Instant Ebook Access, One Click Away – Begin at ebooknice.

com

(Ebook) Business Optimization Using Mathematical

Programming: An Introduction with Case Studies and
Solutions in Various Algebraic Modeling Languages
... Research & Management Science, 307) by Josef
Kallrath ISBN 9783030732363, 3030732363
https://ebooknice.com/product/business-optimization-using-
mathematical-programming-an-introduction-with-case-studies-
and-solutions-in-various-algebraic-modeling-languages-
research-management-science-307-34418854

OR CLICK BUTTON

DOWLOAD EBOOK

Get Instant Ebook Downloads – Browse at https://ebooknice.com

 Instant digital products (PDF, ePub, MOBI) ready for you
Download now and discover formats that fit your needs...

Start reading on any device today!

(Ebook) Introduction to Management Science A Modeling and Case Studies Approach with
Spreadsheets by Frederick Hillier ISBN 9781265040055, 1265040052

https://ebooknice.com/product/introduction-to-management-science-a-modeling-and-
case-studies-approach-with-spreadsheets-49113838

ebooknice.com

(Ebook) Biota Grow 2C gather 2C cook by Loucas, Jason; Viles, James ISBN
9781459699816, 9781743365571, 9781925268492, 1459699815, 1743365578, 1925268497

https://ebooknice.com/product/biota-grow-2c-gather-2c-cook-6661374

ebooknice.com

(Ebook) Introduction to Management Science: A Modeling and Case Studies Approach

with Spreadsheets by Frederick Hillier, Mark Hillier ISBN 9780078096600, 007809660X

https://ebooknice.com/product/introduction-to-management-science-a-modeling-and-
case-studies-approach-with-spreadsheets-10953826

ebooknice.com

(Ebook) Introduction to Management Science: A Modeling and Case Studies Approach

with Spreadsheets by Hillier, Frederick, Hillier, Mark ISBN 9781260091854,
1260091856

https://ebooknice.com/product/introduction-to-management-science-a-modeling-and-
case-studies-approach-with-spreadsheets-28404354

ebooknice.com
(Ebook) Introduction to Management Science: A Modeling and Case Studies Approach
with Spreadsheets 3rd Ed by Frederick S. Hillier ISBN 9780073129037, 0073129038

https://ebooknice.com/product/introduction-to-management-science-a-modeling-and-
case-studies-approach-with-spreadsheets-3rd-ed-2030074

ebooknice.com

(Ebook) Introduction to Management Science: A Modeling and Case Studies Approach

with Spreadsheets by Frederick S. Hillier, Mark S. Hillier ISBN 9781265040055,
1265040052

https://ebooknice.com/product/introduction-to-management-science-a-modeling-and-
case-studies-approach-with-spreadsheets-57858332

ebooknice.com

(Ebook) Fundamentals of Algebraic Modeling: An Introduction to Mathematical Modeling

with Algebra and Statistics by Daniel L. Timmons, Catherine W. Johnson, Sonya McCook
ISBN 9780495555094, 0495555096

https://ebooknice.com/product/fundamentals-of-algebraic-modeling-an-
introduction-to-mathematical-modeling-with-algebra-and-statistics-2330222

ebooknice.com

(Ebook) Learning and Researching with Case Studies: A Student Companion for Business
and Management Research by Scott Andrews ISBN 9781032386324, 1032386320

https://ebooknice.com/product/learning-and-researching-with-case-studies-a-
student-companion-for-business-and-management-research-51087140

ebooknice.com

(Ebook) Introduction to Internet of Things in Management Science and Operations

Research: Implemented Studies (International Series in Operations Research &
Management Science, 311) by , ISBN 9783030746438, 3030746437

https://ebooknice.com/product/introduction-to-internet-of-things-in-management-
science-and-operations-research-implemented-studies-international-series-in-
operations-research-management-science-311-34834248

ebooknice.com
International Series in
Operations Research & Management Science

Josef Kallrath

Business Optimization
Using Mathematical
Programming
An Introduction with Case Studies and
Solutions in Various Algebraic Modeling
Languages
Second Edition
International Series in Operations Research
& Management Science
Founding Editor
Frederick S. Hillier
Stanford University, Stanford, CA, USA

Volume 307

Series Editor
Camille C. Price
Department of Computer Science, Stephen F. Austin State University,
Nacogdoches, TX, USA

Associate Editor
Joe Zhu
Foisie Business School, Worcester Polytechnic Institute, Worcester, MA, USA
More information about this series at http://www.springer.com/series/6161
Josef Kallrath

Business Optimization Using

Mathematical Programming
An Introduction with Case Studies and
Solutions in Various Algebraic Modeling
Languages

Second Edition
Josef Kallrath
Department of Astronomy
University of Florida
Gainesville, FL, USA

ISSN 0884-8289 ISSN 2214-7934 (electronic)

International Series in Operations Research & Management Science
ISBN 978-3-030-73236-3 ISBN 978-3-030-73237-0 (eBook)
https://doi.org/10.1007/978-3-030-73237-0

Previously published under the imprint Palgrave

© Springer Nature Switzerland AG 2021

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

This Springer imprint is published by the registered company Springer Nature Switzerland AG.
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Dedication (2nd Edition)

To Albert & Diana

Dedication (1st Edition)

To those who increased my1 pleasure in mathematics:

Wilhelm Braun (1970–1975)
Klaus Reusch and Wilhelm Gieselmann (1976–1979)
Gerard de Beuckelaer (1989–1992).

To2 Helen, Alex, Tim, and Jack.

1 Josef Kallrath.
2 John M. Wilson.
Foreword

Mathematical optimization is an inherent paradigm of modern operational research

(OR) for 80 years already. Although many subfields of OR got a kind of disciplinary
independence over this period of time, “searching for the best” is still the main
challenge for OR. Anyway, all sister disciplines of OR still benefit from more
and more efficient methods of optimization developed within OR. Would machine
learning have such a great trajectory without optimization? Would we hear about the
success of deep learning if a proper optimization method would not be available?
The present book reflects a continued interest in mathematical optimization. Its first
edition written 20 years ago provided a faithful picture of optimization methods
and their applications at the turn of the last century. An important progress in the
methodology and applications of optimization, that we were able to observe over
the last two decades, as well as a continuing interest in the first edition of the book,
encouraged its first author to write the current update.
What one can learn from this book? A reader who remembers the first edition
will quickly acknowledge that this edition goes far beyond the previous content. In
particular, it includes new material on stochastic programming and nonlinear mixed
integer optimization. Global optimization gets also a special place for its capacity of
solving non-convex continuous or mixed integer nonlinear problems. The relation
of global optimization to mixed-integer optimization, Lagrange relaxation, and
polylithic modeling (sequence of models for one complex problem) is illustrated
by interesting real-world examples from the academic publishing industry. Some
topics, like phi-function techniques for cutting and packing, are not covered by other
books on optimization known to me. A valuable feature of this book is also examples
coded in popular algebraic modeling languages, and available online.
The book is written in full respect of the link existing between three necessary
components of practical decision-making: “model-algorithm-software.” In this way,
the reader learns how real-world problems can be modeled in mathematical
terms understood by optimization algorithms, and then solved with commercially
available software. This holistic approach to solving complex problems with opti-
mization methods makes this book self-contained. Moreover, because the content of
theory, didactic examples, and real-world case studies are perfectly balanced in the

ix
x Foreword

book, the reader gets a convincing report on the practical relevance of contemporary
mathematical optimization.
As a coordinating editor of the European Journal of Operational Research
(EJOR) since 20 years, I can say that the character of this book is perfectly consistent
with our editorial policy, in the sense that, similarly to EJOR, it underlines the link
between the state-of-the-art methodology of optimization with practice of decision-
making. Thus, I am pleased to congratulate Josef Kallrath on the excellent update of
the first edition, and I am confidently recommending this book to students, teachers,
researchers, and industrial practitioners who wish to learn how to effectively cope
with complex decision problems using mathematical optimization.

Poznań, Poland Roman Słowiński

August 2020
Preface to the 2nd Edition

With the continuing interest in my book, it is a pleasure, approximately 20 years

after the completion of the first edition, to enhance it with some topics which, after
two decades, have gained importance or led to a change in perspective.
This book, as the first edition, introduces business optimization using mathemat-
ical programming (optimization) not only from a practitioner’s point of view but
also from a researcher’s perspective. It covers the entire process of solving a real-
world decision problem by mathematical optimization: structuring and formulating
the problem (free of mathematics) as well as collecting the input data, translating
it into the mathematical language obtaining an optimization model, solving it, and
validating the results. Linear programming case studies exemplify and showcase
the learnings. Building on these concepts, the book examines mixed integer
linear programming problems and presents problem formulations and case studies
for these applications, and extends into the nonlinear optimization world, both
continuous and discrete. Another series of case studies show optimization in practice
employing integer programming, leading to larger practical examples, e.g., from
production or supply chain planning, network planning, or cutting and packing.
Hints are given on how users can control the optimization process and improve
its efficiency. The book concludes with observations on the impact and implications
of optimization in business. New to the second edition, many examples are coded
in the algebraic modeling languages — AMPL, GAMS, FICO(R) Xpress Mosel, and
SAS/OR — and are available online.
The second edition goes beyond linear optimization, i.e., beyond linear pro-
gramming and mixed integer linear programming. Thus, the previous Chap. 11 has
been renamed in Beyond LP and MILP Problems, and now includes fractional
programming and its transformation into linear programming, successive linear
programming as a special solution technique of nonlinear optimization, and opti-
mization under uncertainty (especially stochastic programming). Also in Chap. 11,
for quadratic programming, which is again a special case of nonlinear optimization,

xi
xii Preface to the 2nd Edition

we provide an equivalent formulation based on special ordered sets. Chapter 12

provides an introduction into nonlinear continuous and mixed-integer optimization.
More attention is paid to the field of Global Optimization, as after 2002, strong
commercial software using deterministic global optimization techniques became
available for solving non-convex continuous or mixed integer nonlinear problems.
The techniques used in this type of optimization are closely related to those used
to solve mixed-integer optimization problems and are described in Chap. 12. They
are illustrated with applications from paper and metals industry in Chap. 13. A
further coupling of global optimization and mixed-integer optimization becomes
apparent in the calculation of optimal breakpoint systems in connection with the
modeling of nonlinear terms using SOS2 variables. In Chap. 14, Polylithic Modeling
and Solution Approaches, we demonstrate how to use these SOS2 variables for
obtaining good approximations for solving nonlinear problems. In the context of
mixed-integer linear problems we also consider the topic of Lagrange relaxation for
the improvement of lower bounds. Overall, as this second edition covers also the
nonlinear optimization world, it is somewhat more demanding on the mathematical
side. Therefore, it also contains Appendix C summarizing some of the fundamentals
of linear algebra and calculus.
As cutting and packing is used in many examples throughout this book, and cut-
ting and packing industry significantly contributes to the GDP of several countries,
a full chapter on it has been added. Most of the phi-function technique material
in Chap. 15 has been provided by Prof. Dr. Yuriy Stoyan and Prof. Dr. Tatiana
Romanova.3 This technique is a generic approach beyond and within mathematical
programming, very suitable for solving large-scale real-world cutting and packing
problems in 2D and 3D.
In Tools around Optimization in Chap. 17 we have added a section on algebraic
modeling languages. Later in this chapter we address the importance of visualization
of input data and output results.
Many examples in this book are coded in the algebraic modeling languages
mp-model, GAMS, Mosel, and SAS/OR — online at www.springer.com — and
referenced as MCOL (Model Collection OnLine). This online access granted to
readers of this book replaces the CD-ROM which came with the first edition. For
compatibility with the first edition, we keep the mp-model files as well. In this
second edition, we avoid using or explaining software specific syntax and keep
implementation issues on a generic level.

3 The National Academy of Sciences of Ukraine, Institute of Mechanical Engineering Problems,

Department of Mathematical Modeling and Optimal Design, Kharkiv, Ukraine and Kharkiv
National University of Radioelectronics, Department of Applied Mathematics.
Preface to the 2nd Edition xiii

Reading flow suggestions by target group and sequence of chapters:

flow target group sequence

F1 novice, beginners 1–17
F2 linear optimization 1−10
.
F3 nonlinear optimization 11−15
F4 practical optimization 1, 5, 10, 13−17
F5 advanced readers 10, 11, 13−15

In the first edition, a few larger case studies had been analyzed and solved with
XPRESS-MP, the predecessor of what is now FICO Xpress Mosel and FICO
Xpress Optimization, or Xpress in short. We have kept the text from the
first edition in this second edition and continue referring to this modeling and solver
XPRESS-MP from the 1990s in this context. We proceed similarly for currencies
as well as the hardware or software equipment used. The real-world case studies
from the 1990s remain unchanged. As XPRESS-MP and its modeling language
mp-model use integer indices but not index sets, the first edition and its case
studies are formulated using numerical indices. In this second edition, we leave
the case studies untouched but use index sets in all material added. The formerly
large MILP problems from 20 years ago have become toy examples for commercial
solvers in 2020 — and the spreadsheet software LOTUS-1-2-3 does not seem
to be in use any longer. This should serve as a good warning and advise for the
future. Only ASCII data has a reasonably long lifetime. Everything else strongly
depends on market changes. The lifetime of software seems to be very limited —
and even if the software still exists, backward compatibility is not guaranteed either.
Fortunately, the first edition has been written in LATEX, and is retained, largely, in
the second edition.
Overall, we hope that the reader benefits from this second edition in various
ways: providing a path to efficient modeling, and learning what is relevant nowa-
days, and which problems can be handled efficiently concerning mixed integer
optimization and non-convex nonlinear optimization. If the reader has fun on the
way reading this book, all the better.
Acknowledgment
It is a pleasure to again thank some friends and colleagues who have accompanied
me for many years in my work or private environment and who in various ways
have directly or indirectly contributed to the success of this book. This applies
to everyone who has already been mentioned in the acknowledgment of the first
edition. Between 1997 and 2020, I established close relationships or friendships
with many new collaboration partners. These are, especially, Dr. Franz Nelißen
and Dr. Michael Bussieck (GAMS Software GmbH, Frechen and Braunschweig,
Germany), Prof. Dr. Christoudoulos A. Floudas (Princeton University, Princeton,
xiv Preface to the 2nd Edition

NJ, USA),4 and Prof. Dr. Panos M. Pardalos (Center of Applied Optimization,
University of Florida, Gainesville, FL). The contacts from my leadership of the
GOR working group Praxis der Mathematischen Optimierung have also contributed
indirectly to this expanded second edition.
A special word of thanks is directed to all the (new) project partners during
the last 20 years: I have cherished many of them because of their enthusiasm and
their in-depth knowledge of their application areas — and I have all of them in
good memory. They all had in common that, for improving the benefits of their
company, they wanted to enhance their methods in their applications areas by
thorough mathematical modeling, which often led to mixed integer optimization.
Interacting and communicating with them has been an important element during
the projects and has some influence on this book. From the many German BASF
project partners who have contributed indirectly to this book over several years, I
would like to mention a few: Dr. Wolfram Schmidt and his team with Dr. Markus
Klumpe and Bernd Heisel-Hoffmann, Norbert Vormbrock with a common history
at Bonn University, and Dr. Gerd Fischer with his exciting rail car projects. It has
been a pleasure to work with such people who feel the need to understand and solve
a problem as deeply and as thoroughly as possible. High quality and sustainable
solutions require deep understanding, dedication to detail, and the will to solve a
problem on one’s own initiative. I strongly hope that the project partners mentioned
above will still have sufficient time left for focusing deeply on their projects and that
they can enjoy their work.
For a thorough examination and proofreading of the manuscript as well as
many constructive comments and suggestions that have improved the book, it is a
pleasure to thank Dr. Jens Schulz and Dr. Susanne Heipcke (FICO, Berlin, Germany
and Marseille, France); Jan-Erik Justkowiak (Siegen University, Siegen, Germany);
Dr. Philipp M. Christophel (SAS Institute, Heidelberg, Germany); Dr. Johannes
Schlöder (IWR, Heidelberg University, Heidelberg, Germany); Prof. Dr. Iiro Har-
junkoski (Hitachi ABB Power Grids, Mannheim, Germany); Prof. Dr. Eugene
F. Milone (University of Calgary, Calgary, Canada); Prof. Dr. Tapio Westerlund
(Abo University, Finland); Prof. Dr. Ivo Nowak (Hochschule für angewandte
Wissenschaften Hamburg, Hamburg, Germany); Prof. Dr. Alexandra Newman and
Prof. Dr. Tulay Flamand and Phillip Bülow and Louis Kamga and Oluwaseun
Ogunmodede (Colorado School of Mines, Golden, CO, USA); John Cox (US Air
Force, Colorado School of Mines, Golden, CO, USA); Prof. Dr. David Morton
(Northwestern University, Evanston, IL, USA); Dr. Joonghyun Ryu (Hanyang
University, Seoul, Korea); Dominik Schweisgut who also wrote a first draft of
Appendix C (Heidelberg University, Heidelberg, Germany); Prof. Eli V. Olinick
(Southern Methodist University in Dallas, TX, USA); Prof. Dr. Ignacio E. Gross-
mann, Can Li, and Prof. Dr. Destenie Nock (Carnegie Mellon University, PA, USA);

4 Unfortunately, for the whole community, Prof. Floudas passed away in August 2016. I lost a

close collaborator and friend, we shared many common ideas and had joint activities since the
early 1990s.
Preface to the 2nd Edition xv

Dr. Michael Bussieck, Frederick Fiand, and Dr. Stefan Vigerske (GAMS Software
GmbH, Braunschweig, Germany); Dr. Anna Schreieck (Neustadt a.d. Weinstraße,
Germany); Prof. Dr. Stefan Helber (Leibniz University Hanover, Hanover, Ger-
many); and Prof. Dr. Siegfried Jetzke (Ostfalia University of Applied Sciences,
Salzgitter, Germany).
I thank Erwin Kalvelagen (www.amsterdamoptimization.com) for his kind
permission to use his GAMS file lagRel.gms in MCOL and parts of his description
on Lagrange relaxation in this book. From Fair Issac Corporation, I received the
permission to use the FICO® Xpress Mosel modeling examples and FICO® Xpress
Insight visualization examples. FICO is a trademark of Fair Isaac Corporation.
Finally, I thank my daughter Diana for producing the cartoon-like illustrations in
this book and proofreading, and Christian Rauscher, the editor in charge at Springer
(Heidelberg), with whom I had been working for many years — and who has
promoted this second edition of the book.

Weisenheim am Berg, Germany Josef Kallrath

September 2020
Preface

This book arose from a realization that modeling using mathematical programming
should be tightly linked with algorithms and their software implementation to solve
optimization problems. Such linkage is necessary for a full appreciation of the meth-
ods used to model problems that will ensure they can be solved successfully. While
there exist textbooks concentrating on the pure mathematics aspects of optimization,
and others which just describe applications without providing sufficient technical
background, we see our book as trying to provide a link between applications and the
mathematics required to solve real-world problems. Few textbooks have integrated
modeling with state-of-the-art commercially available software. Our book will also
incorporate this missing link and will include the software to solve the models
discussed.
Optimization using mathematical programming is an important subject area as
it can determine the dramatic savings available to organizations that could not be
achieved by other means. In this book, examples are cited where organizations are
saving many millions of pounds (sterling) or dollars (US) by using optimization
methods. Mathematical optimization models are part of tools that can help people
in the process of making decisions concerning the use of resources and saving costs.
Mathematical programming also provides a way to solve problems that, because
of their size or other features, would not otherwise be solvable by other methods. In
major cities, e.g., London, mathematical programming models influence the control
of the flow of domestic water through the city as the model is used to determine the
most efficient strategy to move water from source to user as peaks and troughs in the
usage pattern develop. Thus, the results from mathematical programming models
are literally all around many of us.
The need for a source book of material on the subject was recognized while
teaching at Heidelberg University and Loughborough University and while planning
conference sessions on the practical relevance of mixed integer optimization.
Although there is an extensive literature on mathematical programming, the
paucity of instructional materials in the area of efficient modeling and solving real-
world problems is striking. The student, researcher, or industrial practitioner must
read between the lines of material, usually only available in journal articles or

xvii
xviii Preface

similar, to glean the details of the modeling process and the “tricks of the trade”.
Yet the need is acute: as with many other areas of science, the computer revolution
has given many modelers in industry as well as at universities the tools to attempt
to solve realistic and complex models. In this work, we endeavor to provide a
suitable background as an aid to the novice modeler, a useful reference book for
the experienced modeler, and a springboard for the development of new modeling
ideas. In particular, by tailoring this book around a commercially available software
package we are able to illustrate some of the subtle details that determine the success
or failure of the modeling efforts.
Readership
This book has been planned for use by more than one type of readership. Most of this
book is designed to be used by readers who possess fairly elementary mathematical
skills, i.e., the use of algebraic manipulation, and it is made clear which sections
are not of this type. Further mathematical skills required are developed during
the course of the book but the presentation should not prove too daunting. The
material is suitable for use in courses in Business and Management Studies and
operations research environments. Readers with stronger mathematical skills (e.g.,
linear and matrix algebra) and experienced practitioners in the field will still find
much to interest them as the logic of modeling is developed. The book, therefore,
will provide appropriate course material for lecture courses, short courses and self-
teaching on the topics contained in it.
As some material is for the more advanced reader, or for the reader to use on
a second pass through the book, certain sections in chapters have been marked
as “advanced”. These sections may be omitted on a first pass through the book.
The more advanced parts of the book are written in such a way that it is sufficient
if the reader is familiar with the basic concepts and techniques of linear algebra.
A discussion of some foundations of optimization is provided at the end of some
chapters, where it is helpful if the reader has familiarity with calculus techniques. It
is also expected that the later advanced chapters will be read only once the reader
has started to build models in earnest. A glossary at the end of the book will provide
further help.
Scope
The focus of the book is primarily on models, model applications and individual
case studies rather than algorithmic details. However, because the success of
solution of complex problems requires efficient problem solving, it is important
that models and algorithms are tightly connected. Therefore, we also concentrate
on the mathematical formulation of models and the mathematical background of
the algorithms. The understanding of the mathematics involved in a problem or
model explains why certain model formulations work well while others do not.
We have tried to present in this book a self-contained treatment of the subject
where possible. The presentation of the material is not too far away from what
real modeling in business looks like. Most of the case studies have a commercial
Preface xix

or industrial background. For instance, some of the case studies in Chap. 10 stem
from problems recently analyzed and solved in a mathematical consultancy group
in the chemical industry.
Organization
Chapter 1 gives an introduction and overview of the field. Parts of this chapter,
in particular the details on the software used in this book, can be skipped by the
experienced practitioner. An overview on the history of optimization is presented
in the appendix to Chap. 1. It is presented as an appendix because it requires
some familiarity with the terminology of the subject. This chapter and parts of
Chap. 2, illustrating how small linear and integer programming problems may be
formulated, are kept on a very elementary level appropriate to the novice without
a background in mathematics. We provide a systematic overview of mathematical
solution techniques on both linear and mixed integer linear programming in Chap. 3,
while Chap. 12 contains details on nonlinear optimization techniques. Exercises are
included at the ends of chapters. These exercises can be tackled by hand or by using
the software, where appropriate, included with the book.
Types of linear programming problems and their modeling are discussed in
Chap. 4. Chapter 5 is a collection of case studies in the framework of linear
programming. Chapters 6 and 7 cover foundations of integer programming while
in Chap. 8 case studies are discussed. In Chap. 9 we consider how practitioners
may best set up and solve their optimization problems and in Chap. 10 we consider
examples of large cases. Then in Chap. 11 we consider other types of optimization,
e.g., sequential linear, quadratic and mixed integer stochastic programming. New to
the second edition are the chapters devoted to nonlinear optimization (Chap. 12),
deterministic global optimization (Chap. 13), polylithic modeling and solution
approaches (Chap. 14), cutting and packing (Chap. 15), impact of optimization and
especially parallel optimization (Chap. 16). Finally, Chap. 17 reflects the author’s
view on mathematical optimization and modeling, how it is and should be used, and
what is to be expected from it in the future.
Certain sections of chapters may be skipped by readers new to the area of
optimization. They are marked by  in the section heading. These sections should
be read through when required on a subsequent reading.
Instructors Manual
An Instructors Manual is available to bona fide lecturers. Please, contact the author
Josef Kallrath.
Acknowledgments
We would to like to thank colleagues and mentors who have advised and/or inspired
us over the years. These are too numerous to mention but we would like to single
out Beate Brockmüller for providing material on the telecommunication network
problem, Bob Daniel and Gunter Schnabel for the time they spent with us discussing
the manuscript in great detail, Tom Horak, Gernot Sauerborn, Anna Schreieck,
James Tebboth, Christian Timpe and Max Wagner for reading the manuscript (JK),
and inspiration from the work of Peter Hammer, Ailsa Land, Gautam Mitra, Paul
xx Preface

Williams and the late Martin Beale (JMW). We also offer our thanks to Dash
Associates for help, advice and cooperation over the inclusion of the XPRESS-MP
software and related discussion material.5 JK wants to express his special thanks
to Marilyn Dalton for her kind hospitality during numerous visits to Blisworth
House. Finally, JK wants to thank the clients involved in some of the real-world
cases. The interaction and communication with the clients, most of whom were
enthusiastic persons with deep knowledge of the business process they wanted
to improve using mathematical optimization, was an important and irreplaceable
resource which made the solution of challenging problems possible. Although, after
all the years these people might have forgotten6 the work and the exciting time we
spent together and might not be aware how they indirectly contributed to this book I
want to mention them: Peter Bassler, David DeSantis, Andy Hayter, Klaus Kindler,
Jan Orband, Gunter Schnabel, Hubert Smuda and Eckhardt Schubert.

Ludwigshafen, Germany Josef Kallrath

Loughborough, England John M. Wilson
1997

5 Dash holds a copyright on parts of the following Sects. 2.7, 10.2, 10.3 and 16.6.
6 Of course,I hope they have not forgotten! All the special adventures involved in the time working
on their problems could fill a book on its own.
Contents

1 Optimization: Using Models, Validating Models, Solutions,

Answers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 1
1.1 Introduction: Some Words on Optimization . . . .. . . . . . . . . . . . . . . . . . . . 1
1.2 The Scope of this Book . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 6
1.3 The Significance and Benefits of Models . . . . . . .. . . . . . . . . . . . . . . . . . . . 7
1.4 Mathematical Optimization .. . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 12
1.4.1 A Linear Optimization Example .. . . . . .. . . . . . . . . . . . . . . . . . . . 12
1.4.2 A Typical Linear Programming Problem .. . . . . . . . . . . . . . . . . 17
1.5 Using Modeling Systems and Software . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 19
1.5.1 Modeling Systems . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 19
1.5.2 A Brief History of Modeling Systems .. . . . . . . . . . . . . . . . . . . . 20
1.5.3 Modeling Specialists and Applications Experts .. . . . . . . . . . 21
1.5.4 Implementing a Model . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 22
1.5.5 Obtaining a Solution.. . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 23
1.5.6 Interpreting the Output . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 24
1.6 Benefiting from and Extending the Simple Model . . . . . . . . . . . . . . . . . 24
1.7 A Survey of Real-World Problems . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 26
1.8 Summary and Recommended Bibliography . . . .. . . . . . . . . . . . . . . . . . . . 28
1.9 Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 29
1.9.1 Notation, Symbols, and Abbreviations . . . . . . . . . . . . . . . . . . . . 29
1.9.2 A Brief History of Optimization  . . . .. . . . . . . . . . . . . . . . . . . . 30
2 From the Problem to its Mathematical Formulation . . . . . . . . . . . . . . . . . . . 33
2.1 How to Model and Formulate a Problem.. . . . . . .. . . . . . . . . . . . . . . . . . . . 33
2.2 Variables, Indices, Sets, and Domains. . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 35
2.2.1 Indices, Sets, and Domains.. . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 38
2.2.2 Summation .. . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 40
2.3 Constraints .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 43
2.3.1 Types of Constraints .. . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 44
2.3.2 Example .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 47
2.4 Objectives.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 48

xxi
xxii Contents

2.5 Building More Sophisticated Models . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 50

2.5.1 A Simple Production Planning Problem:
The Background . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 50
2.5.2 Developing the Model .. . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 51
2.6 Mixed Integer Programming .. . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 52
2.6.1 Example: A Farmer Buying Calves and Pigs. . . . . . . . . . . . . . 54
2.6.2 A Formal Definition of Mixed Integer Optimization . . . . . 56
2.6.3 Difficult Optimization Problems .. . . . . .. . . . . . . . . . . . . . . . . . . . 58
2.7 Interfaces: Spreadsheets and Databases . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 59
2.7.1 Example: A Blending Problem . . . . . . . .. . . . . . . . . . . . . . . . . . . . 60
2.7.2 Developing the Model .. . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 61
2.7.3 Re-running the Model with New Data .. . . . . . . . . . . . . . . . . . . . 63
2.8 Creating a Production System . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 63
2.9 Collecting Data .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 65
2.10 Modeling Logic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 66
2.11 Practical Solution of LP Models .. . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 66
2.11.1 Problem Size . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 67
2.11.2 Ease of Solution . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 67
2.12 Summary and Recommended Bibliography . . . .. . . . . . . . . . . . . . . . . . . . 69
2.13 Exercises .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 69
3 Mathematical Solution Techniques . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 71
3.1 Introduction .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 71
3.1.1 Standard Formulation of Linear Programming
Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 72
3.1.2 Slack and Surplus Variables .. . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 73
3.1.3 Underdetermined Linear Equations and Optimization .. . . 74
3.2 Linear Programming . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 75
3.2.1 Simplex Algorithm — A Brief Overview . . . . . . . . . . . . . . . . . 76
3.2.2 Solving the Boat Problem with the Simplex
Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 76
3.2.3 Interior-Point Methods — A Brief Overview . . . . . . . . . . . . . 82
3.2.4 LP as a Subroutine .. . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 84
3.3 Mixed Integer Linear Programming .. . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 85
3.3.1 Solving the Farmer’s Problem Using Branch
and Bound.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 85
3.3.2 Solving Mixed Integer Linear Programming Problems .. . 89
3.3.3 Cutting Planes and Branch and Cut (B&C) . . . . . . . . . . . . . . . 92
3.3.4 Branch and Price: Optimization with Column
Generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 93
3.4 Interpreting the Results . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 94
3.4.1 LP Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 94
3.4.2 Outputting Results and Report Writing.. . . . . . . . . . . . . . . . . . . 95
3.4.3 Dual Value (Shadow Price) . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 96
3.4.4 Reduced Costs . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 97
Contents xxiii

3.5 Duality  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 97
3.5.1 Constructing the Dual Problem . . . . . . . .. . . . . . . . . . . . . . . . . . . . 98
3.5.2 Interpreting the Dual Problem . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 100
3.5.3 Duality Gap and Complementarity . . . .. . . . . . . . . . . . . . . . . . . . 102
3.6 Summary and Recommended Bibliography . . . .. . . . . . . . . . . . . . . . . . . . 104
3.7 Exercises .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 104
3.8 Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 105
3.8.1 Linear Programming — A Detailed Description .. . . . . . . . . 105
3.8.2 Computing Initial Feasible LP Solutions .. . . . . . . . . . . . . . . . . 112
3.8.3 LP Problems with Upper Bounds . . . . . .. . . . . . . . . . . . . . . . . . . . 113
3.8.4 Dual Simplex Algorithm . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 118
3.8.5 Interior-Point Methods — A Detailed Description . . . . . . . 118
3.8.6 Branch and Bound with LP Relaxation .. . . . . . . . . . . . . . . . . . . 128
4 Problems Solvable Using Linear Programming .. . . .. . . . . . . . . . . . . . . . . . . . 131
4.1 Cutting Stock: Trimloss Problems .. . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 131
4.1.1 Example: A Trimloss Problem in the Paper Industry . . . . . 132
4.1.2 Example: An Integer Trimloss Problem .. . . . . . . . . . . . . . . . . . 134
4.2 The Food Mix Problem . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 135
4.3 Transportation and Assignment Problems . . . . . .. . . . . . . . . . . . . . . . . . . . 136
4.3.1 The Transportation Problem . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 136
4.3.2 The Transshipment Problem . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 139
4.3.3 The Assignment Problem . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 141
4.3.4 Transportation and Assignment Problems as
Subproblems . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 142
4.3.5 Matching Problems .. . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 142
4.4 Network Flow Problems . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 143
4.4.1 Illustrating a Network Flow Problem ... . . . . . . . . . . . . . . . . . . . 144
4.4.2 The Structure and Importance of Network Flow
Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 145
4.4.3 Case Study: A Telephone Betting Scheduling
Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 146
4.4.4 Other Applications of Network Modeling Technique . . . . 148
4.5 Unimodularity  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 148
4.6 Summary and Recommended Bibliography . . . .. . . . . . . . . . . . . . . . . . . . 149
4.7 Exercises .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 149
5 How Optimization Is Used in Practice: Case Studies in
Linear Programming .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 151
5.1 Optimizing the Production of a Chemical Reactor . . . . . . . . . . . . . . . . . 151
5.2 An Apparently Nonlinear Blending Problem . . .. . . . . . . . . . . . . . . . . . . . 153
5.2.1 Formulating the Direct Problem . . . . . . .. . . . . . . . . . . . . . . . . . . . 154
5.2.2 Formulating the Inverse Problem . . . . . .. . . . . . . . . . . . . . . . . . . . 156
5.2.3 Analyzing and Reformulating the Model.. . . . . . . . . . . . . . . . . 157
xxiv Contents

5.3 Data Envelopment Analysis (DEA) . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 159

5.3.1 Example Illustrating DEA. . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 160
5.3.2 Efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 163
5.3.3 Inefficiency.. . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 163
5.3.4 More Than One Input . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 164
5.3.5 Small Weights . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 164
5.3.6 Applications of DEA . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 165
5.3.7 A General Model for DEA . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 165
5.4 Vector Minimization and Goal Programming .. .. . . . . . . . . . . . . . . . . . . . 167
5.4.1 Solution Approaches for Multi-Criteria
Optimization Problems .. . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 167
5.4.2 A Case Study Involving Soft Constraints.. . . . . . . . . . . . . . . . . 170
5.4.3 A Case Study Exploiting a Hierarchy of Goals . . . . . . . . . . . 172
5.5 Limitations of Linear Programming .. . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 174
5.5.1 Single Objective . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 174
5.5.2 Assumption of Linearity . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 174
5.5.3 Satisfaction of Constraints . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 175
5.5.4 Structured Situations . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 176
5.5.5 Consistent and Available Data . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 176
5.6 Summary.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 177
5.7 Exercises .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 177
6 Modeling Structures Using Mixed Integer Programming . . . . . . . . . . . . . 179
6.1 Using Binary Variables to Model Logical Conditions .. . . . . . . . . . . . . 179
6.1.1 General Integer Variables and Logical Conditions .. . . . . . . 180
6.1.2 Transforming Logical into Arithmetical Expressions.. . . . 181
6.1.3 Logical Expressions with Two Arguments . . . . . . . . . . . . . . . . 182
6.1.4 Logical Expressions with More Than Two Arguments .. . 185
6.2 Logical Restrictions on Constraints . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 187
6.2.1 Bound Implications on Single Variables.. . . . . . . . . . . . . . . . . . 187
6.2.2 Bound Implications on Constraints .. . .. . . . . . . . . . . . . . . . . . . . 188
6.2.3 Disjunctive Sets of Implications . . . . . . .. . . . . . . . . . . . . . . . . . . . 191
6.3 Modeling Non-Zero Variables . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 193
6.4 Modeling Sets of All-Different Elements . . . . . . .. . . . . . . . . . . . . . . . . . . . 194
6.5 Modeling Absolute Value Terms  . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 195
6.6 Nonlinear Terms and Equivalent MILP Formulations .. . . . . . . . . . . . . 198
6.7 Modeling Products of Binary Variables . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 201
6.8 Special Ordered Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 202
6.8.1 Special Ordered Sets of Type 1 . . . . . . . .. . . . . . . . . . . . . . . . . . . . 203
6.8.2 Special Ordered Sets of Type 2 . . . . . . . .. . . . . . . . . . . . . . . . . . . . 205
6.8.3 Linked Ordered Sets . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 209
6.8.4 Families of Special Ordered Sets . . . . . .. . . . . . . . . . . . . . . . . . . . 212
6.9 Improving Formulations by Adding Logical Inequalities . . . . . . . . . . 213
6.10 Summary.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 214
6.11 Exercises .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 215
Contents xxv

7 Types of Mixed Integer Linear Programming Problems .. . . . . . . . . . . . . . 219

7.1 Knapsack and Related Problems . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 219
7.1.1 The Knapsack Problem .. . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 219
7.1.2 Case Study: Float Glass Manufacturing . . . . . . . . . . . . . . . . . . . 221
7.1.3 The Generalized Assignment Problem . . . . . . . . . . . . . . . . . . . . 222
7.1.4 The Multiple Binary Knapsack Problem . . . . . . . . . . . . . . . . . . 224
7.2 The Traveling Salesman Problem . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 224
7.2.1 Postman Problems .. . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 228
7.2.2 Vehicle Routing Problems.. . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 228
7.2.3 Case Study: Heating Oil Delivery . . . . .. . . . . . . . . . . . . . . . . . . . 229
7.3 Facility Location Problems . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 232
7.3.1 The Uncapacitated Facility Location Problem . . . . . . . . . . . . 232
7.3.2 The Capacitated Facility Location Problem .. . . . . . . . . . . . . . 234
7.4 Set Covering, Partitioning, and Packing.. . . . . . . .. . . . . . . . . . . . . . . . . . . . 234
7.4.1 The Set Covering Problem . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 234
7.4.2 The Set Partitioning Problem . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 237
7.4.3 The Set Packing Problem . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 238
7.4.4 Additional Applications .. . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 239
7.4.5 Case Study: Airline Management at Delta Air Lines . . . . . 239
7.5 Satisfiability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 241
7.6 Bin Packing .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 243
7.6.1 The Bin Packing Problem . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 243
7.6.2 The Capacitated Plant Location Problem.. . . . . . . . . . . . . . . . . 244
7.7 Clustering Problems .. . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 245
7.7.1 The Capacitated Clustering Problem . .. . . . . . . . . . . . . . . . . . . . 245
7.7.2 The p-Median Problem .. . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 246
7.8 Scheduling Problems .. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 247
7.8.1 Example A: Scheduling Machine Operations . . . . . . . . . . . . . 248
7.8.2 Example B: A Flowshop Problem . . . . .. . . . . . . . . . . . . . . . . . . . 250
7.8.3 Example C: Scheduling Involving Job Switching.. . . . . . . . 252
7.8.4 Case Study: Bus Crew Scheduling.. . . .. . . . . . . . . . . . . . . . . . . . 253
7.9 Summary and Recommended Bibliography . . . .. . . . . . . . . . . . . . . . . . . . 255
7.10 Exercises .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 256
8 Case Studies and Problem Formulations . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 259
8.1 A Depot Location Problem . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 259
8.2 Planning and Scheduling Across Time Periods .. . . . . . . . . . . . . . . . . . . . 261
8.2.1 Indices, Data, and Variables .. . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 262
8.2.2 Objective Function . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 263
8.2.3 Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 263
8.3 Distribution Planning for a Brewery . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 265
8.3.1 Dimensions, Indices, Data, and Variables . . . . . . . . . . . . . . . . . 266
8.3.2 Objective Function . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 267
8.3.3 Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 268
8.3.4 Running the Model . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 270
xxvi Contents

8.4 Financial Modeling.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 270

8.4.1 Optimal Purchasing Strategies . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 270
8.4.2 A Yield Management Example . . . . . . . .. . . . . . . . . . . . . . . . . . . . 275
8.5 Post-Optimal Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 277
8.5.1 Getting Around Infeasibility . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 278
8.5.2 Basic Concept of Ranging.. . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 280
8.5.3 Parametric Programming .. . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 282
8.5.4 Sensitivity Analysis in MILP Problems . . . . . . . . . . . . . . . . . . . 283
8.6 Summary and Recommended Bibliography . . . .. . . . . . . . . . . . . . . . . . . . 284
9 User Control of the Optimization Process and Improving
Efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 285
9.1 Preprocessing .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 285
9.1.1 Presolve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 286
9.1.2 Disaggregation of Constraints . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 290
9.1.3 Coefficient Reduction . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 291
9.1.4 Clique Generation . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 293
9.1.5 Cover Constraints .. . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 294
9.2 Efficient LP Solving .. . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 296
9.2.1 Warm Starts . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 296
9.2.2 Scaling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 297
9.3 Good Modeling Practice . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 298
9.4 Choice of Branch in Integer Programming.. . . . .. . . . . . . . . . . . . . . . . . . . 302
9.4.1 Control of the Objective Function Cut-Off .. . . . . . . . . . . . . . . 302
9.4.2 Branching Control . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 303
9.4.3 Priorities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 304
9.4.4 Branching on Special Ordered Sets . . . .. . . . . . . . . . . . . . . . . . . . 304
9.4.5 Branching on Semi-Continuous and Partial
Integer Variables .. . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 306
9.5 Symmetry and Optimality . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 307
9.6 Summary.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 308
9.7 Exercises .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 309
10 How Optimization Is Used in Practice: Case Studies in
Integer Programming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 311
10.1 What Can be Learned from Real-World Problems? . . . . . . . . . . . . . . . . 311
10.2 Three Instructive Solved Real-World Problems . . . . . . . . . . . . . . . . . . . . 312
10.2.1 Contract Allocation .. . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 312
10.2.2 Metal Ingot Production .. . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 314
10.2.3 Project Planning . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 315
10.2.4 Conclusions .. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 317
10.3 A Case Study in Production Scheduling . . . . . . . .. . . . . . . . . . . . . . . . . . . . 318
10.4 Optimal Worldwide Production Plans  . . . . . . . .. . . . . . . . . . . . . . . . . . . . 323
10.4.1 Brief Description of the Problem . . . . . .. . . . . . . . . . . . . . . . . . . . 323
10.4.2 Mathematical Formulation of the Model . . . . . . . . . . . . . . . . . . 325
10.4.3 Remarks on the Model Formulation . . .. . . . . . . . . . . . . . . . . . . . 336
Contents xxvii

10.4.4 Model Performance.. . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 340

10.4.5 Reformulations of the Model . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 341
10.4.6 What Can be Learned from This Case Study? . . . . . . . . . . . . 346
10.5 A Complex Scheduling Problem  . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 346
10.5.1 Description of the Problem.. . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 347
10.5.2 Structuring the Problem . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 347
10.5.3 Mathematical Formulation of the Problem .. . . . . . . . . . . . . . . 351
10.5.4 Time-Indexed Formulations .. . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 353
10.5.5 Numerical Experiments . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 357
10.5.6 What Can be Learned from This Case Study? . . . . . . . . . . . . 363
10.6 Telecommunication Service Network  . . . . . . . .. . . . . . . . . . . . . . . . . . . . 364
10.6.1 Description of the Model .. . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 365
10.6.2 Mathematical Model Formulation . . . . .. . . . . . . . . . . . . . . . . . . . 367
10.6.3 Analysis and Reformulations of the Models . . . . . . . . . . . . . . 377
10.7 Synchronization of Batch and Continuous Processes . . . . . . . . . . . . . . 382
10.7.1 Time Sequencing Constraints . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 385
10.7.2 Reactor Availability Constraints . . . . . . .. . . . . . . . . . . . . . . . . . . . 385
10.7.3 Exploiting Free Reactor Time — Delaying
Campaign Starts . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 387
10.7.4 Restricting the Latest Time a Reactor Is Available . . . . . . . 388
10.8 Summary and Recommended Bibliography . . . .. . . . . . . . . . . . . . . . . . . . 388
10.9 Exercises .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 389
11 Beyond LP and MILP Problems  . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 391
11.1 Fractional Programming * . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 392
11.2 Recursion or Successive Linear Programming... . . . . . . . . . . . . . . . . . . . 393
11.2.1 An Example .. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 394
11.2.2 The Pooling Problem .. . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 396
11.3 Optimization Under Uncertainty* . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 401
11.3.1 Motivation and Overview . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 401
11.3.2 Stochastic Programming . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 405
11.3.3 Recommended Literature . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 417
11.4 Quadratic Programming.. . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 418
11.5 Summary and Recommended Bibliography . . . .. . . . . . . . . . . . . . . . . . . . 422
11.6 Exercises .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 422
12 Mathematical Solution Techniques — The Nonlinear World . . . . . . . . . 423
12.1 Unconstrained Optimization .. . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 423
12.2 Constrained Optimization — Foundations and Theorems . . . . . . . . . 427
12.3 Reduced Gradient Methods .. . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 430
12.4 Sequential Quadratic Programming . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 434
12.5 Interior-Point Methods .. . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 435
12.6 Mixed Integer Nonlinear Programming . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 436
12.6.1 Definition of an MINLP Problem .. . . . .. . . . . . . . . . . . . . . . . . . . 436
12.6.2 Some General Comments on MINLP .. . . . . . . . . . . . . . . . . . . . 437
12.6.3 Deterministic Methods for Solving MINLP Problems . . . 439
xxviii Contents

12.6.4 Algorithms and Software for Solving

Non-convex MINLP Problems .. . . . . . . .. . . . . . . . . . . . . . . . . . . . 440
12.7 Global Optimization — Mathematical Background .. . . . . . . . . . . . . . . 441
12.8 Summary and Recommended Bibliography . . . .. . . . . . . . . . . . . . . . . . . . 446
12.9 Exercises .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 446
13 Global Optimization in Practice . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 447
13.1 Global Optimization Applied to Real-World Problems . . . . . . . . . . . . 448
13.2 A Trimloss Problem in Paper Industry . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 449
13.3 Cutting and Packing Involving Convex Objects . . . . . . . . . . . . . . . . . . . . 452
13.3.1 Modeling the Cutting Constraints . . . . .. . . . . . . . . . . . . . . . . . . . 453
13.3.2 Problem Structure and Symmetry.. . . . .. . . . . . . . . . . . . . . . . . . . 457
13.3.3 Some Results . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 459
13.4 Summary and Recommended Bibliography . . . .. . . . . . . . . . . . . . . . . . . . 459
13.5 Exercises .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 459
14 Polylithic Modeling and Solution Approaches . . . . . .. . . . . . . . . . . . . . . . . . . . 461
14.1 Polylithic Modeling and Solution Approaches (PMSAs) . . . . . . . . . . 461
14.1.1 Idea and Foundations of Polylithic Solution
Approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 462
14.1.2 Problem-Specific Preprocessing . . . . . . .. . . . . . . . . . . . . . . . . . . . 464
14.1.3 Mathematical Algorithms . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 467
14.1.4 Primal Heuristics . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 480
14.1.5 Proving Optimality Using PMSAs. . . . .. . . . . . . . . . . . . . . . . . . . 484
14.2 PMSAs Applied to Real-World Problems.. . . . . .. . . . . . . . . . . . . . . . . . . . 485
14.2.1 Cutting Stock and Packing . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 485
14.2.2 Evolutionary Approach .. . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 487
14.2.3 Optimal Breakpoints . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 490
14.3 Summary and Recommended Bibliography . . . .. . . . . . . . . . . . . . . . . . . . 493
14.4 Exercises .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 493
15 Cutting and Packing Beyond and Within Mathematical
Programming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 495
15.1 Introduction .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 495
15.2 Phi-objects .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 498
15.2.1 Phi-objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 498
15.2.2 Primary and Composed Phi-objects . . .. . . . . . . . . . . . . . . . . . . . 499
15.2.3 Geometric Parameters of Phi-objects ... . . . . . . . . . . . . . . . . . . . 500
15.2.4 Position Parameters of Phi-objects . . . .. . . . . . . . . . . . . . . . . . . . 502
15.2.5 Interaction of Phi-objects.. . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 503
15.3 Phi-functions: Relating Phi-objects . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 503
15.3.1 Construction of Phi-functions for Various Situations .. . . . 504
15.3.2 Properties of Phi-functions .. . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 513
15.4 Mathematical Optimization Model .. . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 514
15.4.1 Objective Function . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 515
15.4.2 Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 515
Contents xxix

15.4.3 Simplifying Distance Constraints .. . . . .. . . . . . . . . . . . . . . . . . . . 516

15.4.4 General Remarks . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 517
15.5 Solving the Optimization Problem . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 518
15.6 Numerical Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 519
15.6.1 Arranging Two Triangles .. . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 519
15.6.2 Arranging Two Irregular Objects . . . . . .. . . . . . . . . . . . . . . . . . . . 522
15.7 Conclusions .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 523
15.8 Summary and Recommended Bibliography . . . .. . . . . . . . . . . . . . . . . . . . 524
15.9 Exercises .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 526
16 The Impact and Implications of Optimization . . . . . .. . . . . . . . . . . . . . . . . . . . 527
16.1 Benefits of Mathematical Programming to Users. . . . . . . . . . . . . . . . . . . 527
16.2 Implementing and Validating Solutions . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 528
16.3 Communicating with Management .. . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 529
16.4 Keeping a Model Alive . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 529
16.5 Mathematical Optimization in Small and Medium Size
Business . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 530
16.6 Online Optimization by Exploiting Parallelism? . . . . . . . . . . . . . . . . . . . 531
16.6.1 Parallel Optimization: Status and Perspectives
in 1997 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 531
16.6.2 Parallel Optimization: Status and Perspectives
in 2020 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 536
16.7 Summary.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 541
17 Concluding Remarks and Outlook . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 543
17.1 Learnings from the Examples and Models . . . . . .. . . . . . . . . . . . . . . . . . . . 543
17.2 Future Developments.. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 544
17.2.1 Pushing the Limits . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 544
17.2.2 Cloud Computing .. . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 546
17.2.3 The Importance of Modeling.. . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 546
17.2.4 Tools Around Optimization . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 548
17.2.5 Visualization of Input Data and Output Results . . . . . . . . . . . 550
17.2.6 Increasing Problem Size and Complexity . . . . . . . . . . . . . . . . . 555
17.2.7 The Future of Planning and Scheduling . . . . . . . . . . . . . . . . . . . 558
17.2.8 Simultaneous Operational Planning & Design
and Strategic Optimization .. . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 559
17.3 Mathematical Optimization for a Better World * . . . . . . . . . . . . . . . . . . . 561

A Software Related Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 567

A.1 Accessing Data from Algebraic Modeling Systems . . . . . . . . . . . . . . . . 567
A.2 List of Case Studies and Model Files . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 568
B Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 571
C Mathematical Foundations: Linear Algebra and Calculus . . . . . . . . . . . . 577
C.1 Sets and Quantifiers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 577
C.2 Absolute Value and Triangle Inequality .. . . . . . . .. . . . . . . . . . . . . . . . . . . . 581
xxx Contents

C.3 Vectors in Rn and Matrices in M (m × n, R) . .. . . . . . . . . . . . . . . . . . . . 582

C.4 Vector Spaces, Bases, Linear Independence,
and Generating Systems . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 585
C.5 Rank of Matrices, Determinant, and Criteria for Invertible
Matrices .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 589
C.6 Systems of Linear Equations . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 590
C.7 Some Facts on Calculus .. . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 592
References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 595
Index . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 621
About the Author

Prof. Dr. Josef Kallrath has studied mathematics, physics, and astronomy in
Bonn, where he received a doctorate in 1989 for his dissertation in astrophysics
on the dynamics of colliding double stellar winds. He has worked in companies
(BASF SE, Ludwigshafen; 1989–2019), has been freelancing since 1998 as a
scientific consultant and lecturer, and solving practical problems in industry with
scientific computing and operations research techniques, most of it being modeling
and solving mathematical optimization problems. His work focuses on mathemati-
cal optimization to support decisions in process, paper, metal, and energy industry,
as well as transport infrastructure and the modeling of physical systems. He has
taught at the University of Heidelberg (1991–2001) and at the University of Florida
in Gainesville/USA (since 1997). Since 2002, Prof. Kallrath has been heading the
Mathematical Optimization Practice Group of the German Operations Research
society (GOR).

xxxi
List of Figures

Fig. 1.1 Transforming a real-world decision problem . .. . . . . . . . . . . . . . . . . . . . 3

Fig. 1.2 A powerful modeler . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 11
Fig. 1.3 Boat problem illustrated . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 13
Fig. 1.4 Graphical solution of an LP problem in two variables . . . . . . . . . . . . 16
Fig. 2.1 A modeler thinking about a model of a real-world problem . . . . . . 34
Fig. 2.2 Integer model coded in GAMS. . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 55
Fig. 2.3 Simple spreadsheet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 59
Fig. 3.1 Simplex algorithm versus interior-point methods.. . . . . . . . . . . . . . . . . 83
Fig. 3.2 LP relaxation and the first two subproblems of a B&B tree .. . . . . . 87
Fig. 3.3 B&B tree for the cows and pigs problem . . . . . .. . . . . . . . . . . . . . . . . . . . 88
Fig. 3.4 LP relaxation, convex hull, and a B&B tree. . . .. . . . . . . . . . . . . . . . . . . . 91
Fig. 3.5 Illustrating the idea of Branch and Cut . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 92
Fig. 3.6 Feasible region of an LP problem . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 107
Fig. 3.7 The revised Simplex algorithm . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 108
Fig. 3.8 Logarithmic penalty term . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 120
Fig. 3.9 The Branch-and-Bound algorithm .. . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 129
Fig. 3.10 Branch-and-Bound tree: Updating the value
of the LP relaxation . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 130
Fig. 4.1 Geometry of a trimloss problem .. . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 132
Fig. 4.2 Routes on transportation network (left) and
transshipment network (right) . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 140
Fig. 4.3 Flows between nodes of a network . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 145
Fig. 4.4 A time expanded network .. . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 147
Fig. 6.1 Smallest Big-M coefficient . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 200
Fig. 6.2 Using SOS1 to select capacity size . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 204
Fig. 6.3 Using SOS2 to model a nonlinear function . . . .. . . . . . . . . . . . . . . . . . . . 206
Fig. 7.1 Knapsack problem illustrated using a burglar example.. . . . . . . . . . . 220
Fig. 7.2 Traveling salesman problem with four cities . . .. . . . . . . . . . . . . . . . . . . . 227
Fig. 7.3 Vehicle routing and dispatching . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 229
Fig. 7.4 A set covering problem . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 235
Fig. 7.5 Gantt chart showing the schedule.. . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 250

xxxiii
xxxiv List of Figures

Fig. 8.1 Costs versus number of items . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 271

Fig. 8.2 Sensitivity analysis: objective versus optimal value . . . . . . . . . . . . . . . 280
Fig. 8.3 Sensitivity analysis: optimal value versus unit profit . . . . . . . . . . . . . . 281
Fig. 8.4 Sensitivity analysis: slope of objective function . . . . . . . . . . . . . . . . . . . 281
Fig. 10.1 Production network with three sites . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 324
Fig. 10.2 Illustration of a set-up change . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 330
Fig. 10.3 Production plan .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 337
Fig. 10.4 Precedence relations between jobs . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 360
Fig. 10.5 Gantt chart and personnel occupation diagram . . . . . . . . . . . . . . . . . . . . 362
Fig. 10.6 Cost of bandwidth for POP-to-POP private lines . . . . . . . . . . . . . . . . . . 366
Fig. 10.7 A possible routing via hub sites for a demand Dij . . . . . . . . . . . . . . . . 368
Fig. 10.8 Combination of batch and continuous production process . . . . . . . . 383
Fig. 10.9 Combination of batch and continuous production process . . . . . . . . 385
Fig. 11.1 The pooling problem and a process unit fed by a pool . . . . . . . . . . . . 398
Fig. 11.2 Newsvendor problem .. . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 406
Fig. 12.1 NLP-solution-structure .. . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 434
Fig. 12.2 Convex and non-convex sets and functions . . . .. . . . . . . . . . . . . . . . . . . . 438
Fig. 12.3 Non-convex function and a convex underestimator.. . . . . . . . . . . . . . . 442
Fig. 13.1 Representation of polygons . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 455
Fig. 13.2 Lines separating polygons . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 456
Fig. 14.1 Polylithic versus monolithic modeling . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 463
Fig. 15.1 Strip packing polygons into a rectangle . . . . . . . .. . . . . . . . . . . . . . . . . . . . 496
Fig. 15.2 3D packing examples . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 497
Fig. 15.3 Invalid phi-objects .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 499
Fig. 15.4 Examples of composed phi-objects in 2D . . . . . .. . . . . . . . . . . . . . . . . . . . 500
Fig. 15.5 Examples of composed phi-objects in 3D . . . . . .. . . . . . . . . . . . . . . . . . . . 501
Fig. 15.6 Phi-functions illustrated .. . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 504
Fig. 15.7 Arrangements of two objects A and B . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 504
Fig. 15.8 Phi-function for two circles . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 505
Fig. 15.9 Phi-function for two rectangles .. . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 506
Fig. 15.10 Phi-function for a rectangle and a circle . . . . . . .. . . . . . . . . . . . . . . . . . . . 510
Fig. 15.11 Composed object and a circle .. . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 511
Fig. 15.12 Simplifying distance constraints for A and B . .. . . . . . . . . . . . . . . . . . . . 517
Fig. 15.13 Two triangles attached to each other and its convex hull .. . . . . . . . . 523
Fig. 15.14 Arrangement of two objects A and B . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 524
Fig. 15.15 Convex hull for two objects A and B . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 525
Fig. 15.16 An approximation of the m-polygonal convex hull .. . . . . . . . . . . . . . . 525
Fig. 15.17 Two phi-objects A and B . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 526
Fig. 16.1 Speed-up achieved with eight slaves . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 532
Fig. 17.1 A typical side-by-side scenario comparison . . .. . . . . . . . . . . . . . . . . . . . 551
Fig. 17.2 A horizontal bar chart showing supplier profiles . . . . . . . . . . . . . . . . . . 551
Fig. 17.3 Churn rates. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 552
Fig. 17.4 A production plan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 552
Fig. 17.5 A routing map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 553
Chapter 1
Optimization: Using Models, Validating
Models, Solutions, Answers

This chapter guides the reader into the field of mathematical optimization, dis-
tinguishes optimization from simulation, and introduces the key objects used in
optimization models. It sketches how simple linear programming (LP) problems
may be solved graphically. The chapter contains a survey of real-world problems
and finally a review of the history of optimization. Those readers who want to use
the optimization software attached to this book will find sufficient information in
this chapter to do so. Readers less interested in implementation issues or already
familiar with the use of commercial software may of course skip those parts.

1.1 Introduction: Some Words on Optimization

The title of this book contains the term mathematical programming. So let us
first consider the question: “What is mathematical programming1 or optimization
and what are optimization problems?” Optimization problems arise in almost all
branches of industry or society, e.g., in product and process design, allocation of
scarce resources, production, logistics, scheduling, strategic planning, or traffic
control. They can have different time scales: real-time (just now), operational
(short-term), tactical (mid-term), and strategic (long-term) planning. Unfortunately,
the word optimization, in nontechnical language, is often used in the sense of
improving while the original meaning of the word is related to finding the best.
In an optimization problem (OP), one tries to minimize or maximize an important

1 The term mathematical programming has its historical roots in the first problems solved by linear
programming in the Second World War. In these days, programming had more the meaning of
planning. Nowadays, mathematical optimization seems more appropriate. In this book we will use
the terms mathematical programming and mathematical optimization synonymously.

© Springer Nature Switzerland AG 2021 1

J. Kallrath, Business Optimization Using Mathematical Programming,
International Series in Operations Research & Management Science 307,
https://doi.org/10.1007/978-3-030-73237-0_1
2 1 Optimization: Using Models, Validating Models, Solutions, Answers

characteristic of a process within the problem such as revenue, elapsed time or cost,
by an appropriate choice of decisions which influence this process, e.g., budget
allocations, manpower deployed, or quantities to be manufactured. Such decisions
can be controlled, but are influenced and ultimately limited by a set of constraints,
linked, for example, to physical limits, e.g., limits on the budget or manpower
available. The results of the optimization process will suggest decisions that could
be taken and, if appropriate, the levels of resources that should be utilized.
A traditional way to develop answers to decision problems is to propose a
number of choices for the values of the decisions, using trial and error methods.
The processes under investigation are then evaluated or simulated under these
various choices, and the results are compared. This concept of simulation [cf. Pidd
(1992,[439])] may involve varying certain input data or building scenarios from
investigating the behavior of a model under various sets of conditions. Simulation
can be carried out using sophisticated software systems, e.g., SIMSCRIPT2, and
is also prevalent as the “what if” facility provided by more general purpose
software such as spreadsheets (e.g., LOTUS3 , EXCEL4). Analysts in charge of
simulating these decisions have developed intuition and simple rules to select
appropriate conditions, and simulation software exists to perform the evaluation
of their performance. The “traditional” techniques may lead to useful and usable
results, but there is no guarantee that the best solution (the optimal solution) or
even a solution close to the optimal solution has been found. This is especially
troublesome for large or complex problems, or those which require decisions with
high financial impact. Below we will discuss the difference between simulation and
mathematical optimization. This difference can be more easily discussed when we
have answered the following question:
What do we need when we want to solve a real-world problem by mathematical
optimization?
The first thing we need is to represent the real-world problem by a mathematical
model. A mathematical model of a system is a set of mathematical relationships
(e.g., equalities, inequalities, logical conditions) which represent an abstraction of
the real-world problem under consideration [see Fig. 1.1]. Usually, a mathematical
model in optimization consists of four key objects:
• data or parameters,5
• variables (continuous, semi-continuous, binary, integer),6
• constraints (equalities, inequalities),7 and
• objective function.

2 SIMSCRIPT is a trademark of CACI products.

3 LOTUS is a registered trademark of Lotus Development Corp.
4 EXCEL is a trademark of Microsoft Corp.
5 Engineers often prefer the term parameters to refer to fixed data.
6 Synonyms are decision variables, unknowns, or columns, depending on community.
7 Sometimes also called restrictions.
1.1 Introduction: Some Words on Optimization 3

Fig. 1.1 Transforming a real-world decision problem to a mathematical optimization problem.

The input data and parameters are hidden in the step from the customer’s problem through the
green glasses to the mathematical structure (objective function and constraints)

The data or parameters define an instance of a problem. They may represent costs
or demands, fixed operating conditions of a reactor, capacities of plants, and so on.
The variables represent the degrees of freedom, i.e., our decisions: how much of a
certain product is to be produced, whether a depot is closed or not, or how much
material should be stored in the inventory for later use. The constraints can be a
wide range of mathematical relationships: algebraic, logic, differential, or integral;
in this book, only algebraic and logic constraints are considered. They may represent
4 1 Optimization: Using Models, Validating Models, Solutions, Answers

mass balances, quality relations, capacity limits, and so on. The objective function,
finally, expresses our goals or objectives in a mathematical form: minimize costs,
maximize utilization rate, minimize waste, and so on. When building mathematical
optimization models they usually lead to structured problems such as:
• linear programming (LP) problems,
• mixed integer linear programming (MILP) problems,
• nonlinear programming (NLP) problems, and
• mixed integer nonlinear programming (MINLP) problems.
Besides building a model and classifying the problem one needs a solver, i.e., a
piece of software which has a set of algorithms capable of solving the problems
listed above.
Above, the terms integer or mixed integer occurred. What is mixed integer or
discrete8 optimization? Classical optimization theory (calculus, variational calculus,
optimal control) treats those cases in which the decision variables can be changed
continuously, e.g., the temperature in a chemical reactor or the amount of a product
to be produced. On the other hand, mixed integer, combinatorial, or discrete
optimization addresses degrees of freedom which are limited to integer values,
for example, counts (numbers of containers, ships), decisions (yes–no), or logical
relations (if product A is produced, then product B also needs to be produced). This
discipline, formerly only a marginal discipline within mathematical optimization, is
becoming more and more important (Grötschel, 1992,[243]) as it extends the power
of mathematical optimization to situations which are relevant to practical decision
making where business success is at stake.
Above, simulation is also mentioned as a mean to improve a company’s business.
What is the difference between simulation and mathematical optimization? In
contrast to simulation, optimization methods search directly for an optimal solution
that fulfills all restrictions and relations which are relevant for the real-world
problem, and prove that a certain solution is optimal (proof of optimality). In
simulation, the effects on a model of selected solutions will be examined, but
there will be no guarantee that any of the solutions under consideration is optimal.
Besides optimality, notice the other very substantial difference between simulation
and optimization: the existence of constraints. While in simulation somebody has to
make sure that only those combinations of variables are evaluated which represent
“appropriate conditions,” in optimization models it has to be specified a priori
what makes a feasible solution. Once the problem has been solved to optimality
within the specified limitations or constraints, then it is known that no other set
of decisions or values for quantities suggested by these decisions can provide any
other solution which will give a “better” value to the characteristic that has been
optimized. “Better” here will mean larger if the characteristic is to be maximized
(e.g., discounted revenue being maximized) and smaller if the characteristic is to

8 The terms mixed integer optimization and discrete optimization are used synonymously in this

book. Sometimes, in the literature, the term combinatorial optimization is also used.
1.1 Introduction: Some Words on Optimization 5

be minimized (e.g., total cost being minimized). The optimization process is a

prescriptive one which tells its user what decisions should be taken and what levels
of resources should be used. These levels might eventually not be accepted at face
value in their entirety, for a variety of reasons, but they provide a yardstick against
which all alternatively proposed decisions and levels of recommended resource can
be compared. This is not to suggest that an organization will choose to implement
a policy that is not the best for any frivolous reason, but rather it may choose to
implement another policy which is very near to being optimal but which has certain
other appealing features about it, e.g., it may only involve disturbing 10% of existing
practice, whereas the optimal solution would disturb 90% of existing practice.
Such features may be intangible and difficult to include directly in a representation
of the problem. Based on the two fundamental differences (optimality proof,
incorporation of constraints) between simulation and optimization we conclude:
by using mathematical optimization it becomes possible to control and adjust
complex systems even when they are difficult for a human being to grasp. Therefore,
optimization techniques allow a better exploitation of the advantages inherent in
complex systems.
What commercial potential is in mathematical optimization? To give some idea
of the scope of mathematical programming in helping organizations we cite four
recent examples of benefits achieved. First, at Delta Airlines it is reported in a paper
by Subramanian et al. (1994,[536]) that use of an optimization model is expected to
save the company (US)$300 million over a 3-year period. Similar models in use at
other airlines around the world are likely to achieve comparable savings. Second,
a comprehensive mathematical programming model used by Digital Equipment
Corporation (DEC) is described by Arntzen et al. (1995,[33]). The model has helped
the company to save (US)$100 million. Third, in the UK considerable savings in the
electricity industry are attributable to the use of mathematical programming. Fourth,
in some blending problems in the 1990s BASF-AG saved several hundred thousand
DM/year. The potential saving in complex production networks in chemical industry
is of the order of millions of DM. Pay-back time of projects is usually less than
six months. Blackburn et al. (2014,[83]) have highlighted optimization-centered
applications of Operations Research (OR) at BASF in the area of supply chain
management (SCM) through the use of selected examples with annual cost savings
ranging between several hundred thousand and several millions e. The role of
mathematical optimization and its impact in industrial practice in chemical and
petroleum industry is well described and reviewed by Grossmann & Harjunkoski
(2019,[241]); they report about 30% energy savings in operations.
The reader can find more about the financial impact of optimization (and Oper-
ations Research in the broader picture) by searching for the Franz Edelman Award
on www.informs.org, the webpage of the Institute for Operations Research and
the Management Sciences (INFORMS). First awarded in 1972 as the INFORMS
practice prize TIMS, in 1982 the prize has been renamed in Franz Edelman Award
and is annually presented in honor of Franz Edelman, who founded the Operations
Research division within Radio Corporation of America (RCA), one of the first
corporations to embed operations research as a business imperative. Since its
6 1 Optimization: Using Models, Validating Models, Solutions, Answers

inception, nearly $250 billion in benefits have been tabulated among Franz Edelman
Award finalist teams. Following the competition, INFORMS publishes papers by the
Edelman finalists in the January issue of the INFORMS journal Interfaces.

1.2 The Scope of this Book

In the first edition of this book published in 1997, this section started with the
following paragraph:
This book will consider the two types of mathematical programming problems
introduced, namely the linear programming problem and its extension, the “mixed
integer linear programming problem,” and certain variations on these problems. We
will consider how to set up these types of problem, how they are used, how they are
solved, and how to use software to both model and solve practical situations that
can be tackled using these two types of problems. Some other types of problems,
principally those most closely related to these two, will be considered, but in
very much less detail. That is not to say that problems which can be handled by
techniques such as dynamic programming [cf. Bellman (1957,[59])] or nonlinear
programming [cf. Chap. 12 or Bazaraa et al. (1993,[53])] are not important, but
rather that most commercially available software has concentrated on the linear
programming problem or the mixed integer linear programming problem because
they are the most useful in applications and the most straightforward to solve for
problems of large size. Thus we will tend to “follow the crowd,” but this will have the
advantage that we can benefit from the considerable effort that has gone into solving
such problems over the last 30–40 years, and, in particular, from the developments
in software to solve such problems.
The chapters of the book introduce the concepts necessary for an understanding
of mathematical optimization problems. However, these topics are not introduced in
a purely abstract form, but rather they will be related to using these techniques and
approaches together with a software system, as would be done in practice.
In this second edition, the scope has been significantly extended. It covers
nonlinear optimization as well as mixed integer nonlinear optimization. Polylithic
modeling and solution approaches have been added and optimization under uncer-
tainty has been treated in more detail. Readers interested in cutting and packing will
find more specific material on this interesting topic with relations to computational
geometry. Unless readers want to start with the nonlinear world right away, the easier
path to follow is to start with linear programming, moving to mixed integer linear
programming, and finally entering the world of nonlinear optimizations problems.
The software, i.e., the model files which come with this book, can be solved
with the algebraic modeling systems (AMSs, hereafter) and the algebraic modeling
languages (AMLs) contained therein: GAMS,9 FICO Xpress Mosel, (short

9 www.gams.com; see also Brooke et al. (1988,[100]) and Bussieck & Meeraus (2003,[106]).
1.3 The Significance and Benefits of Models 7

form: Xpress Mosel or just Mosel),10 or SAS/OR;11 it should not be too

difficult if the reader wants to use AMPL,12 instead. Their modeling environment
provides the user with an easy to use yet powerful language for implementing the
problem. It also enables the user to gather the appropriate data from text files and
a range of popular spreadsheet systems and database systems. The optimization
software provides a solver that uses the problem description produced by the
modeler and then solves it to provide the user with the optimal solution. AMLs
and AMSs have been available and updated since the early 1980s and are innovative
optimization systems widely used worldwide; for a timeline and a historical review
see Kallrath (2004,[306]).
Coming with this book are coded models in GAMS, Fico Xpress Mosel,
and SAS/OR hosted at www.springer.com. Some of these models are stand-alone
models, while others will progressively introduce new features so that by the later
stages of the book the reader should be able to formulate, model, and solve business
problems of commercial size.
Apart from the classical AMLs we mention the recent Python based devel-
opment Pyomo13 as well as JuMP (a domain-specific modeling language for
mathematical optimization based on Julia).14 Matlab now also offers modeling
based mathematical programming. Students might like to see optimization problems
solved with software such as R or maybe Python as they are indeed very popular.
We do not include R, Python, Pyomo, or JuMP code in this book, as we feel that
especially the Python based approaches lack the stability required by industrial
needs. That should, however, not discourage students from solving the exercises by
using the tools.
Exercises will be presented at the end of most chapters. These will vary in
difficulty and it is intended that these be worked through by the reader either by
hand or by using an AMS — or by using R, Python, Pyomo, or JuMP if preferred.
Complete solutions to exercises will be given with full instructions relating to the
use of software on such exercises, where appropriate.

1.3 The Significance and Benefits of Models

The term modeling presupposes the term model, which first derives from the Latin
word modelus (scale [diminutive of mode, measure] and then from the word modelo

10 www.fico.com/xpress; see also Heipcke (2002,[259]), Guéret et al. (2002,[245]), Colombani &

Heipcke (2002,[128]), Ciriani et al. (2003,[124]) and Colombani et al. (2004,[130]).

11 http://go.documentation.sas.com; see also SAS/OR User’s Guide (2018,[483]).
12 www.ampl.com; see also Fourer et al. (1987,[198]) or Fourer et al. (1990,[199]) and Fourer et al.

(2003,[200]).
13 www.pyomo.org.
14 www.juliaopt.org.
8 1 Optimization: Using Models, Validating Models, Solutions, Answers

formed in the sixteenth century and which is used today in everyday life and in
scientific language with various meanings, but mostly in the sense of a simplified,
abstracted, or structured representation of an interesting part of reality. However,
the idea and thus the terminology are much older. Even with Pythagoras around 600
BC, classical geometry consciously distinguished between wheel and circle; field
and rectangle. Around 1100 A.D., a wooden model was made as an abstraction of
the later Speyer Cathedral. Astrolabes and celestial spheres were used as manifested
models of the motion sequences in the firmament to calculate the rising and setting
of the sun, moon, and stars. Until the nineteenth century, mechanical models were
understood as images of reality; they attributed all processes to the movement of
the smallest particles, which followed the principles of classical mechanics. Further
attempts were made to reduce all physical and other processes to the mechanistic
model. Nowadays in physics and mathematical sciences one speaks of models, if
1. for reasons of simplification, the investigation is limited to certain phenomena
which, in a given context, are regarded as important (example: in the movement
of the planets, the spatial extent of these bodies is initially neglected),
2. for reasons of didactic illustration, one gives a picture based on classical ideas for
phenomena which are not accessible in a descriptive way (example: the planetary
model to illustrate the conditions in atomic nuclei), or
3. one studies situations in one range in analogy to known situations in another
range (analogy models).
Let us generally define a model as an appropriate abstract representation of a real
system [cf. Williams (1993,[580]) or Bossel (1994,[93])]. Naturally, a mathematical
model of a process or a problem is formulated with the help of mathematical objects
(variables, terms, relations). A (mathematical) model represents a real problem in
the language of mathematics, i.e., using mathematical symbols, variables, equations,
inequalities, and other relations.
A very important aspect, which is connected with modeling and precedes it,
is the model purpose. It results directly in connection with the problem and very
substantially influences the process of modeling. In science (e.g., Physics, Astron-
omy, Chemistry, and Biology) models are used to gain a deeper understanding
of processes occurring in nature (an epistemological argument), e.g., to explain
the movements of planets. The comparison of measurements and observations
with the predictions of a model is used to determine the appropriateness and
quality of the model. Sir Karl Popper (1980,[444]) uses the expressions falsification
and verification in his famous book Logic of Scientific Discovery as tasks when
accomplished deciding on whether a certain model is eliminated, slightly modified
or accepted in improving the scientific process. Later, aspects and questions of
accepting and improving global and fundamental models (e.g., general relativity
or quantum physics) formed part of the discussion of the philosophy of science.
1.3 The Significance and Benefits of Models 9

In the disciplines “Scientific Computing” and “Operations Research” models

often have a rather local meaning; a special part of reality15 is illustrated in
great detail. Here, on the one hand, the motivation for model development is
epistemological. On the other hand, we find military, pragmatic, and commercial
aspects, from which operational instructions for actions can often be derived. What
follows is a list of reasons and motives. The model maps most of the relevant
features and neglects less important aspects to
• provide insight into the problem,
• permit experimentation but avoid expensive and/or dangerous experiments,
• avoid the production of unwanted side products,
• optimize some objective,
• propose careful use of resources.
Let us now have a closer look on how to develop a mathematical model for a real-
world problem, or alternatively, how does one transform a given real-world problem
into a mathematical model? The associated process of modeling is by no means
easy nor unique; its difficulties somewhat resemble the solution of text problems, in
school, cf. Polya (1979,[443]), but the advantage was that from the teaching content
it was always known which mathematical technique was the most suitable one to
solve the problem at hand, e.g., the law of Pythagoras, quadratic equations, linear
systems with several variables, to name a few. In practice, however, one usually
does not know which mathematical technique is needed. Therefore it is important
to be open and creative in the first approaching attempts to understand and solve
the problem. It is helpful to know as many problem solving strategies and heuristics
[cf. Michalewicz & Fogel (2000,[397])] as possible and to be able to use them or
develop new ones from them. The following points are useful to remember when
starting to build a model:
1. There is no precise recipe telling the user how to build a model,
2. Experience and judgment are two important aspects of model building,
3. There is not necessarily a correct model,
4. There may not necessarily be a unique model, as different models focusing
on different aspects may be appropriate. Co-existence of models can be useful
to investigate different aspects (example: wave-particle dualism in physics). It
should be noted that different models for the description of the same facts may
require different data input based on very different databases.
5. Model building should follow the principle “as simple as possible, as complex as
necessary” or in a free translation of the words of Antoine de Saint-Exupéry: “A

15 Reality itself is already as compact as possible to be able to present itself lossless. Consequently

each self-image on a small section (individual human being) is inevitably connected with a loss of
information. So we shape ourselves and our view of the world only by a very small part of reality
which is filtered by our sensory organs and learned behavior. Our terms and languages in general
are products of limited perception — and therefore highly subjective. Modeling then occurs as
reduction in already limited worldviews.
10 1 Optimization: Using Models, Validating Models, Solutions, Answers

model is not ready when you can’t add anything anymore, but when you can’t
leave anything out.”
However, there are some aspects that should be considered when building a
model:
1. Definition of the problem and the purpose of the model: A clearly formulated task
should be used as a basis for defining the purpose of the model.
2. Problem scope: It is necessary to clearly define what belongs to the problem or
model and what does not (example: in a production planning system for medium-
and long-term planning we may neglect detailed planning aspects).
3. Identification of important model objects (examples: products, production sites,
warehouses, transport links, time periods, etc.) and relations between these
objects.
4. Acceptance and users: The model formulation should be adapted to the problem,
the model purpose, and the potential user. For instance, the planner who needs
to accept the model for his planning purposes. It helps if the modeler knows all
relevant aspects — and the motivations of people involved or designated to use
the model. Models for vehicle routing and their use can encounter acceptance
problems among truck drivers, if the client, the routing planner, has named the
minimization of travel costs as a target for the modeler, but each individual truck
driver is paid on the basis of the delivery value of goods. So a driver would rather
take a trip with one expensive piece of furniture to be delivered and mounted
than a trip with many destinations, where only cheap and perhaps even time-
consuming set-up work has to be done.
These points may already indicate that modeling is one of the most important
and critical tasks in optimization. This might suggest the question: is it possible
to learn good and efficient modeling? The answer is both “Yes” “No.” For good
model building, experience is essential; cf. Williams (1993,[580]) or Ashford and
Daniel (1992,[37]). The choice of the right decision variables to include the model
and a good model formulation correlate with one another. It is not necessarily
the number of decisions or restrictions that reflects the complexity of a problem
[concerning complexity, see remarks on page 555]. The criterion that separates a
good formulation from a bad one has a mathematical foundation which is difficult
for a modeler to establish, if at all possible. Nevertheless, rigid approaches exist
which try to improve the model formulation by exploiting certain mathematical
features [see Chaps. 8 and 9, and examples]. The effect is that the computing time
needed to solve the problem is reduced significantly. For further discussion see
Jeroslow & Lowe (1984,[288]) and Williams (1993,[580]). It is important that a
model developer needs to be knowledgeable about both: The “model” (the right
level of abstraction. . . ) and the “mathematical formulation,” which is also often
referred to as the “mathematical model.” We will come back to the importance of
good modeling again and again in this book. At this stage, let us first focus on the
question (Fig. 1.2):
1.3 The Significance and Benefits of Models 11

Fig. 1.2 A powerful modeler. Sometimes, modeling done right seems like magic. Produced for
this book by Diana Kallrath, Copyright ©2020

What is a model good for and how can we benefit from it? A well formulated and
documented model is useful for deduction, i.e., it allows us to derive consequences
and results. The results from the model have to be interpreted with respect to their
consequences in reality. The model building process increases our understanding
of a real-world problem. It documents structural features of that problem and thus
provides the know-how that others can use later on. The evaluation of a model
is coupled to the degree of reality contained in the model and the modeler’s
power. While within mathematics a strict mathematical proof is the ideal measure,
and one speaks of natural scientific models being subjected to falsification or
verification, in economic or business orientated models validation is what comes
12 1 Optimization: Using Models, Validating Models, Solutions, Answers

next to falsification and verification, whereby the relations

Validation ≤ Falsification/Verification ≤ Mathematical Proof

are fulfilled. However, one should always be aware that with a well-validated model,
the model is only a partial aspect of reality, and hence the model is less than reality.
However, it does not necessarily have to be a lot less. Electrodynamics, for example,
reproduces macroscopic electromagnetic phenomena so accurately that it is possible
to speak of a match between model and reality within the scope of measurement
accuracy. Furthermore, for the real-world problems and models we are dealing with,
we have to keep in mind that today’s reality (and model) can be quite different from
tomorrow’s reality. Let us conclude this section with a statement found in Box &
Draper (1987,[95]) “. . . all models are wrong, but some are useful.”

1.4 Mathematical Optimization

Four fundamental problem types are considered in this book. The first is linear
programming (LP), introduced now, the second is mixed integer linear programming
(MILP), to be introduced in Chap. 2, the third and fourth are nonlinear (NLP) and
mixed integer nonlinear programming (MINLP) to be treated in Chap. 3.

1.4.1 A Linear Optimization Example

We now turn to an example where we will identify the main features of this
linear optimization problem. Sunshine Cruises is a national company specializing
in renting out river-going boats to tourists. The company makes use of two types of
boats — the Premier and the Standard. The Premier is the more expensive boat to
hire and each Premier boat brings in an average net revenue of 800 per week, while
for the Standard the corresponding figure is £600 per week. The company is about
to scrap its older boats and lease new ones and has asked for some advice about
decision making. Decisions to be taken will be based on a scenario of “a typical
week.”
The company has maximum berthing accommodation around the country for
350 boats in total, of which not more than 200 can be Premier boats. The company
also wants to have no fewer Premier boats than Standard. Maintenance of boats is a
further restriction. Each Premier craft requires 4 h maintenance per week while each
Standard craft requires 3 h. The maximum number of maintenance hours available
per week is 1400 (Fig. 1.3).
1.4 Mathematical Optimization 13

Fig. 1.3 Boat problem illustrated. The decision maker from Sunshine Cruises pondering the
number of Premier boats and Standard boats. Produced for this book by Diana Kallrath, Copyright
©2020

The problem introduced so far is simple and structured. Many aspects of

complexity16 have been removed in advance and much information in the problem
is in an easily digestible form.
What are the degrees of freedom and what are the decisions?
Berthing and maintenance are consequences of having boats and are not fundamen-
tal decisions. The fundamental decisions are:
How many Premier boats should be leased?
How many Standard boats should be leased?
Once we have the answers to these questions, most other aspects of the problem
fall into place. We now endeavor to express the problem in terms of variables
representing the decisions (decision variables) and subject these decisions to a series
of conditions (constraints). We also associate a “driving force” with the problem
termed the objective (objective function) and we aim to optimize this objective.

16 The advanced reader or those who have had already a glance at Chap. 3 notice that this problem

is technically an integer linear programming problem. But for the moment, we avoid the difference
between Linear Programming involving only continuous variables and Integer Programming
involving integer variables.
14 1 Optimization: Using Models, Validating Models, Solutions, Answers

Let p be the desired number of Premier boats, and s the desired number of
Standard boats. For practical sense it is clear that neither p nor s should ever take
a negative value, but otherwise we can expect large or small values for p and s -
at the moment we are uncertain. It might be more realistic if p and s could avoid
taking values which include fractions. The fractional solution of 18.5 Premier boats
might not be appropriate, even if it were acceptable mathematically. However, we
will assume that for the moment such an answer will be regarded as appropriate, if it
arises, as such an answer is considered admissible in a linear programming problem.
For the example introduced, the variables can now be linked together in
expressions to model the relationships of the problem.
Firstly, there is a limit on accommodation for berthing. The total number of boats
in our model is

p+s (1.4.1)

and this total is limited by 350. We therefore introduce the inequality

p + s ≤ 350 (1.4.2)

to model this restriction. Notice that the restriction takes the form of an inequality
rather than an equation as the total number of boats which is desired to be used may
turn out to be less than 350.
The restriction that not more than 200 can be Premier boats is modeled by the
inequality

p ≤ 200. (1.4.3)

An inequality as (1.4.3), or any constraint (equality or inequality) containing a right-

hand side coefficient and only a single variable with coefficient 1 in front of it is
called a bound. It turns out that mathematical programming software handles bounds
particularly efficiently.
To model the maintenance restriction we first calculate the total number of main-
tenance hours required. This is obtained by taking the number, 4, of maintenance
hours required by each Premier boat and multiplying it by the number, p, of Premier
boats used giving 4p and adding to that maintenance time a similar expression for
the Standard boats, 3s, giving 4p + 3s. We then construct the inequality

4p + 3s ≤ 1400 (1.4.4)

to show that the total number of maintenance hours used is limited by 1400.
To model the restriction that there must be no fewer Premier boats than Standard
we introduce the constraint

p ≥ s,
Another Random Scribd Document
with Unrelated Content
 CHAPTER VI

A FTER the attempt to storm Copenhagen in February of fifty-nine,

the Swedes retired, and contented themselves with keeping the
city invested. The beleaguered townspeople breathed more freely.
The burdens of war were lightened, and they had time to rejoice in
the honors they had won and the privileges that had been conferred
on them. It is true, there were some who had found a zest in the
stirring scenes of war, and felt their spirits flag, as they saw dull
peace unfold its tedious routine, but the great mass of people were
glad and light at heart. Their happiness found vent in merry routs,
for weddings, christenings, and betrothals, long postponed while the
enemy was so oppressively near, gathered gay crowds in every court
and alley of the city.
Furthermore, there was time to take note of the neighbors and
make the mote in their eyes into a beam. There was time to
backbite, to envy and hate. Jealousies, whether of business or love,
shot a powerful growth again, and old enmity bore fruit in new
rancor and new vengeance. There was one who had lately
augmented the number of his enemies, until he had drawn well-nigh
the hate of the whole community upon his head. This man was
Corfitz Ulfeldt. He could not be reached, for he was safe in the camp
of the Swedes, but certain of his relatives and those of his wife, who
were suspected of a friendly regard for him, were subjected to
constant espionage and annoyance, while the court knew them not.
There were but few such, but among them was Sofie Urne, Ulrik
Frederik’s betrothed. The Queen, who hated Ulfeldt’s wife more than
she hated Ulfeldt himself, had from the first been opposed to Ulrik
Frederik’s alliance with a gentlewoman so closely related to Eleonore
Christine, and since the recent actions of Ulfeldt had placed him in a
more sinister light than ever, she began to work upon the King and
others, in order to have the engagement annulled.
 Nor was it long before the King shared the Queen’s view. Sofie
Urne, who was in fact given to intrigue, had been painted as so wily
and dangerous, and Ulrik Frederik as so flighty and easily led, that
the King clearly saw how much trouble might come of such an
alliance. Yet he had given his consent, and was too sensitive about
his word of honor to withdraw it. He therefore attempted to reason
with Ulrik Frederik, and pointed out how easily his present friendly
footing at court might be disturbed by a woman who was so
unacceptable to the King and Queen, and justly so, as her
sympathies were entirely with the foes of the royal house. Moreover,
he said, Ulrik Frederik was standing in his own light, since none
could expect important posts to be entrusted to one who was
constantly under the influence of the enemies of the court. Finally,
he alluded to the intriguing character of Mistress Sofie, and even
expressed doubt of the sincerity of her regard. True love, he said,
would have sacrificed itself rather than bring woe upon its object,
would have hidden its head in sorrow rather than exulted from the
housetops. But Mistress Sofie had shown no scruples; indeed, she
had used his youth and blind infatuation to serve her own ends.
The King talked long in this strain, but could not prevail upon
Ulrik Frederik, who still had a lively recollection of the pleading it had
cost him to make Mistress Sofie reveal her affection. He left the
King, more than ever resolved that nothing should part them. His
courtship of Mistress Sofie was the first serious step he had ever
taken in his life, and it was a point of honor with him to take it fully.
There had always been so many hands ready to lead and direct him,
but he had outgrown all that; he was old enough to walk alone, and
he meant to do it. What was the favor of the King and the court,
what were honor and glory, compared to his love? For that alone he
would strive and sacrifice; in that alone he would live.
The King, however, let it be known to Christoffer Urne that he
was opposed to the match, and the house was closed to Ulrik
Frederik, who henceforth could see Mistress Sofie only by stealth. At
first this merely fed the flame, but soon his visits to his betrothed
grew less frequent. He became more clear-sighted where she was
concerned, and there were moments when he doubted her love, and
even wondered whether she had not led him on, that summer day,
while she seemed to hold him off.
The court, which had hitherto met him with open arms, was cold
as ice. The King, who had taken such a warm interest in his future,
was indifference itself. There were no longer any hands stretched
out to help him, and he began to miss them; for he was by no
means man enough to go against the stream. When it merely
ceased to waft him along, he lost heart instantly. At his birth, a
golden thread had been placed in his hand, and he had but to follow
it upward to happiness and honor. He had dropped this thread to
find his own way, but he still saw it glimmering. What if he were to
grasp it again? He could neither stiffen his back to defy the King nor
give up Sofie. He had to visit her in secret, and this was perhaps the
hardest of all for his pride to stomach. Accustomed to move in pomp
and display, to take every step in princely style, he winced at
crawling through back alleys. Days passed, and weeks passed, filled
with inactive brooding and still-born plans. He loathed his own
helplessness, and began to despise himself for a laggard. Then came
the doubt: perhaps his dawdling had killed her love, or had she
never loved him? They said she was clever, and no doubt she was,
but—as clever as they said? Oh, no! What was love, then, if she did
not love, and yet—and yet....
Behind Christoffer Urne’s garden ran a passage just wide enough
for a man to squeeze through. This was the way Ulrik Frederik had
to take when he visited his mistress, and he would usually have
Hop-o’-my-Thumb mounted on guard at the end of the passage, lest
people in the street should see him climbing the board fence.
On a balmy, moonlit summer night, three or four hours after
bedtime, Daniel had wrapped himself in his cloak and found a seat
for himself on the remains of a pig’s trough, which some one had
thrown out from a neighboring house. He was in a pleasant frame of
mind, slightly drunk, and chuckling to himself at his own merry
conceits. Ulrik Frederik had already scaled the fence and was in the
garden. It was fragrant with elder-blossoms. Linen laid out to bleach
made long white strips across the grass. There was a soft rustling in
the maples overhead and the rose-bushes at his side; their red
blossoms looked almost white in the moonlight. He went up to the
house, which stood shining white, the windows in a yellow glitter.
How quiet everything was—radiant and calm! Suddenly the glassy
whirr of a cricket shivered the stillness. The sharp, blue-black
shadows of the hollyhocks seemed painted on the wall behind them.
A faint mist rose from the bleach-linen. There!—he lifted the latch,
and the next moment he was in the darkness within. Softly he
groped his way up the rickety staircase until he felt the warm, spice-
scented air of the attic. The rotten boards of the floor creaked under
his step. The moon shone through a small window overhead,
throwing a square of light on the flat top of a grain-pile. Scramble
over—the dust whirling in the column of light! Now—the gable-room
at last! The door opened from within, and threw a faint reddish glow
that illuminated for a second the pile of grain, the smoke-yellowed,
sloping chimney, and the roof-beams. The next moment they were
shut out, and he stood by Sofie’s side in the family clothes-closet.
The small, low room was almost filled with large linen-presses.
From the loft hung bags full of down and feathers. Old spinning-
wheels were flung into the corners, and the walls were festooned
with red onions and silver-mounted harness. The window was closed
with heavy wooden shutters, but on a brass-trimmed chest beneath
it stood a small hand-lantern. Sofie opened its tiny horn-pane to get
a brighter light. Her loosened hair hung down over the fur-edged
broadcloth robe she had thrown over her homespun dress. Her face
was pale and grief-worn, but she smiled gaily and poured out a
stream of chatter. She was sitting on a low stool, her hands clasped
around her knees, looking up merrily at Ulrik Frederik, who stood
silent above her, while she talked and talked, lashed on by the fear
his ill-humor had roused in her.
“How now, Sir Grumpy?” she said. “You’ve nothing to say? In all
the hundred hours that have passed, have you not thought of a
hundred things you wanted to whisper to me? Oh, then you have
not longed as I have!” She trimmed the candle with her fingers, and
threw the bit of burning wick on the floor. Instinctively Ulrik Frederik
took a step forward, and put it out with his foot.
“That’s right!” she went on. “Come here, and sit by my side; but
first you must kneel and sigh and plead with me to be fond again,
for this is the third night I’m watching. Yester eve and the night
before I waited in vain, till my eyes were dim.” She lifted her hand
threateningly. “To your knees, Sir Faithless, and pray as if for your
life!” She spoke with mock solemnity, then smiled, half beseeching,
half impatient. “Come here and kneel, come!”
Ulrik Frederik looked around almost grudgingly. It seemed too
absurd to fall on his knees there in Christoffer Urne’s attic. Yet he
knelt down, put his arm around her waist, and hid his face in her
lap, though without speaking.
She too was silent, oppressed with fear; for she had seen Ulrik
Frederik’s pale, tormented face and uneasy eyes. Her hand played
carelessly with his hair, but her heart beat violently in apprehension
and dread.
They sat thus for a long time.
Then Ulrik Frederik started up.
“No, no!” he cried. “This can’t go on! God our Father in heaven is
my witness, that you’re dear to me as the innermost blood of my
heart, and I don’t know how I’m to live without you. But what does
it avail? What can come of it? They’re all against us—every one. Not
a tongue will speak a word of cheer, but all turn from me. When
they see me, ’tis as though a cold shadow fell over them, where
before I brought a light. I stand so utterly alone, Sofie, ’tis bitter
beyond words. True, I know you warned me, but I’m eaten up in this
strife. It sucks my courage and my honor, and though I’m consumed
with shame, I must ask you to set me free. Dearest girl, release me
from my word!”
Sofie had risen and stood cold and unflinching like a statue,
eyeing him gravely, as he spoke.
“I am with child,” she said quietly and firmly.
 If she had consented, if she had given him his freedom, Ulrik
Frederik felt that he would not have taken it. He would have thrown
himself at her feet. Sure of her, he would have defied the King and
all. But she did not. She but pulled his chain to show him how
securely he was bound. Oh, she was clever as they said! His blood
boiled, he could have fallen upon her, clutched her white throat to
drag the truth out of her and force her to open every petal and lay
bare every shadow and fold in the rose of her love, that he might
know the truth at last! But he mastered himself and said with a
smile: “Yes, of course, I know—’twas nothing but a jest, you
understand.”
Sofie looked at him uneasily. No, it had not been a jest. If it had
been, why did he not come close to her and kiss her? Why did he
stand there in the shadow? If she could only see his eyes! No, it was
no jest. He had asked as seriously as she had answered. Ah, that
answer! She began to see what she had lost by it. If she had only
said yes, he would never have left her! “Oh, Ulrik Frederik,” she said,
“I was but thinking of our child, but if you no longer love me, then
go, go at once and build your own happiness! I will not hold you
back.”
“Did I not tell you that ’twas but a jest? How can you think that I
would ask you to release me from my word and sneak off in base
shame and dishonor! Whenever I lifted my head again,” he went on,
“I must fear lest the eye that had seen my ignominy should meet
mine and force it to the ground.” And he meant what he said. If she
had loved him as passionately as he loved her, then perhaps, but
now—never.
Sofie went to him and laid her head on his shoulder, weeping.
“Farewell, Ulrik Frederik,” she said. “Go, go! I would not hold you
one hour after you longed to be gone, no, not if I could bind you
with a hair.”
He shook his head impatiently. “Dear Sofie,” he said, winding
himself out of her arms, “let us not play a comedy with each other. I
owe it both to you and to myself that the pastor should join our
hands; it cannot be too soon. Let it be in two or three days—but
secretly, for it is of no use to set the world against us more than has
been done already.” Sofie dared not raise any objection. They agreed
on the time and the place, and parted with tender good-nights.
When Ulrik Frederik came down into the garden, it was dark, for
the moon had veiled itself, and a few heavy raindrops fell from the
inky sky. The early cocks were crowing in the mews, but Daniel had
fallen asleep on his post.
A week later his best parlor was the scene of Mistress Sofie’s and
Ulrik Frederik’s private marriage by an obscure clergyman. The
secret was not so well guarded, however, but that the Queen could
mention it to the King a few days later. The result was that in a
month’s time the contract was annulled by royal decree, and
Mistress Sofie was sent to the cloister for gentlewomen at Itzehoe.
Ulrik Frederik made no attempt to resist this step. Although he
felt deeply hurt, he was weary, and bowed in dull dejection to
whatever had to be. He drank too much almost every day, and when
in his cups would weep and plaintively describe to two or three boon
companions, who were his only constant associates, the sweet,
peaceful, happy life he might have led. He always ended with
mournful hints that his days were numbered, and that his broken
heart would soon be carried to that place of healing where the
bolsters were of black earth and the worms were chirurgeon.
The King, to make an end of all this, ordered him to accompany
the troops which the Dutch were transferring to Fyen, and thence he
returned in November with the news of the victory at Nyborg. He
resumed his place at the court and in the favor of the King, and
seemed to be quite his old self.
 CHAPTER VII

M ARIE GRUBBE was now seventeen.

On the afternoon when she fled in terror from the death-
bed of Ulrik Christian Gyldenlöve, she had rushed up to her own
chamber and paced the floor, wringing her hands, and moaning as
with intense bodily pain, until Lucie had run to Mistress Rigitze and
breathlessly begged her for God’s sake to come to Miss Marie, for
she thought something had gone to pieces inside of her. Mistress
Rigitze came, but could not get a word out of the child. She had
thrown herself before a chair with face hidden in the cushions, and
to all Mistress Rigitze’s questions answered only that she wanted to
go home, she wanted to go home, she wouldn’t stay a moment
longer, and she had wept and sobbed, rocking her head from side to
side. Mistress Rigitze had finally given her a good beating and
scolded Lucie, saying that between them they had nearly worried
the life out of her with their nonsense, and therewith she left the
two to themselves.
Marie took the beating with perfect indifference. Had any one
offered her blows in the happy days of her love, it would have
seemed the blackest calamity, the deepest degradation, but now it
no longer mattered. In one short hour, her longings, her faith, and
her hopes had all been withered, shrivelled up, and blown away. She
remembered once at Tjele when she had seen the men stone to
death a dog that had ventured within the high railing of the duck-
park. The wretched animal swam back and forth, unable to get out,
the blood running from many wounds, and she remembered how
she had prayed to God at every stone that it might strike deep, since
the dog was so miserable that to spare it would have been the
greatest cruelty. She felt like poor Diana, and welcomed every
sorrow, only wishing that it would strike deep, for she was so
unhappy that the deathblow was her only hope.
 Oh, if that was the end of all greatness—slavish whimpering,
lecherous raving, and craven terror!—then there was no such thing
as greatness. The hero she had dreamed of, he rode through the
portals of death with ringing spurs and shining mail, with head bared
and lance at rest, not with fear in witless eyes and whining prayers
on trembling lips. Then there was no shining figure that she could
dream of in worshipping love, no sun that she could gaze on till the
world swam in light and rays and color before her blinded eyes. It
was all dull and flat and leaden, bottomless triviality, lukewarm
commonplace, and nothing else.
Such were her first thoughts. She seemed to have been
transported for a short time to a fairy-land, where the warm, life-
pregnant air had made her whole being unfold like an exotic flower,
flashing sunlight from every petal, breathing fragrance in every vein,
blissful in its own light and scent, growing and growing, leaf upon
leaf and petal upon petal, in irresistible strength and fullness. But
this was all past. Her life was barren and void again; she was poor
and numb with cold. No doubt the whole world was like that, and all
the people likewise. And yet they went on living in their futile bustle.
Oh, her heart was sick with disgust at seeing them flaunt their
miserable rags and proudly listen for golden music in their empty
clatter.
Eagerly she reached for those treasured old books of devotion
that had so often been proffered her and as often rejected. There
was dreary solace in their stern words on the misery of the world
and the vanity of all earthly things, but the one book that she pored
over and came back to again and again was the Revelation of St.
John the Divine. She never tired of contemplating the glories of the
heavenly Jerusalem; she pictured it to herself down to the smallest
detail, walked through every by-way, peeped in at every door. She
was blinded by the rays of sardonyx and chrysolyte, chrysoprasus
and jacinth; she rested in the shadow of the gates of pearl and saw
her own face mirrored in the streets of gold like transparent glass.
Often she wondered what she and Lucie and Aunt Rigitze and all the
other people of Copenhagen would do when the first angel poured
out the vial of the wrath of God upon earth, and the second poured
out his vial, and the third poured out his—she never got any farther,
for she always had to begin over again.
When she sat at her work she would sing one long passion hymn
after another, in a loud, plaintive voice, and in her spare moments
she would recite whole pages from “The Chain of Prayerful Souls” or
“A Godly Voice for Each of the Twelve Months;” for these two she
knew almost by heart.
Underneath all this piety there lurked a veiled ambition. Though
she really felt the fetters of sin and longed for communion with God,
there mingled in her religious exercises a dim desire for power, a
half-realized hope that she might become one of the first in the
kingdom of heaven. This brooding worked a transformation in her
whole being. She shunned people and withdrew within herself. Even
her appearance was changed, the face pale and thin, the eyes
burning with a hard flame—and no wonder; for the terrible visions of
the Apocalypse rode life-size through her dreams at night, and all
day long her thoughts dwelt on what was dark and dreary in life.
When Lucie had gone to sleep in the evening, she would steal out of
bed and find a mystic ascetic pleasure in falling on her knees and
praying, till her bones ached and her feet were numb with cold.
Then came the time when the Swedes raised the siege, and all
Copenhagen divided its time between filling glasses as host and
draining them as guest. Marie’s nature, too, rebounded from the
strain, and a new life began for her, on a certain day when Mistress
Rigitze, followed by a seamstress, came up to her room and piled
the tables and chairs high with the wealth of sacks, gowns, and
pearl-embroidered caps that Marie had inherited from her mother. It
was considered time that she should wear grown-up clothes.
She was in raptures at being the centre of all the bustle that
broke in on her quiet chamber, all this ripping and measuring,
cutting and basting. How perfectly dear that pounce-red satin,
glowing richly where it fell in long, heavy folds, or shining brightly
where it fitted smoothly over her form! How fascinating the eager
parley about whether this silk chamelot was too thick to show the
lines of her figure or that Turkish green too crude for her
complexion! No scruples, no dismal broodings could stand before
this joyous, bright reality. Ah, if she could but once sit at the festive
board—for she had begun to go to assemblies—wearing this snow-
white, crisp ruff, among other young maidens in just as crisp ruffs,
all the past would become as strange to her as the dreams of
yesternight, and if she could but once tread the saraband and pavan
in sweeping cloth of gold and lace mitts and broidered linen, those
spiritual excesses would make her cheeks burn with shame.
It all came about: she was ashamed, and she did tread the
saraband and pavan; for she was sent twice a week, with other
young persons of quality, to dancing-school in Christen Skeel’s great
parlor, where an old Mecklenburger taught them steps and figures
and a gracious carriage according to the latest Spanish mode. She
learned to play on the lute, and was perfected in French; for
Mistress Rigitze had her own plans.
Marie was happy. As a young prince who has been held captive is
taken straight from the gloomy prison and harsh jailer to be lifted to
the throne by an exultant people, to feel the golden emblem of
power and glory pressed firmly upon his curls, and see all bowing
before him in smiling homage, so she had stepped from her quiet
chamber into the world, and all had hailed her as a queen indeed, all
had bowed, smiling, before the might of her beauty.
There is a flower called the pearl hyacinth; as that is blue so
were her eyes in color, but their lustre was that of the falling
dewdrop, and they were deep as a sapphire resting in shadow. They
could fall as softly as sweet music that dies, and glance up exultant
as a fanfare. Wistful—ay, as the stars pale at daybreak with a veiled,
tremulous light, so was her look when it was wistful. It could rest
with such smiling intimacy that many a man felt it like a voice in a
dream, far away but insistent, calling his name, but when it
darkened with grief it was full of such hopeless woe that one could
almost hear the heavy dripping of blood.
Such was the impression she made, and she knew it, but not
wholly. Had she been older and fully conscious of her beauty, it
might have turned her to stone. She might have come to look upon
it as a jewel to be kept burnished and in a rich setting, that it might
be the desire of all; she might have suffered admiration coldly and
quietly. Yet it was not so. Her beauty was so much older than herself
and she had so suddenly come into the knowledge of its power, that
she had not learned to rest upon it and let herself be borne along by
it, serene and self-possessed. Rather, she made efforts to please,
grew coquettish and very fond of dress, while her ears drank in
every word of praise, her eyes absorbed every admiring look, and
her heart treasured it all.
She was seventeen, and it was Sunday, the first Sunday after
peace had been declared. In the morning she had attended the
thanksgiving service, and in the afternoon she was dressing for a
walk with Mistress Rigitze.
The whole town was astir with excitement; for peace had opened
the city gates, which had been closed for twenty-two long months.
All were rushing to see where the suburb had stood, where the
enemy had been encamped, and where “ours” had fought. They had
to go down into the trenches, climb the barricades, peep into the
necks of the mines, and pluck at the gabions. This was the spot
where such a one had been posted, and here so-and-so had fallen,
and over there another had rushed forward and been surrounded.
Everything was remarkable, from the wheel-tracks of the cannon-
carriages and the cinders of the watch-fires to the bullet-pierced
board-fences and the sun-bleached skull of a horse. And so the
narrating and explaining, the supposing and debating, went on, up
the ramparts and down the barricades.
Gert Pyper was strutting about with his whole family. He stamped
the ground at least a hundred times and generally thought he
noticed a strangely hollow sound, while his rotund spouse pulled him
anxiously by the sleeve and begged him not to be too foolhardy, but
Master Gert only stamped the harder. The grown-up son showed his
little betrothed where he had been standing on the night when he
got a bullet-hole through his duffel great-coat, and where the
turner’s boy had had his head shot off. The smaller children cried,
because they were not allowed to keep the rifle-ball they had found;
for Erik Lauritzen, who was also there, said it might be poisoned. He
was poking the half-rotten straw where the barracks had stood, for
he remembered a story of a soldier who had been hanged outside of
Magdeburg, and under whose pillow seven of his comrades had
found so much money that they had deserted before the official
looting of the city began.
The green fields and grayish white roads were dotted black with
people coming and going. They walked about, examining the well-
known spots like a newly discovered world or an island suddenly
shot up from the bottom of the sea, and there were many who,
when they saw the country stretching out before them, field behind
field and meadow behind meadow, were seized with wanderlust and
began to walk on and on as though intoxicated with the sense of
space, of boundless space.
Toward supper time, however, the crowds turned homeward, and
as moved by one impulse, sought the North Quarter, where the
graveyard of St. Peter’s Church lay surrounded by spacious gardens;
for it was an old-time custom to take the air under the green trees,
after vespers on summer Sundays. While the enemy was encamped
before the ramparts, the custom naturally fell into disuse, and the
churchyard had been as empty on Sundays as on week days; but
this day old habits were revived, and people streamed in through
both entrances from Nörregade: nobles and citizens, high and low,
all had remembered the full-crowned linden trees of St. Peter’s
churchyard.
On the grassy mounds and the broad tombstones sat merry
groups of townspeople, man and wife, children and neighbors,
eating their supper, while in the outskirts of the party stood the
’prentice boy munching the delicious Sunday sandwich, as he waited
for the basket. Tiny children tripped with hands full of broken food
for the beggar youngsters that hung on the wall. Lads thirsting for
knowledge spelled their way through the lengthy epitaphs, while
father listened full of admiration, and mother and the girls scanned
the dresses of the passers-by: for by this time the gentlefolk were
walking up and down in the broad paths. They usually came a little
later than the others, and either supped at home or in one of the
eating-houses in the gardens round about.
Stately matrons and dainty maids, old councillors and young
officers, stout noblemen and foreign ministers, passed in review.
There went bustling, gray-haired Hans Nansen, shortening his steps
to the pace of the wealthy Villem Fiuren and listening to his piping
voice. There came Corfits Trolle and the stiff Otto Krag. Mistress Ide
Daa, famed for her lovely eyes, stood talking to old Axel Urup, who
showed his huge teeth in an everlasting smile, while the shrunken
form of his lady, Mistress Sidsel Grubbe, tripped slowly by the side of
Sister Rigitze and the impatient Marie. There were Gersdorf and
Schack and Thuresen of the tow-colored mane and Peder Retz with
Spanish dress and Spanish manners.
Ulrik Frederik was among the rest, walking with Niels
Rosenkrands, the bold young lieutenant-colonel, whose French
breeding showed in his lively gestures. When they met Mistress
Rigitze and her companions, Ulrik Frederik would have passed them
with a cold, formal greeting, for ever since his separation from Sofie
Urne he had nursed a spite against Mistress Rigitze, whom he
suspected, as one of the Queen’s warmest adherents, of having had
a finger in the matter. But Rosenkrands stopped, and Axel Urup
urged them so cordially to sup with the party in Johan Adolph’s
garden that they could not well refuse.
A few minutes later they were all sitting in the little brick
summer-house, eating the simple country dishes that the gardener
set before them.
“Is it true, I wonder,” asked Mistress Ide Daa, “that the Swedish
officers have so bewitched the maidens of Sjælland with their pretty
manners that they have followed them in swarms out of land and
kingdom?”
“Marry, it’s true enough at least of that minx, Mistress Dyre,”
replied Mistress Sidsel Grubbe.
“Of what Dyres is she?” asked Mistress Rigitze.
 “The Dyres of Skaaneland, you know, sister, those who have such
light hair. They’re all intermarried with the Powitzes. The one who
fled the country she’s a daughter of Henning Dyre of West
Neergaard, he who married Sidonie, the eldest of the Ove Powitzes,
and she went bag and baggage—took sheets, bolsters, plate, and
ready money from her father.”
“Ay,” smiled Axel Urup, “strong love draws a heavy load.”
“Faith,” agreed Oluf Daa, who always struck out with his left hand
when he talked, “love—as a man may say—love is strong.”
“Lo-ove,” drawled Rosenkrands, daintily stroking his moustache
with the back of his little finger, “is like Hercules in female dress,
gentle and charming in appearance and seeming all weak-ness and
mild-ness, yet it has stre-ength and craftiness to complete all the
twelve labors of Hercules.”
“Indeed,” broke in Mistress Ide Daa, “that is plainly to be seen
from the love of Mistress Dyre, which at least completed one of the
labors of Hercules, inasmuch as it cleaned out chests and presses,
even as he cleaned the stable of Uriah—or whatever his name was—
you know.”
“I would rather say”—Ulrik Frederik turned to Marie Grubbe
—“that love is like falling asleep in a desert and waking in a balmy
pleasure-garden, for such is the virtue of love that it changes the
soul of man, and that which was barren now seems a very wonder
of delight. But what are your thoughts about love, fair Mistress
Marie?”
“Mine?” she asked. “I think love is like a diamond; for as a
diamond is beautiful to look upon, so is love fair, but as the diamond
is poison to any one who swallows it, in the same manner love is a
kind of poison and produces a baneful raging distemper in those
who are infected by it—at least if one is to judge by the strange
antics one may observe in amorous persons and by their curious
conversation.”
“Ay,” whispered Ulrik Frederik gallantly, “the candle may well talk
reason to the poor moth that is crazed by its light!”
 “Forsooth, I think you are right, Marie,” began Axel Urup, pausing
to smile and nod to her. “Yes, yes, we may well believe that love is
but a poison, else how can we explain that coldblooded persons may
be fired with the most burning passion merely by giving them
miracle-philtres and love-potions?”
“Fie!” cried Mistress Sidsel; “don’t speak of such terrible godless
business—and on a Sunday, too!”
“My dear Sidse,” he replied, “there’s no sin in that—none at all.
Would you call it a sin, Colonel Gyldenlöve? No? Surely not. Does not
even Holy Writ tell of witches and evil sorceries? Indeed and indeed
it does. What I was about to say is that all our humors have their
seat in the blood. If a man is fired with anger, can’t he feel the blood
rushing up through his body and flooding his eyes and ears? And if
he’s frightened o’ the sudden, does not the blood seem to sink down
into his feet and grow cold all in a trice? Is it for nothing, do you
think, that grief is pale and joy red as a rose? And as for love, it
comes only after the blood has ripened in the summers and winters
of seventeen or eighteen years; then it begins to ferment like good
grape-wine; it seethes and bubbles. In later years it clears and
settles as do other fermenting juices; it grows less hot and fierce.
But as good wine begins to effervesce again when the grape-vine is
in bloom, so the disposition of man, even of the old, is more than
ordinarily inclined to love at certain seasons of the year, when the
blood, as it were, remembers the springtime of life.”
“Ay, the blood,” added Oluf Daa, “as a man may say, the blood
—’tis a subtle matter to understand—as a man may say.”
“Indeed,” nodded Mistress Rigitze, “everything acts on the blood,
both sun and moon and approaching storm, that’s as sure as if
’twere printed.”
“And likewise the thoughts of other people,” said Mistress Ide. “I
saw it in my eldest sister. We lay in one bed together, and every
night, as soon as her eyes were closed, she would begin to sigh and
stretch her arms and legs and try to get out of bed as some one
were calling her. And ’twas but her betrothed, who was in Holland,
and was so full of longing for her that he would do nothing day and
night but think of her, until she never knew an hour’s peace, and her
health—don’t you remember, dear Mistress Sidsel, how weak her
eyesight was all the time Jörgen Bille was from home?”
“Do I remember? Ah, the dear soul! But she bloomed again like a
rosebud. Bless me, her first lying-in—” and she continued the
subject in a whisper.
Rosenkrands turned to Axel Urup. “Then you believe,” he said,
“that an elixir d’am-our is a fermenting juice poured into the blood?
That tallies well with a tale the late Mr. Ulrik Christian told me one
day we were on the ramparts together. ’Twas in Antwerp it
happened—in the Hotellerie des Trois Brochets, where he had
lodgings. That morning at ma-ass he had seen a fair, fair maid-en,
and she had looked quite kind-ly at him. All day long she was not in
his thoughts, but at night when he entered his chamber, there was a
rose at the head of the bed. He picked it up and smelled it, and in
the same mo-ment the coun-ter-feit of the maiden stood before him
as painted on the wall, and he was seized with such sudden and fu-
rious longing for her that he could have cried aloud. He rushed out
of the house and into the street, and there he ran up and down,
wail-ing like one be-witched. Something seemed to draw and draw
him and burn like fire, and he never stopped till day dawned.”
So they talked until the sun went down, and they parted to go
home through the darkening streets. Ulrik Frederik joined but little in
the general conversation; for he was afraid that if he said anything
about love, it might be taken for reminiscences of his relation with
Sofie Urne. Nor was he in the mood for talking, and when he and
Rosenkrands were alone he made such brief, absentminded replies
that his companion soon wearied of him and left him to himself.
Ulrik Frederik turned homeward to his own apartments, which
this time were at Rosenborg. His valet being out, there was no light
in the large parlor, and he sat alone there in the dark till almost
midnight.
He was in a strange mood, divided between regret and
foreboding. It was one of those moods when the soul seems to drift
as in a light sleep, without will or purpose, on a slowly gliding
stream, while mist-like pictures pass on the background of dark
trees, and half-formed thoughts rise from the sombre stream like
great dimly-lit bubbles that glide—glide onward and burst. Bits of the
conversation that afternoon, the motley crowds in the churchyard,
Marie Grubbe’s smile, Mistress Rigitze, the Queen, the King’s favor,
the King’s anger that other time,—the way Marie moved her hands,
Sofie Urne, pale and far away,—yet paler and yet farther away,—the
rose at the head of the bed and Marie Grubbe’s voice, the cadence
of some word,—he sat listening and heard it again and again
winging through the silence.
He rose and went to the window, opened it, and leaned his
elbows on the wide casement. How fresh it all was—so cool and
quiet! The bittersweet smell of roses cooled with dew, the fresh,
pungent scent of new-mown hay, and the spicy fragrance of the
flowering maple were wafted in. A mist-like rain spread a blue,
tremulous dusk over the garden. The black boughs of the larch, the
drooping leafy veil of the birch, and the rounded crowns of the
beech stood like shadows breathed on a background of gliding mist,
while the clipped yew-trees shot upward like the black columns of a
roofless temple.
The stillness was that of a deep grave, save for the raindrops,
falling light as thistledown, with a faint, monotonous sound like a
whisper that dies and begins again and dies there behind the wet,
glistening trunks.
What a strange whisper it was when one listened! How wistful!—
like the beating of soft wings when old memories flock. Or was it a
low rustle in the dry leaves of lost illusions? He felt lonely, drearily
alone and forsaken. Among all the thousands of hearts that beat
round about in the stillness of the night, not one turned in longing to
him! Over all the earth there was a net of invisible threads binding
soul to soul, threads stronger than life, stronger than death; but in
all that net not one tendril stretched out to him. Homeless, forsaken!
Forsaken? Was that a sound of goblets and kisses out there? Was
there a gleam of white shoulders and dark eyes? Was that a laugh
ringing through the stillness?—What then? Better the slow-dripping
bitterness of solitude than that poisonous, sickly sweetness.... Oh,
curses on it! I shake your dust from my thoughts, slothful life, life for
dogs, for blind men, for weaklings.... As a rose! O God, watch over
her and keep her through the dark night! Oh, that I might be her
guard and protector, smooth every path, shelter her against every
wind—so beautiful—listening like a child—as a rose!...
 CHAPTER VIII

A DMIRED and courted though she was, Marie Grubbe soon found
that, while she had escaped from the nursery, she was not fully
admitted to the circles of the grown up. For all the flatteries lavished
on them, such young maidens were kept in their own place in
society. They were made to feel it by a hundred trifles that in
themselves meant nothing, but when taken together meant a great
deal. First of all, the children were insufferably familiar, quite like
their equals. And then the servants—there was a well-defined
difference in the manner of the old footman when he took the cloak
of a maid or a matron, and the faintest shade in the obliging smile of
the chambermaid showed her sense of whether she was waiting on
a married or an unmarried woman. The free-and-easy tone which
the half-grown younkers permitted themselves was most unpleasant,
and the way in which snubbings and icy looks simply slid off from
them was enough to make one despair.
She liked best the society of the younger men, for even when
they were not in love with her, they would show her the most
delicate attention and say the prettiest things with a courtly
deference that quite raised her in her own estimation,—though to be
sure it was tiresome when she found that they did it chiefly to keep
in practice. Some of the older gentlemen were simply intolerable
with their fulsome compliments and their mock gallantry, but the
married women were worst of all, especially the brides. The
encouraging, though a bit preoccupied glance, the slight
condescending nod with head to one side, and the smile—half
pitying, half jeering—with which they would listen to her—it was
insulting! Moreover, the conduct of the girls themselves was not of a
kind to raise their position. They would never stand together, but if
one could humiliate another, she was only too glad to do so. They
had no idea of surrounding themselves with an air of dignity by
attending to the forms of polite society the way the young married
women did.
Her position was not enviable, and when Mistress Rigitze let fall a
few words to the effect that she and other members of the family
had been considering a match between Marie and Ulrik Frederik, she
received the news with joy. Though Ulrik Frederik had not taken her
fancy captive, a marriage with him opened a wide vista of pleasant
possibilities. When all the honors and advantages had been
described to her—how she would be admitted into the inner court
circle, the splendor in which she would live, the beaten track to fame
and high position that lay before Ulrik Frederik as the natural son
and even more as the especial favorite of the King,—while she made
a mental note of how handsome he was, how courtly, and how much
in love,—it seemed that such happiness was almost too great to be
possible, and her heart sank at the thought that, after all, it was
nothing but loose talk, schemes, and hopes.
Yet Mistress Rigitze was building on firm ground, for not only had
Ulrik Frederik confided in her and begged her to be his spokesman
with Marie, but he had induced her to sound the gracious pleasure
of the King and Queen, and they had both received the idea very
kindly and had given their consent, although the King had felt some
hesitation to begin with. The match had, in fact, been settled long
since by the Queen and her trusted friend and chief gentlewoman,
Mistress Rigitze, but the King was not moved only by the
persuasions of his consort. He knew that Marie Grubbe would bring
her husband a considerable fortune, and although Ulrik Frederik held
Vordingborg in fief, his love of pomp and luxury made constant
demands upon the King, who was always hard pressed for money.
Upon her marriage Marie would come into possession of her
inheritance from her dead mother, Mistress Marie Juul, while her
father, Erik Grubbe, was at that time owner of the manors of Tjele,
Vinge, Gammelgaard, Bigum, Trinderup, and Nörbæk, besides
various scattered holdings. He was known as a shrewd manager who
wasted nothing, and would no doubt leave his daughter a large
fortune. So all was well. Ulrik Frederik could go courting without
more ado, and a week after midsummer their betrothal was
solemnized.
Ulrik Frederik was very much in love, but not with the stormy
infatuation he had felt when Sofie Urne ruled his heart. It was a
pensive, amorous, almost wistful sentiment, rather than a fresh,
ruddy passion. Marie had told him the story of her dreary childhood,
and he liked to picture to himself her sufferings with something of
the voluptuous pity that thrills a young monk when he fancies the
beautiful white body of the female martyr bleeding on the sharp
spikes of the torture-wheel. Sometimes he would be troubled with
dark forebodings that an early death might tear her from his arms.
Then he would vow to himself with great oaths that he would bear
her in his hands and keep every poisonous breath from her, that he
would lead the light of every gold-shining mood into her young heart
and never, never grieve her.
Yet there were other times when he exulted at the thought that
all this rich beauty, this strange, wonderful soul were given into his
power as the soul of a dead man into the hands of God, to grind in
the dust if he liked, to raise up when he pleased, to crush down, to
bend.
It was partly Marie’s own fault that such thoughts could rise in
him, for her love, if she did love, was of a strangely proud, almost
insolent nature. It would be but a halting image to say that her love
for the late Ulrik Christian had been like a lake whipped and tumbled
by a storm, while her love for Ulrik Frederik was the same water in
the evening, becalmed, cold, and glassy, stirred but by the breaking
of frothy bubbles among the dark reeds of the shore. Yet the simile
would have some truth, for not only was she cold and calm toward
her lover, but the bright myriad dreams of life that thronged in the
wake of her first passion had paled and dissolved in the drowsy calm
of her present feeling.
She loved Ulrik Frederik after a fashion, but might it not be
chiefly as the magic wand opening the portals to the magnificent
pageant of life, and might it not be the pageant that she really
loved? Sometimes it would seem otherwise. When she sat on his
knee in the twilight and sang little airs about Daphne and Amaryllis
to her own accompaniment, the song would die away, and while her
fingers played with the strings of the cithern, she would whisper in
his waiting ear words so sweet and warm that no true love owns
them sweeter, and there were tender tears in her eyes that could be
only the dew of love’s timid unrest. And yet—might it not be that her
longing was conjuring up a mere mood, rooted in the memories of
her past feeling, sheltered by the brooding darkness, fed by hot
blood and soft music,—a mood that deceived herself and made him
happy? Or was it nothing but maidenly shyness that made her chary
of endearments by the light of day, and was it nothing but girlish
fear of showing a girl’s weakness that made her eyes mock and her
lips jeer many a time when he asked for a kiss or, vowing love,
would draw from her the words all lovers long to hear? Why was it,
then, that when she was alone, and her imagination had wearied of
picturing for the thousandth time the glories of the future, she would
often sit gazing straight before her hopelessly, and feel unutterably
lonely and forsaken?

In the early afternoon of an August day Marie and Ulrik Frederik

were riding, as often before, along the sandy road that skirted the
Sound beyond East Gate. The air was fresh after a morning shower,
the sun stood mirrored in the water, and blue thunder-clouds were
rolling away in the distance.
They cantered as quickly as the road would allow them, a lackey
in a long crimson coat following closely. They rode past the gardens
where green apples shone under dark leaves, past fish-nets hung to
dry with the raindrops still glistening in their meshes, past the King’s
fisheries with red-tiled roof, and past the glue-boiler’s house, where
the smoke rose straight as a column out of a chimney. They jested
and laughed, smiled and laughed, and galloped on.
At the sign of the Golden Grove they turned and rode through
the woods toward Overdrup, then walked their horses through the
underbrush down to the bright surface of the lake. Tall beeches
leaned to mirror their green vault in the clear water. Succulent
marsh-grass and pale pink feather-foil made a wide motley border
where the slope, brown with autumn leaves, met the water. High in
the shelter of the foliage, in a ray of light that pierced the cool
shadow, mosquitoes whirled in a noiseless swarm. A red butterfly
gleamed there for a second, then flew out into the sunlight over the
lake. Steel-blue dragon-flies made bright streaks through the air, and
the darting pike drew swift wavy lines over the surface of the water.
Hens were cackling in the farm-yard beyond the brushwood, and
from the other side of the lake came a note of wood-doves cooing
under the domes of the beech-trees in Dyrehaven.
They slackened their speed and rode out into the water to let
their horses dabble their dusty hoofs and quench their thirst. Marie
had stopped a little farther out than Ulrik Frederik, and sat with reins
hanging in order to let her mare lower its head freely. She was
tearing the leaves from a long branch in her hand, and sent them
fluttering down over the water, which was beginning to stir in soft
ripples.
“I think we may get a thunder-storm,” she said, her eyes
following the course of a light wind that went whirling over the lake,
raising round, dark, roughened spots on the surface.
“Perhaps we had better turn back,” suggested Ulrik Frederik.
“Not for gold!” she answered and suddenly drove her mare to the
shore. They walked their horses round the lake to the road and
entered the tall woods.
“I would I knew,” said Marie, when she felt the cool air of the
forest fan her cheeks and drew in its freshness in long, deep
breaths. “I would I knew—” She got no further, but stopped and
looked up into the green vault with shining eyes.
“What wouldst thou know, dear heart?”
“I’m thinking there’s something in the forest air that makes
sensible folks mad. Many’s the time I have been walking in Bigum
woods, when I would keep on running and running, till I got into the
very thickest of it. I’d be wild with glee and sing at the top of my
voice and walk and pick flowers and throw them away again and call
to the birds, when they flew up—and then, on the sudden, a strange
fright would come over me, and I would feel, oh! so wretched and
so small! Whenever a branch broke I’d start, and the sound of my
own voice gave me more fright than anything else. Hast thou never
felt it?”
Before Ulrik Frederik could answer her song rang out:

“Right merrily in the woods I go

Where elm and apple grow,
And I pluck me there sweet roses two
And deck my silken shoe.
Oh, the dance,
Oh, the dance,
Oh, tra-la-la!
Oh, the red, red berries on the dogrose bush!”

and as she sang, the whip flew down over her horse, she laughed,
hallooed, and galloped at top speed along a narrow forest path,
where the branches swept her shoulders. Her eyes sparkled, her
cheeks burned, she did not heed Ulrik Frederik calling after her. The
whip whizzed through the air again, and off she went with reins
slack! Her fluttering habit was flecked with foam. The soft earth flew
up around her horse. She laughed and cut the tall ferns with her
whip.
Suddenly the light seemed to be lifted from leaf and branch and
to flee from the rain-heavy darkness. The rustling of the bushes had
ceased, and the hoof-beats were silent, as she rode across a stretch
of forest glade. On either side the trees stood like a dark encircling
wall. Ragged gray clouds were scudding over the black, lowering
heavens. Before her rolled the murky blue waters of the Sound, and
beyond rose banks of fog. She drew rein, and her tired mount
stopped willingly. Ulrik Frederik galloped past, swung back in a wide
circle, and halted at her side.
At that moment a shower fell like a gray, heavy, wet curtain
drawn slantwise over the Sound. An icy wind flattened the grass,
whizzed in their ears, and made a noise like foaming waves in the
distant tree-tops. Large flat hailstones rattled down over them in
white sheets, settled like bead strings in the folds of her dress, fell in
a spray from the horses’ manes, and skipped and rolled in the grass
as though swarming out of the earth.
They sought shelter under the trees, rode down to the beach,
and presently halted before the low door of the Bide-a-Wee Tavern.
A stable-boy took the horses, and the tall, bareheaded inn-keeper
showed them into his parlor, where, he said, there was another
guest before them. It proved to be Hop-o’-my-Thumb, who rose at
their entrance, offering to give up the room to their highnesses, but
Ulrik Frederik graciously bade him remain.
“Stay here, my man,” he said, “and entertain us in this
confounded weather. I must tell you, my dear,”—turning to Marie,
—“that this insignificant mannikin is the renowned comedian and
merry-andrew of ale-houses, Daniel Knopf, well learned in all the
liberal arts such as dicing, fencing, drinking, shrovetide sports, and
such matters, otherwise in fair repute as an honorable merchant in
the good city of Copenhagen.”
Daniel scarcely heard this eulogy. He was absorbed in looking at
Marie Grubbe and formulating some graceful words of felicitation,
but when Ulrik Frederik roused him with a sounding blow on his
broad back, his face flushed with resentment and embarrassment.
He turned to him angrily, but mastered himself, and said with his
coldest smile: “We’re scarce tipsy enough, Colonel.”
Ulrik Frederik laughed and poked his side, crying: “Oh, you
sacred knave! Would you put me to confusion, you plaguy devil, and
make me out a wretched braggart who lacks parchments to prove
his boasting? Fie, fie, out upon you! Is that just? Have I not a score
of times praised your wit before this noble lady, till she has time and
again expressed the greatest longing to see and hear your far-famed
drolleries? You might at least give us the blind Cornelius Fowler and
his whistling birds, or play the trick—you know—with the sick cock
and the clucking hens!”
 Marie now added her persuasions, saying that Colonel Gyldenlöve
was quite right, she had often wondered what pastime, what fine
and particular sport, could keep young gentlemen in filthy ale-
houses for half days and whole nights together, and she begged that
Daniel would oblige them without further urging.
Daniel bowed with perfect grace and replied that his poor pranks
were rather of a kind to give fuddled young sparks added occasion
for roaring and bawling than to amuse a dainty and highborn young
maiden. Nevertheless, he would put on his best speed to do her
pleasure, for none should ever say it of him that any command from
her fair ladyship had failed of instant obedience and execution.
“Look ’ee!” he began, throwing himself down by the table and
sticking out his elbows. “Now I’m a whole assembly of your
betrothed’s honorable companions and especial good friends.”
He took a handful of silver dollars from his pocket and laid them
on the table, pulled his hair down over his eyes, and dropped his
lower lip stupidly.
“Devil melt me!” he drawled, rattling the coins like dice. “I’m not
the eldest son of the honorable Erik Kaase for nothing! What! you’d
doubt my word, you muckworm? I flung ten, hell consume me, ten
with a jingle! Can’t you see, you dog? I’m asking if you can’t see?—
you blind lamprey, you! Or d’ye want me to rip your guts with my
stinger and give your liver and lungs a chance to see too? Shall I—
huh? You ass!”
Daniel jumped up and pulled a long face.
“You’d challenge me, would you?” he said hoarsely with a strong
North Skaane accent, “you stinkard, you! D’you know whom you’re
challenging? So take me king o’ hell, I’ll strike your—Nay, nay,” he
dropped into his natural voice, “that’s perhaps too strong a jest to
begin with. Try another!”
He sat down, folded his hands on the edge of his knees as
though to make room for his stomach, puffed himself up, fat and
heavy jowled, then whistled firmly and thoughtfully but in an
altogether too slow tempo the ballad of Roselil and Sir Peter. Then
he stopped, rolled his eyes amorously, and called in fond tones:
“Cockatoo—cockadoodle-doo!” He began to whistle again, but
had some difficulty in combining it with an ingratiating smile. “Little
sugar-top!” he called, “little honey-dew, come to me, little chuck!
P’st! Will it lap wine, little kitty? Lap nice sweet wine from little
cruse?”
Again he changed his voice, leaned forward in his chair, winked
with one eye, and crooked his fingers to comb an imaginary beard.
“Now stay here,” he said coaxingly, “stay here, fair Karen; I’ll
never forsake you, and you must never forsake me,”—his voice grew
weepy,—“we’ll never part, my dear, dear heart, never in the world!
Silver and gold and honor and glory and precious noble blood—
begone! I curse you! Begone! I say. You’re a hundred heavens high
above them, the thing of beauty you are! Though they’ve
scutcheons and emblems—would that make ’em any better? You’ve
got an emblem, too—the red mark on your white shoulder that
Master Anders burned with his hot iron, that’s your coat-of-arms! I
spit on my scutcheon to kiss that mark—that’s all I think of
scutcheons—that’s all! For there isn’t in all the land of Sjælland a
high-born lady as lovely as you are—is there, huh? No, there isn’t—
not a bit of one!”
“That’s—that’s a lie!” he cried in a new voice, jumped up, and
shook his fist over the table. “My Mistress Ide, you blockhead, she’s
got a shape—as a man may say—she’s got limbs—as a man may say
—limbs, I tell you, you slubberdegulleon!”
At this point Daniel was about to let himself fall into the chair
again, but at that moment Ulrik Frederik pulled it away, and he rolled
on the floor. Ulrik Frederik laughed uproariously, but Marie ran to
him with hands outstretched as though to help him up. The little
man, half rising on his knees, caught her hand and gazed at her with
an expression so full of gratitude and devotion that it haunted her
for a long time. Presently they rode home, and none of them
thought that this chance meeting in the Bide-a-Wee Tavern would
lead to anything further.
 CHAPTER IX

T HE States-General that convened in Copenhagen in the late

autumn brought to town many of the nobility, all anxious to
guard their ancient rights against encroachment, but none the less
eager for a little frolic after the busy summer. Nor were they averse
to flaunting their wealth and magnificence in the faces of the
townspeople, who had grown somewhat loud-voiced since the war,
and to reminding them that the line between gentlemen of the realm
and the unfree mob was still firm and immutable, in spite of the
privileges conferred by royalty, in spite of citizen valor and the
glamor of victory, in spite of the teeming ducats in the strong boxes
of the hucksters.
The streets were bright with throngs of noblemen and their
ladies, bedizened lackeys, and richly caparisoned horses in silver-
mounted harness. There was feasting and open house in the homes
of the nobility. Far into the night the violin sounded from well-lit
halls, telling the sleepy citizens that the best blood of the realm was
warming to a stately dance over parquet floors, while the wine
sparkled in ancestral goblets.
All these festivities passed Marie Grubbe by; none invited her.
Because of their ties to the royal family, some of the Grubbes were
suspected of siding with the King against the Estate, and moreover
the good old nobility cordially hated that rather numerous upper
aristocracy formed by the natural children of the kings and their
relatives. Marie was therefore slighted for a twofold reason, and as
the court lived in retirement during the session of the States-
General, it offered her no compensation.
It seemed hard at first, but soon it woke the latent defiance of
her nature and made her draw closer to Ulrik Frederik. She loved
him more tenderly for the very reason that she felt herself being
wronged for his sake. So when the two were quietly married on the
sixteenth of December, sixteen hundred and sixty, there was the best
reason to believe that she would live happily with the Master of the
King’s Hunt, which was the title and office Ulrik Frederik had won as
his share of the favors distributed by triumphant royalty.
This private ceremony was not in accordance with the original
plan, for it had long been the intention of the King to celebrate their
wedding in the castle, as Christian the Fourth had done that of Hans
Ulrik and Mistress Rigitze, but at the eleventh hour he had scruples
and decided, in consideration of Ulrik Frederik’s former marriage and
divorce, to refrain from public display.

So now they are married and settled, and time passes, and time
flies, and all is well—and time slackened its speed, and time crawled;
for it is true, alas! that when Leander and Leonora have lived
together for half a year, the glory is often departed from Leander’s
love, though Leonora usually loves him much more tenderly than in
the days of their betrothal. She is like the small children, who find
the old story new, no matter how often it is told with the very same
words, the same surprises, and the self-same “Snip, snap, snout, my
tale’s out,” while Leander is more exacting and grows weary as soon
as his feeling no longer makes him new to himself. When he ceases
to be intoxicated, he suddenly becomes more than sober. The flush
and glamor of his ecstasy, which for a while gave him the assurance
of a demigod, suddenly departs; he hesitates, he thinks, and begins
to doubt. He looks back at the chequered course of his passion,
heaves a sigh, and yawns. He is beset with longing, like one who
has come home after a lengthy sojourn in foreign parts, and sees
the altogether too familiar though long-forgotten spots before him;
as he looks at them, he wonders idly whether he has really been
gone from this well-known part of the world so long.
In such a mood, Ulrik Frederik sat at home one rainy day in
September. He had called in his dogs and had frolicked with them for
a while, had tried to read, and had played a game of backgammon
with Marie. The rain was pouring. It was impossible to go walking or
riding, and so he had sought his armory, as he called it, thinking he
would polish and take stock of his treasures—this was just the day
for it! It occurred to him that he had inherited a chest of weapons
from Ulrik Christian; he had ordered it brought down from the attic,
and sat lifting out one piece after another.
There were splendid rapiers of bluish steel inlaid with gold, or
silvery bright with dull engraving. There were hunting-knives, some
heavy and one-edged, some long and flexible like tongues of flame,
some three-edged and sharp as needles. There were toledo blades,
many toledos, light as reeds and flexible as willows, with hilts of
silver and jasper agate, or of chased gold or gold and carbuncles.
One had nothing but a hilt of etched steel, and for a sword-knot a
little silk ribbon embroidered in roses and vines with red glass beads
and green floss. It must be either a bracelet, a cheap bracelet, or—
Ulrik Frederik thought—more likely a garter, and the rapier was stuck
through it.
It comes from Spain, said Ulrik Frederik to himself, for the late
owner had served in the Spanish army for nine years. Alack-a-day!
He too was to have entered foreign service with Carl Gustaf; but
then came the war, and now he supposed he would never have a
chance to get out and try his strength, and yet he was but three and
twenty. To live forever here at this tiresome little court,—doubly
tiresome since the nobility stayed at home,—to hunt a little, look to
his estate once in a while, some time in the future by the grace of
the King to be made Privy Councillor of the Realm and be knighted,
keep on the right side of Prince Christian and retain his office, now
and then be sent on a tedious embassy to Holland, grow old, get the
rheumatism, die, and be buried in Vor Frue Church,—such was the
brilliant career that stretched before him. And now they were
fighting down in Spain! There was glory to be won, a life to be lived
—that was where the rapier and the sword-knot came from. No, he
must speak to the King. It was still raining, and it was a long way to
Frederiksborg, but there was no help for it. He could not wait; the
matter must be settled.
The King liked his scheme. Contrary to his custom, he assented
at once, much to the surprise of Ulrik Frederik, who during his whole
ride had debated with himself all the reasons that made his plan
difficult, unreasonable, impossible. But the King said Yes, he might
leave before Christmas. By that time the preparations could be
completed and an answer received from the King of Spain.
The reply came in the beginning of December, but Ulrik Frederik
did not start until the middle of April; for there was much to be
done. Money had to be raised, retainers equipped, letters written.
Finally he departed.
Marie Grubbe was ill pleased with this trip to Spain. It is true, she
saw the justice of Mistress Rigitze’s argument that it was necessary
for Ulrik Frederik to go abroad and win honor and glory, in order that
the King might do something handsome for him; for although his
Majesty had been made an absolute monarch, he was sensitive to
what people said, and the noblemen had grown so captious and
perverse that they would be sure to put the very worst construction
on anything the King might do. Yet women have an inborn dread of
all farewells, and in this case there was much to fear. Even if she
could forget the chances of war and the long, dangerous journey,
and tell herself that a king’s son would be well taken care of, yet she
could not help her foreboding that their life together might suffer
such a break by a separation of perhaps more than a year that it
would never be the same again. Their love was yet so lightly rooted,
and just as it had begun to grow, it was to be mercilessly exposed to
ill winds and danger. Was it not almost like going out deliberately to
lay it waste? And one thing she had learned in her brief married life:
the kind of marriage she had thought so easy in the days of her
betrothal, that in which man and wife go each their own way, could
mean only misery with all darkness and no dawn. The wedge had
entered their outward life; God forbid that it should pierce to their
hearts! Yet it was surely tempting fate to open the door by such a
parting.
Moreover, she was sadly jealous of all the light papistical feminine
rabble in the land and dominions of Spain.
Frederik the Third, who, like many sovereigns of his time, was much
interested in the art of transmuting baser metals into gold, had
charged Ulrik Frederik when he came to Amsterdam to call on a
renowned alchemist, the Italian Burrhi, and to drop a hint that if he
should think of visiting Denmark, the King and the wealthy Christian
Skeel of Sostrup would make it worth his while.
When Ulrik Frederik arrived in Amsterdam, he therefore asked
Ole Borch, who was studying there and knew Burrhi well, to conduct
him to the alchemist. They found him a man in the fifties, below
middle height, and with a tendency to fat, but erect and springy in
his movements. His hair and his narrow moustache were black, his
nose was hooked and rather thick, his face full and yellow in color;
from the corners of his small, glittering black eyes innumerable
furrows and lines spread out like a fan, giving him an expression at
once sly and goodhumored. He wore a black velvet coat with wide
collar and cuffs and crape-covered silver buttons, black knee-
breeches and silk stockings, and shoes with large black rosettes. His
taste for fine lace appeared in the edging on his cravat and shirt
bosom and in the ruffles that hung in thick folds around his wrists
and knees. His hands were small, white, and chubby, and were
loaded with rings of such strange, clumsy shapes that he could not
bring the tips of his fingers together. Large brilliants glittered even
on his thumbs. As soon as they were seated, he remarked that he
was troubled with cold hands and stuck them in a large fur muff,
although it was summer.
The room into which he conducted Ulrik Frederik was large and
spacious, with a vaulted ceiling and narrow Gothic windows set high
in the walls. Chairs were ranged around a large centre table, their
wooden seats covered with soft cushions of red silk, from which
hung long, heavy tassels. The top of the table was inlaid with a
silver plate on which the twelve signs of the zodiac, the planets, and
some of the more important constellations were done in niello.
Above it, a string of ostrich eggs hung from the ceiling. The floor
had been painted in a chequered design of red and gray, and near
the door a triangle was formed by old horseshoes that had been
fitted into the boards. A large coral tree stood under one window,
and a cupboard of dark carved wood with brass mountings was
placed under the other. A life-size doll representing a Moor was set
in one corner, and along the walls lay blocks of tin and copper ore.
The blackamoor held a dried palm leaf in his hand.
When they were seated and the first interchange of amenities
was over, Ulrik Frederik—they were speaking in French—asked
whether Burrhi would not with his learning and experience come to
the aid of the searchers after wisdom in the land of Denmark.
Burrhi shook his head.
“’Tis known to me,” he replied, “that the secret art has many
great and powerful votaries in Denmark, but I have imparted
instruction to so many royal gentlemen and church dignitaries, and
while I will not say that ingratitude or meagre appreciation have
always been my appointed portion, yet have I encountered so much
captiousness and lack of understanding, that I am unwilling to
assume again the duties of a master to such distinguished scholars.
I do not know what rule or method the King of Denmark employs in
his investigations, and my remarks can therefore contain no
disparagement of him, but I can assure you in confidence that I
have known gentlemen of the highest nobility in the land, nay,
anointed rulers and hereditary kings, who have been so ignorant of
their historia naturalis and materia magica that the most lowborn
quacksalver could not entertain such vulgar superstitions as they do.
They even put their faith in that widely disseminated though
shameful delusion that making gold is like concocting a sleeping-
potion or a healing-pillula, that if one has the correct ingredients, ’tis
but to mix them together, set them over the fire, and lo! the gold is
there. Such lies are circulated by catch-pennies and ignoramuses—
whom may the devil take! Cannot the fools understand that if ’twere
so simple a process, the world would be swimming in gold? For
although learned authors have held, and surely with reason, that
only a certain part of matter can be clarified in the form of gold, yet
even so we should be flooded. Nay, the art of the gold-maker is
costly and exacting. It requires a fortunate hand, and there must be
Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade

Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.

Let us accompany you on the journey of exploring knowledge and

personal growth!

ebooknice.com