the “standard” sizes
given in Table 18.3 to minimize the plant’s tank needs while at the same time reducing any
potential water pollution problems3 that can develop from unnecessary oversized tanks.

TABLE 18.1
Feed Data
Feed 1 110,000 lb/day ρ = 49 lb/ft3
Feed 2 50,000 lb/day ρ = 68 lb/ft3

TABLE 18.2
Product Data
Product A 40,000 lb/day ρ = 52 lb/ft3
Product B 25,000 lb/day ρ = 62 lb/ft3
Product C 10,000 lb/day ρ = 52 lb/ft3
Product D 95,000 lb/day ρ = 47 lb/ft3

TABLE 18.3
Tank Data
Standard Tank Sizes (gallons)
2,800 11,200 56,100 561,000
5,600 16,800 140,000 1,123,000
8,400 28,100 281,000
270 Introduction to Optimization for Chemical and Environmental Engineers

TABLE 18.4
Storage Requirements
Material Lb/day Gal/day Days Gals Storage Required
Feed 1 110,000 16,800 28–35 470,000–587,800
Feed 2 50,000 5,500 28–35 154,000–192,500
Product A 40,000 5,750 28–42 161,100–241,700
Product B 25,000 3,020 28–42 84,500–126,700
Product C 10,000 1,440 28–42 40,300–60,400
Product D 95,000 15,120 7–14 105,800–211,700

SOLUTION:
Unfortunately, there is no single, simple method for determining the optimum mix of
storage tanks for a chemical plant. Most often, estimates are made of the minimum and
maximum amounts of feeds, intermediates, and products that must be kept on hand.
Some additional allowance is then made to permit periodic cleaning and maintenance
of the tanks. One solution and some calculation details follow.3
First, determine the maximum and minimum amounts of each material to be stored
(see Table 18.4). The conversion to gallons requires dividing the rate in lb/day by the
density in lb/ft3, and then multiplying by 7.481 gal/ft3.
For each material, select a set of tanks:

Feed 1: Use 2 281,000-gallon tanks

Feed 2: Use 4 56,100-gallon tanks
Product A: Use 2 140,000-gallon tanks
Product B: Use 2 56,100-gallon tanks
Product C: Use 1 56,100-gallon tanks
Product D: Use 2 140,000-gallon tanks

This set is 5% short on the maximum of Feed 1, 11% short on Product B, and 7% short on
Product C. For most situations, this would be acceptable. Select spare and/or “swing”
tanks to provide for maintenance:

Feed 1: Use an additional 281,000-gallon tank

Feed 2 and all products: Use one 56,100-gallon “swing” tank

The recommendation below provides adequate auxiliary storage for maintenance

periods:

Total tanks required:

3 281,000-gallon tanks
4 140,000-gallon tanks
8 56,100-gallon tanks

The number of tanks required will be quite large in this application. If market forces,
such as environmental regulations and fluctuating demand require this much storage,
they may all be necessary. More modern commercial operations, such as “just-in-time”
or “on demand” manufacturing, call for reducing in-plant inventory to the absolute
minimum possible, particularly with toxic, flammable, and explosive3 liquid waste. To
Select Environmental Engineering Applications 271

further complicate matters, the minimum number of tanks may not always be optimum
if the tanks are extremely large. The reader should note that several smaller tanks may
cost somewhat more initially, but they offer more flexibility in use.

18.3 Solid Waste Management

Coal is a combustible black or brownish-black sedimentary rock normally occurring in
layers or veins referred to as coal beds or coal seams. It is composed primarily of elemental
carbon along with variable quantities of other elements – chiefly hydrogen, with smaller
quantities of sulfur and even smaller quantities of oxygen and nitrogen. Coal is hetero-
geneous and can vary in chemical composition with location. In addition to the major
organic ingredients (carbon, hydrogen, and oxygen), coal also contains impurities. The
impurities that are of major concern are ash and the aforementioned sulfur. The ash results
from mineral or inorganic material produced during coalification.4 Ash sources include
inorganic substances, such as silica, which are part of the chemical structure materials.
Dissolved inorganic ions and mineral grains found in swampy water are also captured by
the organic matter during early coalification.4
After coal comes out of the ground, it typically goes on a conveyor belt to a preparation
plant that is located at or near the mining site. The plant cleans and processes the coal to
remove dirt, rock, ash, free sulfur, and other impurities, which increases the heating value
of the coal to be transported.
At a typical power plant, the coal is combusted, liberating its heating value. This occurs
in a boiler where the released energy is transferred to steam at an elevated pressure.
The high-pressure steam is passed into a turbine containing thousands of propeller-like
blades. The steam pushes these blades, causing the turbine shaft to rotate at high speed. A
generator is mounted at one end of the turbine shaft and consists of carefully wound wire
coils. Electricity is generated when these are rapidly rotated in a strong magnetic field.
After passing through the turbine, the steam is condensed and returned to the boiler to be
heated once again. The electricity generated is transformed into voltages as high as 400,000
V, with transmission provided via power line grids. When it nears the point of consump-
tion (e.g., industry or homes), the electricity is transformed down to the safer 100–250 V
systems used in most markets.4
Improvements continue to be made in conventional power station design, and new com-
bustion technologies are being developed. These allow more electricity to be produced
from less coal – known as optimizing the thermal efficiency. Efficiency gains in electricity
generation from coal-fired power stations also result in a reduction of harmful emissions.
Optimizing the performance of coal-fired power plants has been the focus of considerable
efforts by the coal industry because of problems associated with waste disposal, including
siting a disposal facility, transporting the waste, and costs associated with implementation
and operations.

Illustrative Example 18.3

Consider the following application, drawn from the classic work of Happel.5,6 An indus-
trial boiler house contains five coal-fired boilers, each with a nominal rating of 300 boiler
horsepower. If economically justified, each boiler can be operated at a rating up to 300%
272 Introduction to Optimization for Chemical and Environmental Engineers

of nominal. Due to the growth of manufacturing, it is necessary to install additional

boilers. Referring to the data and notes given below, determine the percent of nominal
rating at which the present boilers should be operated to optimize the operation of the
boiler house and reduce the generation of solid waste ash.
Data and notes: The cost of fuel, including the cost of handling coal and removing
cinders, is $7 per ton. The coal has a heating value of 14,000 Btu per lb. The overall effi-
ciency of the boilers, from coal to steam, has been determined from tests of the present
boilers operated at various ratings; this information is provided in Table 18.5.
Annual fixed charges, C, for each boiler are given by the equation:

C = 14, 000 + 0.04R 2 (18.1)

Assume 8550 hr of operation/yr, five coal-fired boilers, and (1 boiler hp = 33,500 Btu/hr
for R = 100).

The hourly cost of fuel (CF) at $7 per ton is given by

( 5)( 300)  100  ( 33, 500)(7 )

R
 R
CF = = 12.56   (18.2)
 E   E
  (14, 000 )( 2, 000 )
100 
The hourly fixed charges, CA, are

CA =
( 5)(14, 000 + 0.04R1 ) = 8.18 + 0.0000234R2 (18.3)
8, 550
The total output, T, in actual boiler hp developed is given by

 R 
T = 5 ( 300 )  = 15R (18.4)
 100 
Optimum results will be obtained when the sum of the previously mentioned variable
costs, divided by the output, is a minimum. Refer to Table 18.6. As shown in Table 18.6
and in Figure 18.2, the optimum costs per unit output are found at approximately
200% of nominal rating, corresponding to a cost of $14.30 per 1000 boiler hp devel-
oped (33.5 million Btu), or $0.428 per million Btu, exclusive of cost of labor. This
cost is higher than that in a larger power plant in which the fixed charges per boiler
horsepower developed would be less than with these small boilers, and the efficiency
would be higher.

TABLE 18.5
Boiler-Efficiency Data
R 1, % of Nominal Rating E, % Overall Thermal Efficiency
100 75
150 76
175 75
200 74
225 72
250 69
275 65
300 61
Select Environmental Engineering Applications 273

TABLE 18.6
Cost Calculations; Illustrative Example 18.3

CF + CA
(1,000 )
R E R/E CF CA C F+C A T T
100 75 1.33 16.7 8.4 25.1 1500 16.7
150 76 1.97 24.8 8.7 33.5 2250 14.9
175 75 2.33 29.3 8.9 38.2 2625 14.6
200 74 2.70 33.9 9.0 42.9 3000 14.3
225 72 3.12 39.2 9.4 48.6 3375 14.4
250 69 3.62 46.5 9.6 56.1 3750 15.0
275 65 4.23 53.2 10.0 63.2 4125 15.3
300 61 4.92 61.8 10.3 72.1 4500 16.0

20
Dollars per 1000 BHP developed

16
Total, T

12 Coal, CF

4
Fixed charges, CA

0
100 140 180 220 250 300
R, % of nominal rating

FIGURE 18.2
Optimum cost for boiler operation; Illustrative Example 18.3.

18.4 Health Risk Assessment

As mentioned earlier in Chapter 11, people are exposed to risks every day whether
they are aware of it or not. Many of the risks became apparent through various earlier
studies involving cancer, and they have led scientists to utilize health risk assess-
ments (HRAs).7 These HRAs are specifically used to analyze the potential adverse
health effects from the environment, but they are often very complicated to quantify
because of the vast array of potential receptors present. However, HRAs provide an
orderly, explicit, and consistent way to deal with issues in evaluating whether a health
problem exists and what the magnitude of the problem may be. They are used in the
process of risk assessment and can estimate the likelihood that humans or ecological
systems will be adversely affected by chemical or physical agents under a specific set
of conditions.7
274 Introduction to Optimization for Chemical and Environmental Engineers

The term risk assessment is not only used to describe the likelihood of an adverse
response to a chemical or physical agent but has also been used to describe the likeli-
hood of any unwanted event. These include risks, such as explosions or injuries in the
workplace; natural catastrophes; injury or death due to various voluntary activities such
as skiing, skydiving, flying, and bungee jumping; diseases, death due to natural causes;
and, many others. This risk scenario also receives treatment in the industrial application
that now follows.

Illustrative Example 18.4

Theodore Associates has been requested to conduct both a risk health and hazard
assessment at a chemical plant that is concerned with the consequences of two inci-
dents that occur at approximately the same location in the plant and that are defined as
follows:8

1. An explosion resulting from the detonation of an unstable nanochemical.

2. A continuous 240 g/s release of a resulting nanochemical at an elevation of 125
m.

Two weather conditions are envisioned, namely, a northeast wind and a southwest
wind (6.0 mph) with Stability Class B.1 Associated with these two wind directions are
Events IIA and IIB, respectively, defined as follows:

IIA – Toxic cloud to the southwest

IIB – Toxic cloud to the northeast

Based on an extensive literature search, the probabilities and conditional probabili-

ties9 of the occurrence of the defined events in any given year have been estimated by
Dupont Consultants as follows:

P (I ) = 10 −6

P (II ) = 1/33, 333

P (IIA|II ) = 0.33

P (IIB|II ) = 0.67

Note that P(IIA|II) represents the probability that Event IIA occurs given that Event
II has occurred.9 The consequences of Events I, IIA, and IIB, in terms of the number of
people killed, are estimated as follows:

I – All persons within 200 m of the explosion center are killed; all persons beyond
this distance are unaffected.
IIA – All persons in a pie-shaped segment of 22.5° width (downwind of the source)
are killed if the concentration of the toxic gas is above 0.33 μg/L; all persons
outside this area are unaffected.
IIB – Same as IIA.

Thirteen people are located within 200 m of the explosion center but not in the pie-
shaped segment described previously. Eight people are located within the pie-shaped
Select Environmental Engineering Applications 275

segment southwest of the discharge center; five are 350 m downwind, and three are
600 m away at the plant fence (boundary). Another six people are located 500 m away
outside the pie-shaped segment but within the plant boundary. All individuals are at
ground level.
Theodore Associates have been specifically requested to calculate the average
annual individual risk (AAIR) based on the number of individuals potentially
affected as well as the average risk based on all the individuals within the plant
boundary. Their solution is presented subsequently. Hint: Perform atmospheric dis-
persion calculations at various distances from the emission source and combine
these predicted concentrations with consequence information from the problem
statement.

SOLUTION:
Draw a line diagram of the plant layout and insert all pertinent data and information
(see Figure 18.3). An event tree9 for the process is presented in Figure 18.4. First, calcu-
late the probability of Event IIA occurring; also calculate the probability of Event IIB
occurring9 as follows:

 1  1
( )
P (IIA ) = P (II ) P IIA II = 
 33, 333 
(0.33) = 100, 000 = 10 −5 (18.5)

 1  2
( )
P (IIB ) = P (II ) P IIB II = 
 33, 333 
(0.67 ) = 100, 000 = 2 × 10 −5 (18.6)
Perform a dispersion calculation to determine the zones where the concentration of
the chemical exceeds 0.33 μg/L. Assume a continuous emission for a point source.1 To
maintain consistent units, convert wind speed from mph to m/s and concentration from
μg/L to g/m3 as follows:

 miles   ft   1h   m m
u =  6.0
   5280     0.3048  = 2.68
h mile 3600s ft s

13 N

≤200 m

100 m

Number of
5
X
350 m
Individuals at location

3 6
600 m 500 m

FIGURE 18.3
Plant layout and pertinent data; Illustrative Example 18.4.
276 Introduction to Optimization for Chemical and Environmental Engineers

Unstable
chemical

Flammable
Detonation toxic gas
release

NE wind SW wind

Explosion Toxic cloud Toxic cloud

to SW to NW

I IIA IIB

FIGURE 18.4
Event tree for process emission and accident; Illustrative Example 18.4.

 0.33 µg   g  L  g
c=   1 6   10 3 3  = 3.3 × 10 −4 3
 L   10 µg   m  m
These values are then used as input to the Pasquill-Gifford model for centerline, ground-
level concentrations of a continuous source of pollutant at an elevated emission height
of H*, as shown in the following equation:8–12

  1  H *  2  
q 
c ( x , 0, 0; 125) = exp  −  
πσ y σ z u   2  σ z   
  (18.7)
 240 g/s   1  125 m  2 
=  exp  −   
( )
 π σ y (σ z ) ( 2.68 m/s )   2  σ z  

The downwind concentrations can be calculated based on the previous equations. The
concentration results for select downwind distances are provided in Table 18.7. A linear
interpolation of these calculations indicates that the maximum ground-level concentra-
tion (GLC) is approximately 1.01 × 10−3 g/m3 and is located at a downwind distance of
about 800m. In addition, the “critical” zone, that is, where the concentration is above
3.3 × 10−3 g/m3, is located between 475 and 1800 m.
It should be noted that only one “average” weather condition was considered in this
example. However, one often selects the worst-case weather condition that corresponds
with a reasonable probability of occurrence in the location of the site being evaluated.
Employing this worst-case condition produces risk results on the conservative side. An
analysis that includes a full spectrum of wind speeds, directions, and stability classes
would obviously provide a more complete set of risk assessment calculations than is
provided here.
Select Environmental Engineering Applications 277

TABLE 18.7
Downtown Concentration Profile for this Site
x (m) σy (m) σz (m) c (g/m3)
300 47 30 3.43 × 10−6
400 60 41 1.11 × 10−4
500 75 2 4.07 × 10−4
550 80 60 6.78 × 10−4
600 90 65 7.67 × 10−4
700 105 77 9.44 × 10−4
800 120 90 1.01 × 10−3
900 150 110 9.06 × 10−4
1000 170 140 8.04 × 10−4
1500 250 240 4.15 × 10−4
1700 275 275 3.40 × 10−4
2000 300 380 2.37 × 10−4

Now, determine which individuals within the pie-shaped segment downwind from
the source will be killed if either accident (I or II) occurs. Referring to Figure 18.3, 13
individuals within the 200 m radius will die from Accident I. Three individuals located
in the pie-shaped segment and 600 m southwest of the emission source will die from
Accident II. The five individuals located 350 m southwest of the emission source are in
the path of the dispersing plume but are all outside the critical zone. The six individuals
located outside of the pie-shaped impact segment are within the plant boundary but
are not potentially affected by either the explosion or the dispersing plume. The plant
health risk and hazard risk assessment line diagrams7,10 are presented in Figures 18.5
and 18.6, respectively.
The total annual deaths (TAD) for the process if the accident occurs are therefore

TAD = 13 + 3 = 16 deaths / yr

The total annual risk (TAR) is obtained by multiplying the number of people in each
impact zone by the probability of the event affecting that zone and summing the result.
Thus,

( ) ( )
TAR = (13 ) P (1) + ( 3 ) P (IIA ) = (13 ) 10 −6 + ( 3 ) 10 −5 = 4.3 × 10 −5

The average annual individual risk (AAIR) is obtained by dividing this result by
the number of people in the impacted zone. The average annual risk (AAR) is calcu-
lated based only on the “potentially affected” people. Since 21 people are “potentially
affected,” that is, are within the impact area of the explosion or are in the path of the
dispersing plume, the AAR is determined to be:

4.3 × 10 −5
AAR = = 2.05 × 10 −6
21
The AAIR is based on all the individuals within the plant boundary. For this case, this
AAIR is now based on 27 rather than 21 individuals. Thus,

4.3 × 10 −5
AAIR = = 1.6 × 10 −6
27
278 Introduction to Optimization for Chemical and Environmental Engineers

Nano-emission in wrong direction at

annual probability of 1 × 10–5

Health problem
identification

Continuous emission of toxic

Exposure to nanomaterial
Three people
nanomaterial
Human exposed to
release at Dose-response or Individuals
response to Exposure emission at
concentration toxicity exposed to
nanomaterial assessment nanomaterial
≥0.33 mg/L assessment emission
exposure concentration
results in
≥0.33 mg/L
death
How many will die
annually from
nanomaterial exposure?

Risk
characterization

Three individuals will die annually from the nanomaterial release

The annual risk of death from the release = 3 × 10–5

FIGURE 18.5
Plant health risk assessment.

Chemical explosion

Hazard
identification

Explosion of chemical

Annual Annual How many at 13 at

probability of Accident probability of explosion Accident explosion
explosion = probability explosion center will be consequence center will
1 × 10–6 evaluation
occuring affected die

Annual risk of death from

chemical explosion

Risk
determination

Annual risk of death = 13 × 10–6 for individuals at

explosion center

FIGURE 18.6
Plant hazard risk assessment.
Select Environmental Engineering Applications 279

18.5 Hazard Risk Assessment

Risk assessment of accidents requires the calculation of the probability that an accident
will occur and also the severity of the consequences of an accident. Consequences may
include damage to the surrounding environment, financial loss, injury, or loss of life. Risk
assessment of accidents (or hazard risk assessment of accidents (HZRA)) provides an effec-
tive way to help ensure that a mishap either does not occur or reduces the likelihood of
severe consequences as a result of an accident. The result of the risk assessment allows
concerned parties to take precautions to prevent an accident before it happens.7
There are other classes of environmental health risks that do not pertain to chemicals
but are an integral part of HZRA. For example, health problems can arise immediately/
soon after a hazard, such as a hurricane or earthquake that can leave local inhabitants
without potable water for an extended period.
Describing failure and failure rate has received significant attention by the environmen-
tal technical community. One of the equations that has withstood the test of time is the
Weibull distribution.4, 9 Simply put, the Weibull distribution describes failure rate as a func-
tion of time. It has served the technical community for over 50 years; the chemical indus-
try, refineries, the Pentagon, NASA, and so on, have all been the beneficiaries. Unlike the
exponential distribution9,11, the failure rate of equipment frequently exhibits three stages: a
break-in (BI) stage with a declining failure rate, a useful life stage characterized by a fairly
constant failure rate, and a wear-out (WO) period characterized by an increasing failure
rate. Many industrial parts and components follow this path. A failure rate curve exhibit-
ing these three phases (see Figure 18.7) has been referred to as a bathtub curve.
Weibull introduced the distribution, which bears his name, principally on empirical
grounds to represent certain life-test data. The Weibull distribution provides a mathemati-
cal model of all the previously mentioned three stages of the bathtub curve. This is now
discussed. An equation describing the failure rate that reflects all of the previously men-
tioned three stages of the bathtub curve is9,11

Z ( t ) = αβtβ−1 ; t > 0 (18.8)

where α and β are constants, referred to by some as shape parameters or curve-fitting

parameters. For β < 1, the failure rate Z(t) decreases with time. For β = 1, the failure rate
Failure rate

Break-in (BI) Useful life Wear-out (WO) Time

FIGURE 18.7
Bathtub curve.
280 Introduction to Optimization for Chemical and Environmental Engineers

is constant and equal to α. For β > 1, the failure rate increases with time. Using Equation
18.8 to translate the assumption about failure rate into a corresponding assumption about
the probability distribution function (pdf)12 of t, time of failure, one obtains Equation 18.9,
which primarily defines the pdf of the Weibull distribution. (The exponential distribution
is a special case of the Weibull distribution with β = 1.)

t 


0
∫
f ( t ) = αβtβ−1 exp  αβtβ−1dt 

 (18.9)

( )
= αβtβ−1 exp −αtβ ; t > 0; α > 0,β
β>0
The variety of assumptions about failure rate and the probability distribution of time to
failure that can be accommodated by the Weibull distribution make it especially attrac-
tive in describing failure-time distributions in industrial and process plant applications.
Estimating the coefficients in the Weibull distribution can be accomplished using a graph-
ing procedure developed by Bury.13
The Weibull distribution has played an important role in helping to reduce and/or elimi-
nate accidents and other hazards from occurring in both the workplace and in everyday
life. Shaefer and Theodore11, and Theodore and Dupont7 have provided both development
material and numerous illustrative examples that are concerned with this distribution. It
has also served as the basis for a host of technical papers prepared by one of the authors.
In recent years, the Weibull distribution has come under fire, due primarily to the
efforts of one of the authors. The last 6 years has provided an opportunity for Dupont and
Theodore14, along with Ricci and others14–17, to carefully analyze the merits and limitations
of the Weibull distribution. This has led to the development of the DRaT models. Details
are given subsequently.
As noted in the previous section, the general two-coefficient Weibull model is repre-
sented by an equation that can be applied to three failure rate periods representing three
failure mode stages. As such, the model consists of six coefficients – two for each of the
three failure mode stages. Theodore and Dupont14 viewed this six-coefficient relationship
as both cumbersome and unnecessary. After some deliberation, it was decided to employ
a new, simpler approach to represent failure behavior – specifically the failure-time (as
opposed to failure rate-time) relationship depicted in Figure 18.8. After even more delib-
eration and analysis, these authors settled on four models that are based on the melding of
either a power function (P) or an exponential relationship (E) for the BI period with either
a power function or exponential relationship for the WO period – the sum of which results
in a curve as shown in Figure 18.8. There are therefore four combinations of the power
function and exponential term equations: BI(P) – WO(P), BI(E) – WO(E), BI(E) – WO(P), and
BI(P) – WO(E), and the corresponding equations resulting from these combinations are
defined as the DRaT II, DRaT III, DRaT IV, and DRaT V models, respectively.
Note that the general relationship for the power function for either the BI or WO period
takes the form

P ( t ) = P1 + P2t + P3t 2 + P4t 3 (18.10)

The general form of the exponential function takes the form of either

( )
E ( t ) = A1 1 − e − A2t (18.11)
Select Environmental Engineering Applications 281

Number of failures, N

Time, t

FIGURE 18.8
Failure-time relationship for the Weibull distribution.

for the BI period, and

( )
E ( t ) = B1 e B2t − 1 (18.12)

for the WO stage.

The failure-time functional relationships for BI and WO periods are represented by
g(t)and h(t), respectively, for the development to follow, with the sum of the two resulting
in the DRaT model, f(t), that is,

f ( t ) = g ( t ) + h ( t ) (18.13)

For example, if both the BI and WO periods are represented by an exponential function
(see also Figure 18.9) it would appear as

( )
g ( t ) = A1 1 − e − A2t (18.14)

and

( )
h ( t ) = B1 e B2t − 1 (18.15)

The sum of Equations 18.14 and 18.15, presented in Figure 18.9, represents the failure model
previously defined as the DRaT III model, that is,

( ) ( )
N ( III ) = g ( t ) + h ( t ) = A1 1 − e − A2t + B1 e B2t − 1 ; N = number of failures (18.16)

Specific details on each of the remaining DRaT models are provided in the next three
subsections.

18.5.1 DRaT II Model
This model assumes that BI(P) and WO(P) relationships apply. Thus,

N ( III ) = N ( BI ) + N ( WO ) (18.17)
282 Introduction to Optimization for Chemical and Environmental Engineers

g(t) = A1(1 – e–A2t)

(a)

h(t) = B1(eB2t – 1)

(b)

f(t) = g(t) + h(t)

g(t)

h(t)

(c)

FIGURE 18.9
DRaT model II representation. (a) BI period, (b) WO period, and (c) combined BI and WO period.

and

N ( II ) = A1 + A2t + A3t 2 + A4t 3 (18.18)

Interestingly, this model only contains three coefficients.

18.5.2 DRaT IV Model
This model assumes that BI(E) and WO(P) relationships apply; thus,

( )
N ( BI ) = A1 1 − e − A2t (18.19)

and

N ( WO ) = B1 + B2t + B3t 2 ; B1 = 0 (18.20)

Select Environmental Engineering Applications 283

so that

( )
N ( IV ) = A1 1 − e − A2t + B2t + B3t 2 (18.21)

This model contains four coefficients.

18.5.3 DRaT V Model
The last model assumes BI(P) and WO(E) relationships apply. Thus,

N ( BI ) = A1 + A2t + A3t 2 ; A1 = 0 (18.22)

and

( )
N ( WO ) = B e B2t − 1 (18.23)

so that

( )
N ( V ) = A2t + A3t 2 + B e B2t − 1 (18.24)

Similar to the last two models, this model also contains four coefficients.
Three improvements in the proposed DRaT models relative to the Weibull model become
immediately apparent:

1. The DRaT model requires either three of four (not six) coefficients to be estimated.
2. The DRaT model is continuous over the entire time range, as opposed to the
Weibull model that is evaluated separately over three compartmentalized failure
stages.
3. The requirement of a constant failure rate period in the Weibull model has been
replaced by a more realistic failure rate that can continue to increase slightly with
time during the supposed constant failure rate period.

The reader should note once again that the failure rate for most applications is a calcu-
lated quantity obtained from the number of failures (N) versus time (t) data. The failure
rate at a specified time is thus approximately equal to the slope (or derivative) of N versus
t at the time point in question. Theodore and Ricci18 provide six different numerical dif-
ferentiation procedures to calculate this derivative, that is the failure rate, and the reader is
directed to that reference for more details.

References
1. L. Theodore, Air Pollution Control Equipment Calculations, John Wiley & Sons, Hoboken, NJ,
2009.
2. L. Theodore and F. Ricci, Mass Transfer Operations for the Practicing Engineer, John Wiley & Sons,
Hoboken, NJ, 2010.
3. D. Kaufmann, Plant Design, A Theodore Tutorial, Theodore Tutorials, East Williston, NY, origi-
nally published by the USEPA/APTI, RTP, NC, 1997.
284 Introduction to Optimization for Chemical and Environmental Engineers

4. K. Skipka and L. Theodore, Energy Resources: Availability, Management, and Environmental

Impacts, CRC Press/Taylor & Francis Group, Boca Raton, FL, 2014.
5. J. Happel, Chemical Engineering Economics, John Wiley & Sons, Hoboken, NJ, 1958.
6. C. Tyler, Chemical Engineering Economics, 3rd edn, McGraw-Hill, New York City, NY 1948.
7. L. Theodore and R. Dupont, Environmental Health and Hazard Risk Assessment: Principles and
Calculations, CRC Press/Taylor & Francis Group, Boca Raton, FL, 2002.
8. D. Hendershot, A simple example problem illustrating the methodology of chemical process
quantitative risk assessment, Paper presented at AIChE Mid-Atlantic Region Day in Industry for
Chemical Engineering Faculty, April 15, 1988, Bristol, PA.
9. L. Theodore and F. Taylor, Probability and Statistics, A Theodore Tutorial, Theodore Tutorials,
East Williston, NY, originally published by the USEPA/APTI, RTP, NC, 1999.
10. J. Reynolds, J. Jeris, and L. Theodore, Handbook of Chemical and Environmental Engineering
Calculations, John Wiley & Sons, Hoboken, NJ, 2002.
11. S. Shaefer and L. Theodore, Probability and Statistics for Environmental Science, CRC Press/Taylor
& Francis Group, Boca Raton, FL, 2007.
12. L. Theodore, Environmental Risk Analysis: Probability Distribution Calculations, CRC Press/Taylor
& Francis Group, Boca Raton, FL, 2016.
13. K. Bury, Statistical Methods in Applied Science, John Wiley & Sons, Hoboken, NJ, 1975.
14. L. Theodore and R. Dupont, Calculating hazard probabilities using the Weibull distribution,
Paper #125, AWMA Conference, Orlando, FL, 2011.
15. R. Dupont, J. McKenna, and L. Theodore, Baghouse failures: applying the Weibull distribution
to estimate bag failure emissions as a function of time, Paper #112, AWMA Conference, Chicago,
IL, 2013.
16. R. Dupont, F. Ricci, and L. Theodore, An improved failure rate model applied to baghouse
failures, AWMA Conference, Long Beach, CA, 2014.
17. R. Dupont, F. Ricci, and L. Theodore, Replacing the Weibull distribution failure model, AWMA
Conference, Raleigh, NC, 2015.
18. L. Theodore and F. Ricci, Mass Transfer Operations for the Practicing Engineer, John Wiley & Sons,
Hoboken, NJ, 2011.
19
Select Chemical Engineering Applications

Five sections complement the presentation in this final chapter:

19.1 Fluid flow

19.2 Chemical reactors
19.3 Heat exchangers
19.4 Mass transfer operation
19.5 Plant design

19.1 Fluid Flow
Fluids are usually transported in pipes or tubes. Generally speaking, pipes are heavy-
walled and have a relatively large diameter. Tubes are thin-walled and often come in coils.
Pipes are specified in terms of their diameter and wall thickness. The nominal diameters
range from 1/8 to 30 inches for steel pipe. Standard dimensions of steel pipe are provided
in the literature and are known as IPS (iron pipe size) or NPS (nominal pipe size). The wall
thickness of the pipe is indicated by the schedule number, which can be approximated
from 1000(P/S) where P is the maximum internal service pressure (psi) and S is the allow-
able bursting stress in the pipe material (psi). (The S value varies by material, grade of
material and temperature; allowable S values may be found in piping handbooks).1–3
Tube sizes are indicated by the outside diameter. The wall thickness is usually given
a BWG (Birmingham Wire Gauge) number. The smaller the BWG, the heavier the tube.
The literature lists the sizes and wall thicknesses of condenser and heat exchanger tubes.
For example, a ¾ inch 16 BWG tube has an outside diameter (OD) of 0.75 inches, an inside
diameter (ID) of 0.62 inches, a wall thickness of 0.065 inches, and a weight of 0.476 lb/ft.
When designing a piping system for an industrial process, savings can be effected by
determining the optimum pipe diameter and hence the economic optimum condition by
means of an economic balance. As mentioned previously, to obtain an economic balance,
one type of cost must increase while the other decreases with an increase in the indepen-
dent variable. For the case under consideration, the primary independent variable is usu-
ally the diameter of the pipe. One set of costs, namely, the fixed charges (cost of the pipe,
fittings, and installation) will increase with an increase in the pipe diameter. On the other
hand, for a given flow rate, the power costs due to pressure drop decrease as the pipe size
increases. In other words, if the pipe is oversized, the fixed charges will be too large, and if
the pipe diameter is too small, the pressure drop will be excessive, resulting in large power
costs. The economic balance therefore is made between the fixed costs and the power costs.

285
286 Introduction to Optimization for Chemical and Environmental Engineers

TABLE 19.1
Pipe Cost Data; Illustrative Example 19.1
Pipe Size, I.D. (in) Installed Cost ($/ft)
1.0 1.80
1.5 1.89
2.0 2.19
2.5 3.18
3.0 3.99

Illustrative Example 19.1

Consider the following application. It is desired to design a pipeline to operate 8700 h/yr
in which water at 60°F is to be fed to a cooling tower at a rate of 30 gal/min. The overall
length of the horizontal piping system is to be 1000 ft. If the annual fixed charges (depre-
ciation, taxes, insurance, etc.) are 30%/yr of the installed cost of the pipe and power costs
$0.06 kW‑h, calculate the economic optimum pipe diameter based solely on operating
costs. Assume that a pump for the installation is already on hand.

SOLUTION:
First, perform the calculations for the 1.0-in pipe assuming turbulent flow and a friction
factor of approximately 0.005.1,2

Initial cost of pipe = (1000)(1.80 ) = $1, 800

Annual initial charges = (1800 )(0.30 ) = $540

Annual power costs will depend on the pressure drop. First, calculate the velocity:

v=
( 30 )( 4 )
( 60 )( 7.48 )(1 / 12 )
2

= 38.5 ft / s
Apply Fanning’s equation. 1–3

4fLv 2 (19.1)
∆P =
2g c D

( 4 )(1000 )( 0.005 )( 38.5 )

2

∆P =
( 2 )( 32.2 )(1 / 12 )
∆P = 5, 523 ft H 2 O

= 28, 732 lbf / ft 2

The annual power cost is

PC =
( 28, 732 )( 30 )( 60 )( 8700 )( 0.06 )
( 7.48 )( 778 )( 3412 )
Annual power costs = $1359

Total annual cost for the 1.0-in pipe = 540 + 1359 = 1899.

Select Chemical Engineering Applications 287

Table 19.2 provides the calculated results for the remaining pipe sizes.
Plotting both the annual fixed costs and power costs versus the pipe size will result
in a graph similar to that in Figure 19.1, as presented earlier in Part III, Figure 13.1. The
optimum pipe diameter is approximately 1.75 inches. Performing the calculations for
pipe sizes of 1.25 inches and 1.75 inches is left as an exercise for the reader.
The reader should note that this topic will be revisited in Part IV, Chapter 19.

19.2 Chemical Reactors
A reactor where mixing is important is the tank flow or continuous stirred tank reactor
(CSTR). It is also referred to as a “back-mix” reactor. This type of reactor, like the batch
reactor,4,5 essentially consists of a tank or kettle equipped with an agitator. As with the
batch reactor discussed previously, perfect mixing is normally assumed so that both the
concentration and temperature within the reactor is the same, that is, there is no spatial
variation so that both terms are not a function of position within the reactor. This per-
fect mixing assumption also requires that both the concentration and temperature at the
discharge point (the outlet pipe) is identical to that in the reactor. As noted in Chapter 16
in Part III, Fogler6 noted that for perfect mixing: “The idea that the composition is identi-
cal everywhere in the reactor and in the exit pipes requires some thought. It might seem
that, since the concentration changes instantly at the entrance where mixing occurs, the
reaction occurs there and nothing else happens in the reactor because nothing is chang-
ing. However, reaction occurs throughout the reactor, but mixing is sufficiently rapid that
TABLE 19.2
Summary of Pipe Diameter Calculations; Illustrative Example 19.1
Pipe Size (in) Initial Charges ($/yr) Power Cost ($/yr) Total Costs ($/yr)
1.0 540 1359 1899
1.5 567 179 746
2.0 657 42 700
2.5 954 14 968

Total cost
Cost $/yr

Operating cost

Capital cost

Pipe diameter

FIGURE 19.1
Economic balance plot.
288 Introduction to Optimization for Chemical and Environmental Engineers

nothing appears to change with time or position.” One can extend this reasoning to also
conclude that the rate of reaction within the reactor is also uniform so that the extent of
conversion of reactant is the same within the reactor and the discharge port.
The CSTR essentially consists of a tank or kettle equipped with an agitator (see Figure 19.2)
and it may be operated under steady or transient conditions. Reactants are fed continu-
ously, and the product(s) are withdrawn continuously. The reactant(s) and product(s) may
be liquid, gas, or solid, or a combination of them all. If the contents are perfectly mixed,
the reactor design problem is greatly simplified for steady conditions because the mixing
results in the aforementioned uniform concentration, temperature, and so on, throughout
the reactor. This means that the rate of reaction is also constant and the describing equa-
tions are not differential and, therefore, do not require integration.
Consider the elementary irreversible reaction between benzoquinone (A) and cyclopen-
tadiene (B):

A + B → products (19.2)

If one employs a feed containing equimolar concentrations of reactants, the reaction rate
expression can be written as

−rB = kC A C B = kC 2B ; k = k B (19.3)

Proceed to calculate the reactor size requirements for one CSTR. Also, calculate the vol-
ume requirements for a cascade composed of two identical CSTRs. Assume isothermal
operation at 25°C where the reaction rate constant is equal to 9.92 m3/(kgmol/ks). Reactant
concentrations in the feed are each equal to 0.08 kgmol/m3, and the liquid feed rate is equal
to 0.278 m3/ks. The desired degree of conversion is 87.5%.
The rate equation, −rB, in terms of conversion variable X is4,5

−rB = −r = kC B2 (19.4)

Motor

Feed line

Impeller
Cooling or
heating jacket

Product

FIGURE 19.2
Typical continuous stirred tank flow reactor.
Select Chemical Engineering Applications 289

C B = C B0 ( 1 − X ) (19.5)

−r = kC 2B ( 1 − X ) (19.6)
2

If only one reactor is employed,4,5

FB0 X
V=
−r
(19.7)
FB0 X
=
kC B2 0 ( 1 − X )
2

Since
FB0 = C B0 q

= (0.08 )(0.278 )
(19.8)
= 0.02224 kgmol / ks
= 0.02224 gmol / s
The volume may be calculated employing Equation (19.7).

V = ( 0.02224 )( 0.875 ) / ( 9.92 )( 0.08 ) ( 1 − 0.875 ) 

2 2

 
3
= 19.6 m
The previous design equation may also be applied to two reactors of equal volume in
series, noting that X1 − 0 = X1.3

FB0 X1
V1 = (19.9)
kC 2B0 ( 1 − X1 )2

FB0 X1
V2 = ; ∆X = X 2 − X1 (19.10)
kC 2B0 ( 1 − X 2 )2
For reactor 1,

kC 2B0 V1 X1
= 2
FB0 ( X1 )
1 −
For reactor 2,

kC 2B0 V2 X − X1
= 2 2
FB0 (1 − X 2 )
Since the LHS of both of the previous equations are equal, (V1 = V2),

X1 X 2 − X1
=
(1 − X1 ) (1 − X 2 )
2 2
290 Introduction to Optimization for Chemical and Environmental Engineers

Solving with X2 = 0.875 yields

X1 = 0.7251

Thus,

V1 =
( 0.278 )( 0.08 )( 0.7251)
( 9.92 )( 0.08 ) (1 − 0.7251)
2 2

= 3.36 m 3
In addition,

VT = V1 + V2

= ( 2 )( 3.36 ) (19.11)

= 6.72 m 3
The previously mentioned results – in terms of the volume requirements – for the two
cases may be compared. Using one reactor requires a volume of

V1 = 19.6 m 3

Use of two equal volume reactors in series produced a volume requirement of

V2 = 3.36 + 3.36 = 6.72 m 3

As expected, the latter case provided the smaller volume requirement.

The previously mentioned calculations may be further extended to determine the
effect of using a cascade of two CSTRs, that differ in size, on the volume requirements for
the reactor network. If the same feed composition and flow rate as previously mentioned
are employed, and if the reactors are operated isothermally at 25°C, the minimum total
volume required and the manner in which the volume should be distributed between
the two reactors can also be calculated if the fractional overall conversion of 0.875 is to
be achieved again.
Set up an equation to determine the total volume requirement if the two reactor volumes
are not equal.4,5

VT = V1 + V2 ; V1 ≠ V2 (19.12)

 F   X1 X 2 − X1 
VT =  B02   2 + 2  ; X 2 = 0.875
(19.13)
 kC B0   (1 − X 1 ) (1 − X 2 ) 
To solve the previous equation for the intermediate conversion X1 under minimum volume
conditions, one needs to minimize VT, by setting

dVT
= 0; VT = VT ( X 1 ) (19.14)
dX 1
Select Chemical Engineering Applications 291

and then solve for X1. Analytical differentiating yields

dVT 1 2X 1
= + − 64 = 0 (19.15)
dX1 ( 1 − X1 )2 ( 1 − X 2 )2
Solving by trial-and-error gives

X1 = 0.702

The volumes of the two individual reactors and the total minimum volume requirement
can now be calculated:

V1 =
( 0.02224 gmol / s ) ( 0.702 )
( 9.92 L / gmol ⋅ s ) (0.08 gmol / L)2 (1 − 0.875 )2
= 2.77 L

V2 =
( 0.02224 gmol / s ) ( 0.875 − 0.702 )
( 9.92 L / gmol ⋅ s ) (0.08 gmol / L)2 (1 − 0.875 )2 (19.16)
= 3.88 L

VT = V1 + V2
= 2.77 L + 3.88 L

= 6.65 L

The previous result is slightly lower than that calculated earlier (VT = 6.72 L).
The effect of using a cascade of three CSTRs that differ in size on the volume require-
ments for the reactor network may also be investigated. If the same feed composition and
flow are employed and the reactors are operated isothermally at 25°C, the minimum total
volume required and the manner in which the volume should be distributed between
the three reactors is another optimization problem. An overall conversion of 0.875 is to be
achieved again.
Proceeding as before, write the equation for the volume of each reactor and total volume
requirement for three reactor volumes that are not equal.
For each reactor

 F   X1 
V1 =  B02   2
(19.17)
 kC B0   (1 − X1 ) 

 F   X − X1 
V2 =  B02   2 2
(19.18)
 kC B0   (1 − X 2 ) 
with

 F   X − X2 
V3 =  B02   3 2
(19.19)
 kC B0   (1 − X 3 ) 
292 Introduction to Optimization for Chemical and Environmental Engineers

The total volume requirement is

VT = V1 + V2 + V3 (19.20)

 F   X 1   FB0   X 2 − X 1   FB0   X 3 − X 2 
VT =  B02   2 +  2 +  2
(19.21)
 kC B0   (1 − X 1 )   kC 2B0   (1 − X 2 )   kC 2B0   (1 − X 3 ) 
Take the partial derivative of the equation obtained previously with respect to X1. Note that
the partial derivative must now be employed since the total volume, VT, depends on two
variables, X1 and X2, that is, VT = VT(X1, X2).7

∂VT  FB0   1 2X 1 1 
= 2   2 + 3 − 2
(19.22)
∂X1  kC B0   (1 − X 1 ) (1 − X 2 ) (1 − X 3 ) 
Set this derivative equal to zero.

∂VT  FB0   1 2X 1 1 
= 2   2 + 3 − 2 = 0
(19.23)
∂X 1  kC B0   (1 − X 1 ) (1 − X 2 ) (1 − X 3 ) 
Take the partial derivative for VT with respect to X2.

∂VT  FB0   1 2(X 2 − X 1 ) 1 

= 2   2 + 3 − 2
(19.24)
∂X 2  kC B0   (1 − X 1 ) (1 − X 2 ) (1 − X 3 ) 
Also, set this derivative equal to zero.

∂VT  FB0   1 2(X 2 − X 1 ) 1 

= 2   2 + 3 − 2 = 0
(19.25)
∂X 2  kC B0   (1 − X 1 ) (1 − X 2 ) (1 − X 3 ) 
Solve the resulting two algebraic equations obtained from the partial derivative equations
simultaneously. For X3 = 0.875, X2 is obtained as follows by trial-and-error (see Table 19.3).
Therefore, X1 = 0.585, X2 = 0.7879. Finally, calculate the volume of the three individual reac-
tors and the total minimum volume requirement.

 F   X1 
V1 =  B02   2
 kC B0   (1 − X1 ) 

 0.02224 gmol / s   0.585 

= 2  2
(19.26)
 (9.92 L / gmol ⋅ s ) (0.08 gmol / L)   (1 − 0.585) 

= 1.19 L

 F   X − X1 
V2 =  B02   2 2
 kC B0   (1 − X 2 ) 

 0.02224 gmol / s   0.7879 − 0.585 

= 2  2 
(19.27)
 (9.92 L / gmol ⋅ s ) (0.08 gmol / L)   (1 − 0.7879) 

= 1.58 L
Select Chemical Engineering Applications 293

TABLE 19.3
Trial-and-Error Calculation for Three
Reactor Set
X1 X2 Residual
0.6 0.8 11
0.59 0.7918 4
0.58 0.7835 −2.5
0.585 0.7879 0.8

 F   X − X2 
V3 =  B02   3 2
 kC B0   (1 − X 3 ) 

 0.02224 gmol / s   0.875 − 0.7879 

= 2  2 
(19.28)
 (9.92 L / gmol ⋅ s ) (0.08 gmol / L)   (1 − 0.875) 

= 1.95 L
VT = V1 + V2 + V3
= 1.19 L + 1.58 L + 1.95 L (19.29)

= 4.72 L
Note the reduction in the volume requirement when three CSTRs in series are employed.
The reader is left the exercise of showing that the total volume requirement for three
CSTRs of equal volume is

V1 = V2 = V3 = 2.17 m 3
VT = V1 + V2 + V3 = 6.51 m 3

with

X1 = 0.6710
X 2 = 0.7782

Theodore4,5 also calculated the reactor size requirement for a cascade composed of two
CSTRs with the first inlet reactor THRICE the size the second, the reactor size require-
ments for a cascade composed of two CSTRs with the first inlet reactor ONE-THIRD the
size of the second, if a very large number (approaching infinity) of CSTRs are employed,
and if the reaction is reversible for the previously mentioned conclusions.

19.3 Mass Transfer Operation

One of the more important mass transfer operations involves the transfer of gases – includ-
ing those that are toxic – within a plant. Exposure to contaminants in a workplace can be
reduced by proper ventilation. Ventilation can be provided either by dilution ventilation
294 Introduction to Optimization for Chemical and Environmental Engineers

or by a local exhaust system. In dilution ventilation, air is brought into the work area to
dilute the contaminant sufficiently to minimize its concentration and subsequently reduce
worker exposure. In a local exhaust system, the contaminant itself is removed from the
source through hoods.
A local exhaust is generally preferred over a dilution ventilation system for health prob-
lems because a local exhaust system removes the contaminants directly from the source,
whereas dilution ventilation merely mixes the contaminant with uncontaminated air to
reduce the contaminant concentration. Dilution ventilation may be acceptable when the
contaminant concentration has a low toxicity, and the rate of contaminant emission is con-
stant and low enough that the quantity of required dilution air is not prohibitively large.
However, dilution ventilation generally is not acceptable when the concentration is less
than 100 ppm.
In determining the quantity of dilution air required, one must also consider mixing char-
acteristics of the work area in addition to the quantity (mass or volume) of contaminant
to be diluted. Thus, the amount of air required in a dilution ventilation system is much
higher than the amount required in a local exhaust system. In addition, if the replacement
air requires heating or cooling to maintain an acceptable workplace temperature, then
the operating cost of a dilution ventilation system may further exceed the cost of a local
exhaust system.

Illustrative Example 19.2

Discuss the optimization of a ventilation system.

SOLUTION:
The major components of an industrial ventilation system include the following:

1. Exhaust hood
2. Ductwork
3. Contaminant control device
4. Exhaust fan
5. Exhaust vent or stack

Several types of hoods are available. One must select the appropriate hood for a specific
operation to effectively remove contaminants from a work area and transport them into
the ductwork. The ductwork must be sized such that the contaminant is transported
without being deposited within the duct; adequate velocity must be maintained in the
duct to accomplish this. Selecting a control device that is appropriate for the contami-
nant removal is important to meet certain pollution control removal efficiency require-
ments. The exhaust fan is the workhorse of the ventilation system. The fan must provide
the volumetric flow at the required static pressure, and it must be capable of handling
contaminated air characteristics such as dustiness, corrosiveness, and moisture in the
air stream. Properly venting the exhaust out of the building is equally necessary to
avoid contaminant recirculation into the air intake or into the building through other
openings. Such problems can be minimized by properly locating the vent pipe in rela-
tion to the aerodynamic characteristics of the building. In addition, all or a portion
of the cleaned air may be recirculated to the workplace. Primary (outside) air may be
added to the workplace and is referred to as makeup air; the temperature and humid-
ity of the makeup air may have to be controlled. It also may be necessary to exhaust a
portion of the room air. Each of these variables needs careful analysis if a ventilation
system’s operation and performance is to be maximized.
Select Chemical Engineering Applications 295

To
stack

Duckwork

Particulate control
equipment

Makeup Return/
air recirculation air
Discharge/
Room/workplace exhaust air
Source

FIGURE 19.3
Industrial ventilation systems.

A line diagram of a typical industrial ventilation system is provided in Figure 19.3.

Note that either the control device or the fan (or both) can be located in the room/
workplace.

19.4 Heat Exchangers
In many environmental, chemical, and petrochemical plants there are cold streams that
must be heated and hot streams that must be cooled. Rather than use steam to do all the
heating and cooling water to do all the cooling, it is often advantageous to have some of the
hot streams heat the cold ones. In addition, the problem of optimizing the heat exchanger
networks that would be required has been extensively studied, but it remains a daunt-
ing exercise for practicing engineers.8–9 What follows is one simple illustration of a heat
exchanger network.
Consider the following application.10 A plant has three streams to be heated and three
streams to be cooled. Cooling water (90°F supply, 155°F return) and steam (saturated at 250
psia) are available. Devise a network of heat exchangers that will make full use of heat-
ing and cooling streams against each other, using utilities only if necessary. Information
on the three streams to be heated are presented in Table 19.4. Information on the three
streams to be cooled are presented in Table 19.5. Saturated stream at 250 psia has a tem-
perature of 401°F.
Here is the solution represented by Kaufman.10 The heating duties for all streams are
first calculated. The results are tabulated in Table 19.6.
The total heating and cooling duties are next compared:

Heating : 7 , 745, 000 + 6, 612, 000 + 9, 984, 000 = 24, 071, 000 Btu / h

Cooling : 12, 600, 000 + 4, 160, 000 + 3, 150, 000 = 19, 910, 000 Btu / h

As a minimum, 4,160,000 Btu/h might have to be supplied by steam.

296 Introduction to Optimization for Chemical and Environmental Engineers

TABLE 19.4
Heated Network Streams
Stream Flowrate (lb/h) cp [Btu/(lb⋅°F)] Tin(°F) Tout (°F)
1 50,000 0.65 70 300
2 60,000 0.58 120 310
3 80,000 0.78 90 250

TABLE 19.5
Cooled Network Streams
Stream Flowrate (lb/h) cp [Btu/(lb⋅°F)] Tin(°F) Tout (°F)
4 60,000 0.70 420 120
5 45,000 0.52 300 100
6 35,000 0.60 240 90

TABLE 19.6
Network Heating and Cooling Duties
Stream Number Stream Type Duty
1 Heating 7,745,000
2 Heating 6,612,000
3 Heating 9,984,000
4 Cooling 12,600,000
5 Cooling 4,160,000
6 Cooling 3,150,000

Figure 19.4 represents a system of heat exchangers that will transfer heat from hot
streams to the cold ones in the amounts desired. It is important to note that this is but
one of many possible schemes. The optimum system would require a trial-and-error
procedure that would examine a host of different schemes. Obviously, the econom-
ics would also come into play. Theodore9, and Abulencia and Theodore,6 discuss the

2 2
120° F 310° F
6,612,000 4
420° F
4 Steam
120° F 158.1° F
1,600,000 4,388,000 1,496,000
3 3
90° F 115.6° F 250° F
2,500,000
6
240° F

6 Steam
90° F 121° F
650,000 2,665,000
1 1
70° F 90° F 218° F 300° F
4,160,000

5 5
100° F 300° F

FIGURE 19.4
Flow diagram for a heat exchanger network.
Select Chemical Engineering Applications 297

ramifications that arise if the heat exchanger cost is a function of the heat exchanger
area; that is, the network should be such that the total cost of the resulting network is
minimized. They also discuss how to solve the problem so that the network’s entropy
increase11,12 and the network’s cost are minimized. The practicing engineer or scien-
tist should consider how to arrive at a reasonable solution to this energy conservation
problem.
Finally, highly interconnected networks of exchangers can save a great deal of energy
in a chemical plant. The more interconnected they are, however, the harder the plant
is to control, startup, and shut down. Often auxiliary heat sources and cooling sources
must be included in the plant design in order to ensure that the plant can operate
smoothly. The  previously mentioned problem and considerations indicate how dif-
ficult it can be to solve some real-world industrial applications. The interested reader
is encouraged to contact the authors if other methods of solution are available and
possible.

19.5 Plant Design
Plant design requires the use of engineering principles and theories combined with a
practical realization of the limits imposed by industrial conditions. The development of a
plant-design project involves a wide variety of talents. Research, market analyses, design
of individual pieces of equipment, plant-location surveys, and many other problems are
involved in the design of industrial plants.10
Although the environmental and chemical engineer must understand the principles
related to the various design phases, the individual parts of the design are of little benefit
until they are combined to give one complete unit. The complete plant, i.e., the complete
unit – that is the design engineer’s goal. However, one major qualification must be added
before the goal is attained. This is the qualification that the operation must be able to pro-
ceed economically, and this often requires that energy be conserved. Concerns and details
regarding the conservation of energy follow.
One of the areas where meaningful energy conservation measures can be realized is in
the design and specification of process (operating) conditions for heat exchangers. This
can be best accomplished by the inclusion of second law principles in the analysis. The
quantity of heat recovered in an exchanger is not alone in influencing size and cost. As the
temperature difference driving force (LMTD) in the exchanger approaches zero, the “qual-
ity” heat recovered increases.5
Most heat exchangers are designed with the requirements/specification that the tem-
perature difference between the hot and cold fluid be at all times positive and be at least
20°F. This temperature difference or driving force is referred to by some as the approach
temperature. However, the corresponding entropy change is also related to the driving
force, with large temperature difference driving forces resulting in large irreversibilities
and the associated large entropy changes.11,12
The individual designing a heat exchanger is faced with two choices. He/she may decide
to design with a large LMTD that results in both a more compact (smaller area) design and
a large entropy increase that is accompanied by the loss of “quality” energy. Alternatively,
a design with a small driving force results in both a large heat exchanger and a small
entropy change with a larger recovery of “quality” energy.9,11
298 Introduction to Optimization for Chemical and Environmental Engineers

Illustrative Example 19.3

Consider the following application. Shannon O’Brien, a recent graduate from
Manhattan College’s prestigious chemical engineering program was given an assign-
ment to design the most cost-effective heat exchanger to recover the energy from a hot
flue gas at 500°F. The design is to be based on pre-heating 100°F incoming air (to be
employed in a boiler) to a temperature that will result in the maximum annual profit
to the utility. A line diagram of the proposed countercurrent exchanger temperature
profile is provided in Figure 19.5.
Having just completed a heat transfer course with Dr. Annmarie Flynn and a thermo-
dynamics course with the nefarious Dr. Theodore, Shannon realizes that there are two
costs that need to be considered:

1. The heat exchanger employed for energy recovery

2. The “quality” (from an entropy perspective) of the recovered energy

She also notes that the higher the discharge temperature of the heated air, t, the smaller
will be the temperature difference driving force, and the higher the area requirement of
the exchanger and, correspondingly the higher the equipment cost. Alternatively, with
a higher t, the “quality” of the recovered energy is higher, thus leading to an increase in
recovered profits (by reducing fuel costs).
Based on similar system designs, Francesco Ricci Consultants (FRC) has provided the
following annual economic models for the heat exchanger,

Recovered energy profit : A ( t − tc ) ; A = $ / yr ⋅°F (19.30)

Recovered energy cost : B / ( TH − t ) ; B = $ / yr ⋅°F (19.31)

For the previously mentioned system, FRC suggests a value for the coefficients in the
cost model be set at

A = 10
B = 100,000

Employing the information provided, Shannon has been asked to calculate a t that will

TH = 500°F

Hot stream

T = 120°F

Cold stream 100°F = tc

FIGURE 19.5
Proposed countercurrent exchanger; Illustrative Example 19.3.
Select Chemical Engineering Applications 299

1. Provide breakeven operation

2. Maximize profits

She is also required to perform the calculation if A = 10, B = 4000, as well as A = 10,
B = 400,000. Finally, an analysis of the results is requested.

SOLUTION:
1. Since there are two contributing factors to the cost model, one may write the
following equation for the profit, P

B
P = A ( t − tc ) − ; TH = 500 and tc = 100 (19.32)
( TH − t )
For breakeven (BE) operation, set P = 0 so that

B
( t − tc ) ( TH − t ) = (19.33)
A
This may be rewritten as

B 
t 2 − ( TH + tc ) T +  + TH tc  = 0 (19.34)
A 
The solution to this quadratic equation for A = 10 and B = 100,000 is

( 600 ) − ( 4 )(1) (10, 000 + 50, 000 )

2
600 ±
t=
2
600 ± 346
=
2
= 473°F , 127°F
2. To maximize the profit, take the first derivative of P with respect to t and set it
equal to zero, that is,

dP B
= A− 2 = 0 (19.35)
dt ( TH − t )
Solving,

B
( TH − t )
2
=
A
B
TH − t =
A

= 10, 000 = 100

Since TH = 500,

t = 400°F
300 Introduction to Optimization for Chemical and Environmental Engineers

B = 4,000

B = 100,000

Profit t
100°F 500°F
B = 400,000

FIGURE 19.6
Heat exchanger results.

Upon analyzing the first derivative with t values greater than and less than 400°F,
Shannon observes that the derivative changes sign from + → − about t = 400, indicating
a relative maximum.
Similarly, for A = 10, B = 4000,

tBE = 499°F , 101°F

tmax = 480°F
For A = 10, B = 400,000,

tBE = 300°F

tmax = 300°F
Graphical results for the three scenarios are shown in Figure 19.6.
Finally, there are a host of problems peripheral to this “quality energy” concern asso-
ciated with heat exchangers that the reader may choose to consider. This includes topics
that involve minimizing capital costs, minimizing operating costs, minimizing entropy
change(s), minimizing area requirements, minimizing energy requirements, and so on,
and/or any combination of these.

References
1. P. Abulencia and L. Theodore, Open-Ended Problems: A Future Chemical Engineering Education
Approach, Wiley‑Scrivener, Salem, MA, 2015.
2. W. Baasel, Preliminary Chemical Engineering Plant Design, 2nd edn, Van Nostrand Reinhold,
New York City, NY, 1990.
3. R. Huntington, High Pressure Pipe Line Research, Clark Bros., Olean, NY, 1947.
4. L. Theodore, Chemical Reactor Analysis and Applications for the Practicing Engineer, John Wiley &
Sons, Hoboken, NJ, 2012.
Select Chemical Engineering Applications 301

5. L. Theodore, Chemical Reactor Kinetics, A Theodore Tutorial, Theodore Tutorials, East Williston,
NY, originally published by the USEPA/APTI, RTP, NC, 1995.
6. H. Fogler, Elements of Chemical Reaction Engineering, 4th edn, Prentice-Hall, Upper Saddle River,
NJ, 2006.
7. H. Mickley, T. Sherwood, and C. Reed, Applied Mathematics in Chemical Engineering, McGraw-
Hill, New York City, NY, 1957.
8. J. Reynolds, Heat Transfer, A Theodore Tutorial, Theodore Tutorials, East Williston, NY, origi-
nally published by the USEPA/APTI, RTP, NC, 1995.
9. L. Theodore, Heat Transfer Applications for the Practicing Engineer, John Wiley & Sons, Hoboken,
NJ, 2011.
10. D. Kauffman, Process Synthesis and Design, A Theodore Tutorial, Theodore Tutorials, East
Williston, NY, originally published by the USEPA/APTI, RTP, NC, 1992.
11. L. Theodore, F. Ricci, and T. VanVliet, Thermodynamics for the Practicing Engineer, John Wiley &
Sons, Hoboken, NJ, 2009.
12. L. Theodore, Thermodynamics, A Theodore Tutorial, Theodore Tutorials, East Williston, NY,
originally published by the USEPA/APTI, RTP, NC, 1995.
20
Advanced Optimization Term Projects

Three sections complement the presentation in this final chapter:

20.1 Sulfuric acid process

20.2 Optimum indoor ventilation flow Rate
20.3 Fluid transportation system
20.4 Mass transfer application

The final chapter (20) in Part IV contains four illustrative examples that could best be
described as advanced optimization term projects. Illustrative Examples 20.3 and 20.4
were drawn, with permission, from the classic work of Happel1 titled, Chemical Process
Economics, John Wiley & Sons, Hoboken, NJ, 1957; one example is concerned with flow/
piping optimization and the other with a distillation process. The first example in this
chapter is concerned with a sulfuric acid process and the second with optimizing indoor
ventilation flowrates. Note that no introductory material for each example is provided in
this final chapter.

20.1 Sulfuric Acid Process

Illustrative Example 20.1
Two sulfuric acid streams (x1, x2) are processed at a small refinery in northern New
Jersey. The first (1) contains a 2.0% (by volume) concentration of fine catalyst particles
while the second (2) contains a concentration of 8.0%. The maximum and minimum
flow rate of each stream that can be processed is 50 gal/hr and 250 gal/hr, respectively.
An earlier economic study by Lake Engineers concluded that for maximum profit, the
flow rate of the two acid streams should be constrained by the relationship

x1 ≤ 194.5 − 0.69x2 , gal/hr (20.1)

x1 ≥ 126.6 − 0.84 x2 , gal/hr (20.2)

In addition, Fitzmaurice Environmental (FE) introduced another constraint based on envi-

ronmental concerns that the platinum laden catalyst discharge be limited to 7.5 gal/hr
(approximately 50 lb/hr). Assuming a worst-case condition, FE concluded that the follow-
ing constraint be met:

0.02x1 + 0.08 x2 ≤ 7.5; gal/hr (20.3)

303
304 Introduction to Optimization for Chemical and Environmental Engineers

Rufus Hampton Jr., the plant engineer recently requested that the authors calculate
the maximum and minimum hourly flow rate of each stream that can be processed that
satisfies all of the aforementioned constraints and maximizes profit.

SOLUTION:
Employ Equations (20.1) and (20.2) in conjunction with the following two equations:

50 ≤ x1 ≤ 250 (20.4)

50 ≤ x2 ≤ 250 (20.5)

Excel provides the following flow rate (gal/hr) solutions:

x1 (max ) = 154.78 with x2 = 57.55

x1 (min ) = 57.91 with x2 = 81.77

x2 (max ) = 81.77 with x1 = 57.91

x2 (min ) = 57.55 with x1 = 81.77

The reader is left the exercise of solving the problem graphically.

20.2 Optimum Indoor Ventilation Flow Rates

Indoor air quality (IAQ) is a major concern because indoor air pollution may present a
greater risk of illness than exposure to outdoor pollutants. People spend nearly 90% of
their time indoors. This situation is compounded as sensitive populations – for example,
the very young, the very old, and sick people who are more vulnerable to disease – spend
many more hours indoors than the average population.
Indoor air quality problems have become more serious and of greater concern now
than in the past because of a number of developments that are believed to have resulted
in increased levels of harmful chemicals in indoor air. Some of these developments are
the result of construction of more tightly sealed buildings to save on energy costs, the
reduction of the ventilation rate standards to save still more energy, the increased use of
synthetic building materials and synthetics in furniture and carpeting that can release
harmful chemicals, and the widespread use of new pesticides, paints, and cleaners.
Some of the immediate health effects of indoor air quality problems are irritation of the
eyes, nose, and throat, headaches, dizziness, fatigue, asthma, pneumonitis, and “humidi-
fier fever.” Some of the long-term health effects of indoor air quality problems are respi-
ratory diseases and cancer. These are most often associated with radon, asbestos, and
second-hand tobacco smoke. The US Environmental Protection Agency (EPA), in a report
to Congress nearly three decades ago, estimated that the costs of IAQ problems were in the
tens of billions of dollars per year. The major types of costs from IAQ problems are direct
medical costs, lost productivity due to absence from the job because of illness, decreased
efficiency on the job, and damage to materials and equipment.
Advanced Optimization Term Projects 305

Illustrative Example 20.2

Refer to Illustrative Example 13.5.

1. Calculate minimum air ventilation flow rate containing 10 ng/m3 nanoparticles

into the room to ensure that the nanoagent concentration does not exceed 35.0
ng/m3. The nanoagents are appearing in the laboratory at a rate of 250 ng/min.
Assume steady-state conditions.
2. Calculate the steady-state concentration in the laboratory; the initial concentra-
tion of the nanochemical is 500 ng/m3. There is no additional source of nano-
agents (generated) and the ventilation air is essentially pure, that is, there is no
background nanoagent concentration.
3. If the room volume is 142 m3, the flow rate of the 10 ng/m3 ventilation air is 12.1
m3/min, and nanoparticles are being generated at a steady rate of 30 ng/min,
calculate how long it would take for the concentration to reach 20.7 ng/m3. The
initial concentration in the laboratory is 85 ng/m3.
4. Refer to part 3. How long would it take to achieve a concentration of 13.65 ng/m3?
How long would it take to reach 12.2 ng/m3?

SOLUTION:
1. The applicable model for this case is (see also Part III, Illustrative Example 13.5):

dc
v0 (c0 − c ) + rV = V (20.6)
dt
Under steady-state, dc/dt = 0. Pertinent information includes

rV = 250 ng / min

c0 = 10 ng / m 3

c = 35 ng / m 3
Substituting gives

− rV
v0 =
c0 − c
rv
=
c0 − c (20.7)
250
=
35 − 10
= 10 m 3 / min = 353 ft 3 / min

2. The applicable model is

dc
v0 (c0 − c ) + rV = V (20.8)
dt
For steady-state conditions dc/dt = 0. In addition, based on the information pro-
vided, r = 0 and c0 = 0. Therefore, and as expected,

c=0
306 Introduction to Optimization for Chemical and Environmental Engineers

3. First, note that

τ = 142 / 12.1 = 11.73 min

k = r / V = 30 / 142 = 0.211 ng / m 3 ⋅ min

The applicable describing equation is

 t
−   −  
 t
c = ci e  τ
+ (c0 + k τ ) 1 − e  τ   (20.9)
 
Substituting gives

 t 
−   t 
−
 11.73  
20.7 = 85e + (10 + 2.48 ) 1 − e  11.73  
 
Solving by trial-and-error gives (approximately)

t = 29 min

4. First, calculate the steady-state concentration for this condition. The applicable
model is obtained from part 2, after setting dc/dt = 0

c0 + k τ c
− = 0 (20.10)
τ τ
Solving and substituting gives

c = c0 + k τ

= 10 + (0.211)(11.73 )

= 12.48 ng / m 3

Since this is the steady-state concentration, it will take an infinite period of times to
reach this value. However, the steady-state concentration represents the minimum con-
centration achievable in the laboratory based on the conditions specified. Therefore, the
concentration will never reach a value of 12.2 ng/m3.

20.3  Fluid Transportation System1

The general subject of fluid flow encompasses a host of topics. One topic of importance to
both environmental and chemical engineers is the optimizing of a fluid transport system.
Happel1 provides one such example that requires the use of optimization principles. His
example and solution (employing his notation), in which the total yearly cost of operating
a pipeline system in the year 1957 is formulated, is provided subsequently.
First, fixed charges on compressors and pipes will be established. The installed cost of
compressors is taken as X dollars/hp. Q cu ft/day of gas are to be pumped a distance of
L miles. Power requirement will be aZ2 bhp/(MM cu ft)/day, where Z2 is the deviation from
Advanced Optimization Term Projects 307

perfect-gas laws at the suction pressure, and a is the reference brake horsepower per mil-
lion cubic feet of gas per day. The investment in compressors is then

 aQZ2 
X ; dollars / mile (20.11)
 L106 
The fixed charge on compressors, α, includes maintenance, depreciation, and allowance
for profit on investment. The yearly fixed charge on the compressors is then

αXQZ2
; dollars / ( yr ) (mile) (20.12)
L106
Pipe costs Y dollars/ton. If W tons of pipe per mile are necessary for construction of the
line, the investment in pipe is YW tons/mile. For steel pipe W = 28.2(D + t)t where D is the
inside diameter of the pipe in inches, and t is the pipe wall thickness in inches. Barlow’s
formula for the wall thickness of pipe in terms of maximum allowable pressure, P lbf/sq
in., and maximum allowable stress, S lbf/sq in., is

t = P1A / 2 (S − P1 ) (20.13)

The direct pipe investment is then

 P1D   P1D 
YW = Y 28.2  D +   ; dollars / mile (20.14)
 2 (S − P1 )   2 (S − P1 ) 
A good designation of laying cost is H dollars/mile for right of way, surveying, clear-
ing, supervision, engineering; G dollars/ton for unloading, hauling, aligning, placing, and
welding or coupling; and, N dollars/(mile)(in. diam.) for ditching, painting, back filling,
and laying equipment. The total pipe cost is then

YW + H + GW + ND ; dollars / mile (20.15)

B is the fractional fixed charge per year on the installed pipe. The yearly charges on the
pipe will be

 P1D   P1D 
B ( Y + G)( 28.2)  D +   + BND + BH dollars / ( yr ) (mile) (20.16)
 2 (S − P1 )   2 (S − P1 ) 
Next, operating expenses must be considered. Labor, supervision, and salaries are taken at
$3850/month for a 4000-hp station. If direct proportionality is assumed for other stations
for approximately the same size, this expense is

3850 × 12  QZ2   QZ2 

  = 11.6   ; dollars / ( yr ) (mile) (20.17)
4000  L106  L106 
Supplies for maintenance of compressors amount to $3.40/hp/yr. The total yearly cost for
these items is thus,

 QZ2 
3.40  ; dollars / ( yr ) (mile) (20.18)
 L106 
308 Introduction to Optimization for Chemical and Environmental Engineers

About 10 std cu ft of fuel/bhp are required at a cost of Cf dollars/M std cu ft. The yearly
cost of fuel is

10 × 24 × 360  QZ2   QZ2 

Cf  6
= 86.4Cf  ; dollars / ( yr ) (mile) (20.19)
1000  L10   L106 
In addition, an allowance of $100/(yr)(mile) is taken to allow for loss of gas by leakage.
The addition of all these fixed and operation costs, taking the constant

a = kln (P1 / P2 ) + b (20.20)

gives the total yearly cost as follows:

KQZ2  P1 P12 
Cy = 
6 
kln ( P / P ) + b 
 + K D 2
 + 2  + BND + BH + 100 (20.21)
 2 (S − P1 ) 4 (S − P1 ) 
1 2 1
L10

where:

K = αX + 86.4Cf + 15.0 (20.22)

K 1 = 28.2B ( Y + G) (20.23)

The following values of various constants mentioned previously are believed to be repre-
sentative of 1957 costs:

α = 5% maintenance plus 5-year payment = (0.05)(5) = 0.25/yr

X = $250/bhp, compressor cost not installed
Cf = $0.10/1000 cu ft
a = 56.8 ln(P1/P2) + 5.1
k = 56.8
b = 5.1
B = 5.4% maintenance plus 5-yr payout = 0.27/yr
Y = $240/ton, pipe cost not installed
G = $30/ton, unloading pipe, and so on.
S = $20,000 lb/sq in, allowable stress
N = $500/(in. diam.)(mile), unloading pipe, and so on.
H = $5900/mile, survey, and so on.

With these constants provided, the cost Cy of operating a pipeline system for a specified
capacity is a function only of the pressures P1 and P2, the distance between stations L, and
diameter of pipe D. It is desired to compute the optimum values for these parameters for
the economic transportation of 150 MM std ft3/day of natural gas. Additional data from
various standard sources is provided by Huntington.2 The flow equation relating Q to the
independent variables involved is
Advanced Optimization Term Projects 309

( )
0.541
D 2.622 P12 − P2 2
Q = K2 (20.24)
( )
0.541
L0.541 Zavg
where:
D = inside diameter of pipe, in.
P1 = discharge pressure from compressors, lb/sq in. abs
P2 = suction pressure to compressors, lb/sq in. abs
L = distance between compressor stations, miles
Zavg = average deviation from perfect-gas laws for terminal pressures P1 and P2

with

0.541
80.8  M 
K2 =   (20.25)
ρ0µ 0.0814  T 
where for methane
ρ0 = density, lb/cu ft at 60°F = 0.0422
μ = viscosity in ln/(ft)(sec) at 60°F = 7 × 10−6
M = molecular weight of methane = 16
T = absolute temperature, °F abs = 520°R

From this data K2 = 2.15 × 105.

The optimum can be established either by direct differentiation of the cost function with
respect to each of the variables involved, setting the derivatives equal to zero, and solving
the resulting simultaneous equations, or by Lagrange’s method. Instead of working the
solution out for all four variables P1, P2, L, and D, Huntington chooses to assume successive
values for P1, reducing the analytical solution to a problem involving three independent
variables, P2, L, and D.
Employment of Lagrange’s method3,4 results in the following relationships if Q = ψ(P2, L, D)
is taken as the restrictive condition imposed on Cy in Equation (20.24).

∂C y ∂ψ
+λ = 0 (20.26)
∂P2 ∂P2

∂C y ∂ψ
+λ = 0 (20.27)
∂L ∂L

∂C y ∂ψ
+λ = 0 (20.28)
∂D ∂D
whence

KQZ2 ∂a KQZ2 ∂Z2  −1.082P2Q Q ∂Zavg 

+ + λ − 0.541 = 0 (20.29)
6 6
L10 ∂P2 L10 ∂P2 2
 P1 − P2
2
Zavg ∂P2 
310 Introduction to Optimization for Chemical and Environmental Engineers

KQZ2a Q
− − λ 0.541 = 0 (20.30)
L2 106 L

 P P12  Q
K 1D  1 + 2  + NB + λ 2.622 = 0 (20.31)
 S − P1 2 (S − P1 )  D

The previous three equations and Equation (20.24) constitute four independent equations
with unknown P1, L, D, and λ. They are solved by elimination and on the assumption that
Z2 and Zavg are sufficiently close to being constant so that their derivatives will equal zero.
With these assumptions it is found that

P1
= 1.33 (20.32)
P2
and

NB
D 6.85 + D 5.85
K 1 [ P1 /(S − P1 ) + P12 / 2 (S − P1 ) ]
2

(20.33)
4.85KQ 2.85Z2Zavg a
=
(
K 12.85 P12 − P2 2 ) K 1 P1 / (S − P1 ) + P12 / 2 (S − P1 )  106

2

The method of solution then involves calculations of P2 from an assumed value of P1.
D is then computed from Equation (20.31), and finally the flow Equation (20.24) is used to
solve for L.
In the present instance, a series of trials were made for a P1 variation between 300 and
800 lb/sq in. abs. Thus, assuming

P1 = 600
P2 = 451
Z2 = 0.933
Zavg = 0.923
P1 P2
+  =  0.0312
S − P1 2 (S − P1 )2
N = 500
B = 0.27
K1 = 28.2 × 0.27(240 + 30) = 2,060
K = 0.25(250) + 26.4(0.1) + 15 = 86.1

from which the right-hand side of Equation (20.31) equals

( 4.85)(86.1)( 2.01 × 1023 ) (0.933)(22.3)

= 7.43 × 108
(2.15)(10 ) ([600] − [ 451] ) (2060)(0.0312)(10 )
5 2 6
Advanced Optimization Term Projects 311

The left-hand side of equation (20.31) equals

D 6.85 +
(500)(0.27 ) D 5.85 = D 6.85 + 2.10D 5.85
(2060)(0.0312)
The value of D is obtained by solving the equation

D 6.85 + 2.10D 5.85 = 7.43 × 108

The solution of the previous equation by trial-and-error gives D = 19.0 in. The value of L
from the flow equation is

L=
(2.15 × 10 )(1.75 × 10 )(3.6 × 10 ) × 0.435 = 48.7 miles
5 6 5

(1.31 × 10 ) (0.923)
15

whence the yearly cost, Cy corresponding to the assumed 600 lb/sq in. abs discharge pres-
sure is

C y =
86.1 × 1.5 × 108 × 0.933 × 22.3
+ 2060 +
(19) × 0.0312 + 0.27 × 500 × 19 + 0.27 × 5900 + 100
2

48.7 × 106 2

C y = $21, 370

The results of the calculations at other discharge pressures for Q = 150 MM ft3/day are
given in Table 20.1. It appears that the optimum lies close to 400 lb/sq in. abs discharge
pressure for the conditions assumed. One notes that spacing between stations does not
vary a great deal.

20.4  Mass Transfer Application1

As previously explained in Chapter 15, distillation may be defined as the separation
of the components of a liquid feed mixture by a process involving partial vaporization
through the application of heat. This separation process is controlled by the properties

TABLE 20.1
Calculation for Optimum Compressor Installation
Discharge Pressure for Inside
Compressor lb/sq in. Diameter of Distance between Annual
abs, P1 Pipe, in., D Stations, miles, L Cost, $/yr, Cy
300 26 51.6 $21,640
400 23 51.9 $21,180
500 21 53.6 $21,250
600 19 48.7 $21,370
800 17 53.1 $21,560
312 Introduction to Optimization for Chemical and Environmental Engineers

of the components involved, and the physical arrangement of the unit used for distilla-
tion. A common technique used for this separation is continuous distillation. This process
removes and absorbs the less volatile (heavier) components from the vapor, thus efficiently
enriching the vapor with the more volatile (lighter) components.6,7
It is often convenient to express the profitability equation for a mass transfer opera-
tion in terms of some dependent process variable other than the venture worth (W)
itself. Simplification that is possible is based on the observation that unit costs of each
process variable are usually expressible as simple functions of the independent vari-
ables. The modified equation to be differentiated graphically may then be expressed
as follows:

W = Aa + Uu + Vv +  + Zz (20.34)

where unit costs of each process variable A, U, V,⋯, Z are expressed as simple functions of
the independent variables. The dependent variable a is taken as a desired process result,
and the independent variables u, v,⋯, z as the operating conditions whereby it is obtained.
Upon differentiation with respect to each of the independent variables, the following
equations result:

∂W ∂A ∂a ∂U ∂V ∂Z
= a +A +u +U+v ++ z =0
∂u ∂u ∂u ∂u ∂u ∂u
∂W ∂A ∂a ∂U ∂V ∂Z
= a +A +u +v + V ++ z =0
∂v ∂v ∂v ∂v ∂v ∂v (20.35)
 
∂W ∂A ∂a ∂U ∂V ∂Z
= a +A +u +v ++ z +Z=0
∂z ∂z ∂z ∂z ∂z ∂z

A system for employing Equation (20.35), using the plan of variation of one variable at a
time is provided in the following mass transfer application1 developed by Harbert5 that
illustrates its employment in a mass transfer problem involving two independent variables
where process correlations are available for the full range of interest of the variables.
Consider the following application. A chemical by-product is obtained at a constant
rate of 28 mole % solution in water. It is desired to recover this by-product at a purity
corresponding to 99.2 mole % as the overhead from a distillation column. One percent
recovery of this chemical for one day’s operation is worth $4.20. Sufficient heat pickup is
available to bring the feed to its boiling point. The relative volatility for dilute water in a
chemical (top of column) is 1.58, and for a dilute chemical in water (bottom of column)
is 1.75.
The cost of one additional MM (million) Btu of reboiler heat to the column is $0.60. This
figure includes the cost of heat itself, the cost of the condenser cooling to remove the heat,
and the cost of additional reboiler and condenser surfaces, allowing for an appropriate
profit for the investment involved for the additional surface.
For the completely installed column, the cost of one additional plate together with the
section of tower it occupies is estimated to be $96 plus $44/sq ft of area. With 18-in. plate
spacing, 1 sq ft of plate area will handle a vapor load corresponding to 20 MM Btu of
reboiler heat per day. Plate efficiency for these conditions is taken at 90%. It is estimated
that maintenance will amount to 2%/yr of installed cost and that operating time will be
300 days/yr. Management expects a payout time at most of two years.
Advanced Optimization Term Projects 313

It is desired to determine the conditions for the most economical design. The operating
cost may be expressed in terms of the variables,

a = recovery of chemical, % of total

u = reboiler heat duty, MM Btu/day
v = number of equilibrium plates in tower

The interrelations between these variables as calculated for given conditions are shown
in Figure 19.3, which provides the product recovery versus the reboiler heat input for tow-
ers of various numbers of plates (a vs. u at constant v). These curves were obtained by a
trial-and-error computation, using McCabe–Thiele diagrams.6,8
The first step in determining the required optimum is to express the costs of additional
units of the variables (values of A, U, and V) on the basis of dollars per operating day. For
the product recovery, the value given for one additional percent for one day is

A = 4.20

For the reboiler heat variable, the cost per day for 1 MM Btu is $0.60 plus the cost of the
additional plate area required. From the data given, the cost of this ($1/20 sq ft) is

( 44.0 / 20) (0.02 + 1 / 2) / (300)(0.9) = $0.00423 / day

where (44.0/20) is the first cost of the area, (0.02 + 1/2) the annual charge per year, 300 the
operating days per year, and 0.9 the plate efficiency. Then, this cost becomes $0.00423 v for
a tower of v plates, so that

U = − (0.60 + 0.00423 v ) (20.36)

The negative sign denotes a cost instead of profit item. For the number of plates variable,
the total area required per plate is (u/20) sq ft, so that the original cost of one additional
plate is 96.0 + 44.0(u/20). The cost per day of an equilibrium plate then becomes

96.0 + 44.0(u / 20)  1

V=−  0.02 +  = − (0.185 + 0.00423u ) (20.37)
300 × 0.9 2
With these values, appropriate values for the partial derivatives in Equation (19.8) may be
computed. ∂A/∂u, ∂A/∂v, ∂U/∂u, ∂V/∂v, all equal zero, and

∂U
= −0.00423 (20.38)
∂v
∂V
= −0.00423 (20.39)
∂u
Therefore, from Equation (20.35)

 ∂a 
4.20   − (0.60 + 0.00423 v + 0.00423 v ) = 0
 ∂u 
(20.40)
 ∂a 
4.20   − (0.60 + 0.00846 v ) = 0
 ∂u 
314 Introduction to Optimization for Chemical and Environmental Engineers

where ∂a/∂u is the slope of the curves of Figure 20.1. Figure 20.2, the cross plot of the data
of Figure 20.1, as percent recovery versus number of plates for constant amounts of reboiler
heat (a vs. v at constant u) gives, as the slope of its curves, values for ∂a/∂v.
Similarly,

 ∂a 
4.20   − (0.185 + 0.00846 v ) = 0 (20.41)
 ∂u 
A set of reasonable values for the independent variables is next assumed: u = 180, v = 50.
Substitution of v = 50 into Equation (20.40) gives

∂a 1.023
= = 0.244 (20.42)
∂u 4.20
The corresponding value of u is 165 MM Btu/day from Figure 20.2. Substitution of u = 180
into Equation (20.40) gives

∂a 1.705
= = 0.406
∂v 4.20
The slope ∂a/∂v = 4.06 corresponding to the line u = 180 on Figure 20.2 gives v = 39. As a
second trial, u = 165 and v = 39 are assumed. This gives, first ∂a/∂v = 0.222 corresponding to
u = 164, and second, ∂a/∂v = 0.382 corresponding to v = 38.
Alternatively, one might have selected a series of values of v and plotted Equation
(20.50) on Figure 20.1 (dashed line). Similarly, Equation (20.40) can be plotted on Figure 20.1
(dot-dash line). Values for Equation (20.35) could be cross-plotted to Figure 20.1 (dashed
line). The two curves on Figure 20.2 would intersect as shown at approximately the same
value as obtained by the two successive trails; namely u = 164, v = 38.

100
∞ ys
mt f
ra
ilib o. o
equ υ = n
riu

60
55
a, product recovery, % of total

98
50

45

96
40

35
94
Eq.
(3.15)
30

92
160 170 180 190 200 210 220
u, reboiler heat duty, MM Btu/day

FIGURE 20.1
Recovery versus number of plates.
Advanced Optimization Term Projects 315

100
day
ler u/
b oi Bt
re M
u, y, M
a, product recovery, % of total
98 t
du
at 0
he 22 0
1
2 Eq. (3.16)

0
Eq. (3.15)
20

0
19
96

0
18
0
17

94

0
16

92
30 35 40 45 50 55 60
υ, number of plates

FIGURE 20.2
Recovery versus number of plates.

The economic design thus corresponds to 38 equilibrium plates with a reboiler duty of
164 MM Btu/day. Resultant chemical recovery is 93.4%. The tower would have 38/0.9 or 42
actual trays with a vapor area of 164/20 = 8.4 sq ft. This design would result in a maximum
overall profit for the operation.

References
1. J. Happel, Chemical Process Economics, John Wiley & Sons, Hoboken, NJ, 1958.
2. R. Huntington et al., High Pressure Pipe Line Research, Clark Bros., Olean, NY, 1947.
3. R. Ketter and S. Prawler, Modern Methods of Engineering Computations, McGraw-Hill, New York
City, NY, 1969.
4. H. Mickley, T. Sherwood, and C. Reed, Applied Mathematics in Chemical Engineering, McGraw-
Hill, New York City, NY, 1957.
5. W. Harbert, title unknown, Petroleum Refinery, 27, 185, location unknown, 1948.
6. L. Theodore and J. Barden, Mass Transfer, A Theodore Tutorial, Theodore Tutorials, East
Williston, NY, originally published the USEPA/APTI, RTP, NC, 1995.
7. L. Theodore and F. Ricci, Mass Transfer Operations for the Practicing Engineer”, John Wiley &
Sons, Hoboken, NJ, 2010.
8. L. Theodore, Chemical Engineering: The Essential Reference, McGraw-Hill, New York City, NY,
originally published by USEPA/APTI, RTP, NC, 2014.
Index

ABC Drug Laboratory, 161 Biological treatment processes, 126–127

Absorber vs. extractor, 227–232 Biological waste treatments, 138
Accidental errors, 22 Birmingham Wire Gauge (BWG), 285
Advanced optimization term projects, 303–315 Bisection method, 56
fluid transportation system, 306–311 Break-even point operation, 83–85, 117, 125–126
mass transfer application, 311–315 BSE, see Behan Structural Engineers (BSE)
optimum indoor ventilation flow rates, BWG, see Birmingham Wire Gauge (BWG)
304–306
sulfuric acid process, 303–304 Canadian Rocky Mountain Water Plant, 131
Air pollution management, 111–123, 265–268 Cancer, 155, 161
coal-fired power plant options, 113–114 Cancer-fighting drugs cost, 155–156
incinerating mercury-contaminated Capital cost, 112
waste, 119–121 Capital recovery factor (CRF), 112, 145, 185
optimizing incinerator performance, 121–123 Carbon dioxide systems, 176
outdated air pollution control device, 111–112 Carcinogen, 155
particulate control equipment Carnahan, B., 30
options, 114–116 Cartesian coordinates, see Rectangular
recovering dust, 117–118 coordinates
AIT, see Auto-ignition temperature (AIT) CCPs, see Coal combustion products (CCPs)
Algebra, 96 Cdf, see Cumulative distribution function (Cdf)
Analytical methods, 83–93 Cement incinerator operation, 177–180
breakeven considerations, 83–85 Cement kiln vs. rotary kiln incinerator, 139–141
formulation of optimum, 88–89 Centrifugal compressors, 189
one independent variable, 85–88 Centrifugal fan, 184
three independent variables, 93 Centrifugal force, 254
two independent variables, 89–92 Centrifugal pumps, 185, 186
Applied engineering concepts, in linear Chemical engineering applications, 285–300
programming, 102–104 chemical reactors, 287–293
Applied mathematical concepts, in linear fluid flow, 285–287
programming, 99–102 heat exchangers, 295–297
Approximate numbers, 20–22, 21 mass transfer operations, 293–295
Aris, R., 1 plant design, 297–300
Arithmetic relations, 96 Chemical foams, 176
Auto-ignition temperature (AIT), 175 Chemical Process Economics, 60, 264, 303
Chemical reactor (CR) applications, 199–215
Back-mix reactor, 200, 287 achieving maximum conversion, 208–210
Backward curved/inclined fans, 184 batch reactor performance, 212–214
BACT, see Best available control continuous stirred tank reactor vs. tubular
technology (BACT) flow reactor, 199–204
BAT, see Behan and Theodore (BAT) Associates fluidized-bed reactor performance, 204–207
Batch plant operation cost, 258–259 maximizing selectivity, 210–212
Batch reactor performance, 212–214 operating schedules, 214–215
Bathtub curve, 279 Chemical reactors (CR), 287–293
Behan and Theodore (BAT) Associates, 176 Chemical stability, 229
Behan Doyle Environmental Inc., 115–116 Chemical treatment processes, 126–127
Behan Structural Engineers (BSE), 261 Chemical waste treatments, 138
Best available control technology (BACT), 113 Classical algebra, 96
BETH Construction Company, 166 Cleaning bath life, 224

317
318 Index

Coal, 271 Electrostatic precipitators (ESPs), 114

Coal combustion products (CCPs), 113 Elliptical PDE, 39–40
Coal-fired power plant options, 113–114 Encapsulation, 138, 144, 147
Combustion systems, 119 Energy quality, 246–247
Computer age, 6 Energy recovery, 247–248
Concrete mix, 149–150 Environmental engineering
Constraint, 96, 97 applications, 265–283
Contaminants, 191–192 air management, 265–268
Contemporary optimization, 8–11 hazard risk assessment, 279–283
Continuous distillation, 218, 312 DRaT II model, 281–282
Continuous stirred tank reactor (CSTR) vs. DRaT IV model, 282–283
tubular flow (TF) reactor, 199–204, DRaT V model, 283
208–209, 288–291, 293 health risk assessment, 273–277
Conventional optimization procedures, 7–8 solid waste management, 271–272
Copper, 224 water management, 268–271
Corrosiveness, 228 Environmental Protection Agency (EPA), 113,
Counting numbers, see Natural numbers 138, 155, 156, 304
CR, see Chemical reactor (CR) applications Equipment lifetime and salvage value, 112
CRF, see Capital recovery factor (CRF) Errors, 22–23
Critical insulation thickness, 242–245 ESPs, see Electrostatic precipitators (ESPs)
Crude oil, 232–233 Extraction, 229
CSTR, see Continuous stirred tank reactor Extractor, absorber vs., 227–232
(CSTR) vs. tubular flow (TF) reactor Extra expenses, 222
Cumulative distribution function (Cdf), 169 Extrapolation, 17–19
Cyclone selection and design, 254–258
Fabric filter, 115
Determinate error, 22 Failure distribution rate, 169–172
Differentiation, 23–26 Failure rate curve, 279
Digital computer, 6, 95 Fans and blowers, 184
Dilutant and solute, 230 Fan selection, 183–185
Dilution ventilation, 294 FE, see Fitzmaurice Environmental (FE)
Direct approach, 51 Feasible solutions/values, 99
Discounted cash flow application, 144–147 Filter press operation, 219–221
Distillation, 218, 232, 312 Filtration process, 219
Double pipe (DP) heat exchangers, 239–240, Finite differences, 39
241, 247 Fire accidents, 174–175
Doyle Engineers, 166 Fire extinguishing systems, 166
Doyle Financial Consultation, 230 Fire protection equipment, 174–177
DP, see Double pipe (DP) heat exchangers First-degree equation, 67
Drag-out reduction, 224 First law of thermodynamics, 246
DRaT II model, 280, 281–282 First-order differential equations, 32
DRaT IV model, 280, 282–283 First-order ODEs, 34, 36
DRaT V model, 280, 283 Fitzmaurice Environmental (FE), 303
Dry chemical systems, 176 Flue gas emissions, 158–160
Dummying process, 224 Fluid flow, 285–287
Dupont, R., 159, 274, 280 Fluid flow applications, 183–196
Dust recovering, 117–118 fan selection, 183–185
Dynamic programming, 9, 11, 95, 96 pipe diameter selection, 188–189
pump selection, 185–187
Economic lot size, 221 three-stage compressor, 195–196
Economics and optimization applications, 10 two-stage compressor, 189–191
Edison, Thomas, 5 ventilation models, 191–194
Edisonian approach, 5 Fluidization, 204
Index 319

Fluidized-bed reactor performance, 204–207 Health risk assessments (HRAs), 153–163,

Fluid transportation system, 306–311 158–159, 273, 273–277
Flynn, Annmarie, 298 cancer-fighting drugs cost, 155–156
Foam insulation, 238 health concerns associated with flue gas
Foam systems, 176 emissions, 158–160
Fogler, S., 200 insurance costs, 157–158
Forward curved fans, 184 minimum purging time, 153–155
Freezing point, 229 pharmaceutical cancer drug, 161–163
Full logarithmic graph, 68, 71 vitamin caloric and protein requirements,
160–161
Gas absorption, 227, 265 Heat exchangers, 239, 295–297
Gaseous pollutants, 112, 265 Heat transfer applications, 235–248
Gas permeation (GP), 133 critical insulation thickness, 242–245
Gas phase reactions, 213 maximizing profit through energy recovery,
Gas solubility, 228 247–248
Gauss elimination, 30 optimum insulation thickness, 238–239
Gauss–Jordan reduction, 30 profitable exchanger, 239–242
Gauss–Seidel approach, 31 recovering quality energy, 246–247
Geometrical patterns, 5 steam options, 235–237
Golden section method, 57–60 Heinsohn, R., 192
GP, see Gas permeation (GP) High-expansion foams, 176
Graphical approaches, 65–80 High-speed computing equipment, 6
applications, 77–80 HRAs, see Health risk assessments (HRAs)
logarithmic-logarithmic (log-log) Hydrocarbon vapors, 172
coordinates, 68–73 Hydrocracker, 232
methods of plotting data, 76 HZRA, see Hazard risk assessment of
rectangular coordinates, 66–68 accidents (HZRA)
intercepts, 68
linear vs. non-linear, 67–68 IAQ, see Indoor air quality (IAQ)
quadrants, 67 Incineration, 119, 138
semi-logarithmic (semi-log) Incinerator performance optimization, 121–123
coordinates, 73–76 Incipient fluidization, see Minimum
plotting straight lines, 73–76 fluidization
triangular (trilinear) coordinates, 76 Independent design variables, 98
Graphical/tabular presentation, 17–18 Indeterminate errors, 22
Gravity filters, 220 Indirect approach, 51
Guidelines for Carcinogen Risk Assessment, 155 Indoor air pollution, 191
Indoor air quality (IAQ), 304
Hampton, Rufus, Jr., 304 Industrial ventilation, 191
Happel, J., 60, 88, 104, 264, 271, 303, 306 Industrial wastewater, 126
Hazard risk assessment, 165–180, 279–283 Inert waste, 9
cement incinerator operation, 177–180 Inherited error, see Propagation error
DRaT II model, 281–282 Insurance costs, 157–158
DRaT IV model, 282–283 Integer linear programming, 8
DRaT V model, 283 Intercepts, 68
failure distribution rate effect, 169–172 Interpolating method, 25
fire protection equipment, 174–177 Interpolation, 17–19
plant safety proposal comparisons, 167–169 Interval halving method, 53–55
safety expenditures vs. return of safety Inverse interpolation, 18, 19
investments, 165–167
worst-case explosions effect, 172–174 Kaufmann, D., 295
Hazard risk assessment of accidents KBLT Power Plant, 168
(HZRA), 279 Ketter, R., 31, 37
320 Index

Latent enthalpy effects, 235 simultaneous linear algebraic equations, 29–31

Law of conservation of energy, 246 Gauss elimination, 30
Leaf filters, 220 Gauss–Jordan reduction, 30
Least squares, 25 Gauss–Seidel approach, 31
LI, see Liquid injection (LI) incinerators Mathematical optimization technique, 103, 104
Linear algebra, 96 McCabe–Thiele diagrams, 313
Linear programming, 8–11, 95–106, 182 Mechanical foams, 176
applied engineering concepts, 102–104 Mechanism of combustion, 120
applied mathematical concepts, 99–102 Membrane construction costs, 133–135
definitions, 96–97 Membrane processes, 133
optimization concepts, 97–99 Mercury-contaminated waste incineration,
real-world applied concepts in linear 119–121
programming, 105–106 Mercury waste emissions, 147–149
Linear vs. non-linear equation, 67–68 Methylation, 147
Liquefied nitrogen gas (LNG), 84 Methyl mercury, 147
Liquid injection (LI) incinerators, 139–140 MF, see Microfiltration (MF)
Liquid–liquid extraction, 229 Microfiltration (MF), 133
Liquid phase reactions, 213 Minimum expenses, 222
Liquid–solid changes, 236 Minimum fluidization, 205
LNG, see Liquefied nitrogen gas (LNG) Mixed solvent, 230
Logarithmic-logarithmic (log-log) Monte Carlo approach, 39, 177, 179
coordinates, 68–73 Multi-independent variable equations, 52
Logarithmic-probability (log-normal) Multilayer insulation, 238
graphics, 76 Multistage decision process, 95
Multivariable optimization problem, 97
Mass transfer applications, 217–233, 311–315
absorber vs. extractor, 227–232 NAHEB Construction Associates, 157
filter press operation, 219–221 Nanoparticles, 192
optimum reflux ratio, 217–219 National Fire Protection Association, 166
process and equipment modifications, National Science Foundation (NSF), 192
223–227 Natural numbers, 13
processing crude oil, 232–233 Negative integers, 13
production rates, 221–223 Newton–Raphson (NR) method, 31, 52
Mass transfer operations, 293–295 Newton’s method of tangents (NMT), 32
Material and waste exchanges, 132–133 Non-inert waste, 9
Material substitution, 223 Nonlinear algebraic equations, 31–32, 52
Mathematical ideas and concepts, 5 Non-linear equation, linear vs., 67–68
Mathematical operations, 13–40 Nonlinear programming, 9, 95
approximate numbers, 20–22 NR, see Newton–Raphson (NR) method
differentiation, 23–26 NSF, see National Science Foundation (NSF)
errors, 22–23 Numerical error, 36
extrapolation, 17–19 Numerical integration, 26–28
interpolation, 17–19 Simpson’s rule, 28
nonlinear algebraic equations, 31–32 trapezoidal rule, 27–28
numerical integration, 26–28 Numerical optimization algorithm, 5
Simpson’s rule, 28 Nypro Ltd. caprolactam factory, 172
trapezoidal rule, 27–28
ordinary differential equations, 32–36 Objective function, 98, 99
partial differential equations, 37–40 O’Brien, Shannon, 298
elliptical, 39–40 ODEs, see Ordinary differential equations (ODEs)
parabolic, 37–38 One independent variable, 42–45, 85–88
quadratic equation, 15–17 One-independent variable search methods, 60
significant figures, 20–22 OPEC oil embargo (1973), 246
Index 321

Operating costs, 112, 222 Plant safety proposal comparisons, 167–169

Operating schedules, 214–215 Plate-and-frame filters, 220
Operations Research, 11, 97 Plotting data methods, 76
Optimization, 1, 3–11, 109 Positive displacement pumps, 185
computer age, 6 Positive integers, 13
concepts, 97–99 Powder insulation, 238
contemporary, 8–11 Prawler, S., 31, 37
definition, 1 Precision and accuracy, 22
history, 4–6 Primitive counting systems, 5
linear programming, 8–11 Probability distribution function (Pdf), 169
model, 95 Propagation error, 23
procedures, 7–8 Protective systems, 165
scope, 7 Protein molecules, 160
Optimum indoor ventilation flow rates, 304–306 Pump selection, 185–187
Optimum insulation thickness, 238–239 Purging time, 153–155
Optimum reflux ratio, 217–219 Pythagorean theorem, 96
Ordinary differential equations (ODEs), 32–36
Organic wastes, 138 Quadrants, 67
Organization costs, 222 Quadratic equation, 15–17
Outdated air pollution control device, 111–112 Quadratic formula, 16
Oxidation, see Incineration
Radial blade fans, 184
Parabolic PDE, 37–38 Raffinate, 230
Partial differential equations (PDEs), 37–40 Random number generators, 177
elliptical, 39–40 RCRA, see Resource Conservation and
parabolic, 37–38 Recovery Act (RCRA)
Particulate control equipment options, 114–116 Real numbers, 13
Particulate pollutants, 112, 265 Real-world applied concepts, in linear
Pasquill-Gifford model, 276 programming, 105–106
Pathogens, 128 Reciprocating compressors, 189
PD, see Plant design (PD) applications Reciprocating pumps, 185, 186
PDEs, see Partial differential equations (PDEs) Rectangular coordinates, 66–68
Pdf, see Probability distribution function (Pdf) intercepts, 68
Performance measure, 97 linear vs. non-linear, 67–68
Perry’s Handbook, 95, 238 quadrants, 67
Perturbation techniques, 41–49 Refractory materials, 238
one independent variable, 42–45 Regression equation, 25
three independent variables, 48–49 Resource Conservation and Recovery Act
two independent variables, 45–48 (RCRA), 132, 138
Pharmaceutical cancer drug, 161–163 Return of safety investments, safety
Physical treatment processes, 126–127 expenditures vs., 165–167
Physical waste treatments, 138 Reverse osmosis (RO), 133
Pipe diameter selection, 188–189 Rezifp Pharmaceuticals and Chemicals
Pivot equation, 30 Corp., 214
Plant design (PD), 297–300 Ricci, F., 176, 280, 283, 298
Plant design (PD) applications, 249–262 Risk, 158, 274
batch plant operation cost, 258–259 RK, see Rotary kiln (RK) incinerator, cement
cyclone selection and design, 254–258 kiln vs.; Runge–Kutta (RK) method
shipping facilities, 249–252 RO, see Reverse osmosis (RO)
structure design, 261–262 Rotary compressors, 189
tank farms, 252–254 Rotary kiln (RK) incinerator, cement kiln vs.,
ventilation models with system variables, 139–141
259–261 Rotary pumps, 185, 186
322 Index

Rotary vacuums filters, 220 ST, see Shell-and-tube (ST) heat exchangers
Round-off error, 23, 30, 36 Statistical methods, 76
Runge–Kutta (RK) method, 32–4 Steam jet systems, 176
Steam options, 235–237
Safety expenditures vs. return of safety Steam Tables, 236
investments, 165–167 Steepest ascent/descent method, 60–63
SCC, see Secondary combustion chamber (SCC) Straight blade fans, see Radial blade fans
Scientific notation, 20 Structure design, plant, 261–262
Search methods, 51–63 Sulfuric acid process, 303–304
bisection, 56 Superinsulation, 238
golden section, 57–60 Superproduction costs, 222
interval halving, 53–55
steepest ascent/descent, 60–63 Tank farms, 252–254
Secondary combustion chamber (SCC), 141 TF, see Tubular flow (TF) reactor, continuous
Second law of thermodynamics, 246, 247 stirred tank reactor (CSTR) vs.
Second-order polynomial equation, 25 Theodore, Lou, 157, 243, 280, 283, 296, 298
Seek methods, see Direct approach Theodore approach, 5
Selectivity, 210–212 Theohan Gas Company, 84
Semi-logarithmic (semi-log) coordinates, 73–76 Thermal conductivity, 238
plotting straight lines, 73–76 Thermal insulation, 238–239
triangular (trilinear) coordinates, 76 Thermal waste treatment, 138
SF, see Slope factor (SF) Three independent variables, 48–49, 93
Shaefer, S, 280 Three-point equal-interval search method, 53
Shell-and-tube (ST) heat exchangers, 239, 241 Three-stage compressor, 195–196
Shipping facilities, 249–252 “Three T’s of combustion,” 120
Significant figures, 20–22 TNT, 172–174
Simpson’s rule, 28 Toxicity, 229
Simultaneous linear algebraic equations, 29–31 Transient operation, 212
Gauss elimination, 30 Trapezoidal rule, 27–28
Gauss–Jordan reduction, 30 Treybal, R., 229
Gauss–Seidel approach, 31 Trial-and-error method, 5, 41, 60, 98, 99, 104, 296,
Single-pass cylindrical tube, 201 313
Single-payment compound amount factor, 144 Triangular coordinates, 76
Single-phase heat exchanger, 241 Truncation error, 23
Single solvent, 230 Tubular flow (TF) reactor, continuous stirred
Single-variable optimization, 97 tank reactor (CSTR) vs., 199–204,
Slope factor (SF), 155, 156 208–209
Slope-intercept method, 65, 70, 73 Tumble burners, 141
Slurry flows, 219 Two independent variables, 45–48, 89–92
Solidification, 138 Two-phase heat exchanger, 241
Solid–liquid extraction, 229 Two-stage compressor, 189–191
Solid waste management, 137–150, 271–272
cement kiln vs. rotary kiln incinerator, Ultrafiltration (UF), 133
139–141 Unconfined vapor cloud explosions, 172
discounted cash flow application, 144–147 Unimodal function, 51, 60
maximum mercury waste emission Unsteady-state operation, see Transient
constraints, 147–149 operation
sludge waste receiving tank size, 142–143 US Department of Agriculture (USDA), 160
solid waste treatment, 137–139
structural considerations of concrete mix, Vacuum insulation, 238
149–150 Vapor–liquid changes, 236
Solid waste treatment, 137–139 Vapor–solid changes, 236
Solvent, 230 Ventilation models, 191–194, 259–261, 293
Index 323

Viscosity, 229 waste revenues from waste streams, 132–133

Vitamin caloric and protein requirements, wastewater treatment equipment options,
160–161 126–127
Volatility, 228 water disinfection options, 127–130
water plant operation, 131
Walas, S., 202 Water quality problem, 127
Wastewater treatment equipment options, Water sprays, 114, 176
126–127 Water systems, 176
Waterborne diseases, 128 Weibull distribution, 279–280
Water disinfection options, 127–130 Wet collectors, 114, 115
Water management, 268–271 Wet scrubbers, see Wet collectors
Water plant operation, 131 Wilkes, J., 30
Water pollution management, 125–135 Worst-case explosions effect, 172–174
break-even point operation, 125–126
membrane construction costs, 133–135 Zeroes, 21