© 2009 by Taylor & Francis Group, LLC
Strategic Planning
Models for Reverse
and ClosedLoop
Supply Chains
© 2009 by Taylor & Francis Group, LLC
Strategic Planning
Models for Reverse
and ClosedLoop
Supply Chains
CRC Press is an imprint of the
Taylor & Francis Group, an informa business
Boca Raton London New York
© 2009 by Taylor & Francis Group, LLC
CRC Press
Taylor & Francis Group
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 334872742
© 2009 by Taylor & Francis Group, LLC
CRC Press is an imprint of Taylor & Francis Group, an Informa business
No claim to original U.S. Government works
Printed in the United States of America on acidfree paper
10 9 8 7 6 5 4 3 2 1
International Standard Book Number13: 9781420054781 (Hardcover)
This book contains information obtained from authentic and highly regarded sources. Reasonable
efforts have been made to publish reliable data and information, but the author and publisher can
not assume responsibility for the validity of all materials or the consequences of their use. The
authors and publishers have attempted to trace the copyright holders of all material reproduced
in this publication and apologize to copyright holders if permission to publish in this form has not
been obtained. If any copyright material has not been acknowledged please write and let us know so
we may rectify in any future reprint.
Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced,
transmitted, or utilized in any form by any electronic, mechanical, or other means, now known or
hereafter invented, including photocopying, microfilming, and recording, or in any information
storage or retrieval system, without written permission from the publishers.
For permission to photocopy or use material electronically from this work, please access www.copy
right.com (http://www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC), 222
Rosewood Drive, Danvers, MA 01923, 9787508400. CCC is a notforprofit organization that pro
vides licenses and registration for a variety of users. For organizations that have been granted a
photocopy license by the CCC, a separate system of payment has been arranged.
Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and
are used only for identification and explanation without intent to infringe.
Library of Congress CataloginginPublication Data
Pochampally, Kishore K.
Strategic planning models for reverse and closedloop supply chains / Kishore
K. Pochampally, Satish Nukala, Surendra M. Gupta.
p. cm.
Includes bibliographical references and index.
ISBN 9781420054781 (hardback : alk. paper)
1. Manufacturing processesEnvironmental aspects. 2. Business
logisticsEnvironmental aspects. 3. Remanufacturing. 4. Waste minimization.
5. Salvage (Waste, etc.) 6. Recycling (Waste, etc.) I. Nukala, Satish. II. Gupta,
Surendra M. III. Title.
TS155.7.P63 2008
658.5’67dc22 2008014566
Visit the Taylor & Francis Web site at
http://www.taylorandfrancis.com
and the CRC Press Web site at
http://www.crcpress.com
© 2009 by Taylor & Francis Group, LLC
Dedication
To our families:
Seema Dasgupta, Sharada Pochampally, and Narasimha Rao Pochampally
—KKP
Somayajulu Nukala, Jagadeeswari Nukala, Sridhar Nukala,
Mythili Nukala, and Padmaja Nukala
—SN
Sharda Gupta, Monica Gupta, and Neil Gupta
—SMG
© 2009 by Taylor & Francis Group, LLC
vii
Contents
Preface ....................................................................................................... xiii
Acknowledgments......................................................................................xv
About the Authors................................................................................... xvii
1 Introduction......................................................................................... 1
1.1 Motivation............................................................................................ 1
1.2 Overview of the Book......................................................................... 5
1.3 Outline of the Book............................................................................. 7
1.4 Conclusions.......................................................................................... 9
References....................................................................................................... 9
2 Strategic Planning of Reverse and ClosedLoop Supply
Chains ................................................................................................ 11
2.1 Introduction....................................................................................... 11
2.2 Selection of Used Products .............................................................. 12
2.3 Evaluation of Collection Centers .................................................... 12
2.4 Evaluation of Recovery Facilities .................................................... 13
2.5 Optimization of Transportation of Goods .................................... 13
2.6 Evaluation of Marketing Strategies ................................................ 14
2.7 Evaluation of Production Facilities................................................. 14
2.8 Evaluation of Futurity of Used Products....................................... 15
2.9 Selection of New Products............................................................... 15
2.10 Selection of Secondhand Markets .................................................. 16
2.11 Synchronization of Supply Chain Processes ................................ 16
2.12 Supply Chain Performance Measurement .................................... 16
2.13 Conclusions........................................................................................ 17
References..................................................................................................... 17
3 Literature Review............................................................................. 19
3.1 Introduction....................................................................................... 19
3.2 Operational Planning of Reverse and ClosedLoop Supply
Chains................................................................................................. 19
3.3 Strategic and Tactical Planning of Reverse and Closed
Loop Supply Chains ......................................................................... 24
3.4 Conclusions........................................................................................ 31
References..................................................................................................... 31
4 Quantitative Modeling Techniques ............................................... 37
4.1 Introduction....................................................................................... 37
4.2 Analytic Hierarchy Process and Eigen Vector Method............... 37
4.3 Analytic Network Process ............................................................... 39
© 2009 by Taylor & Francis Group, LLC
viii Contents
4.4 Fuzzy Logic........................................................................................ 40
4.5 Extent Analysis Method................................................................... 43
4.6 Fuzzy Multicriteria Analysis Method............................................ 44
4.7 Quality Function Deployment ........................................................ 48
4.8 Method of Total Preferences............................................................ 49
4.9 Linear Physical Programming........................................................ 49
4.10 Goal Programming........................................................................... 52
4.11 Technique for Order Preference by Similarity to Ideal
Solution (TOPSIS).............................................................................. 55
4.12 Borda’s Choice Rule .......................................................................... 58
4.13 Expert Systems .................................................................................. 58
4.14 Bayesian Updating............................................................................ 59
4.15 Taguchi Loss Function ..................................................................... 61
4.16 Six Sigma............................................................................................ 63
4.16.1 Process Capability Ratio (C
p
) .............................................. 64
4.16.2 Process Capability Index (C
pk
)............................................ 64
4.16.2.1 Three Sigma Process........................................... 65
4.16.2.2 4.5 Sigma Process ................................................ 66
4.16.2.3 Six Sigma Process................................................ 66
4.17 Neural Networks............................................................................... 67
4.18 Geographical Information Systems................................................ 68
4.19 Linear Integer Programming .......................................................... 69
4.20 Conclusions........................................................................................ 69
References..................................................................................................... 69
5 Selection of Used Products.............................................................. 73
5.1 The Issue............................................................................................. 73
5.2 First Model (Linear Integer Programming) .................................. 73
5.2.1 Nomenclature ........................................................................ 74
5.2.2 Model Formulation ............................................................... 74
5.2.2.1 Modifed CostBeneft Function.......................... 75
5.2.2.2 Linear Integer Programming Model ................. 76
5.2.3 Numerical Example .............................................................. 77
5.3 Second Model (Linear Physical Programming) ........................... 78
5.3.1 Model Formulation................................................................ 78
5.3.1.1 Class 1S Criteria (Smaller Is Better) .................... 78
5.3.1.2 Class 2S Criteria (Larger Is Better)...................... 79
5.3.2 Numerical Example .............................................................. 80
5.4 Conclusions........................................................................................ 80
References..................................................................................................... 85
6 Evaluation of Collection Centers.................................................... 87
6.1 The Issue............................................................................................. 87
6.2 First Model (Eigen Vector Method and Taguchi Loss Function) ..... 88
6.2.1 Evaluation Criteria ................................................................ 88
6.2.2 Model ...................................................................................... 89
© 2009 by Taylor & Francis Group, LLC
Contents ix
6.2.2.1 n Value.................................................................... 90
6.2.2.2 Distance from Residential Area (DH) ............... 91
6.2.2.3 Distance from Roads (DR) .................................. 91
6.2.2.4 Utilization of Incentives from Local
Government (UI)................................................. 91
6.2.2.5 Per Capita Income of People in Residential
Area (PI) ............................................................... 91
6.2.2.6 Space Cost (SC)...................................................... 92
6.2.2.7 Labor Cost (LC) ..................................................... 92
6.2.2.8 Incentives from Local Government (IG)............ 92
6.3 Evaluation Criteria for Second and Third Models....................... 93
6.3.1 Criteria of Consumers........................................................... 93
6.3.2 Criteria of Local Government Offcials.............................. 94
6.3.3 Criteria of Supply Chain Company Executives ................ 94
6.4 Second Model (Eigen Vector Method, TOPSIS, and Borda’s
Choice Rule) ....................................................................................... 95
6.4.1 Phase I (Individual Decision Making) ............................... 95
6.4.2 Phase II (Group Decision Making) ................................... 101
6.5 Third Model (Neural Networks, Fuzzy Logic, TOPSIS,
Borda’s Rule) .................................................................................... 103
6.5.1 Phase I (Derivation of Impacts) ......................................... 103
6.5.2 Phase II (Individual Decision Making)............................ 106
6.5.3 Phase III (Group Decision Making).................................. 109
6.6 Fourth Model (ANP and Goal Programming) ........................... 110
6.6.1 Application of ANP............................................................. 110
6.6.2 Application of Goal Programming................................... 116
6.6.2.1 Nomenclature for Problem Formulation......... 116
6.6.2.2 Problem Formulation ......................................... 116
6.7 Fifth Model (Eigen Vector Method, Taguchi Loss Function,
and Goal Programming)................................................................ 118
6.7.1 Application of Eigen Vector Method and Taguchi
Loss Function ....................................................................... 118
6.7.2 Application of Goal Programming ................................... 121
6.7.2.1 Nomenclature Used in the Methodology......... 122
6.7.2.2 Problem Formulation.......................................... 122
6.8 Conclusions...................................................................................... 124
References................................................................................................... 124
7 Evaluation of Recovery Facilities ................................................. 125
7.1 The Issue........................................................................................... 125
7.2 First Model (Analytic Hierarchy Process) ................................... 126
7.2.1 ThreeLevel Hierarchy........................................................ 126
7.2.2 Numerical Example............................................................. 128
7.3 Second Model (Linear Physical Programming) ......................... 130
7.3.1 Nomenclature for LPP Model ............................................ 130
© 2009 by Taylor & Francis Group, LLC
x Contents
7.3.2 Criteria for Identifcation of Effcient Recovery
Facilities................................................................................. 131
7.3.2.1 Class 1S Criteria (Smaller is Better) .................. 131
7.3.2.2 Class 2S Criteria (Larger Is Better).................... 131
7.3.3 Numerical Example............................................................. 132
7.4 Evaluation Criteria for Third and Fourth Models...................... 132
7.4.1 Criteria of Consumers ......................................................... 134
7.4.2 Criteria of Local Government Offcials ............................ 134
7.4.3 Criteria of Supply Chain Company Executives............... 135
7.5 Third Model (Eigen Vector Method, TOPSIS, and Borda’s
Choice Rule) ..................................................................................... 135
7.5.1 Phase I (Individual Decision Making).............................. 135
7.5.2 Phase II (Group Decision Making).................................... 140
7.6 Fourth Model (Neural Networks, Fuzzy Logic, TOPSIS,
Borda’s Choice Rule) ....................................................................... 140
7.6.1 Phase I (Derivation of Impacts).......................................... 141
7.6.2 Phase II (Individual Decision Making) ............................ 145
7.6.3 Phase III (Group Decision Making) .................................. 147
7.7 Fifth Model (TwoDimensional Chart) ........................................ 148
7.8 Conclusions...................................................................................... 151
References................................................................................................... 151
8 Optimization of Transportation of Products.............................. 153
8.1 The Issue........................................................................................... 153
8.2 First Model (Linear Integer Programming) ................................ 154
8.2.1 Nomenclature ...................................................................... 154
8.2.2 Model Formulation............................................................. 155
8.2.3 Numerical Example ............................................................ 157
8.3 Second Model (Linear Physical Programming) ......................... 158
8.3.1 Model Formulation.............................................................. 158
8.3.2 Numerical Example ............................................................ 160
8.4 Third Model (Goal Programming)............................................... 161
8.4.1 Nomenclature....................................................................... 161
8.4.2 Model Formulation ............................................................. 162
8.4.3 Numerical Example ............................................................ 166
8.5 Fourth Model (Linear Physical Programming) .......................... 168
8.5.1 Model Formulation .............................................................. 168
8.5.2 Numerical Example ............................................................ 171
8.6 Fifth Model (Fuzzy Goal Programming) .................................... 173
8.6.1 Model Formulation.............................................................. 173
8.6.2 Numerical Example ............................................................ 178
8.7 Conclusions...................................................................................... 179
References................................................................................................... 179
9 Evaluation of Marketing Strategies ............................................. 181
9.1 The Issue........................................................................................... 181
© 2009 by Taylor & Francis Group, LLC
Contents xi
9.2 First Model (Fuzzy Logic and TOPSIS)........................................ 182
9.2.1 Drivers of Public Participation........................................... 182
9.2.2 Methodology........................................................................ 183
9.3 Second Model (Fuzzy Logic, Quality Function
Deployment, and Method of Total Preferences) ......................... 188
9.3.1 Performance Aspects and Enablers .................................. 188
9.3.2 Numerical Example............................................................. 190
9.4 Third Model (Fuzzy Logic, Extent Analysis Method, and
Analytic Network Process) ............................................................ 192
9.4.1 Main Criteria and Subcriteria............................................ 193
9.4.2 Numerical Example............................................................. 193
9.5 Conclusions...................................................................................... 198
References................................................................................................... 199
10 Evaluation of Production Facilities .............................................. 201
10.1 The Issue........................................................................................... 201
10.2 First Model (Fuzzy Logic and TOPSIS) ....................................... 202
10.2.1 Evaluation Criteria............................................................. 203
10.2.1.1 Environmentally Conscious Design (ECD) ... 203
10.2.1.2 Environmentally Conscious
Manufacturing (ECM) ...................................... 203
10.2.1.3 Attitude of Management (AMT) ..................... 204
10.2.1.4 Potentiality (POT).............................................. 204
10.2.1.5 Cost (COS) .......................................................... 204
10.2.1.6 Customer Service (CSE).................................... 204
10.2.2 Numerical Example........................................................... 205
10.3 Second Model (Fuzzy Logic, Extent Analysis Method, and
Analytic Network Process) ............................................................ 212
10.4 Third Model (Fuzzy Multicriteria Analysis Method) ............... 215
10.5 Conclusions...................................................................................... 226
References................................................................................................... 226
11 Evaluation of Futurity of Used Products..................................... 227
11.1 The Issue........................................................................................... 227
11.2 Usage of Fuzzy Logic ..................................................................... 229
11.3 Rules Used in Bayesian Updating................................................ 230
11.4 Bayesian Updating.......................................................................... 231
11.5 FLEXBased Expert System........................................................... 232
11.6 Conclusions...................................................................................... 232
References................................................................................................... 233
12 Selection of New Products ............................................................ 235
12.1 The Issue .......................................................................................... 235
12.2 Assumptions ................................................................................... 236
12.3 Nomenclature.................................................................................. 236
12.4 Formulation of Fuzzy CostBeneft Function.............................. 238
© 2009 by Taylor & Francis Group, LLC
xii Contents
12.4.1 Total New Product Sale Revenue per Period (SR)......... 238
12.4.2 Total Reuse Revenue per Period (UR) ............................ 238
12.4.3 Total Recycle Revenue per Period (CR) .......................... 239
12.4.4 Total New Product Production Cost per Period (MC) . 239
12.4.5 Total Collection Cost per Period (CC) ............................ 239
12.4.6 Total Reprocessing Cost per Period (RC) ....................... 239
12.4.7 Total Disposal Cost per Period (DC)............................... 240
12.4.8 LossofSale Cost per Period (LC) ................................... 240
12.4.9 Investment Cost (IC).......................................................... 240
12.5 Model................................................................................................ 241
12.6 Numerical Example........................................................................ 241
12.7 Conclusions ..................................................................................... 243
References................................................................................................... 244
13 Selection of Secondhand Markets................................................ 245
13.1 The Issue........................................................................................... 245
13.2 Performance Aspects and Enablers for Application of QFD.... 245
13.3 Selection of Potential Secondhand Markets ............................... 246
13.4 Conclusions...................................................................................... 250
14 Design of a Synchronized Reverse Supply Chain..................... 251
14.1 The Issue........................................................................................... 251
14.2 Model (Two Design Experiments) ................................................ 251
14.2.1 First Experiment (Determination of Nominal Pool)..... 251
14.2.2 Second Experiment (Determination of Variance Pool)... 253
14.3 Conclusions...................................................................................... 254
References................................................................................................... 255
15 Performance Measurement............................................................ 257
15.1 The Issue........................................................................................... 257
15.2 Application of LPP to QFD Optimization................................... 258
15.2.1 First Step ............................................................................. 258
15.2.2 Second Step ........................................................................ 260
15.3 Reverse/ClosedLoop Supply Chain Performance
Measurement ................................................................................... 261
15.3.1 Performance Aspects and Enablers ................................ 261
15.3.2 Numerical Example........................................................... 263
15.4 Conclusions...................................................................................... 269
References................................................................................................... 269
16 Conclusions ..................................................................................... 271
© 2009 by Taylor & Francis Group, LLC
xiii
Preface
The rapid technological development of new products, along with the grow
ing desire of consumers to acquire the latest technology, has led to a new
environmental problem: products that are discarded prematurely. But behind
every problem lies an opportunity. In this case, the opportunity comes from
reprocessing (processing used products), which leads to:
1. Saving natural resources: We conserve land and reduce the need to drill
for oil and dig for minerals by making products using materials and
components obtained from reprocessing rather than virgin resources.
2. Saving energy: It usually takes less energy to make products from
reprocessed materials and components than from virgin resources.
3. Saving clean air and water: Making products from reprocessed materi
als and components creates less air and water pollution than from
virgin resources.
4. Saving landfll space: When reprocessed materials and components
are used to make a product, they do not go into landflls.
5. Saving money: It costs much less to make products from reprocessed
materials and components than from virgin resources.
Besides the above opportunities, an important driver for companies to
engage in reprocessing is the enforcement of environmental regulations by
local governments.
A reverse supply chain consists of a series of activities required to collect
used products from consumers and reprocess them to either recover their
leftover market values or properly dispose of them. Today, in practice, it has
become common for companies involved in a forward supply chain (series of
activities required to produce new products from virgin resources and dis
tribute them to consumers) to also carry out collection and reprocessing of
used products. This combined practice of forward and reverse supply chains
is called a closedloop supply chain. In the past decade, there has been an
explosive growth of reverse and closedloop supply chains, in both scope
and scale.
Strategic planning (also called designing) primarily involves the structur
ing (which products should be processed/produced in which facilities) of
a supply chain over the next several years. It is longrange planning and is
typically performed every few years when a supply chain needs to expand
its capabilities. The issues faced by strategic planners of reverse and closed
loop supply chains are evaluation and selection of new and used products,
collection centers, recovery facilities, marketing strategies, and production
facilities, as well as evaluation of the futurity of used products, selection of
© 2009 by Taylor & Francis Group, LLC
xiv Preface
secondhand markets, optimization of transportation of goods, synchroniza
tion of supply chain processes, and supply chain performance measurement.
In all of these issues, strategic planners must meet the following challenges:
uncertainty in supply rate of used products, unknown condition of used
products, and imperfect correlation between supply of used products and
demand for reprocessed goods.
This book addresses the above issues amidst the above challenges in a
variety of decisionmaking situations using effcient models. These mod
els implement several quantitative techniques, such as analytic hierarchy
process, eigen vector method, analytic network process, fuzzy logic, extent
analysis method, fuzzy multicriteria analysis method, quality function
deployment, method of total preferences, linear physical programming, goal
programming, technique for order preference by similarity to ideal solution
(TOPSIS), Borda’s choice rule, expert systems, Bayesian updating, Taguchi
loss function, Six Sigma, neural networks, geographical information sys
tems, and linear integer programming.
The issues addressed in this book can serve as foundations for other
researchers to build bodies of knowledge in this new and fastgrowing feld
of research, viz., strategic planning of reverse and closedloop supply chains.
Furthermore, the models proposed in this book for those issues can be utilized
by industrialists for understanding how a particular issue in the strategic plan
ning of reverse and closedloop supply chains can be effectively approached
in a particular decisionmaking situation, using a suitable quantitative tech
nique or a suitable combination of two or more quantitative techniques.
The strategic planning of reverse and closedloop supply chains remains
an important and promising feld of research. This is desirable from both
environmental and economical points of view, because by using innovations
in this feld, environmentalists as well as industrialists can fully exploit their
desired goals. The authors express their hope that this book will inspire
further research and motivate new and rewarding research in this all too
important feld of study.
© 2009 by Taylor & Francis Group, LLC
xv
Acknowledgments
Encouragement for writing this book came from many sources. We thank
hundreds of researchers from whose work we have benefted and many of
whom we have had the good fortune of meeting and interacting with at con
ferences around the world. Many thanks to Dr. Sagar V. Kamarthi (North
eastern University), Dr. Elif Kongar (University of Bridgeport), Dr. Martin
Bradley (Southern New Hampshire University), Dr. Tej S. Dhakar (Southern
New Hampshire University), and Dr. Seamus M. McGovern (National Trans
portation Systems Center).
This book would not have been possible without the commitment of Taylor
& Francis for encouraging innovative ideas. We express our appreciation to
Taylor & Francis and its staff for providing seamless support in making it
possible to complete this timely and important book.
Most importantly, we are indebted to our families, to whom this book is
lovingly dedicated, for constantly providing us their unconditional support
in making this book a reality.
Kishore K. Pochampally, PhD
Manchester, New Hampshire
Satish Nukala, PhD
Houston, Texas
Surendra M. Gupta, PhD
Boston, Massachusetts
© 2009 by Taylor & Francis Group, LLC
xvii
About the Authors
Dr. Kishore K. Pochampally is an assistant professor of quantitative stud
ies and operations management at Southern New Hampshire University
in Manchester. His prior academic experience is as a postdoctoral fellow
at Massachusetts Institute of Technology (MIT) in Cambridge. He holds a
PhD in industrial engineering from Northeastern University in Boston. His
research interests are in the areas of reverse logistics, supply chain design,
and Six Sigma quality management. He is a Six Sigma Green Belt and a Proj
ect Management Professional (PMP
®
).
Dr. Satish Nukala is a senior supply chain analyst with Halliburton Energy
Services in Houston, Texas. He holds a PhD in industrial engineering from
Northeastern University in Boston and an MS in industrial and systems
engineering from the University of Alabama in Huntsville. He is a Six Sigma
Green Belt. His research interests are in the areas of production and opera
tions management and total quality management.
Dr. Surendra M. Gupta, PE, is a professor of mechanical and industrial
engineering and director of the Laboratory for Responsible Manufactur
ing at Northeastern University in Boston. He received his BE in electron
ics engineering from Birla Institute of Technology and Science, MBA from
Bryant University, and MSIE and PhD in industrial engineering from Pur
due University. Dr. Gupta’s research interests are in the areas of production/
manufacturing systems and operations research. He is mostly interested in
environmentally conscious manufacturing, electronics manufacturing, MRP,
JIT, and queueing theory. He has authored and coauthored more than 350
technical papers published in prestigious journals, books, and conference
proceedings. His publications have been cited by thousands of researchers
all over the world in journals, proceedings, books, and dissertations. He has
traveled to all seven continents and presented his work at international con
ferences there (except Antarctica). He is currently serving as the area editor
of “Environmental Issues” for Computers and Industrial Engineering, associate
editor for International Journal of Agile Systems and Management, and editorial
board member of a variety of journals. He has also served as conference chair,
track chair, and member of technical committees of a variety of international
conferences. Dr. Gupta has been elected to the memberships of several honor
societies and is listed in various Who’s Who publications. He is a registered
professional engineer in the state of Massachusetts and a member of ASEE,
DSI, IIE, INFORMS, and POMS. Dr. Gupta is a recipient of the Outstand
ing Research Award and the Outstanding Industrial Engineering Professor
Award (in recognition of teaching excellence) from Northeastern University.
His recent activities can be viewed at http://www1.coe.neu.edu/~smgupta/,
and he can be reached by email at gupta@neu.edu.
© 2009 by Taylor & Francis Group, LLC
1
1
Introduction
1.1 Motivation
The level of consumption by a growing population in this world of fnite
resources and disposal capacities has been continuously increasing. This
is fueled by the growing desire of consumers to acquire the latest technol
ogy, both at home and in the workplace, along with the rapid technological
development of new products. As a result, the world now faces a serious
environmental problem: waste, viz., used products that are discarded pre
maturely. For example, an estimated 60 million computers enter the mar
ket in the United States every year and more than 12 million computers are
discarded every year, out of which fewer than 10% are reprocessed,* while
the rest head to landflls [1]. According to the Environmental Protection
Agency’s Municipal Solid Waste Fact Book, 29 states in the United States have
10 or more years of landfll space left, 15 states between 5 and 10 years, and
6 states fewer than 5 years [2]. Increased consumption results in increased
use of raw material and energy, thereby depleting the world’s fnite natural
resources. The data [3, 4] presented in tables 1.1 and 1.2 give an idea about
the materials and energy consumed in the manufacturing of microchips and
LCD monitors, respectively. Apart from the excessive consumption of the
world’s fnite natural resources and disposal capacities, the presence of toxic
material, such as lead, polybrominated diphenyl ether, mercury, and hexava
lent chromium, in the discarded electronic equipment poses a serious threat
to the environment. This environmental degradation is not sustainable by
Earth’s ecosystem [6].
Reprocessing of used products helps in:
*
Though direct reuse of the used products is infeasible in most cases, remanufacturing and
recycling are the major reprocessing options applied in the industry. Remanufacturing is a
process in which used products are restored to likenew conditions, and recycling is a pro
cess performed to retrieve the material content of used products without retaining the iden
tity of their components [5].
© 2009 by Taylor & Francis Group, LLC
2 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
1. Saving natural resources: We conserve land and reduce the need to drill
for oil and dig for minerals by making products using materials and
components obtained from reprocessing instead of virgin materials.
2. Saving energy: It usually takes less energy to make products from
reprocessed materials and components than from virgin materials.
3. Saving clean air and water: Making products from reprocessed mate
rials and components creates less air pollution and water pollution
than from virgin materials.
4. Saving landfll space: When reprocessed materials and components
are used to make a product, they do not go into landflls.
5. Saving money: It costs much less to make products from reprocessed
materials and components than from virgin materials.
TABLE 1.1
Materials and energy used in manufacturing microchips
Material Description
Amount per
memory chip
Annual use by
industry
worldwide
Amount used to
make chips in
one computer
Silicon wafer 0.25 g 4,400 tons 0.025 kg
Chemicals Dopants 0.016 g 280 tons 0.002 kg
Photolithography 22 g 390,000 tons 2.2 kg
Etchants 0.37 g 6,600 tons 0.037 kg
Acids/bases 50 g 890,000 tons 4.9 kg
Total chemicals 72 g 1.3 million tons 7.1 kg
Elemental gases N
2
, O
2
, H
2
, He, Ar 700 g 12 million tons 69 kg
Energy Electricity 2.9 kWh 52 billion kWh 281 kWh
Direct fossil fuels 1.6 MJ 28 billion MJ 155 MJ
Embodied fossil
fuels
970 g 17 million tons 94 kg
Water 32 liters 570 billion
liters
310 liters
TABLE 1.2
Aggregate chemicals, energy, and water use in manufacture of LCD monitor
Material/input Amount used per monitor
Photolithographic and other chemicals 3.7 kg
Elemental gases (N
2
, O
2
, Ar) 5.9 kg
Electricity 87 kWh
Direct fossil fuels (98% natural gas) 198 kg
Embodied fossil fuels 226 kg
Water 1,290 liters
© 2009 by Taylor & Francis Group, LLC
Introduction 3
As waste reduction is becoming a major concern in industrialized coun
tries, the concept of reprocessing is gradually replacing a oneway perception
of economy [7]. Increasingly, customers are expecting companies to minimize
the environmental impact of their products and processes. Moreover, legisla
tion extending producers’ responsibility has become an important element of
public environmental policy. Several countries, particularly in the European
Union, have introduced environmental legislation charging manufacturers
with responsibility for the whole life cycle of their products. The obligations
of collection of used products and their reprocessing have been enacted or are
under way for a number of product categories, including electronic equipment
in the European Union and Japan, cars in the European Union and Taiwan,
and packaging material in Germany. Also, companies are recognizing oppor
tunities for combining environmental stewardship with proftability, brought
about by production cost savings and access to new market segments. In this
vein, in the past decade, there has been an explosive growth of reprocessing
activities in both scope and scale. However, in companies in the United States,
reprocessing is still in its infancy. In the United States, cities and towns are
responsible for retrieval of used products and proper disposal of the poten
tially environmentally dangerous and waste components. According to a
2003 report [8], in the state of Massachusetts, support is building for a refled
bill that would require manufacturers of electronic products to pay for col
lection and reprocessing of those products after the end of their usage by the
consumers. If passed, the statewide takeback program would be the frst of
its kind in the nation and would relieve cities and towns, which are bracing
for local aid cuts, from the costs associated with collecting and disposing of
electronic waste. The bill’s supporters say that cities and towns in the United
States spend between $6 million and $21 million a year on such endeavors.
A reverse supply chain (also known as reverse logistics) consists of a series of
activities required to collect a used product from a consumer and reprocess
it (used product) to either recover its leftover market value or dispose of it.
Implementation of any reverse supply chain requires at least three parties:
collection centers where consumers return used products, recovery facilities
where reprocessing (remanufacturing or recycling) is performed, and demand
centers where customers buy reprocessed products, viz., outgoing goods from
recovery facilities. Figure 1.1 shows a generic reverse supply chain.
Environmental consciousness suggests the production of new products from
conceptual design to fnal delivery such that the environmental standards
and requirements are satisfed. In the last decade, environmental conscious
ness has become an obligation to many facilities in a traditional/forward
supply chain (i.e., series of activities required to produce new products from
raw material and distribute the former to customers), enforced primarily
by governmental regulations and customer perspective on environmental
issues [9, 10]. At the same time, many of these facilities are driven, mainly by
proftability, to administer the reverse supply chain as well. The combination
of forward and reverse supply chains is called a closedloop supply chain. A
generic closedloop supply chain is shown in fgure 1.2.
© 2009 by Taylor & Francis Group, LLC
4 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
Reverse/closedloop supply chains are implemented in many industries,
including commercial aircraft, automobiles, computers, chemicals, appli
ances, and apparel [11].
A supply chain involves three phases of decision making: supply chain
design, supply chain planning, and supply chain operation [12]:
Supply chain design r : This phase is also called strategic planning. Dur
ing this phase, a company decides how to structure its supply chain
over the next several years. Decisions made by companies include
the location and capacity of production and warehouse facilities, the
products to be manufactured and stored at various locations, the
modes of transportation to be made available, and the type of infor
mation system to be utilized. Supply chain design decisions are typi
cally made for the long term (years) and are very expensive to alter
on short notice. Consequently, when companies make these deci
sions, they must take into account uncertainty in anticipated market
conditions over the next few years.
Collection
Centers
Recovery
Facilities
Consumers
Used Products Used Products
Demand
Centers
Recycled
Goods
Remanufactured
Products
FIGURE 1.1
Generic reverse supply chain.
Collection
Centers
Production
Facilities
Demand
Centers
New Products
Remanufactured Products
Consumers
New Products
Recycled Goods
Remanufactured Products
Recycled
Goods
Used Products
Raw Material
Suppliers
Used Products
FIGURE 1.2
Generic closedloop supply chain.
© 2009 by Taylor & Francis Group, LLC
Introduction 5
Supply chain planning r : This phase is also called tactical planning. Sup
ply chain planning includes decisions regarding subcontracting of
manufacturing, inventory policies to be followed, and timing and
size of marketing promotions. For decisions made during this phase,
the time frame considered is 3 months to a year. Companies must
include uncertainty in demand, exchange rates, and competition
over the time frame. Given a shorter time frame and better forecasts
than the design phase, companies in this phase try to incorporate
any fexibility built into the supply chain in the design phase and
exploit it to optimize performance.
Supply chain operation r : This phase is also called operational planning.
During this phase, companies make decisions regarding individual
customer orders. The goal of supply chain operations is to handle
incoming customer orders in the best possible manner. Decisions
include allocating inventory or production to individual orders,
setting a date for an order to be flled, generating pick lists at a
warehouse, setting delivery schedules of trucks, and placing replen
ishment orders. Because operational decisions are made in the
short term (minutes, hours, or days), there is less uncertainty about
demand information.
This book focuses on strategic planning of reverse and closedloop supply
chains. The various decisionmaking problems faced by strategic planners of
reverse and closedloop supply chains include selection of used products to
reprocess, evaluation of collection centers, evaluation of recovery facilities, selec
tion of new products to produce, and optimization of transportation of goods.
Reverse and closedloop supply chains differ from traditional/forward
supply chains in many aspects and are complex to handle because of the
inherent uncertainty involved in every stage of strategic planning (see
chain cannot be adopted for a reverse or closedloop supply chain.
This chapter is organized as follows: section 1.2 presents the overview of
this book, section 1.3 gives the outline of the book, and section 1.4 gives
some conclusions.
1.2 Overview of the Book
This book presents quantitative models for various issues faced by strategic
planners of reverse and closedloop supply chains amidst many challenges,
such as uncertainty in supply rate of used products, unknown condition of
used products, and imperfect correlation between supply of used products
and demand for reprocessed goods.
table 1.3 [13]). As a result, strategic planning models for a forward supply
© 2009 by Taylor & Francis Group, LLC
6 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
The complete list of issues addressed in this book is the following (see
chapter 2 for further explanation of each of the issues):
Selection of used products r
Evaluation of collection centers r
Evaluation of recovery facilities r
Optimization of transportation of goods r
Evaluation of marketing strategies r
Evaluation of production facilities r
Evaluation of futurity of used products r
Selection of new products r
Selection of secondhand markets r
Synchronization of supply chain processes r
Supply chain performance measurement r
The models for addressing the above strategic planning issues employ sev
eral quantitative techniques, such as:
Analytic hierarchy process r
Eigen vector method r
Analytic network process r
Fuzzy logic r
TABLE 1.3
Comparison between forward and reverse/closedloop supply chains
Forward supply chain Reverse/closedloop supply chain
Product quality uniform Product quality not uniform
Disposition options clear Disposition options unclear
Routing of products unambiguous Routing of products ambiguous
Costs involved easily understood Costs involved not easily understood
Product pricing uniform Product pricing not uniform
Inventory management consistent Inventory management inconsistent
Product life cycle manageable Product life cycle less manageable
Financial management issues clear Financial management issues unclear
Negotiations between parties straightforward Negotiations less straightforward
Customer easily identifable to market Customer less easily identifable to market
Forecasting relatively straight forward Forecasting more diffcult
Onetomany transportation Manytoone transportation
Marketing methods well known Marketing complication by several factors
Process visibility more transparent Process visibility less transparent
© 2009 by Taylor & Francis Group, LLC
Introduction 7
Extent analysis method r
Fuzzy multicriteria analysis method r
Quality function deployment r
Method of total preferences r
Linear physical programming r
Goal programming r
Technique for order preference by similarity to ideal solution (TOPSIS) r
Borda’s choice rule r
Expert systems r
Bayesian updating r
Taguchi loss function r
Six Sigma r
Neural networks r
Geographical information systems r
Linear integer programming r
The issues addressed in this book can serve as foundations for other
researchers to build bodies of knowledge in this new and fastgrowing feld of
research, i.e., strategic planning of reverse and closedloop supply chains. Fur
thermore, the models proposed in this book for those issues can be utilized by
industrialists for understanding how a particular issue in the strategic plan
ning of reverse and closedloop supply chains can be effectively approached
in a particular decisionmaking situation, using a suitable quantitative tech
nique or a suitable combination of two or more quantitative techniques.
1.3 Outline of the Book
This book is organized as follows:
In chapter 2, an overview of the strategic planning issues (see section 1.2)
is presented, and it is explained how the decisionmaking situations could
differ for each of the issues.
In chapter 3, a brief review of literature in the area of strategic planning of
reverse and closedloop supply chains is presented.
In chapter 4, each of the several quantitative techniques that are used in
the strategic planning models presented in this book is briefy described.
In chapter 5, two models for selection of used products in two different deci
sionmaking situations are presented. The frst model employs linear integer
programming, and the second model employs linear physical programming.
© 2009 by Taylor & Francis Group, LLC
8 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
In chapter 6, fve models for evaluation of collection centers in fve differ
ent decisionmaking situations are presented. The frst model employs eigen
vector method and Taguchi loss function. The second model uses eigen vec
tor method, technique for order preference by similarity to ideal solution
(TOPSIS), and Borda’s choice rule. The third model uses neural networks,
fuzzy logic, TOPSIS, and Borda’s choice rule. The fourth model uses analytic
network process (ANP) and goal programming. The ffth model uses eigen
vector method, Taguchi loss function, and goal programming.
In chapter 7, fve models for evaluation of recovery facilities in fve differ
ent decisionmaking situations are presented. The frst model uses analytic
hierarchy process. The second model employs linear physical programming.
The third model uses eigen vector method, technique for order preference
by similarity to ideal solution (TOPSIS), and Borda’s choice rule. The fourth
model uses neural networks, fuzzy logic, TOPSIS, and Borda’s choice rule.
The ffth model uses a simple twodimensional chart.
Chapter 8 focuses on achieving transportation of the right quantities of prod
ucts (used, remanufactured, and new) across a reverse or closedloop supply
chain, while satisfying certain constraints. Five models are presented in fve
different decisionmaking situations. The frst model employs linear integer
programming. The second model employs linear physical programming. The
third and fourth models use goal programming and linear physical program
ming, respectively. The ffth model employs fuzzy goal programming.
In chapter 9, three models for evaluation of the marketing strategy of a
reverse/closedloop supply chain in three different decisionmaking situa
tions are presented. The frst model employs fuzzy logic and technique for
order preference by similarity to ideal solution (TOPSIS). The second model
employs fuzzy logic, quality function deployment (QFD), and method of
total preferences. The third model uses fuzzy logic, extent analysis method,
and analytic network process.
In chapter 10, three models for evaluation of production facilities in fve
different decisionmaking situations are presented. The frst model employs
fuzzy logic and technique for order preference by similarity to ideal solution
(TOPSIS). The second model employs fuzzy logic, extent analysis method,
and analytic network process. The third model uses fuzzy multicriteria anal
ysis method.
In chapter 11, it is shown how an expert system can be built using Bayesian
updating and fuzzy logic, to decide whether it is more sensible to repair a
used product of interest for subsequent sale on a secondhand market than to
disassemble it for subsequent reprocessing.
In chapter 12, a fuzzy costbeneft function is formulated and then used to
perform a multicriteria economic analysis for selecting an economical new
product to produce in a closedloop supply chain.
In chapter 13, fuzzy logic, quality function deployment (QFD), and method
of total preferences are used to select the market with the most potential in
which to sell a used product from a set of candidate secondhand markets.
© 2009 by Taylor & Francis Group, LLC
Introduction 9
In chapter 14, a model consisting of two design experiments that use the Six
Sigma concept to achieve better synchronization in a reverse supply chain is
presented. This model tailors the individual processes in such a way that the
overall delivery performance is maximized.
In chapter 15, appropriate performance aspects and their enablers (drivers
of performance metrics) are identifed for a reverse/closedloop supply chain
environment, and a performance measurement model that uses linear physical
programming (LPP) and quality function deployment (QFD) is presented.
Chapter 16 presents the conclusions of this book.
1.4 Conclusions
This chapter frst presented the factors that motivated this project, viz., driv
ers of reverse and closedloop supply chains. Then the overview and outline
of the book were presented.
References
1. Platt, B., and Hyde, J. 1997. Plug into electronics reuse, 13–38. Washington, DC:
Institute for Local Self Reliance.
2. Knemeyer, A. M., Ponzurick, G. T., and Logar, M. C. 2002. A qualitative exami
nation of factors affecting reverse logistics systems for endoflife computers.
International Journal of Physical Distribution and Logistics Management 32:455–79.
3. Socolof, M., Overly, L. K., and Greibig, J. 2003. Desktop computer displays: A life
cycle assessment. EPA 744R01004. Washington, DC: U.S. Environmental Pro
tection Agency.
4. Williams, E., Ayres, R., and Heller, M. 2002. The 1.7 kg microchip: Energy and
chemical use in the production of semiconductors. Environmental Science and
Technology 36:5504–10.
5. Gungor, A., and Gupta, S. M. 1999. Issues in environmentally conscious manu
facturing and product recovery: A survey. Computers and Industrial Engineering
36:811–53.
6. Beamon, M. B., and Fernandes, C. 2004. Supplychain network confguration
for product recovery. Production Planning and Control 15:270–81.
7. Fleischmann, M. 2001. Quantitative models for reverse logistics: Lecture notes in
economics and mathematical systems. Berlin: SpringlerVerlag.
8. Anonymous. 2003. Computer recycling bill is gaining support. Boston Metro,
January 10–12, p. 6.
9. Lambert, A. J. D., and Gupta, S. M. 2005. Disassembly modeling for assembly, main
tenance, reuse, and recycling. Boca Raton, FL: CRC Press.
© 2009 by Taylor & Francis Group, LLC
10 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
10. Savaskan, R. C., Bhattacharya, S., and Van Wassenhove Luk, N. 2004. Closed
loop supply chain models with product remanufacturing. Management Science
50:239–52.
11. Thierry, M., Salomon, M., van Nunen, J., and van Wassenhove, L. N. 1995. Stra
tegic issues in product recovery management. California Management Review
37:114–35.
12. Chopra, S., and Meindl, P. 2006. Supply chain management: Strategy, planning and
operations. Englewood Cliffs, NJ: Prentice Hall.
13. Rogers, D. 2001. RLEC project plans. Livonia, October, p. 18.
© 2009 by Taylor & Francis Group, LLC
11
2
Strategic Planning of Reverse and
ClosedLoop Supply Chains
2.1 Introduction
As explained in chapter 1, the frst decisionmaking phase in a supply chain
is supply chain design, which is also called strategic planning. Strategic plan
ning primarily involves the structuring of a supply chain over the next sev
eral years. It is longrange planning and is typically performed every few
years when a supply chain needs to expand its capabilities [1].
The various issues faced by strategic planners of reverse and closedloop
supply chains are the following:
1. Selection of used products
2. Evaluation of collection centers
3. Evaluation of recovery facilities
4. Optimization of transportation of goods
5. Evaluation of marketing strategies
6. Evaluation of production facilities
7. Evaluation of futurity of used products
8. Selection of new products
9. Selection of secondhand markets
10. Synchronization of supply chain processes
11. Supply chain performance measurement
In all of these issues, strategic planners must meet the following chal
lenges: uncertainty in supply rate of used products, unknown condition of
used products, and imperfect correlation between supply of used products
and demand for reprocessed goods.
This chapter gives an overview of the aforementioned issues (sections 2.2
through 2.12, respectively) and how they could be addressed in a variety of
decisionmaking situations using effective quantitative models. Section 2.13
gives some conclusions.
© 2009 by Taylor & Francis Group, LLC
12 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
2.2 Selection of Used Products
Although many original equipment manufacturers (OEMs) are obligated to
take products back from the consumers upon the products’ end of use (hence
called used products), there are also many thirdparty companies that collect
used products solely to make proft. These companies select only those used
products for which revenues from recycle or resale of the products’ com
ponents are expected to be higher than the costs involved in collection and
reprocessing of used products and in disposal of waste. The various sce
narios for selecting economical used products could differ as follows:
1. Evaluation criteria could be presented in terms of classical numeri
cal constraints.
2. Evaluation criteria could be presented in terms of ranges of different
degrees of desirability.
2.3 Evaluation of Collection Centers
Designing an effcient reverse or closedloop supply chain requires selec
tion of effcient collection centers where used products are disposed of by
the consumers. These collection centers, after initial processing (for example,
sorting), ship the used products to recovery facilities or production facilities
where reprocessing operations such as disassembly and recycling/remanu
facturing are carried out.
The various scenarios for evaluating collection centers for effciency could
differ as follows:
1. Supply chain company executives, whose primary concern is proft,
could be the sole decision makers.
2. There could exist three different categories of decision makers:
consumers, local government offcials, and supply chain company
executives. The weights (importance values) of evaluation criteria
are given.
3. There could exist the same three different categories of decision
makers given in scenario 2, but weights of evaluation criteria are not
given (hence, must be derived).
4. Evaluation could be made from the perspective of a remanufactur
ing facility interested in buying used products from the candidate
collection centers. The goals are expressed in terms of performance
indices (effciency scores).
© 2009 by Taylor & Francis Group, LLC
Strategic Planning of Reverse and ClosedLoop Supply Chains 13
5. Evaluation could be made in the same manner as in scenario 4,
but the goals are expressed in terms of Taguchi losses (ineffciency
scores).
2.4 Evaluation of Recovery Facilities
Effcient recovery facilities (where reprocessing operations such as disas
sembly and recycling/remanufacturing are carried out) are as essential for a
reverse supply chain as effcient collection centers. Hence, in addition to eval
uation of collection centers (see section 2.3), strategic planning of a reverse
supply chain involves evaluation of recovery facilities. The various scenarios
for evaluating recovery facilities for effciency could differ as follows:
1. Evaluation criteria could be given numerical weights (importance
values).
2. Evaluation criteria could be presented in terms of ranges of different
degrees of desirability.
3. Decision makers could have conficting criteria for evaluation, and
weights for evaluation criteria are given.
4. Decision makers could have conficting criteria for evaluation, and
weights for evaluation criteria are not given (hence, must be derived).
5. A very simple evaluation technique could be desired (where only
the “most important” evaluation criteria are considered).
2.5 Optimization of Transportation of Goods
The focus of this issue is on achieving transportation of the right quantities
of products (used, remanufactured, and new) across a reverse or closedloop
supply chain while satisfying certain constraints. The various scenarios for
this problem could differ as follows:
1. Decisionmaking criteria for a reverse supply chain could be given
in terms of classical supplyanddemand constraints.
2. Decisionmaking criteria for a reverse supply chain could be pre
sented in terms of ranges of different degrees of desirability.
3. Besides optimal transportation of products, there could be a need to
address the following issues in one continuous phase for a closed
loop supply chain: selection of used products and evaluation of
© 2009 by Taylor & Francis Group, LLC
14 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
production facilities. Also, the decisionmaking criteria could be
presented in terms of classic supplyanddemand constraints.
4. The scenario could be the same as scenario 3, except that the deci
sionmaking criteria could be presented in terms of different degrees
of desirability.
5. The scenario could be the same as scenario 3, except that the deci
sionmaking criteria could be imprecise.
2.6 Evaluation of Marketing Strategies
A reverse/closedloop supply chain program is successful only if there is
a high level of public participation in the program. The level of public par
ticipation is shouldered by the marketing strategy of that program. Hence,
evaluating the marketing strategy of a reverse/closedloop supply chain pro
gram is equivalent to evaluating how well the strategy is driving the pub
lic to participate in the program. Whereas the drivers for governments and
companies to implement a reverse/closedloop supply chain program and
evaluate the program’s marketing strategies are environmental conscious
ness and proftability, respectively, the drivers for the public to participate in
the program are numerous and often conficting with each other (for exam
ple, the more regularly a reverse/closedloop supply chain program offers to
collect used products from consumers, the higher the taxes the consumers
will have to pay). The various scenarios for evaluating the marketing strat
egy of a reverse/closedloop supply chain program could differ as follows:
1. The program could be exclusively for reverse supply chain opera
tions, i.e., absence of a closed loop.
2. The program could be for a closedloop supply chain and the deci
sion maker could be uninterested in considering interdependencies
among evaluation criteria.
3. The program could be for a closedloop supply chain and the deci
sion maker could be interested in considering interdependencies
among evaluation criteria.
2.7 Evaluation of Production Facilities
Effcient production facilities (not only new products are produced but also
used products are reprocessed) are as essential for a closedloop supply
chain as effcient collection centers. Hence, in addition to selecting effcient
collection centers (see section 2.3), strategic planning of a closedloop supply
© 2009 by Taylor & Francis Group, LLC
Strategic Planning of Reverse and ClosedLoop Supply Chains 15
chain involves evaluation of production facilities. The various scenarios for
evaluating production facilities for effciency could differ as follows:
1. The decision maker could desire to structure the problem using a
simple hierarchical model, wherein interactions among evaluation
criteria could be ignored.
2. Interactions among evaluation criteria could not be ignored, which
in turn leads to a more complex problem structure.
3. Interactions among evaluation criteria could not be ignored, and the
decision maker could desire to see how close the rating of a candi
date production facility is to the “ideal solution.”
2.8 Evaluation of Futurity of Used Products
Although a major driver for companies interested in collecting used prod
ucts is recoverable value through reprocessing, those companies seldom
know when those products were bought and why they were discarded. Also,
the products do not indicate their remaining life periods. Hence, they often
undergo partial or complete disassembly for subsequent reprocessing. The
focus of this issue is to fnd out whether, for some products, it makes more
sense to make necessary repairs to the products and sell them on second
hand markets than to disassemble them for subsequent reprocessing.
2.9 Selection of New Products
The focus of this issue is to help companies select and produce only those new
products for which revenues in the closedloop supply chain are expected to
be higher than the costs. The revenues include new product sale revenue
(revenue from selling new products, viz., products in the forward fow of the
closedloop supply chain), reuse revenue (revenue from direct sale/usage in
remanufacturing usable components of used products), and recycle revenue
(revenue from selling material obtained from recycling of unusable compo
nents of used products). The costs include new product production cost (cost
to produce new products), collection cost (cost to collect used products from
consumers), reprocessing cost (cost to remanufacture/recycle used products),
disposal cost (cost to dispose of the material left over after remanufacturing
or recycling of used products), lossofsale cost (cost due to loss of sale, which
might occur occasionally, due to lack of supply of used products), and invest
ment cost (capital required for facilities and machinery involved in produc
tion of new products and collection and reprocessing of used products).
© 2009 by Taylor & Francis Group, LLC
16 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
2.10 Selection of Secondhand Markets
The focus of this issue is to select the market with the most potential in which
to sell a repaired used product from a set of candidate secondhand markets.
The main criteria for evaluation are beforesale performance (refects the
ability to attract new customers to the secondhand market), whilesale per
formance (refects the ability to motivate the customers to buy secondhand
products while the customers are in the secondhand market), and aftersale
performance (refects the ability to attract old customers to the secondhand
market).
2.11 Synchronization of Supply Chain Processes
Synchronization among the internal business processes of a supply chain
is essential for effective management of the supply chain. Synchroniza
tion in a supply chain means reducing the variability among the internal
business processes or partners such that each stakeholder in the supply
chain acts in a way that is appropriately timed with the actions of the other
stakeholders [2]. The delivery performance of a supply chain is maximized
largely by synchronizing the internal business processes such that the fnal
product fts in the customerspecifed delivery window with a very high
probability.
The focus of this issue is to achieve synchronization of processes, such
as procurement, inspection, disassembly, remanufacturing, transportation,
and delivery, in a reverse supply chain.
2.12 Supply Chain Performance Measurement
The focus of this issue is to measure the performance of a reverse/closed
loop supply chain with respect to various metrics, such as ontime delivery,
green image, service effciency, and location of facilities. Due to the inher
ent differences in various aspects between forward and reverse/closedloop
supply chains (see chapter 1), the performance metrics and evaluation tech
niques used in a forward supply chain cannot be extended to a reverse/
closedloop supply chain.
© 2009 by Taylor & Francis Group, LLC
Strategic Planning of Reverse and ClosedLoop Supply Chains 17
2.13 Conclusions
This chapter gave an overview of various issues faced by strategic planners of
reverse and closedloop supply chains in various decisionmaking situations.
References
1. Chopra, S., and Meindl, P. 2006. Supply chain management. 3rd ed. New York:
Prentice Hall.
2. Antony, J., Swarnkar, R., and Tiwari, M. K. 2006. Design of synchronized sup
ply chains: A genetic algorithm based Six Sigma constrained approach. Interna
tional Journal of Logistics Systems and Management 2:120–41.
© 2009 by Taylor & Francis Group, LLC
19
3
Literature Review
3.1 Introduction
As explained in chapter 1, a supply chain involves three phases of decision
making: supply chain design (also called strategic planning), supply chain planning
(also called tactical planning), and supply chain operation (also called operational
planning). In most of the literature for reverse and closedloop supply chains,
tactical planning is addressed in conjunction with strategic planning.
In this chapter, a brief review of the literature related to the three phases
of decision making in reverse and closedloop supply chains, viz., strategic
planning, tactical planning, and operational planning, is presented.
This chapter is organized as follows: section 3.2 presents a review of literature
about operational planning in reverse and closedloop supply chains, section
3.3 gives a review of literature about strategic and tactical planning in reverse
and closedloop supply chains, and section 3.4 gives some conclusions.
3.2 Operational Planning of Reverse
and ClosedLoop Supply Chains
Disassembly marks the frst step in reprocessing of used products. Disas
sembly is defned as a systematic method of separating a product into its
constituent parts, components, subassemblies or other groupings through
a series of operations. Disassembly can be of two types: nondestructive or
destructive. Nondestructive disassembly is the process of systematic removal
of the constituent parts from an assembly in a manner by which there is
no impairment of the parts during the process. On the contrary, destructive
disassembly involves separating materials from an assembly with an objec
tive of sorting the different material types for recycling. Disassembly can
be complete or partial. In a complete disassembly, the used product is fully
disassembled, whereas in partial disassembly, the used product is not fully
disassembled, and only certain parts or assemblies are recovered. See [15],
© 2009 by Taylor & Francis Group, LLC
20 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
[29], [48], and [59] for a comprehensive list of issues related to disassembly
planning and scheduling.
Pnueli and Zussman [64] propose a methodology that includes identifying
reprocessing options of all the parts of a product and improving the prod
uct’s design.
Penev and De Ron [63] discuss the determination of optimal disassem
bly level and sequences of a used product, which provide conditions for the
generation of proft while considering the environmental impact. In order to
obtain the optimal disassembly process strategy, the authors utilize graph
theory and cost analysis.
Lambert [50] proposes a methodology to identify the disassembly level in
an economically optimal way. After disassembly, reusable parts/subassem
blies are cleaned, refurbished, tested, and directed to the part/subassembly
inventory for reprocessing operations.
Lambert and Gupta’s book [49] on disassembly modeling presents disas
sembly in the context of the entire product life cycle. It examines disassembly
on the intermediate level, incorporating design for disassembly, concurrent
design, and reverse supply chain. The book also presents a comprehensive
discussion of the theories and methodologies associated with disassembly,
and the authors incorporate realworld case examples to explore the three
main areas of application: assembly optimization, maintenance and repair,
and reprocessing.
Gungor and Gupta [28] provide an exhaustive review of literature in the
areas of environmentally conscious manufacturing and product recovery
that includes several aspects of environmentally conscious manufacturing
(design for environment, design for disassembly, etc.), common issues in
product recovery (recycling, remanufacturing), disassembly process plan
ning (production planning and inventory control issues), etc.
A disassembly line is perhaps the most suitable setting for disassembling
large products. Gungor and Gupta [27] discuss the importance of a disas
sembly line and identify the critical issues and complexities and their effects
on the disassembly line.
Boon et al. [14] study the economic viability of recycling infrastructure in
the United States in relation to the electronics industry. They use a goal pro
gramming technique to study the sensitivity of several parties (disassemblers,
recyclers, etc.) to different factors involved in the endoflife processing.
Kongar and Gupta [40] and Imtanavanich and Gupta [34, 35] propose sev
eral multicriteria decisionmaking methodologies in a disassemblytoorder
setting under different decisionmaking environments, which, when solved,
provide the number of used products to be taken back for reuse, recycling,
storage, or disposal to meet the demand for those products/components.
Guide et al. [22] identify the problems associated with production plan
ning issues in a closedloop supply chain and advocate that each closedloop
supply chain differs from every other closedloop supply chain and a “one
size fts all” approach does not work. Each type of system offers different
managerial concerns. They apply Hayes and Wheelwright’s productprocess
© 2009 by Taylor & Francis Group, LLC
Literature Review 21
matrix framework to examine three cases representing remanufactureto
stock, reassembletoorder, and remanufacturetoorder environments,
thereby extending the productprocess matrix to include the insights on pro
duction planning in closedloop supply chains.
Rubio et al. [70] review an extensive list of articles on reverse logistics pub
lished in the production and operations management feld. They provide a
database of articles published in the area of reverse logistics in the period
1995–2005 by exploring the topic, methodology, and techniques of analysis,
as well as other relevant aspects of research.
Meade et al. [58] provide an extensive review of literature in the area of
reverse logistics as well as an overview of defnitions and research opportu
nities in this area.
Inventory control and production planning methods are thoroughly under
stood and well established for traditional forward supply chains. However,
available techniques for a traditional forward supply chain are not transfer
able and adequate for a reverse supply chain. Guide and Srivastava [24] list
the factors that complicate the inventory control and production planning in
a reverse supply chain. These factors are as follows:
Probabilistic recovery rate of parts from the used products, which r
implies a high degree of uncertainty in material planning
Unknown conditions of recovered parts until inspected, thus lead r
ing to stochastic routings and lead times
Partmatching problem during the assembly process r
Added complexity of a remanufacturing shop structure r
Uncertainty in supply rate of used products r
Problem of imperfect correlation between supply of used products r
and demand for reprocessed goods
Inventory control models in a reverse supply chain are required to keep
track of various goods through the chains, viz., used products, remanufac
tured products, recycled goods, etc. Inventory models for repairable items
and maintenance systems where failed machines are replaced by warm
or cold spares carry some similarities to the reverse supply chain models
[11, 16]. Some authors consider deterministic models in which supplyand
demand rates are constant. Mabini et al. [57] consider such a model with
fxed setup costs for orders and remanufacturing and linear holding costs
for used products and fnished goods. Also, the model includes understock
servicelevel constraint and machine sharing during the remanufacturing
stage. Richter [68, 69] proposes a similar model under a different control
policy where he provides formulation to calculate optimal values for con
trol parameters and analyzes their dependence on the supply rate of used
products. In a later paper, Richter [67] considers a mixture of two pure poli
cies, total disposal and total remanufacturing in a deterministic system. In a
© 2009 by Taylor & Francis Group, LLC
22 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
more recent paper, Teunter [78] studies a deterministic system with continu
ous review and no lead times for outside procurement or remanufacturing.
Although it is assumed that remanufactured products are as good as new
ones, the associated inventory holding costs for remanufactured products
are lower than those for new items. The author proposes an Economic Order
Quantity (EOQ) control mechanism with fxed batch sizes and shows that
in a given time period, the optimal batch size should be equal to 1 for either
manufacturing or remanufacturing.
In repairable item inventory and remanufacturing systems, stochastic
supply and demand occurrences are controlled by periodic and continuous
review models. Periodic review models aim to minimize expected costs over
a fnite period and obtain optimal control policies under various assump
tions. On the other hand, continuous review models monitor the inventory
level incessantly and aim to fnd static control policies to obtain optimal con
trol parameters to minimize the expected system costs. Simpson [75] con
siders a periodic review model for separate inventories for serviceable and
recoverable parts. He shows that a threeparameter control policy that con
trols demand, remanufacturing, and disposal activities is optimal. Cohen
et al. [17] consider a periodic review model in which used products can be
placed in serviceable inventory directly. The authors assume that a fxed per
centage of supplied products is returned to the manufacturer after a fxed
lead time. The model optimizes the tradeoff between holding and shortage
costs. For a similar model, Kelle and Silver [38] utilize an integer program
based on the net demand per period under chance constraints with fxed
setup costs. Simpson’s model [75] is extended by Inderfurth [36] by consider
ing the effects of nonzero lead times for the reprocessing process and orders.
He considers a push system for remanufacturing to avoid the storage of used
products and shows that the difference between these two lead times is an
important complexity factor of the system. Kiesmuller and Van der Laan [39]
consider a periodically reviewed system in a fnite horizon with supply rates
dependent on demand rates. The authors assume deterministic lead times
and an independent stochastic demand process. They show that dependent
supply rates provide a better inventory control performance. Teunter and
Vlachos [77] consider the impact of disposal on the discounted average cost.
Their model considers deterministic and equal lead times for remanufac
tured and new items procurement.
Heyman [31] analyzes the continuous review inventory control case where
incoming used products are disposed of whenever the inventory position
reaches a predetermined level. The author assumes zero repair times and
does not consider procurement lead times. In a following paper, Muckstadt
and Isaac [60] present a production planning and inventory control model
for remanufacturing. The authors develop an approximate control strategy
with respect to reorder points and order quantities for a singleitem prod
uct case where used products are remanufactured. They consider fxed lead
times and no disposal of used products. Van der Laan et al. [83] consider a
singleproduct, singleechelon production and inventory system with used
© 2009 by Taylor & Francis Group, LLC
Literature Review 23
products, remanufacturing, and disposal. In their paper, the authors compare
the performance of three different procurement and inventory control strat
egies. Salomon et al. [72] and Van der Laan and Salomon [82] develop and
compare PUSH and PULL strategies for a joint production and inventory sys
tem using both remanufactured and new parts. Korugan and Gupta [41] con
sider a twoechelon inventory system with stochastic lead times and arrival
rates. They model the system as a PUSHtype maketostock open queueing
network, where along with the manufacturing and remanufacturing pro
cesses, the demand arrivals are modeled as service. The authors report that
increasing supply rates decrease the expected total cost, as the holding cost
of a used product at the lower echelon is lower than the lost sales cost. Then
Korugan and Gupta [42] look at the CONWIP problem of the hybrid produc
tion systems with mutually exclusive manufacturing and remanufacturing
processes. The authors generalize dynamic and static routing methods by
applying an adaptive kanban control policy. The experiments conducted by
the authors demonstrate that dedicated kanban control performs better with
reasonable utilization rates, whereas adaptive kanban policy always outper
forms nonadaptive policies. Udomsawat and Gupta [80] focus on the pro
duction control aspect of the disassembly environment. The author adapts
the kanban control mechanism to a disassembly line and develops a sys
tem that uses several types of kanbans attached to various components and
subassemblies.
Applicability of traditional production planning and scheduling methods
for remanufacturing systems is extremely limited. Therefore, either new
methodologies have to be developed or classical methods have to be modi
fed to manage the complications of the product reprocessing process [81].
A number of authors propose that material requirements planning (MRP)
techniques utilize a reverse bill of materials (BOM) to allow the manage
ment and control of inventories in the remanufacturing environment [30, 47,
62]. Guide and Srivastava [23] propose a specifc structure for MRP mechan
ics and evaluate a method to calculate safety stock for material recovery
uncertainty. A multiobjective mathematical model formulation for a recycle
oriented manufacturing system is presented by Hoshino et al. [32]. The opti
mization model for the recycleoriented manufacturing system is designed
for a single product with m number of parts. Each part has three attributes
associated with it: (1) reusable part, (2) not reusable but reproducible part,
and (3) not reproducible or reusable part. The authors apply a goal program
ming approach with two objectives: proft maximization and recycling rate
maximization. Guide [25] proposes scheduling using the drum–buffer–rope
concept as an alternative to the MRP method. Guide et al. [26] evaluate vari
ous part orderrelease strategies in a remanufacturing environment using a
simulation model. The authors focus on the problem of uncontrolled release
of parts from disassembly operations, which may cause long queues at
machine centers. This situation may increase lead times and their variabil
ity, making customer service levels decline. Various types of orderrelease
strategies are discussed, viz., a level strategy, a batch strategy, a local order
© 2009 by Taylor & Francis Group, LLC
24 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
strategy, and a global orderrelease strategy, and the fndings are mixed as
to what is the best disassembly release mechanism. Ferrer and Whybark [19]
describe the frst fully integrated material planning system to facilitate man
aging a remanufacturing facility. The approach extends the methods used
for material planning in a remanufacturing environment in several ways.
First, it explicitly links the volume of used products with the volume of
sales. Second, it uses the bill of material for each component directly, with no
need for modifcation. Third, the system derives the need for parts and uses
optimizing procedures to determine the disassembly schedule. Finally, part
commonality and different yield factors are explicitly included.
Aksoy and Gupta [2] model a hybrid manufacturing system with distinct
cells for disassembly, testing, and remanufacturing activities. The hybrid sys
tem is modeled as an open queueing network (OQN) with unreliable servers
and fnite buffers. Then Aksoy and Gupta [1] look at the effect of reusable
rate variation on the performance of a remanufacturing system. The authors
report that when the supply rate of used products is low, the total cost is rela
tively insensitive to the reusability rate. However, as the supply rate becomes
higher, the reusability rate has a signifcant effect on the total cost. The higher
the reusability rate, the lower the total cost. Also, mean process time (average
remanufacturing time) is sensitive to the buffer size, breakdown rate, repair
rate, and service rate of the stations in the remanufacturing system. Vari
ability in the buffer sizes in the network causes a signifcant variation in the
mean process time. The authors also examine the tradeoff between expand
ing the buffer sizes and increasing the service capacity of machines [3, 4].
3.3 Strategic and Tactical Planning of Reverse
and ClosedLoop Supply Chains
Strategic and tactical planning of reverse and closedloop supply chains is
a relatively new area of research, and hence only a few quantitative models
and case studies have been reported in the literature. In this section, we pres
ent a brief survey of those models and studies (see Fleischmann’s book [20]
for a detailed survey of many of the case studies):
Veerakamolmal and Gupta [84] propose a costbeneft function and r
a measure called “design for disassembly index” that analyzes the
tradeoff between the costs and benefts of endoflife disassembly
to fnd the optimum costbeneft ratio for endoflife retrieval. Their
study helps in comparing two product designs by assessing feasible
combinations of components to be retrieved from a used product.
Further, it compares the combination with the highest costbeneft
from one design with those from the others.
© 2009 by Taylor & Francis Group, LLC
Literature Review 25
Louwers et al. [55] consider the design of a reverse supply chain for r
carpet waste in Europe. A continuous location model in which all
costs are considered volume dependent is proposed. The nonlin
ear model, when solved, determines the appropriate locations and
capacities for the regional recovery facilities, taking into consider
ation transportation, investment, and processing costs.
Ravi et al. [65] propose a balanced scorecard and analytic network r
process (ANP)–based approach to evaluate the alternative reverse
logistics operations for used computers. It is a holistic approach that
links fnancial and nonfnancial, tangible and intangible, and inter
nal and external factors, for the selection of an alternative.
Beamon and Fernandes [10] propose a multiperiod mixedinteger r
linear programming model to study a closedloop supply chain
where manufacturers produce new products and remanufacture
used products. Their model addresses issues that include which
warehouses and collection centers should be open and which ware
houses should have sorting capabilities.
Barros et al. [8] address the design of a supply chain for recycling r
sand from processing construction waste in the Netherlands. A four
level sand recycling network is considered: (1) crushing companies
yielding sieved sand from construction waste, (2) regional depots
specifying the pollution level and storing cleaned and halfcleaned
sand, (3) treatment facilities cleaning and storing polluted sand, and
(4) infrastructure projects where sand can be reused. The locations
of the sand sources are known and their supply volumes are esti
mated based on historical data. The optimal number, capacities, and
locations of depots and cleaning facilities are to be determined. The
authors propose a multilevel capacitated facility location model for
this problem formulated as a mixedinteger linear programming
model and solved via iterative rounding of LP relaxations strength
ened by valid inequalities. Listes and Dekker [53] propose a stochas
tic programming–based approach for the sand recycling network to
account for the uncertainties. The stochastic model seeks to fnd a
solution that is approximately balanced between some alternative
scenarios identifed by feld experts.
Savaskan et al. r [74] study the problem of choosing the appropriate
reverse channel structure for collecting the used products from cus
tomers. They compare three decentralized closedloop supply chain
models with the manufacturer collecting the used products, the
retailer collecting the used products, and a third party collecting the
used products. The models are compared with respect to the whole
sale price, product return rate, and total supply chain profts.
Ammons et al. [7] address carpet recycling in the United States. r
A logistics network that includes collection of used carpets from
© 2009 by Taylor & Francis Group, LLC
26 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
carpet dealerships, as well as separation of nylon and other reusable
materials while landflling the remainder, is investigated. Although
the delivery sites for recovered materials are assumed to be known,
the optimal number and location of both collection sites and pro
cessing plants for alternative confgurations are to be determined.
In addition, the amount of carpet collected from each site is to be
determined. Facility capacity constraints are the main restrictions in
view of the vast volume landflled. The authors propose a multilevel
capacitated mixedinteger linear programming model to address
this problem.
Biehl et al. r [13] simulate a reverse logistics network for carpet recy
cling to manage highly variable return fows. They use an experi
mental design technique to study the effect of system design factors
as well as environmental factors that affect the operational perfor
mance of such a reverse logistics network. From their study, the
authors conclude that even with the design of an effcient reverse
logistics network and use of sophisticated recycling technologies,
return fows cannot meet demand for nearly a decade. They also
discuss possible managerial options to address this problem, which
include legal responses to require return fows and utilization of
market incentives for carpet recycling.
Hu et al. [33] present a costminimization model for a multitimestep, r
multitype hazardous wastes reverse logistics system. The authors
formulate a discretetime analytical model that minimizes total haz
ardous waste reverse logistics costs subject to constraints, including
business operating strategies and government regulations. The criti
cal activities that include waste collection, storage, processing, and
distribution are considered in their model. By using the proposed
methodology, coupled with operational strategies, it is found that
the total reverse logistics costs can be reduced by up to 49%.
Lieckens and Vandaele [51] combine queuing models with the tradi r
tional reverse logistics location model and formulate mixedinteger
linear programming models to determine which facilities to open
while minimizing the total cost of investment, transportation, dis
posal, procurement, etc. By combining the queuing models, some
dynamic aspects such as lead time and inventory positions, and
the high degree of uncertainty associated with reverse logistics
networks, are accounted for. With these extensions, the problem
is defned as a mixedinteger nonlinear programming model. The
model is presented for a singleproduct, singlelevel network, and
several case examples are solved using genetic algorithms based on
the technique of differential evolution.
Alshamrani et al. [5] study the reverse logistics network for blood dis r
tribution with the American Red Cross, where containers in which
blood is delivered from a central processing unit to customers in one
© 2009 by Taylor & Francis Group, LLC
Literature Review 27
time period are available for return to the central processing unit
the following period. Containers not picked in the period following
their delivery incur a penalty cost. This leads to a dynamic logistics
planning problem, where in each period the vehicle dispatcher needs
to design a multistop vehicle route while determining the number
of containers to be picked up at each stop. A heuristic procedure is
developed to solve the route design–pickup strategy problem.
Vlachos et al. [85] address the capacity planning issues in remanu r
facturing facilities in reverse supply chains through a simulation
model based on the principles of system dynamics methodology.
Apart from considering economic issues, their study takes into
account environmental issues such as takeback obligations and the
“green image” effect on customer demand. The simulation model
serves as an experimental tool that helps in evaluating longterm
capacity planning policies using total supply chain proft as a mea
sure of effectiveness.
Lu and Bostel [56] study the facility location problem in a reman r
ufacturing network that has a strong interaction between the for
ward and reverse fows of products. Remanufactured products are
introduced as new ones into the forward fow. The authors assume
that the demand is deterministic and the facilities are of three dif
ferent types: producers, remanufacturing centers, and intermediate
centers. The problem is modeled as a 01 mixedinteger program
ming problem that is solved using an algorithm based on Lagrang
ian heuristics.
Spengler et al. [76] study the recycling networks for industrial by r
products in the German steel industry. Recycling facilities with vari
able capacity levels and corresponding fxed and variable processing
costs can be installed at a set of potential locations. Thus, one needs
to determine which recycling processes or process chains to install
at which locations and their capacity levels. The authors propose a
multilevel warehouse location model with piecewise linear costs,
which is used for optimizing several scenarios.
Thierry et al. [79] propose a conceptual model for a closedloop sup r
ply chain that addresses the situation of a manufacturing company
collecting used products for recovery in addition to producing and
distributing new products. The recovered products are sold under
the same conditions as new ones to satisfy a given market demand.
The distribution network encompasses three levels: plants, ware
houses, and markets. All facilities are fxed externally, and hence no
fxed costs are considered in the model. The objective is to determine
the costoptimal fow of goods in the network under the given capac
ity constraints. The problem is formulated as a linear programming
model, which can be solved for optimality.
© 2009 by Taylor & Francis Group, LLC
28 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
Berger and Debaillie [12] address a situation similar to that of Thi r
erry et al. [79] and propose a conceptual model for extending an
existing forward supply chain with disassembly centers to allow for
recovery of used products. The model is illustrated in a fctitious
case of a computer manufacturer. Although the facilities in the for
ward supply chain are fxed, the number, locations, and capacities
of the disassembly centers are to be determined. In a variant of this
model, the recovery network is extended to another level by sepa
rating inspection and disassembly/repair centers. After inspection,
rejected items are disposed of, whereas recoverable items are sent to
the disassembly centers. To this end, the authors propose a multi
level capacitated mixedinteger linear programming model.
Lim et al. [52] propose a mixedinteger programming model that r
takes into account multiperiod planning horizons with uncertain
ties for a product with modular design and multiproduct confgura
tions. Besides maximizing overall proft, minimizing environmental
impacts by minimizing energy consumption is considered.
Reimer et al. r [66] model the economics of electronics recycling from the
perspective of recyclers, generators, and material processors, individu
ally. They propose a nonlinear mixedinteger programming model for
optimizing processing decisions in electronics recycling operations.
Jayaraman et al. r [37] analyze the logistics network of an electronic
equipment remanufacturing company in the United States. The com
pany’s activities include collection of used products from customers,
remanufacturing, and distribution of remanufactured products. In
this network, the optimal number and locations of remanufacturing
facilities, and the number of used products collected, are to be deter
mined while considering investment, transportation, processing,
and storage costs. The authors propose a multiproduct capacitated
warehouse location mixedinteger linear programming model that
is solved to optimality for different supplyanddemand scenarios.
Krikke et al. r [43] report a case study concerning the implementation
of a remanufacturing process at a copier manufacturing company in
the Netherlands. The reverse supply chain is subdivided into three
main stages: (1) disassembly of return products to a fxed level; (2)
preparation, which encompasses the inspection and replacement
of critical components; and (3) reassembly of the remaining carcass
together with repaired and new components into a remanufactured
machine. Although the supplying processes and disassembly are
fxed, optimal locations and fow of goods are to be determined
for both the preparation and the reassembly operations. Based on
a mixedinteger linear programming model, the optimal solution
minimizing operational costs is compared with a number of prese
lected managerial solutions.
© 2009 by Taylor & Francis Group, LLC
Literature Review 29
Kroon and Vrijens [46] consider the design of a logistics system for r
reusable transportation packaging. More specifcally, a closedloop,
depositbased system is considered for collapsible plastic contain
ers that can be rented as secondary packaging material. The actors
involved in the system include a central agency owning a pool of
reusable containers; a logistics service provider responsible for stor
ing, delivering, and collecting the empty containers; senders and
recipients of full containers; and carriers transporting full contain
ers from sender to recipient. In addition to determining the number
of containers required for running the system and an appropriate
fee per shipment, a major question is where to locate the depots
for empty containers. An additional requirement is balancing the
number of containers at the depots. The problem is formulated as a
mixedinteger linear programming model that is closely related to a
classical uncapacitated warehouse location model.
Salema et al. [71] point out that the majority of the quantitative mod r
els that exist in the area of reverse supply chain design are case spe
cifc and hence lack generality. To this end, they propose a generic
reverse supply chain model that incorporates multiproduct man
agement, capacity limits, and uncertainty in product demands and
returns, and they propose a mixedinteger formulation that is solved
using standard Branch and Bound (B&B) techniques.
Krikke et al. [45] focus on commercial returns that have nothing to do r
with environmental legislation. These returns are increasing due to
trends such as product leasing, catalog/Internet sales, shorter product
replacement cycles, and increased warranty claims. The authors pro
pose several options for a closedloop supply chain: reusing the prod
uct as a whole, reusing the components, or reusing the materials.
Listes [54] presents a generic stochastic model for the design of net r
works constituting both supply and return channels, organized in
a closedloop system. The model accounts for a number of alterna
tive scenarios, which may be constructed based on critical levels of
design parameters such as demand or returns. A decomposition
approach is suggested to solve this model, based on the branchand
cut procedure known as the integer Lshaped method.
Bautista and Pereira [9] address reverse logistics problems arising r
in municipal waste management. The usual collection system in the
European Union countries is composed of two phases. First, citizens
leave their refuse at special collection areas where different types of
waste (glass, paper, plastic, organic material) are stored in special
refuse bins. Subsequently, each type of waste is collected separately
and moved to its fnal destination (a recycling plant or refuse dump).
The study focuses on the problem of locating these collection areas.
The authors propose a genetic algorithm and a GRASP heuristic to
solve the problem.
© 2009 by Taylor & Francis Group, LLC
30 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
Dowlatshahi [18] identifes the present state of theory in reverse r
logistics by formulating the propositions for strategic factors. The
approach used is grounded theory development. The strategic fac
tors are delineated and evaluated in terms of specifc subfactors asso
ciated with each factor by the use of interview protocol and within
the context of an indepth analysis of two companies in different
industries that are engaged in remanufacturing/recycling opera
tions within reverse logistics systems. Based on these insights and
strategic factors and subfactors, a framework for effective design and
implementation of remanufacturing/recycling operations in reverse
logistics is provided. This framework allows for the determination
of the viability of used products/parts.
Amini et al. [6] discuss the competitive value of service management r
activities, particularly repair services, as well as the importance of
the supporting role of effective reverse logistics operations for the
successful and proftable execution of repair service activities. Also,
they present a case study of a major international medical diagnos
tics manufacturer to illustrate how a reverse logistics operation for a
repair service supply chain was designed for both effectiveness and
proftability by achieving a rapidcycle time goal for repair service
while minimizing total capital and operational costs.
Savaskan and Van Wassenhove [73] focus on the interaction between r
a manufacturer’s reverse channel choice to collect postconsumer
goods and the strategic product pricing decisions in the forward
channel when retailing is competitive. To this end, they model a
direct product collection system, in which the manufacturer col
lects used products directly from the consumers, and an indirect
product collection system, in which the retailers act as product return
points. The authors show that the buyback payments transferred
to the retailers for postconsumer goods provide a wholesale pricing
fexibility that can be used to pricediscriminate between retailers of
different proftability.
Wojanowski et al. [86] study the interplay between industrial frms r
and government concerning the collection of used products from
households. The authors focus on the use of a depositrefund require
ment by the government when the collection rate voluntarily achieved
by the frms is deemed insuffcient. A continuous modeling frame
work is presented for designing a dropoff facility network and deter
mining the sales price that maximizes the frm’s proft under a given
depositrefund. The customers’ preferences with regards to purchas
ing and returning the product are incorporated via a discrete choice
model with stochastic utilities.
Krikke et al. [44] develop quantitative modeling to support decision r
making concerning both the design structure of a product and the
design structure of the logistic network. Environmental impacts are
© 2009 by Taylor & Francis Group, LLC
Literature Review 31
measured by linear energy and waste functions. Economic costs
are modeled as linear functions of volumes with a fxed setup com
ponent for facilities. This model is applied to a closedloop supply
chain design problem for refrigerators using reallife data of a Japa
nese consumer electronics company.
Nagurney and Toyasaki [61] develop an integrated framework for the r
modeling of reverse supply chain management of electronic waste,
which includes recycling. The decision makers consist of the sources
of electronic wastes, the recyclers, the processors, and consumers asso
ciated with the demand markets for the distinct products. A multi
tiered electronic recycling network equilibrium model is constructed.
The authors also establish the variational inequality formulation. The
variational inequality formulation allows for the formulation of the
complex reverse supply chain network to obtain the endogenous
equilibrium prices and material fows between tiers.
Gautam and Kumar [21] describe a multiobjective evaluation of the r
tradeoffs among the number and size of dropoff stations, popu
lation covered in a service network, average walking distance to
dropoff stations by the population, and the distance traveled by
collection vehicles. A geographical information system (GIS)–based
model is proposed for the design of a solidwaste system consider
ing waste generation, allocation, recycling options, and location of
dropoff stations.
3.4 Conclusions
In this chapter, a brief review of the literature related to the three phases of
decision making in reverse and closedloop supply chains—strategic plan
ning, tactical planning, and operational planning—was presented.
References
1. Aksoy, H. K., and Gupta, S. M. 2000. Effect of reusable rate variation on the per
formance of remanufacturing systems. In Proceedings of the SPIE International
Conference on Environmentally Conscious Manufacturing, Boston, pp. 13–20.
2. Aksoy, H. K., and Gupta, S. M. 1999. An open queueing network model for
remanufacturing systems. In Proceedings of the 25th Conference on Computers and
Industrial Engineering, New Orleans, pp. 62–65.
3. Aksoy, H. K., and Gupta, S. M. 2005. Buffer allocation plan for a remanufactur
ing cell. Computers & Industrial Engineering, vol. 48, No. 3, 657–677.
© 2009 by Taylor & Francis Group, LLC
32 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
4. Aksoy, H. K., and Gupta, S. M. 2001. Capacity and buffer tradeoffs in a reman
ufacturing system. In Proceedings of the SPIE International Conference on Environ
mentally Conscious Manufacturing II, Newton, MA, pp. 167–74.
5. Alshamrani, A., Mathur, K., and Ballou, R. H. 2007. Reverse logistics: Simulta
neous design of delivery routes and return strategies. Computers and Operations
Research 34:595–619.
6. Amini, M. M., RetzlaffRoberts, D., and Bienstock, C. C. 2005. Designing a
reverse logistics operation for short cycle time repair services. International
Journal of Production Economics 96:367–80.
7. Ammons, J. C., Realff, M. J., and Newton, D. J. 1999. Carpet recycling: Deter
mining the reverse production system design. PolymerPlastics Technology and
Engineering 38:547–67.
8. Barros, A. I., Dekker, R., and Scholten, V. 1998. A twolevel network for recy
cling sand: A case study. European Journal of Operational Research 110:199–214.
9. Bautista, J., and Pereira, J. 2006. Modeling the problem of locating collection
areas for urban waste management. An application to the metropolitan area of
Barcelona. Omega 34:617–29.
10. Beamon, M. B., and Fernandes, C. 2004. Supplychain network confguration
for product recovery. Production Planning and Control 15:270–81.
11. Berg, M. A. 1980. A marginal cost analysis for preventive replacement policies.
European Journal of Operations Research 4:136–42.
12. Berger, T., and Debaillie, B. 1997. Location of disassembly centers for reuse to
extend an existing distribution network (in Dutch). Master’s thesis, University
of Leuven, Belgium.
13. Biehl, M., Prater, E., and Realff, M. J. 2007. Assessing performance and uncer
tainty in developing carpet reverse logistics systems. Computers and Industrial
Engineering 34:443–63.
14. Boon, J. E., Isaacs, J. A., and Gupta, S. M. 2002. Economic sensitivity for end
of life planning and processing of PCs. Journal of Electronics Manufacturing
11:81–93.
15. Brennan, L., Gupta, S. M., and Taleb, K. N. 1994. Operations planning issues in
an assembly/disassembly environment. International Journal of Operations and
Production Planning 14:57–67.
16. Cho, D. I., and Parlar, M. 1991. A survey of maintenance models for multiunit
systems. European Journal of Operations Research 51:1–23.
17. Cohen, M. A., Nahmias, S., and Pierskalla, W. P. 1980. A dynamic inventory
system with recycling. Naval Research Logistics Quarterly 27:289–96.
18. Dowlatshahi, S. 2005. A strategic framework for the design and implementa
tion of remanufacturing operations in reverse logistics. International Journal of
Production Research 43:3455–80.
19. Ferrer, G., and Whybark, D. C. 2001. Material planning for a remanufacturing
facility. Production and Operations Management 10:112–24.
20. Fleischmann, M. 2001. Quantitative models for reverse logistics: Lecture notes in
economics and mathematical systems. Berlin: SpringlerVerlag.
21. Gautam, A. K., and Kumar, S. 2005. Strategic planning of recycling options
by multiobjective programming in a GIS environment. Clean Technologies and
Environmental Policy 7:306–16.
22. Guide, Jr., V. D. R., Jayaraman, V., and Linton, J. D. 2003. Building contingency
planning for closedloop supply chains with product recovery. Journal of Opera
tions Management 21:259–79.
© 2009 by Taylor & Francis Group, LLC
Literature Review 33
23. Guide, Jr., V. D. R., and Srivastava, R. 1997. An evaluation of order release strat
egies in a remanufacturing environment. Computers and Operations Research
24:37–47.
24. Guide, Jr., V. D. R., and Srivastava, R. 1997. Repairable inventory theory: Models
and applications. European Journal of Operations Research 102:1–20.
25. Guide, Jr., V. D. R. 1996. Scheduling using drumbufferrope in a remanufactur
ing environment. International Journal of Production Research 34:1081–91.
26. Guide, Jr., V. D. R., Srivastava, R., and Spencer, M. S. 1997. An evaluation of
capacity planning techniques in a remanufacturing environment. International
Journal of Production Research 35:67–82.
27. Gungor, A., and Gupta, S. M. 2002. Disassembly line in product recovery. Inter
national Journal of Production Research 40:2569–89.
28. Gungor, A., and Gupta, S. M. 1999. Issues in environmentally conscious manu
facturing and product recovery: A survey. Computers and Industrial Engineering
36:811–53.
29. Gupta, S. M., and McLean, C. R. 1996. Disassembly of products. Computers and
Industrial Engineering 31:225–28.
30. Gupta, S. M., and Taleb, K. N. 1994. Scheduling disassembly. International Jour
nal of Production Research 32:1857–66.
31. Heyman, D. P. 1997. Optimal disposal policies for a single item inventory sys
tem with returns. Naval Research Logistics Quarterly 24:385–405.
32. Hoshino, T., Yura, K., and Hitomi, K. 1995. Optimization analysis for recycle
oriented manufacturing systems. International Journal of Production Research
33:855–60.
33. Hu, T., Sheu, J., and Huan, K. 2002. A reverse logistics cost minimization model
for the treatment of hazardous wastes. Transportation Research 38E:457–73.
34. Imtanavanich, P., and Gupta, S. M. 2004. Multicriteria decision making for dis
assemblytoorder system under stochastic yields. In Proceedings of SPIE—The
International Society for Optical Engineering, Environmentally Conscious Manufac
turing IV, vol. 5583, pp. 147–62.
35. Imtanavanich, P., and Gupta, S. M. 2005. Multicriteria decision making
approach in multiple periods for a disassemblytoorder system under prod
uct’s deterioration and stochastic yields. In Proceedings of SPIE—The Interna
tional Society for Optical Engineering, Environmentally Conscious Manufacturing V,
vol. 5997, pp. 599–702.
36. Inderfurth, K. 1997. Simple optimal replenishment and disposal policies for a
product recovery system with leadtimes. OR Spectrum 19:111–22.
37. Jayaraman, V., Guide, Jr., V. D. R., and Srivastava, R. 1999. A closedloop logis
tics model for remanufacturing. Journal of the Operational Research Society
50:497–508.
38. Kelle, P., and Silver, E. A. 1989. Purchasing policy of new containers considering
the random returns of previously issued containers. IIE Transactions 21:349–54.
39. Kiesmuller, G. P., and Van der Laan, E. 2001. An inventory model with depen
dent product demands and returns. International Journal of Production Economics
72:73–87.
40. Kongar, E., and Gupta, S. M. 2002. A multicriteria decision making approach for
disassemblytoorder systems. Journal of Electronics Manufacturing 11:171–83.
41. Korugan, A., and Gupta, S. M. 1998. A multiechelon inventory system with
returns. Computers and Industrial Engineering 35:145–48.
© 2009 by Taylor & Francis Group, LLC
34 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
42. Korugan, A., and Gupta, S. M. 2001. An adaptive kanban control mechanism
for a single stage hybrid system. In Proceedings of the International Conference on
Environmentally Conscious Manufacturing II, pp. 175–82.
43. Krikke, H. R., Van Harten, A., and Schuur, P. C. 1999. Business case: Reverse
logistic network redesign for copiers. OR Spectrum 21:381–409.
44. Krikke, H., BloemhofRuwaard, J., and Van Wassenhove, L. N. 2003. Concurrent
product and closedloop supply chain design with an application to refrigera
tors. International Journal of Production Research 41:3689–719.
45. Krikke, H., Leblanc, I., and van de Velde, S. 2004. Product modularity and the
design of closedloop supply chains. California Management Review 46:23–39.
46. Kroon, L., and Vrijens, G. 1995. Returnable containers: An example of reverse
logistics. International Journal of Physical Distribution and Logistics Management
25:56–68.
47. Krupp, J. 1993. Structuring bills of material for automobile remanufacturing.
Production and Inventory Management Journal 34:46–52.
48. Lambert, A. J. D., and Gupta, S. M. 2002. Demanddriven disassembly optimi
zation for electronic products. Journal of Electronics Manufacturing 11:121–35.
49. Lambert, A. J. D., and Gupta, S. M. 2005. Disassembly modeling for assembly, main
tenance, reuse, and recycling. Boca Raton, FL: CRC Press.
50. Lambert, A. J. D. 1997. Optimal disassembly of complex products. International
Journal of Production Research 35:2509–23.
51. Lieckens, K., and Vandaele, N. 2007. Reverse logistics network design with sto
chastic lead times. Computers and Operations Research 34:395–416.
52. Lim, G. H., Kasumastuti, R. D., and Piplani, R. 2005. Designing a reverse sup
ply chain network for product refurbishment. In Proceedings of the International
Conference of Simulation and Modeling.
53. Listes, O., and Dekker, R. 2005. A stochastic approach to a case study for a
product recovery network design. European Journal of Operational Research
160:268–87.
54. Listes, O. 2007. A generic stochastic model for supplyandreturn network
design. Computers and Operations Research 34:417–42.
55. Louwers, D., Kip, B. J., Peters, E., Souren, F., and Flapper, S. D. P. 1999. A facility
location allocation model for reusing carpet materials. Computers and Industrial
Engineering 36:855–69.
56. Lu, Z., and Bostel, N. 2007. A facility location model for logistics system includ
ing reverse fows: The case of remanufacturing activities. Computers and Opera
tions Research 34:299–323.
57. Mabini, M. C., Pintelon, L. M., and Gelders, L. F. 1992. EOQ type formulation
for controlling repairable inventories. International Journal of Production Econom
ics 28:21–33.
58. Meade, L., Sarkis, J., and Presley, A. 2007. The theory and practice of reverse
logistics. International Journal of Logistics Systems and Management 3:56–84.
59. Moyer, L. K., and Gupta, S. M. 1997. Environmental concerns and recycling/dis
assembly efforts in the electronic industry. Journal of Electronics Manufacturing
7:1–22.
60. Muckstadt, J., and Isaac, M. H. 1981. An analysis of single item inventory sys
tems with returns. Naval Research Logistics Quarterly 28:237–54.
61. Nagurney, A., and Toyasaki, F. 2005. Reverse supply chain management and
electronic waste recycling: A multitiered network equilibrium framework for
ecycling. Transportation Research 41E:1–28.
© 2009 by Taylor & Francis Group, LLC
Literature Review 35
62. Panisset, B. 1988. MRP II for repair/refurbish industries. Production and Inven
tory Management Journal 29:12–15.
63. Penev, K. D., and De Ron, A. J. 1996. Determining of disassembly strategy. Inter
national Journal of Production Research 34:495–506.
64. Pnueli, Y., and Zussman, E. 1997. Evaluating the endoflife value of a prod
uct and improving it by redesign. International Journal of Production Research
35:921–42.
65. Ravi, V., Ravi, S., and Tiwari, M. K. 2005. Analyzing alternatives in reverse
logistics for endoflife computers: ANP and balanced scorecard approach.
Computers and Industrial Engineering 48:327–56.
66. Reimer, B., Sodhi, M. S., and Knight, W. A. 2000. Optimizing electronics endof
life disposal costs. In Proceedings of IEEE Symposium on Electronics and the Envi
ronment, pp. 342–47.
67. Richter, K. 1997. Pure and mixed strategies for the EOQ repair and waste dis
posal problem. OR Spectrum 19:123–29.
68. Richter, K. 1996. The EOQ repair and waste disposal model with variable setup
numbers. European Journal of Operational Research 95:313–24.
69. Richter, K. 1996. The extended EOQ repair and waste disposal model. Interna
tional Journal of Production Economics 45:443–48.
70. Rubio, S., Chamarro, A., and Miranda, J. F. 2008. Characteristics of research
on reverse logistics (1999–2005). International Journal of Production Research
46:1099–1120.
71. Salema, M. I. G., BarbosaPovoa, A. P., and Novais, A. Q. 2007. An optimization
model for the design of a capacitated multiproduct reverse logistics network
with uncertainty. European Journal of Operational Research 179:1063–77.
72. Salomon, M., van der Laan, E. A., Dekker, R., Thierry, M., and Ridder, A. A. N. 1994.
Product remanufacturing and its effect on production and inventory control. ERASM Man
agement Report Series 172, Erasmus University, Rotterdam, The Netherlands.
73. Savaskan, R., and Van Wassenhove, L. N. 2006. Reverse channel design: The
case of competing retailers. Management Science 52:1–14.
74. Savaskan, R. C., Bhattacharya, S., and Van Wassenhove Luk, N. 2004. Closed
loop supply chain models with product remanufacturing. Management Science
50:239–52.
75. Simpson, V. P. 1978. Optimum solution structure for a repairable inventory
problem. Operations Research 26:270–81.
76. Spengler, T., Puchert, H., Penkuhn, T., and Rentz, O. 1997. Environmental inte
grated production and recycling management. European Journal of Operations
Research 97:308–26.
77. Teunter, R. H., and Vlachos, D. 2002. On the necessity of a disposal option for
returned items that can be remanufactured. International Journal of Production
Economics 75:257–66.
78. Teunter, R. H. 2001. Economic ordering quantities for remanufacturable item
inventory system. Naval Research Logistics 48:484–95.
79. Thierry, M., Salomon, M., van Nunen, J., and van Wassenhove, L. N. 1995. Stra
tegic issues in product recovery management. California Management Review
37:114–35.
80. Udomsawat, G. and Gupta, S. M. 2008. MultiKanban system for disk assembly
line. In S. M. Gupta and A. J. D. Lambert (Eds.), Environment conscious manufac
turing. Boca Raton, FL: CRC Press..
© 2009 by Taylor & Francis Group, LLC
36 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
81. Uzsoy, R. 2007. Production planning for companies with product recovery and
remanufacturing capability. In Proceedings of the IEEE International Symposium
on Electronics and the Environment, pp. 285–90.
82. Van der Laan, E., and Salomon, M. 1997. Production planning and inventory
control with remanufacturing and disposal. European Journal of Operational
Research 102:264–78.
83. Van der Laan, E., Dekker, R., Salomon, M., and Ridder, A. 1996. An (s, Q) inven
tory model with remanufacturing and disposal. International Journal of Produc
tion Economics 46:339–50.
84. Veerakamolmal, P., and Gupta, S. M. 1999. Analysis of design effciency for the
disassembly of modular electronic products. Journal of Electronics Manufacturing
9:79–95.
85. Vlachos, D., Georgiadis, P., and Iakovou, E. 2007. A systems dynamic model for
dynamic capacity planning of remanufacturing in closedloop supply chains.
Computers and Operations Research 34:367–94.
86. Wojanowski, R., Verter, V., and Boyaci, T. 2007. Retailcollection network design
under depositrefund. Computers and Operations Research 34:324–45.
© 2009 by Taylor & Francis Group, LLC
37
4
Quantitative Modeling Techniques
4.1 Introduction
In this book, quantitative models for a number of crucial issues in strategic
planning of reverse and closedloop supply chains are presented. Depend
ing on the decisionmaking situation, each model makes use of one or more
of the quantitative techniques introduced in this chapter. It must be noted
that only the basic concepts of each technique are presented here.
This chapter is organized as follows: Sections 4.2 through 4.19 introduce
the concepts of analytic hierarchy process (including eigen vector method),
analytic network process, fuzzy logic, extent analysis method, fuzzy multi
criteria analysis method, quality function deployment, method of total pref
erences, linear physical programming, goal programming, technique for
order preference by similarity to ideal solution (TOPSIS), Borda’s choice rule,
expert systems, Bayesian updating, Taguchi loss function, Six Sigma, neural
networks, geographical information systems, and linear integer program
ming, respectively. Finally, section 4.20 gives some conclusions.
4.2 Analytic Hierarchy Process and Eigen Vector Method
The analytic hierarchy process (AHP) is a tool supported by simple math
ematics that enables decision makers to explicitly weigh tangible and intan
gible criteria against each other for the purpose of resolving confict or
setting priorities. The process has been formalized by Saaty [1] and is used
in a wide variety of problem areas, for example, siting landflls [2], evaluat
ing employee performance [3], and selecting a doctoral program [4].
In a large number of cases (for example, [5]), the tangible and intangible
criteria are considered independent of each other; in other words, those crite
ria do not in turn depend upon subcriteria and so on. The AHP in such cases
is conducted in two steps: (1) weigh independent criteria, each of which can
compare two or more decision alternatives, using pairwise judgments, and
© 2009 by Taylor & Francis Group, LLC
38 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
(2) compute the relative ranks of decision alternatives using pairwise judg
ments with respect to each independent criterion.
1. Computation of relative weights of criteria: AHP enables a person to
make pairwise judgments of importance between independent
criteria with respect to the scale shown in table 4.1. The resulting
matrix of comparative importance values is used to weigh the inde
pendent criteria by employing mathematical techniques like eigen
value, mean transformation, or row geometric mean. This step is
called the eigen vector method if an eigen vector is the employed
mathematical technique.
2. Computation of the relative ranks: Pairwise judgments of importance
using the scale shown in table 4.1 are computed for the decision alter
natives as well. These judgments are obtained with respect to each
independent criterion considered in step 1. The resulting matrix of
comparative importance values is used to rank the decision alterna
tives by employing mathematical techniques like eigen value, mean
transformation, or row geometric mean.
The degrees of consistency of pairwise judgments in steps 1 and 2 are
measured using an index called the consistency ratio (CR). Perfect consis
tency implies a value of zero for CR. However, perfect consistency cannot be
demanded because, as human beings, we are often biased and inconsistent
in our subjective judgments. Therefore, it is considered acceptable if CR is
less than or equal to 0.1. For CR values greater than 0.1, the pairwise judg
ments must be revised before the weights of criteria and the ranks of deci
sion alternatives are computed. CR is computed using the formula
CR
n
n R
=
~
~
( max )
( )( )
ì
1
(4.1)
where Mmax is the principal eigen value of the matrix of comparative impor
tance values, n is the number of rows (or columns) in the matrix, and R is
TABLE 4.1
Scale for pairwise judgments
Comparative
importance Defnition
1 Equally important
3 Moderately more important
5 Strongly important
7 Very strongly more important
9 Extremely more important
2, 4, 6, 8 Intermediate judgment values
© 2009 by Taylor & Francis Group, LLC
Quantitative Modeling Techniques 39
the random index for each n value that is greater than or equal to 1. Table 4.2
shows various R values for n values ranging from 1 to 10.
The AHP is illustrated in the form of a hierarchy of three levels, where the
frst level contains the primary objective, the second level contains the inde
pendent criteria, and the last level contains the decision alternatives. Also, an
important feature of the AHP is that the tangible and intangible criteria in
the second level must be chosen in such a way that they can somehow help
the decision maker in comparing two or more decision alternatives.
4.3 Analytic Network Process
Analytic network process [6] (ANP) generalizes the AHP. AHP assumes inde
pendence among the criteria and subcriteria considered in the decision mak
ing, but reallife situations warrant against such assumption. ANP allows for
dependence within a set of criteria (inner dependence) as well as between
sets of criteria (outer dependence); therefore, ANP goes beyond AHP [7].
Whereas AHP assumes a unidirection hierarchical relationship among the
decision levels, ANP allows for a more complex relationship among decision
levels and attributes, as it does not require a strict hierarchical structure. The
looser network structure in ANP allows the representation of any decision
problem, irrespective of which criteria come frst or which come next. Com
pared to AHP, ANP requires more calculations and requires a more careful
track of the pairwise judgment matrices. ANP is used in a wide variety of
problem areas (for example, analyzing alternatives in reverse logistics for
endoflife computers [7]; evaluating connection types in design for disas
sembly [8]; and modeling the metrics of lean, agile, and leagile supply chain
[9]). The steps involved in the ANP methodology are as follows:
Step 1: Model development and problem formulation: In this step, the deci
sion problem is structured into its constituent components. The
relevant criteria, the subcriteria, and alternatives are chosen and
structured in the form of a control hierarchy.
Step 2: Pairwise comparisons: In this step, the decision maker is asked
to carry out a series of pairwise comparisons with respect to the
scale shown in table 4.1, where two main criteria are simultaneously
compared with respect to the problem objective, two subcriteria are
TABLE 4.2
Random index value for each n value
n 1 2 3 4 5 6 7 8 9 10
R 0 0 0.58 0.90 1.12 1.24 0.32 1.41 1.45 1.49
© 2009 by Taylor & Francis Group, LLC
40 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
simultaneously compared with respect to their main criteria, and
pairwise comparisons are performed to address the interdepen
dencies among the subcriteria. The relative matrix of comparative
importance values is then used to weigh the criteria using math
ematical techniques like eigen vector, mean transformation, or row
geometric mean.
Step 3: Super matrix formulation: The super matrix allows for a resolu
tion of interdependencies that exist among the subcriteria. It is a
partitioned matrix, where each submatrix is composed of a set of
relationships between and within the levels as represented by the
decision maker’s model. The super matrix M is made to converge to
obtain a longterm stable set of weights. For convergence, M must be
made column stochastic; that is done by raising M to the power of
2
k+1
, where k is an arbitrarily large number.
Step 4: Selection of the best alternative: The selection of the best alterna
tive depends on the desirability index. The desirability index, DI, for
alternative i is defned as
DI P A A S
j kj
D
kj
I
ikj
k
K
j
J j
=
= ÷
` `
1 1
(4.2)
where P
j
is the relative importance weight of main criterion j; A
kj
D
is the relative importance weight for subcriterion k of main crite
rion j for the dependency (D) relationships among subcriteria; A
kj
I
is
the stabilized relative importance weight (determined by the super
matrix) for subcriterion k of main criterion j for interdependency
(I) relationships among subcriteria; and S
ikj
is the relative impact of
alternative i on subcriterion k of main criterion j.
4.4 Fuzzy Logic
Expressions such as “probably so,” “not very clear,” and “very likely,” which
are often heard in daily life, carry a touch of imprecision with them. This
imprecision or vagueness in human judgments is referred to as fuzziness in
the scientifc literature. As the decisionmaking problem’s intensity grows,
this imprecision leads to results that can often be misleading, if the fuzziness
is not taken into account. Zadeh [10] frst proposed fuzzy logic, after which
an increasing number of studies have dealt with fuzziness in problems by
applying fuzzy logic.
© 2009 by Taylor & Francis Group, LLC
Quantitative Modeling Techniques 41
When dealing with factors with uncertain or imprecise values, people use
linguistic values like high, low, good, and medium, to describe those factors.
For example, height may be a factor with an imprecise value, so its linguistic
value can be “very tall” or “very short.” Fuzzy logic is primarily concerned
with quantifying vagueness in human perceptions and thoughts. The tran
sition from vagueness to quantifcation is performed by the application of
fuzzy logic, as shown in fgure 4.1.
Zadeh proposed a membership function to deal with quantifying vague
ness. Each quantifed linguistic value is associated with a grade membership
value belonging to the interval [0, 1] by means of a membership function.
Thus, a fuzzy set can be defned as \ = = x X x
A
, ( ) [ , ] M 0 1 , where M
A
is the
degree of membership, ranging from 0 to 1, of a quantity x of the linguistic
value, A, over the universe of quantifed linguistic values, X. X is essentially
a set of real numbers. The more x fts A, the larger the degree of member
ship of x. If a quantity has a degree of membership equal to 1, this refects a
complete ftness between the quantity and the linguistic value. On the other
hand, if the degree of membership of a quantity is 0, then that quantity does
not belong to the linguistic value. The membership function looks like a typ
ical cumulative probability function; however, the value of a membership
function represents the possibility of a fuzzy event, whereas the value of a
cumulative probability function represents the cumulative probability of a
statistical event.
A triangular fuzzy number (TFN) [11] is a fuzzy set with three parameters
(l, m, u), each representing a quantity of a linguistic value associated with a
degree of membership of either 0 or 1. Figure 4.2 shows a graphical depic
tion of a TFN. The parameters l, m, and u denote the smallest possible, most
promising, and largest possible quantities that describe the linguistic value.
Vagueness Quantiﬁcation
Fuzzy Logic
FIGURE 4.1
Application of fuzzy logic.
Parameters of TFN
l m
Degree of Membership
0.0
1.0
FIGURE 4.2
Triangular fuzzy number.
© 2009 by Taylor & Francis Group, LLC
42 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
Each TFN, P, has linear representations on its left and righthand side such
that its membership function can be defned as
M
P
x l
x l m l l x m
u x u m m x u
=
<
÷ ÷ _ _
÷ ÷ _ _
0
0
,
( ) / ( ),
( ) / ( ),
,, x u
'
¦
1
1
1
1
1
1
1
+
1
1
1
1
1
1
1
(4.3)
For each quantity x increasing from l to m, the corresponding membership
function linearly increases from 0 to 1, and while x increases from m to u, the
corresponding membership function decreases linearly from 1 to 0.
The basic operations on TFNs are as follow [11, 12]; for example, P
1
= (l, m,
u) and P
2
= (x, y, z):
P P l x m y u z
1 2
+ = + + + ( , , )
addition (4.4)
P P l z m y u x
1 2
~ = ~ ~ ~ ( , , )
subtraction (4.5)
P P l x m y u z
1 2
× = × × × ( , , )
multiplication (4.6)
P
P
l
z
m
y
u
x
1
2
=
(
(
·
·
·
·
\
)
, ,
division (4.7)
Defuzzifcation is a technique to convert a fuzzy number into a crisp real
number. There are several methods available for this purpose. The centerof
area method [13] converts a fuzzy number P = (l, m, u) into a crisp number
Q, where
Q
u l m l
l =
~ + ~
+
( ) ( )
3
(4.8)
Defuzzifcation might be necessary in two situations: (1) when comparison
between two fuzzy numbers is diffcult to perform, and (2) when a fuzzy
number to be operated on has negative parameters (in other words, we make
sure that upon performing an arithmetic operation on a TFN, we get a TFN
only; for example, squaring the TFN (–1, 0, 1) using equation (4.6) leads to (1, 0,
1), which is not a TFN, and hence we defuzzify (–1, 0, 1) before squaring it).
© 2009 by Taylor & Francis Group, LLC
Quantitative Modeling Techniques 43
4.5 Extent Analysis Method
Chang [14] proposed a new approach to handle situations that require use
of both fuzzy logic and AHP/ANP. First, TFNs (see section 4.4) are used for
pairwise comparisons; then by using extent analysis method [15], the syn
thetic extent value of the pairwise comparison is introduced, and by apply
ing the principle of comparison of fuzzy numbers, the weight vectors with
respect to each element under a certain criterion can be computed. The steps
involved in the methodology are as follows.
Let X = {x
1
, x
2
, …, x
n
} be an object set and U = {u
1
, u
2
, …, u
m
} be a goal set.
According to the extent analysis method, each object is taken and an extent
analysis for each goal, g
i
, is performed. Therefore, m extent analysis values
for each object can be obtained, with the following signs: M
1
gi
, M
2
gi
, …, M
m
gi
,
i = 1, 2, …, n, where all the M
j
gi
(j = 1, 2, …, m) are TFNs.
Step 1: The value of fuzzy synthetic extent with respect to the ith object
is defned as
S M M
i gi
j
j
m
j
m
gi
j
i
n
= .
'
¦
1
1
1
+
1
1
1
'
¦
1
1
1
+
1
= = =
` ` `
1 1 1
11
1
÷1
(4.9)
In order to obtain M
gi
j
j
m
=
`
1
, perform the fuzzy addition operation of
m extent analysis values for a particular matrix such that
M l m u
gi
j
j j j
j
m
j
m
j
m
=
(
(
·
·
·
·
·
\
)
= = =
` ` `
, ,
1 1 1
=
`
j
m
1
(4.10)
To obtain
j
m
gi
j
i
n
M
= =
÷
` `
'
¦
1
1
1
+
1
1
1
'
¦
1
1
1
+
1
1
1
1 1
1
,
perform the fuzzy addition operation of M
gi
j
(j = 1, 2, …, m) values such
that
© 2009 by Taylor & Francis Group, LLC
44 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
M l m u
gi
j
j
m
i
n
i i i
i
n
i
n
i
n
= = = = =
` ` ` ` `
=
(
(
·
·
·
1 1 1 1 1
, ,
··
·
\
)
(4.11)
and then compute the inverse of the vector.
Step 2: The degree of possibility of M
2
= (l
2
, m
2
, u
2
) ≥ M
1
= (l
1
, m
1
, u
1
) is
expressed as
V(M
2
≥ M
1
)
= hgt (M
1
≥ M
2
)
= {1, if m
2
≥ m
1
; 0, if l
1
≥ u
2
; (l
1
– u
2
)/((m
2
– u
2
) – (m
1
– l
1
))} (4.12)
To compare M
1
and M
2
, both V(M
2
≥ M
1
) and V(M
1
≥ M
2
) are required.
Step 3: The degree of possibility for a convex fuzzy number to be greater
than k convex fuzzy numbers M
i
(i = 1, 2, …, k) can be defned as
V (M ≥ M
1
, M
2
, M
k
)
= V [(M ≥ M
1
) and (M ≥ M
2
) and … and (M ≥ M
k
)]
= min V (M≥ M
i
), i =1, 2, k (4.13)
Let d’ (A
i
) = min V (S
i
≥ S
k
), for k = 1, 2, n; k ≠ i. Then the weight vector
is given by
W’ = (d’ (A
1
), d’ (A
2
), d’ (A
n
))
T
(4.14)
Step 4: The weight vector obtained in step 3 is normalized to get the
normalized weights.
These steps are applied to deduce the weights of main criteria with respect
to the goal, subcriteria with respect to the main criteria, and alternatives
with respect to the main and subcriteria. The rest of the procedure is similar
to the traditional AHP/ANP.
4.6 Fuzzy Multicriteria Analysis Method
Multicriteria analysis problems require the decision maker to make qualita
tive assessments regarding the performance of the decision alternatives with
respect to each independent criterion and the relative importance of each
independent criterion with respect to the overall objective of the problem. As
a result, uncertain subjective data are present that make the decisionmaking
© 2009 by Taylor & Francis Group, LLC
Quantitative Modeling Techniques 45
process complex [16]. AHP enables a person to make pairwise judgments of
importance between the independent criteria as well as the decision alterna
tives. However, traditional AHP is criticized for its unbalanced scale of judg
ment and failure to precisely handle the inherent uncertainty and vagueness
in carrying out pairwise comparisons.
Deng [17] proposed a multicriteria analysis approach that extends Saa
ty’s AHP (see section 4.2) to deal with the imprecision and subjectiveness
in the pairwise comparisons. TFNs (see section 4.4) are used for pairwise
comparisons, and the concept of extent analysis method (see section 4.5)
is applied to solve the fuzzy reciprocal matrix for determining the criteria
importance and alternative performance. The Bcut concept is used to trans
form the fuzzy performance matrix representing the overall performance of
all alternatives with respect to each criterion into an interval performance
matrix. An overall performance index for each alternative across all criteria
that incorporates the decision maker’s attitude toward risk is obtained by
applying the concept of similarity to the ideal solution [18] using the vector
matching function.
The selection process starts with determining the criteria of importance
and performance of alternatives. By using TFNs, a fuzzy reciprocal matrix
for criteria importance (W) or alternative performance with respect to a spe
cifc criterion (C
j
) can be determined as
W C
a a a
a a a
j
k
k
...
...
... ... .. ...
or =
11 12 1
21 22 2
aa a a
k k kk 1 2
...
l
l
l
l
l
l
l
l
l
(4.15)
where
a
l s
l s l s k k m or n
ls
=
<
= = =
1 3 5 9
1 1 2
, , , , ,
, , , , , ... ; , , ,,
/
,
1 a l s
sl
(
(
·
·
·
·
·
·
·
·
·
·
·
(4.16)
By applying the extent analysis method, the corresponding criteria weights
(w
j
) or alternative performance ratings (x
ij
) with respect to a specifc criterion
C
j
can be determined as
x w a a
ij j
ls ls
s
k
l
k
s
k
or = ÷
= = =
` ` `
1 1 1
(4.17)
© 2009 by Taylor & Francis Group, LLC
46 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
where i = 1, 2, …, n; j = 1, 2, …, m; and k = m or n depending on whether
the reciprocal judgment matrix is for assessing the performance ratings of
alternatives or weights of criteria involved. The decision matrix (X) and the
weight vector (W) can be respectively determined as
X
x x x
x x x
x x
m
m
n n
=
11 12 1
21 22 2
1
...
...
... ... ... ...
22
... x
nm
l
l
l
l
l
l
l
l
l
(4.18)
W =( , , ..., ) w w w
m 1 2
(4.19)
where x
ij
represents the resultant fuzzy performance assessment of alterna
tive A
i
(i = 1, 2, n) with respect to criterion C
j
, and w
j
is the resultant fuzzy
weight of criterion C
j
(j = 1, 2, …, m) with respect to the overall goal of the
problem. A fuzzy performance matrix Z representing the overall perfor
mance of all alternatives with respect to each criterion is obtained by multi
plying the weight vector by the decision matrix.
Z
w x w x w x
w x w x w x
m m
m m
=
1 11 2 12 1
1 21 2 22 2
...
...
... .... ... ...
... w x w x w x
n n m nm 1 1 2 2
l
l
l
l
l
l
l
l
ll
(4.20)
An interval performance matrix [16] is derived by using an Bcut on the
performance matrix, where 0 ≤ B ≤ 1. The value of B represents the decision
maker’s degree of confdence in his or her fuzzy assessments regarding the
alternative ratings and criteria weights. The larger the value of B, the more
confdent the decision maker is about the fuzzy assessments, viz., the assess
ments are closer to the most possible value a
2
of the triangular fuzzy number
(a
1
, a
2
, a
3
).
Z
z z z z z
l r l r m
A
A A A A
=
l
l
l
l
l
l
11 11 12 12 1
, , ...
ll mr
n l
z
z
A E
,
... ... ... ...
... ... ... ...
1
1
l
l
l
AA A A A A
, , ... , z z z z z
n r n l n r nml nm 1 2 2
l
l
l
l
l
l
rr
A
l
l
l
l
l
l
l
l
l
l
l
l
l
..
(4.21)
An overall crisp performance matrix that incorporates the decision maker’s
attitude toward risk, using an optimism index M (M = 1 implies the decision
© 2009 by Taylor & Francis Group, LLC
Quantitative Modeling Techniques 47
maker has an optimistic view, 0 implies a pessimistic view, and 0.5 implies a
moderate view), is calculated.
z
z z z
m
A
L
A
L
A
L
A
L
'
' ' '
...
... ... ... ...
... ..
=
11 12 1
.. ... ...
...
' ' '
z z z
n n nm 1 2 A
L
A
L
A
L
l
l
l
l
ll
l
l
l
l
l
(4.22)
where
z z z
ij ijr ijl A
L A A
L L L
'
( ) , [ , ] = ÷ ÷ = 1 0 1
(4.23)
A normalized performance matrix with respect to each criterion is calcu
lated from equation (4.22).
Z
z z z
z z z
m
m
A
A
L
A
L
A
L
A
L
A
L
A
L
=
11 12 1
21 22 2
...
...
... .... ... ...
... z z z
n n nm 1 2 A
L
A
L
A
L
l
l
l
l
l
l
l
ll
l
(4.24)
where
z z z
ij ij ij
i
n
A
L
A
L
A
L
= ÷
=
`
' '
( )
2
1
(4.25)
Zeleny [18] introduced the concept of ideal solution in multiattribute deci
sion analysis that was further extended by Hwang and Yoon [19], including
negative solution to avoid the worst decision outcome. In line with this con
cept, the positive and negativeideal solutions, respectively, can be deter
mined by selecting maximum and minimum values across all alternatives
with respect to each criterion as follows:
A z z z
A z z
m A
L
A
L
A
L
A
L
A
L
A
L
A
÷ ÷ ÷ ÷
÷ ÷
=
=
( , , ..., )
( ,
1 2
1 2
LL
A
L ÷ ÷
, ...., ) z
m
(4.26)
© 2009 by Taylor & Francis Group, LLC
48 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
where
z z z z
z z
j j j nj
j j
A
L
A
L
A
L
A
L
A
L
÷
÷
=
=
max( , , ..., )
min(
2
AA
L
A
L
A
L
, , ...., ) z z
j nj 2
(4.27)
By applying the vectormatching function, the degree of similarity between
each alternative and the positive and negativeideal solutions can be calcu
lated as
S A A A A A A
S
i i i i i
i
o
ì
o
ì
o
ì
o
ì
o
ì
o
ì
o
ì
o
ì
+ + + +
~
= / max( , )
==
~ ~ ~
A A A A A A
i i i i o
ì
o
ì
o
ì
o
ì
o
ì
o
ì
/ max( , )
(4.28)
where A z z z
i i i im o
ì
o
ì
o
ì
o
ì
= ( , , ...., )
1 2
is the ith row of the overall performance matrix,
which represents the corresponding performance of alternative A
i
with
respect to criterion C
j
. The larger the value of S S
i i o
ì
o
ì + ~
, , the higher the degree
of similarity between each alternative and the positiveideal and negative
ideal solutions [20]. A preferred alternative should have a higher degree of
similarity to the positiveideal solution and a lower degree of similarity to
the negativeideal solution. Hence, an overall performance index for each
alternative with the decision maker’s B level of confdence and M degree of
optimism toward risk can be determined as
P S S S i n
i i i i A
L
A
L
A
L
A
L
= ÷ ÷ .
÷ ÷
/ ( ), , , , 1 2
(4.29)
The larger the performance index, the most preferred the alternative is.
4.7 Quality Function Deployment
Erol and Ferrell [21] defne performance aspects as the features that the deci
sion maker wishes to consider in the selection process and enablers as the
characteristics possessed by the alternatives, which can be used to satisfy the
performance aspects.
The absolute technical importance ratings (ATIRs), which measure how
effectively each enabler can satisfy all of the performance aspects, are com
puted by
ATIR =1, ..., j
i
I
d R j J
i ij
= \
=
`
1
(4.30)
© 2009 by Taylor & Francis Group, LLC
Quantitative Modeling Techniques 49
where d
i
is the importance value of performance aspect i relative to the other
performance aspects, and R
ij
is the relationship score for performance aspect
i and enabler j. Because there is an ATIR for each enabler j, for the compari
son of all enablers, it is normalized to form the relative technical importance
rating (RTIR
j
) as follows:
RTIR
ATIR
ATIR
=1, ..., j
j
j
j
J
j J = \
=
`
1
(4.31)
4.8 Method of Total Preferences
RTIRs (see section 4.7), together with additional human expert opinions, are
used to develop a single measure that refects the rating of each alternative
as follows [21]:
TUP RTIR WA
n j nj
j
J
n = \
=
`
1
(4.32)
where TUP
n
is the total user preference for alternative n, and WA
nj
is the
(defuzzifed) degree to which alternative n can deliver enabler j.
For the purpose of comparison of all alternatives, TUP of each alternative
is then normalized as follows:
NTUP
TUP
TUP
n
n
n
n
N
n = \
=
`
1
(4.33)
where NTUP
n
is the normalized total preference for alternative n, and N is
the total number of alternatives.
The alternative with the highest NTUP is considered the one with the
highest potential.
4.9 Linear Physical Programming
In the linear physical programming (LPP) method [22], four distinct classes
(1S, 2S, 3S, and 4S) are used to allow the decision maker to express his or her
© 2009 by Taylor & Francis Group, LLC
50 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
preferences for the value of each criterion (for decision making) in a more
detailed, quantitative, and qualitative way than when using a weightbased
method like analytic hierarchy process (see section 4.2). These classes are
defned as follows: smaller is better (1S), larger is better (2S), value is better
(3S), and range is better (4S). Figure 4.3 depicts these different classes.
The value of the pth criterion, g
p
, for evaluating the alternative of interest is
categorized according to the preference ranges shown on the horizontal axis.
Consider, for example, the case of class 1S. The preference ranges are:
Class1S
"Smaller is Better"
Cla ss 2S
"Larger is Better"
Cla ss 3S
"Value is Better"
I
d
e
a
l
D
e
s
i
r
a
b
l
e
T
o
l
e
r
a
b
l
e
U
n
d
e
s
i
r
a
b
l
e
H
i
g
h
l
y
u
n
d
e
s
i
r
a
b
l
e
t

p5
t

p4
t

p3
t

p2
t

p1
z
5
z
4
z
3
z
2
z
1
U
n
a
c
c
e
p
t
a
b
l
e
D
e
s
i
r
a
b
l
e
T
o
l
e
r
a
b
l
e
U
n
d
e
s
i
r
a
b
l
e
H
i
g
h
l
y
u
n
d
e
s
i
r
a
b
l
e
U
n
a
c
c
e
p
t
a
b
l
e
Class4S
"Range is Better"
z
p
t
+
p1
t
+
p2
t
+
p3
t
+
p4
t
+
p5
g
p
(x)
U
n
a
c
c
e
p
t
a
b
l
e
I
d
e
a
l
D
e
s
i
r
a
b
l
e
T
o
l
e
r
a
b
l
e
U
n
d
e
s
i
r
a
b
l
e
H
i
g
h
l
y
u
n
d
e
s
i
r
a
b
l
e
D
e
s
i
r
a
b
l
e
T
o
l
e
r
a
b
l
e
U
n
d
e
s
i
r
a
b
l
e
H
i
g
h
l
y
u
n
d
e
s
i
r
a
b
l
e
U
n
a
c
c
e
p
t
a
b
l
e t
p1
t
+
p2
t
+
p3
t
+
p4
t
+
p5
g
p
(x) t

p5
t

p4
t

p3
t

p2
U
n
a
c
c
e
p
t
a
b
l
e
H
i
g
h
l
y
u
n
d
e
s
i
r
a
b
l
e
U
n
d
e
s
i
r
a
b
l
e
T
o
l
e
r
a
b
l
e
D
e
s
i
r
a
b
l
e
I
d
e
a
l
t

p5
t

p4
t

p3
t

p2
t

p1
g
p
(x)
U
n
a
c
c
e
p
t
a
b
l
e
H
i
g
h
l
y
u
n
d
e
s
i
r
a
b
l
e
U
n
d
e
s
i
r
a
b
l
e
T
o
l
e
r
a
b
l
e
D
e
s
i
r
a
b
l
e
I
d
e
a
l
t
+
p1
t
+
p2
t
+
p3
t
+
p4
t
+
p5
g
p
(x)
z
5
z
4
z
3
z
2
z
1
z
5
z
4
z
3
z
2
z
1
z
5
z
4
z
3
z
2
z
1
z
p
z
p
z
p
FIGURE 4.3
Soft class functions for linear physical programming.
© 2009 by Taylor & Francis Group, LLC
Quantitative Modeling Techniques 51
Ideal range: g t
p p
_
÷
1
Desirable range: t g t
p p p 1 2
÷ ÷
_ _
Tolerable range: t g t
p p p 2 3
÷ ÷
_ _
Undesirable range: t g t
p p p 3 4
÷ ÷
_ _
Highly undesirable range: t g t
p p p 4 5
÷ ÷
_ _
Unacceptable range: g t
p p
_
÷
5
The quantities t
p1
+
through t
p5
+
represent the physically meaningful values
that quantify the preferences associated with the pth generic criterion. Con
sider, for example, the cost criterion for class 1S. The decision maker could
specify a preference vector by identifying t
p1
+
through t
p5
+
in dollars as (10
20 30 40 50). Thus, an alternative having a cost of $15 would lie in the desir
able range, an alternative with a cost of $45 would lie in the highly undesir
able range, and so on. We can accomplish this for a nonnumerical criterion
such as color as well by (1) specifying a numerical preference structure and
(2) quantitatively assigning each alternative a specifc criterion value from
within a preference range (e.g., desirable, tolerable).
The class function, Z
p
, on the vertical axis in fgure 4.3 is used to map the
criterion value, g
p
, into a real, positive, and dimensionless parameter (Z
p
is, in
fact, a piecewise linear function of g
p
). Such a mapping ensures that different
criteria values, with different physical meanings, are mapped to a common
scale. Consider class 1S again. If the value of a criterion, g
p
, is in the ideal
range, then the value of the class function is small (in fact, zero), whereas
if the value of the criterion is greater than t
p5
+
, that is, in the unacceptable
range, then the value of the class function is very high. Class functions have
several important properties, including (1) that they are nonnegative, con
tinuous, piecewise linear, and convex, and (2) that the value of the class func
tion, Z
p
, at a given range intersection (say, desirable–tolerable) is the same for
all class types.
Basically, ranking of the alternatives is performed in four steps, as fol
lows [23]:
Step 1: Identify criteria for evaluating each of the alternatives.
Step 2: Specify preferences for each criterion, based on one of the four classes
(see fgure 4.3).
Step 3: Calculate incremental weights: Based on the preference structures
for the different criteria, the LPP weight algorithm [22] determines
incremental weights, ^w
pr
÷
and ^w
pr
÷
(used in step 4), that represent
the incremental slopes of the class functions, Z
p
. Here, r denotes the
range intersection.
© 2009 by Taylor & Francis Group, LLC
52 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
Step 4: Calculate total score for each alternative: The formula for the total
score, J, of the alternative of interest is constructed as a weighted
sum of deviations over all ranges (r = 2 to 5) and criteria (p = 1 to P),
as follows:
J w d w d
r p
P
pr pr pr pr
= ÷
= =
÷ ÷ ÷ ÷
` `
( ) ^ ^
2
5
1
(4.34)
where P represents the total number of criteria (each belonging to
one of the four classes in fgure 4.3), ^w
pr
÷
and ^w
pr
÷
are the incre
mental weights for the pth criterion, and d
pr
+
and d
pr
~
represent the
deviations of the pth criterion value of the alternative of interest
from the corresponding target values. An alternative with a lower
total score is more desirable than one with a higher total score.
The most signifcant advantage of using LPP is that no weights need to be
specifed for the criteria for evaluation. The decision maker only needs to
specify a preference structure for each criterion, which has more physical
meaning than a physically meaningless weight that is arbitrarily assigned
to the criterion.
Note that there are no decision variables in the above ranking procedure.
LPP can be used in a problem consisting of decision variables as well, by
minimizing J in equation (4.34) and subjecting (if necessary) each criterion,
g
p
, to a constraint that falls into either one of the four classes (also called soft
classes) in fgure 4.3 or one of the following four hard classes:
Class 1H: Must be smaller, i.e., g
p
≤ t
p,max
Class 2H: Must be larger, i.e., g
p
≥ t
p,min
Class 3H: Must be equal, i.e., g
p
= t
p,val
Class 4H: Must be in range, i.e., t
p,min
≤ g
p
≤ t
p,max
4.10 Goal Programming
Linear programming [24] assumes that the objectives of an organization can
be encompassed within a single objective function, such as maximizing the
total proft or minimizing the total cost. However, this assumption is not
always realistic, and there are several cases where the management focuses
on a variety of objectives simultaneously. Goal programming provides a way
of tackling such situations.
© 2009 by Taylor & Francis Group, LLC
Quantitative Modeling Techniques 53
Goal programming (GP; see [25]), generally applied to linear problems,
deals with the achievement of specifc targets/goals. The basic approach
involves formulating an objective function for each objective and seeks a
solution that minimizes the sum (weighted sum in case of fuzzy goal pro
gramming) of the deviations of these objective functions from their respec
tive goals. To this end, several criteria are to be considered in the problem
situation on hand. For each criterion, a target value is determined. Next,
the deviation variables are introduced, which may be positive or negative
(represented by S
k
and I
k
, respectively). The negative deviation variable,
I
k
, represents the underachievement of the kth goal. Similarly, the posi
tive deviation variable, S
k
, represents the overachievement of the kth goal.
Finally, for each criterion, the desire to overachieve (minimize I
k
) or under
achieve (minimize S
k
) or to satisfy the target value exactly (minimize S
k
+ I
k
)
is articulated [26].
Goal programming problems can be categorized according to the type of
mathematical programming model. Another categorization is according to
how the goals compare in importance. In the case of preemptive goal pro
gramming, there is a hierarchy of priority levels for the goals, so the goals
of primary importance receive frst attention and so forth. In case of nonpre
emptive goal programming, all the goals are of roughly comparable impor
tance [24].
In goal programming, it is necessary to specify aspiration levels for the
goals, and the overall deviation from the aspiration levels is minimized. In
most realworld scenarios, the aspiration levels and weights/importance lev
els of goals are imprecise in nature. In such situations, fuzzy GP comes in
handy, allowing the decision maker to obtain compromising results for mul
tiple goals with varying aspiration levels. In fuzzy GP, the aspiration levels
are either in the “more is better” form or “less is better” form. [27]. A linear
membership function μ
i
that represents goal fuzziness for the “more is bet
ter” form is expressed as [28]
c
i
i i
i i
i i
i i i
if G X g
G X L
g L
if L G X g
if
=
_
÷
÷
_ _
1
0
( )
( )
( )
GG X L
i i
( )_
'
¦
1
1
1
1
1
1
1
+
1
1
1
1
1
1
1
(4.35)
while a linear membership fraction μ
i
that represents goal fuzziness for the
“less is better” form is expressed as
© 2009 by Taylor & Francis Group, LLC
54 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
c
i
i i
i i
i i
i i i
if G X g
U G X
U g
if g G X U
if
=
_
÷
÷
_ _
1
0
( )
( )
( )
GG X U
i i
( )_
'
¦
1
1
1
1
1
1
1
+
1
1
1
1
1
1
1
(4.36)
where g
i
, L
i
, and U
i
are the aspiration level, lower tolerance limit, and upper
tolerrance limit respectively for the fuzzy goal G
i
(X).
The simple form of the fuzzy GP problem with p fuzzy goals can be stated as:
Maximize V
i
i
p
( ) M M =
=
`
1
(4.37)
subject to
c c
i
i i
i i
i
i i
i i
G X L
g L
or
U G X
U g
=
÷
÷
=
÷
÷
( ) ( )
(4.38)
AX b _ (4.39)
c
i
_1 (4.40)
X
i
, M _0 (4.41)
where V( ) c is the fuzzy achievement function or fuzzy decision function.
The objective is to obtain the μ
i
value as close to 1 as possible. The weighted
additive model is widely used in GP and multiobjective optimization prob
lems to refect the relative importance of goals. In this approach, the deci
sion maker assigns weights as coeffcients of individual terms in the simple
additive fuzzy achievement function to refect their relative importance. The
objective function for the weighted additive model is expressed as
Maximize
V w
i i
i
p
( ) M M =
=
`
1
(4.42)
where w
i
is the relative weight of the ith fuzzy goal. Fibonacci numbers are
used to assign weights to the goals. Fibonacci numbers are 0, 1, 1, 2, 3, 5, 8, etc.
(the next number is a result of the summation of the previous two numbers).
© 2009 by Taylor & Francis Group, LLC
Quantitative Modeling Techniques 55
The concept is applied by starting with numbers 1 and 2. For example, for
two goals, the weights would be in the ratio of 1:2, which approximately are
0.33 and 0.66. These weights are assigned to the two goals according to their
priority levels.
In some situations, the goals/objectives are not commensurable, or the goals
are such that unless a particular goal or subset of goals is achieved, other
goals should not be considered. In such situations, the weighting scheme is
not appropriate. The problem is divided into k subproblems, where k is the
number of priority levels. In the frst subproblem, the fuzzy goals belong
ing to the frst priority level will be considered and solved using the simple
additive model. At other priority levels, the membership values achieved at
earlier priority levels are added as additional constraints. In general, the ith
subproblem becomes:
Maximize ( ) c
s pi
s
`
(4.43)
subject to
M
s
s s
s s
G L
g L
=
÷
÷
(4.44)
AX ≤ b (4.45)
( ) ( ) , , , ,
*
M M
pr pr
r j = = . ÷ 1 2 1
(4.46)
μ
s
≤ 1 (4.47)
X, μ
i
≥ 0, i = 1,2,…, p (4.48)
where (μ
s
)
pi
refers to the membership functions of the goals in the ith priority
level and (μ
*
)
pr
is the achieved membership function value in the rth (r ≤ j – 1)
priority level.
4.11 Technique for Order Preference by
Similarity to Ideal Solution (TOPSIS)
The basic concept of the TOPSIS method [19] is that the rating of the alterna
tive selected as the best from a set of different alternatives should have the
© 2009 by Taylor & Francis Group, LLC
56 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
shortest distance from the ideal solution and the greatest distance from the
negativeideal solution in a geometrical (i.e., Euclidean) sense.
The TOPSIS method evaluates the following decision matrix, which refers
to m alternatives that are evaluated in terms of n criteria [29]:
Criteria
C
1
C
2
C
3
… C
n
Alternatives w
1
w
2
w
3
… w
n
A
1
z
11
z
12
z
13
… z
1n
A
2
z
21
z
22
z
23
… z
2n
A
3
z
31
z
32
z
33
… z
3n
. …
. …
. …
. …
A
m
z
m1
z
m2
z
m3
… z
mn
where A
i
is the ith alternative, C
j
is the jth criterion, w
j
is the weight (impor
tance value) assigned to the jth criterion, and z
ij
is the rating (for example, on
a scale of 1–10, the higher the rating, the better it is) of the ith alternative in
terms of the jth criterion.
The following steps are performed:
Step 1: Construct the normalized decision matrix. This step converts the
various dimensional measures of performance into nondimensional
attributes. An element r
ij
of the normalized decision matrix R is cal
culated as follows:
r
z
z
ij
ij
ij
i
m
=
=
`
2
1
(4.49)
Step 2: Construct the weighted normalized decision matrix. A set of weights
W = (w
1
, w
2
, …, w
n
) (such that ∑w
j
= 1), specifed by the decision
maker, is used in conjunction with the normalized decision matrix
R to determine the weighted normalized matrix V defned by V =
(v
ij
) = (r
ij
w
j
).
Step 3: Determine the ideal and the negativeideal solutions. The ideal (A*)
and the negativeideal (A–) solutions are defned as follows:
A v i m
i
ij
* max = =
¦ ¦
for 1, 2, 3, .....,
= {p
1
, p
2
, p
3
, …, p
n
} (4.50)
© 2009 by Taylor & Francis Group, LLC
Quantitative Modeling Techniques 57
A v i m
i
ij ~ = =
¦ }
min for 1, 2, 3, .....,
= {q
1
, q
2
, q
3
, …, q
n
} (4.51)
With respect to each criterion, the decision maker desires to choose
the alternative with the maximum rating (it is important to note that
this choice varies with the way he or she awards ratings to the alter
natives). Obviously, A* indicates the most preferable (ideal) solution.
Similarly, A– indicates the least preferable (negativeideal) solution.
Step 4: Calculate the separation distances. In this step, the concept of the
ndimensional Euclidean distance is used to measure the separation
distances of the rating of each alternative from the ideal solution and
the negativeideal solution. The corresponding formulae are
S v p i m
i ij j *
( ) = ÷ =
`
2
for 1, 2, 3, ...,
(4.52)
where S
i*
is the separation (in the Euclidean sense) of the rating of
alternative i from the ideal solution, and
S v q i m i ij j ÷= ÷ =
`
( )
2
for 1, 2, 3, ...,
(4.53)
where S
i–
is the separation (in the Euclidean sense) of the rating of
alternative i from the negativeideal solution.
Step 5: Calculate the relative coeffcient. The relative closeness coeffcient for
alternative A
i
with respect to the ideal solution A* is defned as follows:
C
S
S S
i
i
i i
*
*
=
÷
÷
÷
(4.54)
Step 6: Rank the preference order. The best alternative can now be decided
according to the preference order of C
i*
. It is the one with the rat
ing that has the shortest distance to the ideal solution. The way the
alternatives are processed in the previous steps reveals that if an
alternative has the rating with the shortest distance to the ideal solu
tion, then that rating is guaranteed to have the longest distance to the
negativeideal solution. That means the higher the C
i*
, the better the
alternative.
© 2009 by Taylor & Francis Group, LLC
58 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
4.12 Borda’s Choice Rule
Borda proposed a method in which marks of m – 1, m – 2, …, 1, 0 are assigned
to the best, secondbest, …, worst alternatives, for each decision maker [30].
That means that a larger mark corresponds to greater preference. The Borda
score (maximized consensus mark) for each alternative is then determined
as the sum of the individual marks for that alternative, and the alternative
with the highest Borda score is declared the winner. That means that the dif
ferent decision makers unanimously choose the alternative that obtains the
largest Borda score as the most preferred one.
4.13 Expert Systems
Expert systems are computer programs that can represent human exper
tise (knowledge) in a particular domain (area of expertise) and then use a
reasoning mechanism (applying logical deduction and induction processes)
to manipulate this knowledge in order to provide advice in this domain.
Although conventional computer programs also contain knowledge, their
main function is to retrieve information and carry out statistical analysis
and numerical calculations. They do not reason with this knowledge or
make inferences as to what actions to take or conclusions to reach. Thus,
what mainly distinguishes expert systems from conventional programs is
the capability to reason with knowledge. The main components of an expert
system are the following [31]:
Knowledge base r : This is where the knowledge is stored. Typically, this
consists of a set of rules of the form: if EVIDENCE, then HYPOTH
ESIS. The knowledge is written in the knowledge base using the
syntax of what is termed the knowledge representation language (e.g.,
Lisp and Prolog) of the system.
Inference engine r : This reasons with the knowledge resident in the
knowledge base using certain mechanisms.
Reasoning mechanism r : This traces the path or the knowledge steps
used to arrive at a conclusion and can relay it back to the user as
the justifcation for this conclusion. Examples of this mechanism are
deduction (cause + rule n effect), abduction (effect + rule n cause),
and induction (cause + effect n rule).
Uncertainty modeling process r : This aids the inference engine when
dealing with uncertainty.
© 2009 by Taylor & Francis Group, LLC
Quantitative Modeling Techniques 59
A shell is an expert system that is complete except for the knowledge base
[31]. Thus, a shell includes an inference engine, a user interface for pro
gramming, and a user interface for running the system. Typically, the pro
gramming interface comprises a specialized editor for creating rules in a
predetermined format and some debugging tools. The user of the shell enters
rules in a declarative fashion (if X, then Y) and ideally should not need to be
concerned with the working of the inference engine. Expert system shells are
easy to use and allow a simple expert system to be constructed quickly.
4.14 Bayesian Updating
Bayesian updating [32] is an uncertainty modeling technique that assumes
that it is possible for an expert in a domain to guess a probability to every
hypothesis or assertion in that domain and that this probability can be
updated in light of evidence for or against the hypothesis or assertion.
Suppose the probability of a hypothesis H is P(H). Then the formula for the
odds of that hypothesis, O(H), is given by
O(H)
P(H)
1 P(H)
=
–
(4.55)
A hypothesis that is absolutely certain, i.e., has a probability of 1, has inf
nite odds. In practice, limits are often set on odds values so that, for example,
if O(H) > 1,000, then H is true, and if O(H) < 0.01, then H is false.
The standard formula for updating the odds of hypothesis H, given that
evidence E is observed, is
O(HE) (A).O(H) = (4.56)
where O(HE) is the odds of H, given the presence of evidence E, and A is the
affrms weight of E. The defnition of A is
A
P(EH)
P(E~H)
=
(4.57)
where P(EH) is the probability of E, given that H is true, and P(E~H) is the
probability of E, not given that H is true.
Bayesian updating assumes that the absence of supporting evidence is
equivalent to the presence of opposing evidence. The standard formula for
updating the odds of a hypothesis H, given that the evidence E is absent, is
© 2009 by Taylor & Francis Group, LLC
60 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
O(H~E) D).O(H) =( (4.58)
where O(H~E) is the odds of H, given the absence of evidence E, and D is the
denies weight of E. The defnition of D is
D
P(~EH)
P(~E~H)
1–P(EH)
1–P(E~H)
= =
(4.59)
If a given piece of evidence E has an affrms weight A that is greater than 1,
then its denies weight must be less than 1, and vice versa. Also, if A > 1 and D
< 1, then the presence of evidence E is supportive of hypothesis H. Similarly,
if A < 1 and D > 1, then the absence of E is supportive of H.
For example, while controlling a power station boiler, a rule “IF (tempera
ture is high) and NOT (water level is low) THEN (pressure is high)” can also be
written as “IF (temperature is high—AFFIRMS A
1
, DENIES D
1
) AND (water
level is low—AFFIRMS A
2
, DENIES D
2
) THEN (pressure is high).” Here,
A
P(Temperature is high  Pressure is high
1
=
))
P(Temperature is high  ~Pressure is high))
D
P(~Temperature is high  Pressure is hig
1
=
hh)
P(~Temperature is high  ~Pressure is higgh)
A
P(Water level is low Pressure is high)
2
=
PP(Water level is low ~Pressure is high)
D
P(~Temperature is high  Pressure is hig
2
=
hh)
P(~Temperature is high  ~Pressure is higgh)
Sometimes, evidence is neither defnitely present nor defnitely absent. For
example, if one is diagnosing a TV set that is not functioning properly, it is
not defnite if this is due to a malfunctioning picture tube. In such a case,
depending upon the value of the probability of the evidence P(E), the affrms
and denies weights are modifed using the following formulae:
A' =[2.(A1).P(E)]+2–A (4.60)
D' =[2.(1–D).P(E)]+D (4.61)
When P(E) is greater than 0.5, the affrms weight is used to calculate O(HE),
and when P(E) is less than 0.5, the denies weight is used.
© 2009 by Taylor & Francis Group, LLC
Quantitative Modeling Techniques 61
If n statistically independent pieces of evidence are found that support or
oppose a hypothesis H, then the updating equations are given by
O(HE &E &E ......E ) =(A ).(A ).(A ).....
1
2 3 1 2 3
n ...(A ).O(H)
n
(4.62)
and
O(H E & E & E ... E ) =(D ).(D ).(D )....
1 2 3 1 3
2
n . ...(D ).O(H)
n
(4.63)
A
i
and D
i
are given by equations (4.64) and (4.65), respectively.
A
P(E H)
P(E ~H)
i
i
i
=
(4.64)
D
P(~E H)
P(~E ~H)
i
i
i
=

(4.65)
4.15 Taguchi Loss Function
In traditional systems, the product is accepted if the product measurement
falls within the specifcation limits [33]. Otherwise, the product is rejected.
The quality losses occur only when the product deviates beyond the specif
cation limits, thereby becoming unacceptable. These costs tend to be constant
and relate to the costs of bringing the product back into the specifcation
range. Taguchi [33] suggests a narrower view of characteristic acceptability
by indicating that any deviation from a characteristic’s target value results
in a loss. If a characteristic measurement is the same as the target value, the
loss is zero. Otherwise, the loss can be measured using a quadratic function,
after which actions are taken to reduce systematically the variation from the
target value.
There are three types of Taguchi loss functions: “target is best” (see
fgure 4.4), “smaller is better” (see fgure 4.5), and “larger is better” (see
fgure 4.6).
If L(y) is the loss associated with a particular value of characteristic y, m is
the target value of the specifcation, and k is the loss coeffcient whose value
is constant depending on the cost at the specifcation limits and width of the
specifcation, for the “target is best” type,
© 2009 by Taylor & Francis Group, LLC
62 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
L y k y m ( ) ( ) = ÷
2
(4.66)
For the “smaller is better” type,
L y k y ( ) ( ) =
2
(4.67)
For the “larger is better” type,
L y k y ( ) / ( ) =
2
(4.68)
m
Lower Speciﬁcation Upper Speciﬁcation
T
a
g
u
c
h
i
L
o
s
s
T
a
g
u
c
h
i
L
o
s
s
FIGURE 4.4
“Target is best” Taguchi loss.
Upper Speciﬁcation
T
a
g
u
c
h
i
L
o
s
s
FIGURE 4.5
“Smaller is better” Taguchi loss.
© 2009 by Taylor & Francis Group, LLC
Quantitative Modeling Techniques 63
4.16 Six Sigma
Statistical process control techniques help managers achieve and maintain a
process distribution that does not change in terms of its mean and variance
[34]. The control limits on the control charts signal when the mean or vari
ability of the process changes. However, a process that is in statistical con
trol may not be producing outputs according to their design specifcations
because the control limits are based on the mean and variability of sampling
distribution, not the design specifcations.
Process capability refers to the ability of the process to meet the design
specifcations for an output. Design specifcations are often expressed as a
target value ( t ) and a tolerance (T). For example, the administrator of an
intensive care unit lab might have a target value for the turnaround time
of results to the attending physicians of 25 minutes and a tolerance of  5
minutes because of the need for speed under lifethreatening conditions.
The tolerance gives an upper specifcation (U) of 30 minutes and a lower
specifcation (L) of 20 minutes. The lab process must be capable of provid
ing the results of analyses within these specifcations (see fgure 4.7); oth
erwise, it will produce a certain proportion of “defects.”
Note that in most situations,
T
U L
=
~
2
and
t =
+ U L
2
,
and hence it is assumed so here.
Lower Speciﬁcation
T
a
g
u
c
h
i
L
o
s
s
FIGURE 4.6
“Larger is better” Taguchi loss.
© 2009 by Taylor & Francis Group, LLC
64 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
Two essential quantitative measures to assess the capability of a process
are process capability ratio (C
p
) and process capability index (C
pk
).
4.16.1 Process Capability Ratio (C
p
)
Assume that o is the standard deviation of a process that produces a certain
dimension of interest for an output (good or service). This certain dimension
of interest will hereafter be called critical dimension. The process capability
ratio (C
p
) is defned as
C
U L
p =
~
6o
(4.69)
The numerator represents the specifcation width and the denominator cap
tures the total width of the 3o limits of the process distribution. We consider
two examples, one for C
p
= 1 and the other for C
p
= 2.
If C
p
= 1, the specifcation width is the same as the distribution width.
When the process mean ( ) is centered at (U+L)/2 without any shift from the
target value, t , the probability that the actual critical dimension is within
the specifcation limits (assuming that the process distribution is normal)
is 0.9973 (2,700 ppm defect rate). Similarly, if C
p
= 2, the specifcation width
is twice that of the distribution. When the process mean ( ) is centered at
(U+L)/2 without any shift from t , the probability that the actual critical
dimension is within the specifcation limits is 0.999999998 (0.002 ppm defect
rate).
4.16.2 Process Capability Index (C
pk
)
The process capability ratio (C
p
) is enough to fnd out whether a process is
capable, only if is centered at (U+L)/2 without any shift from t . For exam
L = 20 min τ = 25 min U = 30 min
FIGURE 4.7
Capable process.
© 2009 by Taylor & Francis Group, LLC
Quantitative Modeling Techniques 65
ple, the lab process may have a good C
p
value (i.e., more than the critical
value of, say, 1.5), but if is closer to U, lengthy turnaround times may still
be generated. Likewise, if is closer to L, very quick results may be gener
ated. Thus, in order to check whether is not far away from t , there is a
need for an additional capability ratio, called the process capability index
(C
pk
).
C
pk
is defned in [35] as
C C k
pk p
= ~ ( ) 1
, where k
T
=
÷ f c
(4.70)
This defnition allows consideration of a mean shift, i.e., a shift of from t . The
fraction k is the fraction of tolerance consumed by the mean shift. The Motorola
convention uses a onesided mean shift of 1.5 o . This is motivated by com
mon physical phenomena such as tool wear. If C
p
= 2 and C
pk
= 1.5 (i.e., mean
shift consumes 25% of the tolerance), the probability that the actual critical
dimension is within the specifcation limits is 0.9999966 (i.e., 3.4 ppm defect
rate).
A process is said to be capable only if the process has good values (viz.,
more than the respective critical values) of both C
p
and C
pk
. If C
p
is less than
the critical value, o is too high. If C
pk
is less than the critical value, either
is too close to U or L or o is too high.
Six Sigma is an art of management that originated at Motorola in the early
1980s and is a businessdriven, multifaceted approach to process improve
ment, cost reduction, and proft increase. Its fundamental principle is to
improve customer satisfaction by reducing defects in processes.
Traditionally, one needs both C
p
and C
pk
values in order to investigate
whether the process of interest is a Six Sigma process. We illustrate this by
calculating C
p
and C
pk
values (equations (4.69) and (4.70), respectively) for an
n Sigma process (where n is any positive real number; the higher the value
of n, the better the process is). We consider three different cases, viz., n = 3,
4.5, and 6. It must be noted that the mean shift in each case is allowed to be
up to 1.5 o .
4.16.2.1 Three Sigma Process
See fgure 4.8.
C
U L
p =
~
= =
6
6
6
1
o
o
o
C C k
pk p
= ÷
( )
_ ÷
(
(
·
·
·
·
\
)
= 1 1 1
1 5
3
0 5
.
.
o
o
Hence, if C
p
= 1 and C
pk
~ 0.5, it is considered a Three Sigma process.
© 2009 by Taylor & Francis Group, LLC
66 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
4.16.2.2 4.5 Sigma Process
See fgure 4.9.
C
U L
p =
~
= =
6
9
6
1 5
o
o
o
.
C C k
pk p
= ÷
( )
_ ÷
(
(
·
·
·
·
\
)
= 1 1 5 1
1 5
4 5
1 .
.
.
o
o
Hence, if C
p
= 1.5 and C
pk
~ 1, it is considered a 4.5 Sigma process.
4.16.2.3 Six Sigma Process
See fgure 4.10.
C
U L
p =
~
= =
6
12
6
2
o
o
o
C C k
pk p
= ÷
( )
_ ÷
(
(
·
·
·
·
\
)
= 1 2 1
1 5
6
1 5
.
.
o
o
Hence, if C
p
= 2 and C
pk
~ 1.5, it is considered a Six Sigma process.
U = + 3
L =  3
1.5
T= 3
T= 3
FIGURE 4.8
Three Sigma process.
© 2009 by Taylor & Francis Group, LLC
Quantitative Modeling Techniques 67
4.17 Neural Networks
A neural network is made up of simple processing units (called neurons)
combined in a parallel computer system following implicit instructions based
on recognizing patterns in data inputs from external sources [41]. Neural
networks can model complex relationships between inputs and outputs or
fnd patterns in data. Their usefulness derives from their ability to embody
inferential algorithms that alter the strengths or weights of the network con
nections to produce a desired signifcant fow. In this book, the following
equation [42] is used (see chapters 6 and 7) to calculate the weights (impor
tance values) of evaluation criteria considered in a strategic planning issue.
8 ÷ ¹ 4.5
/ ÷ 4.5
1.5
7 ÷ 4.5
7 ÷ 4.5
FIGURE 4.9
4.5 Sigma process.
8 ÷ ¹ 6
/ ÷  6
1.5
7 ÷ 6
7 ÷ 6
FIGURE 4.10
Six Sigma process.
© 2009 by Taylor & Francis Group, LLC
68 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
 
 
 
W
I
I
O
I
I
O
v
vj
ij
i
n
j
j
n
vj
ij
i
n
j
V
H
V
=
(
(
·
·
·
·
`
`
`
··
·
·
·
·
·
·
·
·
\
)
(
(
·
·
·
·
·
·
·
·
·
·
`
j
n
H
··
·
·
\
)
`
i
n
V
(4.71)
Here, the absolute value of W
v
is the weight of the vth input node (evalua
tion criterion) upon the output node (rating of the decision alternative), n
V
is
the number of input nodes (evaluation criteria), n
H
is the number of hidden
nodes (can be any arbitrary number), I
ij
is the connection weight from the ith
input node to the jth hidden node, and O
j
is the connection weight from the
jth hidden node to the output node. The connection weights [43] are obtained
upon training the respective neural network.
4.18 Geographical Information Systems
A geographical information system (GIS) is a computer system with a set of
processes for obtaining, managing, analyzing, and displaying data that have
been located geographically [37]. In a more generic sense, GIS is a tool that
allows users to create interactive queries (usercreated searches), analyze the
spatial information, edit data, and map and present the results of all these
operations [38]. GISs have become powerful operation tools in the business
world.
In this book (see chapter 6), GIS is applied for displaying data. The motivation
for this application is a paper [39] that addresses the necessity for building
the strategic planning process “around a picture,” to make the chairperson
of the concerned supply chain company easily understand the “dense docu
ments flled with numbers” and to convince him or her that it is important to
implement the proposed action. To this end, an excellent GISbased business
mapping application, MapLand [40], is used to map the results obtained in
different phases of a model.
© 2009 by Taylor & Francis Group, LLC
Quantitative Modeling Techniques 69
4.19 Linear Integer Programming
Linear programming problems are about optimization of a linear objective
function, subject to linear equality and inequality constraints [36]. For exam
ple, in canonical form, a linear programming problem can be expressed as
Maximize C
T
X (objective function)
subject to AX ≤ B and X ≥ 0 (constraints)
Here, X represents the vector of variables, whereas C and B are vectors of coef
fcients and A is the matrix of coeffcients. If the variables of a linear program
ming problem are restricted to being integers, the programming is called linear
integer programming.
Linear integer programming can be applied to various felds. Most exten
sively, it is applied to business, economic, and engineering problems.
4.20 Conclusions
This chapter gave an introduction to various quantitative techniques
employed by strategic planning models presented in different chapters of
the book. Depending on the decisionmaking situation, each of the models
uses one or more of the techniques introduced in this chapter.
References
1. Saaty, T. L. 1980. The analytic hierarchy process. New York: McGrawHill.
2. Siddiqui, M. Z., Everett, J. W., and Vieux, B. E. 1996. Landfll siting using geo
graphic information systems: A demonstration. Journal of Environmental Engi
neering 122:515–23.
3. Saaty, T. L. 1990. How to make a decision: The analytic hierarchy process. Euro
pean Journal of Operational Research 48:9–26.
4. Tadisina, S. K., Troutt, M. D., and Bhasin, V. 1991. Selecting a doctoral pro
gramme using the analytic hierarchy process: The importance of perspective.
Journal of the Operational Research Society 42:631–38.
5. Prakash, G. P., Ganesh, L. S., and Rajendran, C. 1994. Criticality analysis of
spare parts using the analytic hierarchy process. International Journal of Produc
tion Economics 35:293–97.
6. Saaty, T. L. 1996. Decision making with dependence and feedback: The analytic net
work process. Pittsburgh, PA: RWS Publications.
© 2009 by Taylor & Francis Group, LLC
70 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
7. Ravi, V., Shankar, R., and Tiwari, M. K. 2005. Analyzing alternatives in reverse
logistics for endoflife computers: ANP and balanced scorecard approach.
Computers and Industrial Engineering 48:327–56.
8. Gungor, A. 2006. Evaluation of connection types in design for disassembly
(DFD) using analytic network process. Computers and Industrial Engineering
50:35–54.
9. Agarwal, A., Shankar, R., and Tiwari, M. K. 2006. Modeling the metrics of lean,
agile and leagile supply chain: An ANP based approach. European Journal of
Operational Research 173:211–25.
10. Zadeh, L. A. 1965. Fuzzy sets. Information and Control 8:338–53.
11. Tsaur, S., Chang, T., and Yen, C. 2002. The evaluation of airline service quality
by fuzzy MCDM. Tourism Management 23:107–15.
12. Chan, F. T., Chan, H. K., and Chan, M. H. 2003. An integrated fuzzy decision
support system for multicriterion decision making problems. Journal of Engi
neering Manufacture 217:11–27.
13. Wang, M., and Liang, G. 1995. Beneft/cost analysis using fuzzy concept. Engi
neering Economist 40:359–76.
14. Chang, D. Y. 1996. Applications of the extent analysis method on fuzzy AHP.
European Journal of Operations Research 95:649–55.
15. Chang D. Y. 1992. Extent analysis and synthetic decision, optimization techniques and
applications, p. 352. Vol. 1. Singapore: World Scientifc.
16. Chen, S. J., and Hwang, C. L. 1992. Fuzzy multiple attribute decision making: Meth
ods and applications. New York: Springer.
17. Deng, H. 1999. Multicriteria analysis with fuzzy pairwise comparison. Interna
tional Journal of Approximate Reasoning 21:215–31.
18. Zeleny, M. 1982. Multiple criteria decision making. New York: McGrawHill.
19. Hwang, C. L., and Yoon, K. S. 1981. Multiple attribute decision making: Methods and
applications. Berlin: Springer.
20. Yeh, C. H., and Deng, H. 1997. An algorithm for fuzzy multicriteria decision
making. In Proceedings of the IEEE First International Conference on Intelligent Pro
cessing Systems, pp. 1564–68.
21. Erol, I., and Ferrell, Jr., W. G. 2003. A methodology for selecting problems with
multiple, conficting objectives and both qualitative and quantitative data.
International Journal of Production Economics 86:187–99.
22. Messac, A., Gupta, S. M., and Akbulut, B. 1996. Linear physical programming:
A new approach to multiple objective optimization. Transactions on Operational
Research 8:39–59.
23. Martinez, M., Messac, A., and RaisRohani, M. 2001. Manufacturabilitybased
optimization of aircraft structures using physical programming. AIAA Journal
39:517–25.
24. Hillier, F. S., and Lieberman, G. J. 2001. Introduction to operations research. New
York: McGrawHill.
25. Ignizio, J. P. 1976. Goal programming and extensions. Lexington, MA: Lexington
Books, D. C. Heath and Company.
26. Ignizio, J. P. 1982. Linear programming in single and multi objective systems. Engle
wood Cliffs, NJ: Prentice Hall.
27. Tiwari, R. N., Dharmar, S., and Rao, J. R. 1987. Fuzzy goal programming—An
additive model. Fuzzy Sets and Systems 24:27–34.
28. Zimmermann, H. J. 1978. Fuzzy programming and linear programming with
several objective functions. Fuzzy Sets and Systems 1:45–55.
© 2009 by Taylor & Francis Group, LLC
Quantitative Modeling Techniques 71
29. Triantaphyllou, E., and Lin, C. 1996. Development and evaluation of fve fuzzy
multiattribute decisionmaking methods. International Journal of Approximate
Reasoning 14:281–310.
30. Hwang, C. L. 1987. Group decision making under multicriteria: Methods and appli
cations. New York: SpringerVerlag.
31. Rich, E., and Knight, K. 1992. Artifcial intelligence. New York: McGrawHill.
32. Hopgood, A. A. 1993. Knowledgebased systems for engineers and scientists. Boca
Raton, FL: CRC Press.
33. Pi, W., and Low, C. 2006. Supplier evaluation and selection via Taguchi loss
functions and an AHP. International Journal of Advanced Manufacturing Technol
ogy 27:625–30.
34. Krajewski, L. J., and Ritzman, L. P. 2004. Operations management: Processes and
value chains. 7th ed. Englewood Cliffs, NJ: Pearson Prentice Hall.
35. Narahari, Y., Viswanadham, N., and Bhattacharya, R. 2000. Design of synchro
nized supply chains: A Six Sigma tolerancing approach. In IEEE International
Conference on Robotics and Automation, pp. 1151–56.
36. Hillier, F. S., and Lieberman, G. J. 2005. Introduction to operations research. 8th ed.
New York: McGrawHill.
37. Estaville, L. E. 2007. GIS and colleges of business: A curricular exploration. Jour
nal of Real Estate Literature 15:443–48.
38. http://en.wikipedia.org/wiki/Geographic_information_system.
39. Kim, C. W., and Mauborgne, R. 2002. Charting your company’s future. Harvard
Business Review, June, pp. 77–82.
40. http://www.softwareillustrated.com/mapland.html.
41. Masi, C. G. 2007. Fuzzy neural control systems—explained. Control Engineering
54:62–66.
42. Cha, Y., and Jung, M. 2003. Satisfaction assessment of multiobjective schedules
using neural fuzzy methodology. International Journal of Production Research
41:1831–49.
43. Haykin, S. 1998. Neural networks: A comprehensive foundation. 2nd ed. Englewood
Cliffs, NJ: Prentice Hall.
© 2009 by Taylor & Francis Group, LLC
73
5
Selection of Used Products
5.1 The Issue
In many countries, especially in Europe, although many original equipment
manufacturers (OEMs) are obligated to take products back from the consum
ers upon the products’ end of use (hence called used products), there are also
many thirdparty companies that collect used products solely to make proft.
These companies select only those used products for which revenues from
recycle or resale of the products’ components are expected to be higher than
the costs involved in collection and reprocessing of used products and in dis
posal of waste. The various scenarios for selecting economical used products
could differ as follows:
1. Evaluation criteria could be presented in terms of classical numeri
cal constraints.
2. Evaluation criteria could be presented in terms of ranges of different
degrees of desirability.
In this chapter, two models to address the above scenarios are presented. The
frst model addresses scenario 1 and employs linear integer programming. The
second model is for scenario 2 and employs linear physical programming.
This chapter is organized as follows: section 5.2 presents the frst model,
section 5.3 shows the second model, and section 5.4 gives some conclusions.
5.2 First Model (Linear Integer Programming)
Section 5.2.1 presents the nomenclature for formulation of the linear integer
programming model, section 5.2.2 presents the formulation of the model,
and section 5.2.3 gives a numerical example to illustrate the model.
© 2009 by Taylor & Francis Group, LLC
74 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
5.2.1 Nomenclature
C
df
Disposal cost factor (cost per unit weight)
C
dx
Disposal cost of product x
C
r
Reprocessing cost per unit time
C
rf
Recycling revenue factor (revenue per unit weight)
C
rpx
Total reprocessing cost of used product x
CC
x
Collection cost of used product x
D
xy
Disposal cost index of component y in used product x (0 =
lowest, 10 = highest)
E
xk
Subassembly k in used product x
m
xy
Probability of missing component y in used product x
M
x
Number of subassemblies in used product x
N
xy
Multiplicity of component y in used product x
p
xy
Probability of breakage of component y in used product x
P
xy
Component y in used product x
PRC
xy
Percent of recyclable contents by weight in component y of
used product x
R
rcx
Total recycling revenue of used product x
R
rsx
Total resale revenue of used product x
R
rsxy
Resale value of component y in used product x
RC
xy
Recycling revenue index of component y in used product x (0
= lowest, 10 = highest)
Root
x
Root node of used product x
SU
x
Supply of used product x per period
T(E
xk
) Time to disassemble subassembly k in used product x
T(Root
x
) Time to disassemble Root
x
W
xy
Weight of component y in used product x
x Used product type
X
xy
Decision variable signifying selection of component y to be
retrieved from used product x for reuse (X
xy
= 1 for reuse, 0 for
recycle)
y Component type
Z Overall proft
5.2.2 Model Formulation
The costbeneft function [1] in the literature presents a technique to select
the best used product for reprocessing from a set of candidate used prod
ucts. The function consists of four terms: resale revenue, recycling revenue,
© 2009 by Taylor & Francis Group, LLC
Selection of Used Products 75
reprocessing cost, and disposal cost. The difference between the sum of the
revenue terms and the sum of the cost terms gives the costbeneft function
value. The used product with the maximum costbeneft function value is
selected as the optimal one for reprocessing. The optimal solution (for proft
maximization) also gives the best feasible set of components of retrieval from
the selected used product. However, there is serious risk of making a bad
choice of the used product using the costbeneft function because the func
tion implicitly makes two assumptions that are seldom valid in a reverse
supply chain scenario: (1) every component selected for resale will be in a
reusable state after disassembling the used product, and (2) that all the com
ponents in the used product are in their original multiplicities. Therefore,
here, a modifed costbeneft function is presented. This function incorpo
rates the probability of breakage and the probability of missing components
in the used product of interest. Then a linear integer programming model is
formulated and implemented to the data of each candidate used product, in
order to select the most economical used product to reprocess. Section 5.2.2.1
presents the modifed costbeneft function, and section 5.2.2.2 gives the lin
ear integer programming model.
5.2.2.1 Modified CostBenefit Function
The modifed costbeneft function for used product x consists of fve terms:
total resale revenue (R
rsx
), total recycling revenue (R
rcx
), total reprocessing
cost (C
rpx
), total disposal cost (C
dx
), and collection cost. It can be written as
Z R R C C CC
rsx rcx rpx dx x
= ÷ ÷ ÷ ÷ (5.1)
The terms in the function are described as follows:
Total resale revenue: R
rsx
is infuenced by the resale value of individual
components of the used product (R
rsxy
), the number of components
(N
xy
), the probability of breakage (p
xy
), and the probability of missing
components (m
xy
). The revenue equation can be written as follows:
R R N p m X
rsx rsxy xy xy xy xy
j Pxy Root
= ÷ ÷
¦ ¦
¬ =
. .( ).
(
1
xx
)
`
(5.2)
Total recycling revenue: R
rcx
is infuenced by the percentage of recyclable
contents in each component (PRC
xy
); the weight of the components
(W
xy
); the recycling revenue index (RC
xy
), which is a number in the
range 1–10 representing the degree of beneft generated by recycling
a component of type y (the higher the value, the more proftable it is
to recycle); the number of components (N
xy
); the probability of break
© 2009 by Taylor & Francis Group, LLC
76 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
age (p
xy
); and the probability of missing components (m
xy
). The recy
cling revenue equation can be written as follows:
R PRC W RC N m N p m
rcx xy xy xy xy xy xy xy x
= ÷ ÷ ÷ ÷ . . ( ( ) ( 1 1
yy xy
j Pxy Root
rf
X C
x
) )
( )
¦ ¦
¬ = ¦
`
(5.3)
Total reprocessing cost: C
rpx
can be calculated from the disassembly time
of the root node of the used product (T(Root
x
)), the disassembly time
of each subassembly in the used product (T(E
xk
)), and the reprocess
ing cost per unit time (C
r
). The total reprocessing cost equation can
be written as follows:
C T Root T E
rpx x xk
k
M
x
= ÷
'
¦
1
1
1
+
1
1
1
'
¦
1
1
1
+
1
1
=
`
( ) ( )
1
11
.C
r
(5.4)
Total disposal cost: C
dx
can be calculated from the disposal cost index
(D
xy
), which is a number in the range 1–10 representing the degree
of diffculty in disposing component y of used product x (the higher
the number, the more diffcult it is to dispose), the percentage of
recyclable contents in each component (PRC
xy
), the weight of the
components (W
xy
), the number of components (N
xy
), the probability
of breakage (p
xy
), the probability of missing components (m
xy
), and
the disposal cost factor (C
df
). The disposal cost equation can be writ
ten as follows:
C D W PRC N m N p
dx xy xy xy xy xy xy xy
= ÷ ÷ ÷ ÷ . .( ).( ( ) ( 1 1 1 ÷÷
¦ ¦
¬ =
`
m X C
xy xy
j Pxy Root
df
x
). ) .
( ) (5.5)
Collection cost: CC
x
is the average cost of collecting used product x from
the consumers.
5.2.2.2 Linear Integer Programming Model
The following linear integer programming model maximizes the costben
eft obtained from reprocessing used product x:
Maximize Z R R C C CC
x rsx rcx rpx dx x
= ÷ ÷ ÷ ÷ (5.6)
subject to X
xy
= 0 or 1 for all x and y (5.7)
© 2009 by Taylor & Francis Group, LLC
Selection of Used Products 77
The above formulation assesses the feasible combinations of components’
retrieval from a used product and compares the combination with the high
est costbeneft from one product against others.
5.2.3 Numerical Example
Two used products (1 and 2) are considered in this numerical example (see
fgures 5.1 and 5.2).
The data required to implement the model for the products are shown in
tables 5.1 and 5.2, respectively.
Also, CC
1
= 25, CC
2
= 40, C
rf
= 0.5, C
r
= 0.8/min, C
df
= 0.25/lb, T(Root
1
) = 6
min, T(Root
2
) = 4 min, T(E
11
) = 3 min, T(E
12
) = 5 min, T(E
21
) = 2 min, and T(E
22
)
= 4 min.
Root
1
E
11
P
11
P
1
P
1
E
12
P
14
P
15
FIGURE 5.1
Structure of used product 1 (frst model).
Root
2
P
21
P
23
E
22
E
21
P
22
P
24
P
25
FIGURE 5.2
Structure of used product 2 (frst model).
© 2009 by Taylor & Francis Group, LLC
78 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
Upon solving the model with the above data using LINGO (v4), one gets
total proft for used product 1 as $46.16 and total proft for used product 2 as
$68.50. Hence, the decision maker will select used product 2 in this case.
5.3 Second Model (Linear Physical Programming)
Section 5.3.1 presents the formulation of the linear physical programming
model, and section 5.3.2 gives a numerical example to illustrate the model. The
nomenclature used to formulate the model is the same as in section 5.2.1.
5.3.1 Model Formulation
The criteria considered in the model fall into either class 1S or class 2S. Section
5.3.1.1 presents the class 1S criteria, and section 5.3.1.2 gives the class 2S criteria.
5.3.1.1 Class 1S Criteria (Smaller Is Better)
Total collection cost per period (g
1
): g
1
of used product x is calculated by
multiplying the supply of x per period (SU
x
) by the cost of collecting
one product from consumers (CC
x
):
TABLE 5.1
Data of used product 1 (frst model)
Part R
rs1y
N
1y
W
1y
Rc1y PRC
1y
D
1y
p
1y
m
1y
P
11
10 4 3 7 75% 3 0.02 0.0
P
12
2 2 2.5 9 60% 2 0.05 0.0
P
13
3.75 1 5 5 50% 6 0.0 0.5
P
14
5 5 7 3 70% 1 0.1 0.3
P
15
3 6 4 6 40% 7 0.4 0.02
TABLE 5.2
Data of used product 2 (frst model)
Part R
rs2y
N
2y
W
2y
R
c2y
PRC
2y
D
2y
p
2y
m
2y
P
21
2.5 2 4 5 40% 2 0.0 0.05
P
22
5 3 1 6 35% 6 0.0 0.1
P
23
3 1 3 2 70% 8 0.1 0.0
P
24
0.5 2 4.5 8 80% 5 0.15 0.2
P
25
2 2 5 7 25% 7 0.1 0.25
© 2009 by Taylor & Francis Group, LLC
Selection of Used Products 79
g SU CC
x x 1
= (5.8)
Total reprocessing cost per period (g
2
): g
2
of used product x is calculated
using the disassembly time of the root node (T(Root
x
)), disassembly
time of each subassembly (T(E
xk
)), supply of x per period (SU
x
), and
the remanufacturing cost per unit time (C
r
):
SU T Root T E C
x x xk
k
M
r
i
( ) ( ) ÷
l
l
l
l
l
=
`
1
(5.9)
Total disposal cost per period (g
3
): g
3
of used product x is calculated by
multiplying the component disposal cost by the number of units of
components disposed, as follows:
SU DI W PRC N m N b m
x xy xy xy xy xy xy xy x
. ( ) ( ) ( 1 1 1 ÷ ÷ ÷ ÷ ÷
yy
y
df
C )
¦ ¦
l
l
l `
(5.10)
DI
xy
is the disposal cost index that varies in value from 1 to 10 repre
senting the degree of nuisance created by the disposal of component
y of product x, and C
df
is the disposal cost factor.
Lossofsale cost (g
4
): g
4
of used product x represents the periodic worth of
not meeting the demand on time. It can occur because of the unpre
dictability in the supply of used products. This can be obtained from
an expert in the feld.
Worth of investment cost (g
5
): g
5
of used product x represents the periodic
worth of the fxed cost of the production facility and the machinery
required to reprocess. This can also be obtained from an expert in
the feld.
5.3.1.2 Class 2S Criteria (Larger Is Better)
Total reuse revenue per period (g
6
): g
6
of used product x is infuenced by
the supply of x per period (SU
x
), the resale value of component y
(RSR
xy
), the multiplicity of component y (N
xy
), and the breakage and
missing probabilities of component y (p
xy
, m
xy
). The reuse revenue
equation can be written as follows:
SU RSRxy N m b
x xy xy xy
y
. . .( ) 1÷ ÷
l
l
l `
(5.11)
© 2009 by Taylor & Francis Group, LLC
80 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
Total recycling revenue per period (g
7
): g
7
of used product x is infuenced
by the supply of x per period (SU
x
), the recycling revenue index of
component y (RCRI
xy
), the percentage of recyclable content in com
ponent y (PRC
xy
), the multiplicity of component y (N
xy
), the weight of
component y (W
xy
), the recycling revenue factor (C
rf
), and the break
age and missing probabilities of component y (p
xy
, m
xy
). The recycling
revenue equation can be written as
SU RCRI W PRC N N m N m
x xy xy xy xy xy xy xy
. . . . . ( ) ( 1 1 ÷ ÷ ÷
xxy xy rf
y
b C ÷
¦ ¦
l
l
l `
) (5.12)
5.3.2 Numerical Example
Consider three used products (3, 4, and 5), whose structures are shown in
fgures 5.3–5.5, respectively.
The data for the three products are given in tables 5.3–5.5, respectively.
The target values for the criteria are given in table 5.6, and table 5.7 shows
the criteria values for each product. Table 5.8 shows the incremental weights
obtained using the LPP weight algorithm [2].
Tables 5.9–5.11 show the deviations of criteria values from target values
for the three products, respectively. For example, for criteria g
1
of product 3,
deviation at s = 2 is the absolute value of the number obtained by subtracting
the criteria value (i.e., 7.5; shown in bold in table 5.7) from the target value
(i.e., 10; shown in bold in table 5.6).
The total score for each product is calculated using equation (4.34) (using
the incremental weights from the LPP algorithm and deviations from target
values) and is shown in table 5.12. Because alternatives with lower scores are
more desirable than ones with higher scores, used product 3 is the best of
the lot.
5.4 Conclusions
In this chapter, two models are presented for selecting the most economical
product to reprocess from a set of candidate used products. The frst model
uses linear integer programming, and the second model employs linear
physical programming.
© 2009 by Taylor & Francis Group, LLC
Selection of Used Products 81
Root
4
P
41
P
43
E
42
E
41
P
42
P
44
P
45
FIGURE 5.4
Structure of used product 4 (second model).
Root
5
P
52
P
53
E
52
E
51
P
51
P
54
P
55
FIGURE 5.5
Structure of used product 5 (second model).
Root
3
E
31
P
31
P
3
P
3
E
32
P
34
P
35
FIGURE 5.3
Structure of used product 3 (second model).
© 2009 by Taylor & Francis Group, LLC
82 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
TABLE 5.3
Data of used product 3 (second model)
Part RSR
xy
N
xy
W
1y
RCRI
xy
PRC
xy
DI
xy
p
xy
m
xy
P
31
6 3 3.5 5 0.65 6 0.1 0.3
P
32
7 2 4.5 4 0.4 4 0.2 0.5
P
33
5 4 5 5 0.3 4 0.4 0.1
P
34
7.5 2 6 3 0.25 5 0.2 0.2
P
35
5.5 1 5.5 2 0.45 1 0.5 0.0
TABLE 5.4
Data of used product 4 (second model)
Part RSR
xy
N
xy
W2
y
RCRI
xy
PRC
xy
DI
xy
p
xy
m
xy
P
41
4 2 9 2 0.56 3 0.2 0.5
P
42
6 5 3 1 0.5 6 0.2 0.2
P
43
7 4 5 2 0.48 6 0.4 0.1
P
44
2 2 6 3 0.2 1 0.1 0.1
P
45
4.5 3 1 4 0.25 3 0.1 0.3
TABLE 5.5
Data of used product 5 (second model)
Part RSR
xy
N
xy
W3
y
RCRI
xy
PRC
xy
DI
xy
p
xy
m
xy
P
51
4.5 3 8.5 5 0.4 5 0.1 0.3
P
52
4 4 6 4 0.65 6 0.1 0.1
P
53
5 6 4.1 2 0.2 3 0.2 0.0
P
54
2 5 3.5 3 0.35 2 0.3 0.2
P
55
3.5 1 2 1 0.25 7 0.5 0.5
© 2009 by Taylor & Francis Group, LLC
Selection of Used Products 83
TABLE 5.6
Target values of criteria (second model)
Criteria t
p1
+ t
p2
+ t
p3
+ t
p4
+ t
p5
+
g
1
10 12 15 17.5 20
g
2
3 5 9 10 15
g
3
1 4 5.5 7.5 8
g
4
2 5 7.5 9 10
g
5
1 4 8 9 10
Criteria t
p1
– t
p2
– t
p3
– t
p4
– t
p5
–
g
6
10 12.5 15 20 25.5
g
7
5 13 17.5 20 25
TABLE 5.7
Criteria values for each product (second model)
Criteria Used product 3 Used product 4 Used product 5
g
1
7.5 7.5 10
g
2
3.5 2.75 3
g
3
2.97 1.6 1.8
g
4
3.5 4 5
g
5
2 2.5 2
g
6
18.3 22.8 24.9
g
7
18.4 15.7 19
TABLE 5.8
Output of LPP weight algorithm (second model)
Criteria Δw
p2
+ Δw
p3
+ Δw
p4
+ Δw
p5
+ Δw
p2
– Δw
p3
– Δw
p4
– Δw
p5
–
g
1
0.05 0.0967 0.62773 2.63296 — — — —
g
2
0.05 0.076 2.41416 0.020321 — — — —
g
3
0.0333 0.3027 0.93408 24.33473 — — — —
g
4
0.0333 0.16827 1.49184 11.10897 — — — —
g
5
0.0333 0.09267 2.41416 10.26225 — — — —
g
6
— — — — 0.04 0.0492 0.010258 0.122333
g
7
— — — — 0.0125 0.037056 0.14936 0.022875
© 2009 by Taylor & Francis Group, LLC
84 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
TABLE 5.9
Deviations of criteria values from targets, for used product 3 (second model)
Criteria s = 2 s = 3 s = 4 s = 5
g
1
2.5 4.5 7.5 10
g
2
0.5 1.5 5.5 6.5
g
3
1.9 1.1 2.6 4.6
g
4
1.5 1.5 4 5.5
g
5
1 2 6 7
g
6
6.7 1.7 3.3 5.8
g
7
6.6 1.6 0.9 5.4
TABLE 5.10
Deviations of criteria values from targets, for used product 4 (second model)
Criteria s = 2 s = 3 s = 4 s = 5
g
1
2.5 4.5 7.5 10
g
2
0.25 2.25 6.25 7.25
g
3
0.6 2.4 3.9 5.9
g
4
2 1 3.5 5
g
5
1.5 1.5 5.5 6.5
g
6
2.2 2.8 7.8 10.3
g
7
9.3 4.3 1.8 2.7
TABLE 5.11
Deviations of criteria values from targets, for used product 5 (second model)
Criteria s = 2 s = 3 s = 4 s = 5
g
1
0 2 5 7.5
g
2
0 2 6 7
g
3
0.8 2.2 3.7 5.7
g
4
3 0 2.5 4
g
5
1 2 6 7
g
6
0.1 4.9 9.9 12.4
g
7
5.93 0.93 1.57 6.07
© 2009 by Taylor & Francis Group, LLC
Selection of Used Products 85
References
1. Veerakamolmal, P., and Gupta, S. M. 1999. Analysis of design effciency for the
disassembly of modular electronic products. Journal of Electronics Manufacturing
9:79–95.
2. Messac, A., Gupta, S. M., and Akbulut, B. 1996. Linear physical programming:
A new approach to multiple objective optimization. Transactions of Operational
Research 8:39–59.
TABLE 5.12
Total scores and ranks of products (second model)
Used product Score Rank
3 315.3144 1
4 339.1954 3
5 317.8653 2
© 2009 by Taylor & Francis Group, LLC
87
6
Evaluation of Collection Centers
6.1 The Issue
Strategic planning of an effcient reverse or closedloop supply chain requires
selection of effcient collection centers where used products are disposed
of by the consumers. These collection centers, after initial processing (for
example, sorting), ship the used products to recovery facilities or production
facilities where reprocessing operations, such as disassembly and recycling/
remanufacturing, are carried out.
The various scenarios for evaluating collection centers for effciency could
differ as follows:
1. Supply chain company executives, whose primary concern is proft,
could be the sole decision makers.
2. There could exist three different categories of decision makers: con
sumers, local government offcials, and supply chain company exec
utives. Impacts* of evaluation criteria are given.
3. The situation could be the same as in scenario 2, but the impacts of
evaluation criteria are not given (and hence must be derived).
4. Evaluation could be made from the perspective of a remanufactur
ing facility interested in buying used products from the candidate
collection centers. The goals are expressed in terms of performance
indices (effciency scores).
5. The situation could be the same as in scenario 4, but the goals are
expressed in terms of Taguchi losses (ineffciency scores).
In this chapter, various models to address the above scenarios are pre
sented. The frst model addresses scenario 1 and employs the eigen vector
method and Taguchi loss function. The second and third models are for sce
narios 2 and 3, respectively. The main difference between these two models
*
In this chapter, we use the term impacts instead of the conventional term weights for impor
tance values of evaluation criteria. This is to avoid confusion in the presence of the connec
tion weights between different nodes of the neural network used in section 6.5.
© 2009 by Taylor & Francis Group, LLC
88 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
lies in the way the impacts are assigned to the criteria for evaluation of col
lection centers. Whereas the second model uses the eigen vector method,
technique for order preference by similarity to ideal solution (TOPSIS), and
Borda’s choice rule, the third model uses neural networks, fuzzy logic, TOP
SIS, and Borda’s choice rule. The fourth model is for scenario 4 and uses
analytic network process (ANP) and goal programming. The ffth model,
which is for scenario 5, uses the eigen vector method, Taguchi loss function,
and goal programming. The main difference between the last two models is
in the way the goals are expressed.
This chapter is organized as follows: Section 6.2 presents the frst model.
Section 6.3 gives the different decision makers (and their criteria) that are
considered in the second and third models. Sections 6.4–6.7 present the sec
ond, third, fourth, and ffth models, respectively. Finally, section 6.8 gives
some conclusions.
6.2 First Model (Eigen Vector Method
and Taguchi Loss Function)
In this model, the eigen vector method and Taguchi loss function are
employed to identify effcient centers from a set of candidate collection cen
ters operating in a region where a reverse supply chain is to be designed.
Section 6.2.1 presents the evaluation criteria used in the model, and section
6.2.2 illustrates the model using a numerical example.
6.2.1 Evaluation Criteria
This model assumes that the supply chain company executives, whose pri
mary concern is proft, are the sole decision makers. Hence, the evaluation
criteria do not consider factors such as “Is the collection center able to pro
vide employment opportunities to the local community?” and “Is the col
lection process convenient to the consumers?” (For factors such as these, see
sections 6.4 and 6.5, where criteria of two additional different decision mak
ers, viz., consumers and local government offcials, are considered.)
The following are the evaluation criteria used in the model:
n r value (the higher the n value, the lower the process defects, and
hence the higher the proft; see section 4.16 for the concept of n Sigma).
The n value of a collection center represents the center’s quality that
could be a function of factors, such as functionality of used products
collected, effciency of collection, effciency of delivery (to recovery
facilities), and effectiveness of customer service. For example (see
fgure 6.1), let the upper specifcation (U) of the service time (critical
© 2009 by Taylor & Francis Group, LLC
Evaluation of Collection Centers 89
dimension) at a candidate collection center be 20 min (i.e., the con
sumers discarding used products must not be made to wait too long
at the collection center), the lower specifcation (L) be 10 units (i.e., the
collection center personnel must spend at least some time inspect
ing used products before accepting them from consumers), and the
target value (τ) be 15 min. If the standard deviation (σ) of the service
time is 2 min and the mean shift ( T M ÷ ) is 2.5 min, then the service
process in the candidate collection center is a 2.5 Sigma process.
Per capita income of people in residential area ( r PI) (the higher it is,
the greater the number of resourceful discarded products and the
less the people will care about the incentives from the collection
center)
Space cost ( r SC) (the lower, the better)
Labor cost ( r LC) (the lower, the better)
Utilization of incentives from local government ( r UI) (the higher,
the better)
Distance from residential area ( r DH) (lower distance implies greater
collection and hence greater proft)
Distance from roads ( r DR) (lower distance implies greater collection
and hence greater proft)
Incentives from local government ( r IG) (higher incentives from
local government imply higher incentives to consumers, and hence
greater collection)
6.2.2 Model
The model is presented using a numerical example. Table 6.1 shows the pair
wise comparison matrix for the criteria for evaluation of candidate collection
20
10 15
2.5 min
÷ 2 min
FIGURE 6.1
Critical dimension of service process at candidate collection center (frst model).
© 2009 by Taylor & Francis Group, LLC
90 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
centers. For example (see table 6.1), the per capita income of the people in the
residential area (PI) is given fve times more importance than the n value and
three times more importance than the utilization of incentives from the local
government (UI).
Table 6.2 shows the impacts of the respective evaluation criteria. These
impacts are the elements of the normalized eigen vector of the pairwise
comparison matrix shown in table 6.1. For example (see table 6.2), the per
capita income of the people in the residential area (PI) is given an impact of
18% and the n value is given an impact of 11%.
The k value of the Taguchi loss function, in the case of each of the evalua
tion criteria, is calculated as follows.
6.2.2.1 n Value
The loss function that applies to this criterion is “larger is better” (see equa
tion (4.68)). If the decision maker considers 100% loss for an n value less than
or equal to 4, then the value of k is 1,600%. Then, for a candidate collection
center, if n = 5, L y ( ) / ( ) % = = 1600 5 64
2
. That means, with respect to quality
TABLE 6.1
Pairwise comparison matrix (frst model)
Criteria n DH DR UI PI SC LC IG
n 1 1 1 3 1/5 1 1 1
DH 1 1 1 1 1 1 1 1
DR 1 1 1 1 1 1 1 1
UI 1/3 1 1 1 1/3 0.2 1 1
PI 5 1 1 3 1 1 1 1
SC 1 1 1 5 1 1 1 0.2
LC 1 1 1 1 1 1 1 1
IG 1 1 1 1 1 5 1 1
TABLE 6.2
Impacts of criteria (frst model)
Criteria Impacts
n 0.11
DH 0.11
DR 0.11
UI 0.07
PI 0.18
SC 0.15
LC 0.11
IG 0.16
© 2009 by Taylor & Francis Group, LLC
Evaluation of Collection Centers 91
(i.e., n value), the collection center is 36% short of the worst performance level
(which, in this case, is n ≤ 4).
6.2.2.2 Distance from Residential Area (DH)
DH is considered the distance of the collection center from the center of grav
ity [1] of all the residential areas around the center. The loss function that
applies to this criterion is “smaller is better” (see equation (4.67)). If the deci
sion maker considers 100% loss for a DH value more than or equal to 4 miles,
then the value of k is 6.25%. Then, for a candidate collection center, if DH =
3, L y ( ) . ( ) . % = = 6 25 3 56 25
2
. That means, with respect to the distance from
the residential area, the collection center is 43.75% short of the worstcase
scenario (which, in this case, is DH ≥ 4 miles).
6.2.2.3 Distance from Roads (DR)
DR is considered the average distance of all the roads in the region from the
collection center of interest. The loss function that applies to this criterion is
“smaller is better” (see equation (4.67)). If the decision maker considers 100%
loss for a DR value more than or equal to 5 miles, then the value of k is 4%.
Then, for a candidate collection center, if DR = 4, L y ( ) ( ) % = = 4 4 64
2
. That
means, with respect to the distance from the roads, the collection center is
36% short of the worstcase scenario (which, in this case, is DR ≥ 5 miles).
6.2.2.4 Utilization of Incentives from Local Government (UI)
Because this is a subjective criterion, the decision maker can obtain ratings
(for example, on a 1–10 scale, where 1 is the worst and 10 is the best) of the
candidate collection centers from experts in the feld of reverse supply chain.
The loss function that applies to this criterion is “larger is better” (see equa
tion (4.68)). If the decision maker considers 100% loss for a UI value less than
or equal to 5, then the value of k is 2,500%. Then, for a candidate collection
center, if UI = 7, L y ( ) / ( ) . % = = 2500 7 51 02
2
. That means, with respect to the
utilization of incentives from the local government, the collection center is
about 49% short of the worstcase scenario (which, in this case, is UI ≤ 5).
6.2.2.5 Per Capita Income of People in Residential Area (PI)
The loss function that applies to this criterion is “larger is better” (see equation
(4.68)). If the decision maker considers 100% loss for a PI value less than or equal
to $30,000 per year, then the value of k is 90,000,000,000%. Then, for a candidate
collection center, if PI = $50,000 per year, L y ( ) , , , / ( , ) % = = 90 000 000 000 50 000 36
2
.
That means, with respect to the per capita income of the people in the residen
tial area, the collection center is 64% short of the worstcase scenario (which,
in this case, is PI ≤ $30,000 per year).
© 2009 by Taylor & Francis Group, LLC
92 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
6.2.2.6 Space Cost (SC)
The loss function that applies to this criterion is “smaller is better” (see equa
tion (4.67)). If the decision maker considers 100% loss for a SC value more than
or equal to $1,000 per day, then the value of k is 0.0001%. Then, for a candi
date collection center, if SC = $800 per day, L y ( ) . ( ) % = = 0 0001 800 64
2
. That
means, with respect to the space cost, the collection center is 36% short of the
worstcase scenario (which, in this case, is SC ≥ $1,000 per day).
6.2.2.7 Labor Cost (LC)
The loss function that applies to this criterion is “smaller is better” (see equa
tion (4.67)). If the decision maker considers 100% loss for a LC value more
than or equal to $15 per hour, then the value of k is 0.44%. Then, for a can
didate collection center, if LC = $10 per hour, L y ( ) . ( ) % = = 0 44 10 44
2
. That
means, with respect to the labor cost, the collection center is 56% short of the
worstcase scenario (which, in this case, is LC ≥ $15 per hour).
6.2.2.8 Incentives from Local Government (IG)
Because this is a subjective criterion, the decision maker can obtain ratings
(for example, on a 1–10 scale, where 1 is the worst and 10 is the best) of the
candidate collection centers from experts in the feld of reverse supply chain.
The loss function that applies to this criterion is “larger is better” (see equa
tion (4.68)). If the decision maker considers 100% loss for an IG value less than
or equal to 7, then the value of k is 4,900%. Then, for a candidate collection
center, if IG = 9, L y ( ) / ( ) . % = = 4900 9 60 49
2
. That means, with respect to the
incentives from the local government, the collection center is about 39.5%
short of the worstcase scenario (which, in this case, is IG ≤ 7).
Four candidate collection centers are considered in the numerical example:
A, B, C, and D. Table 6.3 presents the L(y) for the evaluation criteria for each
of the collection centers. For example, the Taguchi loss of B with respect to
the per capita income of the people in the residential area is 52% (i.e., 48%
short of the worstcase scenario).
The weighted loss of each collection center j is calculated by using the fol
lowing equation and is presented in table 6.4.
Weighted loss of collection center j = WL
i
i
ij
`
(6.1)
where W
i
is the impact of criterion i (see table 6.2) and L
ij
is the Taguchi loss
(see table 6.3) of collection center j with respect to criterion i. For example,
the weighted loss of D (see tables 6.2–6.4) is 0.11×63 + 0.11×55 + … + 0.11×67
+ 0.16×41 = 64.31.
The decision maker will select C because it has the lowest weighted loss.
© 2009 by Taylor & Francis Group, LLC
Evaluation of Collection Centers 93
6.3 Evaluation Criteria for Second and Third Models
In evaluation of collection centers, one may have a scenario with three dif
ferent categories of decision makers with multiple, conficting, and incom
mensurate goals, as follows:
Consumers r whose primary concern is convenience
Local government offcials r whose primary concern is environmental
consciousness
Supply chain company executives r whose primary concern is proft
Therefore, the effciency of a candidate collection center must be evaluated
based on the maximized consensus among decision makers of the three cate
gories. Sections 6.3.1, 6.3.2, and 6.3.3 present the lists of criteria that are consid
ered for the three categories. Note that the criteria for the third category (supply
chain company executives) are the same as those listed in section 6.2.1.
6.3.1 Criteria of Consumers
Incentives from collection center ( r IC) (higher incentives imply higher
motivation to participate)
TABLE 6.3
L(y) values (%) of collection centers (frst model)
Criteria A B C D
n (0.11) 35 54 40 63
DH (0.11) 25 15 37 55
DR (0.11) 32 10 9 90
UI (0.07) 100 75 64 50
PI (0.18) 65 52 40 50
SC (0.15) 20 10 5 100
LC (0.11) 15 18 20 67
IG (0.16) 78 64 36 41
TABLE 6.4
Weighted losses of collection centers (frst model)
Collection center Weighted loss
A 45.95
B 37.02
C 29.85
D 64.31
© 2009 by Taylor & Francis Group, LLC
94 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
Distance from residential area ( r DH) (lower distance implies higher
motivation to participate)
Distance from roads ( r DR) (lower distance implies higher motivation
to participate)
Simplicity of collection process ( r SP) (simpler process implies higher
motivation to participate)
Employment opportunity ( r EO) (the higher, the better)
Salary offered to employees at collection center ( r SA) (the higher,
the better)
6.3.2 Criteria of Local Government Officials
Distance from residential area ( r DH) (lower distance implies greater
collection and hence lower disposal)
Distance from roads ( r DR) (lower distance implies greater collection
and hence lower disposal)
6.3.3 Criteria of Supply Chain Company Executives
n r value (the higher the n value, the lower the process defects, and
Sigma)
Per capita income of people in residential area ( r PI) (the higher it is,
the more the number of resourceful discarded products, and the
less the people will care about the incentives from the collection
center)
Space cost ( r SC) (the lower, the better)
Labor cost ( r LC) (the lower, the better)
Utilization of incentives from local government ( r UI) (the higher,
the better)
Distance from residential area ( r DH) (lower distance implies greater
collection and hence greater proft)
Distance from roads ( r DR) (lower distance implies greater collection
and hence greater proft)
Incentives from local government ( r IG) (higher incentives from local
government imply higher incentives to consumers and hence greater
collection)
hence the higher the proft; see section 4.16 for the concept of n
© 2009 by Taylor & Francis Group, LLC
Evaluation of Collection Centers 95
6.4 Second Model (Eigen Vector Method,
TOPSIS, and Borda’s Choice Rule)
This model to select effcient collection centers is implemented in two phases.
In the frst phase, using the eigen vector method, impacts are given to the
criteria identifed for each category of decision makers (see section 6.3), and
then TOPSIS (technique for order preference by similarity to ideal solution) is
employed to fnd the effciency of each candidate collection center, as evalu
ated by that category. In the second phase, Borda’s choice rule is used to
combine individual evaluations for each candidate collection center into a
group evaluation or maximized consensus ranking. Furthermore, motivated
by a paper [3] that addresses the necessity for building the strategic planning
process “around a picture” to make the chairman of the concerned supply
chain company easily understand the “dense documents flled with num
bers” and to convince him that it is important to implement the proposed
action (here, selection of particular collection centers), a GISbased business
mapping application, MapLand [4], is used to map the results obtained in
both phases of this model.
The model to select effcient collection centers is presented using a numerical
example. Three recovery facilities, E, F, and G, are considered for evaluation.
6.4.1 Phase I (Individual Decision Making)
Tables 6.5–6.7 show the pairwise comparison matrices as formed for the con
sumers, local government offcials, and supply chain company executives,
respectively. Note that not all of the criteria are considered here, because the
focus is on the methodology.
Tables 6.8–6.10 show the impacts given for the criteria of the consumers,
local government offcials, and supply chain company executives, respec
tively. These sets of impacts, calculated using the eigen vector method, are
the elements of the normalized eigen vectors of pairwise comparison matri
ces shown in tables 6.5–6.7, respectively. For example, the impacts of the
criteria DH (distance from residential area) and DR (distance from roads),
shown in table 6.9, are the elements of the normalized eigen vector of the
pairwise comparison matrix shown in table 6.6.
The decision matrices for the consumers, local government offcials, and
supply chain company executives, for implementation of the TOPSIS for each
category of decision makers, are shown in tables 6.11–6.13, respectively. The
elements of these matrices are the ranks (ranging from 1 to 10) assigned to
the collection centers with respect to each criterion for evaluation. A lower
rank implies higher effciency (with respect to that criterion).
To facilitate the construction of the pairwise comparison matrices (see
tables 6.8–6.10) and the decision matrices (see tables 6.11–6.13), representa
tives from each category of decision makers could be invited to participate
© 2009 by Taylor & Francis Group, LLC
96 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
TABLE 6.5
Pairwise comparison matrix for consumers (second model)
Criteria IC DH DR SP EO SA
IC 1 1 1 2 1/2 1
DH 1 1 2 1 1 1/2
DR 1 1/2 1 1 1/7 1/2
SP 1/2 1 1 1 1 1
EO 2 1 7 1 1 1
SA 1 2 2 1 1 1
TABLE 6.6
Pairwise comparison matrix for local government offcials (second model)
Criteria DH DR
DH 1 2
DR 1/2 1
TABLE 6.7
Pairwise comparison matrix for supply chain company executives (second model)
Criteria PI SC LC UI DH DR IG
PI 1 2 4 6 8 9 4
SC 1/2 1 2 6 8 9 1
LC 1/4 1/2 1 1 1 1 1
UI 1/6 1/6 1 1 1 1 1
DH 1/8 1/8 1 1 1 1 1
DR 1/9 1/9 1 1 1 1 1
IG 1/4 1 1 1 1 1 1
TABLE 6.8
Impacts for consumers (second model)
Criteria Impacts
IC 0.1621
DH 0.1515
DR 0.0960
SP 0.1434
EO 0.2533
SA 0.1938
© 2009 by Taylor & Francis Group, LLC
Evaluation of Collection Centers 97
TABLE 6.9
Impacts for local government offcials (second model)
Criteria Impacts
DH 0.6667
DR 0.3333
TABLE 6.10
Impacts for supply chain company executives (second model)
Criteria Impacts
PI 0.3876
SC 0.2599
LC 0.0781
UI 0.0634
DH 0.0598
DR 0.0585
IG 0.0927
TABLE 6.11
Decision matrix for consumers (second model)
Collection
centers IC DH DR SP EO SA
E 8 1 6 2 2 3
F 2 1 7 3 2 3
G 3 2 4 1 1 5
TABLE 6.12
Decision matrix for local government offcials (second model)
Collection centers DH DR
E 2 1
F 2 1
G 3 2
TABLE 6.13
Decision matrix for supply chain company executives (second model)
Collection
centers PI SC LC UI DH DR IG
E 1 3 2 1 3 4 3
F 2 4 7 1 2 2 2
G 10 9 8 7 5 1 6
© 2009 by Taylor & Francis Group, LLC
98 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
in relevant survey questionnaires, individual interviews, focus groups, and
onsite observations.
Now one is ready to perform the six steps in the TOPSIS for each category
of decision makers. The following steps show the implementation of the
TOPSIS for the consumers to evaluate E, F, and G.
Step 1: Construct the normalized decision matrix. Table 6.14 shows the
normalized decision matrix formed by applying equation (4.49) on
each element of table 6.11 (decision matrix for the consumers). For
example, the normalized rank of collection center F with respect to
criterion DH (see tables 6.11 and 6.14) is calculated as follows:
r
22
2 2 2
1
1 1 2
0 4082 =
÷ ÷
= . .
Step 2: Construct the weighted normalized decision matrix. Table 6.15 shows
the weighted normalized decision matrix for the consumers. This
is constructed using the impacts of the criteria listed in table 6.8
and the normalized decision matrix in table 6.14. For example, the
weighted normalized rank of collection center F with respect to cri
terion DH, i.e., 0.0619 (see table 6.15), is calculated by multiplying the
impact of DH, i.e., 0.1515 (see table 6.8), with the normalized rank of
F with respect to DH, i.e., 0.4082 (see table 6.14).
Step 3: Determine the ideal and negativeideal solutions. Each column in
the weighted normalized decision matrix shown in table 6.15 has a
TABLE 6.14
Normalized decision matrix for consumers (second model)
Collection
centers IC DH DR SP EO SA
E 0.9117 0.4082 0.5970 0.5345 0.6667 0.4575
F 0.2279 0.4082 0.6965 0.8018 0.6667 0.4575
G 0.3419 0.8165 0.3980 0.2673 0.3333 0.7625
TABLE 6.15
Weighted normalized decision matrix for consumers (second model)
Collection
centers IC DH DR SP EO SA
E 0.1478 0.0619 0.0573 0.0767 0.1689 0.0887
F 0.0370 0.0619 0.0669 0.1150 0.1689 0.0887
G 0.0554 0.1237 0.0382 0.0383 0.0844 0.1478
© 2009 by Taylor & Francis Group, LLC
Evaluation of Collection Centers 99
minimum rank and a maximum rank. They are the ideal and nega
tiveideal solutions, respectively, for the corresponding criterion.
solution (minimum rank) is 0.0619, and the negativeideal solution
(maximum rank) is 0.1237.
Step 4: Calculate the separation distances. The separation distances (see
table 6.16) for each collection center are calculated using equations
(4.52) and (4.53). For example, the positive separation distance for
collection center F (see table 6.16) is calculated using equation (4.52),
which contains the weighted normalized ranks of F (see table 6.15)
and the ideal solutions (obtained in step 3) for the criteria.
Step 5: Calculate the relative closeness coeffcient. Using equation (4.54),
the relative closeness coeffcient is calculated for each collection
center (see table 6.17). For example, the relative closeness coeffcient
(i.e., 0.5435) for collection center F (see table 6.17) is the ratio of F’s
negative separation distance (i.e., 0.1400) to the sum (i.e., 0.1400 +
0.1176 = 0.2576) of its negative and positive separation distances (see
table 6.16).
Step 6: Form the preference order. Because the best alternative is the one
with the highest relative closeness coeffcient, the preference order
for the collection centers is G, F, and E (that means G is the best col
lection center, as evaluated by the consumers).
The TOPSIS is implemented for the local government offcials and the sup
ply chain company executives in a similar manner. The relative closeness
coeffcients of the collection centers, as calculated for those two categories,
are shown in table 6.18.
Figure 6.2 shows the mapping classifcation, as created using MapLand
[4], for relative closeness coeffcients. Figures 6.3–6.5 show the mapping of
TABLE 6.16
Separation distances for consumers (second model)
Collection centers S* S–
E 0.1458 0.0942
F 0.1176 0.1400
G 0.0875 0.1495
TABLE 6.17
Relative closeness coeffcients for consumers (second model)
Collection centers C*
E 0.3926
F 0.5435
G 0.6308
For example (see table 6.15), with respect to criterion DH, the ideal
© 2009 by Taylor & Francis Group, LLC
100 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
TABLE 6.18
Relative closeness coeffcients for local government offcials and supply chain
company executives (second model)
Collection centers Local government offcials
Supply chain company
executives
E 1.000 0.902
F 1.000 0.851
G 0.739 0.091
0.00 to 0.20 0.21 to 0.40
0.41 to 0.60
0.61 to 0.80
0.81 to 1.00
FIGURE 6.2
Relative closeness mapping classifcation (second model).
C1
C2
C3
FIGURE 6.3
Mapping of collection centers for consumers (second model).
© 2009 by Taylor & Francis Group, LLC
Evaluation of Collection Centers 101
the relative closeness coeffcients as calculated for the collection centers by
the consumers, local government offcials, and supply chain company execu
tives, respectively.
It is assumed that each collection center is in one of the three regions, and
the three regions are mapped with respect to the relative closeness coeffcients.
The darker the region is, the higher the effciency the corresponding collection
center has (with respect to the corresponding category of decision makers).
6.4.2 Phase II (Group Decision Making)
Table 6.19 shows the marks of the collection centers as given using Borda’s
choice rule for the consumers, local government offcials, and supply chain
company executives. Borda scores (group evaluations) calculated for E, F, and
G (viz., 3, 5, and 3, respectively) are also shown. For example, the Borda score
C1 C2
C3
FIGURE 6.4
Mapping of collection centers for local government offcials (second model).
C1
C3
C2
FIGURE 6.5
Mapping of collection centers for supply chain company executives (second model).
© 2009 by Taylor & Francis Group, LLC
102 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
for F (i.e., 5) is calculated by sum
ming the marks of F for consumers,
local government offcials, and sup
ply chain company executives (i.e., 1
+ 2 + 2). Because F has the highest
Borda score, it is the best of the lot.
Figure 6.6 shows the mapping clas
sifcation, as created using MapLand
[4], for the Borda scores. Figure 6.7
shows the mapping of the Borda
score of each collection center. It is
obvious that F is the one with the
highest effciency.
TABLE 6.19
Marks and Borda scores of collection centers (second model)
Collection
centers Consumers
Local
government
offcials
Supply chain
company
executives Borda scores
E 0 2 1 3
F 1 2 2 5
G 2 1 0 3
0.0 to 2.0
2.1 to 4.0
4.1 to 6.0
FIGURE 6.6
Borda score mapping classifcation (second
model).
C1
C3
C2
FIGURE 6.7
Mapping of collection centers with respect to Borda score (second model).
© 2009 by Taylor & Francis Group, LLC
Evaluation of Collection Centers 103
6.5 Third Model (Neural Networks,
Fuzzy Logic, TOPSIS, Borda’s Rule)
This model assumes that neither numerical nor linguistic impacts are avail
able for the evaluation criteria. It employs a neural network [11] to evaluate
the effciency of a collection center of interest (which is being considered for
inclusion in a reverse supply chain), using linguistic performance measures
of collection centers that already exist in the reverse supply chain. To this
end, the model to evaluate the effciency of a collection center of interest is
carried out in three phases. In the frst phase, the ratings of existing collec
tion centers are used to construct a neural network that, in turn, calculates
impacts of criteria identifed for each category of decision makers given in
section 6.3. In the second phase, the impacts obtained in the frst phase are
used in a fuzzy TOPSIS (combination of fuzzy logic and TOPSIS) method to
obtain the overall rating of the collection center of interest, as calculated for
each category. Finally, in the third phase, Borda’s choice rule is employed
to calculate the maximized consensus rating (among the categories consid
ered), i.e., effciency, of the recovery facility of interest.
6.5.1 Phase I (Derivation of Impacts)
Suppose that one has the linguistic ratings of ten existing collection centers,
as given by an expert in each category of decision makers described in section
6.3. Using fuzzy logic, these linguistic ratings are converted into triangular
fuzzy numbers (TFNs). Table 6.20 shows not only one of the many ways for
conversion of linguistic ratings into TFNs but also the defuzzifed ratings of
the corresponding TFNs. Tables 6.21–6.23 show the defuzzifed overall rating
of each existing collection center as well as the collection center’s defuzzifed
rating with respect to each criterion, as evaluated by the consumers, local
government offcials, and supply chain company executives, respectively.
Defuzzifcation of a TFN can be performed using equation (4.8).
A neural network is constructed and trained for each category of decision
makers, using the defuzzifed ratings of the existing collection centers with
respect to criteria as input sets and the collection centers’ defuzzifed overall
TABLE 6.20
Conversion table for ratings (third model)
Linguistic ratings TFNs Defuzzifed ratings
Very good (VG) (7, 10, 10) 9
Good (G) (5, 7, 10) 7.3
Fair (F) (2, 5, 8) 5
Poor (P) (1, 3, 5) 3
Very poor (VP) (0, 0, 3) 1
© 2009 by Taylor & Francis Group, LLC
104 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
ratings as corresponding outputs. In the example, there are ten input–out
put pairs for each neural network because there are ten existing collection
centers. Also, three layers are considered in each network, with fve nodes in
the hidden layer. The number of nodes in the output layer is one (for overall
rating), and the number in the input layer is the number of criteria consid
ered by the corresponding category. For example, fgure 6.8 shows the neural
network constructed and trained for the category of consumers.
After each neural network is trained, the following equation [2] is used to
calculate the impacts of criteria considered by the corresponding category.
Here, the absolute value of W
v
is the impact of the vth input node upon the out
put node, n
V
is the number of input nodes, n
H
is the number of hidden nodes, I
ij
is the connection weight from the ith input node to the jth hidden node, and O
j
is the connection weight from the jth hidden node to the output node:
TABLE 6.21
Ratings for consumers (third model)
Collection
centers IC DH DR SP EO SA Overall
C1 1 3 5 3 5 9 5
C2 9 1 3 5 7.3 9 7.3
C3 3 1 3 1 9 1 3
C4 3 9 1 7.3 1 7.3 5
C5 5 1 3 5 1 3 7.3
C6 9 3 7.3 3 5 7.3 3
C7 5 7.3 9 1 7.3 9 1
C8 1 5 1 5 3 1 9
C9 1 5 5 9 9 5 5
C10 5 9 5 3 9 3 1
TABLE 6.22
Ratings for local government offcials (third model)
Collection centers DH DR Overall
C1 1 3 5
C2 9 1 7.3
C3 3 1 3
C4 3 9 5
C5 5 1 7.3
C6 9 3 3
C7 5 7.3 1
C8 1 5 9
C9 1 5 5
C10 5 9 1
© 2009 by Taylor & Francis Group, LLC
Evaluation of Collection Centers 105
 
 
 
W
I
I
O
I
I
O
v
vj
ij
i
n
j
j
n
vj
ij
i
n
j
V
H
V
=
(
(
·
·
·
·
`
`
`
··
·
·
·
·
·
·
·
·
\
)
(
(
·
·
·
·
·
·
·
·
·
·
`
j
n
H
··
·
·
\
)
`
i
n
V
(6.2)
TABLE 6.23
Ratings for supply chain company executives (third model)
Collection
centers PI SC LC UI DH DR IG Overall
C1 1 3 1 3 5 1 3 5
C2 9 1 3 7.3 3 5 7.3 7.3
C3 3 1 7.3 9 1 7.3 9 3
C4 3 9 5 1 5 3 1 5
C5 5 1 5 5 9 9 5 7.3
C6 9 3 9 5 3 9 3 3
C7 5 7.3 3 1 7.3 9 1 1
C8 1 5 1 3 1 3 5 9
C9 1 5 3 5 9 9 1 5
C10 5 9 1 3 5 7.3 7.3 1
IC
PH
PR
SP
EO
SA
H
H
H
H
H
O
Input layer Hidden layer Output layer
FIGURE 6.8
Neural network for consumers (third model).
© 2009 by Taylor & Francis Group, LLC
106 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
Tables 6.24–6.26 show the impacts of the criteria considered for the consumers,
local government offcials, and supply chain company executives, respectively.
6.5.2 Phase II (Individual Decision Making)
Suppose that there are three collection centers, C11, C12, and C13, of inter
est. A fuzzy TOPSIS method uses the impacts obtained in the frst phase to
calculate the overall ratings of the three collection centers.
The decision matrices formed for the consumers, local government off
cials, and supply chain company executives (with defuzzifed ratings for
C11, C12, and C13) in this example are shown in tables 6.27–6.29, respectively
(Table 6.20 is used here, as well, to convert linguistic ratings given by each
category into TFNs.) As in the second model (see section 6.3), the matrices
can be constructed by inviting representatives from each category of deci
sion makers to participate in relevant survey questionnaires, individual
interviews, focus groups, and onsite observations.
Now one is ready to perform the six steps in the TOPSIS for each category
of decision makers. The following steps show the implementation of the
TOPSIS for the consumers, to evaluate C11, C12, and C13.
Step 1: Construct the normalized decision matrix. Table 6.30 shows the
normalized decision matrix formed by applying equation (4.49) on
each element of table 6.27 (decision matrix for the consumers). For
example, the normalized rating of collection center C12 with respect
to criterion DH (see tables 6.27 and 6.30) is calculated as follows:
TABLE 6.24
Impacts for consumers (third model)
Criteria IC DH DR SP EO SA
Impacts 0.01 0.13 0.06 0.18 0.19 0.43
TABLE 6.25
Impacts for local government offcials (third model)
Criteria DH DR
Impacts 0.33 0.67
TABLE 6.26
Impacts for supply chain company executives (third model)
Criteria PI SC LC UI DH DR IG
Impacts 0.24 0.09 0 0.18 0.25 0.1 0.13
© 2009 by Taylor & Francis Group, LLC
Evaluation of Collection Centers 107
r
22
2 2 2
1
9 1 3
0 105 =
÷ ÷
= . .
Step 2: Construct the weighted normalized decision matrix. Table 6.31 shows
the weighted normalized decision matrix for the consumers. This
is constructed using the impacts of the criteria listed in table 6.24
and the normalized decision matrix in table 6.30. For example, the
TABLE 6.27
Decision matrix for consumers (third model)
Collection
centers IC DH DR SP EO SA
C11 3 9 1 7.33 1 7.33
C12 5 1 3 5 1 3
C13 9 3 7.33 3 5 7.33
TABLE 6.28
Decision matrix for local government offcials (third model)
Collection centers DH DR
C11 3 9
C12 5 1
C13 9 3
TABLE 6.29
Decision matrix for supply chain company executives (third model)
Collection
centers PI SC LC UI DH DR IG
C11 5 1 5 5 9 9 5
C12 9 3 9 5 3 9 3
C13 5 7.33 3 1 7.33 9 1
TABLE 6.30
Normalized decision matrix for consumers (third model)
Collection
centers IC DH DR SP EO SA
C11 0.278 0.943 0.125 0.783 0.192 0.679
C12 0.466 0.105 0.376 0.534 0.192 0.278
C13 0.839 0.314 0.918 0.320 0.962 0.679
© 2009 by Taylor & Francis Group, LLC
108 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
weighted normalized rank of collection center C12 with respect to
criterion DH, i.e., 0.014 (see table 6.31), is calculated by multiplying
the impact of DH, i.e., 0.129 (see table 6.24), with the normalized rat
ing of C12 with respect to DH, i.e., 0.105 (see table 6.30).
Step 3: Determine the ideal and negativeideal solutions. Each column in
the weighted normalized decision matrix shown in table 6.31 has a
maximum rating and a minimum rating. They are the ideal and neg
ativeideal solutions, respectively, for the corresponding criterion.
For example (see table 6.31), with respect to criterion DH, the ideal
solution (maximum rating) is 0.122, and the negativeideal solution
(minimum rating) is 0.014.
Step 4: Calculate the separation distances. The separation distances (see
table 6.32) for each collection center are calculated using equations
(4.52) and (4.53). For example, the positive separation distance for col
lection center C12 (see table 6.32) is calculated using equation (4.52),
which contains the weighted normalized ratings of C12 (see table 6.31)
and the ideal solutions (obtained in step 3) for the criteria.
Step 5: Calculate the relative closeness coeffcient. Using equation (4.54),
the relative closeness coeffcient for each collection center is calcu
lated (see table 6.33). For example, the relative closeness coeffcient
(i.e., 0.137) for collection center C12 (see table 6.33) is the ratio of
C12’s negative separation distance (i.e., 0.041) to the sum (i.e., 0.041
+ 0.257 = 0.298) of its negative and positive separation distances (see
table 6.32).
Step 6: Rank the preference order. Because the best alternative is the one
with the highest relative closeness coeffcient, the preference order
TABLE 6.31
Weighted normalized decision matrix for consumers (third model)
Collection
centers IC DH DR SP EO SA
C11 0.004 0.122 0.007 0.140 0.036 0.294
C12 0.006 0.014 0.022 0.095 0.036 0.120
C13 0.011 0.041 0.053 0.057 0.181 0.294
TABLE 6.32
Separation distances for consumers (third model)
Collection centers S* S–
C11 0.152 0.221
C12 0.257 0.041
C13 0.116 0.233
© 2009 by Taylor & Francis Group, LLC
Evaluation of Collection Centers 109
for the collection centers is C13, C11, and C12 (that means C13 is the
best collection center, as evaluated by the consumers).
The TOPSIS is implemented for the local government offcials and the sup
ply chain company executives in a similar manner. The relative closeness
coeffcients of the collection centers, as calculated for those two categories,
are shown in table 6.34.
6.5.3 Phase III (Group Decision Making)
Table 6.35 shows the marks of the collection centers as given using Borda’s
choice rule for the consumers, local government offcials, and supply chain
company executives. Borda scores (group evaluations) calculated for C11,
C12, and C13 (viz., 5, 1, and 3, respectively) are also shown. For example, the
Borda score for C12 (i.e., 1) is calculated by summing the marks of C12 for
the consumers, local government offcials, and supply chain company execu
tives (i.e., 0 + 0 + 1). Because C11 has the highest Borda score, it is the best of
the lot.
TABLE 6.33
Relative closeness coeffcients for consumers (third model)
Collection centers C*
C11 0.592
C12 0.137
C13 0.668
TABLE 6.34
Relative closeness coeffcients for local government offcials and supply chain
company executives (third model)
Collection centers Local government offcials
Supply chain company
executives
C11 0.754 0.619
C12 0.096 0.502
C13 0.354 0.415
TABLE 6.35
Marks and Borda scores of collection centers (third model)
Collection
centers Consumers
Local
government
offcials
Supply chain
company
executives Borda scores
C11 1 2 2 5
C12 0 0 1 1
C13 2 1 0 3
© 2009 by Taylor & Francis Group, LLC
110 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
6.6 Fourth Model (ANP and Goal Programming)
In this model, frst the analytic network process (ANP) is used to calculate
the performance indices (effciency scores) of candidate collection centers,
with respect to qualitative criteria taken from the perspective of a reman
ufacturing facility interested in buying used products from the collection
centers. Then goal programming is employed to determine the quantities of
used products to be transported from the candidate collection centers to the
remanufacturing facility while satisfying two important goals of the reman
ufacturing facility: to maximize total value of purchase and minimize total
cost of purchase. Sections 6.6.1 and 6.6.2 present the applications of ANP and
goal programming, respectively.
6.6.1 Application of ANP
The problem of evaluating the effciencies of the candidate collection centers
is framed as a fourlevel hierarchy (see fgure 6.9). The frst level contains
the objective of evaluation of the candidate collection centers. The second
level consists of the main evaluation criteria taken from the perspective of a
remanufacturing facility. The third level contains the subcriteria under each
Reliability
Capabilities/
Responsiveness
Financial Issues
Cultural and
Strategic Issues
Others
Evaluating
Suppliers
Delivery Reliability
Conformance to
Speciﬁcations
Order Fulﬁllment Lead
Time
Flexibility (Ability to adapt
to demand ﬂuctuations)
Suppliers Design
Capability
Return/Warranty
processing costs
Pricing Structure
Quantity Discounts
Financial Stability &
Economic Performance
Level of coop and
Information Exchange
Technical Capability
Reputation in Industry
Proximity
Safety and
Environmental Issues
Collection
Center 1
Collection
Center 2
Collection
Center 3
FIGURE 6.9
Hierarchical structure for ANP (fourth model).
© 2009 by Taylor & Francis Group, LLC
Evaluation of Collection Centers 111
main criterion. The fourth level contains the candidate collection centers.
The main and subcriteria considered are the following (see [5–9]):
Reliability r : This criterion relates to a collection center delivering the
used products at the right time, at the right remanufacturing facility,
in the right quantity, and in the promised condition. The subcriteria
considered under this main criterion are (1) delivery reliability and
(2) conformance to standards.
Capability/responsiveness r : This criterion refects the velocity at which a
collection center supplies the used products to the remanufacturing
facility and the collection center’s ability to adapt to sudden demand
fuctuations. The subcriteria considered here are (1) order fulfllment
lead time, (2) fexibility in adapting to demand fuctuations, and (3)
design capabilities.
Financial issues r : This criterion refects the costs and other fnancial
aspects involved. The subcriteria are (1) return/warranty processing
costs, (2) pricing structure, (3) quantity discounts, and (4) fnancial
stability and economic performance of the collection center.
Cultural and strategic issues r : This criterion consists of the follow
ing subcriteria: (1) level of cooperation and information exchange
between the collection center and the remanufacturing facility, (2)
the collection center’s reputation in the industry, and (3) the collec
tion center’s technical capability (how knowledgeable the collection
center is about the product).
Others r : This criterion considers miscellaneous aspects that are not
considered in the other criteria. These aspects are (1) proximity of the
collection center to the remanufacturing facility (it affects the trans
portation cost and the transit time) and (2) safety and environmental
aspects (because the collection center and the remanufacturing facil
ity are closely involved, any safety issues with the collection center
directly refect on the remanufacturing facility’s reputation; envi
ronmental aspects are concerned about the collection center’s effort
in pursuing environmental consciousness or a “green” image).
It is assumed that there exist interdependencies among the subcriteria on
the third level in the hierarchy.
Three collection centers, S1, S2, and S3, are considered in the numerical
example to illustrate the application of ANP. Table 6.36 shows the pairwise
comparison matrix for the main criteria (second level in the hierarchy) and
also the normalized eigen vector of the matrix. The elements of the normal
ized eigen vector are the impacts given to the main criteria with respect to
the objective (frst level in the hierarchy).
Tables 6.37–6.41 show the pairwise comparison matrices of subcriteria with
respect to their main criteria and also the corresponding normalized eigen
vectors of the matrices. Note that each of the matrices in tables 6.36–6.41 has
© 2009 by Taylor & Francis Group, LLC
112 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
a consistency ratio CR less than or equal to 0.1. For example, for the matrix in
table 6.39, CR, using equation (4.1), is
( )
( )( )
.
( ) .
.
max L ÷
÷
=
÷
÷
=
n
n R 1
4 01037 4
4 1 0 90
0 0038..
Table 6.42 shows the matrix of interdependencies (called the super matrix
M) among the subcriteria with respect to their main criteria. This super
matrix M is made to converge to obtain a longterm stable set of impacts. For
convergence, M must be made column stochastic, which is done by raising M
to the power of 2
k+1
, where k is an arbitrarily large number. In the example, k
= 59. Table 6.43 shows the converged super matrix.
TABLE 6.36
Comparative importance values of main criteria (fourth model)
Criteria Reliability
Responsive
ness
Financial
issues
Cultural
and
strategic
issues Others
Normalized
eigen vector
Reliability 1 1/4 3 1 1/6 0.114
Responsiveness 4 1 1 5 3 0.431
Financial issues 1/3 1/5 1 1/2 1/3 0.065
Cultural and
strategic issues
1 1/3 2 1 1/3 0.102
Others 6 1/3 3 3 1 0.285
TABLE 6.37
Comparative importance values of subcriteria under reliability (fourth model)
Subcriteria
Delivery
reliability
Conformance to
specs
Normalized eigen
vector
Delivery reliability (DR) 1 1/3 0.25
Conformance to specs (CS) 3 1 0.75
TABLE 6.38
Comparative importance values of subcriteria under responsiveness (fourth
model)
Subcriteria
Order
fulfllment lead
time (OFT) Flexibility (F)
Design
capability (DC)
Normalized
eigen vector
Order fulfllment
lead time (OFT)
1 3 4 0.623
Flexibility (F) 1/3 1 2 0.239
Design
capability (DC)
1/4 1/2 1 0.137
© 2009 by Taylor & Francis Group, LLC
Evaluation of Collection Centers 113
TABLE 6.39
Comparative importance values of subcriteria under “fnancial issues” (fourth
model)
Subcriteria
Returns/
warranty
processing
costs (RC)
Pricing
structure (PS)
Quantity
discounts
(QD)
Economic
performance
and fnancial
stability (EP)
Normalized
eigen vector
Returns/
warranty
processing
costs (RC)
1 1/3 1/2 1/4 0.099
Pricing
structure
(PS)
3 1 2 1 0.345
Quantity
discounts
(QD)
2 1/2 1 1/2 0.185
Economic
performance
and fnancial
stability (EP)
4 1 2 1 0.37
TABLE 6.40
Comparative importance values of subcriteria under cultural and strategic issues
(fourth model)
Subcriteria
Level of coop
and information
exchange
(Coop)
Technical
capability
(TC) Reputation
Normalized
eigen vector
Level of coop and
information exchange
(Coop)
1 5 6 0.722
Technical capability (TC) 1/5 1 2 0.174
Reputation (R) 1/6 1/2 1 0.103
TABLE 6.41
Comparative importance values of subcriteria under others (fourth model)
Subcriteria Proximity (P)
Safety and
environment (SE)
Normalized eigen
vector
Proximity (P) 1 1/3 0.25
Safety and
environment (SE)
3 1 0.75
© 2009 by Taylor & Francis Group, LLC
114 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
TABLE 6.42
Matrix of interdependencies (super matrix M) (fourth model)
DR CS OFT F DC RC PS QD EP Coop TC R P SE
DR 0 1 0 0 0 0 0 0 0 0 0 0 0 0
CS 1 0 0 0 0 0 0 0 0 0 0 0 0 0
OFT 0 0 0 0.8 0.75 0 0 0 0 0 0 0 0 0
F 0 0 0.75 0 0.25 0 0 0 0 0 0 0 0 0
DC 0 0 0.25 0.2 0 0 0 0 0 0 0 0 0
RC 0 0 0 0 0 0 0.29 0.13 0.201 0 0 0 0 0
PS 0 0 0 0 0 0.53 0 0.62 0.6 0 0 0 0 0
QD 0 0 0 0 0 0.29 0.16 0 0.11 0 0 0 0 0
EP 0 0 0 0 0 0.163 0.53 0.23 0 0 0 0 0 0
Coop 0 0 0 0 0 0 0 0 0 0 0.75 0.83 0 0
TC 0 0 0 0 0 0 0 0 0 0.8 0 0.16 0 0
R 0 0 0 0 0 0 0 0 0 0.2 0.25 0 0 0
P 0 0 0 0 0 0 0 0 0 0 0 0 0 1
SE 0 0 0 0 0 0 0 0 0 0 0 0 1 0
TABLE 6.43
Converged super matrix (fourth model)
Subcriteria Stabilized relative impact
Delivery reliability 1
Conformance to specs 1
Order fulfllment LT 0.44
Flexibility 0.38
Design capability 0.18
Returns/warranty 0.19
Pricing 0.38
Quantity discounts 0.15
Stability and economic performance 0.27
Coop and information exchange 0.44
Technical capability 0.38
Reputation 0.18
Proximity 1
Safety and environment 1
© 2009 by Taylor & Francis Group, LLC
Evaluation of Collection Centers 115
Table 6.44 shows the relative ratings of the candidate collection centers, S1,
S2, and S3, with respect to the subcriteria. These ratings are obtained after car
rying out pairwise comparisons between the candidate collection centers with
respect to the subcriteria and then obtaining the normalized eigen vector.
To obtain pairwise comparisons, interdependencies, and relative ratings
(see tables 6.37–6.44), decision makers could be invited to participate in rel
evant survey questionnaires, individual interviews, focus groups, and on
site observations.
Table 6.45 shows the desirability index calculated for each candidate col
lection center using equation (4.2).
The overall performance index for each of the three collection centers
is calculated by multiplying the desirability index (see table 6.45) of each
collection center for each main criterion by the impact of that criterion (see
TABLE 6.44
Relative ratings of collection centers with respect to subcriteria (fourth model)
Subcriteria/alternate
collection centers S1 S2 S3
DR 0.33 0.141 0.524
CS 0.345 0.543 0.11
OFT 0.274 0.068 0.657
F 0.109 0.309 0.581
DC 0.09 0.25 0.652
RC 0.681 0.216 0.102
PS 0.309 0.581 0.109
QD 0.376 0.151 0.471
EP 0.137 0.623 0.239
Coop 0.33 0.075 0.59
TC 0.137 0.239 0.623
R 0.67 0.23 0.12
P 0.292 0.092 0.615
SE 0.137 0.239 0.623
TABLE 6.45
Desirability indices (fourth model)
Criteria/collection
centers S1 S2 S3
Reliability 0.3429 0.4432 0.2138
Responsiveness 0.0874 0.0528 0.2488
Financial issues 0.0783 0.148551 0.054
Cultural and strategic
issues
0.1271 0.0438 0.2299
Others 0.176 0.2027 0.6211
© 2009 by Taylor & Francis Group, LLC
116 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
table 6.36) and summing over all the criteria. Table 6.46 shows the overall
performance indices (effciencies) for the three collection centers.
6.6.2 Application of Goal Programming
This section presents the application of goal programming to determine the
quantities of used products to be transported from the candidate collection
centers to a remanufacturing facility of interest while satisfying two impor
tant goals of the remanufacturing facility: to maximize total value of purchase
and minimize total cost of purchase. Section 6.6.2.1 gives the nomenclature
used in the methodology, and section 6.6.2.2 presents the problem formula
tion and a numerical example.
6.6.2.1 Nomenclature for Problem Formulation
c
i
Unit purchasing cost of used product at collection center i
d
j
Demand for used product j
g Goal index
i Collection center index, i = 1, 2, …, s
k
i
Capacity of collection center i
p
i
Probability of breakage of used products purchased from collec
tion center i
p
max
Maximum allowable probability of breakage
Q
i
Decision variable representing the quantity to be purchased from
collection center i
s Number of candidate collection centers
w
i
Performance index of collection center i obtained by carrying out
ANP
6.6.2.2 Problem Formulation
The following two goals of the remanufacturing facility are considered:
1. Maximize the total value of purchase (TVP).
2. Minimize the total cost of purchase (TCP).
TABLE 6.46
Overall performance indices (fourth model)
Collection center Performance index
S1 0.2318
S2 0.2322
S3 0.5359
© 2009 by Taylor & Francis Group, LLC
Evaluation of Collection Centers 117
Whereas the frst goal involves minimizing the underachievement of the
target, the second goal involves minimizing the overachievement of the tar
get. It is at the discretion of the decision maker to add any other goals that
are relevant to the situation.
Goal 1: Maximize TVP: w Q
i i
i
s
=
`
1
(6.3)
Goal 2: Maximize TCP:
c Q
i i
`
(6.4)
Capacity constraint: Q k
i i
_ (6.5)
Demand constraint: Q d
i
i
j
`
= (6.6)
Quality constraint: d p Q p
j i i
i
s
_
=
`
max
1
(6.7)
Nonnegativity constraint: Q
i
_0 (6.8)
The three candidate collection centers from section 6.6.1 are considered
in this numerical example as well. Table 6.47 shows the data used for the
goal programming problem (note that only one used product type is consid
ered), and table 6.48 shows the results obtained by solving the problem using
LINGO (v4).
From the application of ANP (see section 6.6.1), S3 is the highestranked
collection center. When no other system constraints are in place, 750 units
might be ordered from S3 before considering other collection centers. How
ever, from the results obtained from the application of goal programming,
it can be noticed that 650 units are ordered from S2 and the remaining 350
units are ordered from S3. This may be attributed to the fact that the unit
purchasing cost at S3 is higher than that at S2 (this was not considered in the
application of ANP, where collection centers were evaluated with respect to
qualitative criteria).
The total cost of purchase is found to be $935, and the total value of pur
chase is 338.54 (aspiration level = 250).
© 2009 by Taylor & Francis Group, LLC
118 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
6.7 Fifth Model (Eigen Vector Method, Taguchi
Loss Function, and Goal Programming)
In this model, frst the eigen vector method and Taguchi loss function are
used to calculate the weighted Taguchi losses (ineffciency scores) of can
didate collection centers, with respect to qualitative criteria taken from the
perspective of a remanufacturing facility interested in buying used products
from the collection centers. Then goal programming is employed to deter
mine the quantities of used products to be transported from the candidate
collection centers to the remanufacturing facility while satisfying two impor
tant goals of the remanufacturing facility: to minimize total loss of proft and
minimize total cost of purchase. Section 6.7.1 presents the application of the
eigen vector method and Taguchi loss function. Then section 6.7.2 presents
the application of goal programming.
6.7.1 Application of Eigen Vector Method and Taguchi Loss Function
The following four qualitative criteria (taken from the perspective of a remanu
facturing facility) are considered to evaluate the candidate collection centers:
Quality of used products (the smaller the defect rate of the used r
products supplied, the better)
TABLE 6.47
Data for goal programming model (fourth model)
Collection center S1 S2 S3
Capacity 300 650 750
Unit purchasing cost 1.2 0.9 1.0
Breakage probability 0.03 0.015 0.01
Net demand for the product = 1,000
Maximum acceptable breakage probability = 0.025
TABLE 6.48
Results (fourth model)
Collection center ANP rating Quantity ordered
S1 0.2318 0
S2 0.2322 650
S3 0.5359 350
Total value of purchase (TVP) = 338.54
Total cost of purchase (TCP) = 935
© 2009 by Taylor & Francis Group, LLC
Evaluation of Collection Centers 119
Ontime delivery (the smaller the number of delayed deliveries, the r
better)
Proximity (the closer the collection center from the remanufacturing r
facility, the better)
Cultural and strategic issues, such as fexibility in adapting to r
demand fuctuations, level of cooperation and information exchange,
green image, and fnancial stability/economic performance (see sec
tion 6.6.1 for explanations of these issues)
For the numerical example, three candidate collection centers, S4, S5, and
S6, are considered. Table 6.49 shows the pairwise comparison matrix for the
criteria* and also the normalized eigen vector of the matrix. The elements
of the normalized eigen vector represent the impacts given to the criteria.
Table 6.49 has a consistency ratio CR of less than 0.1. Specifcally, it is
( )
( )( )
.
( ) .
. .
max L ÷
÷
=
÷
÷
=
n
n R 1
4 13 4
4 1 0 90
0 048
It could be diffcult to quantify the cultural and strategic issues criterion
for the calculation of Taguchi losses. To this end, Monczka and Trecha’s [10]
service factor rating (SFR) is used. SFR includes performance factors diffcult
to quantify but decisive in the selection process. In practice, experts rate these
performance factors. For a given collection center, these ratings on all factors
are summed and averaged to obtain a total service rating. The collection
center’s service factor percentage is obtained by dividing the total service
rating by the total number of points possible. Table 6.50 shows the service
factor ratings for the various aspects of the cultural and strategic issues crite
rion. The ratings are given on a scale of 1–10, the level of performance being
directly proportional to the rating.
*
Only the main criteria are considered in this model.
TABLE 6.49
Comparative importance values of criteria (ffth model)
Criteria Quality
Ontime
delivery Proximity
Cultural and
strategic
issues
Normalized
eigen vector
Quality 1 3 7 1 0.384899
Ontime
delivery
0.33 1 4 0.2 0.137363
Proximity 0.14 0.25 1 0.166 0.052674
Cultural and
strategic
issues
1 5 6 1 0.425064
© 2009 by Taylor & Francis Group, LLC
120 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
Table 6.51 shows the decision variables for calculating the Taguchi losses
for the three collection centers.
Consider the quality criterion. The target defect rate/breakage probability
is zero, at which there is no loss to the remanufacturing facility, and the
upper specifcation limit for the defect rate/breakage probability is 30%,
at which there is 100% loss to the manufacturer. For ontime delivery, the
remanufacturing facility will incur losses if the products are delivered late
and before the scheduled requirement. The specifcation limit of delivery
delay is 5 days, and early delivery is 10 days, meaning that the remanufac
turing facility will incur 100% loss if the deliveries are delayed by 5 days
and are delivered 10 days before the scheduled delivery date. For proximity,
loss will be zero at the closest collection center and the specifcation limit is
up to 40% of the closest collection center, meaning that the remanufacturing
facility will incur 100% loss when the collection center’s distance reaches the
specifcation limit. For cultural and strategic issues, the specifcation limit is
50% for the service factor percentage, at which the loss will be 100%, whereas
there will be no loss incurred at a service factor percentage of 100%. The
values of the loss coeffcient, k (see equations 4.66–4.68), are 1111.11, 625, and
25 for quality, proximity, and cultural and strategic issues, respectively. For
ontime delivery, k
1
= 4 and k
2
= 1 (because an unequal twosided specifca
tion limit is considered for ontime delivery, there exist two loss coeffcients,
k
1
and k
2
).
Table 6.52 shows the characteristic value and the relative value of each
criterion for the three collection centers, S4, S5, and S6. For S4, the quality
TABLE 6.50
Service factor ratings for cultural and strategic issues (ffth model)
Collection
center Flexibility
Level of co
op and
information
exchange
Green
image
Financial
stability and
economic
performance Average
Average ÷
10
S4 7 6 6 4 5.75 57.5%
S5 5 7 8 5 6.25 62.5%
S6 6 5 8 8 6.75 67.5%
TABLE 6.51
Decision variables for selecting collection centers (ffth model)
Criteria Target value Range Specifcation limit
Quality 0% 0–30% 30%
Ontime delivery 0 10–0–5 10 days earlier, 5 days
delay
Proximity Closest 0–40% 40%
Cultural and strategic
issues
100% 100–50% 50%
© 2009 by Taylor & Francis Group, LLC
Evaluation of Collection Centers 121
characteristic value is 15% defect rate, which translates to 15% deviation from
the target value. The relative values, together with the value of the loss coef
fcient, k, are used to calculate (see equations (4.66)–(4.68)) the Taguchi loss
for each collection center for each criterion (see table 6.53).
The weighted Taguchi loss (sum product of criteria impacts and Taguchi
losses) is then calculated from the Taguchi losses of the collection centers
(see table 6.53) and the impacts of the evaluation criteria (see table 6.49).
Table 6.54 shows the weighted Taguchi loss and the normalized Taguchi loss
for each collection center.
6.7.2 Application of Goal Programming
This section presents the application of goal programming to determine the
quantities of used products to be transported from the candidate collection
centers to a remanufacturing facility of interest while satisfying two impor
tant goals of the remanufacturing facility: to minimize total loss of proft
and minimize total cost of purchase. Section 6.7.2.1 gives the nomenclature
TABLE 6.52
Characteristic and relative values of criteria (ffth model)
Quality Ontime delivery Proximity
Cultural and
strategic issues
Collection
center Value
Relative
value Value
Relative
value Value
Relative
value Value
Relative
value
S4 15% 15% +3 +3 8 33.33% 57.5% 57.5%
S5 20% 20% +1 +1 6 0 62.5% 62.5%
S6 10% 10% –8 –8 9 50% 67.5% 67.5%
TABLE 6.53
Taguchi losses (ffth model)
Collection
center Quality
Ontime
delivery Proximity
Cultural and
strategic issues
S4 24.99 36 69.43 75.61
S5 44.44 4 0 64
S6 11.11 64 156.25 54.86
TABLE 6.54
Weighted Taguchi losses (ffth model)
Collection center Weighted Taguchi loss Normalized Taguchi loss
S4 50.37 0.36
S5 44.86 0.32
S6 44.62 0.32
© 2009 by Taylor & Francis Group, LLC
122 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
used in the methodology, and section 6.7.2.2 presents the problem formula
tion and a numerical example.
6.7.2.1 Nomenclature Used in the Methodology
B
j
Budget allocated for collection center j
c
j
Unit purchasing cost of used product at collection center j
d
k
Demand for used product k
g Goal index
j Collection center index, j = 1, 2, …, s
Loss
j
Total loss of collection center j for all the evaluation criteria
r
j
Capacity of collection center j
p
j
Probability of breakage of used products purchased from collec
tion center j
p
max
Maximum allowable probability of breakage
Q
j
Decision variable representing the purchasing quantity from col
lection center j
s Number of candidate collection centers
w
i
Impact of criterion i calculated by the eigen vector method
X
ij
Taguchi loss of collection center j for criterion i
6.7.2.2 Problem Formulation
The following two goals of the remanufacturing facility are considered:
1. Minimize the total loss of proft (TLP).
2. Minimize the total cost of purchase (TCP).
It is at the discretion of the decision maker to add any other goals that are
relevant to the situation.
Goal 1: Minimize TLP: Loss Q TLP
j j
j
s
=
=
`
1
(6.9)
Goal 2: Minimize TCP: c Q TCP
j j
j
s
+ =
=
`
1
(6.10)
© 2009 by Taylor & Francis Group, LLC
Evaluation of Collection Centers 123
Capacity constraint: Q r
i i
_ (6.11)
Demand constraint: Q d
j
j
j
`
= (6.12)
Quality constraint: d p Q p
k j j
j
s
_
=
`
max
1
(6.13)
Nonnegativity constraint: Q
j
_0 (6.14)
The three candidate collection centers from section 6.7.1 are considered in
this numerical example as well. Table 6.55 shows the data used for the goal pro
gramming problem (note that only one used product type is considered), and
table 6.56 shows the results obtained by solving the problem using LINGO (v4).
From the application of eigen vector method and Taguchi loss function (see
section 6.7.1), S6 is the highestranked collection center. When no other system
constraints are in place, 750 units might be ordered from S6 before considering
other collection centers. However, from the results obtained from the applica
tion of goal programming, it can be noticed that 543 units are ordered from S5
and the remaining 457 units are ordered from S6. This may be attributed to
TABLE 6.55
Data for goal programming model (ffth model)
Collection center S4 S5 S6
Capacity 300 650 750
Unit purchasing cost 1.2 0.9 1.0
Breakage probability 0.03 0.015 0.01
Net demand for the product = 1,000
Maximum acceptable breakage probability = 0.025
TABLE 6.56
Results (ffth model)
Collection center Normalized Taguchi loss Quantity ordered
S4 0.360148 0
S5 0.32078 543
S6 0.319072 457
Total loss of purchase (TLP) = 320
Total cost of purchase (TCP) = 945.7
© 2009 by Taylor & Francis Group, LLC
124 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
the fact (besides the system constraints considered in the goal programming
problem) that the unit purchasing cost at S6 is higher than that at S5.
6.8 Conclusions
In this chapter, fve models are presented for identifying effcient collection
centers in a region where a reverse supply chain is to be designed. The frst
model employs eigen vector method and Taguchi loss function. The second
model uses eigen vector method, technique for order preference by similarity to
ideal solution (TOPSIS), and Borda’s choice rule. The third model employs neu
ral networks, fuzzy logic, TOPSIS, and Borda’s choice rule. The fourth model
uses analytic network process (ANP) and goal programming. The ffth model
applies eigen vector method, Taguchi loss function, and goal programming.
References
1. Krajewski, L. J., and Ritzman, L. P. 2004. Operations management: Processes and
value chains, chap. 10. 7th ed. Englewood Cliffs, NJ: Pearson Prentice Hall.
2. Cha, Y., and Jung, M. 2003. Satisfaction assessment of multiobjective schedules
using neural fuzzy methodology. International Journal of Production Research
41:1831–49.
3. Kim, C. W., and Mauborgne, R. 2002. Charting your company’s future. Harvard
Business Review, June, pp. 77–82.
4. http://www.softwareillustrated.com/mapland.html.
5. Choi, T. Y., and Hartley, J. L. 1996. An exploration of supplier selection practices
across the supply chain. Journal of Operations Management 14:333–43.
6. Kannan, V. R., and Tan, K. C. 2002. Supplier selection and assessment: Their
impact on business performance. Journal of Supply Chain Management 38:11–21.
7. Lee, E. K., Ha, S., and Kim, S. K. 2001. Supplier selection and management sys
tem considering relationships in SCM. IEEE Transactions on Engineering Manage
ment 48:307–18.
8. Muralidharan, C., Anantharaman, N., and Deshmukh, S. G. 2002. A multicri
teria group decision making model for supplier rating. Journal of Supply Chain
Management 38:22–33.
9. Weber, C. A., Current, J. R., and Benton, W. C. 1991. Vendor selection criteria
and methods. European Journal of Operational Research 50:2–18.
10. Monczka, R. M., and Trecha, S. J. 1998. Costbased supplier performance evalu
ation. Journal of Purchasing and Material Management 2–7.
11. Haykin, S. 1998. Neural networks: A comprehensive foundation. 2nd ed. Englewood
Cliffs, NJ: Prentice Hall.
© 2009 by Taylor & Francis Group, LLC
125
7
Evaluation of Recovery Facilities
7.1 The Issue
In addition to selecting effcient collection centers (see chapter 6), strate
gic planning of a reverse supply chain involves selecting effcient recovery
facilities where reprocessing operations, such as disassembly and recycling/
remanufacturing, are carried out. The various scenarios for evaluating recov
ery facilities for effciency could differ as follows:
1. Evaluation criteria could be given numerical impacts* (impor
tance values).
2. Evaluation criteria could be presented in terms of ranges of different
degrees of desirability.
3. Decision makers could have conficting criteria for evaluation, and
impacts for evaluation criteria are given.
4. Decision makers could have conficting criteria for evaluation, and
impacts for evaluation criteria are not given (hence must be derived).
5. A very simple evaluation technique could be desired (where only
the “most important” evaluation criteria are considered).
In this chapter, various models to address the above scenarios are pre
sented. The frst model addresses scenario 1 and employs the analytic hierar
chy process. The second model is for scenario 2 and employs linear physical
programming. The third and fourth models are for scenarios 3 and 4, respec
tively. The main difference between these two models lies in the way the
impacts are assigned to the criteria for evaluation of recovery facilities.
Whereas the third model uses the eigen vector method, technique for order
preference by similarity to ideal solution (TOPSIS), and Borda’s choice rule,
the fourth model uses neural networks, fuzzy logic, TOPSIS, and Borda’s
*
In this chapter, we use the term impacts instead of the conventional term weights for impor
tance values of evaluation criteria. This is to avoid confusion in the presence of the connec
tion weights between different nodes of the neural network used in section 7.6.
© 2009 by Taylor & Francis Group, LLC
126 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
choice rule. The ffth model is for scenario 5 and uses a simple twodimen
sional chart to identify effcient recovery facilities.
This chapter is organized as follows: Section 7.2 presents the frst model.
Section 7.3 shows the second model. Section 7.4 presents the different decision
makers (and their criteria) that are considered in the third and fourth models.
Section 7.5 presents the third model. Section 7.6 gives the fourth model. Sec
tion 7.7 presents the ffth model. Finally, section 7.8 gives some conclusions.
7.2 First Model (Analytic Hierarchy Process)
In this section, analytic hierarchy process (AHP) is employed to identify eff
cient facilities from a set of candidate recovery facilities operating in a region
where a reverse supply chain is to be designed. Section 7.2.1 presents the
threelevel hierarchy for the AHP model, and section 7.2.2 gives a numerical
example for the model.
7.2.1 ThreeLevel Hierarchy
The frst level in the hierarchy contains the primary objective, i.e., to identify
effcient facilities from a set of candidate recovery facilities. The last level in
the hierarchy contains the candidate recovery facilities. The level in the mid
dle contains criteria that must somehow be useful in comparing the candi
date recovery facilities. For example, one of the criteria used in the hierarchy
(see fgure 7.1) is the fxed cost of the facility (CO). This criterion can compare
the candidate facilities on the third level.
Though the criteria to be considered in a reverse supply chain seem simi
lar to those considered in a forward supply chain (for example, [3] and [4]),
there are three special factors in a reverse supply chain, that need to be incor
porated in AHP in such a way that the hierarchy levels are not disturbed.
Identiﬁcation of potential recovery facilities
CO QOQI TP/SU TP*DT CS
A B C D
Objective
Criteria
Alternatives
FIGURE 7.1
Threelevel hierarchy for AHP (frst model).
© 2009 by Taylor & Francis Group, LLC
Evaluation of Recovery Facilities 127
Those special factors are average quality of used products, average supply of
used products, and average disassembly time of used products.
Average quality of used products r : Unlike in a forward supply chain,
components of incoming goods (used products) of even the same
type in a recovery facility are likely to be of varied quality (worn
out, low performing, etc). Though the average quality of reprocessed
(recycled/remanufactured) goods (QO) is a criterion that can com
pare two or more candidate facilities, it is not justifed to use QO as
an independent criterion for comparison because QO depends on
average quality of incoming products (QI). Thus, the idea is to take
the difference between QO and QI as a criterion in the hierarchy.
Average supply of used products r : The only driver to design a forward
supply chain is the demand for new products, so, if there is low
demand for new products, there is practically no forward supply
chain. However, this is not the case in some reverse supply chains
where, even if there is low supply of used products (SU), a reverse
supply chain must be administered due to the drivers, such as envi
ronmental regulations and asset recovery. In supplydriven cases
like these, it is unfair to judge a recovery facility without considering
SU in the hierarchy. Though throughput (TP) is a criterion that can
compare two or more candidate recovery facilities, it is not justifed
to use TP as an independent criterion because TP depends on SU. In
other words, a low SU might lead to a low TP, and a high SU might
lead to a high TP. Thus, the idea is to take the ratio of TP to SU as
a criterion in the hierarchy. The effect of a low TP is compensated
for by dividing TP with a possibly low SU (by doing so, the facility
under consideration is not underestimated). Similarly, the effect of
a high TP is dampened by dividing TP with a possibly high SU (by
doing so, the facility under consideration is not overestimated).
Average disassembly time of used products r : The average disassembly
time (DT) is not exactly the inverse of TP because TP takes into
account the whole reprocessing (disassembly plus recycling/reman
ufacturing) time. Unlike in a forward supply chain, components of
incoming goods (used products) in a recovery facility are likely to
be deformed or broken or different in number, even for the same
type of products. Hence, incoming products of the same type might
have different reprocessing times (unlike in a forward supply chain,
where manufacturing time and assembly time are predetermined
and equal for products of the same type). Because TP of a recovery
facility depends upon DT, it is unfair not to consider DT in the hier
archy. In other words, a high DT might lead to a low TP, and a low
DT might lead to a high TP. Thus, the idea is to use the multiplica
tion of TP and DT as a criterion in the hierarchy. The effect of a low
TP is compensated for by multiplying TP with a possibly high DT
© 2009 by Taylor & Francis Group, LLC
128 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
(by doing so, the facility under consideration is not underestimated).
Similarly, the effect of a high TP is dampened by multiplying TP
with a possibly low DT (by doing so, the facility under consideration
is not overestimated).
An intangible criterion considered in this model is termed customer ser
vice (CS). CS is a measure of how well a recovery facility utilizes the incen
tives provided by the government, to what extent it meets the environmental
regulations, what kind of incentives it gives the collection centers that supply
the used products, and what kind of incentives it is giving the customers
buying the reprocessed goods. The allencompassing term customer service
is used here because any benefciary is a customer, be it the government, the
collection center, or the actual customer buying the reprocessed goods.
7.2.2 Numerical Example
Consider the following example: Table 7.1 shows the pairwise comparison
matrix for evaluation of candidate recovery facilities. It also shows the nor
malized eigen vector of the matrix. This vector represents the impacts given
by the decision maker to the criteria.
Tables 7.2–7.6 show comparative ratings of the candidate recovery facilities,
A, B, C, and D, with respect to the evaluation criteria CO, QO–QI, TP/SU,
TP×DT, and CS, respectively. They also show the normalized eigen vectors
representing the relative ratings with respect to the criteria. Note that each of
the matrices in tables 7.1–7.6 has a consistency ratio CR of less than or equal
to 0.1. For example, for the matrix in table 7.4, CR, using equation (4.1), is
( )
( )( )
.
( ) .
. .
max L
r
n
n R 1
4 15 4
4 1 0 90
0 06
Table 7.7 shows the aggregate matrix of relative ratings of recovery facili
ties with respect to each criterion in the second level of the hierarchy. This
matrix is the collection of the eigen vectors obtained in tables 7.2–7.6.
TABLE 7.1
Comparative matrix for criteria (frst model)
Criteria CO QO–QI TP/SU TP×DT CS
Normalized
eigen vector
CO 1 1/5 3 1 1/5 0.104
QO–QI 5 1 7 3 5 0.491
(TP)/(SU) 1/3 1/7 1 1/2 1/3 0.055
(TP)×(DT) 1 1/3 2 1 1/3 0.110
CS 5 1/5 3 3 1 0.240
© 2009 by Taylor & Francis Group, LLC
Evaluation of Recovery Facilities 129
TABLE 7.2
Comparative ratings of recovery facilities with respect to CO (frst model)
CO
Facilities A B C D
Normalized
eigen vector
A 1 3 6 2 0.460
B 1/3 1 7 3 0.310
C 1/6 1/7 1 1/4 0.050
D 1/2 1/3 4 1 0.180
TABLE 7.3
Comparative ratings of recovery facilities with respect to QO–QI (frst model)
QO–QI
Facilities A B C D
Normalized
eigen vector
A 1 1 7 4 0.380
B 1 1 7 7 0.445
C 1/7 1/7 1 1/5 0.050
D 1/4 1/7 5 1 0.125
TABLE 7.4
Comparative ratings of recovery facilities with respect to TP/SU (frst model)
TP/SU
Facilities A B C D
Normalized
eigen vector
A 1 1/7 1/3 1/2 0.072
B 7 1 2 7 0.574
C 3 1/2 1 1 0.212
D 2 1/7 1 1 0.142
TABLE 7.5
Comparative ratings of recovery facilities with respect to TP×DT (frst model)
TP×DT
Facilities A B C D
Normalized
eigen vector
A 1 1/5 1/2 1/2 0.091
B 5 1 3 7 0.595
C 2 1/3 1 1 0.171
D 2 1/7 1 1 0.143
© 2009 by Taylor & Francis Group, LLC
130 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
By multiplying the matrix in table 7.7 with the normalized eigen vector
obtained in table 7.1, the following normalized ranks are obtained for the
recovery facilities: Rank
A
= 0.26, Rank
B
= 0.42, Rank
C
= 0.09, and Rank
D
=
0.23. If the decision maker desires to choose only those recovery facilities
that have ranks of at least 0.25, he or she would choose recovery facilities A
and B.
7.3 Second Model (Linear Physical Programming)
In this section, the model to identify effcient recovery facilities using linear
physical programming (LPP) is presented. Section 7.3.1 presents the nomen
clature for the LPP model, section 7.3.2 presents the evaluation criteria that are
considered in this model, and section 7.3.3 presents a numerical example.
7.3.1 Nomenclature for LPP Model
CS
v
Customer service rating of recovery facility v (numerical scale)
DT
v
Average disassembly time of products supplied to recovery facil
ity v (time units)
TABLE 7.6
Comparative ratings of recovery facilities with respect to CS (frst model)
CS
Facilities A B C D
Normalized
eigen vector
A 1 1/6 1/3 1/7 0.053
B 6 1 5 1/3 0.298
C 3 1/5 1 1/6 0.101
D 7 3 6 1 0.548
TABLE 7.7
Aggregate of ratings of recovery facilities (frst model)
A B C D
CO 0.460 0.310 0.050 0.180
QO–QI 0.380 0.445 0.050 0.125
TP/SU 0.072 0.574 0.212 0.142
TP×DT 0.091 0.595 0.171 0.143
CS 0.053 0.298 0.101 0.548
© 2009 by Taylor & Francis Group, LLC
Evaluation of Recovery Facilities 131
g
i
ith criterion for evaluation of candidate recovery facilities
IT
uv
Transit time between collection center u and recovery facility v
(time units)
K Transportation cost per unit time ($ per unit time)
L
v
Labor cost at location of recovery facility v ($ per unit time)
M
v
Inventory (space) cost at recovery facility v ($ per unit area)
OT
vw
Transit time between recovery facility v and demand center w
(time units)
QI
v
Average quality of products supplied to recovery facility v
(numerical scale)
QO
v
Average quality of outgoing products from recovery facility v
(numerical scale)
R
v
Reprocessing cost at recovery facility v ($ per unit product)
SU
v
Supply to recovery facility v (products)
TP
v
Throughput of recovery facility v (products)
u Collection center
v Recovery facility
w Demand center
7.3.2 Criteria for Identification of Efficient Recovery Facilities
7.3.2.1 Class 1S Criteria (Smaller is Better)
The cost of transporting goods (used as well as reprocessed), g
1
, through a
recovery facility v across the reverse supply chain is calculated using the fol
lowing equation:
g IT K OT K
uv
u
vw
w
1
£ £
( )( ) ( )( ) (7.1)
The operating cost, g
2
, incurred by a recovery facility v is the sum of the labor
cost, the inventory cost, and the reprocessing cost. Thus,
g L M R
v v v 2
(7.2)
7.3.2.2 Class 2S Criteria (Larger Is Better)
As explained in section 7.2, the difference between QO and QI is taken as a
criterion for evaluation. Thus,
g QO QI
v v 3
(7.3)
© 2009 by Taylor & Francis Group, LLC
132 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
Also, TP is divided by SU and then taken as a criterion for evaluation:
g TP SU
v v 4
/ (7.4)
Similarly,
g TP DT
v v 5
r (7.5)
Customer service (CS) is taken as a criterion for evaluation in the physical
programming model as well. Thus,
g CS
v 6
(7.6)
7.3.3 Numerical Example
Three candidate recovery facilities (E, F, and G) are evaluated using the LPP
model and then ranked to identify the effcient ones.
Table 7.8 shows the target values for each criterion detailed in section 7.3.2.
Table 7.9 shows the criteria values for each recovery facility. Table 7.10 shows
the incremental weights obtained by using the LPP weight algorithm [2].
Tables 7.11–7.13 show the deviations of criteria values from the target val
ues for facilities E, F, and G, respectively. Table 7.14 shows the total scores
(obtained using equation (4.34)) and the ranks of the recovery facilities. It is
obvious from table 7.14 that G is the most desirable facility and F is the least
desirable. If the decision maker has a cutoff limit of, say, 100, he or she will
identify facilities E and G as effcient.
7.4 Evaluation Criteria for Third and Fourth Models
Like in evaluation of collection centers (see chapter 6), in evaluation of recovery
facilities one may have a scenario with three different categories of decision
makers with multiple, conficting, and incommensurate goals, as follows:
TABLE 7.8
Preference table (second model)
Criteria t
p1
+ t
p2
+ t
p3
+ t
p4
+ t
p5
+
g
1
10 15 25 30 45
g
2
12 14 15 16 17
Criteria t
p1
– t
p2
– t
p3
– t
p4
– t
p5
–
g
3
0.6 0.4 0.3 0.2 0
g
4
1.1 0.9 0.7 0.6 0.4
g
5
250 200 140 120 100
g
6
10 7 6 4 3
© 2009 by Taylor & Francis Group, LLC
Evaluation of Recovery Facilities 133
TABLE 7.9
Criteria values for each recovery facility (second model)
Criteria Facility E Facility F Facility G
g
1
22 30 15
g
2
15 17 10
g
3
0.4 0.3 0.1
g
4
0.5 0.8 0.5
g
5
200 220 145
g
6
8 6 4
TABLE 7.10
Output of LPP weight algorithm (second model)
Criteria Δw
p2
+ Δw
p3
+ Δw
p4
+ Δw
p5
+ Δw
p2
– Δw
p3
– Δw
p4
– Δw
p5
–
g
1
0.02 0.012 0.168 0.011 — — — —
g
2
0.05 0.266 0.683 2.157 — — — —
g
3
— — — — 0.05 0.115 0.380 1.252
g
4
— — — — 0.05 0.280 0.215 1.252
g
5
— — — — 0.033 0.077 0.253 0.835
g
6
— — — — 0.033 0.077 0.979 2.505
TABLE 7.11
Deviations of criteria values of recovery facility E from target values (second
model)
Criteria r = 2 r = 3 r = 4 r = 5
g
1
d
12
+ = 12 d
13
+ = 7 d
14
+ = 3 d
15
+ = 8
g
2
d
22
– = 3 d
23
– = 1 d
24
– = 0 d
25
– = 1
g
3
d
32
– = 0.2 d
33
– = 0 d
34
– = 0.1 d
35
– = 0.2
g
4
d
42
– = 0.6 d
43
– = 0.4 d
44
– = 0.2 d
45
– = 0.1
g
5
d
52
– = 50 d
53
– = 0 d
54
– = 60 d
55
– = 80
g
6
d
62
– = 2 d
63
– = 1 d
64
– = 2 d
65
– = 4
TABLE 7.12
Deviations of criteria values of recovery facility F from target values (second
model)
Criteria r = 2 r = 3 r = 4 r = 5
g
1
d
12
+ = 20 d
13
+ = 15 d
14
+ = 5 d
15
+ = 0
g
2
d
22
– = 0.3 d
23
– = 0.1 d
24
– = 0 d
25
– = 0.1
g
3
d
32
– = 0.3 d
33
– = 0.1 d
34
– = 0.1 d
35
– = 0.2
g
4
d
42
– = 30 d
43
– = 20 d
44
– = 80 d
45
– = 100
g
5
d
52
– = 4 d
53
– = 1 d
54
– = 0 d
55
– = 2
© 2009 by Taylor & Francis Group, LLC
134 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
Consumers r whose primary concern is convenience
Local government offcials r whose primary concern is environmental
consciousness
Supply chain company executives r whose primary concern is proft
Therefore, the effciency of a candidate recovery facility must be evaluated
based on the maximized consensus among decision makers of the three cat
egories. Sections 7.4.1, 7.4.2, and 7.4.3 present the lists of criteria that are con
sidered for the three categories.
7.4.1 Criteria of Consumers
Proximity to surface water ( r PS) (the closer the facility is to the sur
face water, the higher the facility’s suitability, i.e., less hazardous)
Proximity to residential area ( r PH) (the closer the facility is to the resi
dential area, the higher the facility’s suitability, i.e., less hazardous)
Employment opportunity ( r EO) (the higher, the better)
Salary offered to employees at recovery facility ( r SA) (the higher,
the better)
7.4.2 Criteria of Local Government Officials
Proximity to surface water ( r PS) (the closer the facility is to surface
water, the higher the facility’s suitability, i.e., less hazardous)
TABLE 7.13
Deviations of criteria values of recovery facility G from target values (second
model)
Criteria r = 2 r = 3 r = 4 r = 5
g
1
d
12
+ = 5 d
13
+ = 0 d
14
+ = 10 d
15
+ = 15
g
2
d
22
– = 0.5 d
23
– = 0.3 d
24
– = 0.2 d
25
– = 0.1
g
3
d
32
– = 0.6 d
33
– = 0.4 d
34
– = 0.2 d
35
– = 0.1
g
4
d
42
– = 105 d
43
– = 55 d
44
– = 5 d
45
– = 25
g
5
d
52
– = 6 d
53
– = 3 d
54
– = 2 d
55
– = 0
TABLE 7.14
Total scores and ranks of recovery facilities (second model)
Recovery facilities Total scores Ranks
E 99.85 II
F 117.95 III
G 52.26 I
© 2009 by Taylor & Francis Group, LLC
Evaluation of Recovery Facilities 135
Proximity to residential area ( r PH) (the closer the facility is to the resi
dential area, the higher the facility’s suitability, i.e., less hazardous)
7.4.3 Criteria of Supply Chain Company Executives
Space cost ( r SC) (the lower, the better)
Labor cost ( r LC) (the lower, the better)
Proximity to roads ( r PR) (the closer the facility is to roads, the easier
is the transportation)
Quality of reprocessed products ( r QO)–quality of used products (QI)
(the higher, the better)
Throughput ( r TP)/supply (SU) (the higher, the better)
Throughput ( r TP)×disassembly time (DT) (the higher, the better)
Utilization of incentives from local government ( r UI) (the higher,
the better)
Pollution control ( r PC) (the higher, the better)
7.5 Third Model (Eigen Vector Method,
TOPSIS, and Borda’s Choice Rule)
This model to select effcient recovery facilities is implemented in two
phases. In the frst phase, using the eigen vector method, impacts are given
to the criteria identifed for each category of decision makers (see section 7.4),
and then TOPSIS (technique for order preference by similarity to ideal solu
tion) is employed to fnd the effciency of each candidate recovery facility, as
evaluated by that category. In the second phase, Borda’s choice rule is used
to combine individual evaluations for each candidate recovery facility into a
group evaluation or maximized consensus ranking.
The model to select effcient recovery facilities is presented using a numerical
example. Three recovery facilities, H, J, and K, are considered for evaluation.
7.5.1 Phase I (Individual Decision Making)
Tables 7.15–7.17 show the pairwise comparison matrices as formed for the
consumers, local government offcials, and supply chain company execu
tives, respectively.
Tables 7.18–7.20 show the impacts given for the criteria of the consumers,
local government offcials, and supply chain company executives, respec
tively. These sets of impacts, calculated using the eigen vector method, are
the elements of the normalized eigen vectors of pairwise comparison matri
ces shown in tables 7.15–7.17, respectively. For example, the relative impacts
© 2009 by Taylor & Francis Group, LLC
136 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
of the criteria PH (proximity to residential area) and PR (proximity to roads)
shown in table 7.19 are the elements of the normalized eigen vector of the
pairwise comparison matrix shown in table 7.16.
The decision matrices formed for the consumers, local government offcials,
and supply chain company executives, for implementation of the TOPSIS for
each category, are shown in tables 7.21–7.23, respectively. The elements of
these matrices are the ranks (ranging from 1 to 10) assigned to the recovery
facilities with respect to each criterion for evaluation. A lower rank implies
higher effciency (with respect to that criterion).
To facilitate the construction of the pairwise comparison matrices (see
tables 7.15–7.17) and the decision matrices (see tables 7.21–7.23), representa
tives from each category of decision makers could be invited to participate
in relevant survey questionnaires, individual interviews, focus groups, and
onsite observations.
TABLE 7.15
Pairwise comparison matrix for consumers (third model)
Criteria PS PH EO SA
PS 1 1/2 1/2 1/3
PH 2 1 1 1
EO 2 1 1 1
SA 3 1 1 1
TABLE 7.16
Pairwise comparison matrix for local
government offcials (third model)
Criteria PS PH
PS 1 1/3
PH 3 1
TABLE 7.17
Pairwise comparison matrix for supply chain company executives (third model)
Criteria SC LC PR QO–QI TP/SU TP×DT UI PC
SC 1 1 1 1 ½ 6 1 1
LC 1 1 1 1 1 5 2 1
PR 1 1 1 1/9 1/7 1 1 1/3
QO–QI 1 1 1 1 2 2 1 1/4
TP/SU 2 1 7 1 1 5 1 1
TP×DT 1/6 1/5 1 1/2 0.2 1 1/9 1
UI 1 1/2 1 1 1 9 1 1
PC 1 1 3 4 1 1 1 1
© 2009 by Taylor & Francis Group, LLC
Evaluation of Recovery Facilities 137
Now one is ready to perform the six steps in the TOPSIS for each category
of decision makers. The following steps show the implementation of the
TOPSIS for the consumers to evaluate H, J, and K.
Step 1: Construct the normalized decision matrix. Table 7.24 shows the nor
malized decision matrix formed by applying equation (4.49) on each
element of table 7.21 (decision matrix formed for the consumers). For
example, the normalized rank of recovery facility J with respect to
criterion PH (see tables 7.21 and 7.24) is calculated as follows:
r
22
2 2 2
1
2 1 3
0 2673
. .
TABLE 7.18
Impacts for consumers (third model)
Criteria Impacts
PS 0.1277
PH 0.2804
EO 0.2804
SA 0.3116
TABLE 7.19
Impacts for local government offcials
(third model)
Criteria Impacts
PS 0.25
PH 0.75
TABLE 7.20
Impacts for supply chain company
executives (third model)
Criteria Impacts
SC 0.1233
LC 0.1437
PR 0.0717
QO–QI 0.1198
TP/SU 0.1905
TP×DT 0.0491
UI 0.1356
PC 0.1663
© 2009 by Taylor & Francis Group, LLC
138 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
Step 2: Construct the weighted normalized decision matrix. Table 7.25 shows
the weighted normalized decision matrix for the consumers. This
is constructed using the impacts of the criteria listed in table 7.18
and the normalized decision matrix in table 7.24. For example, the
weighted normalized rank of recovery facility J with respect to cri
terion PH, i.e., 0.0749 (see table 7.25), is calculated by multiplying the
impact of PH, i.e., 0.2804 (see table 7.18), by the normalized rank of R2
with respect to PH, i.e., 0.2673 (see table 7.24).
TABLE 7.21
Decision matrix for consumers (third model)
Recovery
facilities PS PH EO SA
H 2 2 4 1
J 4 1 7 3
K 5 3 3 1
TABLE 7.22
Decision matrix for local government offcials (third
model)
Recovery facilities PS PH
H 4 2
J 3 1
K 1 5
TABLE 7.23
Decision matrix for supply chain company executives (third model)
Recovery
facilities SC LC PR QO–QI TP/SU TP×DT UI PC
H 3 2 1 3 4 1 1 2
J 4 7 1 2 2 2 3 4
K 9 8 7 5 1 6 5 6
TABLE 7.24
Normalized decision matrix for consumers (third model)
Recovery
facilities PS PH EO SA
H 0.2981 0.5345 0.4650 0.3015
J 0.5963 0.2673 0.8137 0.9045
K 0.7454 0.8018 0.3487 0.3015
© 2009 by Taylor & Francis Group, LLC
Evaluation of Recovery Facilities 139
Step 3: Determine the ideal and negativeideal solutions. Each column in
the weighted normalized decision matrix shown in table 7.25 has a
minimum rank and a maximum rank. They are the ideal and nega
tiveideal solutions, respectively, for the corresponding criterion.
For example (see table 7.25), with respect to criterion PH, the ideal
solution (minimum rank) is 0.0749, and the negativeideal solution
(maximum rank) is 0.2248.
Step 4: Calculate the separation distances. The separation distances (see
table 7.26) for each recovery facility are calculated using equations
(4.52) and (4.53). For example, the positive separation distance for
recovery facility J (see table 7.26) is calculated using equation (4.52),
which contains the weighted normalized ranks of J (see table 7.25)
and the ideal solutions (obtained in step 3) for the criteria.
Step 5: Calculate the relative closeness coeffcient. Using equation (4.54), the
relative closeness coeffcient is calculated for each recovery facility
(see table 7.27). For example, the relative closeness coeffcient (i.e.,
0.3945) for recovery facility J (see table 7.27) is the ratio of J’s negative
separation distance (i.e., 0.1511) to the sum (i.e., 0.1511 + 0.2318 = 0.3829)
of its negative and positive separation distances (see table 7.26).
TABLE 7.25
Weighted normalized decision matrix for consumers (third model)
Recovery
facilities PS PH EO SA
H 0.0381 0.1499 0.1304 0.0940
J 0.0761 0.0749 0.2282 0.2819
K 0.0952 0.2248 0.0978 0.0940
TABLE 7.26
Separation distances for consumers (third model)
Recovery facilities S* S–
H 0.0817 0.2318
J 0.2318 0.1511
K 0.1604 0.2287
TABLE 7.27
Relative closeness coeffcients for consumers (third
model)
Recovery facilities C*
H 0.7394
J 0.3945
K 0.5878
© 2009 by Taylor & Francis Group, LLC
140 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
Step 6: Form the preference order. Because the best alternative is the one
with the highest relative closeness coeffcient, the preference order
for the recovery facilities is H, K, and J (that means H is the best
recovery facility, as evaluated by the consumers).
TOPSIS is implemented for the local government offcials and supply chain
company executives in a similar manner. The relative closeness coeffcients
of the recovery facilities, as calculated for those two categories of decision
makers, are shown in table 7.28.
7.5.2 Phase II (Group Decision Making)
Table 7.29 shows the marks of the recovery facilities as given using Borda’s
choice rule for the consumers, local government offcials, and supply chain
company executives. Borda scores (group evaluations) calculated for H, J,
and K (viz., 5, 4, and 2, respectively) are also shown. For example, the Borda
score for J (i.e., 4) is calculated by summing the marks of J for the consumers,
local government offcials, and supply chain company executives (i.e., 0 + 2 +
2). Because H has the highest Borda score, it is the best of the lot.
7.6 Fourth Model (Neural Networks, Fuzzy
Logic, TOPSIS, Borda’s Choice Rule)
This model assumes that neither numerical nor linguistic impacts are avail
able for the evaluation criteria. It employs a neural network [6] to evaluate
TABLE 7.28
Relative closeness coeffcients for local government offcials and
supply chain company executives (third model)
Recovery facilities
Local government
offcials
Supply chain
company executives
H 0.6715 0.5952
J 0.8487 0.5962
K 0.2941 0.4057
TABLE 7.29
Marks and Borda scores of recovery facilities (third model)
Recovery
facilities Consumers
Local
government
offcials
Supply chain
company
executives Borda scores
H 2 1 2 5
J 0 2 2 4
K 1 0 1 2
© 2009 by Taylor & Francis Group, LLC
Evaluation of Recovery Facilities 141
the effciency of a recovery facility of interest (which is being considered for
inclusion in a reverse supply chain), using linguistic performance measures
of recovery facilities that already exist in the reverse supply chain. To this
end, the model to evaluate the effciency of a recovery facility of interest is
carried out in three phases. In the frst phase, the ratings of existing recov
ery facilities are used to construct a neural network that, in turn, calculates
impacts of criteria identifed for each category of decision makers given in
section 7.4. Then, in the second phase, the impacts obtained in the frst phase
are used in a fuzzy TOPSIS (combination of fuzzy logic and TOPSIS) method
to obtain the overall rating of the recovery facility of interest, as calculated
for each category. Finally, in the third phase, Borda’s choice rule is employed
to calculate the maximized consensus rating (among the categories consid
ered), i.e., effciency, of the recovery facility of interest.
7.6.1 Phase I (Derivation of Impacts)
Suppose that one has the linguistic ratings of ten existing recovery facilities,
as given by an expert in each category of decision makers described in section
7.4. Using fuzzy logic, these linguistic ratings are converted into triangular
fuzzy numbers (TFNs). Table 7.30 shows not only one of the many ways for
conversion of linguistic ratings into TFNs but also the defuzzifed ratings of
the corresponding TFNs. Tables 7.31–7.33 show the defuzzifed overall rating
of each existing recovery facility as well as the recovery facility’s defuzzifed
rating with respect to each criterion, as evaluated by the consumers, local
government offcials, and supply chain company executives, respectively.
Defuzzifcation of a TFN can be performed using equation (4.8).
A neural network is constructed and trained for each category of decision
makers, using the defuzzifed ratings of the existing recovery facilities with
respect to criteria as input sets and the recovery facilities’ defuzzifed over
all ratings as corresponding outputs. In the example, there are ten input–
output pairs for each neural network because there are ten existing recovery
facilities. Also, three layers are considered in each network, with fve nodes
in the hidden layer. The number of nodes in the output layer is one (for
overall rating), and the number in the input layer is the number of criteria
TABLE 7.30
Conversion table for ratings (fourth model)
Linguistic
ratings TFNs
Defuzzifed
ratings
Very good (VG) (7, 10, 10) 9
Good (G) (5, 7, 10) 7.3
Fair (F) (2, 5, 8) 5
Poor (P) (1, 3, 5) 3
Very poor (VP) (0, 0, 3) 1
© 2009 by Taylor & Francis Group, LLC
142 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
TABLE 7.31
Consumer ratings of recovery facilities (fourth model)
Recovery
facilities PS PH EO SA Overall
R1 9 3 7.33 3 3
R2 5 7.33 9 1 1
R3 1 5 1 5 9
R4 1 5 5 9 5
R5 5 9 5 3 7.33
R6 1 3 5 3 5
R7 9 1 3 5 1
R8 3 1 3 1 9
R9 3 9 1 7.33 5
R10 5 1 3 5 7.33
TABLE 7.32
Local government offcials’ ratings of recovery facilities (fourth
model)
Recovery
facilities PS PH Overall
R1 9 3 5
R2 5 7.33 3
R3 1 5 9
R4 1 5 7.33
R5 5 9 1
R6 1 3 1
R7 9 1 7.33
R8 3 1 5
R9 3 9 5
R10 5 1 3
TABLE 7.33
Supply chain company executives’ ratings of recovery facilities (fourth model)
Recovery
facilities SC LC PR QO–QI TP/SU TP×DT UI PC Overall
R1 9 3 9 5 3 9 3 3 5
R2 5 7.33 3 1 7.33 9 1 1 3
R3 1 5 1 3 1 3 5 9 1
R4 1 5 3 5 9 9 1 5 7.33
R5 5 9 1 3 5 7.33 7.33 1 9
R6 9 3 9 5 3 9 3 3 9
R7 5 7.33 3 1 7.33 9 1 1 1
R8 1 5 1 3 1 3 5 9 3
R9 1 5 3 5 9 9 1 5 5
R10 5 9 1 3 5 7.33 7.33 1 7.33
© 2009 by Taylor & Francis Group, LLC
Evaluation of Recovery Facilities 143
considered by the corresponding category. For example, fgure 7.2 shows the
neural network constructed and trained for the consumer category.
After each neural network is trained, the following equation [1] is used to
calculate the impacts of criteria considered by the corresponding category.
Here, the absolute value of W
v
is the impact of the vth input node upon the
output node, n
V
is the number of input nodes, n
H
is the number of hidden
nodes, I
ij
is the connection weight from the ith input node to the jth hidden
node, and O
j
is the connection weight from the jth hidden node to the output
node:
 
 
 
W
I
I
O
I
I
O
v
vj
ij
i
n
j
j
n
vj
ij
i
n
j
V
H
V
¤
¦
¥
¥
¥
¥
£
£
£
¥¥
¥
¥
¥
¥
¥
¥
¥
¥
´
¶
µ
µ
µ
µ
µ
µ
µ
µ
µ
µ
µ
µ
µ
µ
µ
¤
¦
¥
¥
¥
¥
¥
¥
¥
¥
¥
¥
£
j
n
H
¥¥
¥
¥
´
¶
µ
µ
µ
µ
µ
µ
µ
µ
µ
µ
µ
µ
µ
µ
µ
µ
£
i
n
V
(7.7)
Tables 7.34–7.36 show the impacts of the criteria considered for the consumers,
local government offcials, and supply chain company executives, respectively.
PS
PH
EO
SA
H
H
H
H
H
O
Input layer Hidden layer Output layer
FIGURE 7.2
Neural network for consumers (fourth model).
© 2009 by Taylor & Francis Group, LLC
144 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
TABLE 7.34
Impacts of criteria of consumers (fourth model)
Criteria PS PH EO SA
Impacts 0.17 0.18 0.27 0.38
TABLE 7.35
Impacts of criteria of local government offcials
(fourth model)
Criteria PS PH
Impacts 0.61 0.39
TABLE 7.36
Impacts of criteria of supply chain company executives (fourth model)
Criteria SC LC PR QO–QI TP/SU TP×DT UI PC
Impacts 0.06 0.02 0.10 0.19 0.11 0.39 0.10 0.02
TABLE 7.37
Decision matrix for consumers (fourth model)
Recovery
facilities PS PH EO SA
R11 1 5 1 5
R12 1 5 5 9
R13 5 9 5 3
TABLE 7.38
Decision matrix for local government offcials
(fourth model)
Recovery
facilities PS PH
R11 3 1
R12 3 9
R13 5 1
TABLE 7.39
Decision matrix for supply chain company executives (fourth model)
Recovery
facilities SC LC PR QO–QI TP/SU TP×DT UI PC
R11 5 9 1 3 5 7.33 7.33 1
R12 9 3 9 5 3 9 3 3
R13 5 7.33 3 1 7.33 9 1 1
© 2009 by Taylor & Francis Group, LLC
Evaluation of Recovery Facilities 145
7.6.2 Phase II (Individual Decision Making)
Suppose that there are three recovery facilities, R11, R12, and R13, of inter
est. A fuzzy TOPSIS method uses the impacts obtained in the frst phase to
calculate the overall ratings of the three recovery facilities.
The decision matrices formed for the consumers, local government off
cials, and supply chain company executives (with defuzzifed ratings for
R11, R12, and R13) in this example are shown in tables 7.37–7.39, respectively
(table 7.30 is used here as well, to convert linguistic ratings given by each cat
egory into TFNs). Like in the third model (see section 7.5), the matrices can be
constructed by inviting representatives from each category of decision mak
ers to participate in relevant survey questionnaires, individual interviews,
focus groups, and onsite observations.
Now one is ready to perform the six steps in the TOPSIS for each category
of decision makers. The following steps show the implementation of the
TOPSIS for the consumers to evaluate R11, R12, and R13.
Step 1: Construct the normalized decision matrix. Table 7.40 shows the
normalized decision matrix formed by applying equation (4.49) on
each element of table 7.37 (decision matrix for the consumers). For
example, the normalized rating of recovery facility R12 with respect
to criterion PH (see tables 7.37 and 7.40) is calculated as follows:
r
22
2 2 2
5
5 5 9
0 4369
. .
Step 2: Construct the weighted normalized decision matrix. Table 7.41 shows
the weighted normalized decision matrix for the consumers. This
is constructed using the impacts of the criteria listed in table 7.34
and the normalized decision matrix in table 7.40. For example, the
weighted normalized rank of recovery facility R12 with respect to
criterion PH, i.e., 0.0793 (see table 7.41), is calculated by multiplying
the impact of PH, i.e., 0.18 (see table 7.34), with the normalized rating
of R12 with respect to PH, i.e., 0.4369 (see table 7.40).
Step 3: Determine the ideal and negativeideal solutions. Each column in
the weighted normalized decision matrix shown in table 7.41 has a
maximum rating and a minimum rating. They are the ideal and neg
ativeideal solutions, respectively, for the corresponding criterion.
For example (see table 7.41), with respect to criterion PH, the ideal
solution (maximum rating) is 0.1428, and the negativeideal solution
(minimum rating) is 0.0793.
Step 4: Calculate the separation distances. The separation distances (see
table 7.42) for each recovery facility are calculated using equations
(4.52) and (4.53). For example, the positive separation distance for
© 2009 by Taylor & Francis Group, LLC
146 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
recovery facility R12 (see table 7.42) is calculated using equation
(4.52), which contains the weighted normalized ratings of R12 (see
table 7.41) and the ideal solutions (obtained in step 3) for the criteria.
Step 5: Calculate the relative closeness coeffcient. Using equation (4.54), the
relative closeness coeffcient is calculated for each recovery facility
(see table 7.43). For example, the relative closeness coeffcient (i.e.,
0.6398) for recovery facility R12 (see table 7.43) is the ratio of R12’s
negative separation distance (i.e., 0.2602) to the sum (i.e., 0.2602 +
0.1464 = 0.4066) of its negative and positive separation distances (see
table 7.42).
Step 6: Rank the preference order. Because the best alternative is the one
with the highest relative closeness coeffcient, the preference order
for the recovery facilities is R12, R13, and R11 (that means R12 is the
best recovery facility, as evaluated by the consumers).
Similarly, TOPSIS is implemented for the local government offcials and
supply chain company executives. The relative closeness coeffcients of
the recovery facilities, as calculated for those two categories, are shown in
table 7.44.
TABLE 7.42
Separation distances for consumers (fourth model)
Recovery
facilities S* S–
R11 0.2526 0.0712
R12 0.1464 0.2602
R13 0.2136 0.2086
TABLE 7.40
Normalized decision matrix for consumers (fourth model)
Recovery
facilities PS PH EO SA
R11 0.1923 0.4369 0.1400 0.4663
R12 0.1925 0.4369 0.7001 0.8393
R13 0.9623 0.7863 0.7001 0.2798
TABLE 7.41
Weighted normalized decision matrix for consumers (fourth model)
Recovery
facilities PS PH EO SA
R11 0.0330 0.0793 0.0371 0.1780
R12 0.0330 0.0793 0.1857 0.3203
R13 0.1650 0.1428 0.1857 0.1068
© 2009 by Taylor & Francis Group, LLC
Evaluation of Recovery Facilities 147
7.6.3 Phase III (Group Decision Making)
Table 7.45 shows the marks of the recovery facilities as given using Borda’s
choice rule for the consumers, local government offcials, and supply chain
company executives. Borda scores (group evaluations) calculated for R11,
R12, and R13 (viz., 1, 6, and 2, respectively) are also shown. For example, the
Borda score for R12 (i.e., 6) is calculated by summing the marks of R12 for
the consumers, local government offcials, and supply chain company execu
tives (i.e., 2 + 2 + 2). Because R12 has the highest Borda score, it is the best of
the lot.
TABLE 7.43
Relative closeness coeffcients for consumers
(fourth model)
Recovery facilities C*
R11 0.2199
R12 0.6398
R13 0.4942
TABLE 7.44
Relative closeness coeffcients for local
government offcials and supply chain company
executives (fourth model)
Recovery
facilities
Local
government
offcials
Supply chain
company
executives
R11 0 0.458016
R12 0.646968 0.695261
R13 0.588066 0.29753
TABLE 7.45
Marks and Borda scores of recovery facilities (fourth model)
Recovery
facilities Consumers
Local
government
offcials
Supply chain
company
executives Borda scores
R11 0 0 1 1
R12 2 2 2 6
R13 1 1 0 2
© 2009 by Taylor & Francis Group, LLC
148 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
7.7 Fifth Model (TwoDimensional Chart)
Here, a simple twodimensional chart, where only the two most impor
tant evaluation criteria are considered, is presented. In this approach, it is
assumed that the two most important criteria are (1) n value in n Sigma of the
recovery facility and (2) TP/SU value.
This approach is similar (but with some important modifcations) to Linn
et al.’s approach [5] for selection of suppliers in a traditional supply chain.
Linn et al.’s approach uses C
pk
(process capability index) and a price com
integrates the process capability and price information of multiple suppliers
(assuming every other criterion value is either unimportant or the same for
all suppliers), provides a method to consider quality and price simultane
ously in the supplier selection process.
C
pk
is drawn on the Yaxis, and the ratio (R) of target price desired to price
quoted by the supplier is drawn on the Xaxis. C
pk
and R of each supplier are
then plotted on the chart. The chart is partitioned into six different zones
representing the quality performance and price levels. The zones are defned
as follows:
C
pk
3.0
2.5
2.0
1.5
1.0
0.5
0.0
H Zone
E Zone
E G S C
G
S
C
U Zone
0.0
0.5 1.0 1.5 2.0
2.5
3.0 3.5
R
FIGURE 7.3
CPC chart
for the concept of process capability). The CPC chart (see fgure 7.3), which
parison (CPC) chart for the selection of effcient suppliers (see section 4.16
© 2009 by Taylor & Francis Group, LLC
Evaluation of Recovery Facilities 149
E zone: Excellent zone, E = {(R, C
pk
)  C
pk
> 2.0 and R > 2.0}
G zone: Good zone, G = {(R, C
pk
)  C
pk
> 1.5 and R > 1.5} – E zone
S zone: Satisfactory zone, S = {(R, C
pk
)  C
pk
> 1.33 and R > 1.33} – G zone
– E zone
C zone: Capable zone, C = {(R, C
pk
)  C
pk
> 1.0 and R > 1.0} – G zone – E
zone – S zone
H zone: Highprice zone, H = {(R, C
pk
)  C
pk
> 1.0 and R < 1.0}
U zone: Unacceptable zone, U = {(R, C
pk
)  C
pk
< 1.0 and R > 0}
Because of their highquality performance and lowcost quotation, those
suppliers falling in the E zone are considered the best group of suppliers to
choose from. Those falling in the U zone are simply not acceptable because
their C
pk
value is too low. The supplier selection should start from the E zone
and follow the sequence of E nG n S n Cn H n U. Within the same zone,
those falling in the lowerhalf zone (below the 45° line) have better cost per
formance. In contrast, those in the upperhalf zone (above the 45° line) have
better quality performance. Therefore, if the objective is to select a better
cost performer, those in the lower half of the zone should be selected. If the
objective is to fnd a betterquality performer, those in the upperhalf zone
should be selected. If quality and cost are equally important, those close to
the 45° line should be selected.
Although the above approach can integrate both quality and price, it must
be noted that the C
pk
value is not enough to judge quality; one needs the
C
p
(process capability ratio) value as well. To overcome this problem in the
selection of effcient recovery facilities, the n value (instead of C
pk
) is used on
the Yaxis of the chart (see section 4.16 for the concepts of process capability
and n Sigma).
The n value for a recovery facility represents the facility’s quality, which
could be a function of factors such as effciency of reprocessing and eff
ciency of delivery (of reprocessed goods to demand centers). For example
(see fgure 7.4), let the upper specifcation (U) of the width (critical dimen
sion) of a reprocessed good at a candidate recovery facility be 100 inches,
the lower specifcation (L) 60 inches, and the target value ( T) 80 inches. If the
standard deviation (T) of the width is 5 inches and the mean shift ( T M ) is
6 inches, then reprocessing in the candidate recovery facility is a Four Sigma
process.
chart that is used for selection of effcient recovery facilities.
The chart is partitioned into fve acceptable regions (I–V) and an unaccept
able region. These zones are defned as follows:
Zone I = {(TP/SU, n)  0.9 ≤ TP/SU ≤ 1 and n ≥ 5.0}
Zone II = {(TP/SU, n)  TP/SU ≥ 0.8 and n ≥ 4.0} – zone I
Zone III = {(TP/SU, n)  TP/SU ≥ 0.7 and n ≥ 3.0} – zone I – zone II
For the chart’s Xaxis, the TP/SU criterion is used. See fgure 7.5 for the
© 2009 by Taylor & Francis Group, LLC
150 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
Zone IV = {(TP/SU, n)  TP/SU ≥ 0.6 and n ≥ 2.0} – zone I – zone II
– zone III
Zone V = {(TP/SU, n)  TP/SU ≥ 0.5 and n ≥ 2.0} – zone I – zone II – zone
III – zone IV
Unacceptable zone = {(TP/SU, n)  TP/SU ≤ 0.5 or n < 2.0}
100
60 80
6 inches
= 5 inches
Inches
FIGURE 7.4
Critical dimension of a reprocessed good at recovery facility (ffth model).
n
6.0
4.0
3.0
2.0
1.0
0.0
5.0
I
II
III
IV
V
Unacceptable Region
Unacceptable Region
0.0 0.1 0.2 0.3 0.4
TP/SU
0.5 0.6 0.7 0.8 0.9 1.0
FIGURE 7.5
n and TP/SU chart (ffth model).
© 2009 by Taylor & Francis Group, LLC
Evaluation of Recovery Facilities 151
Because of their highquality (Five Sigma or higher) performance and high
TP/SU value, those recovery facilities falling in zone I are considered the best
group of recovery facilities to choose from. Those falling in the unacceptable
zone are simply not acceptable because their n value (2 or lower) is too low,
their TP/SU value is low, or both. The recovery facility selection should start
from zone I and follow the sequence of I n II n III n IV n V.
If the decision maker wishes to raise or lower the bar for quality, he or
she can revise the respective charts accordingly. For example, if the decision
maker considers only those recovery facilities with at least Six Sigma quality
(besides high TP/SU ratio) the best, then zone I in his or her chart will be {(n,
TP/SU)  n ≥ 6.0 and 0.9 ≤ TP/SU ≤ 1}.
7.8 Conclusions
In this chapter, fve models are presented for identifying effcient recovery
facilities in a region where a reverse supply chain is to be designed. The
frst model uses analytic hierarchy process in a situation where the evalu
ation criteria can be given numerical impacts. The second model employs
linear physical programming in a situation where the evaluation criteria are
presented in terms of ranges of different degrees of desirability. The third
and fourth models have decision makers with conficting evaluation criteria.
Whereas the third model uses the eigen vector method, TOPSIS, and Borda’s
choice rule, the fourth model uses neural network, fuzzy logic, TOPSIS, and
Borda’s choice rule. Finally, the ffth model presents a twodimensional chart
that can be used when a very simple evaluation technique is desired (only
the two most important evaluation criteria are considered).
References
1. Cha, Y., and Jung, M. 2003. Satisfaction assessment of multiobjective schedules
using neural fuzzy methodology. International Journal of Production Research
41:1831–49.
2. Messac, A., Gupta, S. M., and Akbulut, B. 1996. Linear physical programming:
A new approach to multiple objective optimization. Transactions on Operational
Research 8:39–59.
3. Talluri, S., and Baker, R. C. 2002. A multiphase mathematical programming
approach for effective supply chain design. European Journal of Operational
Research 141:544–58.
4. Talluri, S., Baker, R. C., and Sarkis, J. 1999. Framework for designing effcient
value chain networks. International Journal of Production Economics 62:133–44.
© 2009 by Taylor & Francis Group, LLC
152 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
5. Linn, R. J., Tsung, F., and Ellis, L. W. C. 2006. Supplier selection based on pro
cess capability and price analysis. Quality Engineering 18:123–29.
6. Haykin, S. 1998. Neural networks: A comprehensive foundation. 2nd ed. Englewood
Cliffs, NJ: Prentice Hall.
© 2009 by Taylor & Francis Group, LLC
153
8
Optimization of Transportation of Products
8.1 The Issue
In this chapter, the focus is on achieving transportation of the right quanti
ties of products (used, remanufactured, and new) across a reverse or closed
loop supply chain while satisfying certain constraints. The various scenarios
for this problem could differ as follows:
1. Decisionmaking criteria for a reverse supply chain could be given
in terms of classical supplyanddemand constraints.
2. Decisionmaking criteria for a reverse supply chain could be pre
sented in terms of ranges of different degrees of desirability.
3. Besides optimal transportation of products, there could be a need to
address the following issues in one continuous phase for a closed
loop supply chain: selection of used products and evaluation of
production facilities. Also, the decisionmaking criteria could be
presented in terms of classic supplyanddemand constraints.
4. The situation could be the same as in scenario 3, but the decision
making criteria could be presented in terms of different degrees of
desirability.
5. The situation could be the same as in scenario 3, but the decision
making criteria could be imprecise.
In this chapter, various models to address the above scenarios are pre
sented. The frst model addresses scenario 1 and employs linear integer
programming. The second model is for scenario 2 and employs linear physi
cal programming. The third and fourth models are for scenarios 3 and 4,
respectively, and use goal programming and linear physical programming,
respectively. The ffth model is for scenario 5 and employs fuzzy goal pro
gramming. For the last three models, it is assumed that the manufacturer
has incorporated a remanufacturing process into his or her original produc
tion system, so that products can be manufactured directly from raw materi
als or remanufactured from used products (the fnal demand for the product
is met with either new or remanufactured products).
© 2009 by Taylor & Francis Group, LLC
154 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
This chapter is organized as follows: Section 8.2 presents the frst model.
Section 8.3 shows the second model. Section 8.4 presents the third model.
Section 8.5 gives the fourth model. Section 8.6 presents the ffth model.
Finally, section 8.7 gives some conclusions.
8.2 First Model (Linear Integer Programming)
This model considers a generic reverse supply chain consisting of collection
centers, remanufacturing facilities, and demand centers. Optimal transpor
tation of products (used and remanufactured) is achieved using linear inte
ger programming. See [1] for a review of a number of models similar to the
one presented in this section. Nomenclature for the model is given in section
8.2.1, the model is formulated in section 8.2.2, and a numerical example is
given in section 8.2.3.
8.2.1 Nomenclature
a
1
Space occupied by one unit of remanufactured product
a
2
Space occupied by one unit of used product
CAP
v
Capacity of remanufacturing facility v to remanufacture
products
C
u
Cost per product retrieved at collection center u
d
w
Demand of remanufactured products at demand center w
I
uv
Decision variable representing the number of products to be
transported from collection center u to remanufacturing facility
v
O
vw
Decision variable representing the number of products to be
transported from remanufacturing facility v to demand center
w
R
v
Cost of remanufacturing per product at production facility v
S
1v
Storage capacity of remanufacturing facility v for remanufac
tured products
S
2v
Storage capacity of remanufacturing facility v for used products
S
u
Storage capacity of collection center u for used products
SUP
u
Supply at collection center u
TI
uv
Cost of transporting one product from collection center u to
remanufacturing facility v
TO
vw
Cost of transporting one product from remanufacturing facility
v to demand facility w
© 2009 by Taylor & Francis Group, LLC
Optimization of Transportation of Products 155
u Collection center
v Remanufacturing facility
w Demand center
Y
v
Binary variable (0/1) for selection of recovery facility v
8.2.2 Model Formulation
The following is the singleperiod and singleproduct transshipment model
formulation that is implemented to achieve minimum overall cost, i.e., sum
of the costs of retrieval, inventory, remanufacturing, and transportation of
products (used and remanufactured) across the supply chain (in the formu
lation, it is assumed that the inventory cost of a used product is 25% of its
retrieval cost, C
u
, and that of a remanufactured product is 25% of its remanu
facturing cost, R
v
):
Minimize
Retrieval costs
Transportation costs
Remanufacturing costs
Inventory costs
C I
TI I TO O
R O
u uv
uv uv vw vw
v vw
v u
v u w v
w v
` `
` ` ` `
` `
÷
÷ ÷
÷÷
÷
` ` ` `
( / ). ( / ). C I R Q
u uv v vw
v u w v
4 4
(8.1)
subject to
Demand at each demand center must be met
O d w
vw w
v
`
= \ ;
(8.2)
Total output of each remanufacturing facility is at most its total input
© 2009 by Taylor & Francis Group, LLC
156 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
I O v
uv vw
u w
` `
_ \ ; (8.3)
Total space occupied by remanufactured products at each remanufactur
ing facility is at most its capacity for remanufactured products
Total space occupied by used products at each collection center is at most
its capacity
a O S Y v
vw v v
w
1
1 . . ;
`
_ \ (8.4)
a I S u
uv u
v
2
. ;
`
_ \ (8.5)
Total space occupied by used products at each remanufacturing facility is
at most its capacity for used products
a I S Y v
uv v v
u
2 2
. . ;
`
_ \ (8.6)
Quantities of transported products are nonnegative numbers
I u v
uv
_ \ 0; , (8.7)
Total output of each remanufacturing facility is at most its capacity to
remanufacture
O v w
vw
_ \ 0; , (8.8)
Total quantity of used products supplied to remanufacturing facilities by
each collection center is at most the supply to that collection center
O CAP v
vw v
w
`
_ \ ; (8.9)
I SUP u
uv u
v
`
_ \ ; (8.10)
Note that a
1
and a
2
are possibly different. The reason is that a component
from a used product of one model may be used in a remanufactured product
of a different model that may occupy a different amount of space.
© 2009 by Taylor & Francis Group, LLC
Optimization of Transportation of Products 157
Also, the model assumes that there is always enough supply of the used
products to satisfy the demand for the remanufactured products and that
enough storage space (for used and remanufactured products) is always
available at the remanufacturing facilities. Moreover, inventory costs at
the collection centers and demand centers are not considered in the model
(for example, if the supply of used products is higher than the demand for
remanufactured products, the collection centers are deemed to incur inven
tory costs). However, the model may be revised accordingly in response to
revision of existing constraints or addition of new constraints.
8.2.3 Numerical Example
Three collection centers, two remanufacturing facilities, and three demand
centers are considered in the example (see fgure 8.1). Let the total supply of
the used product per period be a triangular fuzzy number (TFN): (600, 650,
700). The defuzzifed supply is 633.33 per period (see section 4.4 for the con
cepts of TFN and defuzzifcation). Assuming an equal supply rate at all three
collection centers, one gets SUP
1
= SUP
2
= SUP
3
= 211.11. The other data used
for implementation of the model are as follows:
C
1
= 29; C
2
= 25; C
3
= 37; TI
1A
= 3; TI
2A
= 4; TI
3A
= 5.3; TI
1B
= 3.2; TI
2B
= 1.4; TI
3B
= 6.7; TO
A1
= 2.6; TO
B1
= 3.2; TO
A2
= 3.4; TO
B2
= 2.5; TO
B3
= 1.6; TO
B3
= 2.1; R
A
= 4; R
B
= 4.3; d
1
= 100; d
2
= 200; d
3
= 150; a
1
= a
2
= 0.5; S
1A
= 550; S
1B
= 550; S
2A
= 550; S
2B
= 550; S
1
= 550; S
2
= 550; S
3
= 550; CAP
A
= 300; CAP
B
= 250
Upon application of the above data to the model, using LINGO (v4), the
following optimal solution is obtained:
I
1A
= 211, i.e., 211 products are to be transported from collection center
1 to recovery facility A
I
1B
= 0, i.e., no products are to be transported from collection center 1 to
recovery facility B
I
2A
= 0, i.e., no products are to be transported from collection center 2 to
recovery facility A
I
2B
= 211, i.e., 211 products are to be transported from collection center
2 to recovery facility B
Collection
Centers
Remanufacturing
Facilities
A and B
Consumers
Used Products Used Products
Demand
Centers
Remanufactured
Products
FIGURE 8.1
Reverse supply chain (frst and second models).
© 2009 by Taylor & Francis Group, LLC
158 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
I
3A
= 28, i.e., 28 products are to be transported from collection center 3
to recovery facility A
I
3B
= 0, i.e., no products are to be transported from collection center 3 to
recovery facility B
O
A1
= 100, i.e., 100 products are to be transported from recovery facility
A to demand center 1
O
A2
= 0, i.e., no products are to be transported from recovery facility A
to demand center 2
O
A3
= 139, i.e., 139 products are to be transported from recovery facility
A to demand center 3
O
B1
= 0, i.e., no products are to be transported from recovery facility B
to demand center 1
O
B2
= 200, i.e., 200 products are to be transported from recovery facility
B to demand center 2
O
B3
= 11, i.e., 11 products are to be transported from recovery facility B
to demand center 3
8.3 Second Model (Linear Physical Programming)
This model considers a generic reverse supply chain consisting of collection
centers, remanufacturing facilities, and demand centers. Optimal transporta
tion of products (used and remanufactured) is achieved using linear physical
programming. Section 8.3.1 presents the model formulation (using the nomen
clature given in section 8.2.1), and section 8.3.2 gives a numerical example.
8.3.1 Model Formulation
Class 1H criteria
Total retrieval cost per period (h
1
) given by
h
1
=
C I u uv
v u
` `
(8.11)
Class 1S criteria (smaller is better)
Total transportation cost per period (g
1
) given by
g
1
= TI I TO O uv uv
v u
vw vw
w v
` ` ` `
÷ (8.12)
© 2009 by Taylor & Francis Group, LLC
Optimization of Transportation of Products 159
Total remanufacturing cost per period (g
2
) given by
g
2
= R O v
w v
vw
` `
(8.13)
Total inventory cost per period (g
3
) given by
g
3
= ( / ). ( / ). C I R Q u uv
v u
v vw
w v
4 4 ÷
` ` ` `
(8.14)
Goal constraints
h
1
≤ RETMAX (retrieval cost is not more than maximum allowed
value—RETMAX) (8.15)
g
p
– d
pr
÷
≤ t
p r ( ) ÷
÷
1
(deviation is measured from corresponding target value) (8.16)
g
p
≤ t
p5
÷
(criterion value is in acceptable range) (8.17)
d
pr
÷
≥ 0; (deviation is a nonnegative number) (8.18)
System constraints
O d w vw
v
w
`
= \ (demand at each demand center must be met) (8.19)
I SUP v uv
u
u
`
= \ (all products must be transported from each collection
center) (8.20)
O I v vw
w
uv
u
` `
= \ (number of remanufactured products is equal to number
of used ones) (8.21)
© 2009 by Taylor & Francis Group, LLC
160 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
a I S v uv
u
v 2 2 . ;
`
_ \ (space occupied by used products is at most capacity) (8.22)
I u v uv _ \ 0 , (quantities of used products are nonnegative numbers) (8.23)
O v w vw _ \ 0 , (quantities of remanufactured products are
nonnegative numbers) (8.24)
It should be noted here that more constraints may be added to the above
model, as desired by the decision maker.
8.3.2 Numerical Example
Consider the reverse supply chain shown in fgure 8.1 in this example as
well. The data used for implementation of the model are as follows:
C
1
= 0.1; C
2
= 0.1; C
3
= 0.2; TI
1A
= 0.02; TI
1B
= 0.1; TI
2A
= 0.2; TI
2B
= 3; TI
3A
= 0.1;
TI
3B
= 1.2; TO
A1
= 4; TO
A2
= 5; TO
A3
= 0.04; TO
B1
= 1; TO
B2
= 0.1; TO
B3
= 0.2; R
A
= 0.2; R
B
= 0.3; d
1
= 90; d
2
= 80; d
3
= 80; SUP
1
= 75; SUP
2
= 150; SUP
3
= 25; a
1
=
a
2
= 0.5; S
2A
= S
2C
= 400
Furthermore, the target values for each soft criterion are shown in table 8.1,
and the incremental weights obtained by the LPP weight algorithm [2] are
shown in table 8.2.
Upon application of the above data to the model, using LINGO (v4), the
following optimal solution is obtained:
TABLE 8.1
Preference table (second model)
Criteria t
p1
+ t
p2
+ t
p3
+ t
p4
+ t
p5
+
g
1
100 200 300 400 500
g
2
150 250 290 450 600
g
3
70 150 250 300 450
TABLE 8.2
Output of LPP weight algorithm (second model)
Criteria Δw
p2
+ Δw
p3
+ Δw
p4
+ Δw
p5
+
g
1
0.025 0.085 0.132 0.024
g
2
0.017 0.011 0.026 0.479
g
3
0.013 0.031 0.881 0.012
© 2009 by Taylor & Francis Group, LLC
Optimization of Transportation of Products 161
I
1A
= 0; I
2A
= 80; I
3A
= 0; I
1B
= 75; I
2B
= 70; I
3B
= 25; O
A1
= 0; O
A2
= 0; O
A3
= 80;
O
B1
= 90; O
B2
= 80; O
B3
= 0
Interpretation of this solution could be made in a way similar to that in sec
tion 8.2.3.
8.4 Third Model (Goal Programming)
This model considers a generic closedloop supply chain consisting of col
lection centers, production facilities, and demand centers. Goal program
ming is employed to address the following issues in one continuous phase,
besides optimal transportation of products: selection of used products and
evaluation of production facilities. Section 8.4.1 gives the nomenclature for
the model, section 8.4.2 presents the formulation of the model, and section
8.4.3 illustrates the model with a numerical example.
8.4.1 Nomenclature
A
iuv
Decision variable representing number of used products of
type i transported from collection center u to production
facility v
B
ivw
Decision variable representing number of new/remanufac
tured products of type i transported from production facil
ity v to demand center w
b
i
Probability of breakage of product i
TA
uv
Cost to transport one product from collection center u to
production facility v
TB
vw
Cost to transport one product from production facility v to
demand center w
CC
u
Cost per product retrieved at collection center u
CNP
v
Cost to produce one unit of new product at production facil
ity v
CR
v
Cost to produce one unit of remanufactured product at pro
duction facility v
C
di
Disposal cost of product i
DI
i
Disposal cost index of product i (0 = lowest, 10 = highest)
DT
i
Disassembly time for product i
DC Disassembly cost/unit time
i Product type
© 2009 by Taylor & Francis Group, LLC
162 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
MINTPS Minimum throughput per supply
N
ivw
Decision variable representing number of new products of
type i transported from production facility v to demand cen
ter w
Nd
iw
Net demand for product type i (remanufactured or new) at
demand center w
PRC
i
Percent of recyclable contents by weight in product i
RCYR
i
Total recycling revenue of product i
RSR
i
Total resale revenue of product i
RCRI
i
Recycling revenue index of product i
S
1v
Storage capacity of production facility v for used products
S
2v
Storage capacity of production facility v for remanufactured
and new products
S
u
Storage capacity of collection center u
SP
i
Selling price of one unit of new product of type i
SU
iu
Supply of used product i at collection center u
SF
v
Supply of used products that are ft for remanufacturing
(viz., excluding products selected for recycling or disposal)
and new products, at production facility v
TP
v
Throughput (considering only remanufactured products) of
production facility v
u Collection center
v Production facility
w Demand center
W
i
Weight of product i
x
1
Space occupied by one unit of used product
x
2
Space occupied by one unit of remanufactured or new
product
Y
v
Decision variable signifying selection of production facility
v (1 if selected, 0 if not)
Z
iu
Decision variable representing number of units of product
type i picked for remanufacturing at collection center u (SU
iu
– Z
iu
= recycled or disposed)
E
v
Factor that accounts for unassignable causes of variations at
production facility v
8.4.2 Model Formulation
This goal programming model, in one continuous phase, determines the
number of used products of each type to be picked for remanufacturing,
© 2009 by Taylor & Francis Group, LLC
Optimization of Transportation of Products 163
identifes effcient production facilities, and achieves transportation of the
right quantities of products (used, remanufactured, and new) across a closed
loop supply chain.
It is assumed that the inventory costs of a used product and a remanufac
tured product are 20% of the collection and remanufacturing costs, respec
tively, and that of a newly produced product is 25% of the production cost.
The following three goals are considered:
1. Maximize the total proft in the supply chain (TP).
2. Maximize the revenue from recycling (RR).
3. Minimize the number of disposed items (NDIS).
The frst two goals involve minimizing the negative deviation from the
respective target values, whereas the third goal, which has an environmen
tally benign focus rather than a fnancial focus, involves minimizing the posi
tive deviation from the target value.
The revenue and cost criteria and the system constraints considered in the
model are:
Revenues
1. Reuse revenue:
{ } Z RSR
iu i
u i
` `
(8.25)
2. Recycle revenue:
{( ) } SU Z RCRI W PRC
iu iu i i i
u i
÷
` `
(8.26)
3. New product sale revenue:
SP N
i ivw
w v i
` ` `
(8.27)
Costs
1. Collection/retrieval cost:
CC SU
u iu
i u
` `
(8.28)
© 2009 by Taylor & Francis Group, LLC
164 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
2. Processing cost = disassembly cost of used products + reman
ufacturing cost of used products + new products’ production
cost:
DC DT A CR B CN
i iuv
v u i
v ivw
(
(
·
·
·
·
·
\
)
÷ ÷
` ` `
PP N
v ivw
w v i v u i
` ` ` ` ` `
(8.29)
3. Inventory cost = cost of carrying used products’ inventory at
collection centers + cost of carrying remanufactured products’
inventory at production facilities + cost of carrying newly manu
factured products’ inventory at production facilities:
( / ) {( / ) ( CC A CR B CNP
u
v u i
iuv v ivw
w v i
` ` ` ` ` `
÷ ÷ 5 4
vv ivw
N / ) } 4 +
(
(
·
·
·
·
·
\
)
(8.30)
4. Transportation cost = cost of transporting used products from col
lection centers to production facilities + cost of transporting reman
ufactured and new products from production facilities to demand
centers:
TA A TB B N
uv iuv
v u i
vw ivw ivw
w v i
÷ ÷
` ` ` ` ` `
( ) (8.31)
5. Disposal cost: Apart from the number of units disposed of and
the cost to dispose of a product, disposal cost depends on the
percentage of recyclable content in the product, and the disposal
cost index (a number on a scale 0–10; the higher the number, the
more diffcult or expensive it is to dispose of the product):
{( ) ( )} SU Z DI W PRC C
iu iu i i i
u i
di
÷ ÷
` `
1 (8.32)
System constraints
1. The number of used products sent to all production facilities
from collection center u must be equal to the number of used
products picked for remanufacturing at that collection center:
A Z
iuv
v
iu
`
=
(8.33)
© 2009 by Taylor & Francis Group, LLC
Optimization of Transportation of Products 165
2. The demand at each center w must be met by either new or
remanufactured products:
( ) B N Nd w
ivw ivw iw
v
÷ = \
`
(8.34)
3. The number of remanufactured products transported from pro
duction facility v to demand center w equals (number of used
products ft for remanufacturing that are transported from col
lection center u to that production facility) multiplied by E
v
. b
v
is a factor that accounts for the unassignable (common) causes
of variation at production facility v. That is, there is no loss of
products in the supply chain due to reasons other than common
cause variations over which there is no control:
B A v
ivw iuv v
u w
= \
` `
D
(8.35)
4. The number of used products of type i picked for remanufactur
ing at collection center u must be at most equal to the total num
ber of used products ft for remanufacturing:
Z SU b
iu iu i
_ ÷ ( ) 1 (8.36)
5. The total number of used products of all types collected (before
accounting for probability of breakage) at all collection centers
must be at least equal to the net demand (this is to encourage the
use of remanufactured products):
SU Nd
iu iw
w i u i
_
` ` ` `
(8.37)
6. The number of remanufactured products must be at most equal
to the net demand (this is to avoid excess remanufacturing):
Z Nd
iu
u i
iw
w i
` ` ` `
_ (8.38)
7. The space occupied by used products at production facility v
must at most be equal to the space available for used products at
that facility:
© 2009 by Taylor & Francis Group, LLC
166 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
x A S Y
iuv
u i
v v 1 1
` `
_ .
(8.39)
8. The space occupied by new and remanufactured products at pro
duction facility v must at most be equal to the space available
for new and remanufactured products at that production facility
(assuming both new and remanufactured products occupy the
same space):
x B N S Y
ivw ivw v v
w i
2 2
( ) ÷ _
` `
(8.40)
9. The space occupied by used products at collection center u must
at most be equal to the space available for used products at that
collection center:
x a S
iuv u
v i
1
_
` `
(8.41)
10. The ratio of throughput to supply of used products of a produc
tion facility must at least be equal to a preset value for the pro
duction facility to be considered effcient (this is valid only for
remanufactured products):
TP SF Y MINTPS
v v v
/
( )
_ (8.42)
11. Nonnegativity constraints:
A B N Z u v w i
iuv ivw ivw iu
, , , , , , , _ \ 0 (8.43)
12. Y v
v
=
l
l
l
\ 0 1 , , 0 if facility v not selected, 1 if selected (8.44)
8.4.3 Numerical Example
Three collection centers, two production facilities, two demand centers, and
three types of products are considered in the example (see fgure 8.2). The
data used to implement the goal programming model are:
CC
u
= 0.01; SU
11
= 50; SU
12
= 45; SU
1
= 25; SU
21
= 35; SU
22
= 38; SU
23
= 22; SU
31
= 30; SU
32
= 35; SU
33
= 28; DC = 0.05; DT
1
= 10; DT
2
= 12; DT
3
= 9; CR
1
= 13; CR
2
= 10; CNP
1
= 60; CNP
2
= 45; TA
11
= 0.01; TA
12
= 0.09; TA
21
= 0.5; TA
22
= 0.1; TA
31
© 2009 by Taylor & Francis Group, LLC
Optimization of Transportation of Products 167
= 0.02; TA
32
= 0.04; TB
11
= 0.04; TB
12
= 0.03; TB
21
= 0.09; TB
22
= 0.05; DI
1
= 4; DI
2
= 6; DI
3
= 5; W
1
= 0.8; W
2
= 1.0; W
3
= 0.9; PRC
1
= 0.5; PRC
2
= 0.6; PRC
3
= 0.75;
Cd
1
= 0.2; Cd
2
= 0.5; Cd
3
= 0.3; RSR
1
= 30; RSR
2
= 40; RSR
3
= 45; RCYR
1
= 1.5;
RCYR
2
= 2; RCYR
3
= 2.5; RCRI
1
= 7; RCRI
2
= 4; RCRI
3
= 5; SP
1
= 65; SP
2
= 55;
SP
3
= 60; Nd
11
= 20; Nd
12
= 15; Nd
21
= 16; Nd
22
= 22; Nd
31
= 25; Nd
32
= 20; E
1
= 0.4;
E
2
= 0.6; b
1
= 0.2; b
2
= 0.4; b
3
= 0.3; X
1
= 0.7; S
11
= 400; S
12
= 400; S
1
= 150; S
2
= 150;
S
3
= 150; X
2
= 0.7; S
21
= 500; S
22
= 500; MINTPS = 0.25
Upon application of the above data to the model, using LINGO (v4), the
following optimal solution is obtained:
TP = 3,945 (target = 2,500); RR = 951 (target = 750); NDIS = 74 (target = 50);
Z
12
= 35; Z
21
= 2; Z
22
= 23; Z
23
= 13; Z
31
= 20; Z
32
= 12; Z
33
= 12; N
111
= 3; N
112
= 5;
N
211
= 8; N
222
= 1; N
211
= 3; N
312
= 14; A
121
= 15; A
122
= 20; A
212
= 2; A
221
= 5; A
222
=
18; A
231
= 5; A
232
= 8; A
311
= 11; A
312
= 9; A
321
= 12; A
331
= 3; B
111
= 8; B
112
= 4; B
211
= 5; B
212
= 3; B
311
= 15; B
312
= 6; B
121
= 9; B
122
= 6; B
221
= 3; B
222
= 18; B
321
= 7; Y
1
= 1;
Y
2
= 1
It is evident from the above solution that both of the production facilities
are chosen for the network design. Also, 71% of the net demand is satisfed
by remanufactured products and the remaining 29% by newly manufac
tured products.
Collection
Centers
1, 2, and 3
Production
Facilities
1 and 2
Demand
Centers
1 and 2
New Products
Remanufactured Products
Consumers
New Products
Remanufactured Products
Used Products
Used Products
FIGURE 8.2
Closedloop supply chain (third, fourth, and ffth models).
© 2009 by Taylor & Francis Group, LLC
168 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
8.5 Fourth Model (Linear Physical Programming)
This model considers a generic closedloop supply chain consisting of col
lection centers, production facilities, and demand centers. Linear physical
programming is employed to address the following issues in one continuous
phase, besides optimal transportation of products: selection of used prod
ucts and evaluation of production facilities. Section 8.5.1 presents the model
formulation using the nomenclature given in section 8.4.1, and section 8.5.2
gives a numerical example.
8.5.1 Model Formulation
This linear physical programming model, in one continuous phase, deter
mines the number of used products of each type to be picked for remanufac
turing, identifes effcient production facilities, and achieves transportation
of the right quantities of products (used, remanufactured, and new) across a
closedloop supply chain.
It is assumed that the inventory costs of a used product and a remanufac
tured product are 20% of the collection and remanufacturing costs, respec
tively and that the inventory cost of a newly produced product is 25% of the
production cost.
The revenue and cost criteria and system constraints considered in the
model are:
Costs: Class 1S (smaller is better)
1. Collection/retrieval cost:
CC SU
u iu
i u
` `
(8.45)
2. Processing cost = disassembly cost of used products + reman
ufacturing cost of used products + new products’ production
cost:
DC DT A CR B CN
i iuv
v u i
v ivw
(
(
·
·
·
·
·
\
)
÷ ÷
` ` `
PP N
v ivw
w v i v u i
` ` ` ` ` `
(8.46)
3. Transportation cost = cost of transporting used products from col
lection centers to production facilities + cost of transporting reman
ufactured and new products from production facilities to demand
centers:
© 2009 by Taylor & Francis Group, LLC
Optimization of Transportation of Products 169
TA A TB B N
uv iuv
v u i
vw ivw ivw
w v i
÷ ÷
` ` ` ` ` `
( )
(8.47)
4. Disposal cost: Apart from the number of units disposed of and
the cost to dispose of a product, disposal cost depends on the
percentage of recyclable content in the product and the disposal
cost index (a number on a scale 0–10; the higher the number, the
more diffcult or expensive it is to dispose of the product):
{( ) ( )} SU Z DI W PRC C
iu iu i i i
u i
di
÷ ÷
` `
1
(8.48)
Revenues: Class 2S (larger is better)
1. Reuse revenue:
{ } Z RSR
iu i
u i
` `
(8.49)
2. Recycle revenue:
{( ) } SU Z RCRI W PRC
iu iu i i i
u i
÷
` `
(8.50)
3. New product sales revenue:
SP N
i ivw
w v i
` ` `
(8.51)
System constraints
1. The number of used products sent to all production facilities
from collection center u must be equal to the number of used
products picked for remanufacturing at that collection center:
A Z
iuv
v
iu
`
= (8.52)
2. The demand at each center w must be met by either new or
remanufactured products:
( ) B N Nd w
ivw ivw iw
v
÷ = \
`
(8.53)
© 2009 by Taylor & Francis Group, LLC
170 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
3. The number of remanufactured products transported from pro
duction facility v to demand center w equals (number of used
products ft for remanufacturing that are transported from col
lection center u to that production facility) multiplied by E
v
. E
v
is a factor that accounts for the unassignable (common) causes
of variation at production facility v. That is, there is no loss of
products in the supply chain due to reasons other than common
cause variations, over which there is no control:
B A v
ivw iuv v
u w
= \
` `
D (8.54)
4. The number of used products of type i picked for remanufactur
ing at collection center u must be at most equal to the total num
ber of used products ft for remanufacturing:
Z SU b
iu iu i
_ ÷ ( ) 1 (8.55)
5. The total number of used products of all types collected (before
accounting for probability of breakage) at all collection centers
must be at least equal to the net demand (this is to encourage the
use of remanufactured products):
SU Nd
iu iw
w i u i
_
` ` ` `
(8.56)
6. The number of remanufactured products must be at most equal
to the net demand (this is to avoid excess remanufacturing):
Z Nd
iu
u i
iw
w i
` ` ` `
_ (8.57)
7. The space occupied by used products at production facility v
must at most be equal to the space available for used products at
that facility:
x A S Y
iu
u i
v v 1 1
` `
_ . (8.58)
8. The space occupied by new and remanufactured products at pro
duction facility v must at most be equal to the space available
© 2009 by Taylor & Francis Group, LLC
Optimization of Transportation of Products 171
for new and remanufactured products at that production facility
(assuming both new and remanufactured products occupy the
same space):
x B N S Y
ivw ivw v v
w i
2 2
( ) ÷ _
` `
(8.59)
9. The space occupied by used products at collection center u must
at most be equal to the space available for used products at that
collection center:
x a S
iuv u
v i
1
_
` `
(8.60)
10. The ratio of throughput to supply of used products of a produc
tion facility must at least be equal to a preset value for the pro
duction facility to be considered effcient (this is valid only for
remanufactured products):
TP SF Y MINTPS
v v v
/
( )
_ (8.61)
11. Nonnegativity constraints:
A B N Z u v w i
iuv ivw ivw iu
, , , , , , , _ \ 0 (8.62)
12.
Y v
v
=
l
l
l
\ 0 1 ,
, 0 if facility v not selected, 1 if selected (8.63)
8.5.2 Numerical Example
Three collection centers, two production facilities, two demand centers, and
three types of products are considered in the example (see fgure 8.2). The
data used to implement the linear physical programming model are:
CC
u
= 0.01; SU
11
= 20; SU
12
= 25; SU
13
= 15; SU
21
= 25; SU
22
= 18; SU
23
= 15; SU
31
= 17; SU
32
= 9; SU
33
= 15; DC = 0.005; DT
1
= 10; DT
2
= 12; DT
3
= 9; CR
1
= 10; CR
2
= 8; CNP
1
= 55; CNP
2
= 65; TA
11
= 0.001; TA
12
= 0.009; TA
21
= 0.01; TA
22
= 0.002;
TA
31
= 0.004; TA
32
= 0.003; TB
11
= 0.004; TB
12
= 0.003; TB
21
= 0.009; TB
22
= 0.005;
DI
1
= 4; DI
2
= 6; DI
3
= 5; W
1
= 0.8; W
2
= 1.0; W
3
= 0.9; PRC
1
= 0.65; PRC
2
= 0.6;
PRC
3
= 0.75; Cd
1
= 0.02; Cd
2
= 0.05; Cd
3
= 0.03; RSR
1
= 80; RSR
2
= 80; RSR
3
=
65; RCYR
1
= 5; RCYR
2
= 7; RCYR
3
= 105; RCRI
1
= 7; RCRI
2
= 4; RCRI
3
= 6; SP
1
= 100; SP
2
= 110; SP
3
= 95; Nd
11
= 20; Nd
12
= 15; Nd
21
= 16; Nd
22
= 22; Nd
31
= 25;
© 2009 by Taylor & Francis Group, LLC
172 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
Nd
32
= 20; E
1
= 0.85; E
2
= 0.75; b
1
= 0.2; b
2
= 0.4; b
3
= 0.3; X
1
= 0.7; S
11
= 400; S
12
=
400; S
1
= 150; S
2
= 150; S
3
= 150; X
2
= 0.7; S
21
= 500; S
22
= 500; MINTPS = 0.25
The target values for the criteria are given in table 8.3 (target values are
scaled by a factor of 10), and table 8.4 shows the incremental weights obtained
using the LPP weight algorithm [2].
Upon application of the above data to the model, using LINGO (v4), the
following optimal solution is obtained:
Z
11
= 0; Z
12
= 0; Z
13
= 0; Z
21
= 15; Z
22
= 11; Z
23
= 9; Z
31
= 3; Z
32
= 6; Z
33
= 6; N
111
= 15; N
112
= 11; N
211
= 0; N
212
= 0; N
311
= 18; N
312
= 15; N
121
= 5; N
122
= 4; N
221
= 4;
N
222
= 5; N
321
= 0; N
322
= 5; A
111
= 0; A
112
= 0; A
121
= 0; A
122
= 0; A
131
= 0; A
132
= 0;
A
211
= 11; A
212
= 4; A
221
= 8; A
222
= 3; A
231
= 7; A
232
= 2; A
311
= 3; A
312
= 0; A
321
= 5;
A
322
= 2; A
331
= 0; A
332
= 0; B
111
= 0; B
112
= 0; B
211
= 8; B
212
= 14; B
311
= 6; B
312
= 0;
B
121
= 0; B
122
= 0; B
221
= 4; B
222
= 3; B
321
= 1; B
322
= 0; Y
1
= 1; Y
2
= 1
It is evident from the above solution that both production facilities are
chosen for the network design. Also, 69% of the net demand is satisfed by
newly manufactured products and the remaining 31% by remanufactured
TABLE 8.3
Target values of criteria (fourth model)
Criteria t
p1
+ t
p2
+ t
p3
+ t
p4
+ t
p5
+
g
1
1 3 5 7 9
g
2
5 10 13 17 20
g
3
2 2.5 5 7 10
g
4
2 5 7 9 13
Criteria t
p1
– t
p2
– t
p3
– t
p4
– t
p5
–
g
5
10 15 20 25 30
g
6
10 15 17 19 22
g
7
15 17 20 25 35
TABLE 8.4
Output of LPP weight algorithm (fourth model)
Criteria Δw
p2
+ Δw
p3
+ Δw
p4
+ Δw
p5
+ Δw
p2
– Δw
p3
– Δw
p4
– Δw
p5
–
g
1
0.05 0.17 0.75 3.29 — — — —
g
2
0.02 0.13 0.34 2.34 — — — —
g
3
0.20 0.02 1.34 4.29 — — — —
g
4
0.03 0.25 1.29 2.82 — — — —
g
5
— — — — 0.02 0.02 0.05 0.12
g
6
— — — — 0.02 0.09 0.13 0.11
g
7
— — — — 0.05 0.03 0.04 0.02
© 2009 by Taylor & Francis Group, LLC
Optimization of Transportation of Products 173
products. In addition, demand for the frst product type is completely satis
fed by newly manufactured products of that type.
8.6 Fifth Model (Fuzzy Goal Programming)
This model considers a generic closedloop supply chain consisting of col
lection centers, production facilities, and demand centers. Fuzzy goal pro
gramming is employed to address the following issues in one continuous
phase, besides optimal transportation of products: selection of used prod
ucts and evaluation of production facilities. Section 8.6.1 presents the model
formulation using the nomenclature given in section 8.4.1, and section 8.6.2
gives a numerical example.
8.6.1 Model Formulation
This fuzzy goal programming model, in one continuous phase, determines
the number of used products of each type to be picked for remanufactur
ing, identifes effcient production facilities, and achieves transportation of
the right quantities of products (used, remanufactured, and new) across a
closedloop supply chain.
It is assumed that the inventory costs of a used product and a remanufac
tured product are 20% of the collection and remanufacturing costs, respec
tively, and that the inventory cost of a newly produced product is 25% of the
production cost.
Three goals are considered in the model:
1. Maximize the total proft in the closedloop supply chain (TP).
2. Maximize the revenue from recycling (RR).
3. Minimize the number of disposed items (NDIS).
The frst two goals involve minimizing the negative deviation from the
respective target values, whereas the third goal, which has an environmen
tally benign character rather than a fnancial basis, involves minimizing the
positive deviation from the target value. The membership functions for the
three goals are
M
1
1
0
=
_
÷
÷
_ _
_
if TP TP
TP TP
TP TP
if TP TP TP
if TP
L
L
L
*
*
*
TTP
L
'
¦
1
1
1
1
1
1
1
+
1
1
1
1
1
1
1
(8.64)
© 2009 by Taylor & Francis Group, LLC
174 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
where TP
*
is the aspiration level of total proft (TP) and TP
L
is the lower toler
ance level of TP;
M
2
1
0
=
_
÷
÷
_ _
_
if RR RR
RR RR
RR RR
if RR RR RR
if RR
L
L
L
*
*
*
RRR
L
'
¦
1
1
1
1
1
1
1
+
1
1
1
1
1
1
1
(8.65)
where RR
*
is the aspiration level of recycling revenue (RR) and RR
L
is the
lower tolerance level of RR; and
M
3
1
=
_
÷
÷
_ _
if NDI NDI
NDI NDI
NDI NDI
if NDI NDI N
U
U
*
*
*
DDI
if NDI NDI
U
U
0 _
'
¦
1
1
1
1
1
1
1
+
1
1
1
1
1
1
1
(8.66)
where NDI
*
is the aspiration level of number of disposed items (NDI) and
NDI
U
is the upper tolerance level of NDI.
According to the concept of Fibonacci numbers, starting with 1 and 2, the
weight values for μ
1
, μ
2
, and μ
3
are 0.5, 0.33, and 0.17. The objective function
for the weighted fuzzy goal programming model is as follows:
Maximize V( ) ( . ) ( . ) ( . ) M M M M = ÷ ÷ 0 5 0 33 0 17
1 2 3
(8.67)
The cost and revenue criteria and the system constraints considered in the
model include:
Revenues
1. Reuse revenue:
{ } Z RSR
iu i
u i
` `
(8.68)
2. Recycle revenue:
{( ) } SU Z RCYI W PRC
iu iu i i i
u i
÷
` `
(8.69)
© 2009 by Taylor & Francis Group, LLC
Optimization of Transportation of Products 175
3. New product sales revenue:
SP N
i ivw
w v i
` ` `
(8.70)
Costs
1. Collection/retrieval cost:
CC SU
u iu
i u
` `
(8.71)
2. Processing cost = disassembly cost of used products + remanu
facturing cost of used products + new products production cost
in the forward supply chain:
DC DT A CR B CN
i iuv
v u i
v ivw
(
(
·
·
·
·
·
\
)
÷ ÷
` ` `
PP N
v ivw
w v i v u i
` ` ` ` ` `
(8.72)
3. Inventory cost = cost of carrying used products inventory at the
collection center + cost of carrying remanufactured products
inventory at the production facility + cost of carrying newly
manufactured products inventory at the production facility:
( / ) {( / ) ( CC A CR B CNP
u
v u i
iuv v ivw
w v i
` ` ` ` ` `
÷ ÷ 5 4
vv ivw
N / ) } 4
(
(
·
·
·
·
·
\
)
(8.73)
4. Transportation costs = cost of transporting used products from
collection centers to production facility + remanufactured and
new products from production facilities to demand centers:
TA A TB B N
uv iuv
v u i
vw ivw ivw
w v i
÷ ÷
` ` ` ` ` `
( ) (8.74)
5. Disposal cost: Apart from the number of units disposed of and
the cost to dispose of a product, disposal cost depends on the
percentage of recyclable content in the product and the disposal
cost index (a number on a scale 0–10; the higher the number, the
more diffcult or expensive it is to dispose of the product):
© 2009 by Taylor & Francis Group, LLC
176 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
{( ) ( )} SU Z DI W PRC C
iu iu i i i
u i
di
÷ ÷
` `
1
(8.75)
System constraints
1. The number of used products sent to all production facilities
from a collection center u must be equal to the number of used
products picked for remanufacturing at that collection center:
A Z
iuv
v
iu
`
=
(8.76)
2. The demand at each center w must be met by either new or
remanufactured products:
( ) B N Nd w
ivw ivw iw
v
÷ = \
`
(8.77)
3. The number of remanufactured products transported from a pro
duction facility v to a demand center w equals (number of used
products ft for remanufacturing that are transported from col
lection center u to that production facility) multiplied by E
v
v, i.e.,
no loss of products in the supply chain due to reasons other than
common cause variations, over which there is no control. E
v
is a
factor that accounts for the unassignable (common) causes of vari
ation at the production facility v:
B A v
ivw iuv v
u w
= \
` `
D
(8.78)
4. The total number of used products of type i picked for remanu
facturing at collection center u must be at most equal to the total
number of used products ft for remanufacturing:
Z SU b
iu iu i
_ ÷ ( ) 1 (8.79)
5. The total number of used products of all types collected (before
accounting for probability of breakage) at all collection centers
must be at least equal to the net demand (this is to encourage the
use of remanufactured products):
© 2009 by Taylor & Francis Group, LLC
Optimization of Transportation of Products 177
SU Nd
iu iw
w i u i
_
` ` ` `
(8.80)
6. The number of remanufactured products must be at most equal
to the net demand (this is to avoid excess remanufacturing):
Z Nd
iu
u i
i
w i
` ` ` `
_
(8.81)
7. The space occupied by used products at the production facil
ity, v, must at most be equal to the space available for used
products at that facility:
x A S Y
iuv
u i
v v 1 1
` `
_ .
(8.82)
8. The space occupied by new and remanufactured products at the
production facility, v, must at most be equal to the space available
for new and remanufactured products at that production facility
(assuming both new and remanufactured products occupy the
same space):
x B N S Y
ivw ivw v v
w i
2 2
( ) ÷ _
` `
(8.83)
9. The space occupied by used products at the collection center, u,
must at most be equal to the space available for used products at
that collection center:
x a S
iuv u
v i
1
_
` `
(8.84)
10. The ratio of throughput to supply of used products of a produc
tion facility must at least be equal to a preset value for the pro
duction facility to be considered effcient (this is valid only for
remanufactured products):
TP SF Y MINTPS
v v v
/
( )
_ (8.85)
© 2009 by Taylor & Francis Group, LLC
178 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
11. Nonnegativity constraints:
A B N Z u v w i
iuv ivw ivw iu
, , , , , , , _ \ 0 (8.86)
12. Y v
v
=
l
l
l
\ 0 1 , , 0 if facility v not selected, 1 if selected (8.87)
8.6.2 Numerical Example
Three collection centers, two production facilities, two demand centers, and
three types of products are considered in the example (see fgure 8.2). The
data used to implement the fuzzy goal programming model are:
CC
u
= 0.01; SU
11
= 50; SU
12
= 45; SU
1
= 25; SU
21
= 35; SU
22
= 38; SU
23
= 22; SU
31
= 30; SU
2
= 35; SU
33
= 28; DC = 0.05; DT
1
= 10; DT
2
= 12; DT
3
= 9; CR
1
= 13; CR
2
= 10; CNP
1
= 60; CNP
2
= 45; TA
11
= 0.01; TA
12
= 0.09; TA
21
= 0.5; TA
22
= 0.1;
TA
31
= 0.02; TA
32
= 0.04; TB
11
= 0.04; TB
12
= 0.03; TB
21
= 0.09; TB
22
= 0.05; DI
1
=
4; DI
2
= 6; DI
3
= 5; W
1
= 0.8; W
2
= 1.0; W
3
= 0.9; PRC
1
= 0.5; PRC
2
= 0.6; PRC
3
=
0.75; Cd
1
= 0.2; Cd
2
= 0.5; Cd
3
= 0.3; RSR
1
= 30; RSR
2
= 40; RSR
3
= 45; RCYR
1
=
1.5; RCYR
2
= 2; RCYR
3
= 2.5; RCRI
1
= 7; RCRI
2
= 4; RCRI
3
= 5; SP
1
= 65; SP
2
=
53; SP
3
= 60; Nd
11
= 20; Nd
12
= 15; Nd
21
= 16; Nd
22
= 22; Nd
31
= 25; Nd
32
= 20; E
1
= 0.4; E
2
= 0.6; b
1
= 0.2; b
2
= 0.4; b
3
= 0.3; X
1
= 0.7; S
11
= 400; S
12
= 400; S
1
= 150;
S
2
= 150; S
3
= 150; X
2
= 0.7; S
21
= 500; S
22
= 500; TP
*
= 3,500; TP
L
= 2,000; RR
*
=
1,000; RR
L
= 800; NDI
*
= 60; NDI
U
= 100; MINTPS = 0.25
Upon application of the above data to the model using LINGO (v4), the fol
lowing optimal solution is obtained:
Z
12
= 15; Z
13
= 20; Z
21
= 21; Z
22
= 17; Z
31
= 5; Z
32
= 18; Z
33
= 18; N
111
= 7; N
211
= 9;
A
121
= 15; A
131
= 20; A
212
= 21; A
222
= 17; A
311
= 5; A
312
= 9; A
321
= 18; A
331
= 17; A
332
= 20; B
111
= 13; B
112
= 15; B
311
= 25; B
312
= 6; B
221
= 7; B
222
= 22; B
322
= 14; Y
1
= 1; Y
2
= 1; TP = 3,500; RR = 1,000; NDI = 77; μ
1
= 1; μ
2
= 1; μ
3
= 0.59
It should be noted that the achievements of TP and RR (μ
1
and μ
2
) are at
their maximum, whereas that of NDI is not (μ
3
). In addition, both production
facilities are chosen for the network design. Also, 86% of the net demand
is satisfed by remanufactured products and the remaining 14% by newly
manufactured products.
© 2009 by Taylor & Francis Group, LLC
Optimization of Transportation of Products 179
8.7 Conclusions
In this chapter, fve models are presented for achieving transportation of the
right quantities of products (used, remanufactured, and new) across a reverse
or closedloop supply chain while satisfying certain constraints. The frst and
second models are for a reverse supply chain and use linear integer program
ming and linear physical programming, respectively. The third, fourth, and
ffth models are for a closedloop supply chain and employ goal programming,
linear physical programming, and fuzzy goal programming, respectively.
References
1. Fleischmann, M. 2001. Quantitative models for reverse logistics: Lecture notes in
economics and mathematical systems. Berlin: SpringerVerlag.
2. Messac, A., Gupta, S. M., and Akbulut, B. 1996. Linear physical programming:
A new approach to multiple objective optimization. Transactions of Operational
Research 8:39–59.
© 2009 by Taylor & Francis Group, LLC
181
9
Evaluation of Marketing Strategies
9.1 The Issue
The success of a reverse/closedloop supply chain program is heavily depen
dent on the level of public participation (in the program), which in turn is
shouldered by the marketing strategy of that program. Hence, evaluating
the marketing strategy of a reverse/closedloop supply chain program is
equivalent to evaluating how well the strategy is driving the public to par
ticipate in the program. Studies (for example, see [1] and [2]) are conducted in
numerous cities around the world in order to assess the level of participation
of the public in the respective reverse/closedloop supply chain programs.
The offcials of each program painstakingly approach many homes in the
respective city with questions regarding how convenient the program is to
the public and how the program can be improved. Although the drivers for
governments and companies to implement these programs and evaluate the
programs’ marketing strategies are environmental consciousness and proft
ability, respectively, the drivers for the public to participate in the programs
are numerous and often conficting with each other (for example, the more
regularly a reverse/closedloop supply chain program offers to collect used
products from consumers, the higher the taxes the consumers will have to
pay; high regularity of collection and low tax levied on the consumers are
conficting drivers here).
The various scenarios for evaluating the marketing strategy of a reverse/
closedloop supply chain program could differ as follows:
1. The program could be exclusively for reverse supply chain opera
tions, i.e., absence of a closed loop.
2. The program could be for a closedloop supply chain and the deci
sion maker could be uninterested in considering interdependencies
among evaluation criteria.
3. The program could be for a closedloop supply chain and the deci
sion maker could be interested in considering interdependencies
among evaluation criteria.
© 2009 by Taylor & Francis Group, LLC
182 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
In this chapter, various models to address the above scenarios are pre
sented. The frst model addresses scenario 1 and employs fuzzy logic and
technique for order preference by similarity to ideal solution (TOPSIS). The
second model is for scenario 2 and employs fuzzy logic, quality function
deployment (QFD), and method of total preferences. The third model is for
scenario 3 and uses fuzzy logic, extent analysis method, and analytic net
work process. This chapter is organized as follows: Section 9.2 presents the
frst model, section 9.3 shows the second model, section 9.4 presents the third
model, and section 9.5 gives some conclusions.
9.2 First Model (Fuzzy Logic and TOPSIS)
In section 9.2.1, a list of drivers for the public to participate in a reverse sup
ply chain program are identifed. Then, in section 9.2.2, using a numerical
example, fuzzy logic and TOPSIS are employed to evaluate the marketing
strategy of a reverse supply chain program with respect to the drivers identi
fed in section 9.2.1.
9.2.1 Drivers of Public Participation
The following is a fairly exhaustive list of selfexplanatory drivers for the
public to participate in a reverse supply chain:
1. Knowledge of the drivers of implementation of the reverse supply
chain program (KD)
2. Awareness of the reverse supply chain program being imple
mented (AR)
3. Simplicity of the reverse supply chain program (SR)
4. Convenience for disposal of used products at collection centers (CD)
5. Incentives for disposal of used products (ID)
6. Effectiveness of collection methods (EC)
7. Information supplied about used products being collected (IU)
8. Regularity of collection of used products (RC)
9. Design of special methods for abusers of the reverse supply chain
program (AB)
10. Good locations of demand centers where reprocessed goods are
sold (LR)
11. Incentives to buyers of reprocessed goods (IB)
12. Cooperation of the program organizers with the local government (CL)
© 2009 by Taylor & Francis Group, LLC
Evaluation of Marketing Strategies 183
9.2.2 Methodology
Suppose there are three representatives of a community to weigh the drivers
of public participation, depending on which driver greatly motivates them to
participate, which driver is not so important for them, and so on. Because it
is diffcult for them to assign numerical weights, they give linguistic weights
like “very high,” “low,” and “medium.” Table 9.1 illustrates the linguistic
weights. Using fuzzy logic, these linguistic weights are converted into trian
gular fuzzy numbers (TFNs; table 9.2 shows one of the many ways for such
a conversion) and then averaged to form another TFN called the average
weight. For example, the average weight of the driver ID (see tables 9.1 and
9.2) is
Low High Low
3
TABLE 9.1
Linguistic weights of drivers of public participation (frst model)
Driver Rep. 1 Rep. 2 Rep. 3
KD Low Medium High
AR Medium High Very high
SR Low Low Very high
CD Very high Very high Medium
ID Low High Low
EC High Medium Very high
IU Low Low High
RC Medium Low Low
AB Medium Low Medium
LR Very high High High
IB High High Medium
CL Medium High Low
TABLE 9.2
Conversion table for weights of drivers (frst model)
Linguistic weight TFN
Very high (0.7, 0.9, 1.0)
High (0.5, 0.7, 0.9)
Medium (0.3, 0.5, 0.7)
Low (0.1, 0.3, 0.5)
Very low (0.0, 0.1, 0.3)
© 2009 by Taylor & Francis Group, LLC
184 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
which is
0 1 0 5 0 1
3
0 3 0 7 0 3
3
0 5 0 9 0 5
3
. . .
,
. . .
,
. . .
¤
¦
¥
¥
¥
¥¥
´
¶
µ
µ
µ
µ
= (0.23, 0.43, 0.63)
(see table 9.3).
The sum of the average weights of all the drivers is calculated using equa
tion (4.4) as (4.27, 6.33, 8.93). The ratio of the average weight of each driver to
the sum of the average weights of all the drivers gives the corresponding
normalized weight. For example, the normalized weight of ID is
( . , . , . )
( . , . , . )
0 23 0 43 0 63
4 27 6 33 8 93
¤
¦
¥
¥
¥
¥
´
¶
µ
µ
µ
µµ
which is simplifed using equation (4.7) as (0.03, 0.06, 0.15) (see table 9.4).
Suppose that one is interested in evaluating marketing strategies of two dif
ferent reverse supply chain programs. Now that the weights (normalized) of
the drivers of public participation are ready, the two marketing strategies are
linguistically rated by the three representatives with respect to each driver.
Table 9.5 is used for conversion of linguistic ratings into TFNs. Assuming (for
arithmetic simplicity) that the representatives come to a consensus about the
rating of each marketing strategy with respect to each driver, the decision
matrix shown in table 9.6 (S1 and S2 are the marketing strategies) is formed.
For example, the rating (TFN) of marketing strategy S2 with respect to driver
ID is (5, 7, 10) (see table 9.6) because the representatives unanimously rate it
as good with respect to ID (the TFN for the linguistic rating good is (5, 7, 10)
(see table 9.5)).
TABLE 9.3
Average weights of drivers of public participation (frst model)
Driver Average weight
KD (0.3, 0.5, 0.7)
AR (0.5, 0.7, 0.87)
SR (0.3, 0.5, 0.67)
CD (0.57, 0.77, 0.9)
ID (0.23, 0.43, 0.63)
EC (0.5, 0.7, 0.87)
IU (0.23, 0.43, 0.63)
RC (0.17, 0.37, 0.57)
AB (0.17, 0.43, 0.63)
LR (0.57, 0.77, 0.93)
IB (0.43, 0.63, 0.83)
CL (0.3, 0.5, 0.7)
Sum (4.27, 6.73, 8.93)
© 2009 by Taylor & Francis Group, LLC
Evaluation of Marketing Strategies 185
TABLE 9.4
Normalized weights of drivers of public participation (frst model)
Driver Normalized weight
KD (0.03, 0.07, 0.16)
AR (0.06, 0.10, 0.20)
SR (0.03, 0.07, 0.16)
CD (0.06, 0.11, 0.21)
ID (0.03, 0.06, 0.15)
EC (0.07, 0.10, 0.20)
IU (0.03, 0.06, 0.15)
RC (0.02, 0.05, 0.13)
AB (0.02, 0.06, 0.15)
LR (0.06, 0.11, 0.22)
IB (0.05, 0.09, 0.20)
CL (0.03, 0.07, 0.16)
TABLE 9.5
Conversion for ratings of marketing strategies (frst model)
Linguistic rating TFN
Very good (7, 10, 10)
Good (5, 7, 10)
Fair (2, 5, 8)
Poor (0, 3, 5)
Very poor (0, 0, 3)
TABLE 9.6
Decision matrix (frst model)
Driver S1 S2
KD (7, 10, 10) (0, 0, 3)
AR (2, 5, 8) (7, 10, 10)
SR (2, 5, 8) (2, 5, 8)
CD (0, 0, 3) (7, 10, 10)
ID (7, 10, 10) (5, 7, 10)
EC (5, 7, 10) (2, 5, 8)
IU (0, 3, 5) (2, 5, 8)
RC (0, 0, 3) (7, 10, 10)
AB (2, 5, 8) (5, 7, 10)
LR (5, 7, 10) (5, 7, 10)
IB (0, 3, 5) (7, 10, 10)
CL (0, 0, 3) (5, 7, 10)
© 2009 by Taylor & Francis Group, LLC
186 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
Now one is ready to perform the six steps in TOPSIS.
Step 1: Construct the normalized decision matrix. Table 9.7 shows the
normalized decision matrix formed by applying equation (4.49) on
each element of table 9.5. For example, the normalized fuzzy rating
of strategy S2 with respect to driver ID (see table 9.7) is calculated
using equation (4.49) as follows:
r
52
=
( , , )
( , , ) ( , , )
( . , . , .
5 7 10
7 10 10 5 7 10
0 35 0 57 1 1
2 2
66)
Note that equations (4.4), (4.6), and (4.7) are used to perform the basic
operations in the calculation of r
52
above.
Step 2: Construct the weighted normalized decision matrix. Table 9.8 shows
the weighted normalized decision matrix. This is constructed using
the normalized weights of the drivers listed in table 9.6 and the
normalized decision matrix in table 9.7. For example, the weighted
normalized fuzzy rating of strategy S2 with respect to driver ID,
i.e., (0.01, 0.04, 0.17) (see table 9.8), is calculated by multiplying the
normalized weight of ID, i.e., (0.03, 0.07, 0.15) (see table 9.6), with the
normalized fuzzy rating of S2 with respect to ID, i.e., (0.35, 0.57, 1.16)
(see table 9.7). Equation (4.6) is used for the multiplication.
Step 3: Determine the ideal and negativeideal solutions. Each row in the
decision matrix shown in table 9.8 has a maximum rating and a
minimum rating. They are the ideal and negativeideal solutions,
respectively, for the corresponding driver. For arithmetic simplicity,
it is assumed here that the rating with the highest most promising
TABLE 9.7
Normalized decision matrix (frst model)
Driver S1 S2
KD (0.67, 1.00, 1.43) (0.00, 0.00, 0.43)
AR (0.16, 0.45, 1.10) (0.55, 0.89, 1.37)
SR (0.18, 0.71, 2.83) (0.18, 0.71, 2.83)
CD (0.00, 0.00, 0.43) (0.67, 1.00, 1.43)
ID (0.50, 0.82, 1.16) (0.35, 0.57, 1.16)
EC (0.39, 0.81, 1.86) (0.16, 0.58, 1.49)
IU (0.00, 0.51, 2.50) (0.21, 0.86, 4.00)
RC (0.00, 0.00, 0.43) (0.67, 1.00, 1.43)
AB (0.16, 0.58, 1.49) (0.39, 0.81, 1.86)
LR (0.35, 0.71, 1.41) (0.35, 0.71, 1.41)
IB (0.00, 0.29, 0.71) (0.63, 0.96, 1.43)
CL (0.00, 0.00, 0.60) (0.48, 1.00, 2.00)
© 2009 by Taylor & Francis Group, LLC
Evaluation of Marketing Strategies 187
quantity (second parameter in the TFN) is the maximum, and the
rating with the lowest most promising quantity is the minimum. For
example (see table 9.8), with respect to driver ID, the maximum rat
ing is (0.01, 0.05, 0.17) and the minimum rating is (0.01, 0.04, 0.17).
This is because, in the row for that driver, (0.01, 0.05, 0.17) is the TFN
with the highest second parameter and (0.01, 0.04, 0.17) is the TFN
with the lowest second parameter.
Step 4: Calculate the separation distances. The separation distances (see
table 9.9) for each marketing strategy are calculated using equations
(4.52) and (4.53). For example, the positive separation distance for
strategy S2 (see table 9.9) is calculated using the weighted normal
ized fuzzy ratings of S2 (see table 9.8) and the ideal solution (obtained
in step 3) for each driver of public participation.
Step 5: Calculate the relative closeness coeffcient. Using equation (4.54), the
relative closeness coeffcient is calculated for each marketing strat
egy (see table 9.10). For example, the relative closeness coeffcient (i.e.,
0.692) for strategy S2 (see table 9.10) is the ratio of the strategy’s nega
tive separation distance (i.e., 0.215) to the sum (i.e., 0.215 + 0.097 =
0.312) of its negative and positive separation distances (see table 9.9).
Step 6: Rank the preference order. Because the relative closeness coeffcient
of strategy S2 (i.e., 0.692) is much higher than that of strategy S1 (i.e.,
0.308) (see table 9.10), it is evident that S2 is much better than S1.
TABLE 9.8
Weighted normalized decision matrix (frst model)
Driver S1 S2
KD (0.02, 0.07, 0.23) (0.00, 0.00, 0.07)
AR (0.01, 0.05, 0.22) (0.03, 0.09, 0.28)
SR (0.01, 0.05, 0.44) (0.01, 0.05, 0.44)
CD (0.00, 0.00, 0.09) (0.04, 0.11, 0.30)
ID (0.01, 0.05, 0.17) (0.01, 0.04, 0.17)
EC (0.03, 0.08, 0.38) (0.01, 0.06, 0.30)
IU (0.00, 0.03, 0.37) (0.01, 0.06, 0.59)
RC (0.00, 0.00, 0.06) (0.01, 0.05, 0.19)
AB (0.00, 0.04, 0.22) (0.01, 0.05, 0.28)
LR (0.02, 0.08, 0.31) (0.02, 0.08, 0.31)
IB (0.00, 0.03, 0.14) (0.03, 0.09, 0.28)
CL (0.00, 0.00, 0.10) (0.02, 0.07, 0.33)
© 2009 by Taylor & Francis Group, LLC
188 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
9.3 Second Model (Fuzzy Logic, Quality Function
Deployment, and Method of Total Preferences)
In this section, a model to evaluate the marketing strategy planned by a
closedloop supply chain program is presented. Fuzzy logic, quality func
tion deployment (QFD), and method of total preferences are employed in the
approach. Section 9.3.1 presents the performance aspects and their enablers
considered in the use of QFD, and section 9.3.2 gives a numerical example
illustrating the model’s application.
9.3.1 Performance Aspects and Enablers
The following performance aspects of the marketing strategies are consid
ered in the employment of quality function deployment:
1. Program simplicity (PS) (refects the relative ease with which the pub
lic can participate in the program)
2. Manufacturing practices (MP) (refects the production facility’s green
image, innovation, and improvement capability compared to its peers)
3. Government issues (GI) (refects how stringent the local government
regulations are)
4. Incentives (refects the incentives offered by the collection centers and
the local government to the public for participating in the program)
5. Public knowledge (PK) (refects the awareness among the public about
the program being implemented)
The following enablers are considered for the abovelisted performance
aspects:
TABLE 9.9
Separation distances of marketing strategies (frst model)
Marketing strategy Positive distance S* Negative distance S–
S1 0.215 0.097
S2 0.096 0.215
TABLE 9.10
Relative closeness coeffcients of marketing strategies (frst model)
Marketing strategy Relative closeness coeffcient
S1 0.311
S2 0.692
© 2009 by Taylor & Francis Group, LLC
Evaluation of Marketing Strategies 189
1. Program simplicity (PS): Strategic location of collection centers (SL),
effectiveness of collection methods (ECM)
2. Manufacturing practices (MP): Green competency (GC), innovation,
and improvement capability (I&I)
3. Government issues (GI): Regulations (Reg.), company’s level of coop
eration with local governments (Coop)
4. Incentives: Disposal incentives by the local government and the collec
tion centers (DI), incentives for buying remanufactured goods (BI)
5. Public knowledge (PK): Good advertisement of the program being
implemented (AD), socioeconomic status of the society where the
program is implemented (SES)
For the convenience of the reader, the performance aspects and the respec
tive enablers are shown in a fourlevel hierarchy (see fgure 9.1). The frst
level represents the goal (evaluating the marketing strategies), the second
level contains the performance aspects, the third level contains the enablers
of the performance aspects considered in the second level, and the fourth
level contains the candidate marketing strategies (A and B) to be evaluated.
Evaluating
Marketing
Strategv
Program
Simplicitv (PS)
Manufacturing
Practices (MP)
Government
Issues (GI)
Incentives
Public
Knowledge (PK)
Strategic Location of
centers selling used
goods (SL)
Efectiveness of
Collection Methods
(ECM)
Green Competencv
(GC)
Innovation &
Improvement (II)
Regulations
Level of Cooperation
with govt. omcials (co
Op)
Disposal Incentives
(DI)
Incentives for buvers
of used products (BI)
Good Advertisement
of the program
implemented (AD)
SocioEconomic
status of the societv
(SES)
Strategv A
Strategv B
FIGURE 9.1
Performance aspects and enablers of marketing strategies (second model).
© 2009 by Taylor & Francis Group, LLC
190 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
9.3.2 Numerical Example
Suppose that there are three experts (E
1
, E
2
, and E
3
) for giving linguistic val
ues to R
ij
(relationship score for performance aspect i and enabler j) and d
i
(importance value of performance aspect i relative to the other performance
aspects) data. Table 9.11 shows the linguistic scale for R
ij
and d
i
.
Tables 9.12–9.16 show the linguistic relationship scores (R
ij
) as given by
three experts to the enablers of PS, MP, GI, Incentives, and PK, respectively.
Table 9.17 gives the linguistic importance values (d
i
) of PS, MP, GI, Incen
tives, and PK as given by the three experts.
These linguistic relationship scores are converted into TFNs using the con
version scale shown in table 9.11. The TFNs are averaged before defuzzify
ing. For example, consider SL’s linguistic relationship scores given by the
three experts (see table 9.12). The average relationship score for SL, calculated
using equations (4.4) and (4.7), is
7 5 5 7 5
3
10 7 5 10
3
10 10 10
3
. .
,
.
,
¤
¦
¥
¥
¥
¥
´
¶
µ
µ
µ
µ
= (6.67, 9.17, 10)
The defuzzifed relationship score for SL, calculated using equation (4.8), is
8.61. The average relationship scores and the corresponding defuzzifed val
ues for PS, MP, GI, Incentives, and PK are calculated as (6.67, 9.17, 10) (defuzzi
fed value = 8.61), (4.17, 6.67, 9.17) (defuzzifed value = 6.67), (5.83, 8.33, 9.17)
(defuzzifed value = 7.78), (5.83, 8.33, 10) (defuzzifed value = 8.05), and (3.33,
6.83, 8.33) (defuzzifed value = 5.83), respectively.
The ATIRs and RTIRs of the enablers are then calculated using equations
(4.30) and (4.31), respectively (see table 9.18).
Table 9.19 shows the scale for converting linguistic WA
nj
(degree to which
alternative n can deliver enabler j) values given by the three experts into
TFNs (for arithmetic simplicity, we assume consensus among the experts).
The WA
nj
linguistic value and the corresponding defuzzifed TFN (calcu
lated using equation (4.8)) for each alternative are shown in table 9.20.
Using equation (4.32), the TUP value for each alternate marketing strat
egy is calculated. For example, TUP for marketing strategy A is calculated
using the RTIR values from table 9.18 and the defuzzifed WA
nj
values from
table 9.20 as (0.13)×(10) + (0.095)×(7.33) + (0.087)×(10) + (0.077)×(5) + (0.121)×(5)
+ (0.117)×(7.33) + (0.089)×(7.33) + (0.125)×(5) + (0.0821)×(3) + (0.07)×(5) = 6.64.
Finally, using equation (4.33), the NTUP value for each alternative market
ing strategy is calculated. For example, NTUP for marketing strategy A is
calculated as
6 64
6 64 6 62
0 5007
.
. .
.
It is evident that marketing strategy A has more potential than marketing
strategy B, and hence the decision maker would choose strategy A.
© 2009 by Taylor & Francis Group, LLC
Evaluation of Marketing Strategies 191
TABLE 9.11
Linguistic weight conversion table for R
ij
and d
i
(second model)
Linguistic rating TFN
Very strong (VS) (7.5, 10, 10)
Strong (S) (5, 7.5, 10)
Medium (M) (2.5, 5, 7.5)
Weak (W) (0, 2.5, 5)
Very weak (VW) (0, 0, 2.5)
TABLE 9.12
Linguistic relationship scores of PS and its enablers (second model)
E1 E2 E3
SL VS S VS
ECM S M S
TABLE 9.13
Linguistic relationship scores of MP and its enablers (second model)
E1 E2 E3
GC VS VS S
I&I S S VS
TABLE 9.14
Linguistic relationship scores of GI and its enablers (second model)
E1 E2 E3
Reg. S VS M
Coop VS M M
TABLE 9.15
Linguistic relationship scores of incentives and its enablers (second model)
E1 E2 E3
DI VS S VS
BI S M S
TABLE 9.16
Linguistic relationship scores of PK and its enablers (second model)
E1 E2 E3
AD VS M VS
SES M S S
© 2009 by Taylor & Francis Group, LLC
192 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
9.4 Third Model (Fuzzy Logic, Extent Analysis
Method, and Analytic Network Process)
In this section, a model to evaluate the marketing strategy planned by a
closedloop supply chain program is presented. Fuzzy logic and analytic
network process are employed in the approach. The difference between this
model and the second one (see section 9.3) is that interdependencies among
criteria are considered here (unlike in the second model). Section 9.4.1 pres
TABLE 9.17
Linguistic relationship scores of performance aspects (second model)
E1 E2 E3
PS S VS VS
MP S S M
GI VS VS M
Incentives S VS S
PK S M M
TABLE 9.18
ATIRs and RTIRs of enablers (second model)
Enabler ATIR RTIR
SL 74.15 0.13
ECM 52.62 0.10
GC 48.15 0.09
I&I 42.59 0.08
Reg. 66.98 0.12
Coop 64.81 0.12
DI 49.23 0.09
BI 69.37 0.13
AD 45.37 0.08
SES 38.89 0.07
TABLE 9.19
Conversion table for linguistic WA
nj
(second model)
Linguistic rating TFN
Very good (VG) (7, 10, 10)
Good (G) (5, 7, 10)
Fair (F) (2, 5, 8)
Poor (P) (1, 3, 5)
Very poor (VP) (0, 0, 3)
© 2009 by Taylor & Francis Group, LLC
Evaluation of Marketing Strategies 193
ents the main criteria and subcriteria for evaluation, and section 9.4.2 gives a
numerical example illustrating the model’s application.
9.4.1 Main Criteria and Subcriteria
The main criteria and subcriteria are shown in a fourlevel hierarchy (see fg
ure 9.2; differs slightly from fgure 9.1). For a description of these criteria, see
section 9.3.1 (performance aspects and enablers). The frst level represents
the goal (evaluating the marketing strategies), the second level contains the
main criteria, the third level contains the subcriteria of the main criteria con
sidered in the second level, and the fourth level contains the candidate mar
keting strategies (C and D) to be evaluated.
9.4.2 Numerical Example
Suppose that there are three experts to carry out pairwise comparisons among
the main criteria and subcriteria using linguistic weights (high, medium, low,
etc.). These linguistic weights are converted into TFNs using table 9.21.
Table 9.22 illustrates the comparative linguistic weights (H = high, M =
medium, L = low) given to the main criteria. Using fuzzy logic, these lin
guistic weights are converted into TFNs (using table 9.21) and then averaged
to form another TFN called the average weight. For example, the average
weight of “simplicity” criteria with respect to government issues (GI) is
H H M
3
which is
TABLE 9.20
Linguistic and corresponding defuzzifed WA values of marketing strategies
(second model)
Enabler A B
SL VG 10 G 7.33
ECM G 7.33 F 5
GC VG 10 VG 10
I&I F 5 VG 10
Reg F 5 G 7.33
Coop G 7.33 G 7.33
DI G 7.33 F 5
BI F 5 P 3
AD P 3 P 3
SES F 5 VG 10
© 2009 by Taylor & Francis Group, LLC
194 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
0 5 0 5 0 3
3
0 7 0 7 0 5
3
0 9 0 9 0 7
3
. . .
,
. . .
,
. . .
¤
¦
¥
¥
¥
¥¥
´
¶
µ
µ
µ
µ
= (0.43, 0.63, 0.83)
(see table 9.23).
Then the extent analysis method (see equations (4.9)–(4.11)) is applied to the
average weights of the main criteria to calculate the fuzzy synthetic extent
values for the same. By applying equations (4.12)–(4.14) to the fuzzy synthetic
extent values, the weight vectors are obtained for the main criteria, which are
then normalized to get the normalized weights shown in table 9.24.
Evaluating
Marketing
Strategv
Program
Simplicitv (PS)
Manufacturing
Practices (MP)
Government
Issues (GI)
Incentives
Public
Knowledge (PK)
Strategic Location of
centers selling used
goods (SL)
Efectiveness of
Collection Methods
(ECM)
Regularitv of
Collection (RC)
Green Competencv
(GC)
Innovation &
Improvement (II)
Total Qualitv
Programs (TQ)
Regulations
Level of Cooperation
with govt. omcials (co
Op)
Disposal Incentives
(DI)
Incentives for buvers
of used products (BI)
Good Advertisement
of the program
implemented (AD)
SocioEconomic
Status of the Societv
(SES)
Strategv C
Strategv D
FIGURE 9.2
Main criteria and subcriteria (third model).
© 2009 by Taylor & Francis Group, LLC
Evaluation of Marketing Strategies 195
TABLE 9.21
Linguistic weight conversion table for criteria and subcriteria (third model)
Linguistic weight TFN
Very high (VH) (0.7, 0.9, 1.0)
High (H) (0.5, 0.7, 0.9)
Medium (M) (0.3, 0.5, 0.7)
Low (L) (0.1, 0.3, 0.5)
Very low (VL) (0.0, 0.1, 0.3)
TABLE 9.22
Linguistic weights of main criteria (third model)
Criteria PS GI Incentives MP PK
PS (1, 1, 1) (H, H, M) (M, L, M) (L, VL, M) (L, VL, L)
GI 1/(H, H, M) (1, 1, 1) (VH, H, H) (H, M, H) (M, H, M)
Incentives 1/(M, L, M) 1/(VH, H, H) (1, 1, 1) (H, M, H) (M, L, M)
MP 1/(L, VL, M) 1/(H, M, H) 1/(H, M, H) (1, 1, 1) (H, M, H)
PK 1/(L, VL, L) 1/(M, H, M) 1/(M, L, M) 1/(H, M, H) (1, 1, 1)
TABLE 9.23
Average weights of main criteria (third model)
Criteria PS GI Incentives MP PK
PS (1, 1, 1) (0.43, 0.63,
0.83)
(0.23, 0.43,
0.63)
(0.14, 0.3, 0.5) (0.07, 0.23,
0.43)
GI (1.20, 1.58,
2.32)
(1, 1, 1) (0.57, 0.77,
0.63)
(0.43, 0.63,
0.83)
(0.37, 0.57,
0.77)
Incentives (1.58, 2.32,
4.34)
(1.58, 1.29,
1.75)
(1, 1, 1) (0.43, 0.63,
0.83)
(0.23, 0.43,
0.63)
MP (2, 3.33, 7.14) (1.2, 1.58,
2.32)
(1.2, 1.58,
2.32)
(1, 1, 1) (0.43, 0.63,
0.83)
PK (2.32, 4.34,
4.28)
(1.29, 0.07,
2.7)
(1.58, 2.32,
4.34)
(1.2, 1.58,
2.32)
(1, 1, 1)
Sum (8.11, 12.59,
29.1)
(5.52, 4.58,
8.61)
(4.59, 6.11,
8.93)
(3.2, 4.14,
5.58)
(2.1, 2.86,
3.66)
TABLE 9.24
Normalized weights of main criteria (third model)
Criteria Weight
PS 0.015
GI 0.131
Incentives 0.217
MP 0.304
PK 0.331
© 2009 by Taylor & Francis Group, LLC
196 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
In a similar manner, the normalized weight of each subcriterion with
respect to its main criteria is obtained (see table 9.25).
Table 9.26 shows the matrix of interdependencies obtained after carrying
out pairwise comparisons among the subcriteria and the steps involved in
the extent analysis method.
The super matrix M is made to converge to obtain a longterm stable set of
weights. For convergence, M must be made column stochastic, which is done
by raising M to the power of 2
k+1
, where k is an arbitrarily large number; in
the example here, k = 59. Table 9.27 shows the converged super matrix.
TABLE 9.26
Matrix of interdependencies (third model)
SL ECM RC Reg. Coop GC I&I TQ DI BI AD SES
SL 0 0.2 0.08 0 0 0 0 0 0 0 0 0
ECM 0.11 0 0.92 0 0 0 0 0 0 0 0 0
RC 0.889 0.8 0 0 0 0 0 0 0 0 0 0
Regulations 0 0 0 0 0.5 0 0 0 0 0 0 0
Coop 0 0 0 0.5 0 0 0 0 0 0 0 0
GC 0 0 0 0 0 0 0.51 0.259 0 0 0 0
I&I 0 0 0 0 0 0.11 0 0.74 0 0 0 0
TQ 0 0 0 0 0 0.88 0.48 0 0 0 0 0
DI 0 0 0 0 0 0 0 0 0 0.5 0 0
BI 0 0 0 0 0 0 0 0 0.5 0 0 0
AD 0 0 0 0 0 0 0 0 0 0 0 0.5
SES 0 0 0 0 0 0 0 0 0 0 0.5 0
TABLE 9.25
Normalized weights of subcriteria with respect to main criteria
(third model)
Subcriteria Weight
SL 0.37
ECM 0.33
RC 0.30
Regulations 0.12
Coop 0.88
GC 0.80
I&I 0.20
TQ 0.06
DI 0.29
BI 0.65
AD 0.89
SES 0.11
© 2009 by Taylor & Francis Group, LLC
Evaluation of Marketing Strategies 197
Table 9.28 shows the linguistic ratings (and their corresponding TFNs) that
are used to evaluate the marketing strategies with respect to the subcriteria.
Relative ratings of the marketing strategies (C and D) are obtained by
carrying out pairwise comparisons among the marketing strategies with
respect to the subcriteria and applying the steps involved in the extent
analysis method. In the example here, there are seventeen subcriteria that
lead to seventeen such pairwise comparison matrices. Table 9.29 shows the
relative ratings of the marketing strategies.
Using equation (4.2), the desirability indices (DI) for each marketing strat
egy are calculated (see table 9.30).
The overall performance index for each marketing strategy is calculated
by multiplying the desirability indices (table 9.30) of each marketing strategy
by the weights of the criteria (table 9.24) and summing up over all the criteria
and normalizing those indices. Table 9.31 shows the overall weighted indices
for the two marketing strategies.
Strategy D’s overall weighted index is greater; hence, the decision maker
would choose marketing strategy D.
TABLE 9.27
Converged super matrix (third model)
SL 0.121
ECM 0.487
RC 0.45
Regulations 0
Coop 0
GC 0.271
I&I 0.326
TQ 0.401
DI 0
BI 0
AD 0
SES 0
TABLE 9.28
Linguistic weight conversion table for ratings of marketing strategies (third model)
Linguistic weight TFN
Very good (VG) (7, 10, 10)
Good (G) (5, 7, 10)
Fair (F) (2, 5, 8)
Poor (P) (1, 3, 5)
Very poor (VP) (1, 1, 3)
© 2009 by Taylor & Francis Group, LLC
198 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
9.5 Conclusions
In this chapter, three models to evaluate marketing strategies for a reverse/
closed supply chain program are presented. The frst model employs fuzzy
logic and technique for order preference by similarity to ideal solution (TOP
SIS). The second model employs fuzzy logic, quality function deployment
(QFD), and method of total preferences. Finally, the third model uses fuzzy
logic, extent analysis method, and analytic network process. Whereas the
frst model is for a reverse supply chain program, the second and third mod
els are for a closedloop supply chain program. The difference between the
second and third models is that interdependencies among criteria are con
sidered in the third model, but not in the second one.
TABLE 9.29
Relative ratings of marketing strategies with respect to subcriteria (third model)
C D
SL 0.50 0.50
ECM 0.48 0.52
RC 0.50 0.50
Regulations 0.50 0.50
Coop 0.48 0.52
GC 0.48 0.52
I&I 0.48 0.52
TQ 0.49 0.51
DI 0.48 0.52
BI 0.50 0.50
AD 0.49 0.51
SES 0.48 0.52
TABLE 9.30
Desirability indices (third model)
Criterion S1 S2
PS 0.16 0.16
GI 0.00 0.00
Incentives 0.18 0.19
MP 0.00 0.00
PK 0.00 0.00
TABLE 9.31
Overall weighted indices for marketing strategies (third model)
Marketing strategy Overall weighted index
C 0.49
D 0.51
© 2009 by Taylor & Francis Group, LLC
Evaluation of Marketing Strategies 199
References
1. McDonald, S., and Ball, R. 1998. Public participation in plastics recycling
schemes. Resources, Conservation, and Recycling 22:123–41.
2. Williams, I. D., and Kelly, J. 2003. Green waste collection and the public’s recy
cling behavior in the Borough of Wyre England. Resources, Conservation, and
Recycling 38:139–59.
© 2009 by Taylor & Francis Group, LLC
201
10
Evaluation of Production Facilities
10.1 The Issue
In addition to selecting effcient collection centers (see chapter 6), strategic
planning of a closedloop supply chain involves selecting effcient produc
tion facilities where not only new products are produced but also used
products are reprocessed. The various scenarios for evaluating production
facilities for effciency could differ as follows:
1. The decision maker could desire to structure the problem using a
simple hierarchical model, wherein interactions among evaluation
criteria could be ignored.
2. Interactions among evaluation criteria could not be ignored, which
in turn leads to a more complex problem structure.
3. Interactions among evaluation criteria could not be ignored, and the
decision maker could desire to see how close the rating of a candi
date production facility is to the “ideal” solution.
In this chapter, three models to address the above scenarios are presented.
These models evaluate production facilities in terms of both environmen
tal consciousness and potentiality. The frst model addresses scenario 1 and
employs fuzzy logic and technique for order preference by similarity to ideal
solution (TOPSIS). The second model is for scenario 2 and employs fuzzy
logic, extent analysis method, and analytic network process. The third model
is for scenario 3 and employs fuzzy multicriteria analysis method.
This chapter is organized as follows: section 10.2 presents the frst model,
section 10.3 shows the second model, section 10.4 presents the third model,
and section 10.4 gives some conclusions.
© 2009 by Taylor & Francis Group, LLC
202 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
10.2 First Model (Fuzzy Logic and TOPSIS)
The problem of evaluation of production facilities for designing a closedloop
supply chain is framed as a fourlevel hierarchy. The frst level in the hierar
chy contains the objective, i.e., evaluation of the effciency of each candidate
production facility; the second level contains the main criteria for evalua
tion; the third level contains the subcriteria under each main criterion; and
the fourth (last) level contains the candidate production facilities. Figure 10.1
illustrates these hierarchy levels.
valuationofproduction
facilitiesinaclosedloop
supplvchain
EnvironmentallvConscious
Design(ECD)
EnvironmentallvConscious
Manufacturing(ECM)
AttitudeofManagement(AMT)
Potentialitv(POT)
Cost(CO)
Customerervice(CE)
DesignforDisassemblv(DD)
Designforecvcling(DC)
Designforemanufacturing(D)
Energvources(E)
Coolingvstems(C)
WasteManagement(WM)
EnvironmentallvFriendlv/inking
(EF)
FlexibilitvtoHandleUncertainties
(FU)
UsageofAutomatedDisassemblv
vstems(AD)
(/roughput)}(upplv)(TP)}(U)
(/roughput)(Disassemblvtime)
(TP)(DT)
IncrementinQualitvofoods
(QOQI)
FixedCost(FC)
OperationalCost(OC)
IncentivestoCollectionCenters
(IC)
Incentivestocustomers(I)
Utilizationofincentivesfrom
government(U)
Meetingenvironmentalregulations
(E)
A
B
C
D
FIGURE 10.1
Levels of hierarchy.
© 2009 by Taylor & Francis Group, LLC
Evaluation of Production Facilities 203
Section 10.2.1 gives brief descriptions of the main and subcriteria for evalu
ation, and section 10.2.2 illustrates the model using a numerical example.
10.2.1 Evaluation Criteria
The following are brief descriptions of the main criteria on the second level
in the hierarchy and the corresponding subcriteria on the third level (see [1]
for detailed descriptions of some of the criteria).
10.2.1.1 Environmentally Conscious Design (ECD)
ECD is concerned with designing products with certain environmental con
siderations. The following subcriteria fall under ECD.
10.2.1.1.1 Design for Disassembly (DD)
Disassembly is used in both recycling and remanufacturing to increase the
recovery rate by allowing selective separation of parts and materials. Thus,
DD initiatives lead to the correct identifcation of design specifcations to
minimize the complexity of the structure of the product by minimizing the
number of parts, increasing the use of common materials, and choosing a
fastener and joint types that are easily removable.
10.2.1.1.2 Design for Recycling (DC)
DC suggests making better choices for material selection such that the pro
cesses of material selection and material recovery become more effcient.
Some important characteristics of DC are long product life with the mini
mized use of raw materials (source reduction), more adaptable materials for
multiproduct applications, and fewer components within a given material in
an engineered system.
10.2.1.1.3 Design for Remanufacturing (DR)
DR suggests the use of reusable parts and packaging in the design of new
products for source reduction.
10.2.1.2 Environmentally Conscious Manufacturing (ECM)
In addition to environmentally friendly product designs, issues involving
manufacturing must be addressed to have a complete concept of environ
mentally conscious production. These issues (subcriteria on the third level in
the hierarchy) are the following:
Selecting lowpollution energy sources for manufacturing ( r ES)
Designing cooling systems such that the coolant can be reused and r
the heat collected by it can be utilized as an energy source (CS)
Monitoring waste generation as a result of manufacturing ( r WM)
© 2009 by Taylor & Francis Group, LLC
204 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
10.2.1.3 Attitude of Management (AMT)
The attitude of the decision makers (managers) in the facility matters a great
deal when it comes to implementing the above practices (ECD and ECM). All
the managers in the facility must have the following credentials (subcriteria
on the third level in the hierarchy):
Environmentally friendly thoughts ( r EF)
Flexibility to handle uncertainties in the supply and quality of used r
products (FU)
Readiness to usage of automated disassembly systems ( r AD) to avoid
high lead time, expensive labor use, and possible human exposure
to hazardous byproducts
10.2.1.4 Potentiality (POT)
A candidate production facility is also evaluated in terms of its potentiality
to effciently reprocess the incoming used products. The following factors
(subcriteria on the third level in the hierarchy) serve as potentiality measures
(these factors are explained in detail in section 7.2.1):
Throughput ( r TP)/supply (SU)
Throughput ( r TP)×disassembly time (DT)
Quality of reprocessed products ( r QO)–quality of used products (QI)
10.2.1.5 Cost (COS)
Cost incurred by a production facility can be divided into the following
types (subcriteria on the third level in the hierarchy):
Fixed cost ( r FC), which is the sum of space cost, machinery cost, per
sonnel cost, etc.
Operational cost ( r OC), which is the sum of employee salaries, main
tenance cost, etc.
10.2.1.6 Customer Service (CSE)
Customer service basically gives an idea about how well a production facil
ity is:
Giving incentives to the collection centers supplying used products r
(IC)
Giving incentives to the customers buying reprocessed goods ( r IS)
Utilizing incentives provided by the government ( r UG)
Meeting environmental regulations established by the government ( r ER)
© 2009 by Taylor & Francis Group, LLC
Evaluation of Production Facilities 205
Note that the term customer service is used here because any benefciary is a
customer, be it the government, the collection center, or the actual customer
buying reprocessed goods.
10.2.2 Numerical Example
In this example, linguistic (high, medium, etc.) weights are given by three
experts to the main criteria (second level in the hierarchy) and the subcriteria
(third level in the hierarchy). These linguistic weights are quantifed using
triangular fuzzy numbers (TFNs) in table 10.1.
Table 10.2 shows the linguistic weights given by the three experts to the
main criteria. These are quantifed using table 10.1 and then averaged to form
another TFN called the average weight. For example, the average weight of
the main criterion, ECD, is (H + H + M)/3, which is
0 5 0 5 0 3
3
0 7 0 7 0 5
3
0 9 0 9 0 7
3
. . .
,
. . .
,
. . .
¤
¦
¥
¥
¥
¥¥
´
¶
µ
µ
µ
µ
= (0.43, 0.63, 0.83)
(see table 10.2).
The sum of the average weights of all the main criteria is calculated, using
equation (4.4), as (2.29, 3.45, 4.59). The ratio of the average weight of each main
TABLE 10.1
Linguistic weight conversion table for criteria and subcriteria (frst model)
Linguistic weight Triangular fuzzy number
Very high (VH) (0.7, 0.9, 1.0)
High (H) (0.5, 0.7, 0.9)
Medium (M) (0.3, 0.5, 0.7)
Low (L) (0.1, 0.3, 0.5)
Very low (VL) (0.0, 0.1, 0.3)
TABLE 10.2
Relative weights of main criteria (frst model)
Criterion Expert E1 Expert E2 Expert E3 Average weight Relative weight
ECD H H M (0.43, 0.63, 0.83) (0.09, 0.18, 0.36)
ECM VH H VH (0.63, 0.83, 0.97) (0.14, 0.24, 0.42)
AMT L L VL (0.07, 0.23, 0.43) (0.02, 0.07, 0.19)
POT H H H (0.50, 0.70, 0.90) (0.11, 0.20, 0.39)
COS M H H (0.43, 0.63, 0.83) (0.09, 0.18, 0.36)
CSE M L M (0.23, 0.43, 0.63) (0.05, 0.12, 0.28)
© 2009 by Taylor & Francis Group, LLC
206 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
criterion to the sum of the average weights of all the main criteria gives the
corresponding relative weight. For example, the relative weight of ECD is
( . , . , . )
( . , . , . )
0 43 0 63 0 83
2 29 3 45 4 59
¤
¦
¥
¥
¥
¥
´
¶
µ
µ
µ
µµ
which is simplifed, using equation (4.7), as (0.09, 0.18, 0.36) (see table 10.2).
Similarly, linguistic weights, average weights, and relative weights of sub
criteria of ECD, ECM, AMT, POT, COS, and CSE are calculated and shown in
tables 10.3–10.8, respectively.
TABLE 10.3
Relative weights of subcriteria of ECD (frst model)
Subcriterion E1 E2 E3 Average weight Relative weight
DD H H M (0.43, 0.63, 0.83) (0.21, 0.42, 0.89)
DC L L VL (0.07, 0.23, 0.43) (0.03, 0.15, 0.46)
DR M H H (0.43, 0.63, 0.83) (0.21, 0.42, 0.89)
TABLE 10.4
Relative weights of subcriteria of ECM (frst model)
Subcriterion E1 E2 E3 Average weight Relative weight
ES H H H (0.50, 0.70, 0.90) (0.21, 0.40, 0.78)
CS M H H (0.43, 0.63, 0.83) (0.18, 0.36, 0.72)
WM M L M (0.23, 0.43, 0.63) (0.10, 0.24, 0.54)
TABLE 10.5
Relative weights of subcriteria of AMT (frst model)
Subcriterion E1 E2 E3 Average weight Relative weight
EF L L VL (0.07, 0.23, 0.43) (0.04, 0.18, 0.59)
FU M H H (0.43, 0.63, 0.83) (0.23, 0.49, 1.14)
AD M L M (0.23, 0.43, 0.63) (0.12, 0.33, 0.86)
TABLE 10.6
Relative weights of subcriteria of POT (frst model)
Subcriterion E1 E2 E3 Average weight Relative weight
TP/SU H H M (0.43, 0.63, 0.83) (0.21, 0.42, 0.89)
TP×DT L L VL (0.07, 0.23, 0.43) (0.03, 0.15, 0.46)
QO–QI M H H (0.43, 0.63, 0.83) (0.21, 0.42, 0.89)
© 2009 by Taylor & Francis Group, LLC
Evaluation of Production Facilities 207
Because the weights considered in the TOPSIS method must sum to unity,
the weight of each subcriterion on the third level in the hierarchy is multi
plied by the weight of its corresponding main criterion on the second level
in the hierarchy. The weights of the criteria on the third level in the hierar
chy, which are ready for use in the TOPSIS method, are shown in table 10.9.
Table 10.10 shows the linguistic ratings (performance measures) and their
corresponding TFNs, which are used to evaluate the production facilities
with respect to each subcriterion, except TP/SU, TP×DT, QO–QI, FC, and OC.
The reason for this exception is that historical crisp (nonfuzzy) measures of
these subcriteria (viz., TP/SU, TP×DT, QO–QI, FC, and OC) for each produc
tion facility can be easily obtained. Table 10.11 shows the crisp measures that
are considered for these subcriteria for each production facility.
Now one is ready to perform the six steps in the TOPSIS.
Step 1: Construct the normalized decision matrix. Table 10.12 shows the
decision matrix whose elements are the fuzzy ratings of the pro
duction facilities with respect to each subcriterion, as given by the
three experts (for arithmetic simplicity, it is assumed here that the
experts give a consensus rating). For example, the rating of facility
C with respect to subcriterion DD is (1, 3, 5) (see table 10.12) because
the experts unanimously rate it as poor with respect to DD (the TFN
for the linguistic rating “poor” is (1, 3, 5) (see table 10.10)). Also, for
consistency in the TOPSIS, the crisp measures of subcriteria TP/
SU, TP×DT, QO–QI, FC, and OC are converted into TFNs, each of
whose parameters are all equal. For example, crisp measure 10 of
subcriterion FC for facility A is converted into the TFN (10, 10, 10)
(see table 10.12). Table 10.13 shows the normalized decision matrix
formed by applying equation (4.49) to each element of table 10.12. For
example, the normalized fuzzy rating of facility C with respect to
TABLE 10.7
Relative weights of subcriteria of COS (frst model)
Subcriterion E1 E2 E3 Average weight Relative weight
FC H H M (0.43, 0.63, 0.83) (0.34, 0.73, 1.66)
OC L L VL (0.07, 0.23, 0.43) (0.06, 0.27, 0.86)
TABLE 10.8
Relative weights of subcriteria of CSE (frst model)
Subcriterion E1 E2 E3 Average weight Relative weight
IC M H H (0.43, 0.63, 0.83) (0.17, 0.35, 0.78)
IS L L VL (0.07, 0.23, 0.43) (0.03, 0.13, 0.40)
UG H H H (0.50, 0.70, 0.90) (0.19, 0.39, 0.84)
ER L L VL (0.07, 0.23, 0.43) (0.03, 0.13, 0.40)
© 2009 by Taylor & Francis Group, LLC
208 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
TABLE 10.9
Weights of subcriteria for TOPSIS (frst model)
Subcriterion
Weight for TOPSIS (relative weight of
subcriterion × relative weight of
corresponding main criterion)
DD (0.02, 0.08, 0.32)
DC (0.00, 0.03, 0.17)
DR (0.02, 0.08, 0.32)
ES (0.03, 0.10, 0.33)
CS (0.03, 0.09, 0.30)
WM (0.01, 0.06, 0.23)
EF (0.00, 0.01, 0.11)
FU (0.00, 0.03, 0.22)
AD (0.00, 0.02, 0.16)
TP/SU (0.02, 0.08, 0.35)
TP8 × DT (0.00, 0.03, 0.18)
QO–QI (0.02, 0.08, 0.35)
FC (0.03, 0.13, 0.60)
OC (0.00, 0.05, 0.31)
IC (0.01, 0.04, 0.22)
IS (0.00, 0.02, 0.11)
UG (0.01, 0.05, 0.24)
ER (0.00, 0.02, 0.11)
TABLE 10.10
Linguistic rating conversion table for production facilities (frst model)
Linguistic rating Triangular fuzzy number
Very good (VG) (7, 10, 10)
Good (G) (5, 7, 10)
Fair (F) (2, 5, 8)
Poor (P) (1, 3, 5)
Very poor (VP) (0, 0, 3)
TABLE 10.11
Crisp measures of subcriteria for evaluation (frst model)
Subcriterion
Production facilities
A B C D
TP/SU 0.9 0.7 0.9 0.5
TP×DT 25 30 15 40
QO–QI 0.6 0.7 0.3 0.5
FC ($) 100,000 150,000 70,000 200,000
OC ($) 500 300 450 400
© 2009 by Taylor & Francis Group, LLC
Evaluation of Production Facilities 209
subcriterion DD (see table 10.13) is calculated using equation (4.49)
as follows:
r
13
=
( , , )
( , , ) ( , , ) ( , , ) ( , ,
1 3 5
7 10 10 2 5 8 1 3 5 5 7 10
2 2 2
))
( . , . , . )
2
0 06 0 22 0 56
Note that equations (4.4) and (4.7) are used to perform the basic oper
ations in the calculation of r
13
above.
Step 2: Construct the weighted normalized decision matrix. Table 10.14 shows
the weighted normalized decision matrix. This is constructed using
the weights of the subcriteria listed in table 10.9 and the normalized
decision matrix in table 10.13. For example, the weighted normalized
fuzzy rating of facility C with respect to subcriterion DD, i.e., (0, 0.02,
0.18) (see table 10.14), is calculated by multiplying the weight of DD,
i.e., (0.02, 0.08, 0.32) (see table 10.9), by the normalized fuzzy rating
of facility C with respect to DD, i.e., (0.06, 0.22, 0.56) (see table 10.13).
Equation (4.6) is used for the multiplication.
TABLE 10.12
Decision matrix for TOPSIS (frst model)
Subcriterion
Production facilities
A B C D
DD (7, 10, 10) (2, 5, 8) (1, 3, 5) (5, 7, 10)
DC (5, 7, 10) (0, 0, 3) (5, 7, 10) (5, 7, 10)
DR (2, 5, 8) (7, 10, 10) (2, 5, 8) (5, 7, 10)
ES (1, 3, 5) (5, 7, 10) (1, 3, 5) (1, 3, 5)
CS (0, 0, 3) (5, 7, 10) (1, 3, 5) (0, 0, 3)
WM (1, 3, 5) (0, 0, 3) (2, 5, 8) (2, 5, 8)
EF (1, 3, 5) (5, 7, 10) (5, 7, 10) (7, 10, 10)
FU (5, 7, 10) (7, 10, 10) (2, 5, 8) (7, 10, 10)
AD (0, 0, 3) (7, 10, 10) (1, 3, 5) (0, 0, 3)
TP/SU (0.9, 0.9, 0.9) (0.7, 0.7, 0.7) (0.9, 0.9, 0.9) (0.5, 0.5, 0.5)
TP×DT (25, 25, 25) (30, 30, 30) (15, 15, 15) (40, 40, 40)
QO–QI (0.6, 0.6, 0.6) (0.7, 0.7, 0.7) (0.3, 0.3, 0.3) (0.5, 0.5, 0.5)
FC (10, 10, 10) (15, 15, 15) (7, 7, 7) (20, 20, 20)
OC (5, 5, 5) (3, 3, 3) (4.5, 4.5, 4.5) (4, 4, 4)
IC (0, 0, 3) (0, 0, 3) (5, 7, 10) (1, 3, 5)
IS (1, 3, 5) (2, 5, 8) (7, 10, 10) (5, 7, 10)
UG (2, 5, 8) (0, 0, 3) (1, 3, 5) (0, 0, 3)
ER (1, 3, 5) (5, 7, 10) (5, 7, 10) (2, 5, 8)
© 2009 by Taylor & Francis Group, LLC
210 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
Step 3: Determine the ideal and negativeideal solutions. Each row in the
decision matrix shown in table 10.14 has a maximum rating and a
minimum rating. They are the ideal and negativeideal solutions,
respectively, for the corresponding subcriterion. For arithmetic sim
plicity, it is assumed here that the rating with the highest most prom
ising quantity (second parameter in the TFN) is the maximum and
the rating with the lowest most promising quantity is the minimum.
For example (see table 10.14), with respect to subcriterion DD, the
maximum rating is (0.01, 0.06, 0.36) and the minimum rating is (0,
0.02, 0.18). This is because, in the row for that subcriterion, (0.01, 0.06,
0.36) is the TFN with the highest second parameter and (0, 0.02, 0.18)
is the TFN with the lowest second parameter.
Step 4: Calculate the separation distances. The separation distances (see
table 10.15) for each production facility are calculated using equa
tions (4.52) and (4.53). For example, the positive separation distance
for facility C (see table 10.15) is calculated using equation (4.52),
which contains the weighted normalized fuzzy ratings of C (see
table 10.14) and the ideal solution (obtained in step 3) for each subcri
terion. It is important to note that because some TFNs with negative
smallest possible quantities or negative most promising quantities
TABLE 10.13
Normalized decision matrix (frst model)
Subcriterion
Production facilities
A B C D
DD (0.41, 0.74, 1.13) (0.12, 0.37, 0.90) (0.06, 0.22, 0.56) (0.29, 0.52, 1.13)
DC (0.28, 0.58, 1.15) (0, 0, 0.35) (0.28, 0.58, 1.15) (0.28, 0.58, 1.15)
DR (0.11, 0.35, 0.88) (0.39, 0.71, 1.1) (0.11, 0.35, 0.88) (0.28, 0.50, 1.1)
ES (0.11, 0.58, 2.89) (0, 0, 1.73) (0.11, 0.58, 2.89) (0.11, 0.58, 2.89)
CS (0, 0, 0.59) (0.42, 0.92, 1.96) (0.08, 0.39, 0.98) (0, 0, 0.59)
WM (0.08, 0.39, 1.67) (0, 0, 1) (0.16, 0.65, 2.67) (0.16, 0.65, 2.67)
EF (0.06, 0.21, 0.5) (0.28, 0.49, 1) (0.28, 0.49, 1) (0.39, 0.7, 1)
FU (0.26, 0.42, 0.89) (0.37, 0.6, 0.89) (0.1, 0.3, 0.71) (0.37, 0.6, 0.89)
AD (0, 0, 0.42) (0.59, 0.96, 1.41) (0.08, 0.29, 0.71) (0, 0, 0.42)
TP/SU (0.59, 0.59, 0.59) (0.46, 0.46, 0.46) (0.59, 0.59, 0.59) (0.33, 0.33, 0.33)
TP×DT (0.43, 0.43, 0.43) (0.52, 0.52, 0.52) (0.26, 0.26, 0.26) (0.69, 0.69, 0.69)
QO–QI (0.55, 0.55, 0.55) (0.64, 0.64, 0.64) (0.28, 0.28, 0.28) (0.46, 0.46, 0.46)
FC (0.36, 0.36, 0.36) (0.54, 0.54, 0.54) (0.25, 0.25, 0.25) (0.72, 0.72, 0.72)
OC (0.6, 0.6, 0.6) (0.36, 0.36, 0.36) (0.54, 0.54, 0.54) (0.48, 0.48, 0.48)
IC (0, 0, 0.59) (0, 0, 0.59) (0.42, 0.92,1.96) (0.08, 0.39, 0.98)
IS (0.06, 0.22, 0.56) (0.12, 0.37, 0.90) (0.41, 0.74, 1.13) (0.29, 0.52, 1.13)
UG (0.19, 0.86, 3.58) (0, 0, 1.34) (0.1, 0.51, 2.24) (0, 0, 1.34)
ER (0.06, 0.26, 0.67) (0.29, 0.61, 1.35) (0.29, 0.61, 1.35) (0.12, 0.44, 1.08)
© 2009 by Taylor & Francis Group, LLC
Evaluation of Production Facilities 211
are obtained in this step, those TFNs are defuzzifed using equation
(4.8) before squaring them in the process of calculating separation
distances.
Step 5: Calculate the relative closeness to the ideal solution. Using equation
(4.54), the relative closeness coeffcient for each facility in the supply
chain is calculated (see table 10.16). For example, the relative close
ness coeffcient (i.e., 0.592) for facility C (see table 10.16) is the ratio of
the facility’s negative separation distance (i.e., 0.29) to the sum (i.e.,
0.29 + 0.2 = 0.49) of its negative and positive separation distances (see
table 10.15).
Step 6: Rank the preference order. Because the relative closeness coeff
cients of facilities A and C (0.547 and 0.592, respectively) are much
higher than those of facilities B and D (0.394 and 0.434, respectively)
(see table 10.16), it is evident that facilities A and C are much better
than facilities B and D. If the cutoff value of the relative closeness
coeffcient decided by the decision maker is, say, 0.45, he or she will
identify facilities A and C as the effcient ones in the region where
the closedloop supply chain is to be designed.
TABLE 10.14
Weighted normalized decision matrix (frst model)
Subcriterion
Production facilities
A B C D
DD (0.01, 0.06, 0.36) (0, 0.03, 0.29) (0, 0.02, 0.18) (0.01, 0.04, 0.36)
DC (0, 0.02, 0.20) (0, 0, 0.06) (0, 0.02, 0.20) (0, 0.02, 0.20)
DR (0, 0.03, 0.28) (0.01, 0.06, 0.35) (0, 0.03, 0.28) (0.01, 0.04, 0.35)
ES (0, 0.06, 0.95) (0, 0, 0.57) (0, 0.06, 0.95) (0, 0.06, 0.95)
CS (0, 0, 0.18) (0.01, 0.08, 0.59) (0, 0.03, 0.29) (0, 0, 0.18)
WM (0, 0.02, 0.38) (0, 0, 0.23) (0, 0.04, 0.61) (0, 0.04, 0.61)
EF (0, 0, 0.06) (0, 0, 0.11) (0, 0, 0.11) (0, 0.01, 0.11)
FU (0, 0.01, 0.20) (0, 0.02, 0.20) (0, 0.01, 0.16) (0, 0.02, 0.20)
AD (0, 0, 0.07) (0, 0.02, 0.23) (0, 0.01, 0.11) (0, 0, 0.07)
TP/SU (0.01, 0.05, 0.20) (0.01, 0.04, 0.16) (0.01, 0.05, 0.20) (0.01, 0.03, 0.11)
TP×DT (0, 0.01, 0.08) (0, 0.02, 0.09) (0, 0.01, 0.05) (0, 0.02, 0.12)
QO–QI (0.01, 0.04, 0.19) (0.01, 0.05, 0.22) (0.01, 0.02, 0.10) (0.01, 0.04, 0.16)
FC (0.01, 0.05, 0.22) (0.02, 0.07, 0.32) (0.01, 0.03, 0.15) (0.02, 0.09, 0.43)
OC (0, 0.03, 0.18) (0, 0.02, 0.11) (0, 0.03, 0.17) (0, 0.02, 0.15)
IC (0, 0, 0.13) (0, 0, 0.13) (0, 0.04, 0.43) (0, 0.02, 0.22)
IS (0, 0, 0.06) (0, 0.01, 0.10) (0, 0.01, 0.12) (0, 0.01, 0.12)
UG (0, 0.04, 0.86) (0, 0, 0.32) (0, 0.03, 0.54) (0, 0, 0.32)
ER (0, 0.01, 0.07) (0, 0.01, 0.15) (0, 0.01, 0.15) (0, 0.01, 0.12)
© 2009 by Taylor & Francis Group, LLC
212 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
10.3 Second Model (Fuzzy Logic, Extent Analysis
Method, and Analytic Network Process)
In this model, the main criteria and corresponding subcriteria are the same
as those in the frst model (see fgure 10.1); the only difference is that the
IC and IS are combined and simply called incentives. The following numeri
cal example illustrates the use of fuzzy logic, extent analysis method, and
analytic network process in the model. Three experts carry out pairwise
comparisons among the main and subcriteria using linguistic weights (high,
medium, low, etc.). These linguistic weights are converted into TFNs using
table 10.17. Table 10.18 provides the comparative linguistic weights (H = high,
M = medium, L = low) given to the main criteria on the second level of the
hierarchy. Using fuzzy logic, these linguistic weights are converted into
TFNs (using table 10.17) and then averaged to form another TFN called the
average weight. For example, the average weight of criteria ECD with respect
to ECM is
H H M
3
which is
TABLE 10.15
Separation measures of facilities (frst model)
Production facility Positive distance S* Negative distance S–
A 0.239 0.288
B 0.320 0.208
C 0.200 0.290
D 0.305 0.233
TABLE 10.16
Relative closeness coeffcients of production facilities (frst model)
Production facility Relative closeness coeffcient
A 0.547
B 0.394
C 0.592
D 0.434
© 2009 by Taylor & Francis Group, LLC
Evaluation of Production Facilities 213
0 5 0 5 0 3
3
0 7 0 7 0 5
3
0 9 0 9 0 7
3
. . .
,
. . .
,
. . .
¤
¦
¥
¥
¥
¥¥
´
¶
µ
µ
µ
µ
= (0.43, 0.63, 0.83)
(see table 10.19).
The steps of the extent analysis method are applied to the average weights
to get the normalized weight vectors of main criteria. For example, consider
the normalized weight vector of main criterion ECD shown in table 10.20;
applying equations (4.9)–(4.11) to the average weights in table 10.19, the syn
thetic extent value of ECD is calculated as (0.052, 0.12, 0.181). By applying
equations (4.12)–(4.14) to the synthetic extent values, the weight vectors are
obtained for the main criteria, which are then normalized to get the normal
ized weight vectors shown in table 10.20.
Similarly, table 10.21 shows normalized weight vectors of each subcrite
rion with respect to its main criterion, obtained after carrying out pairwise
comparisons among subcriteria with respect to their main criteria and then
applying the steps of the extent analysis method.
Table 10.22 shows the matrix of interdependencies obtained after carrying
out pairwise comparisons among the subcriteria and carrying out the steps
involved in the extent analysis method.
The super matrix M is made to converge to obtain a longterm stable set of
weights. For convergence, M must be made column stochastic, which is done
TABLE 10.17
Linguistic weight conversion table for criteria and subcriteria (second model)
Linguistic weight TFN
Very high (VH) (0.7, 0.9, 1.0)
High (H) (0.5, 0.7, 0.9)
Medium (M) (0.3, 0.5, 0.7)
Low (L) (0.1, 0.3, 0.5)
Very low (VL) (0.0, 0.1, 0.3)
TABLE 10.18
Linguistic weights of main criteria (second model)
Criterion ECD ECM AMT POT COS CSE
ECD (1, 1, 1) (H, H, M) (M, L, M) (L, VL, M) (L, VL, L) (M, L, L)
ECM 1/(H, H, M) (1, 1, 1) (VH, H, H) (H, M, H) (M, H, M) (H, VH, M)
AMT 1/(M, L, M) 1/(VH, H, H) (1, 1, 1) (H, M, H) (M, L, M) (M, H, L)
POT 1/(L, VL, M) 1/(H, M, H) 1/(H, M, H) (1, 1, 1) (H, M, H) (M, L, H)
COS 1/(L, VL, L) 1/(M, H, M) 1/(M, L, M) 1/(H, M, H) (1, 1, 1) (H, M, H)
CSE 1/(M, L, L) 1/(H, VH, M) 1/(M, H, L) 1/(M, L, H) 1/(H, M, H) (1, 1, 1)
© 2009 by Taylor & Francis Group, LLC
214 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
by raising M to the power of 2
k+1
, where k is an arbitrarily large number; in
this example, k = 59. Table 10.23 shows the converged super matrix (
A
kj
I
).
Table 10.24 shows the linguistic ratings and their corresponding TFNs, which
are used to evaluate the production facilities with respect to each subcriterion.
Relative weights of the production facilities are obtained by carrying out
pairwise comparisons among the production facilities with respect to the
subcriteria and applying the steps involved in the extent analysis method.
Table 10.25 shows the relative weights of the production facilities (
S
ikj
).
Using equation (4.2) of analytic network process (ANP), the desirabil
ity indices (DI) for each production facility are calculated and shown in
table 10.26.
The overall performance index for each production facility is calculated
by multiplying the desirability index (table 10.26) of each production facility
for each criterion by the weight of the criterion (table 10.20), summing over
TABLE 10.19
Average weights of main criteria (second model)
Criterion ECD ECM AMT POT COS CSE
ECD (1, 1, 1) (0.43, 0.63,
0.83)
(0.23, 0.43,
0.63)
(0.14, 0.3,
0.5)
(0.07, 0.23,
0.43)
(0.17, 0.37,
0.57)
ECM (1.20, 1.58,
2.32)
(1, 1, 1) (0.57, 0.77,
0.63)
(0.43, 0.63,
0.83)
(0.37, 0.57,
0.77)
(0.5, 0.7,
0.57)
AMT (1.58, 2.32,
4.34)
(1.58, 1.29,
1.75)
(1, 1, 1) (0.43, 0.63,
0.83)
(0.23, 0.43,
0.63)
(0.3, 0.5,
0.7)
POT (2, 3.33,
7.14)
(1.2, 1.58, 2.32) (1.2, 1.58,
2.32)
(1, 1, 1) (0.43, 0.63,
0.83)
(0.3, 0.5,
0.7)
COS (2.32, 4.34,
4.28)
(1.29, 0.07, 2.7) (1.58, 2.32,
4.34)
(1.2, 1.58,
2.32)
(1, 1, 1) (0.43, 0.63,
0.83)
CSE (1.75, 2.7,
5.88)
(1.75, 1.42, 2) (1.42, 2,
3.33)
(1.42, 2,
3.33)
(1.2, 1.58,
2.32)
(1, 1, 1)
Sum (9.87, 15.29,
34.9)
(7.27,6.01,10.6) (6.02, 8.11,
12.2)
(4.63, 6.14,
8.81)
(3.3, 4.44,
5.98)
(2.7, 3.7,
4.37)
TABLE 10.20
Weights of main criteria (P
i
) (second model)
Criterion Weight
ECD 0.008
ECM 0.093
AMT 0.159
POT 0.223
COS 0.253
CSE 0.260
© 2009 by Taylor & Francis Group, LLC
Evaluation of Production Facilities 215
all the criteria, and normalizing those weighted sums. Table 10.27 shows the
overall performance indices for the four production facilities.
Facility S’s overall performance index is the largest; hence, it is the best of
the lot.
10.4 Third Model (Fuzzy Multicriteria Analysis Method)
In this model, the main criteria and corresponding subcriteria are the same
as those in the second model. The model is illustrated using a numerical
example, as follows. Table 10.28 shows the TFNs used for making qualitative
assessments (pairwise comparisons). Table 10.29 shows the pairwise com
parisons among the main criteria with respect to the goal of evaluation of
production facilities using the membership functions shown in table 10.28.
By applying the extent analysis method on the matrix of pairwise compari
sons (table 10.29), the corresponding weights of the main criteria are calcu
lated and shown in table 10.30.
Similarly, pairwise comparisons are carried out among the candidate
production facilities with respect to the subcriteria, and the extent analysis
method is applied on each of those matrices to derive the corresponding
TABLE 10.21
Weights of subcriteria with respect to main criteria ( A
kj
D
) (second model)
Subcriterion Weight
DD 0.365
DC 0.333
DR 0.302
ES 0.057
CS 0.290
WM 0.653
EF 0.067
FU 0.335
AD 0.598
TP/SU 0.136
TP×DT 0.334
QO–QI 0.527
FC 0.121
OC 0.879
Incentives 0.097
UG 0.477
ER 0.427
2
1
6
S
t
r
a
t
e
g
i
c
P
l
a
n
n
i
n
g
M
o
d
e
l
s
f
o
r
R
e
v
e
r
s
e
a
n
d
C
l
o
s
e
d

L
o
o
p
S
u
p
p
l
y
C
h
a
i
n
s
TABLE 10.22
Matrix of interdependencies (super matrix M) (second model)
DD DC DR ES CS WM EF FU AD TP/SU TP×DT QO–QI FC OC Incentives UG ER
DD 0 0.2 0.07 0 0 0 0 0 0 0 0 0 0 0 0 0 0
DC 0.11 0 0.92 0 0 0 0 0 0 0 0 0 0 0 0 0 0
DR 0.88 0.79 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
ES 0 0 0 0 0.11 0.14 0 0 0 0 0 0 0 0 0 0 0
CS 0 0 0 0.51 0 0.85 0 0 0 0 0 0 0 0 0 0 0
WM 0 0 0 0.48 0.88 0 0 0 0 0 0 0 0 0 0 0 0
EF 0 0 0 0 0 0 0 0.5 0.33 0 0 0 0 0 0 0 0
FU 0 0 0 0 0 0 0.51 0 0.66 0 0 0 0 0 0 0 0
AD 0 0 0 0 0 0 0.48 0.46 0 0 0 0 0 0 0 0 0
TP/SU 0 0 0 0 0 0 0 0 0 0 0.5 0.25 0 0 0 0 0
TP×DT 0 0 0 0 0 0 0 0 0 0.11 00.4 0.74 0 0 0 0 0
QO–QI 0 0 0 0 0 0 0 0 0 0.88 0 0 0 0 0 0 0
FC 0 0 0 0 0 0 0 0 0 0 0 0 0 0.5 0 0 0
OC 0 0 0 0 0 0 0 0 0 0 0 0 0.5 0 0 0 0
Incentives 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0.2 0.33
UG 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0.11 0 0.66
ER 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0.88 0.79 0
© 2009 by Taylor & Francis Group, LLC
© 2009 by Taylor & Francis Group, LLC
Evaluation of Production Facilities 217
fuzzy weights. In this example, there are seventeen subcriteria that result
in seventeen such pairwise comparisons among the candidate production
facilities. For example, tables 10.31 and 10.32 show the pairwise compari
sons among the candidate production facilities with respect to the subcrite
ria design for disassembly (DD) and design for recycling (DC), respectively.
Table 10.33 shows the decision matrix (X), the matrix showing the perfor
mance of each candidate production facility with respect to the subcriteria.
These weights are obtained by applying the extent analysis method on each
of the pairwise comparison matrices of candidate production facilities with
respect to the subcriteria.
Pairwise comparisons are carried out among the subcriteria with respect
to their main criteria. The fuzzy weights of the subcriteria with respect to the
TABLE 10.23
Converged super matrix (second model)
Subcriterion Weight
DD 0.121
DC 0.427
DR 0.45
ES 0.11
CS 0.44
WM 0.44
EF 0.29
FU 0.37
AD 0.32
TP/SU 0.27
TP×DT 0.32
QO–QI 0.4
FC 7.89E31
OC 7.89E31
Incentives 0.21
UG 0.32
ER 0.45
TABLE 10.24
Linguistic weight conversion table for production facilities (second model)
Linguistic weight TFN
Very good (VG) (7, 10, 10)
Good (G) (5, 7, 10)
Fair (F) (2, 5, 8)
Poor (P) (1, 3, 5)
Very poor (VP) (1, 1, 3)
© 2009 by Taylor & Francis Group, LLC
218 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
TABLE 10.25
Relative weights of production facilities with respect to subcriteria (second model)
Subcriteria/
production
facilities P Q R S
DD 0.26 0.26 0.24 0.24
DC 0.26 0.25 0.24 0.24
DR 0.26 0.26 0.24 0.24
ES 0.26 0.26 0.24 0.24
CS 0.26 0.25 0.25 0.24
WM 0.26 0.26 0.24 0.24
EF 0.26 0.25 0.25 0.24
FU 0.26 0.26 0.23 0.25
AD 0.26 0.25 0.25 0.24
TP/SU 0.26 0.25 0.25 0.24
TP×DT 0.25 0.27 0.24 0.24
QO–QI 0.26 0.25 0.25 0.24
FC 0.26 0.23 0.26 0.25
OC 0.26 0.23 0.26 0.25
Incentives 0.26 0.25 0.25 0.24
UG 0.26 0.25 0.24 0.25
ER 0.26 0.25 0.25 0.24
TABLE 10.26
Desirability indices (DI) (second model)
Criterion P Q R S
ECD 0.09 0.08 0.08 0.08
ECM 0.11 0.11 0.10 0.10
AMT 0.09 0.09 0.08 0.08
POT 0.09 0.09 0.09 0.15
COS 0.00 0.00 0.00 0.00
CSE 0.10 0.09 0.09 0.09
TABLE 10.27
Overall performance indices for production facilities (second model)
Production facility Overall performance index
P 0.26
Q 0.20
R 0.19
S 0.35
© 2009 by Taylor & Francis Group, LLC
Evaluation of Production Facilities 219
TABLE 10.28
Triangular fuzzy numbers used for qualitative assessments (third
model)
Fuzzy number Membership function
1 (1, 1, 3)
x (x – 2, x, x + 2) for x = 3, 5, 7
9 (7, 9, 11)
TABLE 10.29
Pairwise comparisons among main criteria (third model)
ECD ECM AMT POT COS CSE
ECD (1, 1, 3) (1, 3, 5) (7, 9, 11) (5, 7, 9) (1, 3, 5) (1, 3, 5)
ECM 1/(1, 3, 5) (1, 1, 3) (1, 3, 5) (7, 9, 11) (1, 1, 3) (5, 7, 9)
AMT 1/(7, 9, 11) 1/(1, 3, 5) (1, 1, 3) (3, 5, 7) (5, 7, 9) (7, 9, 11)
POT 1/(5, 7, 9) 1/(7, 9, 11) 1/(3, 5, 7) (1, 1, 3) (7, 9, 11) (1, 3, 5)
COS 1/(1, 3, 5) 1/(1, 1, 3) 1/(5, 7, 9) 1/(7, 9, 11) (1, 1, 3) (5, 7, 9)
CSE 1/(1, 3, 5) 1/(5, 7, 9) 1/(7, 9, 11) 1/(1, 3, 5) 1/(5, 7, 9) (1, 1, 3)
TABLE 10.30
Fuzzy weights of main criteria (third model)
Criterion TFN
ECD (0.12, 0.29, 0.62)
ECM (0.11, 0.23, 0.52)
AMT (0.12, 0.25, 0.51)
POT (0.05, 0.13, 0.29)
COS (0.03, 0.08, 0.20)
CSE (0.01, 0.02, 0.09)
TABLE 10.31
Pairwise comparisons among production facilities with respect to DD (third
model)
P1 P1 P3 P4
P1 (1, 1, 3) (3, 5, 7) (1, 3, 5) (5, 7, 9)
P2 1/(3, 5, 7) (1, 1, 3) (5, 7, 9) (3, 5, 7)
P3 1/(1, 3, 5) 1/(5, 7, 9) (1, 1, 3) (1, 3, 5)
P4 1/(5, 7, 9) 1/(3, 5, 7) 1/(1, 3, 5) (1, 1, 3)
© 2009 by Taylor & Francis Group, LLC
220 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
main criteria are obtained by applying the extent analysis method to those
pairwise comparison matrices. Table 10.34 shows the fuzzy weights of the
subcriteria with respect to their main criteria.
Table 10.35 shows the fuzzy reciprocal judgment matrix (W) that is obtained
by multiplying the fuzzy weights of the subcriteria by the corresponding
main criteria weights.
A fuzzy performance matrix Z representing the overall performance of all
candidate production facilities with respect to each criterion is obtained by
multiplying the weight vector (fuzzy reciprocal judgment matrix, W) by the
decision matrix, X. Table 10.36 shows the fuzzy performance matrix, Z.
An interval performance matrix is derived by using an Bcut on the per
formance matrix, Z, where 0 ≤ B ≤ 1. The value of B represents the decision
TABLE 10.32
Pairwise comparisons among production facilities with respect to DC
(third model)
P1 P1 P3 P4
P1 (1, 1, 3) (1, 3, 5) (5, 7, 9) (7, 9, 11)
P2 1/(1, 3, 5) (1, 1, 3) (3, 5, 7) (5, 7, 9)
P3 1/(5, 7, 9) 1/(3, 5, 7) (1, 1, 3) (1, 3, 5)
P4 1/(7, 9, 11) 1/(5, 7, 9) 1/(1, 3, 5) (1, 1, 3)
TABLE 10.33
Fuzzy decision matrix (X) (third model)
P1 P2 P3 P4
DD (0.17, 0.45, 1.04) (0.16, 0.37, 0.84) (0.04, 0.12, 0.4) (0.025, 0.04, 0.19)
DC (0.22, 0.5, 1.04) (0.15, 0.33, 0.74) (0.03, 0.11, 0.31) (0.02, 0.04, 0.16)
DR (0.12, 0.34, 0.76) (0.17, 0.36, 0.74) (0.13, 0.25, 0.52) (0.02, 0.03, 0.12)
ES (0.17, 0.42, 0.88) (0.16, 0.35, 0.77) (0.07, 0.16, 0.41) (0.02, 0.04, 0.17)
CS (0.14, 0.39, 0.96) (0.12, 0.31, 0.75) (0.1, 0.23, 0.54) (0.02, 0.05, 0.22)
WM (0.38, 0.82, 1.64) (0.35, 0.78, 1.45) (0.2, 0.47, 0.91) (0.04, 0.07, 0.25)
EF (0.38, 0.96, 2.08) (0.22, 0.60, 1.44) (0.13, 0.33, 0.84) (0.04, 0.08, 0.35)
FU (0.19, 0.45, 0.96) (0.15, 0.33, 0.72) (0.06, 0.15, 0.38) (0.02, 0.04, 0.17)
AD (0.18, 0.41, 0.84) (0.14, 0.3, 0.65) (0.12, 0.24, 0.47) (0.02, 0.03, 0.11)
TP/SU (0.38, 0.82, 1.64) (0.35, 0.78, 1.45) (0.2, 0.47, 0.91) (0.04, 0.07, 0.25)
TP×DT (0.17, 0.45, 1.04) (0.16, 0.37, 0.84) (0.04, 0.12, 0.4) (0.02, 0.04, 0.19)
QO–QI (0.17, 0.45, 1.04) (0.16, 0.37, 0.84) (0.04, 0.12, 0.4) (0.02, 0.04, 0.19)
FC (0.12, 0.32, 0.71) (0.2, 0.39, 0.75) (0.12, 0.24, 0.49) (0.02, 0.03, 0.11)
OC (0.22, 0.46, 0.91) (0.14, 0.3, 0.62) (0.09, 0.19, 0.4) (0.022, 0.03, 0.12)
Incentives (0.75, 2.22, 4.66) (0.53, 1.55, 3.37) (0.62, 1.45, 2.95) (0.1, 0.22, 0.84)
UG (0.22, 0.46, 0.91) (0.14, 0.3, 0.62) (0.09, 0.19, 0.4) (0.022, 0.03, 0.12)
ER (0.17, 0.45, 1.04) (0.16, 0.37, 0.84) (0.04, 0.12, 0.4) (0.02, 0.04, 0.19)
© 2009 by Taylor & Francis Group, LLC
Evaluation of Production Facilities 221
TABLE 10.34
Fuzzy weights of subcriteria with respect to main criteria
(third model)
Subcriterion Fuzzy weight
DD (0.14, 0.43, 1.03)
DC (0.24, 0.49, 0.99)
DR (0.03, 0.06, 0.28)
ES (0.16, 0.53, 1.43)
CS (0.13, 0.37, 0.98)
WM (0.04, 0.09, 0.41)
EF (0.28, 0.56, 1.16)
FU (0.12, 0.31, 0.69)
AD (0.04, 0.11, 0.29)
TP/SU (0.15, 0.48, 1.2)
TP×DT (0.19, 0.44, 1.04)
QO–QI (0.039, 0.07, 0.28)
FC (0.3, 0.83, 1.94)
OC (0.085, 0.16, 0.64)
Incentives (0.15, 0.48, 1.2)
UG (0.19, 0.44, 1.04)
ER (0.03, 0.07, 0.28)
TABLE 10.35
Fuzzy reciprocal judgment matrix (W) (third model)
Subcriterion Fuzzy weight
DD (0.017, 0.12, 0.64)
DC (0.02, 0.14, 0.615)
DR (0.004, 0.019, 0.177)
ES (0.01, 0.12, 0.74)
CS (0.015, 0.08, 0.514)
WM (0.005, 0.02, 0.215)
EF (0.03, 0.14, 0.59)
FU (0.015, 0.07, 0.35)
AD (0.005, 0.028, 0.151)
TP/SU (0.008, 0.06, 0.34)
TP×DT (0.01, 0.05, 0.3)
QO–QI (0.002, 0.009, 0.08)
FC (0.01, 0.06, 0.39)
OC (0.002, 0.01, 0.13)
Incentives (0.002, 0.011, 0.028)
UG (0.002, 0.01, 0.024)
ER (0.0005, 0.0001, 0.0006)
© 2009 by Taylor & Francis Group, LLC
222 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
maker’s degree of confdence in his or her fuzzy assessments regarding the
production facility ratings and criteria weights. The larger the value of B, the
more confdent the decision maker is about the fuzzy assessments, i.e., the
assessments are closer to the most possible value a
2
of the triangular fuzzy
number (a
1
, a
2
, a
3
). In this example, B = 0.5. For example, consider the fuzzy
performance rating of candidate production facility P1 (shown in bold in
table 10.36) with respect to subcriterion DD. The Bcut on this performance
rating can be performed as
(( . ) ,( . ) ) ( . , . )
. .
0 003 0 674 0 055 0 82
0 5 0 5
. Table 10.37 shows
the interval performance matrix, Z
B
, obtained by performing an Bcut on the
matrix shown in table 10.36.
An overall crisp performance matrix that incorporates the decision maker’s
attitude toward risk, using an optimism index M (M = 1 implies the decision
maker has an optimistic view, 0 implies a pessimistic view, and 0.5 implies
a moderate view), is calculated using equations (4.22) and (4.23). Table 10.38
shows the crisp performance matrix at an optimism index M = 0.5. For example,
consider the Bcut on the performance rating (shown in bold in table 10.37) of
TABLE 10.36
Fuzzy performance matrix (Z) (third model)
P1 P1 P3 P4
DD (0.003, 0.056, 0.674) (0.002, 0.04, 0.54) (0.0007, 0.015, 0.25) (0.0004, 0.005, 0.12)
DC (0.006, 0.07, 0.64) (0.004, 0.04, 0.45) (0.001, 0.01, 0.195) (0.0006, 0.005, 0.09)
DR (0.0005, 0.006, 0.13) (0.0008, 0.007,
0.13)
(0.0006, 0.005, 0.09) (0.00009, 0.0007,
0.022)
ES (0.003, 0.05, 0.65) (0.003, 0.04, 0.57) (0.001, 0.02, 0.31) (0.0004, 0.005, 0.12)
CS (0.002, 0.03, 0.49) (0.001, 0.02, 0.38) (0.001, 0.02, 0.281) (0.004, 0.004, 0.116)
WM (0.001, 0.017, 0.35) (0.001, 0.016,
0.312)
(0.001, 0.01, 0.19) (0.0002, 0.001, 0.05)
EF (0.012, 0.135, 1.23) (0.007, 0.08, 0.85) (0.004, 0.04, 0.5) (0.001, 0.01, 0.212)
FU (0.003, 0.03, 0.34) (0.002, 0.02, 0.25) (0.001, 0.01, 0.136) (0.0003, 0.003, 0.06)
AD (0.001, 0.01, 0.12) (0.0008, 0.008,
0.09)
(0.0007, 0.006, 0.07) (0.0001, 0.0009,
0.017)
TP/SU (0.003, 0.03, 0.57) (0.002, 0.04, 0.5) (0.001, 0.02, 0.31) (0.0003, 0.004, 0.08)
TP×DT (0.001, 0.02, 0.31) (0.001, 0.02, 0.25) (0.0004, 0.007, 0.12) (0.0002, 0.002, 0.35)
QO–QI (0.0003, 0.0004,
0.08)
(0.0003, 0.003,
0.06)
(0.00008, 0.001,
0.032)
(0.00005, 0.0004,
0.016)
FC (0.001, 0.022, 0.28) (0.002, 0.02, 0.29) (0.001, 0.01, 0.19) (0.0002, 0.002, 0.04)
OC (0.0006, 0.006, 0.11) (0.0004, 0.004,
0.08)
(0.0002, 0.002, 0.05) (0.00006, 0.0005,
0.01)
Incentives (0.001, 0.02, 0.13) (0.001, 0.017, 0.09) (0.001, 0.016, 0.082) (0.0002, 0.002, 0.02)
UG (0.0005, 0.004,
0.022)
(0.0003, 0.003,
0.015)
(0.0002, 0.002,
0.009)
(0.00005, 0.0003,
0.003)
ER (0.00008, 0.0007,
0.006)
((0.00008, 0.0006,
0.005)
(0.00002, 0.0002,
0.002)
(0.00001, 0.00007,
0.001)
© 2009 by Taylor & Francis Group, LLC
Evaluation of Production Facilities 223
TABLE 10.37
Interval performance matrix, Z
B
(third model)
P1 P2 P3 P4
DD (0.055, 0.82) (0.05, 0.73) (0.02, 0.5) (0.02, 0.35)
DC (0.08, 0.8) (0.06, 0.67) (0.032, 0.44) (0.025, 0.31)
DR (0.024, 0.368) (0.028, 0.36) (0.024, 0.3) (0.009, 0.15)
ES (0.05, 0.81) (0.05, 0.76) (0.03, 0.55) (0.02, 0.36)
CS (0.04, 0.7) (0.044, 0.62) (0.04, 0.53) (0.02, 0.34)
WM (0.043, 0.596) (0.04, 0.55) (0.03, 0.44) (0.014, 0.233)
EF (0.11, 1.11) (0.08, 0.92) (0.06, 0.7) (0.03, 0.46)
FU (0.055, 0.58) (0.04, 0.504) (0.032, 0.369) (0.019, 0.247)
AD (0.032, 0.357) (0.028, 0.313) (0.026, 0.267) (0.01, 0.131)
TP/SU (0.05, 0.75) (0.05, 0.7) (0.04, 0.56) (0.02, 0.29)
TP×DT (0.042, 0.561) (0.041, 0.503) (0.02, 0.347) (0.016, 0.243)
QO–QI (0.019, 0.292) (0.018, 0.262) (0.009, 0.181) (0.007, 0.127)
FC (0.03, 0.53) (0.04, 0.54) (0.03, 0.43) (0.01, 0.21)
OC (0.02, 0.345) (0.02, 0.285) (0.017, 0.23) (0.008, 0.128)
Incentives (0.03, 0.36) (0.03, 0.3) (0.03, 0.28) (0.014, 0.15)
UG (0.023, 0.148) (0.018, 0.123) (0.015, 0.099) (0.007, 0.055)
ER (0.009, 0.083) (0.009, 0.07) (0.004, 0.051) (0.003, 0.036)
TABLE 10.38
Crisp performance matrix (third model)
P1 P2 P3 P4
DD 0.57 0.54 0.44 0.37
DC 0.36 0.42 0.54 0.59
DR 0.38 0.39 0.36 0.24
ES 0.57 0.55 0.47 0.37
CS 0.53 0.50 0.47 0.36
WM 0.49 0.48 0.42 0.30
EF 0.70 0.63 0.55 0.44
FU 0.50 0.47 0.39 0.32
AD 0.39 0.36 0.34 0.23
TP/SU 0.55 0.54 0.48 0.34
TP×DT 0.48 0.46 0.37 0.31
QO–QI 0.34 0.32 0.26 0.22
FC 0.46 0.48 0.43 0.29
OC 0.37 0.34 0.31 0.22
Incentives 0.40 0.37 0.36 0.26
UG 0.27 0.24 0.22 0.16
ER 0.19 0.18 0.15 0.13
© 2009 by Taylor & Francis Group, LLC
224 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
production facility P1 with respect to subcriterion DD. Using equation (4.23),
the crisp performance rating is ( . . . . ) .
. .
0 5 0 82 0 5 0 055 0 57
0 5 0 5
r r .
The crisp performance matrix is normalized using equation (4.25) to obtain
the normalized performance matrix. Table 10.39 shows the normalized per
formance matrix. For example, the normalized performance score of produc
tion facility P1 (shown in bold in table 10.39) with respect to subcriterion DD
is calculated using equation (4.25) as
( . )
. . . .
.
0 5707
0 5707 0 5442 0 438 0 3713
0 58
2 2 2 2
55
(Note that the values 0.5707, 0.5442, 0.438, and 0.3713 are in the frst row of
table 10.38.)
Using equations (4.26) and (4.27), the positive and negativeideal solutions
are calculated by selecting maximum and minimum ratings across all candi
date production facilities. Table 10.40 shows the positive and negativeideal
solutions. For example, the positive and negativeideal solutions across all
candidate production facilities with respect to DC are the maximum and
minimum values, respectively, from the second row of table 10.39.
By applying the vectormatching function, the degree of similarity (see
table 10.41) between the normalized performance score of each candidate
TABLE 10.39
Normalized performance matrix (third model)
P1 P2 P3 P4
DD 0.585 0.558 0.449 0.381
DC 0.606 0.554 0.434 0.371
DR 0.550 0.557 0.513 0.352
ES 0.540 0.524 0.445 0.354
CS 0.536 0.509 0.474 0.370
WM 0.544 0.528 0.469 0.335
EF 0.593 0.536 0.469 0.374
FU 0.588 0.547 0.464 0.375
AD 0.577 0.540 0.505 0.346
TP/SU 0.571 0.555 0.492 0.352
TP×DT 0.585 0.558 0.449 0.380
QO–QI 0.585 0.558 0.449 0.381
FC 0.546 0.567 0.508 0.349
OC 0.593 0.537 0.484 0.355
Incentives 0.569 0.525 0.517 0.366
UG 0.593 0.537 0.484 0.354
ER 0.586 0.559 0.447 0.380
© 2009 by Taylor & Francis Group, LLC
Evaluation of Production Facilities 225
production facility and the positive and negativeideal solutions can be cal
culated using equation (4.28).
A preferred candidate production facility should have a higher degree of
similarity to the positiveideal solution and a lower degree of similarity to the
negativeideal solution. Hence, an overall performance index for each candi
date facility with the decision maker’s B level of confdence and M degree of
optimism toward risk is determined using equation (4.29). Table 10.42 shows
the overall performance indices of the four candidate production facilities. The
larger the performance index, the better the candidate. For example, for facil
ity P3, the overall performance index is calculated using equation (4.29) as
( . )
( . ) ( . )
.
0 822
0 822 1 290
0 389
TABLE 10.40
Positive and negativeideal solutions (third model)
Subcriterion Positiveideal solution Negativeideal solution
DD 0.585 0.381
DC 0.606 0.371
DR 0.557 0.352
ES 0.540 0.354
CS 0.536 0.370
WM 0.544 0.335
EF 0.593 0.374
FU 0.588 0.375
AD 0.577 0.346
TP/SU 0.571 0.352
TP×DT 0.585 0.380
QO–QI 0.585 0.381
FC 0.567 0.349
OC 0.593 0.355
Incentives 0.569 0.366
UG 0.593 0.354
ER 0.586 0.380
TABLE 10.41
Degree of similarity to positive and negativeideal solutions (third model)
Production facility
Degree of similarity to
positiveideal solution
Degree of similarity to
negativeideal solution
P1 0.997 1.576
P2 0.946 1.495
P3 0.822 1.290
P4 0.646 1.013
© 2009 by Taylor & Francis Group, LLC
226 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
Because production facility P4’s overall performance index is the largest, it
is the best of the lot.
10.5 Conclusions
In this chapter, three models to evaluate production facilities operating in a
region where a closedloop supply chain is to be designed were presented.
These models evaluate production facilities in terms of both environmen
tal consciousness and potentiality. The frst model employs fuzzy logic and
technique for order preference by similarity to ideal solution (TOPSIS); the
second model employs fuzzy logic, extent analysis method, and analytic net
work process; the third model employs fuzzy multicriteria analysis method.
References
1. Gungor, A., and Gupta, S. M. 1999. Issues in environmentally conscious manu
facturing and product recovery: A survey. Computers and Industrial Engineering
36:811–53.
TABLE 10.42
Overall performance indices (third model)
Production facility Overall performance index
P1 0.388
P2 0.387
P3 0.389
P4 0.390
© 2009 by Taylor & Francis Group, LLC
227
11
Evaluation of Futurity of Used Products
11.1 The Issue
A major driver for companies interested in collecting used products is recov
erable value through reprocessing (remanufacturing or recycling). However,
the companies seldom know when those products were bought and why
they were discarded. Also, the products do not indicate their remaining
life periods. Hence, they often undergo partial or complete disassembly for
subsequent reprocessing. The authors are of the opinion that for some used
products, it might make more sense to make necessary repairs to the prod
ucts and sell them on secondhand markets than to disassemble them for sub
sequent reprocessing. To this end, in this chapter, using a numerical example,
it is shown how an expert system can be built using Bayesian updating and
fuzzy logic to decide whether it is sensible to repair a used product of inter
est for subsequent sale on a secondhand market. It is assumed here that the
used product of interest functions improperly; it is obviously sensible to sell
a properly functioning used product on a secondhand market.
Consider the used product shown in fgure 11.1. Given that this product is
not functioning properly, one shall decide whether it is sensible to repair it
for subsequent sale on a secondhand market. Table 11.1 shows the probabil
ity values used to implement Bayesian updating.
Section 11.2 presents the procedure to use fuzzy logic in Bayesian updat
ing. Section 11.3 gives the list of the rules used in Bayesian updating. Section
11.4 presents the implementation of Bayesian updating. Section 11.5 presents
the employment of FLEX shell [1] to build an expert system that can be used
to decide whether it is sensible to repair an improperly functioning used
product of interest for subsequent sale on a secondhand market. Finally, sec
tion 11.6 gives some conclusions.
© 2009 by Taylor & Francis Group, LLC
228 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
S1
S2
C1 C2 C3 C4 C5
R
FIGURE 11.1
Used product.
TABLE 11.1
Probability values used in Bayesian updating
H E P(H) O(H) P(EH) P(E~H) A D
S1 needs repair Product needs
repair
0.60 1.50 1.00 0.60 1.67 0.00
S2 needs repair Product needs
repair
0.70 2.33 1.00 0.40 2.50 0.00
C1 needs repair S1 needs repair 0.45 0.82 1.00 0.45 2.22 0.00
C2 needs repair S1 needs repair 0.55 1.22 1.00 0.30 3.33 0.00
C3 needs repair S1 needs repair 0.30 0.43 1.00 0.55 1.82 0.00
C4 needs repair S2 needs repair 0.32 0.47 1.00 0.70 1.43 0.00
C5 needs repair S2 needs repair 0.10 0.11 1.00 0.80 1.25 0.00
Sensible to repair
product
C1 needs repair 0.60 1.50 0.70 0.20 3.50 0.38
Sensible to repair
product
C2 needs repair 0.60 1.50 0.60 0.30 2.00 0.57
Sensible to repair
product
C3 needs repair 0.60 1.50 0.45 0.60 0.75 1.38
Sensible to repair
product
C4 needs repair 0.60 1.50 0.10 0.75 0.13 3.60
Sensible to repair
product
C5 needs repair 0.60 1.50 0.85 0.40 2.13 0.25
© 2009 by Taylor & Francis Group, LLC
Evaluation of Futurity of Used Products 229
11.2 Usage of Fuzzy Logic
Because it is diffcult for an expert to guess the probabilities shown in
bold (unlike the rest) in table 11.1, fuzzy logic is used to calculate them as
follows:
1. Ask the expert to assign a linguistic rating to P(EH) for each compo
nent in the used product with respect to each of the following factors
(see table 11.2):
a. Is it economical to repair or replace the component (more eco
nomical implies higher rating)?
b. If disposed of, will the component be harmful to the environ
ment (more harmful implies higher rating)?
c. What is the remaining life period of the component (longer life
implies higher rating)?
d. Is the raw material used to make the component depleting
quickly (faster depletion implies higher rating)?
e. Is it diffcult to repair the component (more diffcult implies lower
rating)?
2. Use the data in table 11.3 to convert the linguistic ratings into TFNs.
3. Calculate the average fuzzy P(EH) value for each component.
TABLE 11.2
Linguistic P(EH) ratings
a b c d e
C1 High High Medium Medium High
C2 Very high High Very high Medium Low
C3 Low Low Very low Very high Medium
C4 High High High High Very low
C5 Medium High High High Medium
TABLE 11.3
Conversion table for linguistic P(EH) ratings in Bayesian updating
Linguistic rating Triangular fuzzy number
Very high (VH) (0.7, 0.9, 1.0)
High (H) (0.5, 0.7, 0.9)
Medium (M) (0.3, 0.5, 0.7)
Low (L) (0.1, 0.3, 0.5)
Very low (VL) (0.0, 0.1, 0.3)
© 2009 by Taylor & Francis Group, LLC
230 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
4. Defuzzify the average P(EH) for each component using equation (4.8).
Apply steps 1–4 to calculate appropriate P(E~H) values for each compo
nent. In order to be saved from tedious calculations, it is assumed here that
the values shown in bold in table 11.1 are the defuzzifed average probabili
ties obtained after performing steps 1–4. For the sake of clarity, however, the
calculation procedure is explained using an example below.
Suppose that one wishes to calculate the P(EH) value (numerical) for a
component in a used product. The four steps are implemented as follows:
1. The expert linguistically rates the component with respect to factors
a–e as very high, high, medium, medium, and low, respectively.
2. Using table 11.3, the linguistic ratings are converted into TFNs.
3. The average fuzzy P(EH) is equal to
0 7 0 5 0 3 0 3 0 1
5
0 9 0 7 0 5 0 5 0 3
5
1 . . . . .
,
. . . . .
,
.. . . . . 0 0 9 0 7 0 7 0 5
5
¤
¦
¥
¥
¥
¥
´
¶
µ
µ
µ
µ
,
i.e., (0.38, 0.58, 0.76).
4. Defuzzifying the average fuzzy P(EH) using equation 4.8, one gets
( . . ) ( . . )
.
0 76 0 38 0 58 0 38
3
0 38
= 0.57
11.3 Rules Used in Bayesian Updating
Rule 1: IF product needs repair (AFFIRMS: 1.67; DENIES: 0.00), THEN
S1 needs repair.
Rule 2: IF product needs repair (AFFIRMS: 2.50; DENIES: 0.00), THEN
S2 needs repair.
Rule 3: IF S1 needs repair (AFFIRMS: 2.22; DENIES: 0.00), THEN C1
needs repair.
Rule 4: IF S1 needs repair (AFFIRMS: 3.33; DENIES: 0.00), THEN C2
needs repair.
Rule 5: IF S1 needs repair (AFFIRMS: 1.82; DENIES: 0.00), THEN C3
needs repair.
Rule 6: IF S2 needs repair (AFFIRMS: 1.43; DENIES: 0.00), THEN C4
needs repair.
© 2009 by Taylor & Francis Group, LLC
Evaluation of Futurity of Used Products 231
Rule 7: IF S2 needs repair (AFFIRMS: 1.25; DENIES: 0.00), THEN C5
needs repair.
Rule 8: IF C1 needs repair (AFFIRMS: 3.50; DENIES 0.38) AND C2
needs repair (AFFIRMS: 2.00; DENIES 0.57) AND C3 needs repair
(AFFIRMS: 0.75; DENIES 1.38) AND C4 needs repair (AFFIRMS: 0.13;
DENIES 3.60) AND C5 needs repair (AFFIRMS: 2.13; DENIES 0.25),
THEN it is sensible to repair the product.
11.4 Bayesian Updating
Refer to table 11.1 while reading this section.
Rule 1: H = S1 needs repair; O(H) = 1.50; E = product needs repair; A =
1.67; O(HE) = O(H)×(A) = 2.51.
Rule 2: H = S2 needs repair; O(H) = 2.33; E = product needs repair; A =
2.50; O(HE) = O(H)×(A) = 5.83.
Rule 3: H = C1 needs repair; O(H) = 0.82; E = S1 needs repair; O(E) =
2.51; P(E) = 0.72; A = 2.22; A’ = [2(A – 1)×P(E)] + 2 – A = 1.54; O(HE) =
O(H)×(A’) = (0.82)×(1.54) = 1.26.
Rule 4: H = C2 needs repair; O(H) = 1.22; E = S1 needs repair; O(E) =
2.51; P(E) = 0.72; A = 3.33; A’ = [2(A – 1)×P(E)] + 2 – A = 2.03; O(HE) =
O(H)×(A’) = (1.22)×(2.03) = 2.48.
Rule 5: H = C3 needs repair; O(H) = 0.43; E = S1 needs repair; O(E) =
2.51; P(E) = 0.72; A = 1.82; A’ = [2(A – 1)×P(E)] + 2 – A = 1.36; O(HE) =
O(H)×(A’) = (0.43)×(1.36) = 0.58.
Rule 6: H = C4 needs repair; O(H) = 0.47; E = S2 needs repair; O(E) =
5.83; P(E) = 0.85; A = 1.43; A’ = [2(A – 1)×P(E)] + 2 – A = 1.30; O(HE) =
O(H)×(A’) = (0.47)×(1.30) = 0.61.
Rule 7: H = C5 needs repair; O(H) = 0.11; E = S2 needs repair; O(E) =
5.83; P(E) = 0.85; A = 1.25; A’ = [2(A – 1)×P(E)] + 2 – A = 1.18; O(HE) =
O(H)×(A’) = (0.11)×(1.18) = 0.13.
Rule 8: H = sensible to repair product; O(H) = 1.50;
E1 = C1 needs repair; O(E1) = 1.26; P(E1) = 0.56; A1 = 3.50; A1’ = [2·(A1
– 1)×P(E1)] + 2 – A1 = 1.30;
E2 = C2 needs repair; O(E2) = 2.48; P(E2) = 0.71; A2 = 2.00; A2’ =
[2·(A2 – 1)×P(E2)] + 2 – A2 = 1.42;
E3 = C3 needs repair; O(E3) = 0.58; P(E3) = 0.37; D3 = 1.38; D3’ = [2·(1
– D3)×P(E3)] + D3 = 1.09;
E4 = C4 needs repair; O(E4) = 0.61; P(E4) = 0.38; D4 = 3.60; D4’ = [2·(1
– D4)×P(E4)] + D4 = 1.88;
© 2009 by Taylor & Francis Group, LLC
232 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
E5 = C5 needs repair; O(E5) = 0.13; P(E5) = 0.12; D5 = 0.25; D5’ =
[2×(1 – D5)×P(E5)] + D5 = 0.43;
O(HE1&E2&E3&E4&E5) = O(H)×(A1’)×(A2’)×(D3’)×(D4’)×(D5’) =
(1.50)×(1.30)×(1.42)×(1.09)×(1.88)×(0.43) = 2.44;
P(HE1&E2&E3&E4&E5) = (2.44)/(3.44) = 0.71.
That is, P(sensible to repair the product) = 0.71.
If the cutoff value as decided by the decision maker is, say, 0.55, he will
decide to send the used product for repair and for subsequent sale on a sec
ondhand market.
11.5 FLEXBased Expert System
An excellent tool called FLEX shell [1] is used to build an expert system that
can decide if it is sensible to repair a particular used product for subsequent
sale on the secondhand market. Figure 11.2 shows the user interface for exe
cuting the expert system.
In fgure 11.2, the probability that it is sensible to repair the product is
calculated by the FLEXbased expert system as 0.61. The difference in the
probability obtained manually in section 11.4 and the one obtained by the
expert system in this section is most likely due to the difference in the for
mulae used to calculate A’ and D’ (they are interpolated values). The user of
an expert system shell cannot know how exactly the inference engine of the
shell works. When there is a signifcant difference in the probability values,
it is advisable to build the expert system using a knowledge representation
language like Lisp or Prolog rather than an expert system shell.
11.6 Conclusions
In this chapter, using a numerical example, it is shown how an expert system
can be built using Bayesian updating and fuzzy logic to decide whether it
is sensible to repair an improperly functioning used product of interest for
subsequent sale on a secondhand market instead of disassembling the prod
uct for subsequent reprocessing (remanufacturing or recycling).
© 2009 by Taylor & Francis Group, LLC
Evaluation of Futurity of Used Products 233
References
1. Vasey, P., Westwood, D., and Johns, N. 1996. FLEX expert system toolkit. Logic
Programming Associated Ltd.
FIGURE 11.2
FLEX user interface for executing expert system.
© 2009 by Taylor & Francis Group, LLC
235
12
Selection of New Products
12.1 The Issue
The focus of this issue is to help companies select and produce only those new
products for which revenues in the closedloop supply chain are expected to
be higher than the costs. In this chapter, a costbeneft function is formulated
and then used to perform a multicriteria economic analysis for selecting an
economical new product to produce in a closedloop supply chain. The cost
beneft function can be defned as the ratio of the equivalent value of benefts
associated with the object of interest to the equivalent value of costs associ
ated with the same object. The equivalent value can be present worth, annual
worth, future worth, etc. In this case, the object of interest is the new product
to be produced in a closedloop supply chain. The costbeneft function (F)
is formulated as
F
B
C
(12.1)
where B represents the equivalent value of the benefts (revenues) and C rep
resents the equivalent value of the costs. An F value greater than 1.0 indicates
that the object is economically advantageous. A notable point here is that due
to uncertainties in supply, quality, and disassembly times in the reverse fow
of the product (as a used product) in the closedloop supply chain, decision
makers must rely on experts’ knowledge to obtain imprecise data for calcu
lating B, C, and F values. Hence, fuzzy logic is used in the model presented
in this chapter, and the costbeneft function will hereafter be referred to as
fuzzy costbeneft function.
The fuzzy costbeneft function consists of equivalent values of the follow
ing terms:
New product sale revenue (revenue from selling new products, viz., r
products in the forward fow of the closedloop supply chain)
Reuse revenue (revenue from direct sale/usage in remanufacturing r
of usable components of used products)
© 2009 by Taylor & Francis Group, LLC
236 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
Recycle revenue (revenue from selling material obtained from recy r
cling of unusable components of used products)
New product production cost (cost to produce new products) r
Collection cost (cost to collect used products from consumers) r
Reprocessing cost (cost to remanufacture/recycle used products) r
Disposal cost (cost to dispose of the material left over after remanu r
facturing or recycling of used products)
Lossofsale cost (cost due to loss of sale, which might occur occa r
sionally due to lack of supply of used products)
Investment cost (capital required for facilities and machinery r
involved in production of new products and collection and repro
cessing of used products)
The chapter is organized as follows. In section 12.2, the assumptions made
while formulating the fuzzy costbeneft function are presented. In section
12.3, the nomenclature for the formulation of the fuzzy costbeneft function is
given. In section 12.4, the fuzzy costbeneft function is formulated. In section
12.5, the model (economic analysis) for selecting an economical new prod
uct to produce in a closedloop supply chain is presented. In section 12.6, a
numerical example is given. Finally, section 12.7 provides some conclusions.
12.2 Assumptions
The following assumptions are made while formulating the fuzzy costben
eft function:
1. The product of interest in the reverse fow (as a used product) of the
closedloop supply chain is completely disassembled.
2. All usable components of the product of interest in the reverse fow
(as a used product) will be reused for direct sale or in remanufactur
ing, and all the remaining ones are recycled/disposed of.
12.3 Nomenclature
b
ij
Probability of bad quality (broken, worn out, low perform
ing, etc.) of component j in used product i
C
i
Cost to produce one new product i
CC
i
Total collection cost of used product i per period
CD Cost of reprocessing per unit time
© 2009 by Taylor & Francis Group, LLC
Selection of New Products 237
CF Recycling revenue factor ($/unit weight)
CR
i
Total recycle revenue of used product i per period
CO
i
Cost to collect one used product i
DC
i
Total disposal cost of used product i per period
D
i
Demand for new product i per period
DI
ij
Disposal cost index of component j in used product i (index
scale 0 = lowest, 10 = highest)
DF Disposal cost factor ($/unit weight)
E
ik
Subassembly k in product i
FCB
i
Fuzzy costbeneft function for product i
i Product type
IC
i
Investment cost of product i
j Component type
LC
i
Lossofsale cost of used product i
M
i
Total number of subassemblies in product i
m
ij
Probability of missing component j in used product i
MC
i
Total production cost of new product i per period
N
ij
Multiplicity of component j in product i
RCP
ij
Percentage of recyclable contents by weight in component j of
used product i
RC
i
Total reprocessing cost of used product i per period
RI
ij
Recycling revenue index of component j in used product i
(index scale 0 = lowest, 10 = highest)
Root
i
Root node (for example, outer casing) of product i
RV
ij
Resale value of component j in used product i
SP
i
Selling price of new product i
SR
i
Total new product sale revenue of product i per period
SU
i
Supply of used product i per period
T(Root
i
) Time to disassemble Root
i
T(E
ik
) Time to disassemble subassembly k in used product i
UR
i
Total reuse revenue of used product i per period
W
ij
Weight of component j in used product i
ΔBZ Incremental total revenues (between the challenger and the
defender)
ΔCZ Incremental total costs (between the challenger and the
defender)
© 2009 by Taylor & Francis Group, LLC
238 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
12.4 Formulation of Fuzzy CostBenefit Function
The fuzzy costbeneft function (FCB) of product i of interest consists of equiv
alent values (EV) of nine terms (total new product sale revenue per period
(SR
i
), total reuse revenue per period (UR
i
), total recycle revenue per period
(CR
i
), total new product production cost per period (MC
i
), total collection cost
per period (CC
i
), total reprocessing cost per period (RC
i
), total disposal cost
per period (DC
i
), lossofsale cost (LC
i
), and investment cost (IC
i
)), as follows:
FCB
SR UR CR
MC CC RC
i
i i i
i i i
EV of (
EV of (
)
DD LC IC
i i i
C )
(12.2)
The following subsections explain how the above nine terms are calcu
lated. Some of the terms are modifed versions of those in [1].
12.4.1 Total New Product Sale Revenue per Period (SR)
SR of product i per period is infuenced by the demand for new products
per period (D
i
) and the selling price of each new product (SP
i
). This revenue
equation can be written as follows:
SR D SP
i i i
. (12.3)
Often, in practice, objective data are available to express D
i
and SP
i
as crisp
real numbers. Hence, SR
i
is a crisp real number as well.
12.4.2 Total Reuse Revenue per Period (UR)
UR of product i is infuenced by the fuzzy supply of the product per period
(SU
i
) and the following data of components of each type j in the product: the
resale value (RV
ij
), the number of components (N
ij
), the fuzzy probability of
missing (m
ij
), and the fuzzy probability of bad quality (broken, worn out, low
performing, etc.) (b
ij
). This revenue equation can be written as follows:
UR SU RV N b m
i i ij ij ij ij
j
£
. . .( ) 1 (12.4)
Because SU
i
, b
ij
, and m
ij
are expressed as fuzzy numbers, the resulting UR
i
is
a fuzzy number as well.
© 2009 by Taylor & Francis Group, LLC
Selection of New Products 239
12.4.3 Total Recycle Revenue per Period (CR)
CR of product i is calculated by multiplying the component recycling rev
enue factors by the number of components recycled per period as follows:
CR
SU RI W RCP
N m N b
i
i ij ij ij
ij ij ij i
. . . .
{ ( ) ( 1 1
jj i
j
m
CF
§
©
¨
¨
¨
¨
·
¹
¸
¸
¸
¸
£
)}
. (12.5)
Note that each component has a percentage of recyclable contents (RCP
ij
). RI
ij
is the recycling revenue index (varying in value from 1 to 10) representing
the degree of beneft generated by the recycling of component of type j (the
higher the value of the index, the more proftable it is to recycle the compo
nent), W
ij
is the weight of the component of type j, and CF is the recycling
revenue factor. Because SU
i
, b
ij
, and m
ij
are expressed as fuzzy numbers, the
resulting CR
i
is a fuzzy number as well.
12.4.4 Total New Product Production Cost per Period (MC)
MC of product i is calculated by multiplying the demand for new products
per period (D
i
) by the cost to produce one new product (C
i
), as follows:
MC D C
i i i
. (12.6)
Often, in practice, objective data are available to express D
i
and C
i
as crisp
real numbers. Hence, MC
i
is a crisp real number as well.
12.4.5 Total Collection Cost per Period (CC)
CC of product i is calculated by multiplying the supply of used products per
period (SU
i
) by the cost of collecting one used product from consumers (CO
i
),
as follows:
CC SU CO
i i i
. (12.7)
Because SU
i
is expressed as a fuzzy number, the resulting CC
i
is a fuzzy
number as well.
12.4.6 Total Reprocessing Cost per Period (RC)
RC of product i can be calculated from the supply of used products per
period (SU
i
), disassembly time of the root node (for example, outer casing) of
the product (T(Root
i
)), disassembly time of each subassembly in the product
(T(E
ik
)), the reprocessing cost per unit time (CD), as follows:
© 2009 by Taylor & Francis Group, LLC
240 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
RC SU T Root T E CD
i i i ik
k
M
i
§
©
¨
¨
¨
·
¹
¸
¸
¸
£
. ( ) ( ) .
1
(12.8)
Depending upon the type (vague or objective) of data available for the disas
sembly times, RC
i
is a fuzzy or crisp real number.
12.4.7 Total Disposal Cost per Period (DC)
DC of product i is calculated by multiplying the component disposal cost by
the number of component units disposed per period, as follows:
DC
SU DI W RCP
N m N
i
i ij ij ij
ij ij ij
. . .( ).
{ ( ) (
1
1 11
§
©
¨
¨
¨
¨
·
¹
¸
¸
¸
¸
£
b m
DF
ij ij
j
)
. (12.9)
Note that DI
ij
is the disposal cost index (varying in value from 1 to 10) rep
resenting the degree of nuisance created by the disposal of components of
type j (the higher the value of the index, the more nuisance the component
creates, and hence it costs more to dispose it), W
ij
is the weight of the compo
nent of type j, and DF is the disposal cost factor. Because SU
i
, b
ij
, and m
ij
are
expressed as fuzzy numbers, the resulting CR
i
is a fuzzy number as well.
12.4.8 LossofSale Cost per Period (LC)
LC of product i represents the cost of not meeting its demand in a timely
manner. This occurs because of the unpredictable supply of used products,
as consumers do not discard them in a predictable manner. LC is diffcult to
predict and thus is usually guessed by experts. Due to the involvement of the
experts’ guesses, LC
i
is expressed as a fuzzy number.
12.4.9 Investment Cost (IC)
IC of product i is the capital required for facilities and machinery involved
in production of new products and collection and reprocessing of used
products. Depending upon the type (vague or objective) of data available
for the product and the location of the facilities, IC
i
is a fuzzy or crisp real
number.
© 2009 by Taylor & Francis Group, LLC
Selection of New Products 241
12.5 Model
In order to select the most economical new product to produce in a closedloop
supply chain, from a set of candidate products, the following steps are used:
Step 1: Eliminate every candidate product whose FCB is less than 1.0.
Step 2: Assign the candidate product that has the lowest IC as the
defender and the product with the nextlowest IC as the challenger.
Step 3: Calculate the ratio of the EV of incremental total revenue ΔBZ
(between the challenger and the defender) to the EV of incremen
tal total cost ΔCZ (between the challenger and the defender). If the
ratio is less than 1.0, eliminate the challenger. Otherwise, eliminate
the defender.
Step 4: Repeat steps 2 and 3 until only one product (which is the most
economical one in the set) is left.
12.6 Numerical Example
Three different products (products 1–3) whose structures are shown in fg
ures 12.1–12.3, respectively, are considered for the example.
It is assumed that the supplies of all these products are perpetual. Hence,
capitalized worth (CW) [2] is taken as the EV. Therefore, FCB is the ratio of
CW of total revenues to CW of total costs. The data necessary to calculate
FCB of products 1–3 are given in tables 12.1–12.2 and 12.3, respectively.
Also, T(Root
1
) = 2 min; T(Root
2
) = 1.5 min; T(Root
3
) = 1.5 min; T(E
11
) = 9 min;
T(E
21
) = 7 min; T(E
22
) = 8 min; T(E
31
) = 7 min; T(E
32
) = 8 min; SU
1
= (200, 230,
250) products per year; SU
2
= (210, 220, 230) products per year; SU
3
= (600, 650,
700) products per year; CO
1
= $20; CO
2
= $21; CO
3
= $18; IC
1
= $20,000; IC
2
=
$25,000; IC
3
= $30,000; D
1
= 900 products per year; D
2
= 850 products per year;
D
3
= 1,000 products per year; SP
1
= $70; SP
2
= $28; SP
3
= $58; C
1
= $25; C
2
= $30;
C
3
= $28; LC
1
= $(300, 500, 700) per year; LC
2
= $(100, 400, 500) per year; LC
3
=
$(900, 1,000, 1,100) per year; CF = 0.2 $/lb; DF = 0.1 $/lb; and CD = 0.55 $/min.
Upon calculating revenues and costs for each product, one gets FCB
1
=
(2.13, 2.45, 2.88), FCB
2
= (0.77, 0.84, 0.91), and FCB
3
= (1.76, 2.11, 2.68). Defuzzi
fying these numbers using equation (4.8), one gets FCB
1
= 2.48, FCB
2
= 0.94,
and FCB
3
= 2.17. Because FCB
2
is less than 1.0, product 2 is eliminated from
further analysis.
Now, because IC
1
is less than IC
3
, product 1 is considered the defender and
product 3 is considered the challenger. The defuzzifed ratio of CW of ΔBZ
to CW of ΔCZ is now calculated as 1.86, which is greater than 1.0. Hence, the
defender (product 1) is eliminated. Therefore, the remaining product (prod
uct 3) is the most economical new product (among the three products) to
produce in the closedloop supply chain.
© 2009 by Taylor & Francis Group, LLC
242 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
Root
1
E
11
P
14
P
15
P
13
P
11
P
12
FIGURE 12.1
Structure of product 1.
Root
2
P
21
E
21
E
22
P
22
P
23
P
24
P
25
FIGURE 12.2
Structure of product 2.
Root
3
P
31
E
31
E
32
P
32
P
33
P
34
P
35
FIGURE 12.3
Structure of product 3.
© 2009 by Taylor & Francis Group, LLC
Selection of New Products 243
12.7 Conclusions
In this chapter, a costbeneft function is formulated and then used to perform
a multicriteria economic analysis for selecting an economical new product to
produce in a closedloop supply chain. Fuzzy logic is used in formulating
the costbeneft function.
TABLE 12.1
Data of product 1
Component RV
1j
($) N
1
j W
1j
(lb) RI
1
j RCP
1j
DI
1j
b
1j
m
1j
P
11
7.0 3 4.5 5 65% 6 (0.1,0.1, 0.2) (0.3, 0.4, 0.4)
P
12
8.0 4 6.5 5 50% 4 (0.5, 0.6, 0.7) (0.1, 0.2, 0.2)
P
13
9.0 2 7.0 3 75% 4 (0.2, 0.3, 0.4) (0.3, 0.4, 0.4)
P
14
6.9 1 2.7 9 35% 5 (0.2, 0.2, 0.3) (0.1, 0.1, 0.2)
P
15
8.4 5 7.5 6 70% 1 (0.1, 0.1, 0.2) (0.3, 0.4, 0.5)
TABLE 12.2
Data of product 2
Component RV
2j
($) N
2j
W
2j
(lb) RI
2j
RCP
2j
DI
2j
b
2j
m
2j
P
21
1.0 1 3.9 2 40% 3 (0.1, 0.1, 0.2) (0.0, 0.1, 0.1)
P
22
1.5 3 1.5 4 20% 1 (0.1, 0.2, 0.2) (0.0, 0.0, 0.0)
P
23
1.2 7 4.1 1 70% 2 (0.2, 0.3, 0.4) (0.2, 0.2, 0.3)
P
24
2.5 4 3.2 5 90% 4 (0.3, 0.4, 0.5) (0.1, 0.1, 0.2)
P
25
3.1 3 2.0 2 50% 2 (0.3, 0.4, 0.4) (0.1, 0.1, 0.2)
TABLE 12.3
Data of product 3
Component RV
3j
($) N
3j
W
3j
(lb) RI
3j
RCP
3j
DI
3j
b
3j
m
3j
P
31
9.0 2 4.0 9 30% 4 (0.2, 0.3, 0.4) (0.1, 0.2, 0.3)
P
32
8.0 5 5.0 7 60% 3 (0.1, 0.2, 0.2) (0.1, 0.2, 0.2)
P
33
9.0 3 2.0 8 70% 1 (0.3, 0.4, 0.4) (0.1, 0.2, 0.2)
P
34
7.0 2 6.0 9 25% 3 (0.2, 0.3, 0.3) (0.3, 0.3, 0.4)
P
35
7.0 1 5.2 6 50% 2 (0.3, 0.3, 0.4) (0.1, 0.1, 0.2)
© 2009 by Taylor & Francis Group, LLC
244 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
References
1. Veerakamolmal, P., and Gupta, S. M. 1999. Analysis of design effciency for the
disassembly of modular electronic products. Journal of Electronics Manufacturing
9:79–95.
2. Sullivan, W. G., Wicks, E. M., and Luxhoj, J. T. 2003. Engineering economy. 12th
ed. Englewood Cliffs, NJ: Prentice Hall.
© 2009 by Taylor & Francis Group, LLC
245
13
Selection of Secondhand Markets
13.1 The Issue
In chapter 11, an expert system is built using Bayesian updating and fuzzy
logic to decide whether it is sensible to repair a used product of interest for
subsequent sale on a secondhand market. In this chapter, fuzzy logic, qual
ity function deployment (QFD), and method of total preferences are used to
select the market with the most potential to sell a used product in, from a set
of candidate secondhand markets.
This chapter is organized as follows: Section 13.2 gives the performance
aspects and enablers for the application of QFD. Section 13.3, using a numeri
cal example, presents the implementation of the model that uses fuzzy logic,
quality function deployment (QFD), and method of total preferences to select
the market with the most potential to sell a used product in, from a set of can
didate secondhand markets. Finally, section 13.4 gives some conclusions.
13.2 Performance Aspects and Enablers
for Application of QFD
The following are the performance aspects of the secondhand markets,*
which are considered in the application of QFD:
1. Beforesale performance (BSP) (refects the ability to attract new cus
tomers to the secondhand market)
2. Whilesale performance (WSP) (refects the ability to motivate the
customers to buy secondhand products while the customers are in
the secondhand market)
*
It should be noted that a secondhand market here means a “store” where secondhand (i.e.,
used) products are sold, along with new products.
© 2009 by Taylor & Francis Group, LLC
246 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
3. Aftersale performance (ASP) (refects the ability to attract old cus
tomers to the secondhand market)
The following are the enablers considered in the application of QFD:
Good advertisement ( r AD)
High difference of prices between new and secondhand products ( r DP)
Greenness of the sale ( r GS)
Incentives (warranty, service, etc.) ( r IC)
Low average price of products ( r LP)
Good location of sale of secondhand products (placement in front of r
the new ones, etc.) (LS)
Proper maintenance of secondhand products (as proper as it is for r
new products) (MN)
Discounts to returning customers ( r RC)
Good return/exchange policy ( r RP)
Reputation of the store ( r RS)
Variety of secondhand products on the shelves ( r VS)
The enablers for BSP are LP, IC, RS, and AD; those for WSP are DP, GS, IC,
MN, AD, VS, RP, and LS; and those for ASP are RC and AD.
13.3 Selection of Potential Secondhand Markets
A numerical example is used to present the approach as follows. Two sec
ondhand markets are compared, and the one that has more potential than
the other is to be selected. Suppose that there are fve experts (M’s) for giving
linguistic values to R
ij
and d
i
data. Table 13.1 shows the linguistic scale for R
ij
as well as for d
i
.
Tables 13.2–13.4 show the linguistic relationship scores (R
ij
) as given by the
fve experts to the enablers of BSP, WSP, and ASP, respectively.
Table 13.5 gives the linguistic importance values (d
i
) of BSP, WSP, and ASP,
as given by the fve experts.
The ATIRs and RTIRs of the enablers are then calculated using equations
(4.30) and (4.31), respectively. It must be noted that the triangular fuzzy
numbers (TFNs) are averaged before defuzzifying. For example, consider
AD, which is an enabler for three performance aspects, BSP, WSP, and ASP.
The linguistic relationship scores, as given by the fve experts, for AD and
BSP (see table 13.2), for AD and WSP (see table 13.3), and for AD and ASP
(see table 13.4) are converted into TFNs using table 13.1. These TFNs are then
© 2009 by Taylor & Francis Group, LLC
Selection of Secondhand Markets 247
TABLE 13.1
Conversion table for linguistic R
ij
data
Linguistic R
ij
Triangular fuzzy number (TFN)
Very strong (VS) (7.5. 10, 10)
Strong (S) (5, 7.5, 10)
Medium (M) (2.5, 5, 7.5)
Weak (W) (0, 2.5, 5)
Very weak (VW) (0, 0, 2.5)
TABLE 13.2
Linguistic relationship scores of BSP and its enablers
M1 M2 M3 M4 M5
LP VS VS M W S
IC S S M VS S
RS M S VS W M
AD VS VS VS VS VS
TABLE 13.3
Linguistic relationship scores of WSP and its enablers
M1 M2 M3 M4 M5
DP V VS M W S
GS N M VS VS S
IC S S M VS S
MN M S VS W M
AD S S VS M W
VS VS VS VS VS VS
RP S VS M VS S
LS M S VS W N
TABLE 13.4
Linguistic relationship scores of ASP and its enablers
M1 M2 M3 M4 M5
RC VS VS M W S
AD VW S M S VS
© 2009 by Taylor & Francis Group, LLC
248 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
averaged as follows: The average relationship score for AD and BSP, as calcu
lated using equations (4.4) and (4.7), is
7 5 7 5 7 5 7 5 7 5
5
10 10 10 10 10
5
10 10 . . . . .
, ,
110 10 10
5
¤
¦
¥
¥
¥
¥
´
¶
µ
µ
µ
µ
= (7.5, 10, 10)
Similarly, the average relationship scores for AD and WSP and AD and
ASP are calculated as (4, 6, 8.5) and (4, 6, 8), respectively. Defuzzifed average
relationship scores for AD and BSP, AD and WSP, and AD and ASP, as cal
culated using equation (4.8), are 9.17, 6.33, and 6, respectively. The linguistic
importance values (d
i
) of BSP, WSP, and ASP, as given by the fve experts
(see table 13.5), are converted into TFNs using table 13.1 again. The average
importance values for BSP, WSP, and ASP are then calculated as (6.5, 9, 10)
(defuzzifed value = 8.5), (5, 7.5, 9.5) (defuzzifed value = 7.33), and (5.5, 8, 9.5)
(defuzzifed value = 7.67), respectively. ATIR of AD is then calculated using
equation (4.30) as (8.5)×(9.17) + (7.33)×(6.33) + (7.67)×(6) = 212.79. The ATIRs of
all the enablers are shown in table 13.6. RTIR of each enabler is then calcu
TABLE 13.5
Linguistic importance values of performance aspects
M1 M2 M3 M4 M5
BSP VS VS VS S S
WSP S S M S VS
ASP VS M S S VS
TABLE 13.6
ATIRs and RTIRs of enablers
Enabler ATIR RTIR
AD 170.36 0.24
DP 48.89 0.07
GS 46.44 0.06
IC 116.11 0.16
LP 56.67 0.08
LS 36.67 0.05
MN 36.67 0.05
RC 51.11 0.07
RP 42.50 0.06
RS 42.50 0.06
VS 67.22 0.10
Sum 715.14 1.00
© 2009 by Taylor & Francis Group, LLC
Selection of Secondhand Markets 249
lated using equation (4.31). For example, RTIR of AD is the ratio of its ATIR
to the sum of the ATIRs of all the enablers, i.e.,
170 36
715 14
.
.
= 0.24. RTIRs are also
shown in table 13.6.
Table 13.7 shows the scale for converting the linguistic WA
nj
value given by
the fve experts (for arithmetic simplicity, we assume here that there is a con
sensus among the experts). The WA
nj
(linguistic) values and the correspond
ing defuzzifed TFNs (calculated using equation (4.8)) for each secondhand
market are shown in table 13.8.
Then, using equation (4.32), the TUPs for the two secondhand markets
are calculated. For example, the TUP for secondhand market 1 is calculated
using the RTIR
j
values (see table 13.7) and the defuzzifed WA
1j
values (see
table 13.8) as follows: (0.24)×(7.33) + (0.07)×(1) + (0.06)×(3) + (0.16)×(7.33) +
(0.08)×(7.33) + (0.05)×(9) + (0.05)×(5) + (0.07)×(1) + (0.06)×(7.33) + (0.06)×(1) +
(0.09)×(3) = 5.35. Similarly, the TUP for secondhand market 2 is calculated as
6.74.
TABLE 13.7
Conversion table for linguistic WA
nj
data
Linguistic WA
nj
TFN
Very good (VG) (7, 10, 10)
Good (G) (5, 7, 10)
Fair (F) (2, 5, 8)
Poor (P) (1, 3, 5)
Very poor (VP) (0, 0, 3)
TABLE 13.8
Linguistic and corresponding defuzzifed WA values of secondhand markets
Enabler
Market 1 Market 2
Linguistic WA
nj
TFN Linguistic WA
nj
TFN
AD G 7.33 G 7.33
DP VP 1 VG 9
GS P 3 VG 9
IC G 7.33 VG 9
LP G 7.33 G 7.33
LS VG 9 VP 1
MN F 5 P 3
RC VP 1 F 5
RP G 7.33 G 7.33
RS VP 1 VP 1
VS P 3 G 7.33
© 2009 by Taylor & Francis Group, LLC
250 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
Finally, using equation (4.33), NTUPs are calculated for the two secondhand
markets. For example, NTUP for secondhand market 1 is calculated as follows:
5 35
5 35 6 74
.
. .
= 0.44
Similarly, NTUP for secondhand market 2 is calculated as 0.56.
It is evident that secondhand market 2 has more potential than second
hand market 1.
13.4 Conclusions
In this chapter, fuzzy logic, quality function deployment (QFD), and method
of total preferences are used to select the market with the most potential to
sell a used product in, from a set of candidate secondhand markets.
© 2009 by Taylor & Francis Group, LLC
251
14
Design of a Synchronized
Reverse Supply Chain
14.1 The Issue
Effective management of a supply chain requires synchronization among
the internal business processes of the supply chain. Synchronization in a
supply chain means reducing the variability among the internal business
processes or partners such that each stakeholder in the supply chain acts in
a way that is appropriately timed with the actions of the other stakeholders
[1]. The delivery performance of a supply chain is maximized largely by syn
chronizing the internal business processes (reducing variability) such that
the fnal product fts in the customerspecifed delivery window with a very
high probability.
In this chapter, a model consisting of two design experiments that use
the Six Sigma concept to achieve better synchronization in a reverse supply
chain is presented. This model tailors the individual processes in such a way
that the overall delivery performance is maximized.
The chapter is organized as follows: section 14.2 presents the model, and
section 14.3 gives some conclusions.
14.2 Model (Two Design Experiments)
In this section, two design experiments are presented. The frst experiment
(section 14.2.1) determines the range of nominal values (for a Six Sigma deliv
ery performance) for the lead times of different individual processes of a
reverse supply chain. The second experiment (section 14.2.2) determines a
variance pool for the lead times of the different individual processes.
14.2.1 First Experiment (Determination of Nominal Pool)
Six processes of a reverse supply chain are considered:
© 2009 by Taylor & Francis Group, LLC
252 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
1. Procurement: Procurement involves obtaining the used products
from the consumers at the collection centers.
2. Inspection/testing: Inspection/testing involves determining the con
dition of the products collected in order to determine whether to
remanufacture, refurbish, or recycle the product.
3. Disassembly
4. Remanufacture/refurbish
5. Transportation: Involves transporting the remanufactured products
to the markets.
6. Delivery
It is assumed that the supply chain deals only with remanufacturing or
refurbishing of the used products (a thirdparty recycler takes care of the
products meant for recycling). It is also assumed that there is no waiting time
between the six processes. Hence, the nominal value of the reverse supply
chain’s lead time, U
y
, is described as the sum of the nominal values of lead
times of the individual processes:
T T
y i
i
£
1
6
(14.1)
Also [2],
S S S
Y i
i
n
i
i
pki
T
C
2 2
1
3
£
; (14.2)
where T
i
and T
i
are the standard deviations and tolerances of the individual
processes and T
Y
is the overall process standard deviation.
In fnding a nominal pool, the tolerances for the lead times of the individual
processes of the reverse supply chain are given in this experiment, as are the
nominal values of the lead times of some of the processes (in this experiment,
U
2
, U
3
, U
4
). The nominal value of the lead time of the reverse supply chain, as
well as its tolerance, is also known. The problem (see table 14.1) is now to fnd
a range of values for the other nominal values (in this case, U
1
, U
5
, U
6
), so as to
achieve a Six Sigma delivery performance. Note that this has its implications
on the choice of suppliers, carriers, and other logistics providers.
Let U
y
= 100 days, T
y
= 12 days, T
1
= 3 days, T
2
= 4 days, T
3
= 1 day, T
4
= 2 days,
T
5
= 2 days, and T
6
= 1 day. Also, let the nominal values of inspection/test
ing, disassembly, and remanufacturing/refurbishing be 20, 25, and 30 days,
respectively. It is now required to fnd a range of values for the pool of other
nominal values (U
1
+ U
5
+ U
6
) such that the probability of delivery is at least
0.9999966 within the delivery window, which is (U
y
– T
y
, U
y
+ T
y
) = (88, 112).
Suppose that the individual processes are Six Sigma processes, which
implies a C
pk
value of 1.5 for each process. Using equation (14.2), the standard
© 2009 by Taylor & Francis Group, LLC
Design of a Synchronized Reverse Supply Chain 253
deviations (σ) of the individual processes as well as the reverse supply chain
are found to be σ
1
= 0.67, σ
2
= 0.87, σ
3
= 0.22, σ
4
= 0.44, σ
5
= 0.44, σ
6
= 0.22, and
σ
y
= 1.299. For example,
S
1
1
1
3
3
3 1 5
0 67
r
T
C
pk
.
.
Also, from equation (14.1), one can see that (τ
1
+ τ
5
+τ
6
) = τ
y
– (τ
2
+ τ
3
+ U
4
)
= 100 – 75 = 25. From equation (14.2), the overall standard deviation for the
lead times of the three processes (1, 5, and 6) is S S S S
156 1
2
5
2
6
2
0 99945 . .
For the set of these three processes to have Six Sigma performance, a shift of
at most 1.5 standard deviations from the target (= 25) is acceptable. Hence,
as long as the nominal value of the lead time of the set falls in the range (τ
156
± 1.5σ
156
), Six Sigma delivery performance is guaranteed. In other words, if
(U
1
+ τ
5
+ τ
6
) is in the range (23.5, 26.5), Six Sigma delivery performance is
guaranteed. That is, for this range of values, the probability of y to be in the
range (88, 112) is at least 0.9999966. The maximum probability is attained at
25 days.
14.2.2 Second Experiment (Determination of Variance Pool)
In this experiment, the nominal values of the lead times of the individual
processes considered in the reverse supply chain are given, as are the toler
ances for the lead times of some of the processes. The nominal value of the
lead time of the reverse supply chain, as well as its tolerance is also known.
The problem now is to fnd a variance pool that can be distributed across the
individual processes whose tolerances are not known (see table 14.2).
Let τ
y
= 100 days, T
y
= 12 days; τ
1
= 20 days, τ
2
= 5 days, τ
3
= 25 days, τ
4
= 30
days, τ
5
= 12 days, and τ
6
= 8 days. Also, let the tolerance of the lead time of
procurement, disassembly, and remanufacturing be 3 days, 2 days, and 1
day, respectively.
Here, too, it is assumed that the individual processes are Six Sigma pro
cesses, which implies a C
pk
value of 1.5 for each process. Also, a Six Sigma
TABLE 14.1
Finding a nominal pool (frst experiment)
Given:
τ
y
= nominal value (target) of the lead time of the reverse supply chain, y
T
y
= tolerance range of the lead time of the reverse supply chain, y
T
i
(i = 1, 2, …, 6) = tolerances of the lead times of the individual processes
τ
2
, τ
3
, τ
4
= nominal values of lead times of individual processes 2, 3, and 4
To compute:
A range of values for each of the nominal values, τ
1
, τ
5
, τ
6
, over which Six Sigma delivery
performance is guaranteed
© 2009 by Taylor & Francis Group, LLC
254 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
delivery performance implies C
p
for the reverse supply chain = 2 and C
pk
for
the reverse supply chain = 1.5. Using equation (14.2), the standard deviations
of the individual processes whose tolerances are known (procurement, disas
sembly, and remanufacturing, in this experiment) as well as the reverse sup
ply chain are found to be T
1
= 2, T
3
= 0.44, T
4
= 0.22, and T
y
= 2.67. Again, using
equation (14.2), one can get the variance pool ( ), S S S
2
2
5
2
6
2
for the three pro
cesses (inspection/testing, transportation, and delivery) whose tolerances
are unknown (inspection/testing, transportation, and delivery in this exam
ple) as 2.89. This variance pool can now be distributed among the individual
processes based on engineering judgment that ensures a Six Sigma delivery
performance for the reverse supply chain. That is, for this variance pool, the
probability of y to be in the range (88, 112) is at least 0.9999966.
14.3 Conclusions
The delivery performance of a supply chain is maximized by synchroniz
ing the internal business processes such that the fnal product fts in the
customerspecifed delivery window with a very high probability. In this
chapter, a model consisting of two design experiments that use the Six
Sigma concept to achieve better synchronization in a reverse supply chain
was presented. Six individual processes—procurement, inspection/testing,
disassembly, remanufacture/refurbish, transportation, and delivery—were
considered in the supply chain. The model tailors the individual processes
in such a way that the overall delivery performance of the reverse supply
chain is maximized.
TABLE 14.2
Finding a variance pool (second experiment)
Given:
τ
y
= nominal value (target) of the lead time of the reverse supply chain, y
T
y
= tolerance range of the lead time of the reverse supply chain, y
τ
i
(i = 1, 2, …, 6) = nominal values of the lead times of the individual processes
T
1
, T
3
, T
4
= tolerances of lead times of the individual processes 1, 3, and 4
To compute:
A variance pool that can be distributed across the individual processes whose tolerances are
not known
© 2009 by Taylor & Francis Group, LLC
Design of a Synchronized Reverse Supply Chain 255
References
1. Antony, J., Swarnkar, R., and Tiwari, M. K. 2006. Design of synchronized sup
ply chains: A genetic algorithm based Six Sigma constrained approach. Interna
tional Journal of Logistics Systems and Management 2:120–41.
2. Narahari, Y., Viswanadham, N., and Bhattacharya, R. 2000. Design of synchro
nized supply chains: A Six Sigma tolerancing approach. IEEE International Con
ference on Robotics and Automation 1151–56.
© 2009 by Taylor & Francis Group, LLC
257
15
Performance Measurement
15.1 The Issue
In the era of globalization of markets and business process outsourcing,
many frms realize the importance of continuous monitoring of their sup
ply chain’s performance for its effectiveness and effciency [1]. Traditionally,
performance measurement is defned as the process of quantifying the effec
tiveness and effciency of action [2]. In the modern era, performance measure
ment has a far more signifcant role than just quantifcation and accounting.
It provides management with important feedback to monitor performance,
reveal progress, diagnose problems, and enhance transparency among the
several tiers of the supply chain, thus making a phenomenal contribution to
decision making, particularly in redesigning business goals and reengineer
ing processes [3, 4].
Developing the performance measurement systems is a diffcult aspect of
performance measure selection. It involves the methods by which an orga
nization creates its measurement system. Important questions that need to
be addressed include [5]: What to measure? How often to measure? How
are multiple individual measures integrated into a measurement system?
Also, the way each performance aspect is weighed is industry specifc [6].
For example, although customer satisfaction is an indication of the standard
level of service, for different industries, customers look at different mea
sures. By the same token, due to the inherent differences in various aspects
between the forward and reverse supply chains (see chapter 1), the perfor
mance aspects and evaluation techniques used in a forward supply chain
cannot be extended to a reverse/closedloop supply chain.
In this chapter, appropriate performance aspects and their enablers (drivers
of performance aspects) are identifed for a reverse/closedloop supply chain
environment, and a performance measurement model that uses linear physical
programming (LPP) and quality function deployment (QFD) is presented.
The chapter is organized as follows. Section 15.2 describes the application
of LPP to QFD optimization. Section 15.3 enlists the performance aspects
and enablers identifed for a reverse/closedloop supply chain environment
and then illustrates the model using a numerical example. Finally, section
15.4 gives some conclusions.
© 2009 by Taylor & Francis Group, LLC
258 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
15.2 Application of LPP to QFD Optimization
Applying LPP to QFD optimization involves two steps [6]. The frst step
involves collecting information for LPP through the traditional QFD approach
and building the house of quality (HOQ); this step helps the design team
to complete a qualitative analysis of the design problem. The second step
involves mathematical modeling using LPP. The complete process is illus
trated in fgure 15.1.
15.2.1 First Step
The house of quality (HOQ) shows the customer requirements, engineer
ing characteristics, and the competitor’s performance analysis (see fgure
15.2). Some engineering characteristics are interrelated; as a result, changing
Collect customer information
Model the body and roof of the HOQ
Classify the satisfaction level of
enablers of performance aspects
Build the HOQ
Construct the LPP class functions and
compute the LPP weights
Formulate and solve the mathematical
model
FIGURE 15.1
Application of LPP to QFD optimization.
© 2009 by Taylor & Francis Group, LLC
Performance Measurement 259
the value of one engineering characteristic may alter its impact on customer
requirements or on other engineering characteristics. Notice that the roof of
the HOQ shows the interrelationships among the engineering characteris
tics. These relationships are based on the knowledge and experience of the
design team. In this chapter, the notation H
jk
, which denotes the elements of
the correlation matrix, is introduced to describe the correlation between the
jth and kth engineering characteristics. The body of the HOQ shows the rela
tionships between engineering characteristics and customer requirements.
R
ik
denotes the relationship between the ith customer requirement and jth
engineering characteristic. These relationships are expressed on a scale of
1–9, where 1 denotes a weak relation and 9 denotes the strongest relationship.
In order to quantify the impact of the dependencies between the engineering
characteristics on the relationship between the customer requirements, the
model transformation shown in equation (15.1) is used:
R
R
R
i
ij
norm
ik kj
k
n
ij jk
k
n
j
n
£
£ £
G
G
1
1 1
1 2 , , , ...., ; , , ..., m j n 1 2
(15.1)
where R
ij
norm
can be interpreted as the incremental change in the level of ful
fllment of the ith customer requirement when the jth engineering character
istic is satisfed to a certain level; m is the number of customer requirements
and n is the number of engineering characteristics.
Interrelationships
Engineering characteristics
Relationships among customer
requirements and
engineering characteristics
Cost Index
C
u
s
t
o
m
e
r
r
e
q
u
i
r
e
m
e
n
t
s
I
m
p
o
r
t
a
n
c
e
w
e
i
g
h
t
C
o
m
p
e
t
i
t
o
r
s
p
e
r
f
o
r
m
a
n
c
e
a
n
a
l
y
s
i
s
FIGURE 15.2
House of Quality
© 2009 by Taylor & Francis Group, LLC
260 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
Using LPP, the satisfaction level of each customer requirement is classi
fed in one of the six different ranges: ideal, desirable, tolerable, undesirable,
highly undesirable, and unacceptable. Each objective is described by one of
the eight subclasses, four soft and four hard.
The roof of the HOQ contains the interrelationships between the engi
neering characteristics, denoted by H
jk
, and the body represents the inter
relationships between the customer requirements and the engineering
characteristics, denoted by R
ij
norm
.
The bottom of the HOQ shows the cost index for each engineering char
acteristic. In most cases, improvements in engineering characteristics
will result in an increase in cost. X
j
(j = 1, 2, …, n) is defned as the value
of engineering characteristic j, and the maximum value of the engineering
characteristic is max {X
j
}. Then the normalized value of engineering charac
teristic j is defned as
x
j
= X
j
/max {X
j
}, j = 1, 2, …, n (15.2)
Suppose that when c
j
is invested in engineering characteristic j, it can
achieve its maximum value, max {X
j
}; however, because of constraints, in
most cases not all engineering characteristics can achieve their maximum
value. To make engineering characteristic j achieve x
j
, the design team needs
to invest c
j
×x
j
in engineering characteristic j, thus making the cost function a
monotonically increasing one.
15.2.2 Second Step
The aim of the design team would be to attain the highest customer satisfac
tion level while meeting the budget limitations. The LPP problem is formu
lated as follows:
Minimize w d w d
is
ps
is
ps
s i
~ ~
§
©
¨
¨
¨
·
¹
¸
¸
¸
£
2
5
£
1
m
(15.3)
where d
ps
represents the deviations of the customer requirements from their
target values, subject to:
R
ij
norm
j
n
£
1
a a
£
d t R t
i
is i s ij
norm
j
n
i ( ) 1
1
5
and
for all in nclasses1 3 4 1 2 2 5 S S S i n s
sc
, , , , , .., ; , .., ;
(15.4)
© 2009 by Taylor & Francis Group, LLC
Performance Measurement 261
R
ij
norm
j
n
£
1
q q
£
d t and R t
i
is i s ij
norm
j
n
i ( ) 1
1
5
for all in nclasses 2 3 4 1 2 2 5 S S S i n s
sc
, , , , , .., ; , .., ;
(15.5)
B c x
j j
j
n
q
£
1
(15.6)
where B is the maximum available budget;
d d
is is
<
* * 0 0 ; (15.7)
0 1 a a x
j
(15.8)
Note that in the above formulation, total budget is the only system con
straint; however, depending on the decision environment, other constraints,
such as minimum satisfaction level for the ith customer requirement or
maximum achievement level for the jth engineering characteristic, can be
added.
15.3 Reverse/ClosedLoop Supply Chain
Performance Measurement
In this section, performance aspects and their enablers for measuring the
performance of a reverse/closedloop supply chain are identifed, and then
the model that uses LPP and QFD for performance measurement is pre
sented. Note that in the LPP–QFD methodology described in section 15.2,
the engineering characteristics are replaced by performance aspects and the
customer requirements are replaced by enablers.
15.3.1 Performance Aspects and Enablers
The following are the performance aspects considered in the model:
1. Reputation: This aspect deals with the frm’s overall reputation in
the industry. It is driven by enablers such as ontime delivery ratio,
returning customers ratio, and the frm’s green image. Green image
© 2009 by Taylor & Francis Group, LLC
262 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
can be measured in a number of different ways, such as the envi
ronmental expenses incurred, the amount of waste disposed of, or
the fairs/symposiums related to environmentally conscious manu
facturing in which the frm participates. In this model, the amount
of waste disposed of is used to measure the green image.
2. Innovation and improvement: This aspect is driven by enablers such
as the R&D expenses ratio or the number of new products and pro
cesses launched.
3. Public participation: This aspect is a measure of the frm’s market
ing capabilities that may be driven by enablers such as fexibility,
or the frm’s ability to handle uncertainties, aftersales service eff
ciency, and the markets targeted (markets targeted is quantifed as
the percentage of customers who are knowledgeable about the frm’s
product; the more knowledgeable the customers are, the better the
participation will be). Aftersales service effciency can be defned as
the ratio of the number of customers served to the number of cus
tomers seeking service.
4. Facility potentiality: Potentiality of the remanufacturing facilities can
be driven by several enablers, such as the location of the facilities,
the increment in the quality of products (quality of outgoing repro
cessed products–quality of incoming used products), disassembly
time multiplied by throughput, throughput divided by the supply
of used products (see section 7.2.1 for a detailed explanation of these
enablers), and usage of automated disassembly systems. Location can
be quantifed by specifying it as the distance from the nearest major
city, and usage of automated disassembly systems can be quantifed
using the investment on automated disassembly systems.
5. Responsiveness: This aspect refects how well the frm responds to the
everchanging customer specifcations and can be driven by enablers
such as fexibility, the frm’s ability to handle uncertainties, and the
frm’s aftersales service effciency. Flexibility can be quantifed using
the number of different varieties of products the frm manufactures/
remanufactures.
6. Delivery reliability: This aspect refects how well the frm meets the
due dates specifed by the customers and can be driven by enablers
such as the effectiveness of the frm’s master production schedule,
the usage of automated disassembly systems, the supply of used
products, and the quality of used products. The effectiveness of the
frm’s master production schedule can be defned as the ratio of the
number of orders delivered no later than the due date to the total
number of orders delivered.
Other operational aspects, such as resource utilization, cost per operating
hour, and manufacturing lead time, are also considered in the model.
© 2009 by Taylor & Francis Group, LLC
Performance Measurement 263
15.3.2 Numerical Example
In this section, LPP–QFD methodology detailed in section 15.2 is illustrated
using a numerical example. Table 15.1 shows the LPP preference ranges for
the enablers that constitute the righthand side of the HOQ.
Table 15.2 shows the relationships between the performance aspects,
denoted as H
jk
, evaluated on a 1 to 9 scale.
Table 15.3 shows the relationships between the ith enabler and the jth per
formance aspect, denoted as R
ij
.
Using equation (15.1), the relationships between performance aspects and
enablers are normalized and the normalized scores form the body of the
HOQ. Table 15.4 shows the HOQ for the performance measurement. For sim
plicity, the roof has been removed after normalizing R
ij
.
TABLE 15.1
LPP preference ranges for enablers
Enabler
LPP
class Unit Ideal Desirable Tolerable
Highly
undesirable Unacceptable
Ontime delivery
ratio
2S % 100 80 75 65 50
Returning
customers ratio
2S % 100 90 80 75 65
Green image 1S $ 0 1,200 1,500 2,000 2,500
R&D expenses
ratio
4S % 50 45 40 35 30
60 65 75 80 100
New products/
processes
4S Unit 20 15 10 8 5
25 30 35 40 45
Flexibility 2S Unit 65 55 50 40 35
Aftersales service
effciency
2S % 100 85 75 60 50
Markets targeted 2S % 100 80 75 60 55
Throughput ×
Disassembly time
(TP × DT)
2S % 100 90 80 75 65
Throughput/
supply of used
products (TP/SU)
2S % 100 95 85 70 65
Quality of
reprocessed
products–quality
of used products
(QO–QI)
2S % 65 60 50 45 40
Effectiveness of
MPS
2S % 100 85 75 70 65
Facility location 4S Unit 35 30 25 20 15
20 25 35 40 50
Labor cost per hour 1S $ 20 25 35 40 50
Usage of automated
DA systems
4S $ 1,000 750 600 500 250
2,000 2,500 3,500 4,000 5,000
© 2009 by Taylor & Francis Group, LLC
264 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
TABLE 15.2
Relationships between performance aspects (H
jk
)
Performance aspect Reputation I&I PP RES FP DR RU C/Oh MLT
Reputation — 7 8 7 6 8 4 3 7
Innovation &
improvement (I&I)
7 — 6 4 4 3 3 3 2
Public participation
(PP)
8 6 — 7 3 2 2 3 2
Responsiveness
(RES)
7 4 7 — 3 4 3 3 2
Facility potentiality
(FP)
6 4 3 3 — 6 8 8 8
Delivery reliability
(DR)
8 3 2 4 6 — 7 5 5
Resource utilization
(RU)
4 3 2 3 8 7 — 6 8
Cost per operating
hour (C/Oh)
3 3 3 2 8 5 6 — 5
Manufacturing lead
time (MLT)
7 2 2 2 8 5 8 5 —
TABLE 15.3
Relationships between performance aspects and enablers (
R
ij
)
Reputation I&I PP RES FP DR RU C/Oh MLT
Ontime delivery
ratio
7 4 5 4 6 9 6 5 7
Returning
customers ratio
7 5 5 7 5 8 6 4 6
Green image 8 3 4 7 6 4 6 3 3
R&D expenses ratio 4 7 3 5 4 5 5 6 3
New
products/processes
6 7 4 4 5 4 4 3 4
Flexibility 7 4 7 8 6 5 4 3 4
Aftersales service
effciency
8 4 8 8 4 4 3 5 3
Markets targeted 5 3 7 5 3 5 3 5 3
TP×DT 5 3 5 4 8 8 8 7 8
TP/SU 5 3 5 4 8 8 8 7 8
QO–QI 7 3 6 5 9 7 8 7 7
Effectiveness of
MPS
4 2 4 5 5 8 7 5 7
Facility location 5 3 6 5 4 6 5 7 5
Labor cost per hour 5 2 5 4 6 6 6 8 5
Usage of automated
DA systems
5 7 5 4 8 8 8 8 8
P
e
r
f
o
r
m
a
n
c
e
M
e
a
s
u
r
e
m
e
n
t
2
6
5
TABLE 15.4
House of quality for performance measurement
Performance aspects Reputation I&I PP RES FP DR RU C/Oh MLT
Enablers t
1
t
2
t
3
t
4
t
5
Ontime delivery ratio 0.138 0.091 0.087 0.093 0.134 0.114 0.12 0.106 0.114 100 80 75 65 50
Returning customers ratio 0.142 0.092 0.097 0.091 0.131 0.101 0.116 0.881 0.111 100 90 80 75 65
Green image 0.117 0.093 0.096 0.084 0.112 0.102 0.098 0.098 0.113 0 1,200 1,500 2,000 2,500
R&D expenses ratio 0.128 0.075 0.091 0.078 0.12 0.101 0.102 0.093 0.102 50 45 40 35 30
60 65 75 80 100
New products/processes 0.122 0.078 0.09 0.085 0.109 0.095 0.102 0.082 0.098 20 15 10 8 5
25 30 35 40 45
Flexibility 0.128 0.091 0.092 0.085 0.105 0.106 0.1 0.083 0.098 65 55 50 40 35
Aftersales service effciency 0.122 0.083 0.08 0.088 0.109 0.106 0.096 0.104 0.095 100 85 75 60 50
Markets targeted 0.128 0.083 0.075 0.0884 0.115 0.109 0.1 0.102 0.096 100 80 75 60 55
TP×DT 0.13 0.087 0.078 0.082 0.126 0.104 0.116 0.1 0.181 100 90 80 75 65
TP/SU 0.135 0.084 0.08 0.08 0.126 0.109 0.116 0.102 0.11 100 95 85 70 65
QO–QI 0.124 0.087 0.078 0.079 0.12 0.104 0.111 0.1 0.111 65 60 50 45 40
Effectiveness of MPS 0.135 0.088 0.075 0.083 0.131 0.102 0.115 0.102 0.107 100 85 75 70 65
Facility location 0.125 0.076 0.079 0.078 0.126 0.102 0.108 0.088 0.102 35 30 25 20 15
20 25 35 40 50
Labor cost 0.125 0.09 0.082 0.079 0.121 0.108 0.113 0.094 0.112 20 25 35 40 50
Usage of automated DA
systems
0.13 0.094 0.078 0.081 0.124 0.103 0.112 0.097 0.105 1,000 750 600 500 250
Cost index 4,000 2,500 2,000 2,500 2,500 3,000 2,000 2,000 2,000
© 2009 by Taylor & Francis Group, LLC
© 2009 by Taylor & Francis Group, LLC
266 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
TABLE 15.5
Normalized LPP weights of enablers
Enablers Weights
Ontime delivery
ratio
w
~
12
<
= 0.000068 w
~
13
<
= 0.00414 w
~
14
<
= 0.028 w
~
15
<
= 0.967
Returning
customers ratio
w
~
22
<
= 0.000224 w
~
23
<
= 0.0034 w
~
24
<
= 0.034 w
~
25
<
= 0.956
Green image
w
~
32
= 0.000114 w
~
33
= 0.006 w
~
34
= 0.05 w
~
35
= 0.93
R&D expenses
ratio
w
~
42
<
= 0.0002 w
~
43
<
= 0.003 w
~
44
<
= 0.06 w
~
45
<
= 0.93
w
~
42
= 0.001 w
~
43
= 0.0073 w
~
44
= 0.251 w
~
45
= 0.74
New
products/processes
w
~
52
<
= 0.00016 w
~
53
<
= 0.0023 w
~
54
<
= 0.094 w
~
55
<
= 0.9
w
~
52
= 0.00027 w
~
53
= 0.0034 w
~
54
= 0.06 w
~
55
= 0.935
Flexibility
w
~
62
<
= 0.00041 w
~
63
<
= 0.0122 w
~
64
<
= 0.084 w
~
65
<
= 0.902
Aftersales service
effciency
w
~
72
<
= 0.000183 w
~
73
<
= 0.0043 w
~
74
<
= 0.039 w
~
75
<
= 0.956
Markets targeted
w
~
82
<
= 0.00006 w
~
83
<
= 0.00414 w
~
84
<
= 0.0174 w
~
85
<
= 0.97
TP×DT
w
~
92
<
= 0.000274 w
~
93
<
= 0.0039 w
~
94
<
= 0.125 w
~
95
<
= 0.87
TP/SU
w
~
102
<
= 0.00028 w
~
103
<
= 0.0018 w
~
104
<
= 0.019 w
~
105
<
= 0.97
QO–QI
w
~
112
<
= 0.00028 w
~
113
<
= 0.0018 w
~
114
<
= 0.06 w
~
115
<
= 0.935
Effectiveness of
MPS
w
~
122
<
= 0.00009 w
~
123
<
= 0.002 w
~
124
<
= 0.06 w
~
125
<
= 0.935
Facility location
w
~
132
<
= 0.000274 w
~
133
<
= 0.0039 w
~
134
<
= 0.06 w
~
135
<
= 0.93
w
~
132
= 0.00054 w
~
133
= 0.0036 w
~
134
= 0.125 w
~
135
= 0.87
(continued)
© 2009 by Taylor & Francis Group, LLC
Performance Measurement 267
Table 15.5 shows the LPP weights of the enablers.
The model is formulated as detailed in section 15.2.3. Apart from the total
budget constraint, we consider three other constraints (see equations (15.9)–
(15.11)). Equations (15.9) and (15.10) limit the maximum achievement levels of
performance aspects, facility potentiality, and resource utilization to realistic
values of 0.85 and 0.8, respectively. Equation (15.11) sets the minimum level
for the enabler, ontime delivery ratio, to 0.65. It is assumed that the total
budget in this example is 19,000 monetary units.
x
potentiality
a0 85 .
(15.9)
x
sourceUtilization Re
. a0 8
(15.10)
R x
Ontime Delivery Ratio j
norm
j
j ,
.
£
q0 65
(15.11)
The model is solved using LINGO (v8), and tables 15.6 and 15.7 show the
respective results. Table 15.6 shows the budget allocation and the achieve
ment levels of performance aspects, and table 15.7 shows the satisfaction lev
els of the enablers
(=
R x
ij
norm
j
j
£ ).
From table 15.6, it is evident that the satisfaction levels of the enablers are
almost the same. This can be attributed to the fact that they are coordinated
with the preference of LPP that always puts more effort into the aspects that
lag behind.
TABLE 15.5 (continued)
Normalized LPP weights of enablers
Enablers Weights
Labor cost per hour
w
~
142
= 0.0006 w
~
143
= 0.006 w
~
144
= 0.05 w
~
145
= 0.94
Usage of
automated DA
systems
w
~
152
<
= 0.0002 w
~
153
<
= 0.006 w
~
154
<
= 0.155 w
~
155
<
= 0.83
w
~
152
= 0.0005 w
~
153
= 0.003 w
~
154
= 0.125 w
~
155
= 0.87
2
6
8
S
t
r
a
t
e
g
i
c
P
l
a
n
n
i
n
g
M
o
d
e
l
s
f
o
r
R
e
v
e
r
s
e
a
n
d
C
l
o
s
e
d

L
o
o
p
S
u
p
p
l
y
C
h
a
i
n
s
TABLE 15.6
Budget allocation and achievement levels of performance aspects
Reputation I&I
Public
participation Responsiveness
Facility
potentiality
Delivery
reliability
Resource
utilization Cost/Oh
Manufacturing
lead time
Budget allocation 1,272 2,500 2,000 2,500 2,125 2,700 1,600 2,000 1,520
Achievement
level (xj)
0.318 1.0 1.0 1.0 0.85 0.9 0.8 1.0 0.76
© 2009 by Taylor & Francis Group, LLC
© 2009 by Taylor & Francis Group, LLC
Performance Measurement 269
15.4 Conclusions
In this chapter, aspects and enablers (drivers of aspects) for measurement of
performance of a reverse/closedloop supply chain were identifed, and then
a model that uses linear physical programming and quality function deploy
ment for performance measurement was presented.
References
1. Gunasekaran, A., Patel, C. and Tirtiroglu, E. 2001. Performance measures and
metrics in a supply chain environment. International Journal of Operations and
Production Management 21:71–87.
2. Neely, A., Gregory, M., and Platts, K. 1995. Performance measurement system
design: A literature review and research agenda. International Journal of Opera
tions and Production Management 15:80–116.
3. Rolstandas, A. 1995. Performance measurement: A business process benchmarking
approach. New York: Chapman & Hall.
4. Waggoner, D. B., Neely, A. D., and Kennerley, M. P. 1999. The forces that shape
organizational performance measurement systems: An interdisciplinary
review. International Journal of Production Economics 60:53–60.
5. Beamon, M. B. 1999. Measuring supply chain performance. International Journal
of Operations and Production Management 19:275–92.
6. Chan, F. T. S. 2003. Performance measurement in a supply chain. International
Journal of Advanced Manufacturing Technology 21:534–48.
TABLE 15.7
Enabler satisfaction levels
Enabler Satisfaction level (%)
Ontime delivery ratio 82.21
Returning customers ratio 82.18
Green image 77.27
R&D expenses ratio 72.84
New products/processes 73.35
Flexibility 73.74
Aftersales service effciency 73.07
Markets targeted 72.41
TP×DT 81.82
TP/SU 76.42
QO–QI 76.53
Effectiveness of MPS 76.73
Facility location 74.49
Labor cost per hour 76.20
Usage of automated DA systems 74.28
© 2009 by Taylor & Francis Group, LLC
271
16
Conclusions
The growing desire of consumers to acquire the latest technology, both at
home and in the workplace, along with the rapid technological development
of new products, has led to a new environmental problem: waste, viz., used
products that are discarded prematurely. But as the saying goes that behind
every problem lies an opportunity, reprocessing of this waste means
1. Saving natural resources: We conserve land and reduce the need to drill
for oil and dig for minerals by making products using materials and
components obtained from reprocessing instead of virgin materials.
2. Saving energy: It usually takes less energy to make products from
reprocessed materials and components than from virgin materials.
3. Saving clean air and water: Making products from reprocessed mate
rials and components creates less air pollution and water pollution
than from virgin materials.
4. Saving landfll space: When reprocessed materials and components
are used to make a product, they do not go into landflls.
5. Saving money: It costs much less to make products from reprocessed
materials and components than from virgin materials.
Besides the above opportunities, an important driver for companies to
engage in reprocessing is the enforcement of environmental regulations by
local governments.
A reverse supply chain consists of a series of activities required to collect
used products from consumers and reprocess them (used products) to either
recover their leftover market values or properly dispose of them. Today, in
practice, it has become common for companies involved in a forward sup
ply chain (series of activities required to produce new products from virgin
materials and distribute the former to consumers) to also carry out collec
tion and reprocessing of used products. This combined practice of forward
and reverse supply chains is called a closedloop supply chain. In the past
decade, there has been an explosive growth of reverse and closedloop sup
ply chains in both scope and scale.
Strategic planning (also called designing) primarily involves the structur
ing (which products should be processed/produced in which facilities) of
a supply chain over the next several years. It is longrange planning and is
typically performed every few years when a supply chain needs to expand
© 2009 by Taylor & Francis Group, LLC
272 Strategic Planning Models for Reverse and ClosedLoop Supply Chains
its capabilities. The issues faced by strategic planners of reverse and closed
loop supply chains are:
Selection of used products r
Evaluation of collection centers r
Evaluation of recovery facilities r
Optimization of transportation of goods r
Evaluation of marketing strategies r
Evaluation of production facilities r
Evaluation of futurity of used products r
Selection of new products r
Selection of secondhand markets r
Synchronization of supply chain processes r
Supply chain performance measurement r
In all of these issues, strategic planners must meet the following chal
lenges: uncertainty in supply rate of used products, unknown condition of
used products, and imperfect correlation between supply of used products
and demand for reprocessed goods.
This book addressed the above issues amidst the above challenges in a
variety of decisionmaking situations using effcient models. These models
implement several quantitative techniques, such as:
Analytic hierarchy process r
Eigen vector method r
Analytic network process r
Fuzzy logic r
Extent analysis method r
Fuzzy multicriteria analysis method r
Quality function deployment r
Method of total preferences r
Linear physical programming r
Goal programming r
Technique for order preference by similarity to ideal solution (TOPSIS) r
Borda’s choice rule r
Expert systems r
Bayesian updating r
Taguchi loss function r
Six Sigma r
© 2009 by Taylor & Francis Group, LLC
Conclusions 273
Neural networks r
Geographical information systems r
Linear integer programming r
The issues addressed in this book can serve as foundations for other
researchers to build bodies of knowledge in this new and fastgrowing feld of
research, i.e., strategic planning of reverse and closedloop supply chains. Fur
thermore, the models proposed in this book for those issues can be utilized by
industrialists for understanding how a particular issue in the strategic plan
ning of reverse and closedloop supply chains can be effectively approached
in a particular decisionmaking situation, using a suitable quantitative tech
nique or a suitable combination of two or more quantitative techniques.
Strategic Planning Models for Reverse and ClosedLoop Supply Chains
© 2009 by Taylor & Francis Group, LLC
Strategic Planning Models for Reverse and ClosedLoop Supply Chains
Boca Raton London New York
CRC Press is an imprint of the Taylor & Francis Group, an informa business
© 2009 by Taylor & Francis Group, LLC
CRC Press Taylor & Francis Group 6000 Broken Sound Parkway NW, Suite 300 Boca Raton, FL 334872742 © 2009 by Taylor & Francis Group, LLC CRC Press is an imprint of Taylor & Francis Group, an Informa business No claim to original U.S. Government works Printed in the United States of America on acidfree paper 10 9 8 7 6 5 4 3 2 1 International Standard Book Number13: 9781420054781 (Hardcover) This book contains information obtained from authentic and highly regarded sources. Reasonable efforts have been made to publish reliable data and information, but the author and publisher cannot assume responsibility for the validity of all materials or the consequences of their use. The authors and publishers have attempted to trace the copyright holders of all material reproduced in this publication and apologize to copyright holders if permission to publish in this form has not been obtained. If any copyright material has not been acknowledged please write and let us know so we may rectify in any future reprint. Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced, transmitted, or utilized in any form by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying, microfilming, and recording, or in any information storage or retrieval system, without written permission from the publishers. For permission to photocopy or use material electronically from this work, please access www.copyright.com (http://www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA 01923, 9787508400. CCC is a notforprofit organization that provides licenses and registration for a variety of users. For organizations that have been granted a photocopy license by the CCC, a separate system of payment has been arranged. Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation without intent to infringe. Library of Congress CataloginginPublication Data Pochampally, Kishore K. Strategic planning models for reverse and closedloop supply chains / Kishore K. Pochampally, Satish Nukala, Surendra M. Gupta. p. cm. Includes bibliographical references and index. ISBN 9781420054781 (hardback : alk. paper) 1. Manufacturing processesEnvironmental aspects. 2. Business logisticsEnvironmental aspects. 3. Remanufacturing. 4. Waste minimization. 5. Salvage (Waste, etc.) 6. Recycling (Waste, etc.) I. Nukala, Satish. II. Gupta, Surendra M. III. Title. TS155.7.P63 2008 658.5’67dc22 Visit the Taylor & Francis Web site at http://www.taylorandfrancis.com and the CRC Press Web site at http://www.crcpress.com 2008014566
© 2009 by Taylor & Francis Group, LLC
Dedication
To our families: Seema Dasgupta, Sharada Pochampally, and Narasimha Rao Pochampally —KKP
Somayajulu Nukala, Jagadeeswari Nukala, Sridhar Nukala, Mythili Nukala, and Padmaja Nukala —SN
Sharda Gupta, Monica Gupta, and Neil Gupta —SMG
© 2009 by Taylor & Francis Group, LLC
.. 19 3......................................................................4 Conclusions.....................5 Optimization of Transportation of Goods ....................................................................................................2 Analytic Hierarchy Process and Eigen Vector Method................................................. 31 Quantitative Modeling Techniques ............................ 37 4....... 19 3........3 Evaluation of Collection Centers ............... 31 References................1 Introduction ..................................................... 37 4..............................3 Outline of the Book...............................3 Strategic and Tactical Planning of Reverse and ClosedLoop Supply Chains .........13 Conclusions.............................................................................................................. xvii 1 Introduction.................................................................................................................................................................................. 39 vii © 2009 by Taylor & Francis Group........2 Selection of Used Products ............................................................................................................................................................................4 Evaluation of Recovery Facilities ..3 Analytic Network Process .... 16 2..............................1 1................................................................................................................................................................................................................12 Supply Chain Performance Measurement ....................... xiii Acknowledgments .................... 14 2.......... 7 1..............11 Synchronization of Supply Chain Processes ................. 1 1.......................................................................................................................................... 15 2....................................................................................................................... 13 2............... 16 2..................................... LLC 2 3 4 ............................4 Conclusions.................. 16 2............ 15 2........................8 Evaluation of Futurity of Used Products..... 12 2................................1 Motivation ....9 References............. 13 2..............................................................1 Introduction ...................................................... 19 3......... 17 References...........................................9 Selection of New Products.............................2 Operational Planning of Reverse and ClosedLoop Supply Chains................... 11 2................................................................... 9 Strategic Planning of Reverse and ClosedLoop Supply Chains ......... 14 2......................................................................................................................................... 17 Literature Review .. 24 3........................... 11 2......................................................................................................... 5 1........2 Overview of the Book..................6 Evaluation of Marketing Strategies .. 12 2..Contents Preface ..........................................xv About the Authors .......................10 Selection of Secondhand Markets ...........................................................7 Evaluation of Production Facilities............ 37 4..............................1 Introduction .........
................1 Evaluation Criteria ......6 4..1 Nomenclature ................64 4.........................................2.................................. 68 4................. 73 5..14 Bayesian Updating.........2.....................................1.... 67 4.....................2 First Model (Linear Integer Programming) ...............16.........2 Model .......................................................9 4..................7 4....................2 Numerical Example ...............................17 Neural Networks................1...............2...........................1 The Issue.................................................16....................18 Geographical Information Systems..................16...64 4...1 The Issue........4 Conclusions.......................11 Contents Fuzzy Logic.......... 85 Evaluation of Collection Centers...................................................................................2.............................2..3 Six Sigma Process.....................................77 5.......2.... 58 4........................... 59 4..............3............................................................... 87 6.....................................................16 Six Sigma .....3......................................2....2........................................................................................................... 66 4........3.............. 74 5.............................................. 88 6......................19 Linear Integer Programming .4 4...2 First Model (Eigen Vector Method and Taguchi Loss Function) .....2............................................................................. 55 4..16............................................8 4................................................................................. 49 Goal Programming..........2.........................................................................................................3...................3 Numerical Example ..............................viii 4.................... 88 6..............44 Quality Function Deployment .....................................................................................1 Model Formulation.............. 89 6 © 2009 by Taylor & Francis Group.......43 Fuzzy Multicriteria Analysis Method........2.......2 Model Formulation .. LLC .........3 Second Model (Linear Physical Programming) ........... 69 4.................. 73 5...... 78 5.............. 78 5....................................................2 Class 2S Criteria (Larger Is Better)...........1 Class 1S Criteria (Smaller Is Better) .........................1 Modified CostBenefit Function................................................................. 48 Method of Total Preferences................................................................80 References......... 78 5............15 Taguchi Loss Function ....................... 79 5............................... 75 5..................... 76 5.......................1 Three Sigma Process....2 Linear Integer Programming Model ....................................... 73 5..........................................................2..................................................... 66 4......................................................1 Process Capability Ratio (Cp) ............................... 58 4.....................................................................65 4.................................... 52 Technique for Order Preference by Similarity to Ideal Solution (TOPSIS)............... 87 6.. 40 Extent Analysis Method.............................. 49 Linear Physical Programming .......20 Conclusions..........................10 4.....16............. 69 5 Selection of Used Products................80 5...................5 Sigma Process .........................12 Borda’s Choice Rule .. 61 4............................. 69 References.....................2 4.........5 4....2 Process Capability Index (Cpk)...............................63 4.................................13 Expert Systems .............................. 74 5.
....................... 118 6....................... 91 6.....................2.................. 106 6........6................. 122 6..............2.... 130 © 2009 by Taylor & Francis Group...... Borda’s Rule) ..2 First Model (Analytic Hierarchy Process) .....5.........................2...3 6.......................................1 Nomenclature for Problem Formulation.1 Phase I (Derivation of Impacts) ...2 Application of Goal Programming ...1 Nomenclature for LPP Model ....5........2 Numerical Example.............................................. 95 6....2.........2....7 Labor Cost (LC) ... 92 6.....6.2....................................................... 93 6............... 125 7............................. 95 6....4 ix n Value .......................................................................3 Evaluation Criteria for Second and Third Models .............. 93 6............1 6................... 91 6.............2............................................................ TOPSIS........7 Fifth Model (Eigen Vector Method........................................3.. 124 7 Evaluation of Recovery Facilities .......... LLC ...............5.... 101 6.2.. 94 6....................... 122 6... 91 Utilization of Incentives from Local Government (UI)................2 Application of Goal Programming..........5 Per Capita Income of People in Residential Area (PI) ......1 Criteria of Consumers........................2...... 116 6.............. 130 7.....3 Phase III (Group Decision Making).............3................ and Borda’s Choice Rule) .... 128 7......................1 Phase I (Individual Decision Making) ..................2................ 116 6........................2.. 126 7........1 ThreeLevel Hierarchy .. 92 6.............7...............................2.. 94 6....6.....................................2.............3.. 125 7....2...................................... 90 Distance from Residential Area (DH) .....1 The Issue.... 118 6.................................................................................2 Criteria of Local Government Officials..........1 Application of ANP.....7.............................. 116 6.............. and Goal Programming)....... 109 6.......................8 Conclusions............2............................. 126 7..3 Criteria of Supply Chain Company Executives ...3............2 Problem Formulation........................... 110 6..1 Application of Eigen Vector Method and Taguchi Loss Function ..3 Second Model (Linear Physical Programming) ......1 Nomenclature Used in the Methodology............................................. 121 6.6 Fourth Model (ANP and Goal Programming) .....2 Phase II (Group Decision Making) ........................... 110 6..........................................7.....2. 103 6..................2.............................2................Contents 6..............2.................... TOPSIS....... 124 References....... 91 Distance from Roads (DR) .......4.............2 Phase II (Individual Decision Making).....4.............................................................. Fuzzy Logic..............2 6................... 103 6............................7............. Taguchi Loss Function....................... 92 6.................2..5 Third Model (Neural Networks...........2..6....6 Space Cost (SC)................................8 Incentives from Local Government (IG)...2 Problem Formulation .2.4 Second Model (Eigen Vector Method.......................
.................................................. Fuzzy Logic.... 131 7.. 179 Evaluation of Marketing Strategies ..........................2 First Model (Linear Integer Programming) ...3...........................6........5..............3 Phase III (Group Decision Making) ..............................2 Phase II (Individual Decision Making) ................ and Borda’s Choice Rule) .........................1 Phase I (Derivation of Impacts)....................... 161 8................ TOPSIS.. 132 7..........................1 Model Formulation ...4 Evaluation Criteria for Third and Fourth Models..2.............. 171 8..... 153 8................4 Third Model (Goal Programming).........................2..... 151 8 Optimization of Transportation of Products ..3......... 168 8........... TOPSIS...........3.....5............................................................... 155 8........................................... 134 7........................1 Model Formulation.. 181 9 © 2009 by Taylor & Francis Group.......... 178 8............. 173 8............................ 135 7................................4....................... Borda’s Choice Rule) ... 141 7.................................................x 7.....1 Criteria of Consumers ..................... 168 8..............3........................... 173 8.......3 Numerical Example ................................................3 Criteria of Supply Chain Company Executives.................. 154 8..........1 Class 1S Criteria (Smaller is Better) ..........................................................................2 Phase II (Group Decision Making).....5 Fourth Model (Linear Physical Programming) .....2.........................................................2.........1 Nomenclature ...2 Class 2S Criteria (Larger Is Better)........6 Fifth Model (Fuzzy Goal Programming) .................. 154 8...................3 Numerical Example ..................................................4...................3................ LLC ..............................................................2 Numerical Example ..................................................................1 Phase I (Individual Decision Making)...... 135 7......8 Conclusions........... 135 7.2 Criteria of Local Government Officials ......................................................................................... 158 8............. 148 7.................................................1 The Issue..6............................1 Model Formulation...................4......7 Fifth Model (TwoDimensional Chart) ................................6.... 151 References........................................................5.....4.................................. 157 8.......................................3................... 179 References............ 162 8.........7 Conclusions........ 140 7...................... 131 7...4................................................................................... 160 8.......6 Fourth Model (Neural Networks.................................. 153 8... 166 8......2 Model Formulation ...................................... 161 8............................................................2 Contents Criteria for Identification of Efficient Recovery Facilities......................6............ 140 7......1 Nomenclature........5 Third Model (Eigen Vector Method............5.....................2 Model Formulation ............................. 147 7.......2........ 134 7................................3 Numerical Example............................ 132 7...... 158 8...............2 Numerical Example ......4...........1 The Issue........................ 131 7........................... 181 9....................................6...........3 Second Model (Linear Physical Programming) .............2 Numerical Example ...... 145 7.....
.2..........................................2................. 203 10........................................................ 203 10....6 Customer Service (CSE). 204 10..................................5 Conclusions...................2.2 Assumptions ....2 Usage of Fuzzy Logic ....................................................................... 235 12....................................................................................................................1 Performance Aspects and Enablers ............1........4 Potentiality (POT)........ 201 10............. 182 9.....4..................1...........Contents 9..... 182 9...... 204 10............................ 192 9.........1 The Issue ....................................................................4 Formulation of Fuzzy CostBenefit Function.............................................5 Conclusions..1 The Issue......................2.............................. 188 9.. Quality Function Deployment..............3 Second Model (Fuzzy Logic........................ 205 10..........................................................3 Rules Used in Bayesian Updating ..2.......................................................................... 212 10..... and Analytic Network Process) ...1 Main Criteria and Subcriteria ..........................2 Numerical Example............ 204 10... 188 9..............2..................... 202 10..................4 Third Model (Fuzzy Logic. 201 10................................................2............................................................... 236 12...............4 Bayesian Updating... and Method of Total Preferences) .. Extent Analysis Method......2 xi First Model (Fuzzy Logic and TOPSIS)............................ 238 11 12 © 2009 by Taylor & Francis Group............... 232 References......................................2................................... 232 11............................ 193 9.................5 Cost (COS) .....................................................3................................... LLC ....1 Drivers of Public Participation........1... 227 11...........................................2 Methodology................ 231 11...................................................... 193 9.............................2 First Model (Fuzzy Logic and TOPSIS) ...................................................1.................................1............................................................. 199 10 Evaluation of Production Facilities .............1 The Issue........................ 190 9........................................ 198 References......................3 Nomenclature...........1 Evaluation Criteria...... 235 12... and Analytic Network Process) ..................4........2 Numerical Example...................4 Third Model (Fuzzy Multicriteria Analysis Method) ............................. 204 10..... 233 Selection of New Products ..........2. 236 12...............................1 Environmentally Conscious Design (ECD) ............ 215 10...3............1........3 Second Model (Fuzzy Logic.................................................2...................................................................... 183 9.......... 227 11.... 203 10........................ 229 11............2 Numerical Example..............2 Environmentally Conscious Manufacturing (ECM) ...............................5 FLEXBased Expert System .................. 226 Evaluation of Futurity of Used Products .........................3 Attitude of Management (AMT) ........... 226 References...........6 Conclusions.................................................................................................... Extent Analysis Method................. 230 11............................................................
. 261 15............................................ 240 12............2...4.......4..........................................................................................2 Total Reuse Revenue per Period (UR) ..................................7 Total Disposal Cost per Period (DC)...........................................9 Investment Cost (IC).....2 Second Experiment (Determination of Variance Pool)............................................... 240 12.............................6 Numerical Example............................................... 258 15..................2..8 LossofSale Cost per Period (LC) .......................... 269 Conclusions ...........................................................................2 Application of LPP to QFD Optimization......2 Model (Two Design Experiments) ......6 Total Reprocessing Cost per Period (RC) ......................2 Performance Aspects and Enablers for Application of QFD.......4............................................................. 246 13......... 238 12......... 238 12......................... 245 13...........................................................1 Total New Product Sale Revenue per Period (SR)...................................................................... 269 References...........................2.........................................3..............................1 The Issue.........................................2........................... 243 References. 245 13....1 The Issue..............................1 Performance Aspects and Enablers ..1 First Step ... 244 13 Selection of Secondhand Markets.............3 Total Recycle Revenue per Period (CR) ..... 263 15.........................4 Conclusions....4.............................................................................................. 255 Performance Measurement....3 Selection of Potential Secondhand Markets ............. 239 12.. 258 15..........................................4........................................2 Numerical Example............ LLC ........................................................ 245 13........................1 First Experiment (Determination of Nominal Pool).......................... 239 12.............................................................4..........................................xii Contents 12........... 241 12..............................4 Conclusions............... 271 14 15 16 © 2009 by Taylor & Francis Group..........4.3 Reverse/ClosedLoop Supply Chain Performance Measurement .................................................. 239 12........7 Conclusions .............................................................................................................. 240 12.3 Conclusions.. 251 14. 251 14..... 260 15.....5 Model...............4 Total New Product Production Cost per Period (MC) ...............254 References............ 261 15........... 251 14................................. 253 14..... 251 14................... 257 15.....................................3........................... 257 15............................................4.....4............... 241 12..........5 Total Collection Cost per Period (CC) ........................................... 250 Design of a Synchronized Reverse Supply Chain ...1 The Issue................................... 239 12.........2 Second Step ................
along with the growing desire of consumers to acquire the latest technology. marketing strategies. as well as evaluation of the futurity of used products. an important driver for companies to engage in reprocessing is the enforcement of environmental regulations by local governments. Saving energy: It usually takes less energy to make products from reprocessed materials and components than from virgin resources. it has become common for companies involved in a forward supply chain (series of activities required to produce new products from virgin resources and distribute them to consumers) to also carry out collection and reprocessing of used products. 2. in both scope and scale. This combined practice of forward and reverse supply chains is called a closedloop supply chain. The issues faced by strategic planners of reverse and closedloop supply chains are evaluation and selection of new and used products. Saving clean air and water: Making products from reprocessed materials and components creates less air and water pollution than from virgin resources. In this case. Besides the above opportunities. collection centers. Strategic planning (also called designing) primarily involves the structuring (which products should be processed/produced in which facilities) of a supply chain over the next several years. in practice. recovery facilities. 3. selection of xiii © 2009 by Taylor & Francis Group. LLC . A reverse supply chain consists of a series of activities required to collect used products from consumers and reprocess them to either recover their leftover market values or properly dispose of them. and production facilities. It is longrange planning and is typically performed every few years when a supply chain needs to expand its capabilities. But behind every problem lies an opportunity. In the past decade. there has been an explosive growth of reverse and closedloop supply chains. Saving money: It costs much less to make products from reprocessed materials and components than from virgin resources. 4. Saving natural resources: We conserve land and reduce the need to drill for oil and dig for minerals by making products using materials and components obtained from reprocessing rather than virgin resources. Today. has led to a new environmental problem: products that are discarded prematurely. the opportunity comes from reprocessing (processing used products). which leads to: 1.Preface The rapid technological development of new products. they do not go into landfills. 5. Saving landfill space: When reprocessed materials and components are used to make a product.
fuzzy multicriteria analysis method. Furthermore. using a suitable quantitative technique or a suitable combination of two or more quantitative techniques. environmentalists as well as industrialists can fully exploit their desired goals. method of total preferences. fuzzy logic. These models implement several quantitative techniques. © 2009 by Taylor & Francis Group. technique for order preference by similarity to ideal solution (TOPSIS). LLC . strategic planners must meet the following challenges: uncertainty in supply rate of used products. linear physical programming. The strategic planning of reverse and closedloop supply chains remains an important and promising field of research. In all of these issues. Bayesian updating. and supply chain performance measurement. The issues addressed in this book can serve as foundations for other researchers to build bodies of knowledge in this new and fastgrowing field of research. strategic planning of reverse and closedloop supply chains. The authors express their hope that this book will inspire further research and motivate new and rewarding research in this all too important field of study. analytic network process.. unknown condition of used products. extent analysis method. and imperfect correlation between supply of used products and demand for reprocessed goods. This is desirable from both environmental and economical points of view. synchronization of supply chain processes. such as analytic hierarchy process. optimization of transportation of goods. This book addresses the above issues amidst the above challenges in a variety of decisionmaking situations using efficient models. expert systems. the models proposed in this book for those issues can be utilized by industrialists for understanding how a particular issue in the strategic planning of reverse and closedloop supply chains can be effectively approached in a particular decisionmaking situation.xiv Preface secondhand markets. viz. and linear integer programming. Taguchi loss function. geographical information systems. because by using innovations in this field. quality function deployment. eigen vector method. neural networks. Six Sigma. goal programming. Borda’s choice rule.
We thank hundreds of researchers from whose work we have benefited and many of whom we have had the good fortune of meeting and interacting with at conferences around the world. New Hampshire Satish Nukala. Most importantly. Gupta. Sagar V. Tej S. Dr. PhD Boston. We express our appreciation to Taylor & Francis and its staff for providing seamless support in making it possible to complete this timely and important book. Kishore K. Martin Bradley (Southern New Hampshire University). PhD Manchester.Acknowledgments Encouragement for writing this book came from many sources. Kamarthi (Northeastern University). Massachusetts xv © 2009 by Taylor & Francis Group. Texas Surendra M. for constantly providing us their unconditional support in making this book a reality. This book would not have been possible without the commitment of Taylor & Francis for encouraging innovative ideas. Many thanks to Dr. Elif Kongar (University of Bridgeport). PhD Houston. Seamus M. Dr. Pochampally. we are indebted to our families. and Dr. McGovern (National Transportation Systems Center). Dr. LLC . Dhakar (Southern New Hampshire University). to whom this book is lovingly dedicated.
MBA from Bryant University. Dr. MRP. IIE. He is a registered professional engineer in the state of Massachusetts and a member of ASEE. He holds a PhD in industrial engineering from Northeastern University in Boston and an MS in industrial and systems engineering from the University of Alabama in Huntsville. and dissertations. His publications have been cited by thousands of researchers all over the world in journals. Dr. His research interests are in the areas of reverse logistics.About the Authors Dr.neu. supply chain design. INFORMS. xvii © 2009 by Taylor & Francis Group. He is mostly interested in environmentally conscious manufacturing. PE. He has authored and coauthored more than 350 technical papers published in prestigious journals. Dr. track chair. and MSIE and PhD in industrial engineering from Purdue University. He has also served as conference chair. JIT.edu/~smgupta/. and member of technical committees of a variety of international conferences. Gupta’s research interests are in the areas of production/ manufacturing systems and operations research. He is currently serving as the area editor of “Environmental Issues” for Computers and Industrial Engineering. Surendra M. and conference proceedings.edu. He holds a PhD in industrial engineering from Northeastern University in Boston. He has traveled to all seven continents and presented his work at international conferences there (except Antarctica). Gupta. His recent activities can be viewed at http:/ /www1. Gupta has been elected to the memberships of several honor societies and is listed in various Who’s Who publications. books.coe. proceedings. He is a Six Sigma Green Belt and a Project Management Professional (PMP®). His research interests are in the areas of production and operations management and total quality management. Dr. DSI. Dr. Satish Nukala is a senior supply chain analyst with Halliburton Energy Services in Houston. and Six Sigma quality management. electronics manufacturing. He received his BE in electronics engineering from Birla Institute of Technology and Science. is a professor of mechanical and industrial engineering and director of the Laboratory for Responsible Manufacturing at Northeastern University in Boston. and POMS. and he can be reached by email at gupta@neu. books. LLC . Gupta is a recipient of the Outstanding Research Award and the Outstanding Industrial Engineering Professor Award (in recognition of teaching excellence) from Northeastern University. Kishore K. He is a Six Sigma Green Belt. His prior academic experience is as a postdoctoral fellow at Massachusetts Institute of Technology (MIT) in Cambridge. Pochampally is an assistant professor of quantitative studies and operations management at Southern New Hampshire University in Manchester. associate editor for International Journal of Agile Systems and Management. and editorial board member of a variety of journals. and queueing theory. Texas.
As a result. remanufacturing and recycling are the major reprocessing options applied in the industry. This environmental degradation is not sustainable by Earth’s ecosystem [6]. and recycling is a process performed to retrieve the material content of used products without retaining the identity of their components [5]. 29 states in the United States have 10 or more years of landfill space left.1 Introduction 1. For example. in the discarded electronic equipment poses a serious threat to the environment.. used products that are discarded prematurely.2 give an idea about the materials and energy consumed in the manufacturing of microchips and LCD monitors. both at home and in the workplace. the presence of toxic material. Reprocessing of used products helps in: * Though direct reuse of the used products is infeasible in most cases. out of which fewer than 10% are reprocessed. mercury. and 6 states fewer than 5 years [2]. viz. the world now faces a serious environmental problem: waste. Remanufacturing is a process in which used products are restored to likenew conditions. along with the rapid technological development of new products. polybrominated diphenyl ether. an estimated 60 million computers enter the market in the United States every year and more than 12 million computers are discarded every year.* while the rest head to landfills [1]. 1 © 2009 by Taylor & Francis Group. respectively. Increased consumption results in increased use of raw material and energy. 4] presented in tables 1. 15 states between 5 and 10 years. The data [3.1 and 1. This is fueled by the growing desire of consumers to acquire the latest technology. Apart from the excessive consumption of the world’s finite natural resources and disposal capacities. LLC . According to the Environmental Protection Agency’s Municipal Solid Waste Fact Book. thereby depleting the world’s finite natural resources. and hexavalent chromium. such as lead.1 Motivation The level of consumption by a growing population in this world of finite resources and disposal capacities has been continuously increasing.
2.1 Materials and energy used in manufacturing microchips Amount per memory chip 0. H2. 3.2 Strategic Planning Models for Reverse and ClosedLoop Supply Chains TABLE 1. Ar) Electricity Direct fossil fuels (98% natural gas) Embodied fossil fuels Water Amount used per monitor 3. 4.9 kWh 1.290 liters 1. He. energy. Ar Energy Electricity Direct fossil fuels Embodied fossil fuels Water 0.25 g Dopants Photolithography Etchants Acids/bases Total chemicals Elemental gases N2. and water use in manufacture of LCD monitor Material/input Photolithographic and other chemicals Elemental gases (N2.002 kg 2.600 tons 890. Saving natural resources: We conserve land and reduce the need to drill for oil and dig for minerals by making products using materials and components obtained from reprocessing instead of virgin materials.016 g 22 g 0.2 kg 0.7 kg 5. LLC . Saving money: It costs much less to make products from reprocessed materials and components than from virgin materials. Saving energy: It usually takes less energy to make products from reprocessed materials and components than from virgin materials. O2.2 Aggregate chemicals. they do not go into landfills.400 tons 280 tons 390.037 kg 4. Saving landfill space: When reprocessed materials and components are used to make a product.9 kg 87 kWh 198 kg 226 kg 1.9 kg 7. Saving clean air and water: Making products from reprocessed materials and components creates less air pollution and water pollution than from virgin materials. O2. 5. © 2009 by Taylor & Francis Group.37 g 50 g 72 g 700 g 2.025 kg 0.000 tons 1.1 kg 69 kg 281 kWh 155 MJ 94 kg 310 liters Material Silicon wafer Chemicals Description TABLE 1.000 tons 6.6 MJ 970 g 32 liters Annual use by Amount used to industry make chips in worldwide one computer 4.3 million tons 12 million tons 52 billion kWh 28 billion MJ 17 million tons 570 billion liters 0.
and packaging material in Germany. series of activities required to produce new products from raw material and distribute the former to customers). In the United States. 10]. from the costs associated with collecting and disposing of electronic waste. cities and towns are responsible for retrieval of used products and proper disposal of the potentially environmentally dangerous and waste components. At the same time. environmental consciousness has become an obligation to many facilities in a traditional/forward supply chain (i. customers are expecting companies to minimize the environmental impact of their products and processes. In the last decade. many of these facilities are driven.. © 2009 by Taylor & Francis Group. cars in the European Union and Taiwan. reprocessing is still in its infancy. The bill’s supporters say that cities and towns in the United States spend between $6 million and $21 million a year on such endeavors.1 shows a generic reverse supply chain. viz. particularly in the European Union. If passed. In this vein. in the state of Massachusetts. mainly by profitability. The obligations of collection of used products and their reprocessing have been enacted or are under way for a number of product categories. A reverse supply chain (also known as reverse logistics) consists of a series of activities required to collect a used product from a consumer and reprocess it (used product) to either recover its leftover market value or dispose of it. However. Figure 1. support is building for a refiled bill that would require manufacturers of electronic products to pay for collection and reprocessing of those products after the end of their usage by the consumers. Implementation of any reverse supply chain requires at least three parties: collection centers where consumers return used products. outgoing goods from recovery facilities. A generic closedloop supply chain is shown in figure 1. recovery facilities where reprocessing (remanufacturing or recycling) is performed.Introduction 3 As waste reduction is becoming a major concern in industrialized countries. brought about by production cost savings and access to new market segments. have introduced environmental legislation charging manufacturers with responsibility for the whole life cycle of their products. which are bracing for local aid cuts. Several countries. Environmental consciousness suggests the production of new products from conceptual design to final delivery such that the environmental standards and requirements are satisfied. including electronic equipment in the European Union and Japan. According to a 2003 report [8].. The combination of forward and reverse supply chains is called a closedloop supply chain. enforced primarily by governmental regulations and customer perspective on environmental issues [9. and demand centers where customers buy reprocessed products.e. Moreover. legislation extending producers’ responsibility has become an important element of public environmental policy. the concept of reprocessing is gradually replacing a oneway perception of economy [7]. Increasingly. in companies in the United States. the statewide takeback program would be the first of its kind in the nation and would relieve cities and towns. to administer the reverse supply chain as well. in the past decade.2. Also. companies are recognizing opportunities for combining environmental stewardship with profitability. there has been an explosive growth of reprocessing activities in both scope and scale. LLC .
Supply chain design decisions are typically made for the long term (years) and are very expensive to alter on short notice. Raw Material Suppliers New Products Recycled Goods Remanufactured Products Recycled Goods Demand Centers New Products Remanufactured Products Production Facilities Consumers Used Products Collection Centers Used Products FIGURE 1. computers. Consequently. A supply chain involves three phases of decision making: supply chain design. appliances. automobiles. including commercial aircraft. chemicals. supply chain planning. and supply chain operation [12]: Supply chain design: This phase is also called strategic planning. Decisions made by companies include the location and capacity of production and warehouse facilities. the products to be manufactured and stored at various locations.1 Generic reverse supply chain. the modes of transportation to be made available. LLC . when companies make these decisions. © 2009 by Taylor & Francis Group. During this phase.4 Strategic Planning Models for Reverse and ClosedLoop Supply Chains Recycled Goods Remanufactured Products Demand Centers Recovery Facilities Used Products Collection Centers Used Products Consumers FIGURE 1. they must take into account uncertainty in anticipated market conditions over the next few years. a company decides how to structure its supply chain over the next several years. Reverse/closedloop supply chains are implemented in many industries. and apparel [11].2 Generic closedloop supply chain. and the type of information system to be utilized.
1. there is less uncertainty about demand information. strategic planning models for a forward supply chain cannot be adopted for a reverse or closedloop supply chain. generating pick lists at a warehouse. 5 This book focuses on strategic planning of reverse and closedloop supply chains. section 1. As a result. Given a shorter time frame and better forecasts than the design phase.2 presents the overview of this book.3 gives the outline of the book. During this phase. Because operational decisions are made in the short term (minutes. This chapter is organized as follows: section 1. © 2009 by Taylor & Francis Group. and placing replenishment orders.3 [13]). Reverse and closedloop supply chains differ from traditional/forward supply chains in many aspects and are complex to handle because of the inherent uncertainty involved in every stage of strategic planning (see table 1. exchange rates. companies make decisions regarding individual customer orders. setting a date for an order to be filled.2 Overview of the Book This book presents quantitative models for various issues faced by strategic planners of reverse and closedloop supply chains amidst many challenges. and optimization of transportation of goods. Companies must include uncertainty in demand. For decisions made during this phase. companies in this phase try to incorporate any flexibility built into the supply chain in the design phase and exploit it to optimize performance. and timing and size of marketing promotions. Decisions include allocating inventory or production to individual orders. hours. selection of new products to produce. such as uncertainty in supply rate of used products. and competition over the time frame. inventory policies to be followed. Supply chain planning includes decisions regarding subcontracting of manufacturing. or days).4 gives some conclusions. and section 1. setting delivery schedules of trucks. unknown condition of used products. LLC . The various decisionmaking problems faced by strategic planners of reverse and closedloop supply chains include selection of used products to reprocess. Supply chain operation: This phase is also called operational planning. evaluation of recovery facilities. and imperfect correlation between supply of used products and demand for reprocessed goods.Introduction Supply chain planning: This phase is also called tactical planning. evaluation of collection centers. The goal of supply chain operations is to handle incoming customer orders in the best possible manner. the time frame considered is 3 months to a year.
3 Comparison between forward and reverse/closedloop supply chains Forward supply chain Product quality uniform Disposition options clear Routing of products unambiguous Costs involved easily understood Product pricing uniform Inventory management consistent Product life cycle manageable Financial management issues clear Customer easily identifiable to market Forecasting relatively straight forward Onetomany transportation Marketing methods well known Process visibility more transparent Reverse/closedloop supply chain Product quality not uniform Disposition options unclear Routing of products ambiguous Costs involved not easily understood Product pricing not uniform Inventory management inconsistent Product life cycle less manageable Financial management issues unclear Customer less easily identifiable to market Forecasting more difficult Manytoone transportation Marketing complication by several factors Process visibility less transparent Negotiations between parties straightforward Negotiations less straightforward The complete list of issues addressed in this book is the following (see chapter 2 for further explanation of each of the issues): Selection of used products Evaluation of collection centers Evaluation of recovery facilities Optimization of transportation of goods Evaluation of marketing strategies Evaluation of production facilities Evaluation of futurity of used products Selection of new products Selection of secondhand markets Synchronization of supply chain processes Supply chain performance measurement The models for addressing the above strategic planning issues employ several quantitative techniques. such as: Analytic hierarchy process Eigen vector method Analytic network process Fuzzy logic © 2009 by Taylor & Francis Group.6 Strategic Planning Models for Reverse and ClosedLoop Supply Chains TABLE 1. LLC .
In chapter 4.3 Outline of the Book This book is organized as follows: In chapter 2. strategic planning of reverse and closedloop supply chains.2) is presented. Furthermore. In chapter 3. the models proposed in this book for those issues can be utilized by industrialists for understanding how a particular issue in the strategic planning of reverse and closedloop supply chains can be effectively approached in a particular decisionmaking situation. a brief review of literature in the area of strategic planning of reverse and closedloop supply chains is presented. LLC . The first model employs linear integer programming. an overview of the strategic planning issues (see section 1. each of the several quantitative techniques that are used in the strategic planning models presented in this book is briefly described. i. and it is explained how the decisionmaking situations could differ for each of the issues.Introduction Extent analysis method Fuzzy multicriteria analysis method Quality function deployment Method of total preferences Linear physical programming Goal programming Technique for order preference by similarity to ideal solution (TOPSIS) Borda’s choice rule Expert systems Bayesian updating Taguchi loss function Six Sigma Neural networks Geographical information systems Linear integer programming 7 The issues addressed in this book can serve as foundations for other researchers to build bodies of knowledge in this new and fastgrowing field of research.e. using a suitable quantitative technique or a suitable combination of two or more quantitative techniques. © 2009 by Taylor & Francis Group. and the second model employs linear physical programming.. two models for selection of used products in two different decisionmaking situations are presented. In chapter 5. 1.
Chapter 8 focuses on achieving transportation of the right quantities of products (used. and Borda’s choice rule. The third model uses fuzzy logic.8 Strategic Planning Models for Reverse and ClosedLoop Supply Chains In chapter 6. and Borda’s choice rule. to decide whether it is more sensible to repair a used product of interest for subsequent sale on a secondhand market than to disassemble it for subsequent reprocessing. and method of total preferences. while satisfying certain constraints. The fifth model employs fuzzy goal programming. TOPSIS. respectively. fuzzy logic. The second model employs linear physical programming. The second model employs linear physical programming. Five models are presented in five different decisionmaking situations. three models for evaluation of production facilities in five different decisionmaking situations are presented. extent analysis method. The third model uses neural networks. quality function deployment (QFD). quality function deployment (QFD). In chapter 10. five models for evaluation of collection centers in five different decisionmaking situations are presented. fuzzy logic. © 2009 by Taylor & Francis Group. and analytic network process. The fourth model uses neural networks. and Borda’s choice rule. LLC . In chapter 9. The first model uses analytic hierarchy process. The first model employs eigen vector method and Taguchi loss function. remanufactured. Taguchi loss function. extent analysis method. technique for order preference by similarity to ideal solution (TOPSIS). The third and fourth models use goal programming and linear physical programming. In chapter 12. fuzzy logic. In chapter 7. The fourth model uses analytic network process (ANP) and goal programming. it is shown how an expert system can be built using Bayesian updating and fuzzy logic. The second model employs fuzzy logic. The third model uses eigen vector method. The first model employs linear integer programming. The second model uses eigen vector method. and analytic network process. In chapter 13. technique for order preference by similarity to ideal solution (TOPSIS). and goal programming. three models for evaluation of the marketing strategy of a reverse/closedloop supply chain in three different decisionmaking situations are presented. In chapter 11. and method of total preferences are used to select the market with the most potential in which to sell a used product from a set of candidate secondhand markets. TOPSIS. The second model employs fuzzy logic. and Borda’s choice rule. a fuzzy costbenefit function is formulated and then used to perform a multicriteria economic analysis for selecting an economical new product to produce in a closedloop supply chain. The fifth model uses eigen vector method. and new) across a reverse or closedloop supply chain. The third model uses fuzzy multicriteria analysis method. five models for evaluation of recovery facilities in five different decisionmaking situations are presented. The fifth model uses a simple twodimensional chart. The first model employs fuzzy logic and technique for order preference by similarity to ideal solution (TOPSIS). The first model employs fuzzy logic and technique for order preference by similarity to ideal solution (TOPSIS).
.. Chapter 16 presents the conclusions of this book. © 2009 by Taylor & Francis Group. 2001.Introduction 9 In chapter 14.S. M. 2005. 5. 2. Plug into electronics reuse. Disassembly modeling for assembly. M.. LLC . 1999. K.7 kg microchip: Energy and chemical use in the production of semiconductors. S. C. a model consisting of two design experiments that use the Six Sigma concept to achieve better synchronization in a reverse supply chain is presented. Beamon.. M. 1997. T. 2003. Socolof. A. Overly. Boston Metro.. 6. and a performance measurement model that uses linear physical programming (LPP) and quality function deployment (QFD) is presented.. S. 2002. 13–38. and recycling. A. and Greibig.. 6. M. C. B.. Computer recycling bill is gaining support. Washington. R.. reuse. and Gupta. D. Lambert. Production Planning and Control 15:270–81. 2004. M. 2003. drivers of reverse and closedloop supply chains. Ayres. DC: U. 8. Washington. and Hyde. Berlin: SpringlerVerlag. 3. DC: Institute for Local Self Reliance. Anonymous. L. G. 2002. and Fernandes. International Journal of Physical Distribution and Logistics Management 32:455–79. J.. Environmental Science and Technology 36:5504–10. 1.. Quantitative models for reverse logistics: Lecture notes in economics and mathematical systems. Environmental Protection Agency. The 1. References 1. Issues in environmentally conscious manufacturing and product recovery: A survey. viz. and Heller. appropriate performance aspects and their enablers (drivers of performance metrics) are identified for a reverse/closedloop supply chain environment. Platt. M.4 Conclusions This chapter first presented the factors that motivated this project. Williams. 4. 9. This model tailors the individual processes in such a way that the overall delivery performance is maximized. J. J. Desktop computer displays: A lifecycle assessment. p. A. B. Boca Raton. E. A qualitative examination of factors affecting reverse logistics systems for endoflife computers. Computers and Industrial Engineering 36:811–53. Ponzurick. Supplychain network configuration for product recovery. Fleischmann. maintenance. January 10–12. In chapter 15. Then the overview and outline of the book were presented. FL: CRC Press. M. Knemeyer. Gungor. EPA 744R01004. and Gupta. 7. and Logar. M.
2001. J. N. P. M. California Management Review 37:114–35. 18. D. Rogers. 1995. and van Wassenhove. NJ: Prentice Hall. Chopra... Closedloop supply chain models with product remanufacturing. Englewood Cliffs. Management Science 50:239–52. van Nunen.10 Strategic Planning Models for Reverse and ClosedLoop Supply Chains 10.. Thierry. and Van Wassenhove Luk.. October. 2004. Strategic issues in product recovery management. 2006.. planning and operations.. p. LLC . Bhattacharya. 11. S. Salomon. C. Savaskan. © 2009 by Taylor & Francis Group. 12. RLEC project plans. S. L. Livonia. N. M. 13. Supply chain management: Strategy. and Meindl. R.
5. Section 2.13 gives some conclusions. 4. the first decisionmaking phase in a supply chain is supply chain design. 9. 11.2 Strategic Planning of Reverse and ClosedLoop Supply Chains 2. The various issues faced by strategic planners of reverse and closedloop supply chains are the following: 1.1 Introduction As explained in chapter 1. Selection of used products Evaluation of collection centers Evaluation of recovery facilities Optimization of transportation of goods Evaluation of marketing strategies Evaluation of production facilities Evaluation of futurity of used products Selection of new products Selection of secondhand markets Synchronization of supply chain processes Supply chain performance measurement In all of these issues. 10. strategic planners must meet the following challenges: uncertainty in supply rate of used products. 2. respectively) and how they could be addressed in a variety of decisionmaking situations using effective quantitative models. unknown condition of used products. It is longrange planning and is typically performed every few years when a supply chain needs to expand its capabilities [1]. 6. This chapter gives an overview of the aforementioned issues (sections 2. 7. 3. 8. and imperfect correlation between supply of used products and demand for reprocessed goods. 11 © 2009 by Taylor & Francis Group. Strategic planning primarily involves the structuring of a supply chain over the next several years.12. which is also called strategic planning.2 through 2. LLC .
Supply chain company executives. 2. LLC . 3. and supply chain company executives.12 Strategic Planning Models for Reverse and ClosedLoop Supply Chains 2. must be derived). 4. The various scenarios for evaluating collection centers for efficiency could differ as follows: 1. Evaluation criteria could be presented in terms of ranges of different degrees of desirability. sorting). © 2009 by Taylor & Francis Group. ship the used products to recovery facilities or production facilities where reprocessing operations such as disassembly and recycling/remanufacturing are carried out. There could exist three different categories of decision makers: consumers. These collection centers. after initial processing (for example.3 Evaluation of Collection Centers Designing an efficient reverse or closedloop supply chain requires selection of efficient collection centers where used products are disposed of by the consumers. local government officials. These companies select only those used products for which revenues from recycle or resale of the products’ components are expected to be higher than the costs involved in collection and reprocessing of used products and in disposal of waste. could be the sole decision makers. Evaluation criteria could be presented in terms of classical numerical constraints. 2. Evaluation could be made from the perspective of a remanufacturing facility interested in buying used products from the candidate collection centers. The weights (importance values) of evaluation criteria are given.2 Selection of Used Products Although many original equipment manufacturers (OEMs) are obligated to take products back from the consumers upon the products’ end of use (hence called used products). there are also many thirdparty companies that collect used products solely to make profit. There could exist the same three different categories of decision makers given in scenario 2. 2. but weights of evaluation criteria are not given (hence. The goals are expressed in terms of performance indices (efficiency scores). The various scenarios for selecting economical used products could differ as follows: 1. whose primary concern is profit.
Decision makers could have conflicting criteria for evaluation. in addition to evaluation of collection centers (see section 2. 2. 4. Evaluation could be made in the same manner as in scenario 4. strategic planning of a reverse supply chain involves evaluation of recovery facilities. A very simple evaluation technique could be desired (where only the “most important” evaluation criteria are considered). but the goals are expressed in terms of Taguchi losses (inefficiency scores). 13 2. 2. Decisionmaking criteria for a reverse supply chain could be presented in terms of ranges of different degrees of desirability. Decisionmaking criteria for a reverse supply chain could be given in terms of classical supplyanddemand constraints. remanufactured. must be derived). Evaluation criteria could be given numerical weights (importance values). Hence. The various scenarios for evaluating recovery facilities for efficiency could differ as follows: 1. LLC . Evaluation criteria could be presented in terms of ranges of different degrees of desirability. there could be a need to address the following issues in one continuous phase for a closedloop supply chain: selection of used products and evaluation of © 2009 by Taylor & Francis Group.3).4 Evaluation of Recovery Facilities Efficient recovery facilities (where reprocessing operations such as disassembly and recycling/remanufacturing are carried out) are as essential for a reverse supply chain as efficient collection centers. The various scenarios for this problem could differ as follows: 1. Besides optimal transportation of products. 5.Strategic Planning of Reverse and ClosedLoop Supply Chains 5.5 Optimization of Transportation of Goods The focus of this issue is on achieving transportation of the right quantities of products (used. Decision makers could have conflicting criteria for evaluation. and weights for evaluation criteria are given. and new) across a reverse or closedloop supply chain while satisfying certain constraints. 3. 2. and weights for evaluation criteria are not given (hence. 3.
3). Whereas the drivers for governments and companies to implement a reverse/closedloop supply chain program and evaluate the program’s marketing strategies are environmental consciousness and profitability. in addition to selecting efficient collection centers (see section 2. Also. i.6 Evaluation of Marketing Strategies A reverse/closedloop supply chain program is successful only if there is a high level of public participation in the program. absence of a closed loop. The program could be exclusively for reverse supply chain operations. The program could be for a closedloop supply chain and the decision maker could be uninterested in considering interdependencies among evaluation criteria. The program could be for a closedloop supply chain and the decision maker could be interested in considering interdependencies among evaluation criteria. respectively. except that the decisionmaking criteria could be presented in terms of different degrees of desirability.14 Strategic Planning Models for Reverse and ClosedLoop Supply Chains production facilities. 3. 5. The various scenarios for evaluating the marketing strategy of a reverse/closedloop supply chain program could differ as follows: 1. the higher the taxes the consumers will have to pay). the more regularly a reverse/closedloop supply chain program offers to collect used products from consumers.7 Evaluation of Production Facilities Efficient production facilities (not only new products are produced but also used products are reprocessed) are as essential for a closedloop supply chain as efficient collection centers. 2. the decisionmaking criteria could be presented in terms of classic supplyanddemand constraints. 2. the drivers for the public to participate in the program are numerous and often conflicting with each other (for example. 2. Hence. LLC .e. The scenario could be the same as scenario 3. The scenario could be the same as scenario 3. except that the decisionmaking criteria could be imprecise. The level of public participation is shouldered by the marketing strategy of that program. Hence. evaluating the marketing strategy of a reverse/closedloop supply chain program is equivalent to evaluating how well the strategy is driving the public to participate in the program.. strategic planning of a closedloop supply © 2009 by Taylor & Francis Group. 4.
which might occur occasionally. and the decision maker could desire to see how close the rating of a candidate production facility is to the “ideal solution.8 Evaluation of Futurity of Used Products Although a major driver for companies interested in collecting used products is recoverable value through reprocessing. it makes more sense to make necessary repairs to the products and sell them on secondhand markets than to disassemble them for subsequent reprocessing. Interactions among evaluation criteria could not be ignored. © 2009 by Taylor & Francis Group. The various scenarios for evaluating production facilities for efficiency could differ as follows: 1. reprocessing cost (cost to remanufacture/recycle used products).. The revenues include new product sale revenue (revenue from selling new products. for some products. lossofsale cost (cost due to loss of sale. wherein interactions among evaluation criteria could be ignored. products in the forward flow of the closedloop supply chain). disposal cost (cost to dispose of the material left over after remanufacturing or recycling of used products). those companies seldom know when those products were bought and why they were discarded. and investment cost (capital required for facilities and machinery involved in production of new products and collection and reprocessing of used products). 2. Hence. reuse revenue (revenue from direct sale/usage in remanufacturing usable components of used products). collection cost (cost to collect used products from consumers). and recycle revenue (revenue from selling material obtained from recycling of unusable components of used products).” 2. the products do not indicate their remaining life periods.Strategic Planning of Reverse and ClosedLoop Supply Chains 15 chain involves evaluation of production facilities. Also. viz. 3. The costs include new product production cost (cost to produce new products). due to lack of supply of used products). 2. The decision maker could desire to structure the problem using a simple hierarchical model.9 Selection of New Products The focus of this issue is to help companies select and produce only those new products for which revenues in the closedloop supply chain are expected to be higher than the costs. which in turn leads to a more complex problem structure. LLC . Interactions among evaluation criteria could not be ignored. The focus of this issue is to find out whether. they often undergo partial or complete disassembly for subsequent reprocessing.
Due to the inherent differences in various aspects between forward and reverse/closedloop supply chains (see chapter 1). © 2009 by Taylor & Francis Group. The focus of this issue is to achieve synchronization of processes. service efficiency. whilesale performance (reflects the ability to motivate the customers to buy secondhand products while the customers are in the secondhand market). such as ontime delivery. LLC .11 Synchronization of Supply Chain Processes Synchronization among the internal business processes of a supply chain is essential for effective management of the supply chain. the performance metrics and evaluation techniques used in a forward supply chain cannot be extended to a reverse/ closedloop supply chain. The delivery performance of a supply chain is maximized largely by synchronizing the internal business processes such that the final product fits in the customerspecified delivery window with a very high probability. green image. Synchronization in a supply chain means reducing the variability among the internal business processes or partners such that each stakeholder in the supply chain acts in a way that is appropriately timed with the actions of the other stakeholders [2]. and delivery. and aftersale performance (reflects the ability to attract old customers to the secondhand market).16 Strategic Planning Models for Reverse and ClosedLoop Supply Chains 2. and location of facilities. remanufacturing.12 Supply Chain Performance Measurement The focus of this issue is to measure the performance of a reverse/closedloop supply chain with respect to various metrics. transportation. in a reverse supply chain. 2.10 Selection of Secondhand Markets The focus of this issue is to select the market with the most potential in which to sell a repaired used product from a set of candidate secondhand markets. The main criteria for evaluation are beforesale performance (reflects the ability to attract new customers to the secondhand market). 2. disassembly. such as procurement. inspection.
Strategic Planning of Reverse and ClosedLoop Supply Chains 17 2. 2006.. P. 2. References 1. 3rd ed. S. International Journal of Logistics Systems and Management 2:120–41. New York: Prentice Hall. LLC ... Antony. J. Chopra. 2006. R. Design of synchronized supply chains: A genetic algorithm based Six Sigma constrained approach. and Meindl. © 2009 by Taylor & Francis Group.13 Conclusions This chapter gave an overview of various issues faced by strategic planners of reverse and closedloop supply chains in various decisionmaking situations. K. Supply chain management. M. Swarnkar. and Tiwari.
4 gives some conclusions. Disassembly can be of two types: nondestructive or destructive.. See [15]. and supply chain operation (also called operational planning). LLC .1 Introduction As explained in chapter 1. Nondestructive disassembly is the process of systematic removal of the constituent parts from an assembly in a manner by which there is no impairment of the parts during the process. whereas in partial disassembly. Disassembly is defined as a systematic method of separating a product into its constituent parts. destructive disassembly involves separating materials from an assembly with an objective of sorting the different material types for recycling.3 Literature Review 3. a supply chain involves three phases of decision making: supply chain design (also called strategic planning). This chapter is organized as follows: section 3.2 presents a review of literature about operational planning in reverse and closedloop supply chains. tactical planning. subassemblies or other groupings through a series of operations. 19 © 2009 by Taylor & Francis Group. strategic planning.3 gives a review of literature about strategic and tactical planning in reverse and closedloop supply chains. viz. a brief review of the literature related to the three phases of decision making in reverse and closedloop supply chains. 3. supply chain planning (also called tactical planning). and operational planning. Disassembly can be complete or partial. On the contrary. the used product is not fully disassembled. and section 3. is presented. In most of the literature for reverse and closedloop supply chains. section 3. and only certain parts or assemblies are recovered. tactical planning is addressed in conjunction with strategic planning. components. the used product is fully disassembled. In a complete disassembly. In this chapter.2 Operational Planning of Reverse and ClosedLoop Supply Chains Disassembly marks the first step in reprocessing of used products.
Lambert and Gupta’s book [49] on disassembly modeling presents disassembly in the context of the entire product life cycle. They use a goal programming technique to study the sensitivity of several parties (disassemblers. storage.20 Strategic Planning Models for Reverse and ClosedLoop Supply Chains [29]. Guide et al. maintenance and repair. Boon et al. and reprocessing. when solved. In order to obtain the optimal disassembly process strategy. or disposal to meet the demand for those products/components. After disassembly. Pnueli and Zussman [64] propose a methodology that includes identifying reprocessing options of all the parts of a product and improving the product’s design. design for disassembly. concurrent design. Penev and De Ron [63] discuss the determination of optimal disassembly level and sequences of a used product. Gungor and Gupta [28] provide an exhaustive review of literature in the areas of environmentally conscious manufacturing and product recovery that includes several aspects of environmentally conscious manufacturing (design for environment. [22] identify the problems associated with production planning issues in a closedloop supply chain and advocate that each closedloop supply chain differs from every other closedloop supply chain and a “one size fits all” approach does not work. refurbished. etc. remanufacturing). etc. recyclers. They apply Hayes and Wheelwright’s productprocess © 2009 by Taylor & Francis Group.) to different factors involved in the endoflife processing. and reverse supply chain. The book also presents a comprehensive discussion of the theories and methodologies associated with disassembly. the authors utilize graph theory and cost analysis. and directed to the part/subassembly inventory for reprocessing operations. recycling. Each type of system offers different managerial concerns. and the authors incorporate realworld case examples to explore the three main areas of application: assembly optimization. [14] study the economic viability of recycling infrastructure in the United States in relation to the electronics industry. LLC . and [59] for a comprehensive list of issues related to disassembly planning and scheduling. which. It examines disassembly on the intermediate level. A disassembly line is perhaps the most suitable setting for disassembling large products. [48]. etc. Lambert [50] proposes a methodology to identify the disassembly level in an economically optimal way. which provide conditions for the generation of profit while considering the environmental impact. reusable parts/subassemblies are cleaned. Gungor and Gupta [27] discuss the importance of a disassembly line and identify the critical issues and complexities and their effects on the disassembly line. disassembly process planning (production planning and inventory control issues). 35] propose several multicriteria decisionmaking methodologies in a disassemblytoorder setting under different decisionmaking environments. tested. common issues in product recovery (recycling. provide the number of used products to be taken back for reuse. incorporating design for disassembly. Kongar and Gupta [40] and Imtanavanich and Gupta [34.).
In a © 2009 by Taylor & Francis Group. In a later paper. Mabini et al. etc. LLC . Richter [67] considers a mixture of two pure policies. Some authors consider deterministic models in which supplyanddemand rates are constant. remanufactured products. thereby extending the productprocess matrix to include the insights on production planning in closedloop supply chains. They provide a database of articles published in the area of reverse logistics in the period 1995–2005 by exploring the topic. total disposal and total remanufacturing in a deterministic system. Guide and Srivastava [24] list the factors that complicate the inventory control and production planning in a reverse supply chain. and remanufacturetoorder environments. Richter [68. recycled goods. [70] review an extensive list of articles on reverse logistics published in the production and operations management field. thus leading to stochastic routings and lead times Partmatching problem during the assembly process Added complexity of a remanufacturing shop structure Uncertainty in supply rate of used products Problem of imperfect correlation between supply of used products and demand for reprocessed goods Inventory control models in a reverse supply chain are required to keep track of various goods through the chains. 69] proposes a similar model under a different control policy where he provides formulation to calculate optimal values for control parameters and analyzes their dependence on the supply rate of used products. Inventory control and production planning methods are thoroughly understood and well established for traditional forward supply chains. as well as other relevant aspects of research. However. viz. available techniques for a traditional forward supply chain are not transferable and adequate for a reverse supply chain.Literature Review 21 matrix framework to examine three cases representing remanufacturetostock. [58] provide an extensive review of literature in the area of reverse logistics as well as an overview of definitions and research opportunities in this area. which implies a high degree of uncertainty in material planning Unknown conditions of recovered parts until inspected. [57] consider such a model with fixed setup costs for orders and remanufacturing and linear holding costs for used products and finished goods. the model includes understock servicelevel constraint and machine sharing during the remanufacturing stage. methodology. used products.. Also. Meade et al. reassembletoorder. Inventory models for repairable items and maintenance systems where failed machines are replaced by warm or cold spares carry some similarities to the reverse supply chain models [11. 16]. and techniques of analysis. Rubio et al. These factors are as follows: Probabilistic recovery rate of parts from the used products.
The author assumes zero repair times and does not consider procurement lead times. They consider fixed lead times and no disposal of used products. The model optimizes the tradeoff between holding and shortage costs. Van der Laan et al. the associated inventory holding costs for remanufactured products are lower than those for new items. Teunter [78] studies a deterministic system with continuous review and no lead times for outside procurement or remanufacturing. Simpson [75] considers a periodic review model for separate inventories for serviceable and recoverable parts. He shows that a threeparameter control policy that controls demand. In repairable item inventory and remanufacturing systems. Simpson’s model [75] is extended by Inderfurth [36] by considering the effects of nonzero lead times for the reprocessing process and orders. continuous review models monitor the inventory level incessantly and aim to find static control policies to obtain optimal control parameters to minimize the expected system costs. and disposal activities is optimal. The authors develop an approximate control strategy with respect to reorder points and order quantities for a singleitem product case where used products are remanufactured. For a similar model. Teunter and Vlachos [77] consider the impact of disposal on the discounted average cost. Heyman [31] analyzes the continuous review inventory control case where incoming used products are disposed of whenever the inventory position reaches a predetermined level. Kelle and Silver [38] utilize an integer program based on the net demand per period under chance constraints with fixed setup costs. LLC . stochastic supply and demand occurrences are controlled by periodic and continuous review models. The authors assume that a fixed percentage of supplied products is returned to the manufacturer after a fixed lead time. Kiesmuller and Van der Laan [39] consider a periodically reviewed system in a finite horizon with supply rates dependent on demand rates. [17] consider a periodic review model in which used products can be placed in serviceable inventory directly.22 Strategic Planning Models for Reverse and ClosedLoop Supply Chains more recent paper. Although it is assumed that remanufactured products are as good as new ones. He considers a push system for remanufacturing to avoid the storage of used products and shows that the difference between these two lead times is an important complexity factor of the system. Cohen et al. The author proposes an Economic Order Quantity (EOQ) control mechanism with fixed batch sizes and shows that in a given time period. The authors assume deterministic lead times and an independent stochastic demand process. Their model considers deterministic and equal lead times for remanufactured and new items procurement. On the other hand. singleechelon production and inventory system with used © 2009 by Taylor & Francis Group. [83] consider a singleproduct. Periodic review models aim to minimize expected costs over a finite period and obtain optimal control policies under various assumptions. In a following paper. They show that dependent supply rates provide a better inventory control performance. the optimal batch size should be equal to 1 for either manufacturing or remanufacturing. remanufacturing. Muckstadt and Isaac [60] present a production planning and inventory control model for remanufacturing.
viz. a batch strategy. Then Korugan and Gupta [42] look at the CONWIP problem of the hybrid production systems with mutually exclusive manufacturing and remanufacturing processes. The authors report that increasing supply rates decrease the expected total cost. and (3) not reproducible or reusable part. In their paper. a level strategy. 47. Each part has three attributes associated with it: (1) reusable part. as the holding cost of a used product at the lower echelon is lower than the lost sales cost. LLC . [32]. A multiobjective mathematical model formulation for a recycleoriented manufacturing system is presented by Hoshino et al. The experiments conducted by the authors demonstrate that dedicated kanban control performs better with reasonable utilization rates. This situation may increase lead times and their variability. The author adapts the kanban control mechanism to a disassembly line and develops a system that uses several types of kanbans attached to various components and subassemblies..Literature Review 23 products. [26] evaluate various part orderrelease strategies in a remanufacturing environment using a simulation model. Guide et al. Guide [25] proposes scheduling using the drum–buffer–rope concept as an alternative to the MRP method. The authors focus on the problem of uncontrolled release of parts from disassembly operations. either new methodologies have to be developed or classical methods have to be modified to manage the complications of the product reprocessing process [81]. making customer service levels decline. Salomon et al. Various types of orderrelease strategies are discussed. They model the system as a PUSHtype maketostock open queueing network. Korugan and Gupta [41] consider a twoechelon inventory system with stochastic lead times and arrival rates. [72] and Van der Laan and Salomon [82] develop and compare PUSH and PULL strategies for a joint production and inventory system using both remanufactured and new parts. Therefore. The authors generalize dynamic and static routing methods by applying an adaptive kanban control policy. Applicability of traditional production planning and scheduling methods for remanufacturing systems is extremely limited. The optimization model for the recycleoriented manufacturing system is designed for a single product with m number of parts. where along with the manufacturing and remanufacturing processes. Guide and Srivastava [23] propose a specific structure for MRP mechanics and evaluate a method to calculate safety stock for material recovery uncertainty. the authors compare the performance of three different procurement and inventory control strategies. the demand arrivals are modeled as service. which may cause long queues at machine centers. 62]. a local order © 2009 by Taylor & Francis Group. The authors apply a goal programming approach with two objectives: profit maximization and recycling rate maximization. A number of authors propose that material requirements planning (MRP) techniques utilize a reverse bill of materials (BOM) to allow the management and control of inventories in the remanufacturing environment [30. whereas adaptive kanban policy always outperforms nonadaptive policies. Udomsawat and Gupta [80] focus on the production control aspect of the disassembly environment. and disposal. (2) not reusable but reproducible part. remanufacturing.
Third. the system derives the need for parts and uses optimizing procedures to determine the disassembly schedule. the total cost is relatively insensitive to the reusability rate. Variability in the buffer sizes in the network causes a significant variation in the mean process time. Their study helps in comparing two product designs by assessing feasible combinations of components to be retrieved from a used product. Then Aksoy and Gupta [1] look at the effect of reusable rate variation on the performance of a remanufacturing system. The authors also examine the tradeoff between expanding the buffer sizes and increasing the service capacity of machines [3.24 Strategic Planning Models for Reverse and ClosedLoop Supply Chains strategy. However. Second. we present a brief survey of those models and studies (see Fleischmann’s book [20] for a detailed survey of many of the case studies): Veerakamolmal and Gupta [84] propose a costbenefit function and a measure called “design for disassembly index” that analyzes the tradeoff between the costs and benefits of endoflife disassembly to find the optimum costbenefit ratio for endoflife retrieval. Also. The higher the reusability rate. it uses the bill of material for each component directly. part commonality and different yield factors are explicitly included. with no need for modification. repair rate. and a global orderrelease strategy. LLC . The approach extends the methods used for material planning in a remanufacturing environment in several ways. breakdown rate. and service rate of the stations in the remanufacturing system. the lower the total cost. 3. testing. the reusability rate has a significant effect on the total cost. Ferrer and Whybark [19] describe the first fully integrated material planning system to facilitate managing a remanufacturing facility. © 2009 by Taylor & Francis Group. it compares the combination with the highest costbenefit from one design with those from the others. Further. The hybrid system is modeled as an open queueing network (OQN) with unreliable servers and finite buffers. mean process time (average remanufacturing time) is sensitive to the buffer size. The authors report that when the supply rate of used products is low. and remanufacturing activities. Finally. and hence only a few quantitative models and case studies have been reported in the literature. Aksoy and Gupta [2] model a hybrid manufacturing system with distinct cells for disassembly. as the supply rate becomes higher. it explicitly links the volume of used products with the volume of sales.3 Strategic and Tactical Planning of Reverse and ClosedLoop Supply Chains Strategic and tactical planning of reverse and closedloop supply chains is a relatively new area of research. In this section. 4]. and the findings are mixed as to what is the best disassembly release mechanism. First.
and processing costs. product return rate. determines the appropriate locations and capacities for the regional recovery facilities. investment. They compare three decentralized closedloop supply chain models with the manufacturer collecting the used products. taking into consideration transportation. for the selection of an alternative. Listes and Dekker [53] propose a stochastic programming–based approach for the sand recycling network to account for the uncertainties. Barros et al. Ravi et al. A logistics network that includes collection of used carpets from 25 © 2009 by Taylor & Francis Group. It is a holistic approach that links financial and nonfinancial. and locations of depots and cleaning facilities are to be determined. Their model addresses issues that include which warehouses and collection centers should be open and which warehouses should have sorting capabilities. Ammons et al. The stochastic model seeks to find a solution that is approximately balanced between some alternative scenarios identified by field experts.Literature Review Louwers et al. [8] address the design of a supply chain for recycling sand from processing construction waste in the Netherlands. [65] propose a balanced scorecard and analytic network process (ANP)–based approach to evaluate the alternative reverse logistics operations for used computers. the retailer collecting the used products. [74] study the problem of choosing the appropriate reverse channel structure for collecting the used products from customers. The locations of the sand sources are known and their supply volumes are estimated based on historical data. The authors propose a multilevel capacitated facility location model for this problem formulated as a mixedinteger linear programming model and solved via iterative rounding of LP relaxations strengthened by valid inequalities. Beamon and Fernandes [10] propose a multiperiod mixedinteger linear programming model to study a closedloop supply chain where manufacturers produce new products and remanufacture used products. and total supply chain profits. [55] consider the design of a reverse supply chain for carpet waste in Europe. [7] address carpet recycling in the United States. A fourlevel sand recycling network is considered: (1) crushing companies yielding sieved sand from construction waste. The optimal number. capacities. (3) treatment facilities cleaning and storing polluted sand. and a third party collecting the used products. The models are compared with respect to the wholesale price. The nonlinear model. LLC . (2) regional depots specifying the pollution level and storing cleaned and halfcleaned sand. and (4) infrastructure projects where sand can be reused. Savaskan et al. when solved. A continuous location model in which all costs are considered volume dependent is proposed. tangible and intangible. and internal and external factors.
it is found that the total reverse logistics costs can be reduced by up to 49%. By combining the queuing models. Facility capacity constraints are the main restrictions in view of the vast volume landfilled. as well as separation of nylon and other reusable materials while landfilling the remainder. They also discuss possible managerial options to address this problem. the amount of carpet collected from each site is to be determined. coupled with operational strategies. and the high degree of uncertainty associated with reverse logistics networks. Hu et al. transportation. [5] study the reverse logistics network for blood distribution with the American Red Cross. which include legal responses to require return flows and utilization of market incentives for carpet recycling. [13] simulate a reverse logistics network for carpet recycling to manage highly variable return flows. the authors conclude that even with the design of an efficient reverse logistics network and use of sophisticated recycling technologies. Alshamrani et al. and distribution are considered in their model. multitype hazardous wastes reverse logistics system. The critical activities that include waste collection. storage.26 Strategic Planning Models for Reverse and ClosedLoop Supply Chains carpet dealerships. the optimal number and location of both collection sites and processing plants for alternative configurations are to be determined. [33] present a costminimization model for a multitimestep. etc. Although the delivery sites for recovered materials are assumed to be known. LLC . The authors formulate a discretetime analytical model that minimizes total hazardous waste reverse logistics costs subject to constraints. some dynamic aspects such as lead time and inventory positions. including business operating strategies and government regulations. Biehl et al. The model is presented for a singleproduct. where containers in which blood is delivered from a central processing unit to customers in one © 2009 by Taylor & Francis Group. From their study. and several case examples are solved using genetic algorithms based on the technique of differential evolution. They use an experimental design technique to study the effect of system design factors as well as environmental factors that affect the operational performance of such a reverse logistics network. is investigated. The authors propose a multilevel capacitated mixedinteger linear programming model to address this problem. By using the proposed methodology. singlelevel network. Lieckens and Vandaele [51] combine queuing models with the traditional reverse logistics location model and formulate mixedinteger linear programming models to determine which facilities to open while minimizing the total cost of investment. the problem is defined as a mixedinteger nonlinear programming model. return flows cannot meet demand for nearly a decade. With these extensions. procurement. disposal. In addition. processing. are accounted for.
remanufacturing centers. The recovered products are sold under the same conditions as new ones to satisfy a given market demand. Thus. Containers not picked in the period following their delivery incur a penalty cost. Vlachos et al. where in each period the vehicle dispatcher needs to design a multistop vehicle route while determining the number of containers to be picked up at each stop. This leads to a dynamic logistics planning problem. Spengler et al. The distribution network encompasses three levels: plants. and intermediate centers. The authors assume that the demand is deterministic and the facilities are of three different types: producers. The authors propose a multilevel warehouse location model with piecewise linear costs. their study takes into account environmental issues such as takeback obligations and the “green image” effect on customer demand. The simulation model serves as an experimental tool that helps in evaluating longterm capacity planning policies using total supply chain profit as a measure of effectiveness. and hence no fixed costs are considered in the model. [76] study the recycling networks for industrial byproducts in the German steel industry. Thierry et al. A heuristic procedure is developed to solve the route design–pickup strategy problem. The problem is formulated as a linear programming model. Apart from considering economic issues. which is used for optimizing several scenarios. The objective is to determine the costoptimal flow of goods in the network under the given capacity constraints. [85] address the capacity planning issues in remanufacturing facilities in reverse supply chains through a simulation model based on the principles of system dynamics methodology. one needs to determine which recycling processes or process chains to install at which locations and their capacity levels. LLC . The problem is modeled as a 01 mixedinteger programming problem that is solved using an algorithm based on Lagrangian heuristics. Lu and Bostel [56] study the facility location problem in a remanufacturing network that has a strong interaction between the forward and reverse flows of products.Literature Review time period are available for return to the central processing unit the following period. All facilities are fixed externally. Recycling facilities with variable capacity levels and corresponding fixed and variable processing costs can be installed at a set of potential locations. warehouses. [79] propose a conceptual model for a closedloop supply chain that addresses the situation of a manufacturing company collecting used products for recovery in addition to producing and distributing new products. and markets. 27 © 2009 by Taylor & Francis Group. Remanufactured products are introduced as new ones into the forward flow. which can be solved for optimality.
transportation. the optimal number and locations of remanufacturing facilities. Besides maximizing overall profit. © 2009 by Taylor & Francis Group. [37] analyze the logistics network of an electronic equipment remanufacturing company in the United States. minimizing environmental impacts by minimizing energy consumption is considered. remanufacturing. processing. (2) preparation. [43] report a case study concerning the implementation of a remanufacturing process at a copier manufacturing company in the Netherlands. The model is illustrated in a fictitious case of a computer manufacturer. The authors propose a multiproduct capacitated warehouse location mixedinteger linear programming model that is solved to optimality for different supplyanddemand scenarios. [52] propose a mixedinteger programming model that takes into account multiperiod planning horizons with uncertainties for a product with modular design and multiproduct configurations. which encompasses the inspection and replacement of critical components. The company’s activities include collection of used products from customers. They propose a nonlinear mixedinteger programming model for optimizing processing decisions in electronics recycling operations. are to be determined while considering investment. The reverse supply chain is subdivided into three main stages: (1) disassembly of return products to a fixed level. [79] and propose a conceptual model for extending an existing forward supply chain with disassembly centers to allow for recovery of used products. LLC . the number.28 Strategic Planning Models for Reverse and ClosedLoop Supply Chains Berger and Debaillie [12] address a situation similar to that of Thierry et al. Lim et al. the authors propose a multilevel capacitated mixedinteger linear programming model. individually. Although the facilities in the forward supply chain are fixed. the optimal solution minimizing operational costs is compared with a number of preselected managerial solutions. and storage costs. and capacities of the disassembly centers are to be determined. and the number of used products collected. and (3) reassembly of the remaining carcass together with repaired and new components into a remanufactured machine. Krikke et al. In this network. Reimer et al. Although the supplying processes and disassembly are fixed. and material processors. To this end. After inspection. and distribution of remanufactured products. generators. In a variant of this model. optimal locations and flow of goods are to be determined for both the preparation and the reassembly operations. [66] model the economics of electronics recycling from the perspective of recyclers. the recovery network is extended to another level by separating inspection and disassembly/repair centers. Jayaraman et al. locations. Based on a mixedinteger linear programming model. rejected items are disposed of. whereas recoverable items are sent to the disassembly centers.
plastic. or reusing the materials. The usual collection system in the European Union countries is composed of two phases. The authors propose a genetic algorithm and a GRASP heuristic to solve the problem. and increased warranty claims. organic material) are stored in special refuse bins. based on the branchandcut procedure known as the integer Lshaped method. delivering. Salema et al. The problem is formulated as a mixedinteger linear programming model that is closely related to a classical uncapacitated warehouse location model. A decomposition approach is suggested to solve this model. capacity limits. a major question is where to locate the depots for empty containers. In addition to determining the number of containers required for running the system and an appropriate fee per shipment. [71] point out that the majority of the quantitative models that exist in the area of reverse supply chain design are case specific and hence lack generality. The authors propose several options for a closedloop supply chain: reusing the product as a whole. These returns are increasing due to trends such as product leasing. Krikke et al. which may be constructed based on critical levels of design parameters such as demand or returns. An additional requirement is balancing the number of containers at the depots. citizens leave their refuse at special collection areas where different types of waste (glass. each type of waste is collected separately and moved to its final destination (a recycling plant or refuse dump). Listes [54] presents a generic stochastic model for the design of networks constituting both supply and return channels. a logistics service provider responsible for storing. To this end. [45] focus on commercial returns that have nothing to do with environmental legislation. shorter product replacement cycles. senders and recipients of full containers. and they propose a mixedinteger formulation that is solved using standard Branch and Bound (B&B) techniques. depositbased system is considered for collapsible plastic containers that can be rented as secondary packaging material. LLC . catalog/Internet sales. and carriers transporting full containers from sender to recipient. The actors involved in the system include a central agency owning a pool of reusable containers. they propose a generic reverse supply chain model that incorporates multiproduct management. The model accounts for a number of alternative scenarios. More specifically. a closedloop. The study focuses on the problem of locating these collection areas. First. and uncertainty in product demands and returns. Bautista and Pereira [9] address reverse logistics problems arising in municipal waste management. reusing the components. 29 © 2009 by Taylor & Francis Group. paper. Subsequently. and collecting the empty containers.Literature Review Kroon and Vrijens [46] consider the design of a logistics system for reusable transportation packaging. organized in a closedloop system.
The authors show that the buyback payments transferred to the retailers for postconsumer goods provide a wholesale pricing flexibility that can be used to pricediscriminate between retailers of different profitability. Based on these insights and strategic factors and subfactors.30 Strategic Planning Models for Reverse and ClosedLoop Supply Chains Dowlatshahi [18] identifies the present state of theory in reverse logistics by formulating the propositions for strategic factors. [44] develop quantitative modeling to support decision making concerning both the design structure of a product and the design structure of the logistic network. in which the manufacturer collects used products directly from the consumers. The strategic factors are delineated and evaluated in terms of specific subfactors associated with each factor by the use of interview protocol and within the context of an indepth analysis of two companies in different industries that are engaged in remanufacturing/recycling operations within reverse logistics systems. The customers’ preferences with regards to purchasing and returning the product are incorporated via a discrete choice model with stochastic utilities. [86] study the interplay between industrial firms and government concerning the collection of used products from households. as well as the importance of the supporting role of effective reverse logistics operations for the successful and profitable execution of repair service activities. LLC . particularly repair services. Savaskan and Van Wassenhove [73] focus on the interaction between a manufacturer’s reverse channel choice to collect postconsumer goods and the strategic product pricing decisions in the forward channel when retailing is competitive. This framework allows for the determination of the viability of used products/parts. Wojanowski et al. The approach used is grounded theory development. in which the retailers act as product return points. A continuous modeling framework is presented for designing a dropoff facility network and determining the sales price that maximizes the firm’s profit under a given depositrefund. [6] discuss the competitive value of service management activities. Also. The authors focus on the use of a depositrefund requirement by the government when the collection rate voluntarily achieved by the firms is deemed insufficient. Krikke et al. and an indirect product collection system. they model a direct product collection system. Environmental impacts are © 2009 by Taylor & Francis Group. a framework for effective design and implementation of remanufacturing/recycling operations in reverse logistics is provided. Amini et al. they present a case study of a major international medical diagnostics manufacturer to illustrate how a reverse logistics operation for a repair service supply chain was designed for both effectiveness and profitability by achieving a rapidcycle time goal for repair service while minimizing total capital and operational costs. To this end.
Gautam and Kumar [21] describe a multiobjective evaluation of the tradeoffs among the number and size of dropoff stations. allocation. © 2009 by Taylor & Francis Group. Boston. recycling options. pp. 2000. New Orleans. Effect of reusable rate variation on the performance of remanufacturing systems. and Gupta. A geographical information system (GIS)–based model is proposed for the design of a solidwaste system considering waste generation. 3. and operational planning—was presented. The variational inequality formulation allows for the formulation of the complex reverse supply chain network to obtain the endogenous equilibrium prices and material flows between tiers. LLC . population covered in a service network. M. The decision makers consist of the sources of electronic wastes. Aksoy. 2. S.. 13–20. the recyclers. H. H. pp. H. and consumers associated with the demand markets for the distinct products. Buffer allocation plan for a remanufacturing cell. M. Economic costs are modeled as linear functions of volumes with a fixed setup component for facilities. a brief review of the literature related to the three phases of decision making in reverse and closedloop supply chains—strategic planning. S. 1999. S.. 2005. 31 3. This model is applied to a closedloop supply chain design problem for refrigerators using reallife data of a Japanese consumer electronics company. and Gupta. A multitiered electronic recycling network equilibrium model is constructed. 48. and Gupta. average walking distance to dropoff stations by the population. No. K. 62–65. which includes recycling. 3.. An open queueing network model for remanufacturing systems. Aksoy. the processors. K. tactical planning. The authors also establish the variational inequality formulation. and location of dropoff stations. Nagurney and Toyasaki [61] develop an integrated framework for the modeling of reverse supply chain management of electronic waste. M.4 Conclusions In this chapter. References 1. In Proceedings of the 25th Conference on Computers and Industrial Engineering. vol. 657–677. K.Literature Review measured by linear energy and waste functions. and the distance traveled by collection vehicles. Aksoy. In Proceedings of the SPIE International Conference on Environmentally Conscious Manufacturing. Computers & Industrial Engineering.
and Bienstock. Building contingency planning for closedloop supply chains with product recovery. V. Alshamrani. M.. and Whybark. B. M. K. Production and Operations Management 10:112–24.. 13. An application to the metropolitan area of Barcelona. Jayaraman. and Pereira. Dekker. In Proceedings of the SPIE International Conference on Environmentally Conscious Manufacturing II. 19. H. V. J. and Taleb. Boon. J. Berlin: SpringlerVerlag. Guide. S... Designing a reverse logistics operation for short cycle time repair services.. 5. European Journal of Operations Research 51:1–23. A. M. and Parlar. © 2009 by Taylor & Francis Group. J. and Gupta. Master’s thesis. Ferrer. D. European Journal of Operations Research 4:136–42. European Journal of Operational Research 110:199–214. 11. Modeling the problem of locating collection areas for urban waste management. Quantitative models for reverse logistics: Lecture notes in economics and mathematical systems... and Scholten. University of Leuven.. 6. International Journal of Operations and Production Planning 14:57–67. 18. J. 2005. Ammons. Gautam. 2005. A. Material planning for a remanufacturing facility.32 Strategic Planning Models for Reverse and ClosedLoop Supply Chains 4. Belgium. Mathur. M. Cho. M. Capacity and buffer tradeoffs in a remanufacturing system... I. M. Isaacs. E. M. 10. Production Planning and Control 15:270–81. Naval Research Logistics Quarterly 27:289–96. Journal of Electronics Manufacturing 11:81–93. Gupta. C. R. 12. LLC . D. P.. H. D. 1991. A twolevel network for recycling sand: A case study. Beamon. Clean Technologies and Environmental Policy 7:306–16. 2006. D. 167–74. Computers and Operations Research 34:595–619. M.. S. 9.. S.. 2004. and Kumar. Bautista. Prater. 2001. Reverse logistics: Simultaneous design of delivery routes and return strategies.. and Realff. M. 20. T. 1998. Berg. 2002. Berger.. S. Supplychain network configuration for product recovery. MA. A survey of maintenance models for multiunit systems. S. 2001. Cohen. W. Economic sensitivity for end of life planning and processing of PCs. R. Location of disassembly centers for reuse to extend an existing distribution network (in Dutch). Carpet recycling: Determining the reverse production system design. Dowlatshahi. and Ballou. K. Fleischmann. 21. C. A strategic framework for the design and implementation of remanufacturing operations in reverse logistics. Aksoy. International Journal of Production Research 43:3455–80. Barros.. A. M.. 1994. Biehl. and Gupta. I. Strategic planning of recycling options by multiobjective programming in a GIS environment. 22. 16. and Fernandes. Brennan. J. V. Computers and Industrial Engineering 34:443–63. J. Assessing performance and uncertainty in developing carpet reverse logistics systems. G. D.. PolymerPlastics Technology and Engineering 38:547–67. C. J. Realff. M. A. N. A.. Newton. 17. 2007. and Pierskalla... International Journal of Production Economics 96:367–80. D. J. 1999. C.. 1980. Nahmias. 2007. 1997. 2003. Journal of Operations Management 21:259–79. Amini. M. B. 1980. L. 2005. 14. K. C. Jr. A. A marginal cost analysis for preventive replacement policies. Operations planning issues in an assembly/disassembly environment. 7.. pp. and Linton. M. S. J. and Debaillie. R.. A dynamic inventory system with recycling. and Newton. 8. 2001.. RetzlaffRoberts. 15. K. E. Omega 34:617–29.
S. M. Srivastava.. Purchasing policy of new containers considering the random returns of previously issued containers. M. M. and Srivastava. and Gupta. 1994. Imtanavanich. 1998. 1996. 35. K. S. T.. LLC . Guide. Guide.. Computers and Industrial Engineering 31:225–28. P. T.. V. S. International Journal of Production Research 33:855–60. M. 1999. and Srivastava. International Journal of Production Research 40:2569–89. International Journal of Production Research 32:1857–66. 2002. M. Transportation Research 38E:457–73.. vol. Repairable inventory theory: Models and applications. vol. 26. Journal of the Operational Research Society 50:497–508. A multiechelon inventory system with returns. S. N. R. P. Inderfurth.Literature Review 33 23. E. M. D. and Hitomi. A closedloop logistics model for remanufacturing. 41. Scheduling using drumbufferrope in a remanufacturing environment. Yura. International Journal of Production Economics 72:73–87. S. P. D. P. and Gupta. Jr. International Journal of Production Research 34:1081–91. Jayaraman. S. D. and McLean. 32. K. Guide. R. European Journal of Operations Research 102:1–20.. 5583.. A. Guide. S. R. © 2009 by Taylor & Francis Group. J. R.. S. In Proceedings of SPIE—The International Society for Optical Engineering. Gupta.. V. G. Guide. Jr. A. pp.. Disassembly line in product recovery. An inventory model with dependent product demands and returns. In Proceedings of SPIE—The International Society for Optical Engineering.. V. K. and Taleb. Gungor. Imtanavanich. R. Environmentally Conscious Manufacturing IV. 2005. Naval Research Logistics Quarterly 24:385–405. 33. Optimal disposal policies for a single item inventory system with returns. 2001. 27. Disassembly of products. K.. and Gupta. 1997.. M. C. An evaluation of capacity planning techniques in a remanufacturing environment.. 1989. R. 147–62. 31.. 599–702. Heyman. A. 24. An evaluation of order release strategies in a remanufacturing environment. R. 29.. International Journal of Production Research 35:67–82.. M. Jr. Computers and Operations Research 24:37–47. D. Optimization analysis for recycleoriented manufacturing systems. 34. 25. Korugan. Kiesmuller. 38. 1997. Kelle. Simple optimal replenishment and disposal policies for a product recovery system with leadtimes. S. and Huan. 1996. R. Gupta.. A reverse logistics cost minimization model for the treatment of hazardous wastes. 2002. and Gupta. K. Hoshino. E. Jr. and Van der Laan. 39. 30.. and Gupta. R. 37. D. Computers and Industrial Engineering 35:145–48.. Environmentally Conscious Manufacturing V. 1999. 2002. M. 2004.. 28. and Spencer. V. Kongar. 40. and Srivastava. E. 5997.. V. Journal of Electronics Manufacturing 11:171–83. Sheu. pp.. Multicriteria decision making approach in multiple periods for a disassemblytoorder system under product’s deterioration and stochastic yields. 1995. P. Jr. V.. A. 1997. Hu. IIE Transactions 21:349–54. Scheduling disassembly. and Gupta. OR Spectrum 19:111–22. Multicriteria decision making for disassemblytoorder system under stochastic yields. 1997. R. 1997. 36. D. Gungor. Issues in environmentally conscious manufacturing and product recovery: A survey. Computers and Industrial Engineering 36:811–53.. A multicriteria decision making approach for disassemblytoorder systems. and Silver.
Reverse supply chain management and electronic waste recycling: A multitiered network equilibrium framework for ecycling. and Gupta. Listes. and Vandaele. Journal of Electronics Manufacturing 11:121–35. and Vrijens. F. pp. 2005. Moyer. Krikke. N. Boca Raton. reuse. Business case: Reverse logistic network redesign for copiers. 44. In Proceedings of the International Conference of Simulation and Modeling. C. A. 1995. K.. International Journal of Production Economics 28:21–33. EOQ type formulation for controlling repairable inventories. Returnable containers: An example of reverse logistics. F. Optimal disassembly of complex products. 1999. Computers and Operations Research 34:299–323. Environmental concerns and recycling/disassembly efforts in the electronic industry. European Journal of Operational Research 160:268–87.. 46.34 Strategic Planning Models for Reverse and ClosedLoop Supply Chains 42. and recycling. N.. Kasumastuti. Concurrent product and closedloop supply chain design with an application to refrigerators. R. J. Z. and Isaac. Meade. 1999. O. 58. Kroon. H... Computers and Industrial Engineering 36:855–69. Lieckens.. 48. Lim. 2003. and Gupta. M.. Product modularity and the design of closedloop supply chains. and Piplani. 43. 1993... D. H. The theory and practice of reverse logistics. A stochastic approach to a case study for a product recovery network design. J. Muckstadt. 2005. Lambert. L.. H. P. D. M. FL: CRC Press. An analysis of single item inventory systems with returns. L. Korugan. and Dekker. and Schuur. maintenance. Lambert. J. Designing a reverse supply chain network for product refurbishment. In Proceedings of the International Conference on Environmentally Conscious Manufacturing II. R. and Bostel.. L. 1997.. 50.. Structuring bills of material for automobile remanufacturing. 47. D. A. 2002. J. 2007.. D. S. 2007. California Management Review 46:23–39. A. and Presley. M. 2004. International Journal of Production Research 35:2509–23. L. S. Pintelon. Nagurney.. 51. and Gupta. L. R. Leblanc. P.. 175–82. 52. Production and Inventory Management Journal 34:46–52. Reverse logistics network design with stochastic lead times.. N. Krikke. 2005. S. and van de Velde. R. LLC . Krikke. S. Souren. 45. 57. H. Lu. M... D. Computers and Operations Research 34:417–42. S. J. Louwers. 61.. G. L. M. International Journal of Logistics Systems and Management 3:56–84. 60. 2007. 53. Krupp. J. Sarkis. A. OR Spectrum 21:381–409. A facility location allocation model for reusing carpet materials. B.. 1992. A. F. Listes. I. International Journal of Production Research 41:3689–719. H. A. and Flapper. BloemhofRuwaard. 56. Mabini. J. and Gupta.. A generic stochastic model for supplyandreturn network design. 54. K.. C. G. Transportation Research 41E:1–28. A facility location model for logistics system including reverse flows: The case of remanufacturing activities. S. O. and Toyasaki. Disassembly modeling for assembly. 2007. and Van Wassenhove. 2005. J. M. 59. Computers and Operations Research 34:395–416. D. Naval Research Logistics Quarterly 28:237–54. © 2009 by Taylor & Francis Group. A. 1997. Journal of Electronics Manufacturing 7:1–22. E. 1981. 55. M. 49. 2001. Lambert.. and Gelders. Demanddriven disassembly optimization for electronic products.. Kip. Van Harten. Peters. An adaptive kanban control mechanism for a single stage hybrid system. International Journal of Physical Distribution and Logistics Management 25:56–68..
The EOQ repair and waste disposal model with variable setup numbers. D. 2006. E. 65.. 1997. M. Gupta and A. 1988.. Reverse channel design: The case of competing retailers. Richter. T. M.. LLC . Pure and mixed strategies for the EOQ repair and waste disposal problem.. and Miranda. W. Boca Raton. Teunter. 69. Naval Research Logistics 48:484–95. Operations Research 26:270–81. 2007. 2004.. J. O. European Journal of Operations Research 97:308–26. Thierry.. A. L. and Van Wassenhove. J. K. N. Computers and Industrial Engineering 48:327–56.. On the necessity of a disposal option for returned items that can be remanufactured. P. Penev. N. 70. Pnueli. 2008. FL: CRC Press.. and Van Wassenhove Luk. H. Richter. D. R. K. Analyzing alternatives in reverse logistics for endoflife computers: ANP and balanced scorecard approach. J. Udomsawat. An optimization model for the design of a capacitated multiproduct reverse logistics network with uncertainty. K. Simpson. 68. Spengler. 80.. B.. MRP II for repair/refurbish industries. International Journal of Production Research 34:495–506. N. California Management Review 37:114–35. Ravi. R. Erasmus University.. N. V. and Rentz. Puchert. Environmental integrated production and recycling management. Environment conscious manufacturing. K. and Knight. Salomon. and Ridder. Lambert (Eds. Richter. M. A. Rotterdam. M. 1995. 2000. 2001.. Determining of disassembly strategy. Teunter. Production and Inventory Management Journal 29:12–15. 2005. 64. Product remanufacturing and its effect on production and inventory control. L. A. Chamarro. Panisset. and Tiwari.. 2002. Reimer. 63. F. 1996. European Journal of Operational Research 179:1063–77. M. The extended EOQ repair and waste disposal model. Salomon. V. J. Strategic issues in product recovery management. 79. 1997.. pp. International Journal of Production Economics 45:443–48. European Journal of Operational Research 95:313–24. G. Optimizing electronics endoflife disposal costs. H. Evaluating the endoflife value of a product and improving it by redesign. R. 74. In S. 2008. S. © 2009 by Taylor & Francis Group. 76. and van Wassenhove. D. van der Laan. and Novais.. A. A.Literature Review 35 62. B. I. S. Savaskan. 73. R. Characteristics of research on reverse logistics (1999–2005). 75. Optimum solution structure for a repairable inventory problem. Y. BarbosaPovoa. T. The Netherlands. 1997. and Gupta.. A.. Rubio.. 342–47. 1996... MultiKanban system for disk assembly line. 71. K. OR Spectrum 19:123–29. Sodhi. Economic ordering quantities for remanufacturable item inventory system. M. S. 1994. S... M. G.. Q. and De Ron. Thierry. A. and Vlachos. and Zussman. 67. C. Management Science 50:239–52. S. Bhattacharya. In Proceedings of IEEE Symposium on Electronics and the Environment. 1996. Closedloop supply chain models with product remanufacturing. Management Science 52:1–14. 1978. 78. R.). E. van Nunen.. M. 77. Savaskan. H. M. ERASM Management Report Series 172. P. International Journal of Production Research 46:1099–1120. Penkuhn. Ravi. Dekker. 66. International Journal of Production Economics 75:257–66. 72. A. Salema. International Journal of Production Research 35:921–42..
Computers and Operations Research 34:367–94. 2007.. and Iakovou. T. 1996. 83. 82. LLC . R. and Boyaci. pp. A systems dynamic model for dynamic capacity planning of remanufacturing in closedloop supply chains. Georgiadis. 86. and Salomon. Dekker. 1999.. 85... Van der Laan. Vlachos. An (s. D. M. Q) inventory model with remanufacturing and disposal.. 1997. M. In Proceedings of the IEEE International Symposium on Electronics and the Environment. E. International Journal of Production Economics 46:339–50. 285–90. E. 2007. P. Analysis of design efficiency for the disassembly of modular electronic products. and Ridder.. and Gupta. Production planning and inventory control with remanufacturing and disposal. 84. Production planning for companies with product recovery and remanufacturing capability.36 Strategic Planning Models for Reverse and ClosedLoop Supply Chains 81. R. R. © 2009 by Taylor & Francis Group. V. E. 2007. M.. Journal of Electronics Manufacturing 9:79–95. Wojanowski. P. Salomon. Uzsoy. Verter. S.. Veerakamolmal. A. European Journal of Operational Research 102:264–78. Retailcollection network design under depositrefund. Computers and Operations Research 34:324–45.. Van der Laan.
expert systems. goal programming. respectively. those criteria do not in turn depend upon subcriteria and so on. and selecting a doctoral program [4]. fuzzy logic. siting landfills [2]. quantitative models for a number of crucial issues in strategic planning of reverse and closedloop supply chains are presented. [5]). extent analysis method. method of total preferences. and linear integer programming. The AHP in such cases is conducted in two steps: (1) weigh independent criteria. The process has been formalized by Saaty [1] and is used in a wide variety of problem areas. It must be noted that only the basic concepts of each technique are presented here. using pairwise judgments. In a large number of cases (for example. LLC . Bayesian updating. section 4. Borda’s choice rule.1 Introduction In this book. and 37 © 2009 by Taylor & Francis Group. quality function deployment. technique for order preference by similarity to ideal solution (TOPSIS). 4. Depending on the decisionmaking situation. in other words. This chapter is organized as follows: Sections 4. each of which can compare two or more decision alternatives.2 Analytic Hierarchy Process and Eigen Vector Method The analytic hierarchy process (AHP) is a tool supported by simple mathematics that enables decision makers to explicitly weigh tangible and intangible criteria against each other for the purpose of resolving conflict or setting priorities. analytic network process. neural networks. fuzzy multicriteria analysis method.2 through 4. Six Sigma. linear physical programming. Taguchi loss function. Finally.19 introduce the concepts of analytic hierarchy process (including eigen vector method). geographical information systems. for example.20 gives some conclusions. each model makes use of one or more of the quantitative techniques introduced in this chapter. evaluating employee performance [3]. the tangible and intangible criteria are considered independent of each other.4 Quantitative Modeling Techniques 4.
1.1. This step is called the eigen vector method if an eigen vector is the employed mathematical technique.38 Strategic Planning Models for Reverse and ClosedLoop Supply Chains (2) compute the relative ranks of decision alternatives using pairwise judgments with respect to each independent criterion. Computation of the relative ranks: Pairwise judgments of importance using the scale shown in table 4. or row geometric mean.1.1) where max is the principal eigen value of the matrix of comparative importance values. These judgments are obtained with respect to each independent criterion considered in step 1. 4. mean transformation. The degrees of consistency of pairwise judgments in steps 1 and 2 are measured using an index called the consistency ratio (CR). For CR values greater than 0. The resulting matrix of comparative importance values is used to rank the decision alternatives by employing mathematical techniques like eigen value.1 are computed for the decision alternatives as well. Perfect consistency implies a value of zero for CR. as human beings. perfect consistency cannot be demanded because. or row geometric mean. However. it is considered acceptable if CR is less than or equal to 0. Therefore. 8 Definition Equally important Moderately more important Strongly important Very strongly more important Extremely more important Intermediate judgment values © 2009 by Taylor & Francis Group. the pairwise judgments must be revised before the weights of criteria and the ranks of decision alternatives are computed.1 Scale for pairwise judgments Comparative importance 1 3 5 7 9 2. and R is TABLE 4. n is the number of rows (or columns) in the matrix. CR is computed using the formula CR ( max n ) ( n 1)( R ) (4. The resulting matrix of comparative importance values is used to weigh the independent criteria by employing mathematical techniques like eigen value. 2. LLC . 6. we are often biased and inconsistent in our subjective judgments. Computation of relative weights of criteria: AHP enables a person to make pairwise judgments of importance between independent criteria with respect to the scale shown in table 4. mean transformation. 1.
The AHP is illustrated in the form of a hierarchy of three levels. Compared to AHP. where two main criteria are simultaneously compared with respect to the problem objective. evaluating connection types in design for disassembly [8]. Whereas AHP assumes a unidirection hierarchical relationship among the decision levels. as it does not require a strict hierarchical structure. but reallife situations warrant against such assumption.41 9 1.12 6 1. The relevant criteria. ANP allows for dependence within a set of criteria (inner dependence) as well as between sets of criteria (outer dependence). ANP goes beyond AHP [7].2 shows various R values for n values ranging from 1 to 10.24 7 0. The looser network structure in ANP allows the representation of any decision problem. ANP is used in a wide variety of problem areas (for example. and leagile supply chain [9]). analyzing alternatives in reverse logistics for endoflife computers [7].32 8 1. ANP requires more calculations and requires a more careful track of the pairwise judgment matrices. irrespective of which criteria come first or which come next. Also.2 Random index value for each n value n R 1 0 2 0 3 0.49 the random index for each n value that is greater than or equal to 1. LLC . and the last level contains the decision alternatives. the subcriteria.90 5 1. the decision maker is asked to carry out a series of pairwise comparisons with respect to the scale shown in table 4.3 Analytic Network Process Analytic network process [6] (ANP) generalizes the AHP. two subcriteria are © 2009 by Taylor & Francis Group. Table 4. AHP assumes independence among the criteria and subcriteria considered in the decision making.58 4 0. The steps involved in the ANP methodology are as follows: Step 1: Model development and problem formulation: In this step. ANP allows for a more complex relationship among decision levels and attributes.1. 4. and alternatives are chosen and structured in the form of a control hierarchy. and modeling the metrics of lean. Step 2: Pairwise comparisons: In this step. therefore.Quantitative Modeling Techniques TABLE 4.45 10 39 1. where the first level contains the primary objective. agile. the decision problem is structured into its constituent components. the second level contains the independent criteria. an important feature of the AHP is that the tangible and intangible criteria in the second level must be chosen in such a way that they can somehow help the decision maker in comparing two or more decision alternatives.
It is a partitioned matrix. The super matrix M is made to converge to obtain a longterm stable set of weights. where k is an arbitrarily large number. and pairwise comparisons are performed to address the interdependencies among the subcriteria.” “not very clear. carry a touch of imprecision with them. if the fuzziness is not taken into account. This imprecision or vagueness in human judgments is referred to as fuzziness in the scientific literature. The desirability index. Step 3: Super matrix formulation: The super matrix allows for a resolution of interdependencies that exist among the subcriteria. for alternative i is defined as J Kj D I Pj Akj Akj Sikj j 1 k 1 DI (4. where each submatrix is composed of a set of relationships between and within the levels as represented by the decision maker’s model. © 2009 by Taylor & Francis Group. or row geometric mean. and Sikj is the relative impact of alternative i on subcriterion k of main criterion j.40 Strategic Planning Models for Reverse and ClosedLoop Supply Chains simultaneously compared with respect to their main criteria. As the decisionmaking problem’s intensity grows. Step 4: Selection of the best alternative: The selection of the best alternative depends on the desirability index. 4. DI. LLC . For convergence. this imprecision leads to results that can often be misleading. The relative matrix of comparative importance values is then used to weigh the criteria using mathematical techniques like eigen vector.” and “very likely. that is done by raising M to the power of 2k+1. M must be made column stochastic. Akj is the stabilized relative importance weight (determined by the super matrix) for subcriterion k of main criterion j for interdependency (I) relationships among subcriteria.” which are often heard in daily life.2) D where Pj is the relative importance weight of main criterion j. mean transformation. Zadeh [10] first proposed fuzzy logic. after which an increasing number of studies have dealt with fuzziness in problems by applying fuzzy logic. Akj is the relative importance weight for subcriterion k of main criteI rion j for the dependency (D) relationships among subcriteria.4 Fuzzy Logic Expressions such as “probably so.
low. the larger the degree of membership of x. each representing a quantity of a linguistic value associated with a degree of membership of either 0 or 1. m. as shown in figure 4. A ( x ) [0. Zadeh proposed a membership function to deal with quantifying vagueness.2 shows a graphical depiction of a TFN. good. If a quantity has a degree of membership equal to 1. Thus. A.” Fuzzy logic is primarily concerned with quantifying vagueness in human perceptions and thoughts. On the other hand. A triangular fuzzy number (TFN) [11] is a fuzzy set with three parameters (l.0 0. X is essentially a set of real numbers. X. The transition from vagueness to quantification is performed by the application of fuzzy logic. and largest possible quantities that describe the linguistic value. and medium. For example. 1] by means of a membership function. so its linguistic value can be “very tall” or “very short. this reflects a complete fitness between the quantity and the linguistic value.2 Triangular fuzzy number. and u denote the smallest possible. LLC . where A is the degree of membership.1] . the value of a membership function represents the possibility of a fuzzy event. © 2009 by Taylor & Francis Group. people use linguistic values like high. u). The parameters l. m. Figure 4. then that quantity does not belong to the linguistic value. to describe those factors.0 l m Parameters of TFN FIGURE 4. most promising. over the universe of quantified linguistic values. height may be a factor with an imprecise value. The membership function looks like a typical cumulative probability function. The more x fits A. Each quantified linguistic value is associated with a grade membership value belonging to the interval [0. of a quantity x of the linguistic value.1 Application of fuzzy logic. if the degree of membership of a quantity is 0. whereas the value of a cumulative probability function represents the cumulative probability of a statistical event.1. Degree of Membership 1. a fuzzy set can be defined as x X . ranging from 0 to 1. however. Fuzzy Logic Vagueness Quantiﬁcation FIGURE 4.Quantitative Modeling Techniques 41 When dealing with factors with uncertain or imprecise values.
and (2) when a fuzzy number to be operated on has negative parameters (in other words. and while x increases from m to u. u z) addition (4. m y.7) Defuzzification is a technique to convert a fuzzy number into a crisp real number.3) ( x l ) / (m l ). m 0.6) P1 P2 l m u . z y x division (4. The basic operations on TFNs are as follow [11. m. which is not a TFN. P x l x x m u (4. for example. 0.5) P1 P2 (l x.6) leads to (1. . The centerofarea method [13] converts a fuzzy number P = (l. we get a TFN only. the corresponding membership function linearly increases from 0 to 1. y. and hence we defuzzify (–1. m. 12]. we make sure that upon performing an arithmetic operation on a TFN. has linear representations on its left. l (u x ) / (u m). u) and P2 = (x. There are several methods available for this purpose. the corresponding membership function decreases linearly from 1 to 0. LLC . 1) using equation (4. © 2009 by Taylor & Francis Group. u z) multiplication (4.8) Defuzzification might be necessary in two situations: (1) when comparison between two fuzzy numbers is difficult to perform. u) into a crisp number Q. where Q (u l ) (m 3 l) l (4. u x) subtraction (4. P1 = (l. P. z): P1 P2 (l x. m y. 0. 1) before squaring it). 0.4) P1 P2 (l z. 1). x u For each quantity x increasing from l to m.42 Strategic Planning Models for Reverse and ClosedLoop Supply Chains Each TFN.and righthand side such that its membership function can be defined as 0. squaring the TFN (–1. m y. for example.
u2. m) are TFNs.9) m In order to obtain j 1 j M gi . i = 1.5 Extent Analysis Method Chang [14] proposed a new approach to handle situations that require use of both fuzzy logic and AHP/ANP. j 1 j 1 mj .Quantitative Modeling Techniques 43 4. j perform the fuzzy addition operation of M gi (j = 1. m) values such that © 2009 by Taylor & Francis Group. Step 1: The value of fuzzy synthetic extent with respect to the ith object is defined as m n m 1 Si j 1 M j gi i 1 j 1 M j gi (4. j 1 uj (4.10) To obtain n m 1 M i 1 j 1 j gi . perform the fuzzy addition operation of m extent analysis values for a particular matrix such that m j M gi j 1 m m m lj . TFNs (see section 4. each object is taken and an extent analysis for each goal. …. then by using extent analysis method [15]. 2. Therefore. M2gi. xn} be an object set and U = {u1. First. um} be a goal set. …. the synthetic extent value of the pairwise comparison is introduced. LLC . 2. n. where all the Mjgi (j = 1. Mmgi. is performed. 2. gi. …. m extent analysis values for each object can be obtained. According to the extent analysis method. The steps involved in the methodology are as follows. …. and by applying the principle of comparison of fuzzy numbers.4) are used for pairwise comparisons. …. Let X = {x1. the weight vectors with respect to each element under a certain criterion can be computed. with the following signs: M1gi. …. x2.
14) Step 4: The weight vector obtained in step 3 is normalized to get the normalized weights. k) can be defined as V (M ≥ M1. i =1. 2. i 1 mi . m2. (l1 – u2)/((m2 – u2) – (m1 – l1))} (4. both V (M2 ≥ M1) and V (M1 ≥ M2) are required. k (4. As a result. 2.12) To compare M1 and M2. 0. 4. Then the weight vector is given by W’ = (d’ (A1). and alternatives with respect to the main and subcriteria. Step 2: The degree of possibility of M2 = (l2. uncertain subjective data are present that make the decisionmaking © 2009 by Taylor & Francis Group. m1.13) Let d’ (Ai) = min V (Si ≥ Sk).44 Strategic Planning Models for Reverse and ClosedLoop Supply Chains n m j M gi i 1 j 1 i 1 n n n li . The rest of the procedure is similar to the traditional AHP/ANP. n. M2. i 1 ui (4. u2) ≥ M1 = (l1. d’ (An))T (4.6 Fuzzy Multicriteria Analysis Method Multicriteria analysis problems require the decision maker to make qualitative assessments regarding the performance of the decision alternatives with respect to each independent criterion and the relative importance of each independent criterion with respect to the overall objective of the problem. subcriteria with respect to the main criteria. …. LLC . for k = 1. d’ (A2). These steps are applied to deduce the weights of main criteria with respect to the goal. Mk) = V [(M ≥ M1) and (M ≥ M2) and … and (M ≥ Mk)] = min V (M ≥ Mi). 2. if m2 ≥ m1. if l1 ≥ u2. k ≠ i. u1) is expressed as V (M2 ≥ M1) = hgt (M1 ≥ M2) = {1. Step 3: The degree of possibility for a convex fuzzy number to be greater than k convex fuzzy numbers Mi (i = 1.11) and then compute the inverse of the vector.
akk (4. n. l s..Quantitative Modeling Techniques 45 process complex [16]. . The cut concept is used to transform the fuzzy performance matrix representing the overall performance of all alternatives with respect to each criterion into an interval performance matrix.5) is applied to solve the fuzzy reciprocal matrix for determining the criteria importance and alternative performance.16) a12 a22 .. l als 1. 1. 3. ak1 where 1.. However.17) © 2009 by Taylor & Francis Group.k . (4.. a fuzzy reciprocal matrix for criteria importance (W) or alternative performance with respect to a specific criterion (Cj) can be determined as a11 W or C j a21 . Deng [17] proposed a multicriteria analysis approach that extends Saaty’s AHP (see section 4. traditional AHP is criticized for its unbalanced scale of judgment and failure to precisely handle the inherent uncertainty and vagueness in carrying out pairwise comparisons. 9. s s s.l By applying the extent analysis method. 5.. The selection process starts with determining the criteria of importance and performance of alternatives.. 2. AHP enables a person to make pairwise judgments of importance between the independent criteria as well as the decision alternatives. l.2) to deal with the imprecision and subjectiveness in the pairwise comparisons... and the concept of extent analysis method (see section 4. TFNs (see section 4. ak 2 .. k m.4) are used for pairwise comparisons.. . An overall performance index for each alternative across all criteria that incorporates the decision maker’s attitude toward risk is obtained by applying the concept of similarity to the ideal solution [18] using the vectormatching function... LLC . a1k a2 k . .... the corresponding criteria weights (wj) or alternative performance ratings (xij) with respect to a specific criterion Cj can be determined as k k k xij or w j = s 1 als l 1 s 1 als (4. By using TFNs.. or .15) 1 / a sl .
z nml . z n1l ..18) X W = (w1 .... . wm ) (4... z12r . .. .. xn1 x12 x 22 . z n1r z12l ... where 0 ≤ ≤ 1.. 2.... z11r Z . xnm (4.. LLC . wm x1m w m x 2m . x1m x 2m .. and k = m or n depending on whether the reciprocal judgment matrix is for assessing the performance ratings of alternatives or weights of criteria involved.... j = 1.....19) where xij represents the resultant fuzzy performance assessment of alternative Ai (i = 1. . z n 2l .. . . z nmr .. z11l .. 2. (4. wm xnm (4.... w 2 xn 2 . the assessments are closer to the most possible value a2 of the triangular fuzzy number (a1. The decision matrix (X) and the weight vector (W) can be respectively determined as x11 x 21 . …... .. m.... w 2 .. . A fuzzy performance matrix Z representing the overall performance of all alternatives with respect to each criterion is obtained by multiplying the weight vector by the decision matrix.. The value of represents the decision maker’s degree of confidence in his or her fuzzy assessments regarding the alternative ratings and criteria weights. w1 x11 Z w1 x 21 . and wj is the resultant fuzzy weight of criterion Cj (j = 1... …... 2.... a3). The larger the value of .46 Strategic Planning Models for Reverse and ClosedLoop Supply Chains where i = 1. using an optimism index ( = 1 implies the decision © 2009 by Taylor & Francis Group. . m) with respect to the overall goal of the problem. . the more confident the decision maker is about the fuzzy assessments.... 2. z1mr .. n) with respect to criterion Cj. n.20) An interval performance matrix [16] is derived by using an cut on the performance matrix.. w1 xn1 w 2 x12 w 2 x 22 .... z n 2r . xn 2 . .. . ….. . a2.21) An overall crisp performance matrix that incorporates the decision maker’s attitude toward risk. z1ml . viz.
z11 z ' ' z12 ...... z m ) (z1 . ....and negativeideal solutions..... the positive......Quantitative Modeling Techniques 47 maker has an optimistic view.. zn2 . z 2 . .. z1m ... and 0. z1m z 2m . z n1 ' (4.. z nm ' ' .25) Zeleny [18] introduced the concept of ideal solution in multiattribute decision analysis that was further extended by Hwang and Yoon [19]. . z 2 .22). z m ) (4... ..1] (4. 0 implies a pessimistic view.26) © 2009 by Taylor & Francis Group.. zn2 ' ' .. [0. .... respectively.. including negative solution to avoid the worst decision outcome.5 implies a moderate view).23) A normalized performance matrix with respect to each criterion is calculated from equation (4. LLC . . is calculated. can be determined by selecting maximum and minimum values across all alternatives with respect to each criterion as follows: A A (z1 .. z n1 z12 z 22 ...22) where z ij ' z ijr (1 )z ijl . z nm Z (4.. In line with this concept...... .24) where n ' z ij z ij (z ij )2 i 1 ' (4. .. . . z11 z 21 ...
A A ) (4.. the degree of similarity between each alternative and the positive... i 1. Hence..7 Quality Function Deployment Erol and Ferrell [21] define performance aspects as the features that the decision maker wishes to consider in the selection process and enablers as the characteristics possessed by the alternatives.. where Ai which represents the corresponding performance of alternative Ai with respect to criterion Cj. A preferred alternative should have a higher degree of similarity to the positiveideal solution and a lower degree of similarity to the negativeideal solution. z nj ) min(z j .29) The larger the performance index.. the higher the degree of similarity between each alternative and the positiveideal and negativeideal solutions [20].. z 2 j .. A A ) Ai Ai / max( Ai Ai .. which measure how effectively each enabler can satisfy all of the performance aspects. the most preferred the alternative is. 4.30) © 2009 by Taylor & Francis Group.. are computed by I ATIRj i 1 di Rij j = 1. Si .28) ( zi1 .. z 2 j . .and negativeideal solutions can be calculated as Si Si Ai Ai / max( Ai Ai . LLC .48 where Strategic Planning Models for Reverse and ClosedLoop Supply Chains zj zj max(z j . z nj ) (4. zi 2 . zim ) is the ith row of the overall performance matrix.. . which can be used to satisfy the performance aspects. an overall performance index for each alternative with the decision maker’s level of confidence and degree of optimism toward risk can be determined as Pi Si / (Si Si ).. J (4.n (4..27) By applying the vectormatching function.. The absolute technical importance ratings (ATIRs)... 2. The larger the value of Si .
it is normalized to form the relative technical importance rating (RTIRj) as follows: RTIRj ATIRj J j = 1. for the comparison of all enablers.Quantitative Modeling Techniques 49 where di is the importance value of performance aspect i relative to the other performance aspects. and N is the total number of alternatives.. and 4S) are used to allow the decision maker to express his or her © 2009 by Taylor & Francis Group. together with additional human expert opinions. J (4. 3S.32) where TUPn is the total user preference for alternative n. Because there is an ATIR for each enabler j.. LLC . TUP of each alternative is then normalized as follows: NTUPn TUPn N n (4. and Rij is the relationship score for performance aspect i and enabler j.. The alternative with the highest NTUP is considered the one with the highest potential.8 Method of Total Preferences RTIRs (see section 4. 4.9 Linear Physical Programming In the linear physical programming (LPP) method [22]. .33) TUPn n 1 where NTUPn is the normalized total preference for alternative n. four distinct classes (1S. and WAnj is the (defuzzified) degree to which alternative n can deliver enabler j. are used to develop a single measure that reflects the rating of each alternative as follows [21]: J TUPn j 1 RTIR j WAnj n (4. 2S.7). For the purpose of comparison of all alternatives.31) ATIRj j 1 4.
and range is better (4S).2S "Larger is Better" Unac ce ptable Unacce ptable Highly unde sirabl e Unde sira ble Unde sira ble Desira ble Tolerable Highly undesirable Tolerable Ideal Desirable t+p1 t+p2 t+p3 t+p4 t+p5 g p(x) tp5 tp4 tp3 tp2 tp1 g p(x) zp z5 Cla ss.50 Strategic Planning Models for Reverse and ClosedLoop Supply Chains preferences for the value of each criterion (for decision making) in a more detailed. Consider. Figure 4. for example. These classes are defined as follows: smaller is better (1S).3S "Value is Better" z4 z3 z2 z1 zp z5 Class4S "Range is Better" z4 z3 z2 z1 Unac ce ptable FIGURE 4. The value of the pth criterion. the case of class 1S.3 Soft class functions for linear physical programming. LLC Unac ce ptable Highly undesi rable Undesirable Undesirable Highly undesira ble Desirable Tolerable Desirable Tolerable tp5 tp4 tp3 tp2 tp1 t+p1 t+p2 t+p3 t+p4 Unac ce ptable Unac ceptable Highly unde sirable Unde sirabl e Undesirabl e Desira ble Desira ble Highly undesirabl e Tolerable Tolerable tp5 tp4 tp3 tp2 tp1 t+p2 t+p3 t+p4 t+p5 g p(x) Idea l t+p5 g p(x) Ideal Idea l . gp.2). larger is better (2S). and qualitative way than when using a weightbased method like analytic hierarchy process (see section 4.3 depicts these different classes. © 2009 by Taylor & Francis Group. quantitative. value is better (3S). for evaluating the alternative of interest is categorized according to the preference ranges shown on the horizontal axis. The preference ranges are: zp z5 zp z5 Class1S "Smaller is Better" z4 z3 z2 z1 z4 z3 z2 z1 Cla ss.
We can accomplish this for a nonnumerical criterion such as color as well by (1) specifying a numerical preference structure and (2) quantitatively assigning each alternative a specific criterion value from within a preference range (e. The class function. zero).. that represent the incremental slopes of the class functions. then the value of the class function is very high. a piecewise linear function of gp). Step 2: Specify preferences for each criterion. with different physical meanings. ranking of the alternatives is performed in four steps. desirable. Basically. and (2) that the value of the class function. and dimensionless parameter (Zp is. Here. then the value of the class function is small (in fact. w pr and w pr (used in step 4). r denotes the range intersection. tolerable). into a real. is in the ideal range. gp. Step 3: Calculate incremental weights: Based on the preference structures for the different criteria. Zp. the LPP weight algorithm [22] determines incremental weights.Quantitative Modeling Techniques 51 Ideal range: g p t p1 gp gp t p2 t p3 gp t p4 gp t p5 Desirable range: t p1 Tolerable range: t p 2 Undesirable range: t p 3 Highly undesirable range: t p 4 Unacceptable range: g p t p5 The quantities t p1 through t p5 represent the physically meaningful values that quantify the preferences associated with the pth generic criterion. for example.3 is used to map the criterion value. at a given range intersection (say. in the unacceptable range. an alternative having a cost of $15 would lie in the desirable range. Class functions have several important properties. as follows [23]: Step 1: Identify criteria for evaluating each of the alternatives. Consider. in fact. © 2009 by Taylor & Francis Group. an alternative with a cost of $45 would lie in the highly undesirable range. Thus. Zp. The decision maker could specify a preference vector by identifying t p1 through t p5 in dollars as (10 20 30 40 50). piecewise linear. based on one of the four classes (see figure 4. the cost criterion for class 1S. If the value of a criterion. whereas if the value of the criterion is greater than t p5 . continuous. Such a mapping ensures that different criteria values. Zp. and so on. desirable–tolerable) is the same for all class types. are mapped to a common scale. Consider class 1S again. gp. on the vertical axis in figure 4. LLC .3).g. that is. including (1) that they are nonnegative. and convex. positive.
max Class 2H: Must be larger. tp.34) and subjecting (if necessary) each criterion. and there are several cases where the management focuses on a variety of objectives simultaneously.3 or one of the following four hard classes: Class 1H: Must be smaller. gp. i.val Class 4H: Must be in range.. However. The most significant advantage of using LPP is that no weights need to be specified for the criteria for evaluation. © 2009 by Taylor & Francis Group.. w pr and w pr are the incremental weights for the pth criterion. gp ≤ tp.34) where P represents the total number of criteria (each belonging to one of the four classes in figure 4.min Class 3H: Must be equal.52 Strategic Planning Models for Reverse and ClosedLoop Supply Chains Step 4: Calculate total score for each alternative: The formula for the total score..e. An alternative with a lower total score is more desirable than one with a higher total score. by minimizing J in equation (4.max 4. of the alternative of interest is constructed as a weighted sum of deviations over all ranges (r = 2 to 5) and criteria (p = 1 to P). gp = tp.e. i. gp ≥ tp. i. i.e. LPP can be used in a problem consisting of decision variables as well. this assumption is not always realistic. to a constraint that falls into either one of the four classes (also called soft classes) in figure 4. which has more physical meaning than a physically meaningless weight that is arbitrarily assigned to the criterion. as follows: P 5 J p 1 r 2 ( w pr d pr w pr d pr ) (4.10 Goal Programming Linear programming [24] assumes that the objectives of an organization can be encompassed within a single objective function.min ≤ gp ≤ tp. and d pr and d pr represent the deviations of the pth criterion value of the alternative of interest from the corresponding target values. such as maximizing the total profit or minimizing the total cost. Goal programming provides a way of tackling such situations. The decision maker only needs to specify a preference structure for each criterion. LLC .e. Note that there are no decision variables in the above ranking procedure.3).. J.
deals with the achievement of specific targets/goals. Another categorization is according to how the goals compare in importance. k.35) Gi ( X ) Li g i Li 0 if L i Gi ( X ) if Gi ( X ) Li while a linear membership fraction μi that represents goal fuzziness for the “less is better” form is expressed as © 2009 by Taylor & Francis Group. several criteria are to be considered in the problem situation on hand. In case of nonpreemptive goal programming. the aspiration levels are either in the “more is better” form or “less is better” form. LLC . the desire to overachieve (minimize k) or underachieve (minimize k) or to satisfy the target value exactly (minimize k + k) is articulated [26]. In the case of preemptive goal programming. the deviation variables are introduced. Goal programming problems can be categorized according to the type of mathematical programming model. [27]. which may be positive or negative (represented by k and k. Finally. To this end.Quantitative Modeling Techniques 53 Goal programming (GP. generally applied to linear problems. In goal programming. and the overall deviation from the aspiration levels is minimized. The basic approach involves formulating an objective function for each objective and seeks a solution that minimizes the sum (weighted sum in case of fuzzy goal programming) of the deviations of these objective functions from their respective goals. represents the underachievement of the kth goal. represents the overachievement of the kth goal. fuzzy GP comes in handy. In such situations. Similarly. For each criterion. In most realworld scenarios. so the goals of primary importance receive first attention and so forth. respectively). Next. A linear membership function μi that represents goal fuzziness for the “more is better” form is expressed as [28] 1 i if Gi ( X ) gi gi (4. the aspiration levels and weights/importance levels of goals are imprecise in nature. see [25]). all the goals are of roughly comparable importance [24]. In fuzzy GP. allowing the decision maker to obtain compromising results for multiple goals with varying aspiration levels. it is necessary to specify aspiration levels for the goals. k. for each criterion. a target value is determined. the positive deviation variable. there is a hierarchy of priority levels for the goals. The negative deviation variable.
41) where V( ) is the fuzzy achievement function or fuzzy decision function. Li. Fibonacci numbers are used to assign weights to the goals.38) (4. The weighted additive model is widely used in GP and multiobjective optimization problems to reflect the relative importance of goals. © 2009 by Taylor & Francis Group. i 0 (4. In this approach.40) X. and Ui are the aspiration level. Fibonacci numbers are 0.36) if Gi ( X ) U i where gi. 2. etc. 5. The objective function for the weighted additive model is expressed as p V( ) Maximize i 1 wi i (4. 1. 3. The simple form of the fuzzy GP problem with p fuzzy goals can be stated as: p Maximize V( ) i 1 i (4.37) subject to i Gi ( X ) Li or g i Li AX b i U i Gi ( X ) Ui g i (4. 1.39) i 1 (4. (the next number is a result of the summation of the previous two numbers). lower tolerance limit.54 Strategic Planning Models for Reverse and ClosedLoop Supply Chains 1 i if Gi ( X ) gi U i Gi ( X ) if g i Ui g i 0 Gi ( X ) U i (4. the decision maker assigns weights as coefficients of individual terms in the simple additive fuzzy achievement function to reflect their relative importance.42) where wi is the relative weight of the ith fuzzy goal. The objective is to obtain the μi value as close to 1 as possible. LLC . and upper tolerrance limit respectively for the fuzzy goal Gi (X). 8.
33 and 0. other goals should not be considered.46) μs ≤ 1 (4. These weights are assigned to the two goals according to their priority levels. for two goals.66. 4. the ith subproblem becomes: Maximize s ( s ) pi (4. which approximately are 0. For example. the weights would be in the ratio of 1:2.47) X.44) AX ≤ b (4. the fuzzy goals belonging to the first priority level will be considered and solved using the simple additive model. p (4. the membership values achieved at earlier priority levels are added as additional constraints. LLC . 2. or the goals are such that unless a particular goal or subset of goals is achieved.j 1 (4. .45) ( ) pr ( * ) pr .2. The problem is divided into k subproblems. In some situations. r 1. where k is the number of priority levels. In general. In the first subproblem.…. At other priority levels. In such situations. μi ≥ 0.43) subject to s Gs gs Ls Ls (4. i = 1.11 Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) The basic concept of the TOPSIS method [19] is that the rating of the alternative selected as the best from a set of different alternatives should have the © 2009 by Taylor & Francis Group. the weighting scheme is not appropriate. the goals/objectives are not commensurable.48) where (μs)pi refers to the membership functions of the goals in the ith priority level and (μ*)pr is the achieved membership function value in the rth (r ≤ j – 1) priority level.Quantitative Modeling Techniques 55 The concept is applied by starting with numbers 1 and 2.
3.49) Step 2: Construct the weighted normalized decision matrix. on a scale of 1–10. An element rij of the normalized decision matrix R is calculated as follows: rij z ij m i 1 2 z ij (4. specified by the decision maker. wn) (such that ∑wj = 1). The TOPSIS method evaluates the following decision matrix. p3. m (4. The following steps are performed: Step 1: Construct the normalized decision matrix. …. LLC . A set of weights W = (w1. . ..50) = {p1.. w2. The ideal (A*) and the negativeideal (A–) solutions are defined as follows: A* max vij i for i 1... . the better it is) of the ith alternative in terms of the jth criterion. is used in conjunction with the normalized decision matrix R to determine the weighted normalized matrix V defined by V = (vij) = (rijwj). 2. Cj is the jth criterion.. .e. ….56 Strategic Planning Models for Reverse and ClosedLoop Supply Chains shortest distance from the ideal solution and the greatest distance from the negativeideal solution in a geometrical (i. the higher the rating. wj is the weight (importance value) assigned to the jth criterion. pn} © 2009 by Taylor & Francis Group. Euclidean) sense. This step converts the various dimensional measures of performance into nondimensional attributes. which refers to m alternatives that are evaluated in terms of n criteria [29]: Criteria Alternatives A1 A2 A3 . Step 3: Determine the ideal and the negativeideal solutions. p2. and zij is the rating (for example.. Am C1 w1 z11 z21 z31 … … … … zm1 C2 w2 z12 z22 z32 C3 w3 z13 z23 z33 … … … … … Cn wn z1n z2n z3n zm2 zm3 … zmn where Ai is the ith alternative.
LLC . m (4. Similarly. A– indicates the least preferable (negativeideal) solution. .Quantitative Modeling Techniques 57 A min vij i for i 1. The corresponding formulae are Si* (vij p j )2 for i 1. 3.54) Step 6: Rank the preference order... q3. 2.53) where Si– is the separation (in the Euclidean sense) of the rating of alternative i from the negativeideal solution. qn} (4. 3. the decision maker desires to choose the alternative with the maximum rating (it is important to note that this choice varies with the way he or she awards ratings to the alternatives). m = {q1. In this step.. . the concept of the ndimensional Euclidean distance is used to measure the separation distances of the rating of each alternative from the ideal solution and the negativeideal solution. q2. A* indicates the most preferable (ideal) solution. Obviously. the better the alternative. ….52) where Si* is the separation (in the Euclidean sense) of the rating of alternative i from the ideal solution. 3.51) With respect to each criterion. The relative closeness coefficient for alternative Ai with respect to the ideal solution A* is defined as follows: Si Si* Si Ci* (4... Step 5: Calculate the relative coefficient... The way the alternatives are processed in the previous steps reveals that if an alternative has the rating with the shortest distance to the ideal solution. 2. The best alternative can now be decided according to the preference order of Ci*... then that rating is guaranteed to have the longest distance to the negativeideal solution. That means the higher the Ci*. Step 4: Calculate the separation distances. It is the one with the rating that has the shortest distance to the ideal solution. .. 2. © 2009 by Taylor & Francis Group.. m (4. and Si (vij qj )2 for i 1.
…. Lisp and Prolog) of the system. Thus. Uncertainty modeling process: This aids the inference engine when dealing with uncertainty. and the alternative with the highest Borda score is declared the winner.13 Expert Systems Expert systems are computer programs that can represent human expertise (knowledge) in a particular domain (area of expertise) and then use a reasoning mechanism (applying logical deduction and induction processes) to manipulate this knowledge in order to provide advice in this domain. their main function is to retrieve information and carry out statistical analysis and numerical calculations. The Borda score (maximized consensus mark) for each alternative is then determined as the sum of the individual marks for that alternative. The knowledge is written in the knowledge base using the syntax of what is termed the knowledge representation language (e. Reasoning mechanism: This traces the path or the knowledge steps used to arrive at a conclusion and can relay it back to the user as the justification for this conclusion. worst alternatives. 4. The main components of an expert system are the following [31]: Knowledge base: This is where the knowledge is stored.12 Borda’s Choice Rule Borda proposed a method in which marks of m – 1. this consists of a set of rules of the form: if EVIDENCE. Typically. © 2009 by Taylor & Francis Group. what mainly distinguishes expert systems from conventional programs is the capability to reason with knowledge. Although conventional computer programs also contain knowledge.g. then HYPOTHESIS. Inference engine: This reasons with the knowledge resident in the knowledge base using certain mechanisms. secondbest.58 Strategic Planning Models for Reverse and ClosedLoop Supply Chains 4. 0 are assigned to the best.. LLC . That means that the different decision makers unanimously choose the alternative that obtains the largest Borda score as the most preferred one. m – 2. …. Examples of this mechanism are deduction (cause + rule effect). 1. That means that a larger mark corresponds to greater preference. for each decision maker [30]. and induction (cause + effect rule). They do not reason with this knowledge or make inferences as to what actions to take or conclusions to reach. abduction (effect + rule cause).
In practice. then Y) and ideally should not need to be concerned with the working of the inference engine. and if O(H) < 0. is O(HE) (A). limits are often set on odds values so that. Then the formula for the odds of that hypothesis. The user of the shell enters rules in a declarative fashion (if X. The standard formula for updating the odds of a hypothesis H. for example. is © 2009 by Taylor & Francis Group. and A is the affirms weight of E. if O(H) > 1. has infinite odds.55) A hypothesis that is absolutely certain. then H is true. Bayesian updating assumes that the absence of supporting evidence is equivalent to the presence of opposing evidence. given that evidence E is observed. 4.Quantitative Modeling Techniques 59 A shell is an expert system that is complete except for the knowledge base [31]. Typically.01. The standard formula for updating the odds of hypothesis H. Suppose the probability of a hypothesis H is P(H).O(H) (4. has a probability of 1. and a user interface for running the system. i. given that H is true. given the presence of evidence E.000. is given by O(H) P(H) 1 – P(H) (4.14 Bayesian Updating Bayesian updating [32] is an uncertainty modeling technique that assumes that it is possible for an expert in a domain to guess a probability to every hypothesis or assertion in that domain and that this probability can be updated in light of evidence for or against the hypothesis or assertion. given that the evidence E is absent. then H is false. O(H).56) where O(HE) is the odds of H. Thus. a shell includes an inference engine. the programming interface comprises a specialized editor for creating rules in a predetermined format and some debugging tools.57) where P(EH) is the probability of E.. The definition of A is A P(EH) P(E~H) (4. not given that H is true. LLC .e. Expert system shells are easy to use and allow a simple expert system to be constructed quickly. a user interface for programming. and P(E~H) is the probability of E.
In such a case. then the presence of evidence E is supportive of hypothesis H.(1–D).(A1). Similarly. the affirms and denies weights are modified using the following formulae: A' = [2. For example. Also. given the absence of evidence E.59) If a given piece of evidence E has an affirms weight A that is greater than 1. a rule “IF (temperature is high) and NOT (water level is low) THEN (pressure is high)” can also be written as “IF (temperature is high—AFFIRMS A1. For example.” Here. then its denies weight must be less than 1. © 2009 by Taylor & Francis Group. The definition of D is D P(~EH) P(~E~H) 1–P(EH) 1–P(E~H) (4.5. the denies weight is used.O(H) (4.5. it is not definite if this is due to a malfunctioning picture tube.60) D' = [2. depending upon the value of the probability of the evidence P(E).P(E)]+2–A (4. DENIES D2) THEN (pressure is high).61) When P(E) is greater than 0.58) where O(H~E) is the odds of H.P(E)]+D (4. and D is the denies weight of E. DENIES D1) AND (water level is low—AFFIRMS A2.60 Strategic Planning Models for Reverse and ClosedLoop Supply Chains O(H~E) (D). and vice versa. A1 D1 P(Temperature is high  Pressure is high) ) P(Temperature is high  ~Pressure is high) P(~Temperature is high  Pressure is high) g P(~Temperature is high  ~Pressure is high) A2 D2 P(Water level is low  Pressure is high) P(Water level is low  ~Pressure is high) P(~Temperature is high  Pressure is high) g P(~Temperature is high  ~Pressure is high) Sometimes. LLC . if A < 1 and D > 1. if A > 1 and D < 1. if one is diagnosing a TV set that is not functioning properly. the affirms weight is used to calculate O(HE). while controlling a power station boiler. then the absence of E is supportive of H. evidence is neither definitely present nor definitely absent. and when P(E) is less than 0.
© 2009 by Taylor & Francis Group.63) (4.6). There are three types of Taguchi loss functions: “target is best” (see figure 4...(A 3 ). If L(y) is the loss associated with a particular value of characteristic y..(Dn ). The quality losses occur only when the product deviates beyond the specification limits. Otherwise. for the “target is best” type...65) 4.(D3 ).15 Taguchi Loss Function In traditional systems. and k is the loss coefficient whose value is constant depending on the cost at the specification limits and width of the specification. the product is rejected. If a characteristic measurement is the same as the target value.62) Ai and Di are given by equations (4.O(H) (4. then the updating equations are given by O(HE1 &E 2 &E 3 . m is the target value of the specification. the loss can be measured using a quadratic function. LLC .. the loss is zero. Taguchi [33] suggests a narrower view of characteristic acceptability by indicating that any deviation from a characteristic’s target value results in a loss.. respectively..5).O(H) and O(H E1 & E2& E 3 . the product is accepted if the product measurement falls within the specification limits [33]...Quantitative Modeling Techniques 61 If n statistically independent pieces of evidence are found that support or oppose a hypothesis H.. Otherwise.4).65). thereby becoming unacceptable.64) Di P(~E i  H) P(~E i ~H) (4..(D2 )....64) and (4. Ai P(E i H) P(E i ~H) (4.. and “larger is better” (see figure 4... after which actions are taken to reduce systematically the variation from the target value.. These costs tend to be constant and relate to the costs of bringing the product back into the specification range.(A 2 ). “smaller is better” (see figure 4.(An ).En ) = (A1 ). En ) = (D1 ).
FIGURE 4. L( y ) k( y m)2 (4.67) k / ( y )2 (4.5 “Smaller is better” Taguchi loss.4 “Target is best” Taguchi loss. L( y ) For the “larger is better” type. LLC Taguchi Loss m Upper Speciﬁcation Upper Speciﬁcation Taguchi Loss Taguchi Loss .68) Lower Speciﬁcation FIGURE 4.62 Strategic Planning Models for Reverse and ClosedLoop Supply Chains L( y ) For the “smaller is better” type.66) k( y )2 (4. © 2009 by Taylor & Francis Group.
Taguchi Loss Lower Speciﬁcation 4. Process capability refers to the ability of the process to meet the design specifications for an output. The lab process must be capable of providing the results of analyses within these specifications (see figure 4. LLC .” Note that in most situations. it will produce a certain proportion of “defects. the administrator of an intensive care unit lab might have a target value for the turnaround time of results to the attending physicians of 25 minutes and a tolerance of 5 minutes because of the need for speed under lifethreatening conditions.7). The tolerance gives an upper specification (U) of 30 minutes and a lower specification (L) of 20 minutes. Design specifications are often expressed as a target value ( ) and a tolerance (T).Quantitative Modeling Techniques 63 FIGURE 4. For example.6 “Larger is better” Taguchi loss. However. The control limits on the control charts signal when the mean or variability of the process changes. © 2009 by Taylor & Francis Group.16 Six Sigma Statistical process control techniques help managers achieve and maintain a process distribution that does not change in terms of its mean and variance [34]. otherwise. a process that is in statistical control may not be producing outputs according to their design specifications because the control limits are based on the mean and variability of sampling distribution. T U 2 L and U 2 L . and hence it is assumed so here. not the design specifications.
16. the probability that the actual critical dimension is within the specification limits is 0. the specification width is the same as the distribution width.64 Strategic Planning Models for Reverse and ClosedLoop Supply Chains L = 20 min FIGURE 4. For exam © 2009 by Taylor & Francis Group. When the process mean ( ) is centered at (U+L)/2 without any shift from the target value. 4. the specification width is twice that of the distribution. When the process mean ( ) is centered at (U+L)/2 without any shift from .16. τ = 25 min U = 30 min Two essential quantitative measures to assess the capability of a process are process capability ratio (Cp) and process capability index (Cpk).002 ppm defect rate). if Cp = 2. 4. We consider two examples.700 ppm defect rate).999999998 (0. the probability that the actual critical dimension is within the specification limits (assuming that the process distribution is normal) is 0. The process capability ratio (Cp) is defined as U L 6 Cp (4. . one for Cp = 1 and the other for Cp = 2.9973 (2.7 Capable process.2 Process Capability Index (Cpk) The process capability ratio (Cp) is enough to find out whether a process is capable. only if is centered at (U+L)/2 without any shift from .69) The numerator represents the specification width and the denominator captures the total width of the 3 limits of the process distribution. LLC . Similarly.1 Process Capability Ratio (Cp) Assume that is the standard deviation of a process that produces a certain dimension of interest for an output (good or service). If Cp = 1. This certain dimension of interest will hereafter be called critical dimension.
16. a shift of from . Thus.1 Three Sigma Process See figure 4. and profit increase. We illustrate this by calculating Cp and Cpk values (equations (4. more than the critical is closer to U.5. 1.69) and (4.5). The Motorola convention uses a onesided mean shift of 1. A process is said to be capable only if the process has good values (viz. mean shift consumes 25% of the tolerance)..5 (i. viz. multifaceted approach to process improvement. It must be noted that the mean shift in each case is allowed to be up to 1. n = 3. say. We consider three different cases.9999966 (i. it is considered a Three Sigma process. called the process capability index (Cpk). the better the process is). lengthy turnaround times may still value of.5 0. if is not far away from . cost reduction. 4. and 6.. Cpk is defined in [35] as C pk C p 1 k .e.. If Cpk is less than the critical value. i. in order to check whether need for an additional capability ratio.70) T This definition allows consideration of a mean shift. the higher the value of n. very quick results may be generbe generated. This is motivated by common physical phenomena such as tool wear. LLC . one needs both Cp and Cpk values in order to investigate whether the process of interest is a Six Sigma process. If Cp is less than the critical value. Cp U L 6 6 6 1 C pk Hence. Its fundamental principle is to improve customer satisfaction by reducing defects in processes...4 ppm defect rate). If Cp = 2 and Cpk = 1. either is too close to U or L or is too high. 4. Six Sigma is an art of management that originated at Motorola in the early 1980s and is a businessdriven.5 .5. Likewise. if Cp = 1 and Cpk Cp 1 k 11 1. more than the respective critical values) of both Cp and Cpk. Traditionally. The fraction k is the fraction of tolerance consumed by the mean shift. but if is closer to L.2. 3..e. there is a ated. where k (4. is too high.Quantitative Modeling Techniques 65 ple. respectively) for an n Sigma process (where n is any positive real number.70). the probability that the actual critical dimension is within the specification limits is 0.5 . the lab process may have a good Cp value (i.5 3 0. © 2009 by Taylor & Francis Group.e.8.e.
5 Sigma process. U= +3 4. it is considered a 4.5 1. if Cp = 1. Six Sigma Process See figure 4.2. © 2009 by Taylor & Francis Group.8 Three Sigma process.66 Strategic Planning Models for Reverse and ClosedLoop Supply Chains T= 3 T= 3 1.10.2 4.5 4.5 C pk Hence. Cp U L 6 9 6 1.3 Cp 1 k 1.16.9.5 and Cpk 4.5 L= 3 FIGURE 4.5 1 1.5. LLC .5 1 1.16. if Cp = 2 and Cpk Cp 1 k 21 1.5 Sigma Process See figure 4. it is considered a Six Sigma process.2.5 6 1. Cp U L 6 12 6 2 C pk Hence.
LLC . FIGURE 4.10 Six Sigma process. 4. In this book. Neural networks can model complex relationships between inputs and outputs or find patterns in data. © 2009 by Taylor & Francis Group.Quantitative Modeling Techniques 67 FIGURE 4.17 Neural Networks A neural network is made up of simple processing units (called neurons) combined in a parallel computer system following implicit instructions based on recognizing patterns in data inputs from external sources [41].5 Sigma process. the following equation [42] is used (see chapters 6 and 7) to calculate the weights (importance values) of evaluation criteria considered in a strategic planning issue. Their usefulness derives from their ability to embody inferential algorithms that alter the strengths or weights of the network connections to produce a desired significant flow.9 4.
© 2009 by Taylor & Francis Group. LLC . GIS is applied for displaying data. In a more generic sense. an excellent GISbased business mapping application. edit data.18 Geographical Information Systems A geographical information system (GIS) is a computer system with a set of processes for obtaining. is used to map the results obtained in different phases of a model. 4.68 Strategic Planning Models for Reverse and ClosedLoop Supply Chains nH nV j I vj  Iij  i Oj Wv  nV nH I vj nV Oj (4. managing. analyzing. In this book (see chapter 6). analyze the spatial information. and Oj is the connection weight from the jth hidden node to the output node. MapLand [40]. GIS is a tool that allows users to create interactive queries (usercreated searches). nV is the number of input nodes (evaluation criteria). The motivation for this application is a paper [39] that addresses the necessity for building the strategic planning process “around a picture.” to make the chairperson of the concerned supply chain company easily understand the “dense documents filled with numbers” and to convince him or her that it is important to implement the proposed action. nH is the number of hidden nodes (can be any arbitrary number). the absolute value of Wv is the weight of the vth input node (evaluation criterion) upon the output node (rating of the decision alternative). Iij is the connection weight from the ith input node to the jth hidden node. GISs have become powerful operation tools in the business world. To this end.71) i j i  Iij  Here. and displaying data that have been located geographically [37]. and map and present the results of all these operations [38]. The connection weights [43] are obtained upon training the respective neural network.
it is applied to business. D. L. S. in canonical form. C. If the variables of a linear programming problem are restricted to being integers. 4. V. 6. P. E. 1996. Pittsburgh.20 Conclusions This chapter gave an introduction to various quantitative techniques employed by strategic planning models presented in different chapters of the book. 1991. each of the models uses one or more of the techniques introduced in this chapter.19 Linear Integer Programming Linear programming problems are about optimization of a linear objective function. 2. and Vieux. The analytic hierarchy process. Tadisina. G. 4. T. L. Landfill siting using geographic information systems: A demonstration. S. L. Journal of the Operational Research Society 42:631–38. Journal of Environmental Engineering 122:515–23. M.. For example. Z. J.. Ganesh. Decision making with dependence and feedback: The analytic network process.. 1980. Most extensively. and Bhasin. B. K. Saaty. 1990. © 2009 by Taylor & Francis Group. 1996. International Journal of Production Economics 35:293–97. New York: McGrawHill. Prakash. Everett.Quantitative Modeling Techniques 69 4. PA: RWS Publications. whereas C and B are vectors of coefficients and A is the matrix of coefficients. How to make a decision: The analytic hierarchy process.. X represents the vector of variables. Selecting a doctoral programme using the analytic hierarchy process: The importance of perspective. Saaty. 3. Saaty. economic. subject to linear equality and inequality constraints [36]. LLC . Siddiqui. Troutt. References 1. L. Linear integer programming can be applied to various fields. a linear programming problem can be expressed as Maximize CTX subject to AX ≤ B and X ≥ 0 (objective function) (constraints) Here. the programming is called linear integer programming. European Journal of Operational Research 48:9–26. and engineering problems. M. Depending on the decisionmaking situation. W.. T. 5.. and Rajendran. T. 1994. Criticality analysis of spare parts using the analytic hierarchy process.
and Yen. Modeling the metrics of lean. Martinez.. 1999. S. and Lieberman. Information and Control 8:338–53. Tiwari. W. 12. C. R. AIAA Journal 39:517–25. NJ: Prentice Hall. C. An algorithm for fuzzy multicriteria decision making. C. © 2009 by Taylor & Francis Group. 2001. Ignizio. Engineering Economist 40:359–76. Analyzing alternatives in reverse logistics for endoflife computers: ANP and balanced scorecard approach. Fuzzy Sets and Systems 24:27–34. M. and RaisRohani. Y. 1. S. I. J. conflicting objectives and both qualitative and quantitative data. Manufacturabilitybased optimization of aircraft structures using physical programming. Zeleny. New York: McGrawHill. Fuzzy Sets and Systems 1:45–55. G. Zimmermann. Hwang. Agarwal.. H. European Journal of Operations Research 95:649–55. Goal programming and extensions. S. G. M. Fuzzy sets. and Hwang. Chan. Tsaur.. A.. K. Heath and Company. Computers and Industrial Engineering 48:327–56. Benefit/cost analysis using fuzzy concept. Multiple attribute decision making: Methods and applications.. Chan. Multiple criteria decision making. International Journal of Approximate Reasoning 21:215–31. 18. Erol. 1978. 9. P.. and Deng. New York: Springer. and Akbulut. p.. A. R. Messac. Hillier. and Yoon. Yeh. L.. H. agile and leagile supply chain: An ANP based approach. M. T.70 Strategic Planning Models for Reverse and ClosedLoop Supply Chains 7. A. 2003. 21. 1982. F. J. 26. Berlin: Springer. International Journal of Production Economics 86:187–99. F.. M. H. and Tiwari. 20. LLC . 1976... An integrated fuzzy decision support system for multicriterion decision making problems. Extent analysis and synthetic decision. K. Vol.. T. H. S. Fuzzy programming and linear programming with several objective functions. New York: McGrawHill. 1992. S. Linear physical programming: A new approach to multiple objective optimization.. Shankar. MA: Lexington Books. 28. Introduction to operations research. Chang D. The evaluation of airline service quality by fuzzy MCDM. 1997. S. A methodology for selecting problems with multiple. M. Applications of the extent analysis method on fuzzy AHP. Computers and Industrial Engineering 50:35–54. J. H. A. Zadeh. 19. K. 24. and Liang. R. Transactions on Operational Research 8:39–59. 1564–68. J. 2003. 1965. Ravi. Linear programming in single and multi objective systems. European Journal of Operational Research 173:211–25. Gungor.. Singapore: World Scientific. L. L. Fuzzy multiple attribute decision making: Methods and applications. 10. Fuzzy goal programming—An additive model.. Messac.. R.. G.. Journal of Engineering Manufacture 217:11–27. M. 1981. In Proceedings of the IEEE First International Conference on Intelligent Processing Systems. J. pp. and Tiwari. 17. Ignizio. Englewood Cliffs. M. 2006. 8. Chen. 11. 13. B. 22. H. and Ferrell. D.. Evaluation of connection types in design for disassembly (DFD) using analytic network process. Wang. Deng. 25. 27. and Rao. 1992. 2001. P.. Dharmar. 15. Tourism Management 23:107–15. Chang. Y. V. A. optimization techniques and applications. J.. Multicriteria analysis with fuzzy pairwise comparison. Chang. 1987. M. Gupta. Lexington. K. C. 2005. 23. 1995. 352. 1982. 2006. 2002. 14. D. 1996. Shankar. N. 1996. 16. C. Jr. and Chan.
N.Quantitative Modeling Techniques 71 29.. C. Artificial intelligence. Journal of Real Estate Literature 15:443–48. and Lin. 2007. G.org/wiki/Geographic_information_system. 34. Viswanadham. 8th ed.html. 32. G. http://en. S. Group decision making under multicriteria: Methods and applications. and Knight. Masi.. Knowledgebased systems for engineers and scientists. 37. and Mauborgne. Triantaphyllou. 2002. and Lieberman. J. 2nd ed. R. 42. Hwang. 33. New York: SpringerVerlag. 40. C. 1998.. J. W. 36. and Bhattacharya. Narahari. 2007. Englewood Cliffs. Operations management: Processes and value chains. Neural networks: A comprehensive foundation. 77–82. Hopgood. New York: McGrawHill. Hillier. Introduction to operations research. P. © 2009 by Taylor & Francis Group. E. and Ritzman. E. 38. E. 1993. A..softwareillustrated. International Journal of Approximate Reasoning 14:281–310.wikipedia. 1151–56. Y. Supplier evaluation and selection via Taguchi loss functions and an AHP.. C. 2004. Estaville. 2006. W. http://www. 43. S. June. NJ: Prentice Hall. Charting your company’s future. 31. 2003. Satisfaction assessment of multiobjective schedules using neural fuzzy methodology. International Journal of Production Research 41:1831–49. International Journal of Advanced Manufacturing Technology 27:625–30. 2005. 2000. 39.. LLC . 7th ed. Rich. L. A. pp. 30. NJ: Pearson Prentice Hall. 1996. M.. New York: McGrawHill. L. Englewood Cliffs.. Control Engineering 54:62–66. 1987. GIS and colleges of business: A curricular exploration. Boca Raton. and Low. Fuzzy neural control systems—explained. K.. C. In IEEE International Conference on Robotics and Automation. Krajewski. FL: CRC Press. Cha. Harvard Business Review. R. Y. and Jung. pp. Development and evaluation of five fuzzy multiattribute decisionmaking methods. Kim. Design of synchronized supply chains: A Six Sigma tolerancing approach. C. L.com/mapland. F. 1992. 35. 41. L. Haykin. Pi.
section 5. In this chapter.5 Selection of Used Products 5. there are also many thirdparty companies that collect used products solely to make profit.2. section 5. LLC . Evaluation criteria could be presented in terms of ranges of different degrees of desirability. This chapter is organized as follows: section 5. 73 © 2009 by Taylor & Francis Group.2 presents the formulation of the model. especially in Europe.2 First Model (Linear Integer Programming) Section 5. 5. and section 5. 2. although many original equipment manufacturers (OEMs) are obligated to take products back from the consumers upon the products’ end of use (hence called used products).2 presents the first model. two models to address the above scenarios are presented. These companies select only those used products for which revenues from recycle or resale of the products’ components are expected to be higher than the costs involved in collection and reprocessing of used products and in disposal of waste.4 gives some conclusions.1 presents the nomenclature for formulation of the linear integer programming model.3 gives a numerical example to illustrate the model. The various scenarios for selecting economical used products could differ as follows: 1.1 The Issue In many countries.2. and section 5. Evaluation criteria could be presented in terms of classical numerical constraints.3 shows the second model. The first model addresses scenario 1 and employs linear integer programming.2. The second model is for scenario 2 and employs linear physical programming.
© 2009 by Taylor & Francis Group. LLC . recycling revenue.2 The costbenefit function [1] in the literature presents a technique to select the best used product for reprocessing from a set of candidate used products. 10 = highest) Root node of used product x Supply of used product x per period Time to disassemble subassembly k in used product x Weight of component y in used product x Used product type Decision variable signifying selection of component y to be retrieved from used product x for reuse (Xxy = 1 for reuse. 10 = highest) Subassembly k in used product x Probability of missing component y in used product x Number of subassemblies in used product x Multiplicity of component y in used product x Probability of breakage of component y in used product x Component y in used product x Percent of recyclable contents by weight in component y of used product x Total recycling revenue of used product x Total resale revenue of used product x Resale value of component y in used product x Recycling revenue index of component y in used product x (0 = lowest. 0 for recycle) Component type Overall profit Model Formulation PRCxy Rrcx Rrsx Rrsxy RCxy Rootx SUx T(Exk) Wxy x Xxy T(Rootx) Time to disassemble Rootx y Z 5.2.1 Cdf Cdx Cr Crf Crpx CCx Dxy Exk mxy Mx Nxy pxy Pxy Strategic Planning Models for Reverse and ClosedLoop Supply Chains Nomenclature Disposal cost factor (cost per unit weight) Disposal cost of product x Reprocessing cost per unit time Recycling revenue factor (revenue per unit weight) Total reprocessing cost of used product x Collection cost of used product x Disposal cost index of component y in used product x (0 = lowest.74 5. The function consists of four terms: resale revenue.2.
and disposal cost.2. total disposal cost (Cdx). the number of components (Nxy).N xy . LLC . the weight of the components (Wxy).2.1) The terms in the function are described as follows: Total resale revenue: Rrsx is influenced by the resale value of individual components of the used product (Rrsxy). The optimal solution (for profit maximization) also gives the best feasible set of components of retrieval from the selected used product.2. the number of components (Nxy).1 presents the modified costbenefit function. total reprocessing cost (Crpx). The used product with the maximum costbenefit function value is selected as the optimal one for reprocessing.2. X xy (5.(1 pxy mxy ). in order to select the most economical used product to reprocess. and the probability of missing components (mxy). Then a linear integer programming model is formulated and implemented to the data of each candidate used product.2. and (2) that all the components in the used product are in their original multiplicities.Selection of Used Products 75 reprocessing cost. the recycling revenue index (RCxy). a modified costbenefit function is presented. The difference between the sum of the revenue terms and the sum of the cost terms gives the costbenefit function value. This function incorporates the probability of breakage and the probability of missing components in the used product of interest. However. Therefore.2 gives the linear integer programming model. which is a number in the range 1–10 representing the degree of benefit generated by recycling a component of type y (the higher the value.1 Modified CostBenefit Function The modified costbenefit function for used product x consists of five terms: total resale revenue (Rrsx). 5. total recycling revenue (Rrcx). the more profitable it is to recycle). The revenue equation can be written as follows: Rrsx j Pxy ( Root x ) Rrsxy .2. there is serious risk of making a bad choice of the used product using the costbenefit function because the function implicitly makes two assumptions that are seldom valid in a reverse supply chain scenario: (1) every component selected for resale will be in a reusable state after disassembling the used product. and section 5.2) Total recycling revenue: Rrcx is influenced by the percentage of recyclable contents in each component (PRCxy). It can be written as Z Rrsx Rrcx Crpx Cdx CCx (5. here. the probability of break © 2009 by Taylor & Francis Group. the probability of breakage (pxy). and collection cost. Section 5.
6) subject to Xxy = 0 or 1 for all x and y (5.( N xy (1 mxy ) N xy (1 pxy mxy ).4) Total disposal cost: Cdx can be calculated from the disposal cost index (Dxy).RCxy ( N xy (1 mxy ) N xy (1 pxy mxy )X xy ) Crf (5.2. the weight of the components (Wxy). LLC .76 Strategic Planning Models for Reverse and ClosedLoop Supply Chains age (pxy). the probability of missing components (mxy). the disassembly time of each subassembly in the used product (T(Exk)). the more difficult it is to dispose).2. which is a number in the range 1–10 representing the degree of difficulty in disposing component y of used product x (the higher the number. and the disposal cost factor (Cdf). and the probability of missing components (mxy).3) Total reprocessing cost: Crpx can be calculated from the disassembly time of the root node of the used product (T(Rootx)). The disposal cost equation can be written as follows: Cdx j Pxy ( Root x ) Dxy .Wxy . X xy ) . the probability of breakage (pxy). the number of components (Nxy).Cr (5.Wxy .7) © 2009 by Taylor & Francis Group. The recycling revenue equation can be written as follows: Rrcx j Pxy ( Root x ) PRCxy .2 Linear Integer Programming Model The following linear integer programming model maximizes the costbenefit obtained from reprocessing used product x: Maximize Z x Rrsx Rrcx Crpx Cdx CCx (5.(1 PRCxy ).5) Collection cost: CCx is the average cost of collecting used product x from the consumers. and the reprocessing cost per unit time (Cr). The total reprocessing cost equation can be written as follows: Mx Crpx T ( Root x ) k 1 T ( Exk ) .Cdf (5. 5. the percentage of recyclable contents in each component (PRCxy).
© 2009 by Taylor & Francis Group. The data required to implement the model for the products are shown in tables 5.3 Numerical Example Two used products (1 and 2) are considered in this numerical example (see figures 5.2.2 Structure of used product 2 (first model). 5. T(E21) = 2 min. T(E12) = 5 min. T(Root1) = 6 min.1 and 5. Cr = 0. CC2 = 40.1 and 5.2).1 Structure of used product 1 (first model). Root2 P 21 E 22 P 23 E 21 P 24 P 25 P 22 FIGURE 5. Also.25/lb. and T(E22) = 4 min. LLC . respectively. T(Root2) = 4 min. Root1 E 11 P 11 E 12 P1 P 14 P 15 P1 FIGURE 5.Selection of Used Products 77 The above formulation assesses the feasible combinations of components’ retrieval from a used product and compares the combination with the highest costbenefit from one product against others. Cdf = 0.2. T(E11) = 3 min.5.8/min. Crf = 0. CC1 = 25.
3 Second Model (Linear Physical Programming) Section 5.1.2. 5.3.2 0.78 Strategic Planning Models for Reverse and ClosedLoop Supply Chains TABLE 5.1 Model Formulation The criteria considered in the model fall into either class 1S or class 2S. and section 5.3.1 0.0 0. LLC .05 0. Hence.5 5 3 0.1 Class 1S Criteria (Smaller Is Better) Total collection cost per period (g1): g1 of used product x is calculated by multiplying the supply of x per period (SUx) by the cost of collecting one product from consumers (CCx): © 2009 by Taylor & Francis Group.1 m2y 0.2 gives a numerical example to illustrate the model.5 2 N2y 2 3 1 2 2 W2y 4 1 3 4.4 m1y 0.1.5 5 Rc2y 5 6 2 8 7 PRC2y 40% 35% 70% 80% 25% D2y 2 6 8 5 7 p2y 0.1 Data of used product 1 (first model) Part P11 P12 P13 P14 P15 Rrs1y 10 2 3.02 TABLE 5.0 0.3 0. one gets total profit for used product 1 as $46. Section 5.1 0. 5.0 0.2 gives the class 2S criteria.16 and total profit for used product 2 as $68.1.3.3.5 5 7 4 Rc1y 7 9 5 3 6 PRC1y 75% 60% 50% 70% 40% D1y 3 2 6 1 7 p1y 0.75 5 3 N1y 4 2 1 5 6 W1y 3 2.25 Upon solving the model with the above data using LINGO (v4).2 Data of used product 2 (first model) Part P21 P22 P23 P24 P25 Rrs2y 2. and section 5. the decision maker will select used product 2 in this case.0 0.50.0 0.05 0. 5.1 presents the class 1S criteria.3.15 0.02 0.1 presents the formulation of the linear physical programming model.5 0.0 0.1 0. The nomenclature used to formulate the model is the same as in section 5.3.1.
disassembly time of each subassembly (T(Exk)).(1 mxy y bxy ) (5. and the remanufacturing cost per unit time (Cr): Mi SU x T ( Root x ) k 1 T ( Exk ) Cr (5. and Cdf is the disposal cost factor. and the breakage and missing probabilities of component y (pxy. It can occur because of the unpredictability in the supply of used products. This can be obtained from an expert in the field. as follows: SU x . mxy).8) Total reprocessing cost per period (g2): g2 of used product x is calculated using the disassembly time of the root node (T(Rootx)). supply of x per period (SUx).N xy . the resale value of component y (RSRxy).1. This can also be obtained from an expert in the field.9) Total disposal cost per period (g3): g3 of used product x is calculated by multiplying the component disposal cost by the number of units of components disposed. LLC .10) DIxy is the disposal cost index that varies in value from 1 to 10 representing the degree of nuisance created by the disposal of component y of product x.2 Class 2S Criteria (Larger Is Better) Total reuse revenue per period (g6): g6 of used product x is influenced by the supply of x per period (SUx). Worth of investment cost (g5): g5 of used product x represents the periodic worth of the fixed cost of the production facility and the machinery required to reprocess.Selection of Used Products 79 g1 SU x CCx (5. 5. Lossofsale cost (g4): g4 of used product x represents the periodic worth of not meeting the demand on time.11) © 2009 by Taylor & Francis Group.DI xyWxy (1 PRCxy ) N xy (1 mxy ) N xy (1 bxy y mxy ) Cdf (5.3. the multiplicity of component y (Nxy). The reuse revenue equation can be written as follows: SU x .RSRxy .
12) 5.9–5.34) (using the incremental weights from the LPP algorithm and deviations from target values) and is shown in table 5.. the recycling revenue factor (Crf). respectively.80 Strategic Planning Models for Reverse and ClosedLoop Supply Chains Total recycling revenue per period (g7): g7 of used product x is influenced by the supply of x per period (SUx).12. the percentage of recyclable content in component y (PRCxy).11 show the deviations of criteria values from target values for the three products. and the second model employs linear physical programming. 7. respectively.3–5. Because alternatives with lower scores are more desirable than ones with higher scores. shown in bold in table 5.RCRI xy . and table 5.N xy . respectively.e.8 shows the incremental weights obtained using the LPP weight algorithm [2].7 shows the criteria values for each product. shown in bold in table 5. 10. The target values for the criteria are given in table 5. N xy (1 mxy ) N xy (1 mxy y bxy ) Crf (5. the multiplicity of component y (Nxy). used product 3 is the best of the lot.. The data for the three products are given in tables 5. and the breakage and missing probabilities of component y (pxy.Wxy .3. mxy). 5. two models are presented for selecting the most economical product to reprocess from a set of candidate used products. deviation at s = 2 is the absolute value of the number obtained by subtracting the criteria value (i. for criteria g1 of product 3.3–5.5. © 2009 by Taylor & Francis Group. and 5).6. Table 5.e. Tables 5.5.7) from the target value (i. the recycling revenue index of component y (RCRIxy). The recycling revenue equation can be written as SU x . The total score for each product is calculated using equation (4. For example. The first model uses linear integer programming. whose structures are shown in figures 5. the weight of component y (Wxy).5. 4.2 Numerical Example Consider three used products (3.4 Conclusions In this chapter. LLC .6).PRCxy .
LLC . Root5 E 52 E 51 P 52 P 53 P 51 P 54 P 55 FIGURE 5. Root4 P 41 E 42 P 43 E 41 P 44 P 45 P 42 FIGURE 5.4 Structure of used product 4 (second model).5 Structure of used product 5 (second model). © 2009 by Taylor & Francis Group.3 Structure of used product 3 (second model).Selection of Used Products 81 Root3 E 31 P 31 E 32 P3 P 34 P 35 P3 FIGURE 5.
82
Strategic Planning Models for Reverse and ClosedLoop Supply Chains
TABLE 5.3 Data of used product 3 (second model)
Part P31 P32 P33 P34 P35 RSRxy 6 7 5 7.5 5.5 Nxy 3 2 4 2 1 W1y 3.5 4.5 5 6 5.5 RCRIxy 5 4 5 3 2 PRCxy 0.65 0.4 0.3 0.25 0.45 DIxy 6 4 4 5 1 pxy 0.1 0.2 0.4 0.2 0.5 mxy 0.3 0.5 0.1 0.2 0.0
TABLE 5.4 Data of used product 4 (second model)
Part P41 P42 P43 P44 P45 RSRxy 4 6 7 2 4.5 Nxy 2 5 4 2 3 W2y 9 3 5 6 1 RCRIxy 2 1 2 3 4 PRCxy 0.56 0.5 0.48 0.2 0.25 DIxy 3 6 6 1 3 pxy 0.2 0.2 0.4 0.1 0.1 mxy 0.5 0.2 0.1 0.1 0.3
TABLE 5.5 Data of used product 5 (second model)
Part P51 P52 P53 P54 P55 RSRxy 4.5 4 5 2 3.5 Nxy 3 4 6 5 1 W3y 8.5 6 4.1 3.5 2 RCRIxy 5 4 2 3 1 PRCxy 0.4 0.65 0.2 0.35 0.25 DIxy 5 6 3 2 7 pxy 0.1 0.1 0.2 0.3 0.5 mxy 0.3 0.1 0.0 0.2 0.5
© 2009 by Taylor & Francis Group, LLC
Selection of Used Products
TABLE 5.6 Target values of criteria (second model)
Criteria g1 g2 g3 g4 g5 Criteria g6 g7 tp1– 10 5 tp1+ 10 3 1 2 1 tp2– 12.5 13 tp2+ 12 5 4 5 4 tp3– 15 17.5 tp3+ 15 9 5.5 7.5 8 tp4– 20 20 tp4+ 17.5 10 7.5 9 9 tp5+ 20 15 8 10 10 tp5– 25.5 25
83
TABLE 5.7 Criteria values for each product (second model)
Criteria g1 g2 g3 g4 g5 g6 g7 Used product 3 7.5 3.5 2.97 3.5 2 18.3 18.4 Used product 4 7.5 2.75 1.6 4 2.5 22.8 15.7 Used product 5 10 3 1.8 5 2 24.9 19
TABLE 5.8 Output of LPP weight algorithm (second model)
Criteria g1 g2 g3 g4 g5 g6 g7 Δwp2+ 0.05 0.05 0.0333 0.0333 0.0333 — — Δwp3+ 0.0967 0.076 0.3027 0.16827 0.09267 — — Δwp4+ 0.62773 2.41416 0.93408 1.49184 2.41416 — — Δwp5+ 2.63296 0.020321 24.33473 11.10897 10.26225 — — Δwp2– — — — — — 0.04 0.0125 Δwp3– — — — — — 0.0492 Δwp4– — — — — — Δwp5– — — — — — 0.022875
0.010258 0.122333
0.037056 0.14936
© 2009 by Taylor & Francis Group, LLC
84
Strategic Planning Models for Reverse and ClosedLoop Supply Chains
TABLE 5.9 Deviations of criteria values from targets, for used product 3 (second model)
Criteria g1 g2 g3 g4 g5 g6 g7 s=2 2.5 0.5 1.9 1.5 1 6.7 6.6 s=3 4.5 1.5 1.1 1.5 2 1.7 1.6 s=4 7.5 5.5 2.6 4 6 3.3 0.9 s=5 10 6.5 4.6 5.5 7 5.8 5.4
TABLE 5.10 Deviations of criteria values from targets, for used product 4 (second model)
Criteria g1 g2 g3 g4 g5 g6 g7 s=2 2.5 0.25 0.6 2 1.5 2.2 9.3 s=3 4.5 2.25 2.4 1 1.5 2.8 4.3 s=4 7.5 6.25 3.9 3.5 5.5 7.8 1.8 s=5 10 7.25 5.9 5 6.5 10.3 2.7
TABLE 5.11 Deviations of criteria values from targets, for used product 5 (second model)
Criteria g1 g2 g3 g4 g5 g6 g7 s=2 0 0 0.8 3 1 0.1 5.93 s=3 2 2 2.2 0 2 4.9 0.93 s=4 5 6 3.7 2.5 6 9.9 1.57 s=5 7.5 7 5.7 4 7 12.4 6.07
© 2009 by Taylor & Francis Group, LLC
Selection of Used Products
TABLE 5.12 Total scores and ranks of products (second model)
Used product 3 4 5 Score 315.3144 339.1954 317.8653 Rank 1 3 2
85
References
1. Veerakamolmal, P., and Gupta, S. M. 1999. Analysis of design efficiency for the disassembly of modular electronic products. Journal of Electronics Manufacturing 9:79–95. 2. Messac, A., Gupta, S. M., and Akbulut, B. 1996. Linear physical programming: A new approach to multiple objective optimization. Transactions of Operational Research 8:39–59.
© 2009 by Taylor & Francis Group, LLC
6
Evaluation of Collection Centers
6.1
The Issue
Strategic planning of an efficient reverse or closedloop supply chain requires selection of efficient collection centers where used products are disposed of by the consumers. These collection centers, after initial processing (for example, sorting), ship the used products to recovery facilities or production facilities where reprocessing operations, such as disassembly and recycling/ remanufacturing, are carried out. The various scenarios for evaluating collection centers for efficiency could differ as follows: 1. Supply chain company executives, whose primary concern is profit, could be the sole decision makers. 2. There could exist three different categories of decision makers: consumers, local government officials, and supply chain company executives. Impacts* of evaluation criteria are given. 3. The situation could be the same as in scenario 2, but the impacts of evaluation criteria are not given (and hence must be derived). 4. Evaluation could be made from the perspective of a remanufacturing facility interested in buying used products from the candidate collection centers. The goals are expressed in terms of performance indices (efficiency scores). 5. The situation could be the same as in scenario 4, but the goals are expressed in terms of Taguchi losses (inefficiency scores). In this chapter, various models to address the above scenarios are presented. The first model addresses scenario 1 and employs the eigen vector method and Taguchi loss function. The second and third models are for scenarios 2 and 3, respectively. The main difference between these two models
*
In this chapter, we use the term impacts instead of the conventional term weights for importance values of evaluation criteria. This is to avoid confusion in the presence of the connection weights between different nodes of the neural network used in section 6.5.
87
© 2009 by Taylor & Francis Group, LLC
4–6.5. The main difference between the last two models is in the way the goals are expressed. whose primary concern is profit.1 presents the evaluation criteria used in the model.2 illustrates the model using a numerical example. the eigen vector method and Taguchi loss function are employed to identify efficient centers from a set of candidate collection centers operating in a region where a reverse supply chain is to be designed.7 present the second. are considered. the evaluation criteria do not consider factors such as “Is the collection center able to provide employment opportunities to the local community?” and “Is the collection process convenient to the consumers?” (For factors such as these. fuzzy logic. This chapter is organized as follows: Section 6. respectively. The n value of a collection center represents the center’s quality that could be a function of factors. Whereas the second model uses the eigen vector method.) The following are the evaluation criteria used in the model: n value (the higher the n value. and goal programming. technique for order preference by similarity to ideal solution (TOPSIS).1 Evaluation Criteria This model assumes that the supply chain company executives. For example (see figure 6.. where criteria of two additional different decision makers. such as functionality of used products collected. The fifth model. and effectiveness of customer service. Section 6. uses the eigen vector method. TOPSIS. efficiency of delivery (to recovery facilities). and Borda’s choice rule. let the upper specification (U) of the service time (critical © 2009 by Taylor & Francis Group. Taguchi loss function. efficiency of collection.88 Strategic Planning Models for Reverse and ClosedLoop Supply Chains lies in the way the impacts are assigned to the criteria for evaluation of collection centers.3 gives the different decision makers (and their criteria) that are considered in the second and third models. and fifth models. Hence. Sections 6. the lower the process defects. The fourth model is for scenario 4 and uses analytic network process (ANP) and goal programming. LLC .8 gives some conclusions.2. and Borda’s choice rule. consumers and local government officials. and section 6.16 for the concept of n Sigma). the third model uses neural networks.2.1). Section 6. viz.2 First Model (Eigen Vector Method and Taguchi Loss Function) In this model. and hence the higher the profit. which is for scenario 5.2. are the sole decision makers. see section 4. fourth. Finally.2 presents the first model.4 and 6. section 6. 6. 6. see sections 6. third.
5 min.. the collection center personnel must spend at least some time inspecting used products before accepting them from consumers). If the standard deviation (σ) of the service time is 2 min and the mean shift ( ) is 2. Table 6. Per capita income of people in residential area (PI) (the higher it is.1 Critical dimension of service process at candidate collection center (first model). the better) Labor cost (LC) (the lower.1 shows the pairwise comparison matrix for the criteria for evaluation of candidate collection © 2009 by Taylor & Francis Group. the consumers discarding used products must not be made to wait too long at the collection center). the better) Distance from residential area (DH) (lower distance implies greater collection and hence greater profit) Distance from roads (DR) (lower distance implies greater collection and hence greater profit) Incentives from local government (IG) (higher incentives from local government imply higher incentives to consumers.. the lower specification (L) be 10 units (i.Evaluation of Collection Centers 89 FIGURE 6.5 Sigma process. and the target value (τ) be 15 min. and hence greater collection) 6.2 Model The model is presented using a numerical example.e. then the service process in the candidate collection center is a 2. LLC . the better) Utilization of incentives from local government (UI) (the higher.2. dimension) at a candidate collection center be 20 min (i.e. the greater the number of resourceful discarded products and the less the people will care about the incentives from the collection center) Space cost (SC) (the lower.
90
Strategic Planning Models for Reverse and ClosedLoop Supply Chains
TABLE 6.1 Pairwise comparison matrix (first model)
Criteria n DH DR UI PI SC LC IG n 1 1 1 1/3 5 1 1 1 DH 1 1 1 1 1 1 1 1 DR 1 1 1 1 1 1 1 1 UI 3 1 1 1 3 5 1 1 PI 1/5 1 1 1/3 1 1 1 1 SC 1 1 1 0.2 1 1 1 5 LC 1 1 1 1 1 1 1 1 IG 1 1 1 1 1 0.2 1 1
centers. For example (see table 6.1), the per capita income of the people in the residential area (PI) is given five times more importance than the n value and three times more importance than the utilization of incentives from the local government (UI). Table 6.2 shows the impacts of the respective evaluation criteria. These impacts are the elements of the normalized eigen vector of the pairwise comparison matrix shown in table 6.1. For example (see table 6.2), the per capita income of the people in the residential area (PI) is given an impact of 18% and the n value is given an impact of 11%. The k value of the Taguchi loss function, in the case of each of the evaluation criteria, is calculated as follows. 6.2.2.1 n Value
The loss function that applies to this criterion is “larger is better” (see equation (4.68)). If the decision maker considers 100% loss for an n value less than or equal to 4, then the value of k is 1,600%. Then, for a candidate collection center, if n = 5, L( y ) 1600 / (5)2 64% . That means, with respect to quality
TABLE 6.2 Impacts of criteria (first model)
Criteria n DH DR UI PI SC LC IG Impacts 0.11 0.11 0.11 0.07 0.18 0.15 0.11 0.16
© 2009 by Taylor & Francis Group, LLC
Evaluation of Collection Centers
91
(i.e., n value), the collection center is 36% short of the worst performance level (which, in this case, is n ≤ 4). 6.2.2.2 Distance from Residential Area (DH)
DH is considered the distance of the collection center from the center of gravity [1] of all the residential areas around the center. The loss function that applies to this criterion is “smaller is better” (see equation (4.67)). If the decision maker considers 100% loss for a DH value more than or equal to 4 miles, then the value of k is 6.25%. Then, for a candidate collection center, if DH = 3, L( y ) 6.25 (3)2 56.25%. That means, with respect to the distance from the residential area, the collection center is 43.75% short of the worstcase scenario (which, in this case, is DH ≥ 4 miles). 6.2.2.3 Distance from Roads (DR)
DR is considered the average distance of all the roads in the region from the collection center of interest. The loss function that applies to this criterion is “smaller is better” (see equation (4.67)). If the decision maker considers 100% loss for a DR value more than or equal to 5 miles, then the value of k is 4%. Then, for a candidate collection center, if DR = 4, L( y ) 4 (4 )2 64%. That means, with respect to the distance from the roads, the collection center is 36% short of the worstcase scenario (which, in this case, is DR ≥ 5 miles). 6.2.2.4 Utilization of Incentives from Local Government (UI)
Because this is a subjective criterion, the decision maker can obtain ratings (for example, on a 1–10 scale, where 1 is the worst and 10 is the best) of the candidate collection centers from experts in the field of reverse supply chain. The loss function that applies to this criterion is “larger is better” (see equation (4.68)). If the decision maker considers 100% loss for a UI value less than or equal to 5, then the value of k is 2,500%. Then, for a candidate collection center, if UI = 7, L( y ) 2500 / (7 )2 51.02%. That means, with respect to the utilization of incentives from the local government, the collection center is about 49% short of the worstcase scenario (which, in this case, is UI ≤ 5). 6.2.2.5 Per Capita Income of People in Residential Area (PI)
The loss function that applies to this criterion is “larger is better” (see equation (4.68)). If the decision maker considers 100% loss for a PI value less than or equal to $30,000 per year, then the value of k is 90,000,000,000%. Then, for a candidate collection center, if PI = $50,000 per year, L( y ) 90, 000, 000, 000 / (50, 000)2 36%. That means, with respect to the per capita income of the people in the residential area, the collection center is 64% short of the worstcase scenario (which, in this case, is PI ≤ $30,000 per year).
© 2009 by Taylor & Francis Group, LLC
92 6.2.2.6
Strategic Planning Models for Reverse and ClosedLoop Supply Chains Space Cost (SC)
The loss function that applies to this criterion is “smaller is better” (see equation (4.67)). If the decision maker considers 100% loss for a SC value more than or equal to $1,000 per day, then the value of k is 0.0001%. Then, for a candidate collection center, if SC = $800 per day, L( y ) 0.0001 (800)2 64% . That means, with respect to the space cost, the collection center is 36% short of the worstcase scenario (which, in this case, is SC ≥ $1,000 per day). 6.2.2.7 Labor Cost (LC)
The loss function that applies to this criterion is “smaller is better” (see equation (4.67)). If the decision maker considers 100% loss for a LC value more than or equal to $15 per hour, then the value of k is 0.44%. Then, for a candidate collection center, if LC = $10 per hour, L( y ) 0.44 (10)2 44% . That means, with respect to the labor cost, the collection center is 56% short of the worstcase scenario (which, in this case, is LC ≥ $15 per hour). 6.2.2.8 Incentives from Local Government (IG)
Because this is a subjective criterion, the decision maker can obtain ratings (for example, on a 1–10 scale, where 1 is the worst and 10 is the best) of the candidate collection centers from experts in the field of reverse supply chain. The loss function that applies to this criterion is “larger is better” (see equation (4.68)). If the decision maker considers 100% loss for an IG value less than or equal to 7, then the value of k is 4,900%. Then, for a candidate collection center, if IG = 9, L( y ) 4900 / (9)2 60.49% . That means, with respect to the incentives from the local government, the collection center is about 39.5% short of the worstcase scenario (which, in this case, is IG ≤ 7). Four candidate collection centers are considered in the numerical example: A, B, C, and D. Table 6.3 presents the L(y) for the evaluation criteria for each of the collection centers. For example, the Taguchi loss of B with respect to the per capita income of the people in the residential area is 52% (i.e., 48% short of the worstcase scenario). The weighted loss of each collection center j is calculated by using the following equation and is presented in table 6.4. Weighted loss of collection center j =
i
Wi Lij
(6.1)
where Wi is the impact of criterion i (see table 6.2) and Lij is the Taguchi loss (see table 6.3) of collection center j with respect to criterion i. For example, the weighted loss of D (see tables 6.2–6.4) is 0.11×63 + 0.11×55 + … + 0.11×67 + 0.16×41 = 64.31. The decision maker will select C because it has the lowest weighted loss.
© 2009 by Taylor & Francis Group, LLC
Evaluation of Collection Centers
TABLE 6.3 L(y) values (%) of collection centers (first model)
Criteria n (0.11) DH (0.11) DR (0.11) UI (0.07) PI (0.18) SC (0.15) LC (0.11) IG (0.16) A 35 25 32 100 65 20 15 78 B 54 15 10 75 52 10 18 64 C 40 37 9 64 40 5 20 36 D 63 55 90 50 50 100 67 41
93
TABLE 6.4 Weighted losses of collection centers (first model)
Collection center A B C D Weighted loss 45.95 37.02 29.85 64.31
6.3
Evaluation Criteria for Second and Third Models
In evaluation of collection centers, one may have a scenario with three different categories of decision makers with multiple, conflicting, and incommensurate goals, as follows: Consumers whose primary concern is convenience Local government officials whose primary concern is environmental consciousness Supply chain company executives whose primary concern is profit Therefore, the efficiency of a candidate collection center must be evaluated based on the maximized consensus among decision makers of the three categories. Sections 6.3.1, 6.3.2, and 6.3.3 present the lists of criteria that are considered for the three categories. Note that the criteria for the third category (supply chain company executives) are the same as those listed in section 6.2.1. 6.3.1 Criteria of Consumers Incentives from collection center (IC) (higher incentives imply higher motivation to participate)
© 2009 by Taylor & Francis Group, LLC
94
Strategic Planning Models for Reverse and ClosedLoop Supply Chains Distance from residential area (DH) (lower distance implies higher motivation to participate) Distance from roads (DR) (lower distance implies higher motivation to participate) Simplicity of collection process (SP) (simpler process implies higher motivation to participate) Employment opportunity (EO) (the higher, the better) Salary offered to employees at collection center (SA) (the higher, the better)
6.3.2
Criteria of Local Government Officials Distance from residential area (DH) (lower distance implies greater collection and hence lower disposal) Distance from roads (DR) (lower distance implies greater collection and hence lower disposal)
6.3.3
Criteria of Supply Chain Company Executives n value (the higher the n value, the lower the process defects, and hence the higher the profit; see section 4.16 for the concept of n Sigma) Per capita income of people in residential area (PI) (the higher it is, the more the number of resourceful discarded products, and the less the people will care about the incentives from the collection center) Space cost (SC) (the lower, the better) Labor cost (LC) (the lower, the better) Utilization of incentives from local government (UI) (the higher, the better) Distance from residential area (DH) (lower distance implies greater collection and hence greater profit) Distance from roads (DR) (lower distance implies greater collection and hence greater profit) Incentives from local government (IG) (higher incentives from local government imply higher incentives to consumers and hence greater collection)
© 2009 by Taylor & Francis Group, LLC
and G. because the focus is on the methodology. 6. respectively. as evaluated by that category.8–6. a GISbased business mapping application. Note that not all of the criteria are considered here. are the elements of the normalized eigen vector of the pairwise comparison matrix shown in table 6. E.3). Three recovery facilities. the impacts of the criteria DH (distance from residential area) and DR (distance from roads). are the elements of the normalized eigen vectors of pairwise comparison matrices shown in tables 6.13). LLC .10 show the impacts given for the criteria of the consumers.Evaluation of Collection Centers 95 6. respectively. To facilitate the construction of the pairwise comparison matrices (see tables 6. and Borda’s Choice Rule) This model to select efficient collection centers is implemented in two phases.1 Phase I (Individual Decision Making) Tables 6. for implementation of the TOPSIS for each category of decision makers. impacts are given to the criteria identified for each category of decision makers (see section 6. is used to map the results obtained in both phases of this model. using the eigen vector method.7 show the pairwise comparison matrices as formed for the consumers. Furthermore. and supply chain company executives. local government officials. are shown in tables 6. respectively. local government officials. MapLand [4]. selection of particular collection centers).8–6.11–6. For example.4 Second Model (Eigen Vector Method. In the second phase. A lower rank implies higher efficiency (with respect to that criterion). are considered for evaluation.4. The decision matrices for the consumers. local government officials.13.6. representatives from each category of decision makers could be invited to participate © 2009 by Taylor & Francis Group. and supply chain company executives.5–6.5–6. motivated by a paper [3] that addresses the necessity for building the strategic planning process “around a picture” to make the chairman of the concerned supply chain company easily understand the “dense documents filled with numbers” and to convince him that it is important to implement the proposed action (here. shown in table 6.10) and the decision matrices (see tables 6.7. Borda’s choice rule is used to combine individual evaluations for each candidate collection center into a group evaluation or maximized consensus ranking.9. The model to select efficient collection centers is presented using a numerical example. respectively. TOPSIS. and then TOPSIS (technique for order preference by similarity to ideal solution) is employed to find the efficiency of each candidate collection center. The elements of these matrices are the ranks (ranging from 1 to 10) assigned to the collection centers with respect to each criterion for evaluation. and supply chain company executives.11–6. In the first phase. calculated using the eigen vector method. F. These sets of impacts. Tables 6.
LLC .1938 © 2009 by Taylor & Francis Group.1434 0.1515 0.96 Strategic Planning Models for Reverse and ClosedLoop Supply Chains TABLE 6.7 Pairwise comparison matrix for supply chain company executives (second model) Criteria PI SC LC UI DH DR IG PI 1 1/2 1/4 1/6 1/8 1/9 1/4 SC 2 1 1/2 1/6 1/8 1/9 1 LC 4 2 1 1 1 1 1 UI 6 6 1 1 1 1 1 DH 8 8 1 1 1 1 1 DR 9 9 1 1 1 1 1 IG 4 1 1 1 1 1 1 TABLE 6.6 Pairwise comparison matrix for local government officials (second model) Criteria DH DR DH 1 1/2 DR 2 1 TABLE 6.8 Impacts for consumers (second model) Criteria IC DH DR SP EO SA Impacts 0.2533 0.0960 0.5 Pairwise comparison matrix for consumers (second model) Criteria IC DH DR SP EO SA IC 1 1 1 1/2 2 1 DH 1 1 1/2 1 1 2 DR 1 2 1 1 7 2 SP 2 1 1 1 1 1 EO 1/2 1 1/7 1 1 1 SA 1 1/2 1/2 1 1 1 TABLE 6.1621 0.
0927 TABLE 6.0634 0. LLC .Evaluation of Collection Centers TABLE 6.0781 0.0598 0.3333 97 TABLE 6.10 Impacts for supply chain company executives (second model) Criteria PI SC LC UI DH DR IG Impacts 0.0585 0.11 Decision matrix for consumers (second model) Collection centers E F G IC 8 2 3 DH 1 1 2 DR 6 7 4 SP 2 3 1 EO 2 2 1 SA 3 3 5 TABLE 6.9 Impacts for local government officials (second model) Criteria DH DR Impacts 0.3876 0.12 Decision matrix for local government officials (second model) Collection centers E F G DH 2 2 3 DR 1 1 2 TABLE 6.13 Decision matrix for supply chain company executives (second model) Collection centers E F G PI 1 2 10 SC 3 4 9 LC 2 7 8 UI 1 1 7 DH 3 2 5 DR 4 2 1 IG 3 2 6 © 2009 by Taylor & Francis Group.6667 0.2599 0.
Step 3: Determine the ideal and negativeideal solutions.6667 0.14 shows the normalized decision matrix formed by applying equation (4.1478 0.0669 0.4082 (see table 6.6965 0. Each column in the weighted normalized decision matrix shown in table 6.. individual interviews..2673 EO 0.0382 SP 0.15 has a TABLE 6.1689 0.49) on each element of table 6. Now one is ready to perform the six steps in the TOPSIS for each category of decision makers.1150 0.8018 0.1515 (see table 6. F. i.1237 DR 0.1689 0.14).6667 0. Table 6. LLC . the normalized rank of collection center F with respect to criterion DH (see tables 6.0554 DH 0.3980 SP 0.8 and the normalized decision matrix in table 6. is calculated by multiplying the impact of DH. The following steps show the implementation of the TOPSIS for the consumers to evaluate E. 0. 0.14) is calculated as follows: r22 1 1 2 12 22 0. the weighted normalized rank of collection center F with respect to criterion DH.4082 0. 0.0767 0.0619 0.4575 0.98 Strategic Planning Models for Reverse and ClosedLoop Supply Chains in relevant survey questionnaires. Step 1: Construct the normalized decision matrix. This is constructed using the impacts of the criteria listed in table 6..15). focus groups. and onsite observations.0887 0. For example.0844 SA 0.3419 DH 0.e.3333 SA 0.2279 0.5345 0.0370 0.14. For example.14 Normalized decision matrix for consumers (second model) Collection centers E F G IC 0.5970 0.0887 0.0619 (see table 6.0573 0.e.e.9117 0. i.1478 © 2009 by Taylor & Francis Group.0619 0.15 Weighted normalized decision matrix for consumers (second model) Collection centers E F G IC 0. Table 6. Step 2: Construct the weighted normalized decision matrix.7625 TABLE 6. i.11 and 6.11 (decision matrix for the consumers).15 shows the weighted normalized decision matrix for the consumers. and G.8165 DR 0. with the normalized rank of F with respect to DH.0383 EO 0.4082 0.4082.4575 0.8).
and E (that means G is the best collection center. Step 5: Calculate the relative closeness coefficient.1400 0..0875 S– 0. LLC .16) is calculated using equation (4.1400) to the sum (i. the relative closeness coefficient is calculated for each collection center (see table 6. as evaluated by the consumers)..53).1400 + 0.5435) for collection center F (see table 6. for the corresponding criterion.1176 0. 0. Using equation (4. Figures 6. 99 The TOPSIS is implemented for the local government officials and the supply chain company executives in a similar manner. Step 6: Form the preference order.1237. Step 4: Calculate the separation distances.2576) of its negative and positive separation distances (see table 6. The separation distances (see table 6. 0. They are the ideal and negativeideal solutions.0619. For example (see table 6.Evaluation of Collection Centers minimum rank and a maximum rank.54).18.52).1495 TABLE 6. for relative closeness coefficients.3–6. with respect to criterion DH.15) and the ideal solutions (obtained in step 3) for the criteria. The relative closeness coefficients of the collection centers.17) is the ratio of F’s negative separation distance (i.15).e. 0.3926 0.16 Separation distances for consumers (second model) Collection centers E F G S* 0.2 shows the mapping classification.6308 © 2009 by Taylor & Francis Group. are shown in table 6.16) for each collection center are calculated using equations (4. and the negativeideal solution (maximum rank) is 0.16). the ideal solution (minimum rank) is 0.17).52) and (4. as created using MapLand [4].e. which contains the weighted normalized ranks of F (see table 6.. the relative closeness coefficient (i.5435 0. Because the best alternative is the one with the highest relative closeness coefficient.5 show the mapping of TABLE 6. the positive separation distance for collection center F (see table 6. as calculated for those two categories. respectively. For example.1458 0.1176 = 0. F. Figure 6. the preference order for the collection centers is G.17 Relative closeness coefficients for consumers (second model) Collection centers E F G C* 0. For example.0942 0.e.
18 Relative closeness coefficients for local government officials and supply chain company executives (second model) Collection centers E F G Local government officials 1. LLC .000 1.100 Strategic Planning Models for Reverse and ClosedLoop Supply Chains TABLE 6.3 Mapping of collection centers for consumers (second model). © 2009 by Taylor & Francis Group.851 0.091 FIGURE 6.739 Supply chain company executives 0.000 0.902 0.2 Relative closeness mapping classification (second model). FIGURE 6.
and 3. local government officials. Borda scores (group evaluations) calculated for E.4 Mapping of collection centers for local government officials (second model).2 Phase II (Group Decision Making) Table 6.4. and supply chain company executives. It is assumed that each collection center is in one of the three regions. For example. 3. and the three regions are mapped with respect to the relative closeness coefficients. 6.5 Mapping of collection centers for supply chain company executives (second model). F.. 5. LLC .19 shows the marks of the collection centers as given using Borda’s choice rule for the consumers. the higher the efficiency the corresponding collection center has (with respect to the corresponding category of decision makers). local government officials. respectively. The darker the region is. the Borda score FIGURE 6. and G (viz. © 2009 by Taylor & Francis Group. respectively) are also shown. FIGURE 6. and supply chain company executives.Evaluation of Collection Centers 101 the relative closeness coefficients as calculated for the collection centers by the consumers.
Because F has the highest Borda score.6 Borda score mapping classification (second model).19 Marks and Borda scores of collection centers (second model) Collection centers E F G Local government officials 2 2 1 Supply chain company executives 1 2 0 Consumers 0 1 2 Borda scores 3 5 3 for F (i.7 shows the mapping of the Borda score of each collection center.. FIGURE 6. as created using MapLand [4].. local government officials.6 shows the mapping classification.e.e. LLC . and supply chain company executives (i. FIGURE 6. Figure 6.102 Strategic Planning Models for Reverse and ClosedLoop Supply Chains TABLE 6. Figure 6. it is the best of the lot. © 2009 by Taylor & Francis Group. 5) is calculated by summing the marks of F for consumers. 1 + 2 + 2).7 Mapping of collection centers with respect to Borda score (second model). It is obvious that F is the one with the highest efficiency. for the Borda scores.
Borda’s Rule) This model assumes that neither numerical nor linguistic impacts are available for the evaluation criteria. LLC . Defuzzification of a TFN can be performed using equation (4.23 show the defuzzified overall rating of each existing collection center as well as the collection center’s defuzzified rating with respect to each criterion. 8) (1.Evaluation of Collection Centers 103 6. efficiency. using linguistic performance measures of collection centers that already exist in the reverse supply chain. 3) Defuzzified ratings 9 7. respectively.20 shows not only one of the many ways for conversion of linguistic ratings into TFNs but also the defuzzified ratings of the corresponding TFNs. 5) (0.5. 10) (2.5 Third Model (Neural Networks.21–6. the ratings of existing collection centers are used to construct a neural network that. local government officials. In the first phase. Using fuzzy logic. using the defuzzified ratings of the existing collection centers with respect to criteria as input sets and the collection centers’ defuzzified overall TABLE 6. It employs a neural network [11] to evaluate the efficiency of a collection center of interest (which is being considered for inclusion in a reverse supply chain). 5. in the third phase. 0. 10) (5. the impacts obtained in the first phase are used in a fuzzy TOPSIS (combination of fuzzy logic and TOPSIS) method to obtain the overall rating of the collection center of interest. calculates impacts of criteria identified for each category of decision makers given in section 6.1 Phase I (Derivation of Impacts) Suppose that one has the linguistic ratings of ten existing collection centers. as evaluated by the consumers. Tables 6.. these linguistic ratings are converted into triangular fuzzy numbers (TFNs).3. 3.20 Conversion table for ratings (third model) Linguistic ratings Very good (VG) Good (G) Fair (F) Poor (P) Very poor (VP) TFNs (7. Borda’s choice rule is employed to calculate the maximized consensus rating (among the categories considered). TOPSIS. A neural network is constructed and trained for each category of decision makers. 10.3.8). 7. In the second phase. and supply chain company executives.3 5 3 1 © 2009 by Taylor & Francis Group. 6. Finally. i.e. Fuzzy Logic. as given by an expert in each category of decision makers described in section 6. in turn. the model to evaluate the efficiency of a collection center of interest is carried out in three phases. Table 6. To this end. as calculated for each category. of the recovery facility of interest.
with five nodes in the hidden layer.3 3 5 7.21 Ratings for consumers (third model) Collection centers C1 C2 C3 C4 C5 C6 C7 C8 C9 C10 IC 1 9 3 3 5 9 5 1 1 5 DH 3 1 1 9 1 3 7.3 5 3 1 5 9 3 EO 5 7. and the number in the input layer is the number of criteria considered by the corresponding category. After each neural network is trained.3 9 1 5 3 Overall 5 7. there are ten input–output pairs for each neural network because there are ten existing collection centers. For example. nV is the number of input nodes. figure 6.3 3 1 9 5 1 TABLE 6.3 3 5 7. nH is the number of hidden nodes. In the example.3 3 1 9 5 1 ratings as corresponding outputs.8 shows the neural network constructed and trained for the category of consumers. Here. the following equation [2] is used to calculate the impacts of criteria considered by the corresponding category.3 3 7. three layers are considered in each network.22 Ratings for local government officials (third model) Collection centers C1 C2 C3 C4 C5 C6 C7 C8 C9 C10 DH 1 9 3 3 5 9 5 1 1 5 DR 3 1 1 9 1 3 7.3 3 9 9 SA 9 9 1 7.104 Strategic Planning Models for Reverse and ClosedLoop Supply Chains TABLE 6. Also.3 5 5 9 Overall 5 7.3 9 1 5 5 SP 3 5 1 7. LLC .3 9 1 1 5 7.3 5 5 9 DR 5 3 3 1 3 7. and Oj is the connection weight from the jth hidden node to the output node: © 2009 by Taylor & Francis Group. Iij is the connection weight from the ith input node to the jth hidden node. the absolute value of Wv is the impact of the vth input node upon the output node. The number of nodes in the output layer is one (for overall rating).
3 5 5 9 LC 1 3 7. LLC .2) i j i  Iij  © 2009 by Taylor & Francis Group.8 Neural network for consumers (third model).Evaluation of Collection Centers TABLE 6.3 5 5 9 3 1 3 1 UI 3 7. nH nV j I vj  Iij  i Oj Wv  nV nH I vj nV Oj (6.3 105 Overall 5 7.3 3 1 9 5 1 IC H PH H PR H SP H EO H SA O Input layer Hidden layer Output layer FIGURE 6.3 3 5 7.23 Ratings for supply chain company executives (third model) Collection centers C1 C2 C3 C4 C5 C6 C7 C8 C9 C10 PI 1 9 3 3 5 9 5 1 1 5 SC 3 1 1 9 1 3 7.3 3 9 9 9 3 9 7.3 9 1 5 3 1 5 1 7.3 IG 3 7.3 9 1 5 5 1 3 5 3 DH 5 3 1 5 9 3 7.3 1 9 5 DR 1 5 7.
respectively. to evaluate C11.25 Impacts for local government officials (third model) Criteria Impacts DH 0. Now one is ready to perform the six steps in the TOPSIS for each category of decision makers. Table 6.30) is calculated as follows: © 2009 by Taylor & Francis Group.01 DH 0. and C13. Step 1: Construct the normalized decision matrix. and supply chain company executives (with defuzzified ratings for C11. C12. the normalized rating of collection center C12 with respect to criterion DH (see tables 6. The decision matrices formed for the consumers. C12. and C13.2 Phase II (Individual Decision Making) Suppose that there are three collection centers.29.30 shows the normalized decision matrix formed by applying equation (4. the matrices can be constructed by inviting representatives from each category of decision makers to participate in relevant survey questionnaires. LLC .13 Tables 6. C12.20 is used here.19 SA 0.) As in the second model (see section 6.24 Impacts for consumers (third model) Criteria Impacts IC 0. and C13) in this example are shown in tables 6.06 SP 0.67 TABLE 6.3). C11.106 Strategic Planning Models for Reverse and ClosedLoop Supply Chains TABLE 6. local government officials.26 show the impacts of the criteria considered for the consumers.27 (decision matrix for the consumers). of interest. and supply chain company executives.1 IG 0. The following steps show the implementation of the TOPSIS for the consumers. individual interviews.13 DR 0. 6. respectively (Table 6.43 TABLE 6. as well. local government officials.27–6.33 DR 0.24 SC 0.25 DR 0.27 and 6. and onsite observations.18 EO 0. focus groups. to convert linguistic ratings given by each category into TFNs. A fuzzy TOPSIS method uses the impacts obtained in the first phase to calculate the overall ratings of the three collection centers. For example.24–6.09 LC 0 UI 0.5.49) on each element of table 6.18 DH 0.26 Impacts for supply chain company executives (third model) Criteria Impacts PI 0.
27 Decision matrix for consumers (third model) Collection centers C11 C12 C13 IC 3 5 9 DH 9 1 3 DR 1 3 7.962 SA 0.24 and the normalized decision matrix in table 6.839 DH 0.918 SP 0.192 0. the © 2009 by Taylor & Francis Group.33 107 TABLE 6. This is constructed using the impacts of the criteria listed in table 6.679 r22 1 9 2 12 32 0.125 0.33 3 7.943 0.679 0.33 DR 9 9 9 IG 5 3 1 TABLE 6. For example. Table 6.33 5 3 EO 1 1 5 SA 7.466 0.30.192 0. LLC .30 Normalized decision matrix for consumers (third model) Collection centers C11 C12 C13 IC 0.314 DR 0.Evaluation of Collection Centers TABLE 6.29 Decision matrix for supply chain company executives (third model) Collection centers C11 C12 C13 PI 5 9 5 SC 1 3 7.783 0.33 SP 7.278 0.376 0.105 0.33 LC 5 9 3 UI 5 5 1 DH 9 3 7.31 shows the weighted normalized decision matrix for the consumers.278 0.534 0.105.28 Decision matrix for local government officials (third model) Collection centers C11 C12 C13 DH 3 5 9 DR 9 1 3 TABLE 6.320 EO 0. Step 2: Construct the weighted normalized decision matrix.
140 0.014.e. respectively.041) to the sum (i.e.32). They are the ideal and negativeideal solutions.e..036 0.31 Weighted normalized decision matrix for consumers (third model) Collection centers C11 C12 C13 IC 0..116 S– 0.095 0..181 SA 0.041 DR 0.32 Separation distances for consumers (third model) Collection centers C11 C12 C13 S* 0. Step 3: Determine the ideal and negativeideal solutions.31 has a maximum rating and a minimum rating.54). i. the relative closeness coefficient for each collection center is calculated (see table 6. the preference order TABLE 6. Each column in the weighted normalized decision matrix shown in table 6.257 0.004 0.057 EO 0. Because the best alternative is the one with the highest relative closeness coefficient.31). For example (see table 6. the positive separation distance for collection center C12 (see table 6.52) and (4..e. Step 5: Calculate the relative closeness coefficient.041 + 0.122.52). is calculated by multiplying the impact of DH. The separation distances (see table 6.022 0.014 (see table 6.014 0. 0.120 0.011 DH 0. 0.041 0. Using equation (4.221 0.30). For example.e. 0. LLC .32) for each collection center are calculated using equations (4.31) and the ideal solutions (obtained in step 3) for the criteria.129 (see table 6. and the negativeideal solution (minimum rating) is 0.007 0.e. i. the ideal solution (maximum rating) is 0.036 0. with the normalized rating of C12 with respect to DH.33).053 SP 0.108 Strategic Planning Models for Reverse and ClosedLoop Supply Chains TABLE 6.31).. the relative closeness coefficient (i. 0. with respect to criterion DH.122 0. 0. i.298) of its negative and positive separation distances (see table 6..233 © 2009 by Taylor & Francis Group. For example. which contains the weighted normalized ratings of C12 (see table 6.105 (see table 6.006 0.33) is the ratio of C12’s negative separation distance (i. 0. for the corresponding criterion.32) is calculated using equation (4.294 0.24).257 = 0. Step 4: Calculate the separation distances.152 0.53).294 weighted normalized rank of collection center C12 with respect to criterion DH. Step 6: Rank the preference order.137) for collection center C12 (see table 6.
137 0.415 TABLE 6. are shown in table 6. 5. and C13 (viz. The TOPSIS is implemented for the local government officials and the supply chain company executives in a similar manner. 6.Evaluation of Collection Centers TABLE 6. 0 + 0 + 1). C12.35 Marks and Borda scores of collection centers (third model) Collection centers C11 C12 C13 Local government officials 2 0 1 Supply chain company executives 2 1 0 Consumers 1 0 2 Borda scores 5 1 3 © 2009 by Taylor & Francis Group. local government officials.3 Phase III (Group Decision Making) Table 6. it is the best of the lot.34 Relative closeness coefficients for local government officials and supply chain company executives (third model) Collection centers C11 C12 C13 Local government officials 0. 1.. as evaluated by the consumers). and C12 (that means C13 is the best collection center..354 Supply chain company executives 0.35 shows the marks of the collection centers as given using Borda’s choice rule for the consumers.34. respectively) are also shown. TABLE 6. and supply chain company executives.e. Borda scores (group evaluations) calculated for C11. as calculated for those two categories.592 0.754 0.e.668 109 for the collection centers is C13. 1) is calculated by summing the marks of C12 for the consumers. Because C11 has the highest Borda score. LLC .619 0. and supply chain company executives (i. For example.502 0. The relative closeness coefficients of the collection centers.33 Relative closeness coefficients for consumers (third model) Collection centers C11 C12 C13 C* 0. and 3..5. the Borda score for C12 (i. local government officials. C11.096 0.
9).6. respectively. LLC .6 Fourth Model (ANP and Goal Programming) In this model. The first level contains the objective of evaluation of the candidate collection centers. with respect to qualitative criteria taken from the perspective of a remanufacturing facility interested in buying used products from the collection centers.110 Strategic Planning Models for Reverse and ClosedLoop Supply Chains 6.1 Application of ANP The problem of evaluating the efficiencies of the candidate collection centers is framed as a fourlevel hierarchy (see figure 6.6. first the analytic network process (ANP) is used to calculate the performance indices (efficiency scores) of candidate collection centers.6. Sections 6. 6. Then goal programming is employed to determine the quantities of used products to be transported from the candidate collection centers to the remanufacturing facility while satisfying two important goals of the remanufacturing facility: to maximize total value of purchase and minimize total cost of purchase. © 2009 by Taylor & Francis Group. The third level contains the subcriteria under each Delivery Reliability Reliability Conformance to Speciﬁcations Order Fulﬁllment Lead Time Collection Center 1 Capabilities/ Responsiveness Flexibility (Ability to adapt to demand ﬂuctuations) Suppliers Design Capability Return/Warranty processing costs Collection Center 2 Evaluating Suppliers Financial Issues Pricing Structure Quantity Discounts Financial Stability & Economic Performance Level of coop and Information Exchange Cultural and Strategic Issues Collection Center 3 Technical Capability Reputation in Industry Proximity Others Safety and Environmental Issues FIGURE 6.2 present the applications of ANP and goal programming.1 and 6.9 Hierarchical structure for ANP (fourth model). The second level consists of the main evaluation criteria taken from the perspective of a remanufacturing facility.
(2) flexibility in adapting to demand fluctuations. S1. and in the promised condition. Table 6. S2.36 shows the pairwise comparison matrix for the main criteria (second level in the hierarchy) and also the normalized eigen vector of the matrix. The main and subcriteria considered are the following (see [5–9]): Reliability: This criterion relates to a collection center delivering the used products at the right time. and S3.37–6. any safety issues with the collection center directly reflect on the remanufacturing facility’s reputation. It is assumed that there exist interdependencies among the subcriteria on the third level in the hierarchy. and (4) financial stability and economic performance of the collection center. LLC . The subcriteria considered here are (1) order fulfillment lead time. at the right remanufacturing facility. The fourth level contains the candidate collection centers. Capability/responsiveness: This criterion reflects the velocity at which a collection center supplies the used products to the remanufacturing facility and the collection center’s ability to adapt to sudden demand fluctuations. Tables 6.41 show the pairwise comparison matrices of subcriteria with respect to their main criteria and also the corresponding normalized eigen vectors of the matrices. The subcriteria are (1) return/warranty processing costs. Cultural and strategic issues: This criterion consists of the following subcriteria: (1) level of cooperation and information exchange between the collection center and the remanufacturing facility. in the right quantity. and (3) the collection center’s technical capability (how knowledgeable the collection center is about the product). Three collection centers. Others: This criterion considers miscellaneous aspects that are not considered in the other criteria. (2) the collection center’s reputation in the industry. are considered in the numerical example to illustrate the application of ANP. These aspects are (1) proximity of the collection center to the remanufacturing facility (it affects the transportation cost and the transit time) and (2) safety and environmental aspects (because the collection center and the remanufacturing facility are closely involved. The elements of the normalized eigen vector are the impacts given to the main criteria with respect to the objective (first level in the hierarchy).41 has © 2009 by Taylor & Francis Group. Financial issues: This criterion reflects the costs and other financial aspects involved. (2) pricing structure.36–6. environmental aspects are concerned about the collection center’s effort in pursuing environmental consciousness or a “green” image). Note that each of the matrices in tables 6. and (3) design capabilities. (3) quantity discounts. The subcriteria considered under this main criterion are (1) delivery reliability and (2) conformance to standards.Evaluation of Collection Centers 111 main criterion.
01037 4 (4 1) 0. is ( max n) (n 1)( R ) 4.431 0.25 0. k = 59.065 0. CR.623 0.112 Strategic Planning Models for Reverse and ClosedLoop Supply Chains TABLE 6. where k is an arbitrarily large number. Table 6. for the matrix in table 6. © 2009 by Taylor & Francis Group.285 Criteria Reliability Responsiveness Financial issues Cultural and strategic issues Others Reliability 1 4 1/3 1 6 Others 1/6 3 1/3 1/3 1 TABLE 6. LLC .43 shows the converged super matrix.1).Financial ness issues 1/4 3 1 1 1/5 1 1/3 2 1/3 3 Cultural and strategic issues 1 5 1/2 1 3 Normalized eigen vector 0. .75 TABLE 6.90 0.42 shows the matrix of interdependencies (called the super matrix M) among the subcriteria with respect to their main criteria.38 Comparative importance values of subcriteria under responsiveness (fourth model) Order fulfillment lead time (OFT) 1 1/3 1/4 Design capability (DC) 4 2 1 Normalized eigen vector 0.114 0.39. using equation (4. which is done by raising M to the power of 2k+1.37 Comparative importance values of subcriteria under reliability (fourth model) Subcriteria Delivery reliability (DR) Conformance to specs (CS) Delivery reliability 1 3 Conformance to specs 1/3 1 Normalized eigen vector 0.36 Comparative importance values of main criteria (fourth model) Responsive.0038.137 Subcriteria Order fulfillment lead time (OFT) Flexibility (F) Design capability (DC) Flexibility (F) 3 1 1/2 a consistency ratio CR less than or equal to 0. This super matrix M is made to converge to obtain a longterm stable set of impacts. For convergence. M must be made column stochastic.239 0. In the example. Table 6.1. For example.102 0.
722 1/5 1/6 1 1/2 2 1 0.25 0.41 Comparative importance values of subcriteria under others (fourth model) Subcriteria Proximity (P) Safety and environment (SE) Proximity (P) 1 3 Safety and environment (SE) 1/3 1 Normalized eigen vector 0.40 Comparative importance values of subcriteria under cultural and strategic issues (fourth model) Level of coop and information exchange (Coop) 1 Technical capability (TC) 5 Subcriteria Level of coop and information exchange (Coop) Technical capability (TC) Reputation (R) Reputation 6 Normalized eigen vector 0.103 TABLE 6. LLC .37 TABLE 6.Evaluation of Collection Centers TABLE 6.75 © 2009 by Taylor & Francis Group.185 4 1 2 1 0.174 0.099 3 1 2 1 0.345 2 1/2 1 1/2 0.39 Comparative importance values of subcriteria under “financial issues” (fourth model) Returns/ warranty processing costs (RC) 1 Quantity discounts (QD) 1/2 Economic performance and financial stability (EP) 1/4 113 Subcriteria Returns/ warranty processing costs (RC) Pricing structure (PS) Quantity discounts (QD) Economic performance and financial stability (EP) Pricing structure (PS) 1/3 Normalized eigen vector 0.
LLC .18 0.75 0.6 0.38 0.53 0.8 0 0.83 TABLE 6.53 0.29 0.19 0.25 0 0 0 0 0 0 0 0 0 F 0 0 0.43 Converged super matrix (fourth model) Subcriteria Delivery reliability Conformance to specs Order fulfillment LT Flexibility Design capability Returns/warranty Pricing Quantity discounts Stability and economic performance Coop and information exchange Technical capability Reputation Proximity Safety and environment Stabilized relative impact 1 1 0.23 0.2 0 0 TC 0 0 0 0 0 0 0 0 0 0 0.18 1 1 © 2009 by Taylor & Francis Group.44 0.201 0.16 0 0 0 R 0 0 0 0 P 0 0 0 0 0 0 0 0 0 0 0 0 0 1 SE 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0.75 0.114 Strategic Planning Models for Reverse and ClosedLoop Supply Chains TABLE 6.38 0.25 0 0 0 0 0 0 0 0 0 0 RC 0 0 0 0 0 0 0.25 0 0 0 0 0 0 0.44 0.15 0.2 0 0 0 0 0 0 0 0 0 DC 0 0 0.163 0 0 0 0 0 PS 0 0 0 0 0 0 0.38 0.75 0.62 0 0 0 0 0 0 EP 0 0 0 0 0 0.11 0 0 0 0 0 0 Coop 0 0 0 0 0 0 0 0 0 0 0.8 0.29 0.42 Matrix of interdependencies (super matrix M) (fourth model) DR DR CS OFT F DC RC PS QD EP Coop TC R P SE 0 1 0 0 0 0 0 0 0 0 0 0 0 0 CS 1 0 0 0 0 0 0 0 0 0 0 0 0 0 OFT 0 0 0 0.16 0 0 0 0 0 QD 0 0 0 0 0 0.27 0.13 0.
individual interviews.376 0. The overall performance index for each of the three collection centers is calculated by multiplying the desirability index (see table 6.0783 0.137 S2 0.23 0.054 0.075 0. interdependencies.09 0.25 0.615 0.33 0.176 S2 0.1271 0.2488 0. S2.657 0.623 0.292 0.581 0. LLC .148551 0.33 0.44 Relative ratings of collection centers with respect to subcriteria (fourth model) Subcriteria/alternate collection centers DR CS OFT F DC RC PS QD EP Coop TC R P SE S1 0.543 0.109 0.0528 0.45 Desirability indices (fourth model) Criteria/collection centers Reliability Responsiveness Financial issues Cultural and strategic issues Others S1 0.3429 0. with respect to the subcriteria.2138 0.45 shows the desirability index calculated for each candidate collection center using equation (4.092 0.652 0.216 0.0874 0. and onsite observations. and S3. S1.Evaluation of Collection Centers 115 Table 6.239 S3 0. decision makers could be invited to participate in relevant survey questionnaires.45) of each collection center for each main criterion by the impact of that criterion (see TABLE 6. Table 6.102 0.6211 © 2009 by Taylor & Francis Group.274 0.309 0.2299 0.12 0. These ratings are obtained after carrying out pairwise comparisons between the candidate collection centers with respect to the subcriteria and then obtaining the normalized eigen vector. focus groups.44 shows the relative ratings of the candidate collection centers.524 0.581 0.2).37–6.2027 S3 0.141 0.0438 0. To obtain pairwise comparisons.681 0.4432 0.67 0.109 0. and relative ratings (see tables 6.623 TABLE 6.309 0.44).471 0.623 0.151 0.59 0.137 0.137 0.239 0.239 0.345 0.068 0.11 0.
2 Application of Goal Programming This section presents the application of goal programming to determine the quantities of used products to be transported from the candidate collection centers to a remanufacturing facility of interest while satisfying two important goals of the remanufacturing facility: to maximize total value of purchase and minimize total cost of purchase. Section 6. i = 1.36) and summing over all the criteria. s Capacity of collection center i Probability of breakage of used products purchased from collection center i pmax Maximum allowable probability of breakage Qi Decision variable representing the quantity to be purchased from collection center i s wi Number of candidate collection centers Performance index of collection center i obtained by carrying out ANP Problem Formulation 6.6.6.116 Strategic Planning Models for Reverse and ClosedLoop Supply Chains TABLE 6.2.6. 6.2322 0. © 2009 by Taylor & Francis Group.1 gives the nomenclature used in the methodology. LLC .6.46 shows the overall performance indices (efficiencies) for the three collection centers.2.2318 0. 2. Minimize the total cost of purchase (TCP). Maximize the total value of purchase (TVP).6.5359 table 6.2. 6. …. 2. Table 6.2.1 ci dj g i ki pi Nomenclature for Problem Formulation Unit purchasing cost of used product at collection center i Demand for used product j Goal index Collection center index.2 presents the problem formulation and a numerical example.2 The following two goals of the remanufacturing facility are considered: 1. and section 6.46 Overall performance indices (fourth model) Collection center S1 S2 S3 Performance index 0.
It is at the discretion of the decision maker to add any other goals that are relevant to the situation. from the results obtained from the application of goal programming. LLC .7) Nonnegativity constraint: Qi 0 (6. Table 6. and table 6.6. This may be attributed to the fact that the unit purchasing cost at S3 is higher than that at S2 (this was not considered in the application of ANP.1). s Goal 1: Maximize TVP: i 1 wi Qi (6. S3 is the highestranked collection center.4) Capacity constraint: Qi ki (6.8) The three candidate collection centers from section 6.47 shows the data used for the goal programming problem (note that only one used product type is considered). When no other system constraints are in place. it can be noticed that 650 units are ordered from S2 and the remaining 350 units are ordered from S3. From the application of ANP (see section 6. However.54 (aspiration level = 250). The total cost of purchase is found to be $935. where collection centers were evaluated with respect to qualitative criteria). 750 units might be ordered from S3 before considering other collection centers. and the total value of purchase is 338.1 are considered in this numerical example as well.Evaluation of Collection Centers 117 Whereas the first goal involves minimizing the underachievement of the target.5) Demand constraint: i Qi dj (6. the second goal involves minimizing the overachievement of the target.48 shows the results obtained by solving the problem using LINGO (v4). © 2009 by Taylor & Francis Group.3) Goal 2: Maximize TCP: ci Qi (6.6.6) s Quality constraint: d j pmax i 1 Qi pi (6.
7. Taguchi Loss Function. first the eigen vector method and Taguchi loss function are used to calculate the weighted Taguchi losses (inefficiency scores) of candidate collection centers. and Goal Programming) In this model. Then section 6.2318 0.025 TABLE 6.000 Maximum acceptable breakage probability = 0.2322 0.03 S2 650 0.118 Strategic Planning Models for Reverse and ClosedLoop Supply Chains TABLE 6. the better) © 2009 by Taylor & Francis Group.015 S3 750 1. Section 6.54 Total cost of purchase (TCP) = 935 ANP rating 0. Then goal programming is employed to determine the quantities of used products to be transported from the candidate collection centers to the remanufacturing facility while satisfying two important goals of the remanufacturing facility: to minimize total loss of profit and minimize total cost of purchase.0 0.5359 Quantity ordered 0 650 350 6.2 presents the application of goal programming.9 0. LLC .2 0.7.7 Fifth Model (Eigen Vector Method.1 presents the application of the eigen vector method and Taguchi loss function. with respect to qualitative criteria taken from the perspective of a remanufacturing facility interested in buying used products from the collection centers.1 Application of Eigen Vector Method and Taguchi Loss Function The following four qualitative criteria (taken from the perspective of a remanufacturing facility) are considered to evaluate the candidate collection centers: Quality of used products (the smaller the defect rate of the used products supplied.7.48 Results (fourth model) Collection center S1 S2 S3 Total value of purchase (TVP) = 338. 6.01 Net demand for the product = 1.47 Data for goal programming model (fourth model) Collection center Capacity Unit purchasing cost Breakage probability S1 300 1.
experts rate these performance factors. To this end.33 0. TABLE 6. Table 6. the better) Cultural and strategic issues.052674 0. The collection center’s service factor percentage is obtained by dividing the total service rating by the total number of points possible. these ratings on all factors are summed and averaged to obtain a total service rating. the better) Proximity (the closer the collection center from the remanufacturing facility.90 0. The ratings are given on a scale of 1–10.2 0. LLC . The elements of the normalized eigen vector represent the impacts given to the criteria.1 for explanations of these issues) For the numerical example.49 has a consistency ratio CR of less than 0.425064 Criteria Quality Ontime delivery Proximity Cultural and strategic issues Quality 1 0. green image.6.13 4 (4 1) 0. S4. are considered. Table 6.14 1 Proximity 7 4 1 6 * Only the main criteria are considered in this model. Monczka and Trecha’s [10] service factor rating (SFR) is used.49 Comparative importance values of criteria (fifth model) Ontime delivery 3 1 0. level of cooperation and information exchange.1.137363 0. For a given collection center. In practice.048.49 shows the pairwise comparison matrix for the criteria* and also the normalized eigen vector of the matrix. It could be difficult to quantify the cultural and strategic issues criterion for the calculation of Taguchi losses. and financial stability/economic performance (see section 6. and S6. such as flexibility in adapting to demand fluctuations. © 2009 by Taylor & Francis Group. Specifically. it is ( max n) (n 1)( R ) 4.384899 0. SFR includes performance factors difficult to quantify but decisive in the selection process. three candidate collection centers.25 5 Cultural and strategic issues 1 0. S5.166 1 Normalized eigen vector 0. Table 6.50 shows the service factor ratings for the various aspects of the cultural and strategic issues criterion.Evaluation of Collection Centers 119 Ontime delivery (the smaller the number of delayed deliveries. the level of performance being directly proportional to the rating.
S4. For ontime delivery. 5 days delay 40% 50% Table 6. and the upper specification limit for the defect rate/breakage probability is 30%.75 Average ÷ 10 57. are 1111. meaning that the remanufacturing facility will incur 100% loss when the collection center’s distance reaches the specification limit. loss will be zero at the closest collection center and the specification limit is up to 40% of the closest collection center.51 Decision variables for selecting collection centers (fifth model) Criteria Quality Ontime delivery Proximity Cultural and strategic issues Target value 0% 0 Closest 100% Range 0–30% 10–0–5 0–40% 100–50% Specification limit 30% 10 days earlier. The values of the loss coefficient. at which there is 100% loss to the manufacturer. and cultural and strategic issues.66–4. The specification limit of delivery delay is 5 days. and 25 for quality. and S6. proximity.52 shows the characteristic value and the relative value of each criterion for the three collection centers. the specification limit is 50% for the service factor percentage. at which there is no loss to the remanufacturing facility. k1 and k2). For S4. For proximity. 625. meaning that the remanufacturing facility will incur 100% loss if the deliveries are delayed by 5 days and are delivered 10 days before the scheduled delivery date. For ontime delivery. respectively. Table 6. the remanufacturing facility will incur losses if the products are delivered late and before the scheduled requirement. LLC . the quality © 2009 by Taylor & Francis Group.5% 67. at which the loss will be 100%.75 6. there exist two loss coefficients. and early delivery is 10 days. Consider the quality criterion. k1 = 4 and k2 = 1 (because an unequal twosided specification limit is considered for ontime delivery.5% 62. k (see equations 4. For cultural and strategic issues.51 shows the decision variables for calculating the Taguchi losses for the three collection centers.120 Strategic Planning Models for Reverse and ClosedLoop Supply Chains TABLE 6. The target defect rate/breakage probability is zero.25 6.11. whereas there will be no loss incurred at a service factor percentage of 100%.5% TABLE 6.50 Service factor ratings for cultural and strategic issues (fifth model) Level of coop and information exchange 6 7 5 Financial stability and economic performance 4 5 8 Collection center S4 S5 S6 Flexibility 7 5 6 Green image 6 8 8 Average 5.68). S5.
5% Relative value 57.1 gives the nomenclature TABLE 6.32 © 2009 by Taylor & Francis Group.5% 62.7.5% characteristic value is 15% defect rate.86 44.54 Weighted Taguchi losses (fifth model) Collection center S4 S5 S6 Weighted Taguchi loss 50.11 Ontime delivery 36 4 64 Proximity 69.43 0 156.32 0.54 shows the weighted Taguchi loss and the normalized Taguchi loss for each collection center.Evaluation of Collection Centers TABLE 6.53 Taguchi losses (fifth model) Collection center S4 S5 S6 Quality 24.25 Cultural and strategic issues 75.5% 67.2 Application of Goal Programming This section presents the application of goal programming to determine the quantities of used products to be transported from the candidate collection centers to a remanufacturing facility of interest while satisfying two important goals of the remanufacturing facility: to minimize total loss of profit and minimize total cost of purchase.86 TABLE 6.66)–(4. 6.53).7. The relative values.99 44.53) and the impacts of the evaluation criteria (see table 6.49). which translates to 15% deviation from the target value.5% 62. are used to calculate (see equations (4.62 Normalized Taguchi loss 0. LLC .52 Characteristic and relative values of criteria (fifth model) Quality Collection center S4 S5 S6 Value 15% 20% 10% Relative value 15% 20% 10% Ontime delivery Value +3 +1 –8 Relative value +3 +1 –8 Proximity Value 8 6 9 Relative value 33.61 64 54. k.2.5% 67. Table 6. The weighted Taguchi loss (sum product of criteria impacts and Taguchi losses) is then calculated from the Taguchi losses of the collection centers (see table 6.36 0.33% 0 50% 121 Cultural and strategic issues Value 57. Section 6.44 11.68)) the Taguchi loss for each collection center for each criterion (see table 6. together with the value of the loss coefficient.37 44.
9) s Goal 2: Minimize TCP: j 1 cj Qj TCP (6.2. Minimize the total loss of profit (TLP).2. LLC . and section 6.1 Nomenclature Used in the Methodology Bj Budget allocated for collection center j Unit purchasing cost of used product at collection center j cj dk Demand for used product k g Goal index j Collection center index. 2.7.2 Capacity of collection center j Probability of breakage of used products purchased from collection center j Maximum allowable probability of breakage Decision variable representing the purchasing quantity from collection center j Number of candidate collection centers Impact of criterion i calculated by the eigen vector method Taguchi loss of collection center j for criterion i Problem Formulation The following two goals of the remanufacturing facility are considered: 1.2 presents the problem formulation and a numerical example.10) © 2009 by Taylor & Francis Group. j = 1. It is at the discretion of the decision maker to add any other goals that are relevant to the situation. 2. s Goal 1: Minimize TLP: j 1 Loss j Q j TLP (6. 6. ….2. Minimize the total cost of purchase (TCP).7.7.122 Strategic Planning Models for Reverse and ClosedLoop Supply Chains used in the methodology. s Lossj Total loss of collection center j for all the evaluation criteria rj pj pmax Qj s wi Xij 6.
S6 is the highestranked collection center.12) s Quality constraint: dk pmax j 1 Qj pj (6.0 0. From the application of eigen vector method and Taguchi loss function (see section 6.32078 0. This may be attributed to TABLE 6.360148 0.7.7 Normalized Taguchi loss 0.Evaluation of Collection Centers 123 Capacity constraint: Qi ri (6.1 are considered in this numerical example as well.14) The three candidate collection centers from section 6.2 0. When no other system constraints are in place.000 Maximum acceptable breakage probability = 0.56 Results (fifth model) Collection center S4 S5 S6 Total loss of purchase (TLP) = 320 Total cost of purchase (TCP) = 945. it can be noticed that 543 units are ordered from S5 and the remaining 457 units are ordered from S6.1).55 shows the data used for the goal programming problem (note that only one used product type is considered).01 Net demand for the product = 1. However. 750 units might be ordered from S6 before considering other collection centers.319072 Quantity ordered 0 543 457 © 2009 by Taylor & Francis Group.015 S6 750 1. LLC .56 shows the results obtained by solving the problem using LINGO (v4).025 TABLE 6. from the results obtained from the application of goal programming.11) Demand constraint: j Qj dj (6.55 Data for goal programming model (fifth model) Collection center Capacity Unit purchasing cost Breakage probability S4 300 1.9 0.7.03 S5 650 0. and table 6.13) Nonnegativity constraint: Q j 0 (6. Table 6.
C. S. European Journal of Operational Research 50:2–18. chap. The first model employs eigen vector method and Taguchi loss function. Neural networks: A comprehensive foundation.. L... C. Y. Journal of Operations Management 14:333–43. S.. 1998. P. 2004. Charting your company’s future. 2nd ed. L. and Mauborgne. W. The third model employs neural networks. Operations management: Processes and value chains. The second model uses eigen vector method.softwareillustrated. and goal programming. 2002. Journal of Purchasing and Material Management 2–7. A. 1991. S. C. References 1. T. 7th ed. 6. K. 5. Costbased supplier performance evaluation. C. R. The fifth model applies eigen vector method. Haykin. K. Krajewski. Satisfaction assessment of multiobjective schedules using neural fuzzy methodology. 4.. and Kim. J. R. S. TOPSIS.124 Strategic Planning Models for Reverse and ClosedLoop Supply Chains the fact (besides the system constraints considered in the goal programming problem) that the unit purchasing cost at S6 is higher than that at S5. and Hartley. Kim. 10.. 77–82.. International Journal of Production Research 41:1831–49. Taguchi loss function. 10. Current. Muralidharan. 1996. The fourth model uses analytic network process (ANP) and goal programming. E.. 1998. Harvard Business Review. 2001. fuzzy logic. and Deshmukh. 6. 8. and Tan. and Jung. 2. 2002. and Borda’s choice rule. 9. C. L. pp. M. 11.. and Ritzman. and Benton.. NJ: Pearson Prentice Hall. W.8 Conclusions In this chapter. Englewood Cliffs. Cha. June. J. An exploration of supplier selection practices across the supply chain. S... IEEE Transactions on Engineering Management 48:307–18. and Trecha. R. M. Supplier selection and assessment: Their impact on business performance. 2002. NJ: Prentice Hall. N. 2003. Monczka. five models are presented for identifying efficient collection centers in a region where a reverse supply chain is to be designed. Ha. © 2009 by Taylor & Francis Group. Journal of Supply Chain Management 38:11–21. technique for order preference by similarity to ideal solution (TOPSIS). Anantharaman. J. G. 3. Lee. Y.html.com/mapland. A multicriteria group decision making model for supplier rating. Kannan. and Borda’s choice rule. 7. R. J. LLC . Englewood Cliffs. K. Choi. Journal of Supply Chain Management 38:22–33. Weber. V. http://www. Supplier selection and management system considering relationships in SCM. Vendor selection criteria and methods.
A very simple evaluation technique could be desired (where only the “most important” evaluation criteria are considered). and impacts for evaluation criteria are given. respectively. 125 © 2009 by Taylor & Francis Group. Whereas the third model uses the eigen vector method. The main difference between these two models lies in the way the impacts are assigned to the criteria for evaluation of recovery facilities. Decision makers could have conflicting criteria for evaluation. and Borda’s * In this chapter. and impacts for evaluation criteria are not given (hence must be derived). and Borda’s choice rule. such as disassembly and recycling/ remanufacturing.6.1 The Issue In addition to selecting efficient collection centers (see chapter 6). The various scenarios for evaluating recovery facilities for efficiency could differ as follows: 1. In this chapter. The third and fourth models are for scenarios 3 and 4. LLC . we use the term impacts instead of the conventional term weights for importance values of evaluation criteria. technique for order preference by similarity to ideal solution (TOPSIS). various models to address the above scenarios are presented. 4. strategic planning of a reverse supply chain involves selecting efficient recovery facilities where reprocessing operations. Evaluation criteria could be presented in terms of ranges of different degrees of desirability.7 Evaluation of Recovery Facilities 7. Evaluation criteria could be given numerical impacts* (importance values). 3. The first model addresses scenario 1 and employs the analytic hierarchy process. are carried out. This is to avoid confusion in the presence of the connection weights between different nodes of the neural network used in section 7. 5. The second model is for scenario 2 and employs linear physical programming. fuzzy logic. the fourth model uses neural networks. TOPSIS. 2. Decision makers could have conflicting criteria for evaluation.
one of the criteria used in the hierarchy (see figure 7. and section 7.e.1) is the fixed cost of the facility (CO). The last level in the hierarchy contains the candidate recovery facilities. section 7. Section 7. 7. LLC . there are three special factors in a reverse supply chain. This criterion can compare the candidate facilities on the third level.2 presents the first model. The level in the middle contains criteria that must somehow be useful in comparing the candidate recovery facilities. Section 7.126 Strategic Planning Models for Reverse and ClosedLoop Supply Chains choice rule. The fifth model is for scenario 5 and uses a simple twodimensional chart to identify efficient recovery facilities.2.1 presents the threelevel hierarchy for the AHP model.1 Threelevel hierarchy for AHP (first model).2.2 gives a numerical example for the model. to identify efficient facilities from a set of candidate recovery facilities..1 ThreeLevel Hierarchy The first level in the hierarchy contains the primary objective. © 2009 by Taylor & Francis Group. Though the criteria to be considered in a reverse supply chain seem similar to those considered in a forward supply chain (for example.3 shows the second model.8 gives some conclusions. Section 7. For example.2 First Model (Analytic Hierarchy Process) In this section.7 presents the fifth model. i. Objective Identiﬁcation of potential recovery facilities Criteria CO QOQI TP/SU TP*DT CS Alternatives A B C D FIGURE 7. that need to be incorporated in AHP in such a way that the hierarchy levels are not disturbed. Section 7. 7. Finally. Section 7. analytic hierarchy process (AHP) is employed to identify efficient facilities from a set of candidate recovery facilities operating in a region where a reverse supply chain is to be designed. Section 7. This chapter is organized as follows: Section 7.5 presents the third model.4 presents the different decision makers (and their criteria) that are considered in the third and fourth models. [3] and [4]).2.6 gives the fourth model.
incoming products of the same type might have different reprocessing times (unlike in a forward supply chain. etc). average supply of used products. The effect of a low TP is compensated for by dividing TP with a possibly low SU (by doing so. In supplydriven cases like these. and average disassembly time of used products. it is unfair not to consider DT in the hierarchy. Similarly. it is unfair to judge a recovery facility without considering SU in the hierarchy. Thus. Though throughput (TP) is a criterion that can compare two or more candidate recovery facilities. a high DT might lead to a low TP. Hence. a low SU might lead to a low TP. However. Average supply of used products: The only driver to design a forward supply chain is the demand for new products. even if there is low supply of used products (SU). In other words. components of incoming goods (used products) of even the same type in a recovery facility are likely to be of varied quality (worn out.Evaluation of Recovery Facilities 127 Those special factors are average quality of used products. the facility under consideration is not underestimated). and a high SU might lead to a high TP. The effect of a low TP is compensated for by multiplying TP with a possibly high DT © 2009 by Taylor & Francis Group. if there is low demand for new products. the idea is to use the multiplication of TP and DT as a criterion in the hierarchy. LLC . such as environmental regulations and asset recovery. the idea is to take the difference between QO and QI as a criterion in the hierarchy. a reverse supply chain must be administered due to the drivers. Unlike in a forward supply chain. so. Thus. Average quality of used products: Unlike in a forward supply chain. the facility under consideration is not overestimated). the idea is to take the ratio of TP to SU as a criterion in the hierarchy. and a low DT might lead to a high TP. it is not justified to use QO as an independent criterion for comparison because QO depends on average quality of incoming products (QI). where manufacturing time and assembly time are predetermined and equal for products of the same type). even for the same type of products. Average disassembly time of used products: The average disassembly time (DT) is not exactly the inverse of TP because TP takes into account the whole reprocessing (disassembly plus recycling/remanufacturing) time. low performing. Because TP of a recovery facility depends upon DT. this is not the case in some reverse supply chains where. the effect of a high TP is dampened by dividing TP with a possibly high SU (by doing so. it is not justified to use TP as an independent criterion because TP depends on SU. components of incoming goods (used products) in a recovery facility are likely to be deformed or broken or different in number. Though the average quality of reprocessed (recycled/remanufactured) goods (QO) is a criterion that can compare two or more candidate facilities. there is practically no forward supply chain. Thus. In other words.
2–7.06.1). be it the government.4.104 0. A. and D. 7. and what kind of incentives it is giving the customers buying the reprocessed goods.1 Comparative matrix for criteria (first model) Criteria CO QO–QI (TP)/(SU) (TP)×(DT) CS CO 1 5 1/3 1 5 QO–QI 1/5 1 1/7 1/3 1/5 TP/SU 3 7 1 2 3 TP×DT 1 3 1/2 1 3 CS 1/5 5 1/3 1/3 1 Normalized eigen vector 0.2. For example. Table 7. This vector represents the impacts given by the decision maker to the criteria. the effect of a high TP is dampened by multiplying TP with a possibly low DT (by doing so.6 has a consistency ratio CR of less than or equal to 0. with respect to the evaluation criteria CO. An intangible criterion considered in this model is termed customer service (CS). and CS. C. LLC . is ( max n) (n 1)( R ) 4.055 0.240 © 2009 by Taylor & Francis Group. the collection center.6.6 show comparative ratings of the candidate recovery facilities.2–7. TP×DT.128 Strategic Planning Models for Reverse and ClosedLoop Supply Chains (by doing so. CR. QO–QI.2 Numerical Example Consider the following example: Table 7.15 4 (4 1) 0. TP/SU. They also show the normalized eigen vectors representing the relative ratings with respect to the criteria. what kind of incentives it gives the collection centers that supply the used products. Tables 7. It also shows the normalized eigen vector of the matrix. The allencompassing term customer service is used here because any beneficiary is a customer.1–7. using equation (4. This matrix is the collection of the eigen vectors obtained in tables 7. the facility under consideration is not overestimated). TABLE 7.7 shows the aggregate matrix of relative ratings of recovery facilities with respect to each criterion in the second level of the hierarchy.90 0. Note that each of the matrices in tables 7. respectively.1 shows the pairwise comparison matrix for evaluation of candidate recovery facilities. or the actual customer buying the reprocessed goods. CS is a measure of how well a recovery facility utilizes the incentives provided by the government. the facility under consideration is not underestimated). for the matrix in table 7. Similarly.491 0.110 0.1. to what extent it meets the environmental regulations. B.
Evaluation of Recovery Facilities
TABLE 7.2 Comparative ratings of recovery facilities with respect to CO (first model)
CO Facilities A B C D A 1 1/3 1/6 1/2 B 3 1 1/7 1/3 C 6 7 1 4 D 2 3 1/4 1
129
Normalized eigen vector 0.460 0.310 0.050 0.180
TABLE 7.3 Comparative ratings of recovery facilities with respect to QO–QI (first model)
QO–QI Facilities A B C D A 1 1 1/7 1/4 B 1 1 1/7 1/7 C 7 7 1 5 D 4 7 1/5 1 Normalized eigen vector 0.380 0.445 0.050 0.125
TABLE 7.4 Comparative ratings of recovery facilities with respect to TP/SU (first model)
TP/SU Facilities A B C D A 1 7 3 2 B 1/7 1 1/2 1/7 C 1/3 2 1 1 D 1/2 7 1 1 Normalized eigen vector 0.072 0.574 0.212 0.142
TABLE 7.5 Comparative ratings of recovery facilities with respect to TP×DT (first model)
TP×DT Facilities A B C D A 1 5 2 2 B 1/5 1 1/3 1/7 C 1/2 3 1 1 D 1/2 7 1 1 Normalized eigen vector 0.091 0.595 0.171 0.143
© 2009 by Taylor & Francis Group, LLC
130
Strategic Planning Models for Reverse and ClosedLoop Supply Chains
TABLE 7.6 Comparative ratings of recovery facilities with respect to CS (first model)
CS Facilities A B C D A 1 6 3 7 B 1/6 1 1/5 3 C 1/3 5 1 6 D 1/7 1/3 1/6 1 Normalized eigen vector 0.053 0.298 0.101 0.548
TABLE 7.7 Aggregate of ratings of recovery facilities (first model)
A CO QO–QI TP/SU TP×DT CS 0.460 0.380 0.072 0.091 0.053 B 0.310 0.445 0.574 0.595 0.298 C 0.050 0.050 0.212 0.171 0.101 D 0.180 0.125 0.142 0.143 0.548
By multiplying the matrix in table 7.7 with the normalized eigen vector obtained in table 7.1, the following normalized ranks are obtained for the recovery facilities: RankA = 0.26, RankB = 0.42, RankC = 0.09, and RankD = 0.23. If the decision maker desires to choose only those recovery facilities that have ranks of at least 0.25, he or she would choose recovery facilities A and B.
7.3
Second Model (Linear Physical Programming)
In this section, the model to identify efficient recovery facilities using linear physical programming (LPP) is presented. Section 7.3.1 presents the nomenclature for the LPP model, section 7.3.2 presents the evaluation criteria that are considered in this model, and section 7.3.3 presents a numerical example. 7.3.1 CSv Nomenclature for LPP Model Customer service rating of recovery facility v (numerical scale)
DTv Average disassembly time of products supplied to recovery facility v (time units)
© 2009 by Taylor & Francis Group, LLC
Evaluation of Recovery Facilities gi ITuv K Lv Mv ith criterion for evaluation of candidate recovery facilities
131
Transit time between collection center u and recovery facility v (time units) Transportation cost per unit time ($ per unit time) Labor cost at location of recovery facility v ($ per unit time) Inventory (space) cost at recovery facility v ($ per unit area)
OTvw Transit time between recovery facility v and demand center w (time units) QIv Average quality of products supplied to recovery facility v (numerical scale)
QOv Average quality of outgoing products from recovery facility v (numerical scale) Rv TPv u v w 7.3.2 7.3.2.1 Reprocessing cost at recovery facility v ($ per unit product) Throughput of recovery facility v (products) Collection center Recovery facility Demand center Criteria for Identification of Efficient Recovery Facilities Class 1S Criteria (Smaller is Better) SUv Supply to recovery facility v (products)
The cost of transporting goods (used as well as reprocessed), g1, through a recovery facility v across the reverse supply chain is calculated using the following equation: g1
u
( ITuv )( K )
w
(OTvw )( K )
(7.1)
The operating cost, g2, incurred by a recovery facility v is the sum of the labor cost, the inventory cost, and the reprocessing cost. Thus, g2 Lv Mv Rv (7.2)
7.3.2.2
Class 2S Criteria (Larger Is Better)
As explained in section 7.2, the difference between QO and QI is taken as a criterion for evaluation. Thus, g3 QOv QI v (7.3)
© 2009 by Taylor & Francis Group, LLC
132
Strategic Planning Models for Reverse and ClosedLoop Supply Chains
Also, TP is divided by SU and then taken as a criterion for evaluation: g4 Similarly, g5 TPv DTv (7.5) TPv / SU v (7.4)
Customer service (CS) is taken as a criterion for evaluation in the physical programming model as well. Thus, g6 7.3.3 Numerical Example CSv (7.6)
Three candidate recovery facilities (E, F, and G) are evaluated using the LPP model and then ranked to identify the efficient ones. Table 7.8 shows the target values for each criterion detailed in section 7.3.2. Table 7.9 shows the criteria values for each recovery facility. Table 7.10 shows the incremental weights obtained by using the LPP weight algorithm [2]. Tables 7.11–7.13 show the deviations of criteria values from the target values for facilities E, F, and G, respectively. Table 7.14 shows the total scores (obtained using equation (4.34)) and the ranks of the recovery facilities. It is obvious from table 7.14 that G is the most desirable facility and F is the least desirable. If the decision maker has a cutoff limit of, say, 100, he or she will identify facilities E and G as efficient.
7.4
Evaluation Criteria for Third and Fourth Models
Like in evaluation of collection centers (see chapter 6), in evaluation of recovery facilities one may have a scenario with three different categories of decision makers with multiple, conflicting, and incommensurate goals, as follows:
TABLE 7.8 Preference table (second model)
Criteria g1 g2 Criteria g3 g4 g5 g6 tp1+ 10 12 tp1– 0.6 1.1 250 10 tp2+ 15 14 tp2– 0.4 0.9 200 7 tp3+ 25 15 tp3– 0.3 0.7 140 6 tp4+ 30 16 tp4– 0.2 0.6 120 4 tp5+ 45 17 tp5– 0 0.4 100 3
© 2009 by Taylor & Francis Group, LLC
Evaluation of Recovery Facilities
TABLE 7.9 Criteria values for each recovery facility (second model)
Criteria g1 g2 g3 g4 g5 g6 Facility E 22 15 0.4 0.5 200 8 Facility F 30 17 0.3 0.8 220 6 Facility G 15 10 0.1 0.5 145 4
133
TABLE 7.10 Output of LPP weight algorithm (second model)
Criteria g1 g2 g3 g4 g5 g6 Δwp2+ 0.02 0.05 — — — — Δwp3+ 0.012 0.266 — — — — Δwp4+ 0.168 0.683 — — — — Δwp5+ 0.011 2.157 — — — — Δwp2– — — 0.05 0.05 0.033 0.033 Δwp3– — — 0.115 0.280 0.077 0.077 Δwp4– — — 0.380 0.215 0.253 0.979 Δwp5– — — 1.252 1.252 0.835 2.505
TABLE 7.11 Deviations of criteria values of recovery facility E from target values (second model)
Criteria g1 g2 g3 g4 g5 g6 r=2 d12+ = 12 d22– = 3 d32– = 0.2 d42– = 0.6 d52– = 50 d62– = 2 r=3 d13+ = 7 d23– = 1 d33– = 0 d43– = 0.4 d53– = 0 d63– = 1 r=4 d14+ = 3 d24– = 0 d34– = 0.1 d44– = 0.2 d54– = 60 d64– = 2 r=5 d15+ = 8 d25– = 1 d35– = 0.2 d45– = 0.1 d55– = 80 d65– = 4
TABLE 7.12 Deviations of criteria values of recovery facility F from target values (second model)
Criteria g1 g2 g3 g4 g5 r=2 d12+ = 20 d22– = 0.3 d32– = 0.3 d42– = 30 d52– = 4 r=3 d13+ = 15 d23– = 0.1 d33– = 0.1 d43– = 20 d53– = 1 r=4 d14+ = 5 d24– = 0 d34– = 0.1 d44– = 80 d54– = 0 r=5 d15+ = 0 d25– = 0.1 d35– = 0.2 d45– = 100 d55– = 2
© 2009 by Taylor & Francis Group, LLC
1 d45– = 25 d55– = 0 TABLE 7.6 d42– = 105 d52– = 6 r=3 d13+ = 0 d23– = 0. the higher the facility’s suitability. LLC . i..95 52.3 present the lists of criteria that are considered for the three categories.1.4 d43– = 55 d53– = 3 r=4 d14+ = 10 d24– = 0. less hazardous) Proximity to residential area (PH) (the closer the facility is to the residential area. i. the better) Salary offered to employees at recovery facility (SA) (the higher.5 d32– = 0. less hazardous) Employment opportunity (EO) (the higher. 7.1 d35– = 0. Sections 7.4.2 Criteria of Local Government Officials Proximity to surface water (PS) (the closer the facility is to surface water.134 Strategic Planning Models for Reverse and ClosedLoop Supply Chains TABLE 7.4.4. the efficiency of a candidate recovery facility must be evaluated based on the maximized consensus among decision makers of the three categories.2 d44– = 5 d54– = 2 r=5 d15+ = 15 d25– = 0.2 d34– = 0..1 Criteria of Consumers Proximity to surface water (PS) (the closer the facility is to the surface water. less hazardous) © 2009 by Taylor & Francis Group.3 d33– = 0. the higher the facility’s suitability. the higher the facility’s suitability.2.14 Total scores and ranks of recovery facilities (second model) Recovery facilities E F G Total scores 99.e.e. the better) 7.4. and 7. 7.26 Ranks II III I Consumers whose primary concern is convenience Local government officials whose primary concern is environmental consciousness Supply chain company executives whose primary concern is profit Therefore.4.85 117.e..13 Deviations of criteria values of recovery facility G from target values (second model) Criteria g1 g2 g3 g4 g5 r=2 d12+ = 5 d22– = 0. i.
and then TOPSIS (technique for order preference by similarity to ideal solution) is employed to find the efficiency of each candidate recovery facility. are considered for evaluation. the better) Utilization of incentives from local government (UI) (the higher. local government officials. Tables 7. as evaluated by that category. calculated using the eigen vector method. are the elements of the normalized eigen vectors of pairwise comparison matrices shown in tables 7. less hazardous) 7.20 show the impacts given for the criteria of the consumers.5 Third Model (Eigen Vector Method. TOPSIS.15–7. the easier is the transportation) Quality of reprocessed products (QO)–quality of used products (QI) (the higher.17.e. impacts are given to the criteria identified for each category of decision makers (see section 7.4). Three recovery facilities..3 Criteria of Supply Chain Company Executives Space cost (SC) (the lower. and K. the better) Labor cost (LC) (the lower. Borda’s choice rule is used to combine individual evaluations for each candidate recovery facility into a group evaluation or maximized consensus ranking. the higher the facility’s suitability. and supply chain company executives. the better) Proximity to roads (PR) (the closer the facility is to roads.4. In the first phase. respectively. H.Evaluation of Recovery Facilities 135 Proximity to residential area (PH) (the closer the facility is to the residential area. In the second phase. the better) Throughput (TP)×disassembly time (DT) (the higher. the better) Pollution control (PC) (the higher. respectively. the better) Throughput (TP)/supply (SU) (the higher. J. and supply chain company executives. and Borda’s Choice Rule) This model to select efficient recovery facilities is implemented in two phases. using the eigen vector method.18–7. For example. 7.17 show the pairwise comparison matrices as formed for the consumers. the better) 7. the relative impacts © 2009 by Taylor & Francis Group. The model to select efficient recovery facilities is presented using a numerical example.1 Phase I (Individual Decision Making) Tables 7. local government officials.15–7. respectively.5. i. These sets of impacts. LLC .
focus groups. individual interviews.15–7. representatives from each category of decision makers could be invited to participate in relevant survey questionnaires.15 Pairwise comparison matrix for consumers (third model) Criteria PS PH EO SA PS 1 2 2 3 PH 1/2 1 1 1 EO 1/2 1 1 1 SA 1/3 1 1 1 TABLE 7.17 Pairwise comparison matrix for supply chain company executives (third model) Criteria SC LC PR QO–QI TP/SU TP×DT UI PC SC 1 1 1 1 2 1/6 1 1 LC 1 1 1 1 1 1/5 1/2 1 PR 1 1 1 1 7 1 1 3 QO–QI 1 1 1/9 1 1 1/2 1 4 TP/SU ½ 1 1/7 2 1 0.23. and onsite observations. The decision matrices formed for the consumers. respectively. are shown in tables 7. local government officials.16.2 1 1 TP×DT 6 5 1 2 5 1 9 1 UI 1 2 1 1 1 1/9 1 1 PC 1 1 1/3 1/4 1 1 1 1 of the criteria PH (proximity to residential area) and PR (proximity to roads) shown in table 7.136 Strategic Planning Models for Reverse and ClosedLoop Supply Chains TABLE 7.21–7. The elements of these matrices are the ranks (ranging from 1 to 10) assigned to the recovery facilities with respect to each criterion for evaluation. A lower rank implies higher efficiency (with respect to that criterion). To facilitate the construction of the pairwise comparison matrices (see tables 7. © 2009 by Taylor & Francis Group. LLC .23). and supply chain company executives.16 Pairwise comparison matrix for local government officials (third model) Criteria PS PH PS 1 3 PH 1/3 1 TABLE 7.21–7.17) and the decision matrices (see tables 7.19 are the elements of the normalized eigen vector of the pairwise comparison matrix shown in table 7. for implementation of the TOPSIS for each category.
49) on each element of table 7.0491 0.0717 0.18 Impacts for consumers (third model) Criteria PS PH EO SA Impacts 0.75 TABLE 7.1437 0. For example.21 (decision matrix formed for the consumers). LLC .19 Impacts for local government officials (third model) Criteria PS PH Impacts 0.21 and 7.Evaluation of Recovery Facilities TABLE 7. Table 7.2804 0.1663 Now one is ready to perform the six steps in the TOPSIS for each category of decision makers.24 shows the normalized decision matrix formed by applying equation (4.3116 137 TABLE 7.1905 0. Step 1: Construct the normalized decision matrix. The following steps show the implementation of the TOPSIS for the consumers to evaluate H.20 Impacts for supply chain company executives (third model) Criteria SC LC PR QO–QI TP/SU TP×DT UI PC Impacts 0. © 2009 by Taylor & Francis Group. and K.1356 0.2673.1277 0.1233 0. J.25 0.24) is calculated as follows: r22 1 2 2 12 32 0.2804 0.1198 0. the normalized rank of recovery facility J with respect to criterion PH (see tables 7.
24)..2981 0.5963 0.23 Decision matrix for supply chain company executives (third model) Recovery facilities H J K SC 3 4 9 LC 2 7 8 PR 1 1 7 QO–QI 3 2 5 TP/SU 4 2 1 TP×DT 1 2 6 UI 1 3 5 PC 2 4 6 TABLE 7. © 2009 by Taylor & Francis Group. 0.22 Decision matrix for local government officials (third model) Recovery facilities H J K PS 4 3 1 PH 2 1 5 TABLE 7.7454 PH 0.138 Strategic Planning Models for Reverse and ClosedLoop Supply Chains TABLE 7.3015 Step 2: Construct the weighted normalized decision matrix.24.e.4650 0.8018 EO 0.2804 (see table 7.18).e.5345 0.8137 0.0749 (see table 7. by the normalized rank of R2 with respect to PH. Table 7. i. i. is calculated by multiplying the impact of PH.9045 0.25).18 and the normalized decision matrix in table 7.25 shows the weighted normalized decision matrix for the consumers. 0. 0..2673 0.3487 SA 0..21 Decision matrix for consumers (third model) Recovery facilities H J K PS 2 4 5 PH 2 1 3 EO 4 7 3 SA 1 3 1 TABLE 7.24 Normalized decision matrix for consumers (third model) Recovery facilities H J K PS 0. LLC .e. This is constructed using the impacts of the criteria listed in table 7.2673 (see table 7. For example.3015 0. i. the weighted normalized rank of recovery facility J with respect to criterion PH.
2287 TABLE 7.7394 0.e.1511) to the sum (i. Each column in the weighted normalized decision matrix shown in table 7. Using equation (4.54).5878 Step 3: Determine the ideal and negativeideal solutions. LLC .0761 0.0940 139 TABLE 7.3945) for recovery facility J (see table 7.e. 0. For example..2282 0. Step 4: Calculate the separation distances.0749. and the negativeideal solution (maximum rank) is 0.0940 0. the ideal solution (minimum rank) is 0.27) is the ratio of J’s negative separation distance (i.e.1604 S– 0. © 2009 by Taylor & Francis Group.26) for each recovery facility are calculated using equations (4.52). The separation distances (see table 7.3829) of its negative and positive separation distances (see table 7.2248.Evaluation of Recovery Facilities TABLE 7.26 Separation distances for consumers (third model) Recovery facilities H J K S* 0. the relative closeness coefficient is calculated for each recovery facility (see table 7.3945 0. 0.25 Weighted normalized decision matrix for consumers (third model) Recovery facilities H J K PS 0. For example.2318 = 0..26).2318 0.2318 0.25) and the ideal solutions (obtained in step 3) for the criteria.1511 0. For example (see table 7.0817 0. the relative closeness coefficient (i. the positive separation distance for recovery facility J (see table 7. respectively.1511 + 0.25 has a minimum rank and a maximum rank.53).0749 0. They are the ideal and negativeideal solutions.0381 0.1304 0.27 Relative closeness coefficients for consumers (third model) Recovery facilities H J K C* 0.. which contains the weighted normalized ranks of J (see table 7.25).52) and (4.2819 0.0952 PH 0.2248 EO 0.1499 0. with respect to criterion PH. for the corresponding criterion. Step 5: Calculate the relative closeness coefficient.26) is calculated using equation (4. 0.27).0978 SA 0.
K.140 Strategic Planning Models for Reverse and ClosedLoop Supply Chains Step 6: Form the preference order. 7. 5.e.e. as evaluated by the consumers). and supply chain company executives (i. TOPSIS. and K (viz. as calculated for those two categories of decision makers. the preference order for the recovery facilities is H. For example. The relative closeness coefficients of the recovery facilities. respectively) are also shown. local government officials. Borda’s Choice Rule) This model assumes that neither numerical nor linguistic impacts are available for the evaluation criteria. 4) is calculated by summing the marks of J for the consumers.2941 Supply chain company executives 0. J. TOPSIS is implemented for the local government officials and supply chain company executives in a similar manner. Borda scores (group evaluations) calculated for H. and 2. 0 + 2 + 2).29 Marks and Borda scores of recovery facilities (third model) Recovery facilities H J K Local government officials 1 2 0 Supply chain company executives 2 2 1 Consumers 2 0 1 Borda scores 5 4 2 © 2009 by Taylor & Francis Group.5. 4. and J (that means H is the best recovery facility.6 Fourth Model (Neural Networks.5962 0. 7. Because H has the highest Borda score. Because the best alternative is the one with the highest relative closeness coefficient.28.6715 0..28 Relative closeness coefficients for local government officials and supply chain company executives (third model) Recovery facilities H J K Local government officials 0.. and supply chain company executives.. the Borda score for J (i.5952 0.8487 0. local government officials. LLC . it is the best of the lot. are shown in table 7.29 shows the marks of the recovery facilities as given using Borda’s choice rule for the consumers. Fuzzy Logic.2 Phase II (Group Decision Making) Table 7.4057 TABLE 7. It employs a neural network [6] to evaluate TABLE 7.
as calculated for each category. three layers are considered in each network. in the third phase. To this end. LLC .1 Phase I (Derivation of Impacts) Suppose that one has the linguistic ratings of ten existing recovery facilities.4. 3) Defuzzified ratings 9 7. Finally. there are ten input– output pairs for each neural network because there are ten existing recovery facilities.30 Conversion table for ratings (fourth model) Linguistic ratings Very good (VG) Good (G) Fair (F) Poor (P) Very poor (VP) TFNs (7.6. respectively. 10.Evaluation of Recovery Facilities 141 the efficiency of a recovery facility of interest (which is being considered for inclusion in a reverse supply chain). A neural network is constructed and trained for each category of decision makers. of the recovery facility of interest.31–7. 5) (0. and supply chain company executives. Table 7. calculates impacts of criteria identified for each category of decision makers given in section 7. the model to evaluate the efficiency of a recovery facility of interest is carried out in three phases. efficiency. 7. 5. 3.. Defuzzification of a TFN can be performed using equation (4. the ratings of existing recovery facilities are used to construct a neural network that. Tables 7. i.33 show the defuzzified overall rating of each existing recovery facility as well as the recovery facility’s defuzzified rating with respect to each criterion. in turn. Using fuzzy logic. 0.30 shows not only one of the many ways for conversion of linguistic ratings into TFNs but also the defuzzified ratings of the corresponding TFNs.8). In the example. 8) (1. as given by an expert in each category of decision makers described in section 7.e. Then. using linguistic performance measures of recovery facilities that already exist in the reverse supply chain. in the second phase. The number of nodes in the output layer is one (for overall rating). 10) (2. Borda’s choice rule is employed to calculate the maximized consensus rating (among the categories considered).4. 10) (5. 7. In the first phase. the impacts obtained in the first phase are used in a fuzzy TOPSIS (combination of fuzzy logic and TOPSIS) method to obtain the overall rating of the recovery facility of interest.3 5 3 1 © 2009 by Taylor & Francis Group. Also. and the number in the input layer is the number of criteria TABLE 7. using the defuzzified ratings of the existing recovery facilities with respect to criteria as input sets and the recovery facilities’ defuzzified overall ratings as corresponding outputs. local government officials. with five nodes in the hidden layer. as evaluated by the consumers. these linguistic ratings are converted into triangular fuzzy numbers (TFNs).
33 1 9 5 3 7.33 Supply chain company executives’ ratings of recovery facilities (fourth model) Recovery facilities R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 SC 9 5 1 1 5 9 5 1 1 5 LC 3 7.142 Strategic Planning Models for Reverse and ClosedLoop Supply Chains TABLE 7.33 5 Overall 3 1 9 5 7.33 5 1 9 5 7.33 PC 3 1 9 5 1 3 1 9 5 1 Overall 5 3 1 7.33 UI 3 1 5 1 7.33 9 9 3 9 7.33 5 5 9 3 7.33 9 9 1 3 5 7.31 Consumer ratings of recovery facilities (fourth model) Recovery facilities R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 PS 9 5 1 1 5 1 9 3 3 5 PH 3 7.33 5 5 9 3 1 1 9 1 EO 7.33 3 1 5 1 7.33 © 2009 by Taylor & Francis Group.33 TABLE 7.33 1 1 7. LLC .33 5 5 3 TABLE 7.33 5 5 9 PR 9 3 1 3 1 9 3 1 3 1 QO–QI 5 1 3 5 3 5 1 3 5 3 TP/SU 3 7.33 5 5 9 3 1 1 9 1 Overall 5 3 9 7.32 Local government officials’ ratings of recovery facilities (fourth model) Recovery facilities R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 PS 9 5 1 1 5 1 9 3 3 5 PH 3 7.33 9 1 5 5 5 3 3 1 3 SA 3 1 5 9 3 3 5 1 7.33 1 9 5 TP×DT 9 9 3 9 7.
2 Neural network for consumers (fourth model).2 shows the neural network constructed and trained for the consumer category. LLC . local government officials. © 2009 by Taylor & Francis Group. the absolute value of Wv is the impact of the vth input node upon the output node.36 show the impacts of the criteria considered for the consumers.34–7.Evaluation of Recovery Facilities 143 PS H H PH O H EO H H SA Input layer Hidden layer Output layer FIGURE 7.7) i j i  Iij  Tables 7. considered by the corresponding category. After each neural network is trained. respectively. and Oj is the connection weight from the jth hidden node to the output node: nH nV j I vj  Iij  i Oj Wv  nV nH I vj nV Oj (7. Iij is the connection weight from the ith input node to the jth hidden node. For example. the following equation [1] is used to calculate the impacts of criteria considered by the corresponding category. Here. nH is the number of hidden nodes. and supply chain company executives. nV is the number of input nodes. figure 7.
11 TP×DT 0.35 Impacts of criteria of local government officials (fourth model) Criteria Impacts PS 0. LLC .17 PH 0.61 PH 0.39 TABLE 7.10 PC 0.33 PR 1 9 3 QO–QI 3 5 1 TP/SU 5 3 7.39 Decision matrix for supply chain company executives (fourth model) Recovery facilities R11 R12 R13 SC 5 9 5 LC 9 3 7.34 Impacts of criteria of consumers (fourth model) Criteria Impacts PS 0.38 Decision matrix for local government officials (fourth model) Recovery facilities R11 R12 R13 PS 3 3 5 PH 1 9 1 TABLE 7.19 TP/SU 0.02 TABLE 7.06 LC 0.36 Impacts of criteria of supply chain company executives (fourth model) Criteria Impacts SC 0.33 9 9 UI 7.38 TABLE 7.39 UI 0.33 TP×DT 7.144 Strategic Planning Models for Reverse and ClosedLoop Supply Chains TABLE 7.27 SA 0.18 EO 0.37 Decision matrix for consumers (fourth model) Recovery facilities R11 R12 R13 PS 1 1 5 PH 5 5 9 EO 1 5 5 SA 5 9 3 TABLE 7.10 QO–QI 0.02 PR 0.33 3 1 PC 1 3 1 © 2009 by Taylor & Francis Group.
. respectively (table 7. 0. The separation distances (see table 7.42) for each recovery facility are calculated using equations (4.41).39. A fuzzy TOPSIS method uses the impacts obtained in the first phase to calculate the overall ratings of the three recovery facilities.40). Now one is ready to perform the six steps in the TOPSIS for each category of decision makers.0793 (see table 7.Evaluation of Recovery Facilities 7.e. Each column in the weighted normalized decision matrix shown in table 7.53). the positive separation distance for © 2009 by Taylor & Francis Group. Step 3: Determine the ideal and negativeideal solutions. the weighted normalized rank of recovery facility R12 with respect to criterion PH.4369.40) is calculated as follows: r22 5 5 2 52 92 0. of interest. For example (see table 7.40 shows the normalized decision matrix formed by applying equation (4.52) and (4. For example.41 shows the weighted normalized decision matrix for the consumers. R12. and R13) in this example are shown in tables 7.e. R12.41). The following steps show the implementation of the TOPSIS for the consumers to evaluate R11.5). i. The decision matrices formed for the consumers.37 and 7.37–7.41 has a maximum rating and a minimum rating. for the corresponding criterion.0793. the matrices can be constructed by inviting representatives from each category of decision makers to participate in relevant survey questionnaires. Table 7. Step 2: Construct the weighted normalized decision matrix.. For example.2 Phase II (Individual Decision Making) 145 Suppose that there are three recovery facilities. is calculated by multiplying the impact of PH. to convert linguistic ratings given by each category into TFNs).30 is used here as well.6. Like in the third model (see section 7.1428. R11. the ideal solution (maximum rating) is 0. i. the normalized rating of recovery facility R12 with respect to criterion PH (see tables 7. Table 7. local government officials. and the negativeideal solution (minimum rating) is 0. focus groups. Step 1: Construct the normalized decision matrix. with the normalized rating of R12 with respect to PH. This is constructed using the impacts of the criteria listed in table 7.18 (see table 7. Step 4: Calculate the separation distances. 0.34 and the normalized decision matrix in table 7. individual interviews.e. They are the ideal and negativeideal solutions.4369 (see table 7.40.. and R13.49) on each element of table 7. LLC . with respect to criterion PH. i.34). For example. R12. 0. respectively. and onsite observations. and R13. and supply chain company executives (with defuzzified ratings for R11.37 (decision matrix for the consumers).
.42) is calculated using equation (4. Similarly. as evaluated by the consumers).1857 SA 0..2602 0.41 Weighted normalized decision matrix for consumers (fourth model) Recovery facilities R11 R12 R13 PS 0. © 2009 by Taylor & Francis Group.7001 0. Because the best alternative is the one with the highest relative closeness coefficient.0793 0. Step 5: Calculate the relative closeness coefficient.2136 S– 0.2526 0.44.1464 0.0371 0.1400 0. 0.2602 + 0. LLC .2798 TABLE 7. and R11 (that means R12 is the best recovery facility.52).e.40 Normalized decision matrix for consumers (fourth model) Recovery facilities R11 R12 R13 PS 0.1780 0. TOPSIS is implemented for the local government officials and supply chain company executives. 0. For example.43). R13. Step 6: Rank the preference order.41) and the ideal solutions (obtained in step 3) for the criteria.e.0712 0. The relative closeness coefficients of the recovery facilities.1857 0.7863 EO 0.1464 = 0.6398) for recovery facility R12 (see table 7.4369 0..1923 0. the relative closeness coefficient is calculated for each recovery facility (see table 7. Using equation (4.42 Separation distances for consumers (fourth model) Recovery facilities R11 R12 R13 S* 0.4066) of its negative and positive separation distances (see table 7.e. 0.1925 0. are shown in table 7.7001 SA 0. the preference order for the recovery facilities is R12.2602) to the sum (i. the relative closeness coefficient (i.9623 PH 0.42).1428 EO 0.0793 0.43) is the ratio of R12’s negative separation distance (i.4663 0.146 Strategic Planning Models for Reverse and ClosedLoop Supply Chains TABLE 7.1650 PH 0. as calculated for those two categories.8393 0.2086 recovery facility R12 (see table 7.0330 0.0330 0.54).3203 0. which contains the weighted normalized ratings of R12 (see table 7.1068 TABLE 7.4369 0.
6. TABLE 7..458016 0. and supply chain company executives (i. the Borda score for R12 (i.e.44 Relative closeness coefficients for local government officials and supply chain company executives (fourth model) Recovery facilities R11 R12 R13 Local government officials 0 0.646968 0. local government officials. it is the best of the lot.45 Marks and Borda scores of recovery facilities (fourth model) Recovery facilities R11 R12 R13 Local government officials 0 2 1 Supply chain company executives 1 2 0 Consumers 0 2 1 Borda scores 1 6 2 © 2009 by Taylor & Francis Group..e. 2 + 2 + 2).695261 0. Borda scores (group evaluations) calculated for R11..588066 Supply chain company executives 0. For example.6398 0. 6. R12.Evaluation of Recovery Facilities TABLE 7. respectively) are also shown.45 shows the marks of the recovery facilities as given using Borda’s choice rule for the consumers. local government officials.29753 7. and R13 (viz. and supply chain company executives. and 2.43 Relative closeness coefficients for consumers (fourth model) Recovery facilities R11 R12 R13 C* 0.2199 0. 6) is calculated by summing the marks of R12 for the consumers.3 Phase III (Group Decision Making) Table 7. Because R12 has the highest Borda score. 1.4942 147 TABLE 7. LLC .
In this approach. The zones are defined as follows: Cpk H Zone 3.7 Fifth Model (TwoDimensional Chart) Here. it is assumed that the two most important criteria are (1) n value in n Sigma of the recovery facility and (2) TP/SU value.’s approach uses Cpk (process capability index) and a price comparison (CPC) chart for the selection of efficient suppliers (see section 4. Cpk is drawn on the Yaxis.0 3.5 0.0 2. where only the two most important evaluation criteria are considered. Linn et al.’s approach [5] for selection of suppliers in a traditional supply chain.0 1. which integrates the process capability and price information of multiple suppliers (assuming every other criterion value is either unimportant or the same for all suppliers). LLC .3 CPC chart R 0. The CPC chart (see figure 7.5 © 2009 by Taylor & Francis Group.0 FIGURE 7.0 C S G E 2.5 E Zone 2.5 1.0 0. a simple twodimensional chart.148 Strategic Planning Models for Reverse and ClosedLoop Supply Chains 7. and the ratio (R) of target price desired to price quoted by the supplier is drawn on the Xaxis.0 G 1.3). Cpk and R of each supplier are then plotted on the chart.5 S C 1.16 for the concept of process capability).0 U Zone 0.5 3. is presented. provides a method to consider quality and price simultaneously in the supplier selection process. The chart is partitioned into six different zones representing the quality performance and price levels.5 2. This approach is similar (but with some important modifications) to Linn et al.
The supplier selection should start from the E zone and follow the sequence of E G S C H U. those suppliers falling in the E zone are considered the best group of suppliers to choose from. the TP/SU criterion is used.0 and R > 2. For example (see figure 7. those falling in the lowerhalf zone (below the 45° line) have better cost performance. those in the lower half of the zone should be selected. LLC . n)  0.5} – E zone 149 S zone: Satisfactory zone. Cpk)  Cpk > 1. Cpk)  Cpk > 1. If the objective is to find a betterquality performer. those close to the 45° line should be selected.0 and R > 0} Because of their highquality performance and lowcost quotation. If the standard deviation ( ) of the width is 5 inches and the mean shift ( ) is 6 inches. Therefore. See figure 7. those in the upperhalf zone should be selected. G = {(R. E = {(R. Cpk)  Cpk > 1. U = {(R. n)  TP/SU ≥ 0. let the upper specification (U) of the width (critical dimension) of a reprocessed good at a candidate recovery facility be 100 inches. Within the same zone.4). For the chart’s Xaxis. Those falling in the U zone are simply not acceptable because their Cpk value is too low.0} G zone: Good zone. H = {(R.16 for the concepts of process capability and n Sigma). it must be noted that the Cpk value is not enough to judge quality. Cpk)  Cpk > 2.0 and R > 1.8 and n ≥ 4.0} – zone I – zone II © 2009 by Taylor & Francis Group.0} – zone I Zone III = {(TP/SU. These zones are defined as follows: Zone I = {(TP/SU. Cpk)  Cpk > 1. In contrast.0} – G zone – E zone – S zone H zone: Highprice zone.0 and R < 1. one needs the Cp (process capability ratio) value as well. Cpk)  Cpk < 1.7 and n ≥ 3. those in the upperhalf zone (above the 45° line) have better quality performance. the lower specification (L) 60 inches. if the objective is to select a bettercost performer.5 and R > 1.9 ≤ TP/SU ≤ 1 and n ≥ 5. S = {(R. C = {(R.0} Zone II = {(TP/SU.0} U zone: Unacceptable zone.33 and R > 1. To overcome this problem in the selection of efficient recovery facilities. n)  TP/SU ≥ 0. which could be a function of factors such as efficiency of reprocessing and efficiency of delivery (of reprocessed goods to demand centers). The n value for a recovery facility represents the facility’s quality. and the target value ( ) 80 inches.Evaluation of Recovery Facilities E zone: Excellent zone.33} – G zone – E zone C zone: Capable zone. The chart is partitioned into five acceptable regions (I–V) and an unacceptable region.5 for the chart that is used for selection of efficient recovery facilities. the n value (instead of Cpk) is used on the Yaxis of the chart (see section 4. then reprocessing in the candidate recovery facility is a Four Sigma process. Although the above approach can integrate both quality and price. If quality and cost are equally important.
0 0.0 2.5 and n ≥ 2.5 n and TP/SU chart (fifth model).0 Unacceptable Region TP/SU 0.6 0. © 2009 by Taylor & Francis Group.7 0.6 and n ≥ 2.150 Strategic Planning Models for Reverse and ClosedLoop Supply Chains Zone IV = {(TP/SU.0} – zone I – zone II – zone III – zone IV Unacceptable zone = {(TP/SU. n)  TP/SU ≤ 0.4 Critical dimension of a reprocessed good at recovery facility (fifth model).0} – zone I – zone II – zone III Zone V = {(TP/SU. n 6.0 1.5 0.1 0.0 FIGURE 7.3 0. n)  TP/SU ≥ 0.0} 6 inches = 5 inches Inches 60 80 100 FIGURE 7.5 or n < 2.0 I 5.0 0.0 V 4.0 Unacceptable Region III IV II 3. n)  TP/SU ≥ 0.8 0.9 1.4 0. LLC .2 0.
their TP/SU value is low. A multiphase mathematical programming approach for effective supply chain design. LLC . 4. Those falling in the unacceptable zone are simply not acceptable because their n value (2 or lower) is too low.. Whereas the third model uses the eigen vector method. TOPSIS. 7. Baker. If the decision maker wishes to raise or lower the bar for quality. The first model uses analytic hierarchy process in a situation where the evaluation criteria can be given numerical impacts.Evaluation of Recovery Facilities 151 Because of their highquality (Five Sigma or higher) performance and high TP/SU value. the fourth model uses neural network. S. A. R. Talluri. TOPSIS. the fifth model presents a twodimensional chart that can be used when a very simple evaluation technique is desired (only the two most important evaluation criteria are considered). References 1. if the decision maker considers only those recovery facilities with at least Six Sigma quality (besides high TP/SU ratio) the best. The second model employs linear physical programming in a situation where the evaluation criteria are presented in terms of ranges of different degrees of desirability. Finally. The third and fourth models have decision makers with conflicting evaluation criteria.. or both.. and Jung. five models are presented for identifying efficient recovery facilities in a region where a reverse supply chain is to be designed. and Borda’s choice rule. Linear physical programming: A new approach to multiple objective optimization. he or she can revise the respective charts accordingly. M. Gupta. Talluri. and Akbulut. fuzzy logic. International Journal of Production Economics 62:133–44. S. Y...9 ≤ TP/SU ≤ 1}. R. For example. 1999.. Messac. 2002. and Baker. 3. S. C.8 Conclusions In this chapter. © 2009 by Taylor & Francis Group. The recovery facility selection should start from zone I and follow the sequence of I II III IV V. 2. Cha. and Sarkis. then zone I in his or her chart will be {(n. 2003. C. 1996.0 and 0. J. those recovery facilities falling in zone I are considered the best group of recovery facilities to choose from. M. TP/SU)  n ≥ 6. Framework for designing efficient value chain networks. and Borda’s choice rule. European Journal of Operational Research 141:544–58. B. International Journal of Production Research 41:1831–49. Satisfaction assessment of multiobjective schedules using neural fuzzy methodology. Transactions on Operational Research 8:39–59.
2006. Tsung. F. Englewood Cliffs. Neural networks: A comprehensive foundation. W. Linn. C. Haykin. NJ: Prentice Hall. R. 6. S. and Ellis. LLC .. L. 1998. Quality Engineering 18:123–29.152 Strategic Planning Models for Reverse and ClosedLoop Supply Chains 5. J. Supplier selection based on process capability and price analysis. 2nd ed.. © 2009 by Taylor & Francis Group.
but the decisionmaking criteria could be imprecise. 2. LLC . Besides optimal transportation of products. so that products can be manufactured directly from raw materials or remanufactured from used products (the final demand for the product is met with either new or remanufactured products). 153 © 2009 by Taylor & Francis Group. The second model is for scenario 2 and employs linear physical programming. 4. 3. and use goal programming and linear physical programming.1 The Issue In this chapter. In this chapter. The fifth model is for scenario 5 and employs fuzzy goal programming. but the decisionmaking criteria could be presented in terms of different degrees of desirability. Decisionmaking criteria for a reverse supply chain could be given in terms of classical supplyanddemand constraints. The various scenarios for this problem could differ as follows: 1. it is assumed that the manufacturer has incorporated a remanufacturing process into his or her original production system. For the last three models.8 Optimization of Transportation of Products 8. respectively. and new) across a reverse or closedloop supply chain while satisfying certain constraints. Also. there could be a need to address the following issues in one continuous phase for a closedloop supply chain: selection of used products and evaluation of production facilities. The situation could be the same as in scenario 3. respectively. remanufactured. 5. various models to address the above scenarios are presented. the focus is on achieving transportation of the right quantities of products (used. The first model addresses scenario 1 and employs linear integer programming. The situation could be the same as in scenario 3. the decisionmaking criteria could be presented in terms of classic supplyanddemand constraints. The third and fourth models are for scenarios 3 and 4. Decisionmaking criteria for a reverse supply chain could be presented in terms of ranges of different degrees of desirability.
Section 8.2.2 First Model (Linear Integer Programming) This model considers a generic reverse supply chain consisting of collection centers. Section 8.2.2 presents the first model.2. Optimal transportation of products (used and remanufactured) is achieved using linear integer programming.7 gives some conclusions. and a numerical example is given in section 8. 8. Section 8. and demand centers.1. section 8. Nomenclature for the model is given in section 8.5 gives the fourth model.1 a1 a2 Nomenclature Space occupied by one unit of remanufactured product Space occupied by one unit of used product CAPv Capacity of remanufacturing facility v to remanufacture products Cu dw Iuv Cost per product retrieved at collection center u Demand of remanufactured products at demand center w Decision variable representing the number of products to be transported from collection center u to remanufacturing facility v Decision variable representing the number of products to be transported from remanufacturing facility v to demand center w Cost of remanufacturing per product at production facility v Storage capacity of remanufacturing facility v for remanufactured products Storage capacity of remanufacturing facility v for used products Storage capacity of collection center u for used products Cost of transporting one product from collection center u to remanufacturing facility v Ovw Rv S1v S2v Su TIuv SUPu Supply at collection center u TOvw Cost of transporting one product from remanufacturing facility v to demand facility w © 2009 by Taylor & Francis Group. LLC . 8. the model is formulated in section 8.3 shows the second model.2. remanufacturing facilities. Finally.4 presents the third model. See [1] for a review of a number of models similar to the one presented in this section.6 presents the fifth model.3.154 Strategic Planning Models for Reverse and ClosedLoop Supply Chains This chapter is organized as follows: Section 8. Section 8.2.
inventory. Cu. i.e.Qvw (8.2) Total output of each remanufacturing facility is at most its total input © 2009 by Taylor & Francis Group. and transportation of products (used and remanufactured) across the supply chain (in the formulation.Iuv u v v w ( Rv / 4 )..1) subject to Demand at each demand center must be met Ovw v dw . Rv): Minimize Retrieval costs Transportation costs Remanufacturing costs Inventory costs Cu Iuv u v TIuv Iuv u v v w TOvwOvw RvOvw v w (Cu / 4 ). LLC . and that of a remanufactured product is 25% of its remanufacturing cost. it is assumed that the inventory cost of a used product is 25% of its retrieval cost.Optimization of Transportation of Products u v w Yv 8. w (8.2 Collection center Remanufacturing facility Demand center Binary variable (0/1) for selection of recovery facility v Model Formulation 155 The following is the singleperiod and singleproduct transshipment model formulation that is implemented to achieve minimum overall cost. remanufacturing.2. sum of the costs of retrieval.
Iuv u S2 v . v . LLC .6) Quantities of transported products are nonnegative numbers Iuv 0. v (8. The reason is that a component from a used product of one model may be used in a remanufactured product of a different model that may occupy a different amount of space. v (8. v (8.10) Note that a1 and a2 are possibly different.7) Total output of each remanufacturing facility is at most its capacity to remanufacture Ovw 0.Yv . v (8.3) Total space occupied by remanufactured products at each remanufacturing facility is at most its capacity for remanufactured products Total space occupied by used products at each collection center is at most its capacity a1 .4) a2 .Yv . u.8) Total quantity of used products supplied to remanufacturing facilities by each collection center is at most the supply to that collection center Ovw w CAPv . v (8.Ovw w S1v .Iuv v Su . w (8.9) Iuv v SUPu . © 2009 by Taylor & Francis Group. u (8. u (8.156 Strategic Planning Models for Reverse and ClosedLoop Supply Chains Iuv u w Ovw .5) Total space occupied by used products at each remanufacturing facility is at most its capacity for used products a2 .
5.e.1. CAPA = 300. two remanufacturing facilities. d1 = 100. TI1B = 3. one gets SUP1 = SUP2 = SUP3 = 211. S1 = 550.e.4. TI3A = 5.3.7. S1A = 550. TOB3 = 2.e. 211 products are to be transported from collection center 2 to recovery facility B Remanufactured Remanufacturing Used Products Products Facilities A and B Demand Centers Collection Centers Used Products Consumers FIGURE 8.Optimization of Transportation of Products 157 Also. TOB2 = 2.3 Numerical Example Three collection centers.1). 211 products are to be transported from collection center 1 to recovery facility A I1B = 0.. R A = 4. RB = 4. and three demand centers are considered in the example (see figure 8. a1 = a2 = 0. the collection centers are deemed to incur inventory costs).33 per period (see section 4. no products are to be transported from collection center 1 to recovery facility B I2A = 0. © 2009 by Taylor & Francis Group.6. the model may be revised accordingly in response to revision of existing constraints or addition of new constraints.. inventory costs at the collection centers and demand centers are not considered in the model (for example. d2 = 200. d3 = 150. TI1A = 3. TOB1 = 3. Moreover. i.4.2. TI2B = 1.11. The defuzzified supply is 633. C2 = 25. The other data used for implementation of the model are as follows: C1 = 29.. S1B = 550.1 Reverse supply chain (first and second models).3. no products are to be transported from collection center 2 to recovery facility A I2B = 211. i. S2 = 550.5. 700).2. S2B = 550. 650. 8.2. the following optimal solution is obtained: I1A = 211. TOA1 = 2. TI3B = 6. Let the total supply of the used product per period be a triangular fuzzy number (TFN): (600. LLC .4 for the concepts of TFN and defuzzification). TOB3 = 1. i. CAPB = 250 Upon application of the above data to the model. i. C3 = 37. TOA2 = 3. Assuming an equal supply rate at all three collection centers. if the supply of used products is higher than the demand for remanufactured products. using LINGO (v4).. S3 = 550. S2A = 550.6. the model assumes that there is always enough supply of the used products to satisfy the demand for the remanufactured products and that enough storage space (for used and remanufactured products) is always available at the remanufacturing facilities. However. TI2A = 4.e.
11) Class 1S criteria (smaller is better) Total transportation cost per period (g1) given by g1 = u v TIuvIuv v w TOvwOvw (8.1).e. i.3 Second Model (Linear Physical Programming) This model considers a generic reverse supply chain consisting of collection centers.e.2.e.2 gives a numerical example. 8. remanufacturing facilities. and section 8. i. no products are to be transported from recovery facility A to demand center 2 OA3 = 139.e. no products are to be transported from recovery facility B to demand center 1 OB2 = 200. Optimal transportation of products (used and remanufactured) is achieved using linear physical programming.... and demand centers..3..e. Section 8. i. 139 products are to be transported from recovery facility A to demand center 3 OB1 = 0.1 presents the model formulation (using the nomenclature given in section 8.e.e. i. LLC . i.3. 28 products are to be transported from collection center 3 to recovery facility A I3B = 0. 200 products are to be transported from recovery facility B to demand center 2 OB3 = 11.158 Strategic Planning Models for Reverse and ClosedLoop Supply Chains I3A = 28. 11 products are to be transported from recovery facility B to demand center 3 8.12) © 2009 by Taylor & Francis Group.1 Model Formulation Class 1H criteria Total retrieval cost per period (h1) given by CuIuv h1 = u v (8. i... i.e.. no products are to be transported from collection center 3 to recovery facility B OA1 = 100. i.3. 100 products are to be transported from recovery facility A to demand center 1 OA2 = 0.
17) d pr ≥ 0. LLC .Qvw (8.Optimization of Transportation of Products Total remanufacturing cost per period (g2) given by g2 = v w 159 RvOvw (8.20) center) Ovw w u Iuv v (number of remanufactured products is equal to number (8.21) of used ones) © 2009 by Taylor & Francis Group.19) Iuv u SUPu v (all products must be transported from each collection (8.15) gp – d pr ≤ t p(r 1) (deviation is measured from corresponding target value) (8.Iuv v w ( Rv / 4 ).16) gp ≤ t p5 (criterion value is in acceptable range) (8.13) Total inventory cost per period (g3) given by g3 = u v (Cu / 4 ).14) Goal constraints h1 ≤ RETMAX (retrieval cost is not more than maximum allowed value—RETMAX) (8.18) dw w (demand at each demand center must be met) (8. (deviation is a nonnegative number) System constraints Ovw v (8.
LLC . TI3B = 1. TOA2 = 5.1.2.22) Iuv 0 u. d3 = 80. TOB3 = 0.04. TI3A = 0.132 0.1.011 0. v (quantities of used products are nonnegative numbers) (8.3. S2A = S2C = 400 Furthermore. C2 = 0. RA = 0. d2 = 80. the target values for each soft criterion are shown in table 8.017 0.026 0.2. v (space occupied by used products is at most capacity) (8.2 Numerical Example Consider the reverse supply chain shown in figure 8.013 Δwp3+ 0.031 Δwp4+ 0.024 0.24) It should be noted here that more constraints may be added to the above model.012 © 2009 by Taylor & Francis Group.1. TOA1 = 4. 8.1 Preference table (second model) Criteria g1 g2 g3 tp1+ 100 150 70 tp2+ 200 250 150 tp3+ 300 290 250 tp4+ 400 450 300 tp5+ 500 600 450 TABLE 8. TOB1 = 1. TI1B = 0.881 Δwp5+ 0.160 Strategic Planning Models for Reverse and ClosedLoop Supply Chains a 2.1.23) Ovw 0 v.w (quantities of remanufactured products are nonnegative numbers) (8. TI2A = 0. using LINGO (v4). SUP3 = 25. Upon application of the above data to the model. the following optimal solution is obtained: TABLE 8.1.5. TI1A = 0.2. SUP1 = 75.1 in this example as well. and the incremental weights obtained by the LPP weight algorithm [2] are shown in table 8.085 0. SUP2 = 150. a1 = a2 = 0. The data used for implementation of the model are as follows: C1 = 0.Iuv u S 2v .479 0.025 0. d1 = 90.2. TOB2 = 0.3.2. as desired by the decision maker. TOA3 = 0.1.2. C3 = 0. TI2B = 3. RB = 0.02.2 Output of LPP weight algorithm (second model) Criteria g1 g2 g3 Δwp2+ 0.
LLC .Optimization of Transportation of Products I1A = 0. I3B = 25. OA3 = 80. Section 8. OB2 = 80. I1B = 75. 8.4. 10 = highest) Disassembly time for product i Disassembly cost/unit time Product type Bivw bi TAuv TBvw CCu CNPv CRv Cdi DIi DTi DC i © 2009 by Taylor & Francis Group. I3A = 0. section 8. I2B = 70. and section 8.4.3. Goal programming is employed to address the following issues in one continuous phase. OA1 = 0.2 presents the formulation of the model. production facilities. 8. OB1 = 90. and demand centers.1 gives the nomenclature for the model.2. besides optimal transportation of products: selection of used products and evaluation of production facilities. OA2 = 0. OB3 = 0 161 Interpretation of this solution could be made in a way similar to that in section 8. I2A = 80.4.3 illustrates the model with a numerical example.4.4 Third Model (Goal Programming) This model considers a generic closedloop supply chain consisting of collection centers.1 Aiuv Nomenclature Decision variable representing number of used products of type i transported from collection center u to production facility v Decision variable representing number of new/remanufactured products of type i transported from production facility v to demand center w Probability of breakage of product i Cost to transport one product from collection center u to production facility v Cost to transport one product from production facility v to demand center w Cost per product retrieved at collection center u Cost to produce one unit of new product at production facility v Cost to produce one unit of remanufactured product at production facility v Disposal cost of product i Disposal cost index of product i (0 = lowest.
determines the number of used products of each type to be picked for remanufacturing. in one continuous phase. © 2009 by Taylor & Francis Group. at production facility v Throughput (considering only remanufactured products) of production facility v Collection center Production facility Demand center Weight of product i Space occupied by one unit of used product Space occupied by one unit of remanufactured or new product Decision variable signifying selection of production facility v (1 if selected..162 Strategic Planning Models for Reverse and ClosedLoop Supply Chains MINTPS Minimum throughput per supply Nivw Decision variable representing number of new products of type i transported from production facility v to demand center w Net demand for product type i (remanufactured or new) at demand center w Percent of recyclable contents by weight in product i Total recycling revenue of product i Total resale revenue of product i Recycling revenue index of product i Storage capacity of production facility v for used products Storage capacity of production facility v for remanufactured and new products Storage capacity of collection center u Selling price of one unit of new product of type i Supply of used product i at collection center u Supply of used products that are fit for remanufacturing (viz. LLC . excluding products selected for recycling or disposal) and new products.4. 0 if not) Decision variable representing number of units of product type i picked for remanufacturing at collection center u (SUiu – Ziu = recycled or disposed) Factor that accounts for unassignable causes of variations at production facility v Model Formulation Ndiw PRCi RCYRi RSRi RCRIi S1v S2v Su SPi SUiu SFv TPv u v w Wi x1 x2 Yv Ziu v 8.2 This goal programming model.
25) 2. LLC . remanufactured.Optimization of Transportation of Products 163 identifies efficient production facilities. The revenue and cost criteria and the system constraints considered in the model are: Revenues 1. and achieves transportation of the right quantities of products (used. 3. which has an environmentally benign focus rather than a financial focus. respectively. Maximize the total profit in the supply chain (TP). Recycle revenue: {(SU iu i u Ziu ) RCRIi Wi PRCi } (8.28) © 2009 by Taylor & Francis Group. Minimize the number of disposed items (NDIS). Collection/retrieval cost: CCu SU iu u i (8. involves minimizing the positive deviation from the target value. 2. The following three goals are considered: 1. and that of a newly produced product is 25% of the production cost. New product sale revenue: SPi N ivw i v w (8.26) 3. Maximize the revenue from recycling (RR). and new) across a closedloop supply chain. Reuse revenue: {Ziu RSRi } i u (8. It is assumed that the inventory costs of a used product and a remanufactured product are 20% of the collection and remanufacturing costs.27) Costs 1. The first two goals involve minimizing the negative deviation from the respective target values. whereas the third goal.
Transportation cost = cost of transporting used products from collection centers to production facilities + cost of transporting remanufactured and new products from production facilities to demand centers: TAuv i u v Aiuv TBvw i v w ( Bivw N ivw ) (8.30) 4. The number of used products sent to all production facilities from collection center u must be equal to the number of used products picked for remanufacturing at that collection center: Aiuv v Ziu (8. Processing cost = disassembly cost of used products + remanufacturing cost of used products + new products’ production cost: DC i u v DTi Aiuv i u v CRv Bivw i v w CNPv N ivw (8. and the disposal cost index (a number on a scale 0–10. Inventory cost = cost of carrying used products’ inventory at collection centers + cost of carrying remanufactured products’ inventory at production facilities + cost of carrying newly manufactured products’ inventory at production facilities: (CCu /5) Aiuv i u v i v w {(CRv / 4 ) Bivw (CNPv / 4 ) N ivw } (8.31) 5. disposal cost depends on the percentage of recyclable content in the product. the higher the number. Disposal cost: Apart from the number of units disposed of and the cost to dispose of a product.32) System constraints 1. the more difficult or expensive it is to dispose of the product): {(SU iu i u Ziu ) DIi Wi (1 PRCi )} Cdi (8.33) © 2009 by Taylor & Francis Group. LLC .164 Strategic Planning Models for Reverse and ClosedLoop Supply Chains 2.29) 3.
34) 3. The total number of used products of all types collected (before accounting for probability of breakage) at all collection centers must be at least equal to the net demand (this is to encourage the use of remanufactured products): SU iu i u i w Ndiw (8. The number of used products of type i picked for remanufacturing at collection center u must be at most equal to the total number of used products fit for remanufacturing: Ziu SU iu (1 bi ) (8. The space occupied by used products at production facility v must at most be equal to the space available for used products at that facility: © 2009 by Taylor & Francis Group. The number of remanufactured products transported from production facility v to demand center w equals (number of used products fit for remanufacturing that are transported from collection center u to that production facility) multiplied by v. The number of remanufactured products must be at most equal to the net demand (this is to avoid excess remanufacturing): Ziu i u i w Ndiw (8.37) 6.Optimization of Transportation of Products 165 2. v is a factor that accounts for the unassignable (common) causes of variation at production facility v.36) 5.35) 4. That is. LLC .38) 7. there is no loss of products in the supply chain due to reasons other than common cause variations over which there is no control: Bivw w u Aiuv v v (8. The demand at each center w must be met by either new or remanufactured products: ( Bivw v N ivw ) Ndiw w (8.
2). Yv 8.166 Strategic Planning Models for Reverse and ClosedLoop Supply Chains x1 i u Aiuv S1v . SU12 = 45. SU33 = 28. The data used to implement the goal programming model are: CCu = 0.3 0 . SU23 = 22.01. and three types of products are considered in the example (see figure 8.39) 8. N ivw .4. SU11 = 50. SU22 = 38. CNP1 = 60.1. two demand centers. The space occupied by used products at collection center u must at most be equal to the space available for used products at that collection center: x1 i v aiuv Su (8.43) MINTPS (8. TA31 © 2009 by Taylor & Francis Group. SU32 = 35. TA22 = 0. The space occupied by new and remanufactured products at production facility v must at most be equal to the space available for new and remanufactured products at that production facility (assuming both new and remanufactured products occupy the same space): x 2 ( Bivw i w N ivw ) S2v Yv (8.09. DT1 = 10.01.40) 9. TA12 = 0.5. DT2 = 12.41) 10. CR 2 = 10. two production facilities. LLC . v .Yv (8. TA11 = 0. i (8.05. The ratio of throughput to supply of used products of a production facility must at least be equal to a preset value for the production facility to be considered efficient (this is valid only for remanufactured products): TPv / SFv Yv 11. Nonnegativity constraints: Aiuv .44) Numerical Example Three collection centers. Ziu 0. CR1 = 13. TA21 = 0. CNP2 = 45. 1 if selected (8. Bivw .1 v . DT3 = 9.42) 12. SU1 = 25. SU31 = 30. 0 if facility v not selected. u . DC = 0. SU21 = 35. w .
PRC1 = 0.03. Nd21 = 16. A232 = 8. Nd31 = 25.8. DI2 = 6.2 Closedloop supply chain (third. MINTPS = 0. SP2 = 55. B121 = 9. S22 = 500. S12 = 400. N312 = 14. W3 = 0. B221 = 3. RSR3 = 45. Nd11 = 20. N211 = 3. b2 = 0. Z32 = 12. A122 = 20. Z31 = 20. 2 = 0. © 2009 by Taylor & Francis Group. W2 = 1. B312 = 6. Y2 = 1 It is evident from the above solution that both of the production facilities are chosen for the network design. Nd32 = 20. Nd22 = 22.25 Upon application of the above data to the model. PRC2 = 0. = 0. B222 = 18. Also. RCYR3 = 2. LLC .04. B112 = 4. SP1 = 65.6. TB21 = 0.04. B311 = 15. RCRI3 = 5. SP3 = 60.0. A212 = 2. X2 = 0. 1 = 0.2. RCRI1 = 7. 2. S11 = 400.4. Z22 = 23. RCYR2 = 2.6. A312 = 9. Cd2 = 0. S1 = 150. RCYR1 = 1. and fifth models). S3 = 150.2.75. Z21 = 2. A121 = 15. b1 = 0.4. RSR 2 = 40. B111 = 8. Nd12 = 15.5.3. N111 = 3. W1 = 0. A221 = 5. A321 = 12. Cd3 = 0. RCRI2 = 4. RR = 951 (target = 750).9. TB22 = 0. Z23 = 13. N112 = 5. TA32 = 0.7. S2 = 150. DI1 = 4.5.500). TB12 = 0. X1 = 0.7. N222 = 1.5. the following optimal solution is obtained: TP = 3. A231 = 5.05. A331 = 3. Z33 = 12.945 (target = 2. b3 = 0. Cd1 = 0. B321 = 7. A222 = 18. DI3 = 5. TB11 = 0.Optimization of Transportation of Products New Products Remanufactured Products 167 New Products Remanufactured Products Demand Centers 1 and 2 Production Facilities 1 and 2 Consumers Used Products Collection Centers 1.5. RSR1 = 30. B212 = 3. A311 = 11. fourth. N211 = 8. B122 = 6. Y1 = 1. Z12 = 35. PRC3 = 0. NDIS = 74 (target = 50). B211 = 5.02. 71% of the net demand is satisfied by remanufactured products and the remaining 29% by newly manufactured products. and 3 Used Products FIGURE 8.09. S21 = 500.3. using LINGO (v4).
45) 2. production facilities. 8.1. and new) across a closedloop supply chain. identifies efficient production facilities.5. and demand centers. Linear physical programming is employed to address the following issues in one continuous phase. and section 8. determines the number of used products of each type to be picked for remanufacturing.4. Section 8. It is assumed that the inventory costs of a used product and a remanufactured product are 20% of the collection and remanufacturing costs. remanufactured.2 gives a numerical example.46) 3.5. Collection/retrieval cost: CCu SU iu u i (8. The revenue and cost criteria and system constraints considered in the model are: Costs: Class 1S (smaller is better) 1. Transportation cost = cost of transporting used products from collection centers to production facilities + cost of transporting remanufactured and new products from production facilities to demand centers: © 2009 by Taylor & Francis Group. and achieves transportation of the right quantities of products (used. in one continuous phase. LLC .5 Fourth Model (Linear Physical Programming) This model considers a generic closedloop supply chain consisting of collection centers.1 Model Formulation This linear physical programming model.1 presents the model formulation using the nomenclature given in section 8. respectively and that the inventory cost of a newly produced product is 25% of the production cost.168 Strategic Planning Models for Reverse and ClosedLoop Supply Chains 8. besides optimal transportation of products: selection of used products and evaluation of production facilities.5. Processing cost = disassembly cost of used products + remanufacturing cost of used products + new products’ production cost: DC i u v DTi Aiuv i u v CRv Bivw i v w CNPv N ivw (8.
49) 2. disposal cost depends on the percentage of recyclable content in the product and the disposal cost index (a number on a scale 0–10.50) 3.51) System constraints 1.52) 2. The number of used products sent to all production facilities from collection center u must be equal to the number of used products picked for remanufacturing at that collection center: Aiuv v Ziu (8.Optimization of Transportation of Products 169 TAuv i u v Aiuv TBvw i v w ( Bivw N ivw ) (8. Recycle revenue: {(SU iu i u Ziu ) RCRIi Wi PRCi } (8.53) © 2009 by Taylor & Francis Group. LLC . the higher the number. Disposal cost: Apart from the number of units disposed of and the cost to dispose of a product. Reuse revenue: {Ziu RSRi } i u (8. The demand at each center w must be met by either new or remanufactured products: ( Bivw v N ivw ) Ndiw w (8. the more difficult or expensive it is to dispose of the product): {(SU iu i u Ziu ) DIi Wi (1 PRCi )} Cdi (8.47) 4. New product sales revenue: SPi N ivw i v w (8.48) Revenues: Class 2S (larger is better) 1.
v is a factor that accounts for the unassignable (common) causes of variation at production facility v. The number of remanufactured products must be at most equal to the net demand (this is to avoid excess remanufacturing): Ziu i u i w Ndiw (8. over which there is no control: Bivw w u Aiuv v v (8.55) 5. The space occupied by new and remanufactured products at production facility v must at most be equal to the space available © 2009 by Taylor & Francis Group. The number of remanufactured products transported from production facility v to demand center w equals (number of used products fit for remanufacturing that are transported from collection center u to that production facility) multiplied by v.56) 6.57) 7. That is. there is no loss of products in the supply chain due to reasons other than common cause variations. The total number of used products of all types collected (before accounting for probability of breakage) at all collection centers must be at least equal to the net demand (this is to encourage the use of remanufactured products): SU iu i u i w Ndiw (8.Yv (8.170 Strategic Planning Models for Reverse and ClosedLoop Supply Chains 3. The space occupied by used products at production facility v must at most be equal to the space available for used products at that facility: x1 i u Aiu S1v . The number of used products of type i picked for remanufacturing at collection center u must be at most equal to the total number of used products fit for remanufacturing: Ziu SU iu (1 bi ) (8.58) 8. LLC .54) 4.
Bivw . CR1 = 10. DT1 = 10. Ziu 0.6.01. and three types of products are considered in the example (see figure 8.5.62) MINTPS (8.05.005. DC = 0. Nonnegativity constraints: Aiuv . LLC . RCRI2 = 4. SU33 = 15.1 v . SP3 = 95. RCYR 2 = 7.60) 10. v . TB22 = 0. RCYR3 = 105. CR 2 = 8. DI3 = 5.2 Yv 0 . Nd31 = 25. SP2 = 110. PRC1 = 0. DT3 = 9. TB11 = 0.001. TB12 = 0. 1 if selected (8.03. Nd11 = 20. TA21 = 0. TA12 = 0.8.61) 12. SU21 = 25. The data used to implement the linear physical programming model are: CCu = 0. two production facilities.2). W1 = 0. SU32 = 9. The space occupied by used products at collection center u must at most be equal to the space available for used products at that collection center: x1 i v aiuv Su (8. Cd2 = 0.009. u .009.Optimization of Transportation of Products 171 for new and remanufactured products at that production facility (assuming both new and remanufactured products occupy the same space): x 2 ( Bivw i w N ivw ) S2v Yv (8.005. RSR1 = 80. SP1 = 100. i (8. W3 = 0. RSR3 = 65. two demand centers. DT2 = 12. Cd1 = 0. Nd22 = 22.02. w . DI1 = 4.75. SU13 = 15. CNP2 = 65. SU31 = 17. TA32 = 0.003.01.9.65.59) 9. TA11 = 0. The ratio of throughput to supply of used products of a production facility must at least be equal to a preset value for the production facility to be considered efficient (this is valid only for remanufactured products): TPv / SFv Yv 11. SU23 = 15. PRC2 = 0. Cd3 = 0. RSR 2 = 80. RCYR1 = 5. PRC3 = 0.0. TA22 = 0. N ivw . Nd12 = 15. 8.003.002.004. TB21 = 0. DI2 = 6.63) Numerical Example Three collection centers. SU22 = 18. SU11 = 20.004. 0 if facility v not selected. SU12 = 25. RCRI3 = 6. © 2009 by Taylor & Francis Group. CNP1 = 55. Nd21 = 16. TA31 = 0. W2 = 1. RCRI1 = 7.
34 4. A132 = 0. 69% of the net demand is satisfied by newly manufactured products and the remaining 31% by remanufactured © 2009 by Taylor & Francis Group. A112 = 0. N211 = 0.34 1. A122 = 0. S12 = 400. A131 = 0. B321 = 1. N111 = 15. 1 = 0. S1 = 150.29 2. S22 = 500.02 Nd32 = 20.02 0.05 0. Z33 = 6. Z31 = 3. A121 = 0. B222 = 3.09 0. B211 = 8.13 0. b1 = 0. b3 = 0.7. A322 = 2. A212 = 4.25 The target values for the criteria are given in table 8.4 Output of LPP weight algorithm (fourth model) Criteria g1 g2 g3 g4 g5 g6 g7 Δwp2+ 0.02 0. N221 = 4. A211 = 11. B122 = 0.12 0. N112 = 11. X1 = 0. A222 = 3. B311 = 6. B322 = 0. A231 = 7. the following optimal solution is obtained: Z11 = 0.75. Z23 = 9. N312 = 15.25 — — — Δwp4+ 0. A221 = 8. B221 = 4. MINTPS = 0. N321 = 0. B112 = 0. A321 = 5. X2 = 0. Z13 = 0.85.02 0. A332 = 0.34 1. Y2 = 1 It is evident from the above solution that both production facilities are chosen for the network design. S21 = 500. 2 = 0.03 Δwp4– — — — — 0. B312 = 0.05 Δwp3– — — — — 0.04 Δwp5– — — — — 0. LLC . A331 = 0.3 Target values of criteria (fourth model) Criteria g1 g2 g3 g4 Criteria g5 g6 g7 tp1+ 1 5 2 2 tp1– 10 10 15 tp2+ 3 10 2. N212 = 0. N121 = 5.02 0.3. and table 8. Upon application of the above data to the model. N322 = 5.82 — — — Δwp2– — — — — 0. N222 = 5. A232 = 2.5 5 tp2– 15 15 17 tp3+ 5 13 5 7 tp3– 20 17 20 tp4+ 7 17 7 9 tp4– 25 19 25 tp5+ 9 20 10 13 tp5– 30 22 35 TABLE 8.2.29 2.03 — — — Δwp3+ 0.13 0. Z32 = 6.4.172 Strategic Planning Models for Reverse and ClosedLoop Supply Chains TABLE 8. B111 = 0.75 0.11 0.3 (target values are scaled by a factor of 10). Z12 = 0. A311 = 3. A312 = 0.05 0.7. B121 = 0. Also.20 0. A111 = 0. Y1 = 1. b2 = 0.4 shows the incremental weights obtained using the LPP weight algorithm [2]. N311 = 18.02 0. Z22 = 11. N122 = 4.17 0. using LINGO (v4).29 — — — Δwp5+ 3. B212 = 14. S2 = 150. Z21 = 15. S3 = 150. S11 = 400.
1 presents the model formulation using the nomenclature given in section 8. Maximize the total profit in the closedloop supply chain (TP). and achieves transportation of the right quantities of products (used. and section 8. 8.1. remanufactured.1 Model Formulation This fuzzy goal programming model. involves minimizing the positive deviation from the target value. 8. LLC .Optimization of Transportation of Products 173 products.6. and demand centers. Three goals are considered in the model: 1.2 gives a numerical example. demand for the first product type is completely satisfied by newly manufactured products of that type.64) TP TPL TP * TPL 0 if TP TPL © 2009 by Taylor & Francis Group. The membership functions for the three goals are 1 1 if TP TP * if TPL TP TP * (8. production facilities.6. 3.6. Fuzzy goal programming is employed to address the following issues in one continuous phase. Minimize the number of disposed items (NDIS). in one continuous phase. and new) across a closedloop supply chain. 2. determines the number of used products of each type to be picked for remanufacturing. whereas the third goal. respectively. In addition. The first two goals involve minimizing the negative deviation from the respective target values. identifies efficient production facilities. It is assumed that the inventory costs of a used product and a remanufactured product are 20% of the collection and remanufacturing costs. Section 8. Maximize the revenue from recycling (RR). and that the inventory cost of a newly produced product is 25% of the production cost.4.6 Fifth Model (Fuzzy Goal Programming) This model considers a generic closedloop supply chain consisting of collection centers. which has an environmentally benign character rather than a financial basis. besides optimal transportation of products: selection of used products and evaluation of production facilities.
starting with 1 and 2.5 1 ) (0. 0. According to the concept of Fibonacci numbers.174 Strategic Planning Models for Reverse and ClosedLoop Supply Chains where TP* is the aspiration level of total profit (TP) and TPL is the lower tolerance level of TP.17. 1 2 if RR if RRL if RR RR * RR RR * (8. LLC .69) © 2009 by Taylor & Francis Group.66) NDIU NDI NDIU NDI * 0 NDIU where NDI* is the aspiration level of number of disposed items (NDI) and NDIU is the upper tolerance level of NDI.65) RR RRL RR * RRL 0 RRL where RR* is the aspiration level of recycling revenue (RR) and RRL is the lower tolerance level of RR.67) The cost and revenue criteria and the system constraints considered in the model include: Revenues 1.33 2 ) (0. Reuse revenue: {Ziu RSRi } i u (8. The objective function for the weighted fuzzy goal programming model is as follows: Maximize V( ) (0. μ2. and 0. and μ 3 are 0.68) 2. and 1 3 if NDI if NDI * if NDI NDI * NDI NDIU (8. Recycle revenue: {(SU iu i u Ziu ) RCYIi Wi PRCi } (8.17 3 ) (8. the weight values for μ1.5.33.
Processing cost = disassembly cost of used products + remanufacturing cost of used products + new products production cost in the forward supply chain: DC i u v DTi Aiuv i u v CRv Bivw i v w CNPv N ivw (8. the higher the number. Inventory cost = cost of carrying used products inventory at the collection center + cost of carrying remanufactured products inventory at the production facility + cost of carrying newly manufactured products inventory at the production facility: (CCu /5) Aiuv i u v i v w {(CRv / 4 ) Bivw (CNPv / 4 ) N ivw } (8. Disposal cost: Apart from the number of units disposed of and the cost to dispose of a product. disposal cost depends on the percentage of recyclable content in the product and the disposal cost index (a number on a scale 0–10.Optimization of Transportation of Products 3. New product sales revenue: SPi N ivw i v w 175 (8.71) 2.73) 4. LLC . Collection/retrieval cost: CCu SU iu u i (8.74) 5. the more difficult or expensive it is to dispose of the product): © 2009 by Taylor & Francis Group.72) 3. Transportation costs = cost of transporting used products from collection centers to production facility + remanufactured and new products from production facilities to demand centers: TAuv i u v Aiuv TBvw i v w ( Bivw N ivw ) (8.70) Costs 1.
The demand at each center w must be met by either new or remanufactured products: ( Bivw v N ivw ) Ndiw w (8. The number of used products sent to all production facilities from a collection center u must be equal to the number of used products picked for remanufacturing at that collection center: Aiuv v Ziu (8. v is a factor that accounts for the unassignable (common) causes of variation at the production facility v: Bivw w u Aiuv v v (8. The total number of used products of all types collected (before accounting for probability of breakage) at all collection centers must be at least equal to the net demand (this is to encourage the use of remanufactured products): © 2009 by Taylor & Francis Group.176 Strategic Planning Models for Reverse and ClosedLoop Supply Chains {(SU iu i u Ziu ) DIi Wi (1 PRCi )} Cdi (8.75) System constraints 1. The total number of used products of type i picked for remanufacturing at collection center u must be at most equal to the total number of used products fit for remanufacturing: Ziu SU iu (1 bi ) (8.76) 2..e.77) 3. i.79) 5. LLC .78) 4. no loss of products in the supply chain due to reasons other than common cause variations. over which there is no control. The number of remanufactured products transported from a production facility v to a demand center w equals (number of used products fit for remanufacturing that are transported from collection center u to that production facility) multiplied by v v.
80) 6.82) 8. The number of remanufactured products must be at most equal to the net demand (this is to avoid excess remanufacturing): Ziu i u i w Ndi (8. The space occupied by used products at the collection center.84) 10.Yv (8.Optimization of Transportation of Products 177 SU iu i u i w Ndiw (8.83) 9. u.85) © 2009 by Taylor & Francis Group.81) 7. v. must at most be equal to the space available for used products at that facility: x1 i u Aiuv S1v . must at most be equal to the space available for used products at that collection center: x1 i v aiuv Su (8. The ratio of throughput to supply of used products of a production facility must at least be equal to a preset value for the production facility to be considered efficient (this is valid only for remanufactured products): TPv / SFv Yv MINTPS (8. The space occupied by new and remanufactured products at the production facility. must at most be equal to the space available for new and remanufactured products at that production facility (assuming both new and remanufactured products occupy the same space): x 2 ( Bivw i w N ivw ) S2v Yv (8. LLC . v. The space occupied by used products at the production facility.
B221 = 7. 86% of the net demand is satisfied by remanufactured products and the remaining 14% by newly manufactured products. NDI* = 60. μ1 = 1.2).4. SP2 = 53.7. Z22 = 17. RCYR2 = 2. Nd21 = 16. RCRI2 = 4. A121 = 15. Nd31 = 25. two production facilities. Nd32 = 20. u .86) 0 . whereas that of NDI is not (μ 3). SU2 = 35. 2 = 0. SP1 = 65. MINTPS = 0.5. TP* = 3. SU23 = 22. Also. RCRI3 = 5. S3 = 150.8. SU11 = 50.000. A312 = 9. A321 = 18. Yv 0. RSR3 = 45. A212 = 21. S22 = 500. N211 = 9. S2 = 150. both production facilities are chosen for the network design. SU21 = 35. TA22 = 0. TA11 = 0. TA32 = 0. TB22 = 0.7.02. S12 = 400. S1 = 150. w . PRC2 = 0. 0 if facility v not selected. A311 = 5.05. RCRI1 = 7. S21 = 500. B111 = 13. S11 = 400. 1 if selected (8. DI2 = 6. The data used to implement the fuzzy goal programming model are: CCu = 0. CR1 = 13. B112 = 15. TB21 = 0. TB11 = 0.01.75.01. N ivw .09.500.6.5.2 Numerical Example Three collection centers. RR* = 1. © 2009 by Taylor & Francis Group. TA21 = 0. X1 = 0.1.4. SU12 = 45. SP3 = 60. b2 = 0. B322 = 14. RRL = 800. B222 = 22.178 Strategic Planning Models for Reverse and ClosedLoop Supply Chains 11.03.000.0. Bivw . A131 = 20.6.1 v . μ 3 = 0. TB12 = 0. CNP1 = 60. RSR 2 = 40. NDI = 77. b3 = 0. LLC . two demand centers. Ziu 12. W3 = 0. W1 = 0. Y2 = 1. Z33 = 18. N111 = 7. b1 = 0.9. W2 = 1. TA31 = 0. Cd3 = 0. and three types of products are considered in the example (see figure 8.000.3. NDIU = 100. A222 = 17. DI3 = 5. CNP2 = 45. SU22 = 38.5.04. SU1 = 25. Z31 = 5.87) 8. TA12 = 0. Nd12 = 15. Cd2 = 0.5. DT3 = 9. RR = 1.25 Upon application of the above data to the model using LINGO (v4). PRC3 = 0. SU31 = 30. DC = 0. TPL = 2. B311 = 25. RCYR1 = 1. Nonnegativity constraints: Aiuv .2. v . RSR1 = 30. the following optimal solution is obtained: Z12 = 15. i (8.3. Z13 = 20. CR 2 = 10. TP = 3.5. SU33 = 28. Nd22 = 22. Nd11 = 20. B312 = 6. DI1 = 4.500. 1 = 0. Y1 = 1. Cd1 = 0.04.09. RCYR3 = 2. DT1 = 10. In addition.2. DT2 = 12. A332 = 20. μ2 = 1. A331 = 17. Z21 = 21.05.6. Z32 = 18. PRC1 = 0. X2 = 0.59 It should be noted that the achievements of TP and RR (μ1 and μ2) are at their maximum.
fourth. Linear physical programming: A new approach to multiple objective optimization. 2001. The first and second models are for a reverse supply chain and use linear integer programming and linear physical programming. and Akbulut. M. 2. Gupta.Optimization of Transportation of Products 179 8.. Quantitative models for reverse logistics: Lecture notes in economics and mathematical systems. Fleischmann. M. B. Messac..7 Conclusions In this chapter. linear physical programming. A. respectively. © 2009 by Taylor & Francis Group. The third. and fuzzy goal programming. five models are presented for achieving transportation of the right quantities of products (used. and fifth models are for a closedloop supply chain and employ goal programming. LLC . respectively. Transactions of Operational Research 8:39–59. Berlin: SpringerVerlag. remanufactured. References 1. 1996. S. and new) across a reverse or closedloop supply chain while satisfying certain constraints.
The officials of each program painstakingly approach many homes in the respective city with questions regarding how convenient the program is to the public and how the program can be improved. evaluating the marketing strategy of a reverse/closedloop supply chain program is equivalent to evaluating how well the strategy is driving the public to participate in the program.e. The program could be for a closedloop supply chain and the decision maker could be uninterested in considering interdependencies among evaluation criteria. The program could be for a closedloop supply chain and the decision maker could be interested in considering interdependencies among evaluation criteria. 2. Although the drivers for governments and companies to implement these programs and evaluate the programs’ marketing strategies are environmental consciousness and profitability. the higher the taxes the consumers will have to pay. see [1] and [2]) are conducted in numerous cities around the world in order to assess the level of participation of the public in the respective reverse/closedloop supply chain programs.9 Evaluation of Marketing Strategies 9. Studies (for example. the drivers for the public to participate in the programs are numerous and often conflicting with each other (for example. Hence. high regularity of collection and low tax levied on the consumers are conflicting drivers here). absence of a closed loop.. 3. The various scenarios for evaluating the marketing strategy of a reverse/ closedloop supply chain program could differ as follows: 1. 181 © 2009 by Taylor & Francis Group. which in turn is shouldered by the marketing strategy of that program. i.1 The Issue The success of a reverse/closedloop supply chain program is heavily dependent on the level of public participation (in the program). the more regularly a reverse/closedloop supply chain program offers to collect used products from consumers. The program could be exclusively for reverse supply chain operations. LLC . respectively.
quality function deployment (QFD). Incentives to buyers of reprocessed goods (IB) 12.2. and section 9.2.2.3 shows the second model. Then. Good locations of demand centers where reprocessed goods are sold (LR) 11. section 9. a list of drivers for the public to participate in a reverse supply chain program are identified. and method of total preferences. in section 9. Incentives for disposal of used products (ID) 6. This chapter is organized as follows: Section 9. Effectiveness of collection methods (EC) 7. Knowledge of the drivers of implementation of the reverse supply chain program (KD) 2. The first model addresses scenario 1 and employs fuzzy logic and technique for order preference by similarity to ideal solution (TOPSIS). Convenience for disposal of used products at collection centers (CD) 5. fuzzy logic and TOPSIS are employed to evaluate the marketing strategy of a reverse supply chain program with respect to the drivers identified in section 9.1. various models to address the above scenarios are presented. Awareness of the reverse supply chain program being implemented (AR) 3.2. and analytic network process. The third model is for scenario 3 and uses fuzzy logic. Design of special methods for abusers of the reverse supply chain program (AB) 10.1.5 gives some conclusions. The second model is for scenario 2 and employs fuzzy logic. using a numerical example. Simplicity of the reverse supply chain program (SR) 4. Cooperation of the program organizers with the local government (CL) © 2009 by Taylor & Francis Group. LLC .4 presents the third model.2 presents the first model. 9. 9.2 First Model (Fuzzy Logic and TOPSIS) In section 9. Regularity of collection of used products (RC) 9.2.182 Strategic Planning Models for Reverse and ClosedLoop Supply Chains In this chapter. extent analysis method. Information supplied about used products being collected (IU) 8.1 Drivers of Public Participation The following is a fairly exhaustive list of selfexplanatory drivers for the public to participate in a reverse supply chain: 1. section 9.
0) (0. 0. these linguistic weights are converted into triangular fuzzy numbers (TFNs.1 illustrates the linguistic weights.Evaluation of Marketing Strategies 9. 1 Low Medium Low Very high Low High Low Medium Medium Very high High Medium Rep.1.7) (0.” and “medium.2 Conversion table for weights of drivers (first model) Linguistic weight Very high High Medium Low Very low TFN (0.2. Using fuzzy logic. depending on which driver greatly motivates them to participate. 0.7. 0. For example.1 Linguistic weights of drivers of public participation (first model) Driver KD AR SR CD ID EC IU RC AB LR IB CL Rep. the average weight of the driver ID (see tables 9. 0.1.2 shows one of the many ways for such a conversion) and then averaged to form another TFN called the average weight. 0. 0. and so on. 1. they give linguistic weights like “very high.3) © 2009 by Taylor & Francis Group.7.3. 0.5.0. 2 Medium High Low Very high High Medium Low Low Low High High High Rep.5.5) (0.1 and 9.9. table 9. 0. 0. 3 High Very high Very high Medium Low Very high High Low Medium High Medium Low TABLE 9.2 Methodology 183 Suppose there are three representatives of a community to weigh the drivers of public participation. Because it is difficult for them to assign numerical weights. which driver is not so important for them. LLC .2) is Low High 3 Low TABLE 9.” Table 9.3.” “low.9) (0.
43. 8.5.87) (0.3 Average weights of drivers of public participation (first model) Driver KD AR SR CD ID EC IU RC AB LR IB CL Sum Average weight (0.7) (4. 0.63) (0.23.57. the normalized weight of ID is (0.43. Suppose that one is interested in evaluating marketing strategies of two different reverse supply chain programs.27 . The ratio of the average weight of each driver to the sum of the average weights of all the drivers gives the corresponding normalized weight.4).9 0. 10) (see table 9.3 0.33. 8.57.93) © 2009 by Taylor & Francis Group. 0.4) as (4. 0. 0. 0.83) (0. 0. 0.7.6) because the representatives unanimously rate it as good with respect to ID (the TFN for the linguistic rating good is (5.06. 0. 6.1 0. The sum of the average weights of all the drivers is calculated using equation (4.67) (0.5 is used for conversion of linguistic ratings into TFNs. 0. 0.3.9) (0.17. Table 9. 0.23.63) 3 (see table 9. 0. TABLE 9.3).27.5. 0. For example.63) (0. 10) (see table 9. Assuming (for arithmetic simplicity) that the representatives come to a consensus about the rating of each marketing strategy with respect to each driver.7) as (0.7 . 0. 0.27.3 0. 0.5.43. 0.93) which is simplified using equation (4. 6. 0. Now that the weights (normalized) of the drivers of public participation are ready.5. 0.17. the decision matrix shown in table 9. 0. 0.5 .63) (4.37.93).93) (0.5 0. 0. 7.7. 0. 7.33.57) (0. the two marketing strategies are linguistically rated by the three representatives with respect to each driver.03.87) (0.5.43. 0. 0.3.5 0.5)). 6.77. = (0.63. LLC . 0. 0.3.63) (0. 0.77.7) (0. 8.6 (S1 and S2 are the marketing strategies) is formed.23.23. 3 3 0.43.73.1 0. For example.43.184 which is Strategic Planning Models for Reverse and ClosedLoop Supply Chains 0. 0. 0.15) (see table 9. the rating (TFN) of marketing strategy S2 with respect to driver ID is (5.
20) (0. 7. 0. 0. 10. 0. 8) (7. 5) (0. 0. 10) (5. 10. 10) (2.6 Decision matrix (first model) Driver KD AR SR CD ID EC IU RC AB LR IB CL S1 (7.21) (0.4 Normalized weights of drivers of public participation (first model) Driver KD AR SR CD ID EC IU RC AB LR IB CL Normalized weight (0.16) (0. 10) (2. 10) (5. 10) (5.03.02.10. 7. 0. 5. 10) (7.11. 0. 8) (5. 7. 0. 0. 0. 5) (0.03.03.06. 0.16) 185 TABLE 9.Evaluation of Marketing Strategies TABLE 9.11.02. 3) (2. 0. 0.20) (0. 0. 10. 3) (7. 10) (2. 10. 0. 0.06. 10.07. 0. 0. 0. 5.13) (0. 8) (2. 7. 10. 0. 7. 0. 8) (0. 0. 10) (5.07. 0. 0.06.10. 0.09.07.16) (0.15) (0. 5.05. 10) (0.5 Conversion for ratings of marketing strategies (first model) Linguistic rating Very good Good Fair Poor Very poor TFN (7. 10) © 2009 by Taylor & Francis Group. 3.06. 8) (0.15) (0. 3) (7.06. 10) (5.05.03. 10) (5. 0. 8) (7. 3. 5) (0. 3) S2 (0.06. LLC . 10. 7.15) (0.03.07. 10) (0. 0. 5. 0.22) (0. 3) TABLE 9.20) (0. 5. 7. 8) (2. 5. 10) (2. 0. 3. 5. 0.
51.71.03. respectively.01.81. for the corresponding driver.57. 0.e.18.96. 0. Table 9. Step 3: Determine the ideal and negativeideal solutions.49) as follows: (5. 2.81.7.. Table 9. 0. 0.50) (0.00. 1.07. 0.15) (see table 9.58.00. 1. 1.41) (0.35. 1. 7 .7 shows the normalized decision matrix formed by applying equation (4.8 has a maximum rating and a minimum rating. 0.186 Strategic Planning Models for Reverse and ClosedLoop Supply Chains Now one is ready to perform the six steps in TOPSIS. 1. 0. i.7) is calculated using equation (4.86) (0.6).00.16.60) S2 (0. 0.21. and (4.43) (0.41) (0. 0.00. (0. Step 1: Construct the normalized decision matrix.57 . 1. 0. 0. 4. 1.10)2 (0. 0.00) (0. the normalized fuzzy rating of strategy S2 with respect to driver ID (see table 9.43) (0.e.18. 1. 1. 0. Step 2: Construct the weighted normalized decision matrix. Each row in the decision matrix shown in table 9.1. 0. 0.04. the weighted normalized fuzzy rating of strategy S2 with respect to driver ID. 2.71. i.50.16) (0.45. For example.67. LLC . 0.10) (0.43) (0.86) (0.7) are used to perform the basic operations in the calculation of r52 above. 7 . 1.6) is used for the multiplication.43) (0.00.6 and the normalized decision matrix in table 9.00. 0. For example. 0. 0.7).7 Normalized decision matrix (first model) Driver KD AR SR CD ID EC IU RC AB LR IB CL S1 (0. This is constructed using the normalized weights of the drivers listed in table 9. 0.63.00.16) Note that equations (4.43) (0.58. 0. 1. 0.89.67. 2.10)2 (5.49) (0. 1.00. 0.35. 1.00) © 2009 by Taylor & Francis Group.16) (0. 0.00.83) (0. with the normalized fuzzy rating of S2 with respect to ID.86.57. 1. They are the ideal and negativeideal solutions.. it is assumed here that the rating with the highest most promising TABLE 9.55.35.00. (0.00. 0.48. 1.43) (0.00. 1. is calculated by multiplying the normalized weight of ID.8 shows the weighted normalized decision matrix.4). 0. i.16) (see table 9. 1. For arithmetic simplicity.29.49) (0.. 0. 1.82. 0. 2.71) (0.35.67.00. (4.10.43) (0.37) (0.00.16. 0.10) r52 = (7 . 0.71.39.5.17) (see table 9.16.e.6). 1.49) on each element of table 9.71. Equation (4.39. (0.35.8). 0.83) (0.
0.00. 0.01.00.00.01.59) (0.17) is the TFN with the highest second parameter and (0. 0. 0. 0.215 + 0. 0. 0..00.01.00. 0.08.08.10) S2 (0.02.22) (0.17).04.06. 0. 0. the relative closeness coefficient is calculated for each marketing strategy (see table 9.06.44) (0.215) to the sum (i.10).03. 0.10).692) is much higher than that of strategy S1 (i. For example.e.00.. 0. 0.28) (0.07.28) (0. 0. Using equation (4.01. 0. the relative closeness coefficient (i.02. 0.28) (0. 0. 0.e.09. 0.00.06) (0. it is evident that S2 is much better than S1. 0.01. 0.Evaluation of Marketing Strategies TABLE 9. This is because.05. 0.17) (0. 0. LLC . 0.01.04. 0. 0.02.53). 0. 0. 0. with respect to driver ID. 0.07) (0.8) and the ideal solution (obtained in step 3) for each driver of public participation. in the row for that driver.03.22) (0.9).14) (0. 0. The separation distances (see table 9. 0. 0.9) for each marketing strategy are calculated using equations (4.54).33) 187 quantity (second parameter in the TFN) is the maximum. 0.05.03. 0.01.. 0. 0. 0.52) and (4.31) (0.8). 0.097 = 0. 0.04.17) is the TFN with the lowest second parameter. 0.01.00..00.17) (0. 0. 0.9) is calculated using the weighted normalized fuzzy ratings of S2 (see table 9. 0. 0.38) (0. 0.08.05.312) of its negative and positive separation distances (see table 9.04. Step 6: Rank the preference order. 0.23) (0.30) (0.10) is the ratio of the strategy’s negative separation distance (i. 0.05. Step 5: Calculate the relative closeness coefficient. 0.07. 0. 0.e. (0.. 0.31) (0.17) and the minimum rating is (0.05. 0.09) (0.03.05.37) (0. 0.01. 0.44) (0.8 Weighted normalized decision matrix (first model) Driver KD AR SR CD ID EC IU RC AB LR IB CL S1 (0.00.01.19) (0.09. the maximum rating is (0.03.692) for strategy S2 (see table 9. Because the relative closeness coefficient of strategy S2 (i.05. 0. and the rating with the lowest most promising quantity is the minimum.01. Step 4: Calculate the separation distances. 0. 0.11.00. 0. 0.e.02. © 2009 by Taylor & Francis Group. the positive separation distance for strategy S2 (see table 9. For example.04.30) (0. For example (see table 9.05.e.01.308) (see table 9.01. 0.
a model to evaluate the marketing strategy planned by a closedloop supply chain program is presented. Program simplicity (PS) (reflects the relative ease with which the public can participate in the program) 2.2 gives a numerical example illustrating the model’s application.215 TABLE 9.097 0.1 Performance Aspects and Enablers The following performance aspects of the marketing strategies are considered in the employment of quality function deployment: 1. and section 9. and Method of Total Preferences) In this section.311 0.096 Negative distance S– 0.215 0.9 Separation distances of marketing strategies (first model) Marketing strategy S1 S2 Positive distance S* 0.3.3. innovation. quality function deployment (QFD). Section 9.10 Relative closeness coefficients of marketing strategies (first model) Marketing strategy S1 S2 Relative closeness coefficient 0. LLC .188 Strategic Planning Models for Reverse and ClosedLoop Supply Chains TABLE 9. Public knowledge (PK) (reflects the awareness among the public about the program being implemented) The following enablers are considered for the abovelisted performance aspects: © 2009 by Taylor & Francis Group.692 9. and method of total preferences are employed in the approach.3. Quality Function Deployment. Fuzzy logic. and improvement capability compared to its peers) 3. Incentives (re