Textbooks

Mathematical Modeling
Williams, Gueret a. o.
Tony Hürlimann
Department of Informatics
University of Fribourg
CH – 1700 Fribourg (Switzerland)
tony.huerlimann@unifr.ch
February 1, 2021
Copyright © 2021
All rights reserved. Department of Informatics
University of Fribourg
CH-1700 Fribourg Switzerland
Email: tony.huerlimann@unifr.ch
WWW: http://www.virtual-optima.com
i
ii
C ONTENTS
1 Introduction 3
1.1 Model and Related Concepts . . . . . . . . . . . . . . . . . . 4
1.2 The Model Building Process . . . . . . . . . . . . . . . . . . . 6
1.3 The Modeling Language LPL . . . . . . . . . . . . . . . . . . 10
1.4 Mathematical Notation . . . . . . . . . . . . . . . . . . . . . 16
1.4.1 Set-Names versus Index-Names . . . . . . . . . . . . . 16
1.4.2 Logic-mathematical Notation . . . . . . . . . . . . . . 17
1.4.3 Logical Operators . . . . . . . . . . . . . . . . . . . . 18
2 Problems from William's Book 19

2.1 Food Manufacture I (will01) . . . . . . . . . . . . . . . . . . 20
2.2 Food Manufacture II (will02) . . . . . . . . . . . . . . . . . . 21
2.3 Factory planning I (will03) . . . . . . . . . . . . . . . . . . . 23
2.4 Factory planning II (will04) . . . . . . . . . . . . . . . . . . . 24
2.5 Manpower Planning (will05) . . . . . . . . . . . . . . . . . . 26
2.6 Refinery Optimization I (will06) . . . . . . . . . . . . . . . . 28
2.7 Refinery Optimization II (will06a) . . . . . . . . . . . . . . . 31
2.8 Mining (will07) . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.9 Farm Planning (will08) . . . . . . . . . . . . . . . . . . . . . 34
2.10 Economic Planning (will09) . . . . . . . . . . . . . . . . . . . 36
2.11 Decentralization (will10) . . . . . . . . . . . . . . . . . . . . 38
2.12 Curve Fitting (linear) (will11) . . . . . . . . . . . . . . . . . . 40
2.13 Curve Fitting (quadratic) (will11a) . . . . . . . . . . . . . . . 43
2.14 Logical Design (will12) . . . . . . . . . . . . . . . . . . . . . 44
2.15 Market Sharing I (will13) . . . . . . . . . . . . . . . . . . . . 46
2.16 Market Sharing II (will13a) . . . . . . . . . . . . . . . . . . . 48
2.17 Opencast Mining I (will14) . . . . . . . . . . . . . . . . . . . 50
2.18 Opencast Mining II (will14a) . . . . . . . . . . . . . . . . . . 52
iii
2.19 Opencast Mining III (will14b) . . . . . . . . . . . . . . . . . . 54
2.20 Tariff Rates (Power Generation) (will15) . . . . . . . . . . . . 56
2.21 Hydro power (will16) . . . . . . . . . . . . . . . . . . . . . . 58
2.22 T3-dimensional Noughts and Crosses (will17) . . . . . . . . . 60
2.23 Optimizing a Constraint (will18) . . . . . . . . . . . . . . . . 62
2.24 Distribution I (will19) . . . . . . . . . . . . . . . . . . . . . . 64
2.25 Distribution II (will19a) . . . . . . . . . . . . . . . . . . . . . 66
2.26 Depot Location (Distribution 2) (will20) . . . . . . . . . . . . 68
2.27 Agricultural Pricing I (will21) . . . . . . . . . . . . . . . . . . 70
2.28 Agricultural Pricing II (will21a) . . . . . . . . . . . . . . . . . 73
2.29 Efficiency Analysis (EA) (will22) . . . . . . . . . . . . . . . . 74
2.30 Milk Collection (will23) . . . . . . . . . . . . . . . . . . . . . 76
2.31 Yield Management (will24) . . . . . . . . . . . . . . . . . . . 78
2.32 Car Rental I (will25) . . . . . . . . . . . . . . . . . . . . . . . 80
2.33 Car Rental II (will26) . . . . . . . . . . . . . . . . . . . . . . 83
2.34 Lost Baggage Distribution (will27) . . . . . . . . . . . . . . . 86
2.35 Protein Folding (will28) . . . . . . . . . . . . . . . . . . . . . 88
2.36 Protein Comparison (will29) . . . . . . . . . . . . . . . . . . 90
2.37 Product Mix I (willi005) . . . . . . . . . . . . . . . . . . . . . 91
2.38 Product Mix II (willi005a) . . . . . . . . . . . . . . . . . . . . 92
2.39 Vegetable Oil Blending (willi008) . . . . . . . . . . . . . . . . 93
2.40 Multi-Plant Planning (willi055) . . . . . . . . . . . . . . . . . 94
2.41 Transportation (willi082) . . . . . . . . . . . . . . . . . . . . 95
2.42 Production Planning (willi085) . . . . . . . . . . . . . . . . . 96
2.43 Minimum Cost Flow (willi090) . . . . . . . . . . . . . . . . . 97
2.44 Shortest Path Problem (willi093) . . . . . . . . . . . . . . . . 98
2.45 Maximum Flow (willi094) . . . . . . . . . . . . . . . . . . . . 99
2.46 Critical Path (willi095) . . . . . . . . . . . . . . . . . . . . . 100
2.47 A Small QP (willi142) . . . . . . . . . . . . . . . . . . . . . . 101
2.48 LP-relaxtion versus IP-solution (willi155) . . . . . . . . . . . 102
2.49 LP-relaxtion versus 0-1-IP-solution (willi155z) . . . . . . . . . 105
3 Problems from the Gueret Book 107

3.1 Animal food production (guer04-2) . . . . . . . . . . . . . . . 108
3.2 Production of alloys (guer04-3) . . . . . . . . . . . . . . . . . 109
3.3 Refinery (guer04-4) . . . . . . . . . . . . . . . . . . . . . . . 110
3.4 Cane Sugar Production (guer04-5) . . . . . . . . . . . . . . . 111
3.5 Opencast Mining (guer04-6) . . . . . . . . . . . . . . . . . . 112
3.6 Production of Electricity (guer04-7) . . . . . . . . . . . . . . . 113
3.7 Stadium Construction (guer05-2) . . . . . . . . . . . . . . . . 114
3.8 Flow-shop Scheduling (guer05-3) . . . . . . . . . . . . . . . . 117
3.9 Job Shop: non-generic version (guer05-4a) . . . . . . . . . . . 119
3.10 Job Shop: generic version (guer05-4b) . . . . . . . . . . . . . 120
3.11 Sequencing jobs on a bottleneck machine (guer05-5) . . . . . . 121
3.12 Paint production (guer05-6) . . . . . . . . . . . . . . . . . . . 122
3.13 Assembly line balancing (guer05-7) . . . . . . . . . . . . . . . 123
iv
3.14 Planning the production of bicycles (guer06-2) . . . . . . . . . 124
3.15 Production of drinking glasses (guer06-3) . . . . . . . . . . . 125
3.16 Material Requirement Planning (guer06-4) . . . . . . . . . . . 126
3.17 Production of elec. components (guer06-5) . . . . . . . . . . . 127
3.18 Production of fiberglass (guer06-6) . . . . . . . . . . . . . . . 128
3.19 Assign prod batches to machines (guer06-7) . . . . . . . . . . 129
3.20 Wagon load balancing (guer07-2) . . . . . . . . . . . . . . . . 130
3.21 Barge loading (guer07-3) . . . . . . . . . . . . . . . . . . . . 131
3.22 Tank loading (guer07-4) . . . . . . . . . . . . . . . . . . . . . 132
3.23 Backing up files (guer07-5) . . . . . . . . . . . . . . . . . . . 133
3.24 Cutting sheet metal (guer07-6) . . . . . . . . . . . . . . . . . 134
3.25 Cutting steel bars for desk legs (guer07-7) . . . . . . . . . . . 135
3.26 Car rental (guer08-2) . . . . . . . . . . . . . . . . . . . . . . 136
3.27 Choosing the mode of transport (guer08-3) . . . . . . . . . . . 137
3.28 Depot location (guer08-4) . . . . . . . . . . . . . . . . . . . . 138
3.29 Heating oil delivery (guer08-5) . . . . . . . . . . . . . . . . . 139
3.30 Combining diff modes of transport (guer08-6) . . . . . . . . . 140
3.31 Fleet planning for vans (guer08-7) . . . . . . . . . . . . . . . 141
3.32 Flight connections at a hub (guer09-2) . . . . . . . . . . . . . 142
3.33 Composing flight crews (guer09-3) . . . . . . . . . . . . . . . 143
3.34 Scheduling flight langings (guer09-4) . . . . . . . . . . . . . . 144
3.35 Airline hub location (guer09-5) . . . . . . . . . . . . . . . . . 145
3.36 Planning a flight tour (guer09-6) . . . . . . . . . . . . . . . . 146
3.37 Network reliability (guer10-2) . . . . . . . . . . . . . . . . . 147
3.38 Dimensioning of a mobile phone network (guer10-3) . . . . . 148
3.39 Routing tlephone calls (guer10-4) . . . . . . . . . . . . . . . . 149
3.40 Construction of a cabled networkt (guer10-5) . . . . . . . . . 150
3.41 Scheduling of telecomm. via satellite (guer10-6) . . . . . . . . 151
3.42 Localisation of GSM transmitters (guer10-7) . . . . . . . . . . 152
3.43 Choice of loans (guer11-2) . . . . . . . . . . . . . . . . . . . 153
3.44 Publicity campaign (guer11-3) . . . . . . . . . . . . . . . . . 154
3.45 Portfolio selection (guer11-4) . . . . . . . . . . . . . . . . . . 155
3.46 Finacing an early retirment scheme (guer11-5) . . . . . . . . . 156
3.47 Family budget (guer11-6) . . . . . . . . . . . . . . . . . . . . 157
3.48 Choice of expansion projects (guer11-7) . . . . . . . . . . . . 158
3.49 Mean variance portfolio (guer11-8) . . . . . . . . . . . . . . . 159
3.50 Assigning personnel to machines (guer12-2) . . . . . . . . . . 160
3.51 Scheduling nurses (guer12-3) . . . . . . . . . . . . . . . . . . 161
3.52 Establishing a college timetable (guer12-4) . . . . . . . . . . . 162
3.53 Exam scheduling (guer12-5) . . . . . . . . . . . . . . . . . . 163
3.54 Personnel assignment (guer12-6) . . . . . . . . . . . . . . . . 164
3.55 Personnel at a construction site (guer12-7) . . . . . . . . . . . 166
3.56 Water conveyance mamangement (guer13-2) . . . . . . . . . . 167
3.57 CCTV surveillance (guer13-3) . . . . . . . . . . . . . . . . . . 168
3.58 Rigging elections (guer13-4) . . . . . . . . . . . . . . . . . . 169
v
3.59 Gritting roads (guer13-5) . . . . . . . . . . . . . . . . . . . . 170
3.60 Location of income tax offices (guer13-6) . . . . . . . . . . . . 173
3.61 Efficiency of hospitals (guer13-7) . . . . . . . . . . . . . . . . 174
3.62 MasterMind (guer14-2) . . . . . . . . . . . . . . . . . . . . . 175
3.63 Probleme de logique sur le marathon de Paris (guer14-3) . . . 176
3.64 Chercheurs dans un Congres (guer14-4) . . . . . . . . . . . . 177
3.65 Grille et jetons (guer14-5) . . . . . . . . . . . . . . . . . . . . 178
3.66 Partage de tonneaux entre des neveux (guer14-6) . . . . . . . 179
3.67 Le probleme des n reines (guer14-7) . . . . . . . . . . . . . . 180
4 Various Textbooks 181

4.1 Forestry Production Planning (forestry) . . . . . . . . . . . . 181
4.2 Wyndor Glass Company (hill03-1-1) . . . . . . . . . . . . . . 184
4.3 Mary’s Radiation Therapy (hill03-4-1) . . . . . . . . . . . . . 185
4.4 Kibbutzim Crop Allocation (hill03-4-2) . . . . . . . . . . . . . 186
4.5 Nori/Leets Air Pollution (hill03-4-3) . . . . . . . . . . . . . . 187
4.6 Save-It Company (hill03-4-4) . . . . . . . . . . . . . . . . . . 188
4.7 Save-It Company (hill03-4-4b) . . . . . . . . . . . . . . . . . 189
4.8 Union Airways Personnel (hill03-4-5) . . . . . . . . . . . . . . 190
4.9 Distribution Unlimited (hill03-4-6) . . . . . . . . . . . . . . . 191
4.10 Wyndor Glass (hill07-1-1) . . . . . . . . . . . . . . . . . . . . 192
4.11 Upper Bound (hill07-3-1) . . . . . . . . . . . . . . . . . . . . 193
4.12 Goal Programming (hill07-5-1) . . . . . . . . . . . . . . . . . 194
4.13 P-T Company (hill08-1-1) . . . . . . . . . . . . . . . . . . . . 195
4.14 Northern Airplane (hill08-1-2) . . . . . . . . . . . . . . . . . 196
4.15 Metro Water (hill08-1-3) . . . . . . . . . . . . . . . . . . . . . 197
4.16 Job Shop Company (hill08-3-1) . . . . . . . . . . . . . . . . . 198
4.17 Better Products Company (hill08-3-2a) . . . . . . . . . . . . . 199
4.18 Better Products Company (hill08-3-2b) . . . . . . . . . . . . . 200
4.19 Shortest Path (hill09-3-1) . . . . . . . . . . . . . . . . . . . . 201
4.20 Maximum Flow (hill09-5-1) . . . . . . . . . . . . . . . . . . . 202
4.21 Minimum Cost (hill09-6-1) . . . . . . . . . . . . . . . . . . . 203
4.22 Critical Path (hill10-1-1) . . . . . . . . . . . . . . . . . . . . . 204
4.23 Crashing (hill10-5-1) . . . . . . . . . . . . . . . . . . . . . . . 205
4.24 Calif. Manufacturing (hill12-1-1) . . . . . . . . . . . . . . . . 206
4.25 Production Rates (hill12-4-1) . . . . . . . . . . . . . . . . . . 207
4.26 TV Spots (hill12-4-2a) . . . . . . . . . . . . . . . . . . . . . . 208
4.27 TV Spots (hill12-4-2b) . . . . . . . . . . . . . . . . . . . . . . 209
4.28 Crew Assignments (hill12-4-3) . . . . . . . . . . . . . . . . . 210
4.29 Nonlinear constraint (hill13-2-1) . . . . . . . . . . . . . . . . 211
4.30 Nonlinear Objective (hill13-2-2) . . . . . . . . . . . . . . . . . 212
4.31 Nonlinear Objective 2 (hill13-2-3) . . . . . . . . . . . . . . . . 213
4.32 Convex Programming (hill13-9-1) . . . . . . . . . . . . . . . 214
4.33 Odds And Even (hill14-1-1) . . . . . . . . . . . . . . . . . . . 215
4.34 Political Variaton 1 (hill14-2-1) . . . . . . . . . . . . . . . . . 216
4.35 Political Variation 2 (hill14-2-2) . . . . . . . . . . . . . . . . . 217
vi
4.36 Political Variation 3 (hill14-2-3) . . . . . . . . . . . . . . . . . 218
4.37 School Board Scheduling (hillC3-1) . . . . . . . . . . . . . . . 219
4.38 Manufacturing (hillP3-1-3) . . . . . . . . . . . . . . . . . . . 220
4.39 TV Manufacturing (hillP3-1-4) . . . . . . . . . . . . . . . . . 221
4.40 Resource PQ (hillP3-1-5) . . . . . . . . . . . . . . . . . . . . 222
4.41 Insurance (hillP3-1-6) . . . . . . . . . . . . . . . . . . . . . . 223
4.42 Auto Spare Parts (hillP3-1-7) . . . . . . . . . . . . . . . . . . 224
4.43 Resource QRS (hillP3-2-1) . . . . . . . . . . . . . . . . . . . . 225
4.44 Invest Venture (hillP3-2-3) . . . . . . . . . . . . . . . . . . . 226
4.45 Investment (hillP3-4-10) . . . . . . . . . . . . . . . . . . . . . 227
4.46 Investor ABCD (hillP3-4-11) . . . . . . . . . . . . . . . . . . 228
4.47 Alloy Blending (hillP3-4-12) . . . . . . . . . . . . . . . . . . . 229
4.48 Computer Fac Oper Assignment (hillP3-4-13) . . . . . . . . . 230
4.49 Warehouse Storage (hillP3-4-14) . . . . . . . . . . . . . . . . 231
4.50 Paper Manufacturing (hillP3-4-15) . . . . . . . . . . . . . . . 232
4.51 Steak-Potato Diet (hillP3-4-6) . . . . . . . . . . . . . . . . . . 233
4.52 Pig Feed Blending (hillP3-4-7) . . . . . . . . . . . . . . . . . 234
4.53 Plant Production Planning (hillP3-4-8) . . . . . . . . . . . . . 235
4.54 Cargo Plane Planning (hillP3-4-9) . . . . . . . . . . . . . . . . 236
4.55 MidWest Grain Elevator (midwest) . . . . . . . . . . . . . . . 237
4.56 Fertilizer (murty2-1) . . . . . . . . . . . . . . . . . . . . . . . 242
4.57 Gasoline Blending (murty2-2) . . . . . . . . . . . . . . . . . . 243
4.58 Diet Problem (murty2-3) . . . . . . . . . . . . . . . . . . . . 244
4.59 Transportation (murty2-4) . . . . . . . . . . . . . . . . . . . . 245
4.60 Marriage Problem (murty2-5) . . . . . . . . . . . . . . . . . . 246
4.61 Multi-Period Planning (murty2-6) . . . . . . . . . . . . . . . 247
4.62 Breck and Dapper (shapiro1-1) . . . . . . . . . . . . . . . . . 248
4.63 DowPont Chemical (shapiro1-2) . . . . . . . . . . . . . . . . 249
4.64 Portfolio Selection (shapiro1-3) . . . . . . . . . . . . . . . . . 250
4.65 Portfolio Selection (shapiro1-3b) . . . . . . . . . . . . . . . . 251
4.66 Transportation (shapiro1-4) . . . . . . . . . . . . . . . . . . . 252
4.67 Multi-Period Scheduling (shapiro1-5) . . . . . . . . . . . . . 253
4.68 Giapetto Wood Carving (winst3-1-1) . . . . . . . . . . . . . . 255
4.69 Giapetto Wood Carving II (winst3-1-1s) . . . . . . . . . . . . 256
4.70 Dorian Advertising (winst3-2-2) . . . . . . . . . . . . . . . . 257
4.71 Dorian Advertising (winst3-2-2s) . . . . . . . . . . . . . . . . 258
4.72 Auto Company (winst3-3-3) . . . . . . . . . . . . . . . . . . 259
4.73 Auto Company (winst3-3-3s) . . . . . . . . . . . . . . . . . . 260
4.74 Auto Company (infeasible) (winst3-3-4) . . . . . . . . . . . . 261
4.75 Unbounded Problem (winst3-3-5) . . . . . . . . . . . . . . . 262
4.76 Diet Problem (winst3-4-6) . . . . . . . . . . . . . . . . . . . . 263
4.77 Diet Problem (winst3-4-6s) . . . . . . . . . . . . . . . . . . . 264
4.78 Star Oil (winst3-6-8) . . . . . . . . . . . . . . . . . . . . . . . 265
4.79 Semicond Eletronics (winst3-7-9) . . . . . . . . . . . . . . . . 266
4.80 Rylon Corporation (winst3-9-11) . . . . . . . . . . . . . . . . 268
vii
4.81 Sailco (winst3-10-12) . . . . . . . . . . . . . . . . . . . . . . 270
4.82 Farmer Jones (winstP3-1-1) . . . . . . . . . . . . . . . . . . . 271
4.83 Farmer Jones (winstP3-1-1s) . . . . . . . . . . . . . . . . . . 272
4.84 Truck Corporation (winstP3-1-4) . . . . . . . . . . . . . . . . 273
4.85 Product-Mix (Truck Co) (winstP3-1-4s) . . . . . . . . . . . . . 274
4.86 Leary Chemical (winstP3-2-3) . . . . . . . . . . . . . . . . . . 275
4.87 Chemical Product Processes (winstP3-2-3s) . . . . . . . . . . 276
4.88 Furniture Corporation (winstP3-2-5) . . . . . . . . . . . . . . 277
4.89 Product-Mix (Furniture Corporation) (winstP3-2-5s) . . . . . . 278
4.90 Money Manager (winstP3-3-10s) . . . . . . . . . . . . . . . . 279
4.91 Momiss River Pollution (winstP3-4-1) . . . . . . . . . . . . . 280
4.92 US Lab (winstP3-4-2) . . . . . . . . . . . . . . . . . . . . . . 281
4.93 Diet Problem (winstP3-4-3) . . . . . . . . . . . . . . . . . . . 282
4.94 Gold Mining (winstP3-4-4) . . . . . . . . . . . . . . . . . . . 283
4.95 Capital Budget (winstP3-6-2) . . . . . . . . . . . . . . . . . . 284
4.96 Blending Candy (winstP3-8-1) . . . . . . . . . . . . . . . . . 285
4.97 Blending Oranges (winstP3-8-2) . . . . . . . . . . . . . . . . 286
4.98 Blending Portfolio (winstP3-8-3) . . . . . . . . . . . . . . . . 287
4.99 Blending Investments (winstP3-8-4) . . . . . . . . . . . . . . 288
4.100Blending Oils (winstP3-8-5) . . . . . . . . . . . . . . . . . . . 289
4.101Blending Oils (winstP3-8-5b) . . . . . . . . . . . . . . . . . . 290
4.102Blending fertilizer (winstP3-8-6) . . . . . . . . . . . . . . . . 291
4.103Blending chemicals (winstP3-8-7) . . . . . . . . . . . . . . . . 292
4.104Highland TV and Radio (winstP3-8-8) . . . . . . . . . . . . . 293
4.105Production Process (Sunco Oil) (winstP3-9-1) . . . . . . . . . 294
4.106Production Process (winstP3-9-2) . . . . . . . . . . . . . . . . 296
4.107Production Process (Rylon Corp.) (winstP3-9-3) . . . . . . . . 297
4.108Production Process (winstP3-9-5) . . . . . . . . . . . . . . . . 299
4.109Production Process (Daisy Drug) (winstP3-9-6) . . . . . . . . 300
4.110Production Process (Lizzies Dairy) (winstP3-9-7) . . . . . . . . 301
4.111Production Process (Lizzies Dairy) (winstP3-9-7b) . . . . . . . 303
4.112Product-Mix (Bloomington Breweries) (winstR3-01) . . . . . . 305
4.113Product-mix (Farmer Jones Cake) (winstR3-02) . . . . . . . . 306
4.114Investment (winstR3-03) . . . . . . . . . . . . . . . . . . . . 307
4.115Process Oil (Sunco Oil) (winstR3-04) . . . . . . . . . . . . . . 308
4.116Investment (Finco) (winstR3-05) . . . . . . . . . . . . . . . . 309
4.117Blending Steel (winstR3-06) . . . . . . . . . . . . . . . . . . . 310
4.118Production Planning (winstR3-07) . . . . . . . . . . . . . . . 312
4.119Production Planning with Diet (winstR3-08) . . . . . . . . . . 313
4.120Process Scheduling (winstR3-09) . . . . . . . . . . . . . . . . 314
4.121Product-mix (Carco Advertising) (winstR3-10) . . . . . . . . . 316
4.122Process Oil (winstR3-11) . . . . . . . . . . . . . . . . . . . . 317
4.123Telephone Survey (winstR3-12) . . . . . . . . . . . . . . . . . 319
4.124Feed Blending (winstR3-13) . . . . . . . . . . . . . . . . . . . 320
4.125Feed Blending (winstR3-14) . . . . . . . . . . . . . . . . . . . 321
viii
4.126Feed Blending (winstR3-14b) . . . . . . . . . . . . . . . . . . 322
4.127Production Planning (winstR3-15) . . . . . . . . . . . . . . . 323
4.128Product Mix (winstR3-17) . . . . . . . . . . . . . . . . . . . . 324
4.129Bus-ville SchoolDistricts (winstR3-18) . . . . . . . . . . . . . 325
4.130Brady (winstR3-19) . . . . . . . . . . . . . . . . . . . . . . . 327
4.131Canadian Parks Commission (winstR3-20) . . . . . . . . . . . 328
4.132Chandler Enterprises (winstR3-21) . . . . . . . . . . . . . . . 330
4.133Alden Enterprices (winstR3-22) . . . . . . . . . . . . . . . . . 332
4.134Kiriakis (winstR3-23) . . . . . . . . . . . . . . . . . . . . . . 334
4.135Multi-Period Planning (winstR3-40) . . . . . . . . . . . . . . 336
4.136Production Planning (winstR3-40b) . . . . . . . . . . . . . . . 337
4.137Multi-Period Finance Planning (winstR3-41) . . . . . . . . . . 338
4.138Transporation: Waste disposal (winstR3-42) . . . . . . . . . . 339
4.139Product-mix with blending (winstR3-45) . . . . . . . . . . . . 341
4.140Review Prob (Priceler) (winstR3-46) . . . . . . . . . . . . . . 342
4.141EJ-Korvair DeptStore (winstR3-53) . . . . . . . . . . . . . . . 344
5 Models from the GAMS Library 347

5.1 Linear Quadratic Control Problem (abel) . . . . . . . . . . . . 347
5.2 Agreste Farm Level Model of NE Brazil (agreste) . . . . . . . 349
5.3 Ajax Paper Company Production Schedule (ajax) . . . . . . . 355
5.4 Sample Problem (AMPL) (ampl) . . . . . . . . . . . . . . . . 356
5.5 Bid Evaluation (bid) . . . . . . . . . . . . . . . . . . . . . . . 357
5.6 Bid Evaluation (bid1) . . . . . . . . . . . . . . . . . . . . . . 358
5.7 Blending Problem I (blend) . . . . . . . . . . . . . . . . . . . 360
5.8 Hanging Chain COPS 2.0 3 (chain) . . . . . . . . . . . . . . . 361
5.9 Chance Constraint Feed Mix Problem (chance) . . . . . . . . . 362
5.10 Organic Fertilizer Use in Intensive Farming (china) . . . . . . 363
5.11 Financial Optimization: Financial Engineering (cmo) . . . . . 372
5.12 Peacefully Coexisting Armies of Queens (coex) . . . . . . . . 376
5.13 Peacefully Coexisting Armies of Queens - tight (coexx) . . . . 377
5.14 Alcuin’s River Crossing (cross) . . . . . . . . . . . . . . . . . 378
5.15 3-dimensional Noughts and Crosses (cube) . . . . . . . . . . 379
5.16 Simple Farm Level Model (demo1) . . . . . . . . . . . . . . . 381
5.17 Non-transitive Dice Design (dice) . . . . . . . . . . . . . . . . 384
5.18 Stigler’s Nutrition model (diet1) . . . . . . . . . . . . . . . . 385
5.19 Fertilizer Production (egypt) . . . . . . . . . . . . . . . . . . 387
5.20 House Plan Design (house) . . . . . . . . . . . . . . . . . . . 389
5.21 Tanglewood Chair Manufacturing (tangle) . . . . . . . . . . . 390
5.22 A Travel Optimization Problem (travel) . . . . . . . . . . . . 391
5.23 A Transportation Problem (trnsport) . . . . . . . . . . . . . . 392
6 Billonnet 393
6.1 A MIQP problem (Bill018) . . . . . . . . . . . . . . . . . . . 393
6.2 Localisation Optimal (Bill023) . . . . . . . . . . . . . . . . . . 394
6.3 Optimal Portefolio (Bill026) . . . . . . . . . . . . . . . . . . . 396
ix
6.4 Optimal Positioning on the Market (Bill028) . . . . . . . . . . 398
6.5 An iQP Problem: Clustering (Bill035) . . . . . . . . . . . . . . 402
6.6 Fraction Objective (Bill046) . . . . . . . . . . . . . . . . . . . 404
6.7 Optimal Portefolio (Bill053) . . . . . . . . . . . . . . . . . . . 407
6.8 Piecewise Linear Function (Bill227) . . . . . . . . . . . . . . . 409
6.9 A Cost Flow Problem (bill320) . . . . . . . . . . . . . . . . . 412
7 Miscellaneous 419
7.1 A simple blending problem (alloy) . . . . . . . . . . . . . . . 419
7.2 Assign Consulters (assign1) . . . . . . . . . . . . . . . . . . . 420
7.3 Agriculture Production (crop) . . . . . . . . . . . . . . . . . 421
7.4 Import and Export Goods (export) . . . . . . . . . . . . . . . 422
7.5 A Small Portfolio Model (lin193) . . . . . . . . . . . . . . . . 423
7.6 Exposures in Media Vehicles (media) . . . . . . . . . . . . . . 424
7.7 Mexican Steel Industry (mexican) . . . . . . . . . . . . . . . . 425
7.8 A Production Planning Model (omp) . . . . . . . . . . . . . . 427
7.9 One Machine Scheduling (ordon) . . . . . . . . . . . . . . . . 430
7.10 A Transport Model Fragment (pam) . . . . . . . . . . . . . . 431
7.11 The Beer Game (beergame) . . . . . . . . . . . . . . . . . . . 432
7.12 Production Planning (aggrplan) . . . . . . . . . . . . . . . . 436
7.13 Production Planning (planning) . . . . . . . . . . . . . . . . 438
7.14 Multiperiod Production (prod) . . . . . . . . . . . . . . . . . 439
7.15 Production and Planning (prodplan) . . . . . . . . . . . . . . 442
7.16 Production and Shipment Planning (prodship) . . . . . . . . . 444
7.17 A Production Model (produ) . . . . . . . . . . . . . . . . . . 448
7.18 Multiperodic Production Planning (product) . . . . . . . . . . 450
7.19 Multiperiod Production Scheduling (schedule) . . . . . . . . 454
7.20 Multiperiod Production Scheduling (schedule) . . . . . . . . 455
7.21 Tanglewood Chair Manufacturing (tangle) . . . . . . . . . . . 456
7.22 A Travel Optimization Problem (travel) . . . . . . . . . . . . 457
7.23 Multi-period Blending Problem (xpress) . . . . . . . . . . . . 458
x
L IST OF F IGURES
1.1 The Feasible Space . . . . . . . . . . . . . . . . . . . . . . . . 13
2.1 Linear Solution . . . . . . . . . . . . . . . . . . . . . . . . . . 41

2.2 Quadratic Solution . . . . . . . . . . . . . . . . . . . . . . . . 42
2.3 The nor-operation, and the nor gate . . . . . . . . . . . . . . . 45
2.4 The and/xor circuit . . . . . . . . . . . . . . . . . . . . . . . . 45
2.5 The Folding Structure . . . . . . . . . . . . . . . . . . . . . . 89
2.6 The Feasible Space and the Objective . . . . . . . . . . . . . . 101
2.7 The IP- and the LP-solution . . . . . . . . . . . . . . . . . . . 102
2.8 The IP- and the LP-solution are equal . . . . . . . . . . . . . . 104
2.9 The IP- and the LP-solution . . . . . . . . . . . . . . . . . . . 105
3.1 The village streets ([28] p.324) . . . . . . . . . . . . . . . . . . 170

3.2 The minimal traversing . . . . . . . . . . . . . . . . . . . . . 172
6.1 A Small Example Graph . . . . . . . . . . . . . . . . . . . . . 414

6.2 A piecewise linear function of arc (7, 8) . . . . . . . . . . . . . 416
6.3 A piecewise linear function . . . . . . . . . . . . . . . . . . . 417
7.1 The Supply Chain . . . . . . . . . . . . . . . . . . . . . . . . 432

7.2 Local s-S strategy of ordering . . . . . . . . . . . . . . . . . . 434
7.3 Demand estimated strategy of ordering . . . . . . . . . . . . . 435
7.4 Product Flow Description . . . . . . . . . . . . . . . . . . . . 444
xi
L IST OF TABLES
1.1 Logical Operators . . . . . . . . . . . . . . . . . . . . . . . . 18
3.1 Data for the Stadium Construction . . . . . . . . . . . . . . . 114
4.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265
4.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 266
4.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 266
4.4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
4.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 284
4.6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 310
4.7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312
4.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313
4.9 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 316
4.10 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 317
4.11 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319
4.12 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 325
4.13 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 325
4.14 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 328
4.15 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 330
4.16 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 332
4.17 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 332
4.18 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 334
4.19 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 338
4.20 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339
4.21 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 342
4.22 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344
6.1 Data and solution for the model . . . . . . . . . . . . . . . . . 415
xiii
P REFACE
This text is a collection of many mathematical models formulated in the LPL

modeling language and executable directly though the Internet. The models
are from many sources and are of varying quality. A wide range of applica-
tions is covered, but there is no claim of a systematical order. The collection
has grown over time.
After a short introduction to LPL in Chapter 1, the next Chapter 2 contains

all 29 “case studies” of the famous textbook of P. Williams (see [17]. For each
model only a short description of the problem is given followed by the formu-
lation of the model in LPL format. The name of the model is will?? where ??
means the number (01–29) of the case study in the book. Several small models
spread over the book are also implemented. Their name is willi??? and ???
means the page of the book.
Chapter 3 contains all 65 models from the French book of Guéret Christelle,
Prins Christian and Sevaux Marc (see [5]. The book has been translated to
English by Dash Optimization and is available as [28] (it can be downloaded
from the Internet and is free. The last chapter 14 of [5] has not been trans-
lated, so I give the original French version of the models. Some models that
I used in my master courses have additional comments and explanation, but
most of them only contain the bare bone LPL implementation that can also be
executed directly though the Internet.
Chapter 4 contains small model from exercises in several known operations

research textbooks. There are models from the textbook of Hillier [12], Murty
[19], Shapiro [25], and many from Winston [31]. If possible, a small problem
description was copied from the book.
Chapter 5 implements several models from the GAMS Library.

1
Finally, the last Chapter implements miscellaneous models collected over the
time.
All models are part of an automatic testing of LPL. When a new LPL version
is deployed all these models must run correctly. Selected models can also be
used for self study of modeling concepts or in classes.
The reader does not need to install any software to run and solve the models
compiled in this text. Only an Internet browser and an Internet connection is
needed. A link appears in the title of each model. Click on it, and you are on
the Internet site where the model is prompted in a text box. You can modify
it interactively. Clicking the button “Send” on the web page, sends the model
to the LPL-server, runs it and returns the result after it has been solved in a
new browser page. If you know well the syntax of LPL, you can even write an
entirely new model and present it to the LPL-server as well. Just click “Send”
to solve it.
However, the solution time is limited to 60secs, and large ans complex models
cannot be solved this way.
The book was compiled using the typesetting system LATEX. However, the
documentation of all models was written in LPL’s own “literate documenta-
tion system”. LPL contains a documentation tool (similar to javadoc for Java
programs) to automatically translate a documented model into a full LATEX
code. Additional information for this tool can be found in the Reference Man-
ual of LPL [29].
Availability of the LPL-Modeling-System
The mathematical modeling and all case studies in this book are based on
the software LPL. It is my own contribution to the field of computer-based
mathematical modeling.
Currently, there is an implementation for Windows. A Linux console version
is available – write me an email.
LPL – the software – together with documentation containing the Reference
Manual, several papers and examples can be downloaded at:
www.virtual-optima.com
The paper’s home page can be found at any of the two locations:
textbooks home page (lpl.virtual-optima.com/lpl/mainothers.html)

textbooks home page (lpl.unifr.ch/lpl/mainothers.html)
C HAPTER 1
I NTRODUCTION
“The essence of mathematics is not to make simple things com-

plicated, but to make complicated things simple.”
— S. Gudder
“Ich betrachte diese einleitenden theoretischen Betrachtungen als

die schwierigsten, weil wir die ganze Zeit das Gehirn bemühen
müssen. Nachher können wir an seiner Stelle die Mathematik
verwenden!”
— A. Eddington
This chapter introduces briefly the concept of a model and related notions.
Then it gives a rudimentary overview of the model building process in the
way it is used for the examples in the following chapters. Finally, since all
models are formulated in the executable modeling language LPL, the basic
elements of this language are briefly presented.
3
CHAPTER 1. INTRODUCTION 4
1.1. Model and Related Concepts

Solving a small case studies is like solving a mini research problem. Both
require a certain insight to see the problem from the right angle, and then ex-
plore that idea until one finds a solution. Problem solving is fun! But even
more exciting than solving a particular instance of a problem by any ad-hoc
method or by trial and error is to find a general method and representation to
solve it. In this book, we are not so much interested in finding a solution to a
specific problem in a precise way. We want to write a general statement that
describes it. If we can transform a problem this way, we can give the transfor-
mation to existing software (called a solver in this book) to solve it generally.
It is the process of building and composing a mathematical model. However,
there is not as much “operative” mathematics involved as the reader may
suppose. No complicated derivation or algebraic transformation is used or
has to be learnt. The reader only needs to have an idea of the basic formal
mathematical and logical notation and how to read and use it.
So what is a model? The word model has many meanings: “I want to be a
model father for my children”, “She works as a photo model”, “My son plays
with the railway model”, “This is an exact model of the Titanic”, “The car was
an old model”, “She is an artist, she can model a head with clay”, Rutherford
came up with a model of the atom which had electrons orbiting around a
nucleus, an orrery is a model of our solar system. In operations research (OR)1
we use mathematical models to find the shortest round-trip of a journey or the
“best” way to use resources in production.
We use models everywhere, even if we do not recognize them for what they
are. Many of those we are using every day, are becoming routine instruments:
The most elementary budget involves a simplified representation of the future
– this is a model. Our daily way to the work office makes implicit estimations
about the traffic. All our decisions are based on implicit or explicit, on more
or less correct assumptions and models. In fact, we could say that modeling
itself is a ubiquitous human activity; it is one of the fundamental ways in
which human understand and modify the world.
While a model is a simplified representation of a problem or a situation, model-
ing is the process of building and refining it for greater insight and improved
decision-making. Although the final goal of the modeling process is to obtain
a model to work with, the modeling process itself is often as important, be-
cause its discovery path teaches us more about the problem than the resulting
model. Mostly, we do not use the settled model as a static piece of knowledge.
What makes it more interesting is the fact that it can be used as a vehicle to
1
Operations research is a branch of applied mathematics and decision theory, which
uses mathematical models, statistics, and algorithms to aid in decision-making. It
is most often used to analyze complex real-world systems, typically with the goal
of improving or optimizing performance.
5 CHAPTER 1. INTRODUCTION
further research and queries varying the initial questions and problems.
Models, and especially mathematical models, are capable of giving deeper

insight to understand better how “things tick”. The consequences are that we
can make better decisions and save costs or get better returns. Better models
are also a vehicle to communication and to accumulate and store our knowledge
and to give us an analytic framework to explore the world, testing alternatives
and variants.
Science is modeling – and mathematics as a research of patterns is at the very

core of (better) modeling. Why mathematics? Mathematics is about building
concepts, making new links, searching for new patterns – this is inherently
linked to abstraction. Abstracting is a process of determining the similarities and
properties of different objects that make it possible to address them as one. It
is also a process of finding analogous structures, leaving out the unessential in
different contexts – this is very much linked to modeling. Of course, it is often
difficult to classify, to structure – but it is at the core of all scientific activities.
Modeling is a powerful means to solve problems, however, it is all but easy.

So, how can modeling be learnt? Problems, in practice, do not come neatly
packaged and expressed in mathematical notation; they turn up in messy,
confusing ways, often expressed, if at all, in somebody else’s terminology.
Whereas problem solving can generally be approached by more or less well-
defined techniques, there is seldom such order in the problem posing mode.
Therefore, a modeler needs to learn a number of skills. He must have a good
grasp of the system or the situation which he is trying to model; he has to
choose the appropriate mathematical tools to represent the problem formally;
he must use software tools to solve the models; and finally, he should be able
to communicate the solutions and the results to an audience, who is not nec-
essarily skilled in mathematics.
It is often said that modeling skills can only be acquired in a process of learning-
by-doing; like learning to ride a bike can only be achieved by getting on
the saddle. It is true that the “case study approach” is most helpful, and
many university courses in (mathematical) modeling use this approach. But
it is also true – once some basic skills have been acquired – that theoretical
knowledge about the mechanics of bicycles can deepen our understanding
and enlarge our faculty to ride it. This is even more important for model-
ing industrial processes. It is not enough to exercise these skills, one should
also acquire methodologies and the theoretical background of modeling. In
applied mathematics, more time than it is currently spent, should be given
to the study of discovery, expression and formulation of the problem, initially in
non-mathematical terms. So, the novice first needs to be an observer and then,
very quickly, a doer. Modeling is not learnt only by watching others building
models, but also by being actively and personally involved in the modeling
process.
The use of computer modeling tools to simulate, visualize, and analyze math-
ematical structures has spread steadily. In the eighties, the “micro-computer”
– a word which disappeared as quickly as it emerged – was the archetype of a
whole generation of self-made quick and messy models. Everyone produced
their own simulation tool, mathematical toolbox, etc. This phenomenon has
now almost vanished. We have powerful packages such as Mathematica,
Maple, Axiom, the NAG-library, Matlab, R, Python packages, Julia packages,
and powerful MIP solvers to mention just a few, for solving complex prob-
lems. But the modeling process still needs more than easy-to-use solving
tools: It also needs tools to integrate different tasks: such as data manipulation
tools, representation tools for the model structure, viewing tools to represent
the data and structure in different ways, reporting tools, etc. Some tasks can
be done by different software packages: data are preferably manipulated in
databases and spreadsheets, and are best viewed by different graphic tools.
This brings us to the heart of this book: Modeling is still an art with tools
and methods built in an ad-hoc way for a specific context and for a particu-
lar type of model. While I believe that modeling will always be an art in the
sense of a creative process, I also believe that we can build general computer-
based modeling tools that can be used not only to help to find the solution to a
mathematical model, but also to support the modeling process itself. The ex-
traordinary advances in science, on one hand, and in computing performance,
on the other hand, leave a cruel gap that can be bridged by general and effi-
cient modeling management systems and mathematical modeling languages.
As already mentioned, all models presented in this book are formulated with
the modeling language LPL. After the problem is formulated in this language
specification, our task is finished. The rest – solving a model instance – will
be done by the software.
1.2. The Model Building Process

The immediate approach to solving problems is by trial and error, as men-
tioned. An alternative and much more general approach is by creating a
model – in our cases a mathematical model. This has the advantage that the
designed model can then be applied to a whole class of similar problems and
is not only applicable to the specific problem example at hand.
Building a mathematical model means to conceptualize, to invent, to contrive

a scheme to turn a “specification” into a operational set of (mathematical)
formulas or other concepts. Finding a good model for a problem is not a trivial
task. It is a sloppy process; false steps and dead-ends are common. The idea
that there exists a rational, error-free path from the problem description to a
useful model is quite unrealistic. Model building is also non-deterministic:
In two situations, you can come up with very different models for the same
problem. There does not exist a unique way to design a model and no tool is
right for everything. In short, modeling is a heuristic process.

This being said, there are nonetheless rules and guidelines for designing good
models. The most important ideas are: (1) “identify the essential objects and
their attributes” leaving away what is accidental. (2) “find a consistent ab-
straction”. Abstraction is the ability to engage with the concept and ignore
the irrelevant details, it is the faculty to extract common features from specific
examples.
One of the first books that tried to describe this process of modeling was G.
Polya’s How to Solve It (1957) [13]. He decomposes the process into 4 different
steps:
1. Understanding the problem: You need to understand the problem:
What are the data? What is unknown? What is the condition? Is it
sufficient, redundant, irrelevant, contradictory? Draw a figure! Intro-
duce a suitable notation! Write it down! If you are stuck, start again!
2. Devising a plan: Do you know a related or a similar problem? Find
the connection between the data and the unknowns! Maybe you need
to design auxiliary problems and intermediary steps! Look at the un-
knowns! Go back to the definitions! Decompose the problem and solve
the parts! Did you use all data? Did you use all conditions?
3. Carrying out the plan: Write it down step by step! Can you show that
each step is correct?
4. Looking back: Examine the solution! Can you check the result or the
argument? Can you derive the result differently?
Other methodologies exist. However, most of them decompose the modeling
live cycle into simpler steps, as in:
• understand the problem
• describe the objective in words
• describe each constraint in words
• define the decision variables
• express the objective in terms of the decision variables
• express each constraint in terms of the decision variables
• formalize stepwise and check the result
• iterate and do not stop after the first draft.
The most important rule is not to be fixed on a single approach: if you are
stuck, free your mind and try a different technique.
A small example follows to illustrate the main points. Suppose you have to
solve the following problem:
“Knowing that the length plus the width of a rectangle is 27, while the area
plus the difference of the length and the width is 183, find the length and
width!”
One of the key steps in mathematical modeling is to find out what we are
looking for. In this example, it can easily be stated: we are looking for the
length and the width of the rectangle. So, we immediately introduce two sym-
bols for them: L and W. Passing from the words to the symbols is a crucial
and necessary step and should not be underestimated when building a math-
ematical model. One of the difficulties in modeling often stems from the fact,
that we cannot easily identify the “objects” in the problem. If we have iden-
tified all essential objects, we have reached an important milestone. The rest
often follows straight away.
Now, having L and W, we can easily derive from the first part of the statement
that their sum is 27. Therefore, we write it down as: L + W = 27. Write it down
on a piece of paper! Do not do it mentally only! The paper is patient, you can throw
it away and begin again from scratch! The second part of the statement says, that
the area plus the difference is 183. The area is L·W and the difference is L−W.
Therefore we have L · W + L − W = 183. Did we use all information? Yes! Is
something missing? No! Well, probably not. Can we now derive, what L and
W are? Yes! So, let’s see. We now have:
L + W = 27
L · W + L − W = 183
That’s the model! We are done. To find a solution, we do some algebraic

transformations which is not our modeling business anymore. We leave it to
a program to find the correct values for L and W. (The correct and unique
solution is L = 15 and W = 12, as you can easily verify.)
The problems here and their well polished models in the next chapters give
the impression that modeling is easy and straightforward. It is not! The main
goal of this text is to give you the complete solution on the modeling process,
indeed. But I also try to give you a possible step by step procedure on how
one could model them. Of course, this itself is an elaborated recipe on how to
proceed. In reality, it was a “painful” process that was anything but a straight
path. This has to be kept in mind, when you are going through these steps
of modeling. Nevertheless, it can be helpful to learn techniques and methods
on how to build models, on how to attack a problem. To stimulate further
your capability of modeling, each “case study” contains a number of ques-
tions that you can solve yourself. The proposed answer, also given in this
text, is not necessarily the “right” approach, but gives you an indication how
I did it. Maybe this also has an educational effect. Anyway, the reader can
only learn something from the examples, when he goes through all steps and
questions himself. It is better to find a wrong answer (which can be corrected
afterwards) than to look up the solution right away.
The approach, I propose for the case studies, is as follows: First, the prob-
lem is presented by a short description, this is marked by a title Problem.
The information given in this part should be enough to get started with the
modeling steps.
Next follows a polished procedure of the model building steps in Modeling
Steps. Normally, we begin by identifying the data, that means, the sets and the
numerical data used in the model. The critical step is the introduction of the
variables. The reader should be convinced that he understood this important
part in the model building process. Often the introduction of the variables
determine the rest of the modeling process. In the next step, we develop the
conditions (the constraints) that hold about the variables. Finally, the goal, the
objective function, is put forward.
The presentation of the model building process is normally given by enumer-
ating the steps to suggest to the reader a logical and sequential order he can
follow. Again the reader, nevertheless, should be aware, that this sequential
ordering is only an already polished outline of the real model building pro-
cess. It is itself already an abstraction of the labyrinthine process of modeling.
However, I did not find a better way of presenting this process. I also wanted
to break it down to the main points. To describe the “real process” – of what
is really going on in the brain when modeling – is also not what is actually
needed to reconstruct and comprehend the model building process. It would
be a lengthy explanation of all the intricate and nebulous thoughts that one
goes through while modeling. Modeling can be learnt, but not by following
the connotations and tortuous brainwork that one goes through actually. It
can only be learnt using case studies, if the presentation itself is a clear and
unequivocal synopsis of the main steps. If the classical outline proposed in
this book (data – variables – constraints) is the right one, is an open question.
Surely, other methods exist.
After the presentation of the modeling steps, a short paragraph follows to lay-
out the solution of the problem. This solution is normally found by running
the model on the computer. Sometimes hints are given to check the result and
to verify its correctness.
Finally, each case study ends with some questions. The reader can test his
comprehension and explore model variants. The questions should stimulate
him to investigate various topics.
1.3. The Modeling Language LPL

Mathematical notation is a powerful language to specify formal knowledge.
The LPL modeling language is nothing else than a computer executable form
for a part of that great “language” – created over the last 3000 years. A model
written in LPL can directly be parsed and transformed into various other rep-
resentations by a program called the LPL interpreter. The interpreter can also
communicate with various “solvers” – other programs that solve the problem
numerically, it can retrieve the solution and report it in various forms. It can
also read (write) large numbers of data from (to) databases. LPL is great in
prototyping large and complicated models used in a professional context. It
is not only used for small problems. The design principle behind LPL always
was: “Make it as concise as possible, but not more concise.”
The core of the modeling system LPL is the modeling language. Each model
has the following basic syntactic structure:
model <modelname> ;
<statement list>
end
model and end are keywords and <modelname> is a user-defined identi-

fier. Basically, the statements can be classified into declarations and instructions.
They can be in any order, but the instructions are executed from top to bot-
tom. With this respect the LPL language is like a programming language. The
main declarations are: set, parameter, variable, constraint. The main instructions
are: solve, if, while, for, and the assignment. The instructions are similar to in-
structions of a classical programming language.
Each of the declarations begins with a keyword and ends with a semicolon:
set .... ; --declare sets, lists
parameter .... ; --declare tables of data
variable .... ; --declare tables of unknowns
constraint .... ; --declare tables of formulas
A set declaration is similar to the definition of a mathematical set. For ex-

ample, I, J = {1 . . . 3} (where J is just another name of I) is a set with three
elements in the classical mathematical notation. In LPL, this is formulated
as follows (we prefer lowercase letters to designate sets, since in LPL we use
these names also as indices, if no confusion occurs):
set i, j := [1..3];
There is one important difference between mathematical sets and sets in LPL:
Sets (also called index-sets) in LPL are always ordered. These sets can then be
used to build vectors, matrices, and higher dimensional tables of parameters
(the data), variables, or constraints. For example, the two-dimensional table
ai,j with i, j ∈ I (that is, a 3 × 3 data matrix)

 
1 2 3
ai,j = 4 5 6
7 8 9
can be declared in LPL as follows (the single entries are listed in lexicograph-
ical ordering, row by row from left to right, the commas are optional):
parameter a{i,j} := [1 2 3 , 4 5 6 , 7 8 9];
The main difference between the mathematical notation and LPL’s notation
is that in LPL we do not need to distinguish between the set names and the
index names. Of course, we could also write:
set I; J; -- set name uppercase
parameter a{i in I,j in J}; -- index names lowercase
to make a difference between set name and index name, but this is not nec-
essary in most contexts. A further difference, as already mentioned, between
mathematical sets and sets defined in LPL is that mathematical sets are un-
ordered, while in LPL they are always ordered. However, this difference only
becomes important when we are referencing, for example, the elements by
their relative or absolute position, as in expression i − 1 (an expression that
returns the previous element of element i).
A declaration of a vector of unknown quantities (called variables) is written
in mathematical notation as xj with j ∈ J. In LPL we write it as follows:
variable x{j};
If the variables are integer or binary we write:

integer variable x{j} [0..100];
binary variable y{i};
The expression [0..100] means that the integer variables x{j} are bound
to the range [0 . . . 100]. Binary variable can only be zero or one. This means:
yi ∈ {0, 1} with i ∈ I.
Constraints are defined in the same way. A constraint vector:2
2 P
The formula uses a very common mathematical symbol ( ), which is a shortcut for
a summation. Hence,
X
xj with J = {1 . . . n} is the same as x1 + x2 + . . . + xn
j∈J
where n is a positive integer. It is supposed that the reader is familiar with this
notation. In the context of this book and if it is clear to which set the index j is
X
xj + yi ≤ 10 for all i ∈ I
j∈J
The constraint is written in LPL as follows:

constraint D{i}: sum{j} x[j] + y[i] <= 10;
In LPL, each constraint has a name (here D) and can be indexed like a variable
or a parameter. If the context allows it, the indexes within the expressions
can even be dropped (but this is a matter of style). So the constraint can also
compactly be written as follows:
constraint D{i}: sum{j} x + y <= 10;
The solve statement is identical to the constraint, except that it begins with
the keyword maximize or minimize. If there is no objective function and
we are only looking for a feasible solution then in LPL we can just use the
keyword solve.
maximize obj: sum{j} x;
minimize obj: sum{j} x;
solve;
A Simple Example
These basic syntax elements allows us to write complete mathematical models

in LPL. This is illustrated by a simple example:
Suppose your company produces laptops and printers. A laptop takes 6 hours
of work and generates a revenue of 300= C, while a printer takes 2 hours of
work and generates a revenue of 200= C. Furthermore, the space in our labo-
ratory is limited to 350m2 and each product takes 5m2 of space. We want to
find the “best” production mix, that means, how many of each can be man-
ufactured in a week, if the total hours of work is limited to 300 hours per
week?
referring, we also often use the shorter notation
X X
xj instead of xj with J = {1 . . . n}
j j∈J
In concordance with the LPL syntax, we often do not introduce the symbol for the
set J, we just use a index name j. So we use the notation
j ∈ {1 . . . n} instead of j ∈ J , J = {1 . . . n}
6x+
2y=
300
C
5x
+5y
=3
50
A D x
Figure 1.1: The Feasible Space
To find the optimal production mix, we introduce two symbols x and y for the
(unknown) quantities for the two products. Then the production is limited to
6x + 2y ≤ 300 hours, and to 5x + 5y ≤ 350 m2 . This is best displayed by
a diagram (see Figure 1.1). The points (x, y) within the quadrangle ABCD
are the only points that do not violate the two requirements. The point C
= (40, 30) is particularly interesting, because it defines the production mix
with the highest revenue, that means, 300x + 200y – the revenue of the mix
– cannot be larger without violating either the time or the space constraint.
Mathematically we say, that point C maximizes the expression 300x + 200y.
We can formulate the problem as a linear system in the following way:
max 300x + 200y

subject to 6x + 2y ≤ 300
5x + 5y ≤ 350
x, y ≥ 0
The formulation says to maximize an expression such that the “subject to”
conditions (called constraints) are fulfilled. In LPL the problem can be formu-
lated as follows (see model tutor013 ):

model simple;
variable x; y;
maximize revenue: 300*x + 200*y;
constraint A: 6*x + 2*y <= 300;
constraint B: 5*x + 5*y <= 350;
end
As expected, solving the model gives x = 40 and y = 30.

The simple example is an instance of an important class of problems having a
huge number of real applications. In the operations research (OR) community
the class is called linear programs or LP. The general linear model containing n
constraints and m variables can be compactly formulated as follows:
X
max cj x j
X
j∈J
subject to Ai,j xj ≤ bi for all i ∈ I
j∈J
xj ≥ 0 for all j ∈ J
with J = {1 . . . m}, I = {1 . . . n}, m, n ≥ 0
An even more compact formulation using matrix notation for the model is:
max cx
subject to Ax ≤ b
x≥0
In LPL the general model can be formulated as follows:

model myLP;
set i; j;
parameter A{i,j}; c{j}; b{i};
variable x{j};
constraint C{i}: sum{j} A*x <= b;
maximize obj: sum{j} c*x;
end
This is very close to the mathematical notation. The model first declares the
two sets i and j. The data matrix and the two vectors are compactly declared
as Ai,j , cj and bi .4 There size will depend of the number of elements of the
two sets. The variable vector xj is also declared in a similar way. (The posi-
tivity of the variable is automatically assumed by LPL, if no range is given.)
3
http://lpl.virtual-optima.com/lpl/Solver.jsp?name=/tutor01
4
Note – as already mentioned – in LPL syntax index names can be the same as set
names if no ambiguity is created.
The constraints and the maximizing function are also very close to the math-
ematical notation. The model does not contain data. They can be added by
a separate submodel. For example, to formulate the small example above in
this general form, the data can be added as :
model data;
i := [1..2]; j := [1..2];
A{i,j} := [6 2 , 5 5];
c{j} := [300 200];
b{i} := [300 350];
end
Data – we already gave an example – can be placed within the model code or
it can be read from files using the Read function. The following data matrix:
 

 2 3 −4 1 15  

 
 −5 4 8 1 5  
ai,j = 2 1 14 1 17 with i ∈ I, j ∈ J

 


 3 4 5 1 7 

 
0 8 −3 10 1
could be written in LPL as:

set i, j := [1..5];
parameter a{i,j} := [2 3 -4 1 15, -5 4 8 1 5,
2 1 14 1 17, 3 4 5 1 7, 0 8 -3 10 1 ];
The matrix elements are listed in lexicographical order (other formats exist).
Of course, for large data tables, it is better to read them from files or database
tables. A comma-separated second identifier just introduces another name
for the same set. Hence, i and j represent the same set, but they may be used
as different index names.
Writing the results can be done using the Writep and Write functions. The
simplest way to output the values of the variable vector x{j} is as follows;
Writep(x);
However, sometimes we need to format the output. To write each value of xj

together with the set element name on a separate line, we may write in LPL:
Write{j}('Element: %s Value= %6.2f \n', j,x);
The string in single quotes contains the literal text and formatting instructions
that must be written. %s is a place holder for strings and %6.2f is a place
holder for floating point data (with width 6 and 2 decimals). The values of the
following two parameters ,j,x replace the placeholders in the string. This
expression means to write (for each element in the set j) on a separate line the
following two elements, that is (j,x). More information on this powerful
“reporting instruction” can be found in the reference manual of LPL.
Data can also be assigned by the assignment instruction. The table assignment
ai,j = cj + bi for all i ∈ I, j ∈ J
would be formulated in LPL as follows:

a{i,j} := c[j] + b[i];
Another often used instruction in the presented case studies is the for loop.
Suppose, one would like to optimize a model 10 times (with different param-
eters), then one could write:
set loop := [1..10];
for{loop} do
maximize obj: ... ;
....
end
The loop can contain a sequence of any instructions.
The reader should now be able to read and recognize the most basic state-
ments of LPL in order to work with the case studies in the following chapters.
There is no room here to explain each detail of the language. The interested
user can get a free LPL version together with the definite language reference
guide and tutor examples. It can be downloaded from the LPL-site.
1.4. Mathematical Notation

The mathematical notation in this book deviates in somewhat from the com-
mon notations in three ways.
1.4.1 Set-Names versus Index-Names
The LPL syntax uses some shortcuts in mathematical notation which is re-
flected also in the notation of mathematical expressions used in this book. As
already mentioned, one can use an set name also as an index name and vice
versa. So we use the following shorter notation in LPL :
set i := 1..10;
set j := 1..5;
parameter b{i,j} := i*j;
parameter a{i} := sum{j} b[i,j];
instead of the longer notation (although this second notation is also perfectly
correct in LPL) :
set I := 1..10;
set J := 1..5;
parameter b{i in I,j in J} := i*j;
parameter a{i in I} := sum{j in J} b[i,j];
This is also reflected in the mathematical notation often used in this book as :
i·j
bi,j =X for all i, j
ai = bi,j for all i
j
i ∈ {1, . . . , 10}
j ∈ {1, . . . , 5}
instead of the usual notation :
i·j
bi,j =X for all i ∈ I, j ∈ J
ai = bi,j for all i ∈ I
j∈J
i∈I I = {1, . . . , 10}
j∈J J = {1, . . . , 5}
The difference is that we do not use the set names I or J, if the context is clear,
only index names i or j are written.
1.4.2 Logic-mathematical Notation
LPL does not distinguish between Boolean and mathematical operators. All
Boolean operator return 0 or 1, which might be interpreted as false and true.
Hence, one can mix these operators in an expression. This is also true in
constraints. We may write, for example:
a=0 or b=1
(a=1) + (b=10)
The first expression is common in all kind of programming languages and

is a Boolean expression that is true if a is zero or if b is one. The second
expression seems to be a little bit strange. But on the light that all expressions
are numerical, we have: a = 1 is zero or one, and also b = 10 is zero or one, so
the result of the second expression is zero, one or two. This is also used in the
mathematical notation and in constraints. For example, in the model will075 ,
the constraint (with variables dm,t ∈ {0, 1} and xm,t ≥ 0) is used:
dm,t ← xm,t > 0
The expression means: “the mine is working if ore is extracted from it”, or “if
a certain positive quantity of ore is delivered from a mine then the variabe d
is defined to be true”.
5
http://lpl.virtual-optima.com/lpl/Solver.jsp?name=/will07
1.4.3 Logical Operators
LPL uses various logical and Boolean operators that can also be used in con-
straints, see Table 1.1. Many problems can be formulated using Boolean or
logical constraints.
Meaning Math. notation LPL syntax

and x∧y x and y
or x∨y x or y
not ¬x ~x
not and x nand y (¬(x ∧ y) x nand y
not or x nor y (¬(x ∨ y) x nor y
implication x→y x -> y
rev. impl. x←y x <- y
exclusive or x∨ ˙ y x xor y
equivalence V ⇐⇒ y
x x <-> y
indexed and Wi xi and{i} x[i]
indexed or i xi or{i} x[i]
indexed xor
W˙ xi xor{i} x[i]
i
forall ∀i xi forall{i} x[i]
exists ∃i xi exist{i} x[i]
indexed nand nandi xi nand{i} x[i]
indexed nor nori xi nor{i} x[i]
at least 2 atleast(2)i xi atleast(2){i} x[i]
at most 2 atmost(2)i xi atmost(2){i} x[i]
exactly 2 exactly(2)i xi exactly(2){i} x[i]
Table 1.1: Logical Operators

C HAPTER 2
P ROBLEMS FROM W ILLIAM ’ S

B OOK
“Approach your problem from the right end and begin with the
answer. Then one day, perhaps you will find the final ques-
tion.”
— R. van Gulik
“It isn’t that they can’t see the solution. It is that they can’t see
the problem.”
— G. K. Chesterton
“No mathematician can be a complete mathematician unless he

is also something of a poet.”
— Weierstrass
This chapter presents all the models exposed in the famous modeling book
of Williams [17]. All models are implemented in LPL and can be executed
straight away.
19
CHAPTER 2. PROBLEMS FROM WILLIAM’S BOOK 20
2.1. Food Manufacture I (will01)

Problem: A company manufactures (over the next six months) one food by re-
fining various raw oils, which come in vegetable and non-vegetable oil – each
with a given “hardness”. The company can buy oils in the different months.
The storage capacity is not a limitation but costs 5ct/ton. The refining capac-
ity, however, is limited at 200 tons of veg oil and 250 tons of non-veg oils. The
“hardness” of the food must be in certain given ranges. How many tons of
each oil should the company buy, store, and use in each period in order to
maximize profit by selling the food at a given price ?
Model: The Model is as follows:The problem is from [17] Chapter 13.1.
Listing 2.1: The Complete Model in LPL [29]

model Will01 "Food Manufacture I";
set p := [ January February March April May June ];
r := [ VEG1 VEG2 OIL1 OIL2 OIL3 ] "all oils";
v{r} := [ VEG1 VEG2 ] "veg oils";
parameter pr{p,r} := [
110 120 130 110 115, 130 130 110 90 115
110 140 130 100 95, 120 110 120 120 125
100 120 150 110 105, 90 100 140 80 135 ];
hardness{r} := [ 8.8 6.1 2.0 4.2 5.0 ];
Pr:=150; Sto:=500; vegCapa:=200; oilCapa:=250;
variable
food{p} "Quantity of food produced";
buy{p,r} "Quantity bought in each period";
use{p,r} "Quantity used in each period";
sto{p,r} "Quantity stored in each period";
constraint
Bal{r,p}: if(p=1,Sto,sto[p-1,r]) + buy = use + sto;
Endstore{r}: sto[#p,r] = Sto;
Capacity1{p}: sum{v} use <= vegCapa;
Capacity2{p}: sum{r|~v} use <= oilCapa;
Hardness{p}: 3*food <= sum{r} hardness*use <= 6*food;
Conserve{p}: sum{r} use = food;
maximize profit: sum{p}Pr*food - sum{p,r}(pr*buy+5*sto);
Writep(profit,buy,use,sto,food);
end

21 CHAPTER 2. PROBLEMS FROM WILLIAM’S BOOK
2.2. Food Manufacture II (will02)

Problem: A company manufactures (over the next six months) one food by re-
fining various raw oils, which come in vegetable and non-vegetable oil – each
with a given “hardness”. The company can buy oils in the different months.
The storage capacity is not a limitation but costs 5ct/ton. The refining capac-
ity, however, is limited at 200 tons of veg oil and 250 tons of non-veg oils. The
“hardness” of the food must be in certain given ranges. How many tons of
each oil should the company buy, store, and use in each period in order to
maximize profit by selling the food at a given price ? Furthermore, we wish
to impose the following restrictions :
1. The food may never be made up of more than three different oils.
2. If an oil is used at least 20 tons must be used.
3. If either vegetable oil Veg1 or Veg2 is used, then non-vegetable oil
Oil3 must also be used.
Model: The Model is as follows:The problem is from [17] Chapter 12.2. The
model is the same as will011 , with exception of the additional restrictions. to
impose these conditions we introduce a 0-1 variable d which is 1 only if the
qunatity used (use) of an oil is not zero, that is we must impose:
use > 0 ↔ d = 1

model Will02 "Food Manufacture II";
set p := [ January February March April May June ];
r := [ VEG1 VEG2 OIL1 OIL2 OIL3 ] "all oils";
v{r} := [ VEG1 VEG2 ] "veg oils";
parameter pr{p,r} := [
110 120 130 110 115, 130 130 110 90 115
110 140 130 100 95, 120 110 120 120 125
100 120 150 110 105, 90 100 140 80 135 ];
hardness{r} := [ 8.8 6.1 2.0 4.2 5.0 ];
Pr:=150; Sto:=500; vegCapa:=200; oilCapa:=250;
variable
food{p} "quantity of food produced";
buy{p,r} "quantity bought in each period";
use{p,r} "quantity used in each period";
sto{p,r} "quantity stored in each period";
binary d{p,r}; // is 1 if use>0
constraint
1
Bal{r,p}: if(p=1,Sto,sto[p-1,r])+buy = use+sto;

endstore{r}: sto[#p,r]=Sto;
Capacity1{p}: sum{v} use <= 200;
Capacity2{p}: sum{r|~v} use <= 250;
Hardness{p}: 3*food <= sum{r} hardness*use <= 6*food;
Conserve{p}: sum{r} use = food;
Cond1{p}: sum{r} d <= 3;
Cond2{p,r} :use-20*d >=0; -- use>0 -> d=1
Cond2a{p,r}:use-if(v,200,250)*d <=0; -- d=1 -> use>0
Cond3{p}: d[p,'VEG1']+d[p,'VEG2']-2*d[p,'OIL3']<=0;
maximize profit: sum{p}Pr*food - sum{p,r}(pr*buy+5*sto);
Writep(profit,buy,use,sto,food);
end

2.3. Factory planning I (will03)

Problem: Seven products are manufactured using different machines over the
next month. Each product yields a certain contribution to profit. Several ma-
chines will be down in different months for maintenance. We also want a
given endstock for each product at the end of our time horizont. What must
be manufactured, stored and sold each month in order to maximize profit, if
the storage is limited and the market also has limitations on each product?

model Will03 "Factory planning I";
set
p:=[Prod1 Prod2 Prod3 Prod4 Prod5 Prod6 Prod7] "prod";
m:=[grinder vDrill hDrill borer planer] "mach";
t:=[Jan Feb Mar Apr May Jun] "period";
parameter profit{p} := [10 6 8 4 11 9 3];
time{m,p} "machine time table" :=[.5 .7 . . .3 .2 .5,
.1 .2 . .3 . .6 . , .2 . .8 . . . .6 ,
.05 .03 . .07 .1 . .08 , . . .01 . .05 . .05];
down{t,m} "number of machines down" := [
(Jan,grinder) 1 (Feb,hDrill) 2 (Mar,borer) 1
(Apr,vDrill) 1 (May,grinder) 1 (May,vDrill) 1
(Jun,planer) 1 (Jun,hDrill) 1];
qMach{m} "nr. of machines available" := [4 2 3 1 1];
upper{t,p} "market limitation of sells" := [500 1000
300 300 800 200 100, 600 500 200 0 400 300 150, 300
600 0 0 500 400 100, 200 300 400 500 200 0 100, 0
100 500 100 1000 300 0, 500 500 100 300 1100 500 60];
storeCost := 0.5; storeCapacity := 100;
endStock := 50; hoursMonth := 2*8*24;
variable
manu{t,p} "quantity manufactured";
held{t,p} "quantity stored";
sell{t,p} [0..upper] "quantity sold";
constraint
Balance{t,p}: held[t-1,p] + manu = sell + held;
endstore{p}: held[#t,p] = endStock;
storecapa{t,p|t<#t}: held <= storeCapacity;
capa{t,m}: sum{p} time*manu <= hoursMonth*(qMach-down);
maximize Profit: sum{t,p} (profit*sell - storeCost*held);
Writep(Profit,manu,held,sell);
Write(' Manufacture Sell Store\n');
Write{t,p}('%5s %5s %8d %8d %8d\n',t,p,manu,sell,held);
end

2.4. Factory planning II (will04)

Problem: Seven products are manufactured using different machines over the
next month. Each product yields a certain contribution to profit. Several ma-
chines will be down in different months for maintenance. We also want a
given endstock for each product at the end of our time horizont. What must
be manufactured, stored and sold each month in order to maximize profit, if
the storage is limited and the market also has limitations on each product?
Instead of stipulating when each machine is down, it is desired to find the
best month for each machine to be down. Each machine must be down once
in the six months except two grinders.
Model: The Model is as follows:The problem is from [17]. For a problem de-
scription see Chapter 13.4. Note that this is the same model as will032 except
that down is now a variable.

model Will04 "Factory planning II";
set
p:=[Prod1 Prod2 Prod3 Prod4 Prod5 Prod6 Prod7] "prod";
m:=[grinder vertDrill horiDrill borer planer] "mach";
t:=[January February March April May June] "period";
parameter
profit{p} "profit per p" := [ 10 6 8 4 11 9 3 ];
time{m,p} "machine time table" := [
0.5 0.7 . . 0.3 0.2 0.5
0.1 0.2 . 0.3 . 0.6 .
0.2 . 0.8 . . . 0.6
0.05 0.03 . 0.07 0.1 . 0.08
. . 0.01 . 0.05 . 0.05 ];
qMach{m} "nr of machines" := [ 4 2 3 1 1];
notDown{m} "nr not-down" := [ 2 . . . .];
upper{t,p} "market limitation of sells" := [
500 1000 300 300 800 200 100
600 500 200 0 400 300 150
300 600 0 0 500 400 100
200 300 400 500 200 0 100
0 100 500 100 1000 300 0
500 500 100 300 1100 500 60 ];
storeCost := 0.5; storeCapacity := 100;
endStock := 50; hoursMonth := 2*8*24;
variable
manu{t,p} "quantity manufactured";
held{t,p} "quantity stored";
sell{t,p} [0..upper] "quantity sold";
2
integer down{m,t} [0..qMach-notDown]

"number of machines down in a period";
constraint
Balance{t,p}: held[t-1,p] + manu = sell + held;
endstore{p}: held[#t,p]=endStock;
storecapa{t,p|t<>#t}: held <= storeCapacity;
capa{t,m}: sum{p} time*manu <= hoursMonth*(qMach-down);
mustBeDown{m}: sum{t} down = qMach - notDown;
maximize Profit: sum{t,p} (profit*sell - storeCost*held);
Writep(Profit,manu,held,sell,down);
end

2.5. Manpower Planning (will05)

Problem: A company is undergoing a number of changes over the next 3 years
which will affect its manpower requirements in future years. “Less unskilled”
and “more skilled” men are needed. It also has “semi-skilled”. The company
must decide its policy with regard the following:
1. Recruitment from outside the company which is limited

2. Retraining from less skilled to more skilled
3. Redundancy and overmanning which causes extra costs and
4. Short-time working
The company declared to minimize the overall redundancy and costs.
Model: The Model is as follows:The problem is from [17]. For a problem

description see Chapter 13.5.

model Will05 "Manpower Planning";
set year,y := [1..3];
skill,s,t := [Unskilled,SemiSkilled,Skilled];
parameter
CurrentStrength{s} := [2000,1500,1000];
Requirement{y,s} := [1000, 1400, 1000,
500, 2000, 1500, 0, 2500, 2000];
LeaveFirstYear{s} := [0.25, 0.20, 0.10];
LeaveEachYear{s} := [0.10, 0.05, 0.05];
ContinueFirstYear{s} := 1-LeaveFirstYear;
ContinueEachYear{s} := 1-LeaveEachYear;
LeaveDownGraded := 0.50;
ContinueDownGraded := 1-LeaveDownGraded;
MaxRecruit{s} := [500, 800, 500];
MaxRetrainUnskilled := 200;
MaxOverManning := 150;
MaxShortTimeWorking := 50;
RetrainSemiSkilled := 0.25;
ShortTimeUsage := 0.50;
RetrainCost{s} := [400, 500, 0];
RedundantCost{s} := [200, 500, 500];
ShortTimeCost{s} := [500, 400, 400];
OverManningCost{s} := [1500, 2000, 3000];
variable
LaborForce,L{s,y};
Recruite, R{s,y} [0..MaxRecruit];
Retrain, T{s,t,y};
Redundant, D{s,y};
ShortTime, S{s,y} [0..MaxShortTimeWorking];
OverManned,O{s,y};
expression
TotalRetrainCost : sum{y,s} RetrainCost*T[s,s+1,y];
TotalRedundantCost : sum{s,y} RedundantCost*D;
TotalShortTimeCost : sum{s,y} ShortTimeCost*S;
TotalOverManningCost: sum{s,y} OverManningCost*O;
TotalManpowerCost:TotalRetrainCost+TotalRedundantCost
+TotalOverManningCost + TotalShortTimeCost;
TotalRedundantMen : sum{s,y} D;
constraint
Continuity{s,y}: L =
ContinueEachYear*if(y=1,CurrentStrength,L[s,y-1])
+ ContinueFirstYear*R + 0.95*T[s-1,s,y]
+ sum{t} ContinueDownGraded*T[s+t,s,y]
- sum{t} T[s,s-t,y] - T[s,s+1,y] - D;
RetainMaxUnskilled{y}:
T['Unskilled','SemiSkilled',y]<=200;
RetrainingSemiSkilled{y}: T['SemiSkilled','Skilled',y]
<= RetrainSemiSkilled*L['Skilled',y];
OverManning{y}: sum{s} O <= MaxOverManning;
Requirements{s,y}: L = Requirement+O+ShortTimeUsage*S;
minimize TotalRedundancy: TotalRedundantMen;
Writep(TotalManpowerCost,TotalRedundantMen,
Recruite,Retrain,ShortTime,OverManned);
end

2.6. Refinery Optimization I (will06)

Problem: An oil refinery purchases two crude oil, which are subject to four
processes:
1. Distillation which yields light, medium, and heavy naphtha as well as

light oil, heavy oil, and residuum
2. Reforming which yields reformed gasoline using the naphthas.
3. Cracking which yields cracked oil and gasoline using the oils
4. Blending which yields the final products: regular and premium petrol,
jet fuel, fuel oil, and lube oil. The objective is to maximize profit.


model Will06 "Refinery Optimization I";
set c := [Crude1, Crude2];
f := [LightNaphtha MediumNaphtha HeavyNaphtha
LightOil HeavyOil Residuum];
n{f} := [LightNaphtha MediumNaphtha HeavyNaphtha];
o{f} := [LightOil HeavyOil];
a := [CrackedOil CrackedGasoline];
p:=[PremiumFuel RegularFuel JetFuel FuelOil LubeOil];
m{p} := [PremiumFuel, RegularFuel];
parameter
DistillYld{c,f}:=[0.10 0.20 0.20 0.12 0.20 0.13,
0.15 0.25 0.18 0.08 0.19 0.12];
ReformingYield{n} := [0.60, 0.52, 0.45];
CrackingYield{o,a} := [0.68, 0.28, 0.75, 0.20];
ResiduumLubeOilYield := 0.50;
VaporPressure{o} := [1.0, 0.6];
VaporPressureCrackedOil := 1.5;
VaporPressureResiduum := 0.05;
FuelOilRatio{o} := [.55555, .16666];
FuelOilRatioCrackedOil := 4/18;
FuelOilRatioResiduum := 1/18;
OctaneNumberNaphtha{n} := [90, 80, 70];
OctaneNumberReformedGas := 115;
OctaneNumberCrackedGas := 105;
MinOctaneLevel{m} := [94, 84];
CrudeDailyAvail{c} := [20000, 30000];
DistillCapacity := 45000;
ReformingCapacity := 10000;
CrackingCapacity := 8000;
MinLubeOil := 500;
MaxLubeOil := 1000;
MinPremiumProdPct := 0.40;
Profit{p} := [7.00 6.00 4.00 3.50 1.50];
variable
FinalProduct{p}; PurchaseCrude{c};
DistillOutput{f}; NaphthaForReformGas{n};
ReformedGasoline; OilForCracking{o};
CrackedOutput{a}; NaphthaForMotorFuel{n,m};
ReformedGasForMotorFuel{m};
CrackedGasForMotorFuel{m};
OilForJetFuel{o}; ResiduumForJetFuel;
CrackOilForJetFuel; ResiduumForLubeOil;
maximize TotalProfit: sum{p} Profit*FinalProduct;
constraint
DistillLimit "At most 45000 barrels of crude can be
distilled per day":
sum{c} PurchaseCrude <= DistillCapacity;
ReformingLimit "At most 10000 barrels of naphtha can
be reformed per day":
sum{n} NaphthaForReformGas <= ReformingCapacity;
CrackingLimit "At most 8000 barrels of oil can be
cracked per day":
sum{o} OilForCracking <= CrackingCapacity;
DistillationBalance{f} "Balance crude inputs to
distillation outputs":
sum{c} DistillYld*PurchaseCrude = DistillOutput;
ReformedGasBalance "Balance naphtha inputs to
reformed gas output":
sum{n} ReformingYield*NaphthaForReformGas =
ReformedGasoline;
CrackBalance{a} "Balance oil inputs to cracked output
":
sum{o} CrackingYield*OilForCracking =
CrackedOutput;
ResiduumLubeOil "Each barrel of residuum yields 0.5
barrels of lube oil":
ResiduumLubeOilYield*ResiduumForLubeOil =
FinalProduct['LubeOil'];
DivideReformGas "Divide reformed gas to motor fuels":
ReformedGasoline = sum{m}ReformedGasForMotorFuel;
DivideNaphtha{n} "Divide naphthas to reformed gas and
motor fuels":
DistillOutput = NaphthaForReformGas + sum{m}
NaphthaForMotorFuel;
DivideOil{o} "Divide light/heavy oils to cracking,
jet fuel, and fuel oil":
DistillOutput = OilForCracking + OilForJetFuel +
FuelOilRatio*FinalProduct['FuelOil'];
DivideCrackGas "Divide cracked gasoline to motor
fuels":
CrackedOutput['CrackedGasoline'] = sum{m}
CrackedGasForMotorFuel;
DivideCrackedOil "Divide cracked oil to jet fuel and
fuel oil":
CrackedOutput['CrackedOil'] = CrackOilForJetFuel
+ FuelOilRatioCrackedOil*FinalProduct['FuelOil'];
DivideResiduum "Divide Residuum to jet fuel, fuel oil
, and lube oil":
DistillOutput['Residuum'] = ResiduumForJetFuel
+ FuelOilRatioResiduum*FinalProduct['FuelOil'] +
ResiduumForLubeOil;
BlendMotorFuel{m} "Blending motor fuels from naphtha,
reformed gas, and cracked gas":
FinalProduct = sum{n} NaphthaForMotorFuel +
ReformedGasForMotorFuel +
CrackedGasForMotorFuel;
BlendJetFuel "Blending jet fuel from light/heavy oils
, cracked oil and residuum":
FinalProduct['JetFuel'] = sum{o} OilForJetFuel +
CrackOilForJetFuel + ResiduumForJetFuel;
MaxVaporPressure "Jet fuel maximum vapor pressure":
sum{o} VaporPressure*OilForJetFuel +
VaporPressureCrackedOil*CrackOilForJetFuel
+ VaporPressureResiduum*ResiduumForJetFuel <=
FinalProduct['JetFuel'];
MinPremFuel "Premium fuel at least 40% of Regular":
FinalProduct['PremiumFuel'] >= MinPremiumProdPct*
FinalProduct['RegularFuel'];
MinOctFuel{m} "Motor fuels minimum octane levels":
sum{n} OctaneNumberNaphtha*NaphthaForMotorFuel
+ OctaneNumberReformedGas*ReformedGasForMotorFuel
+ OctaneNumberCrackedGas*CrackedGasForMotorFuel
>= MinOctaneLevel*FinalProduct;
C{c}: PurchaseCrude <= CrudeDailyAvail;
D: MinLubeOil<= FinalProduct['LubeOil'] <=MaxLubeOil;
Writep(TotalProfit);
SetFocus(); //write all variables
while NextFocus(3) do
while NextPosition() do
Write('%-47s %d\n', GetName(0), GetValue(0));
end;
end
end

2.7. Refinery Optimization II (will06a)

Problem: An oil refinery purchases two crude oil, which are subject to four
processes:
1. Distillation which yields light, medium, and heavy naphtha as well as

light oil, heavy oil, and residuum
2. Reforming which yields reformed gasoline using the naphthas.
3. Cracking which yields cracked oil and gasoline using the oils
4. Blending which yields the final products: regular and premium petrol,
jet fuel, fuel oil, and lube oil. The objective is to maximize profit.


model Will06a "Refinery Optimization II";
variable
cr1; cr2; ln; mn; hn; lo; ho; r; lnrg; mnrg; hnrg; rg;
locgo; hocgo; cg; co; lnpmf; lnrmf; mnpmf; mnrmf;hnpmf;
hnrmf; rgpmf; rgrmf; cgpmf; cgrmf; lojf; hojf; rjf;
cojf; lofo; hofo; rfo; cofo; rlbo; pmf; rmf;
jf; fo; lbo;
constraint
r1: cr1 <= 20000; --availabilities
r2: cr2 <= 30000;
r3: cr1+cr2 <= 45000; --capacity
r4: lnrg+mnrg+hnrg <= 10000; --reforming capacity
r5: locgo+hocgo <= 8000; --cracking capacity
r6: 500 <= lbo <= 1000; --bounds
r8: 0.1 *cr1 +0.15*cr2 >= ln; --split under
destillation
r9: +0.2 *cr1 +0.25*cr2 >= mn;
r10: +0.2 *cr1 +0.18*cr2 >= hn;
r11: +0.12*cr1 +0.08*cr2 >= lo;
r12: +0.2 *cr1 +0.19*cr2 >= ho;
r13: +0.13*cr1 +0.12*cr2 >= r;
r14: +0.6*lnrg+0.52*mnrg+0.45*hnrg >= rg; --splitting
under reforming
r15: 0.68*locgo+0.75*hocgo = co; --splitting under
cracking
r16: 0.28*locgo+0.2 *hocgo = cg;
r17: lnrg+lnpmf+lnrmf = ln; --continuities
r18: mnrg+mnpmf+mnrmf = mn;
r19: hnrg+hnpmf+hnrmf = hn;
r20: locgo+lojf+0.55*fo = lo; --fixed proportions on

fuel oil in blending
r21: hocgo+hojf+0.17*fo = ho;
r22: cojf+0.22*fo = co;
r23: rlbo+rjf+0.055*fo = r;
r24: cgpmf+cgrmf = cg; --further continuities
r25: rgrmf+rgpmf = rg;
r26: +lnpmf+mnpmf+hnpmf+rgpmf+cgpmf = pmf;
r27: +lnrmf+mnrmf+hnrmf+rgrmf+cgrmf = rmf;
r28: +lojf+hojf+cojf+rjf = jf;
r29: +0.5*rlbo = lbo; --lube-oil is 0.5 of residuum
used
r30: pmf >= 0.4*rmf; --pmf/rmf must be 40%
r31: +90*lnpmf+80*mnpmf+70*hnpmf+115*rgpmf+105*cgpmf >=
94*pmf; --octane number
r32: +90*lnrmf+80*mnrmf+70*hnrmf+115*rgrmf+105*cgrmf >=
84*rmf;
r33: +lojf+0.6*hojf+1.5*cojf+0.05*rjf <= jf; --
vapour pessure constraints
maximize profit: 7*pmf+6*rmf+4*jf+3.5*fo+1.5*lbo;
Writep(profit);
while NextFocus(3) do
while NextPosition() do
Write('%-10s %d\n', GetName(0), GetValue(0));
end;
end
end

2.8. Mining (will07)

Problem: A mining company extracts ore from several mines over the next
five years. The ore from different mines is of varying quality and the extracted
amount of ore is limited. Furthermore, royalities are payable on each mine
kept open. The company can operate on three of the four available mines at
a time. A mine can be closed and no royalities are payable. A closed mine,
however, must be closed down permanently. Which mines should be oper-
ated each period and how much should they produce? Future revenue and
expentiture must be discounted.


model Will07 "Mining";
set m := [Mine1 Mine2 Mine3 Mine4];
t,t0 := [Year1 Year2 Year3 Year4 Year5];
parameter
Royalities{m} := [5 4 4 5];
ExtractLimit{m}:= [2 2.5 1.3 3];
OreQuality{m} := [1 .7 1.5 .5];
BlendedQuality{t} := [0.9 0.8 1.2 0.6 1.0];
rate := 10/100;
sellPrice := 10;
discountFactor{t} := 1/(prod{t0|t0<t} (1+rate));
variable
x{m,t}[0..ExtractLimit] "output";
q{t} "quantity of blended ore";
binary d{m,t} "mine is working (if x>0)";
binary g{m,t} "mine is open";
constraint
WorkingMines{t}: atmost(3){m} d;
NoWorkOnClosedMines{m,t}: d -> g;
SubsequentYearsClosed{m,t|t<#t}: g[m,t+1] -> g;
Blending{t}: sum{m} OreQuality*x = BlendedQuality*q;
Combine{t}: sum{m} x = q;
Working{m,t}: d <- x>0 "mine is working if x>0";
maximize profit: - (sum{m,t} discountFactor*Royalities*g)
+ sum{t} discountFactor*sellPrice*q;
Writep(profit,x,q,discountFactor,d,g);
end

2.9. Farm Planning (will08)

Problem: A farmer wishes to plan production on his farm for the next five
years. He has heifers and milk-producing cows. They consume grain and
sugar beet which can be grown both on the farm or bought (or sell) on the
market. Heifers produce calves which can be sold (the bollocks are sold
immediately) or which will become milk-producing cows in the third year.
Labour can be recruited from outside and extra capital is also available to
some extent. How should the farmer operate to maximize profit over the five
years? How much grain and sugar beet should he plant, sell or buy; how
much extra labour and capital should he use; and how many cows does he
have in each year and of each age?


model Will08 "Farm Planning";
set t,k := [ Year1 Year2 Yera3 Year4 Year5 ];
i := [ 1 2 3 4 ]; /*land groups*/
j := [ 1 2 3 4 5 6 7 8 9 10 11 12 ]; --cow age
j1{j} := [ 2 3 4 5 6 7 8 9 10 11 ];
parameter
GrainYld{i} "grain growing" := [1.1 0.9 0.8 0.65];
GrainArea{i} "area in arces" := [ 20 30 20 10];
variable
GR{i,t} "grain grown in tons";
SB{t} "sugar beet grown in tons";
GRBUY{t} "bought grain in tons";
GRSELL{t} "sold grain in tons";
SBBUY{t} "bought sugar beet in tons";
SBSELL{t} "sold sugar beet in tons";
EXLAB{t} "extra labour recruited";
EXCAP{t} "capital outlay";
HFSELL{t} "number of heifers sold at birth";
COWS{j,t} "number of cows of age j in year t";
NEWCOW{t} "number of cows of age 0 at year t";
PROF{t} "profit in year t";
constraint
ACCOMM{t} "Accommodation (housing capacity)":
NEWCOW + sum{j|j<#j} COWS <= 130 + sum{k|k<=t} EXCAP;
GRCONS{t} "Grain consumption":
sum{j1} 0.6*COWS[j1,t] <= sum{i} GR + GRBUY - GRSELL;
SBCONS{t} "Sugar beet consumption":
sum{j1} 0.7*COWS[j1,t] <= SB + SBBUY - SBSELL;
BOUNDS{i,t} "Grain growing 'capacity'":
GR <= GrainYld*GrainArea;
ACREAS{t} "Acreage limitation":
sum{i} 1/GrainYld*GR + 2/3*SB + 2/3*NEWCOW
+ 2/3*COWS[1,t] + sum{j1} COWS[j1,t] <= 200;
LABOUR{t} "Labour in 100 hours":
0.1*NEWCOW + 0.1*COWS[1,t] + sum{j1} 0.42*COWS[j1,t]
+ sum{i} 0.04/GrainYld*GR + 0.14/1.5*SB <= 55+EXLAB;
-- continuity constraints
CTA{t|t<#t}: COWS[1,t+1] = 0.95*NEWCOW[t];
CTB{t|t<#t}: COWS[2,t+1] = 0.95*COWS[1,t];
CT{j1,t|t<#t}:COWS[j1+1,t+1] = 0.98*COWS[j1,t];
CTC{t}: NEWCOW - sum{j1} 0.55*COWS[j1,t] + HFSELL = 0;
ENDT "Cows at the end": 50<= sum{j1} COWS[j1,5] <=175;
PROFIT{t}:
sum{j1} 16.5*COWS[j1,t] + 40*HFSELL + 120*COWS[12,t]
+ sum{j1} 370*COWS[j1,t] + 75*GRSELL + 58*SBSELL
- 90*GRBUY - 70*SBBUY - 120*EXLAB - 50*NEWCOW
- 50*COWS[1,t] - sum{j1} 100*COWS[j1,t]
- sum{i}15/GrainYld*GR - 20/3*SB - sum{t}39.71*EXCAP
- PROF = 4000;
-- initial conditions, fixing number of cows of each
age at first period
BOUNDS1{j}: COWS[j,1] = if(j<3 , 9.5 , 9.8);
-- maximize profit minus discounted capital expenditures
maximize TProf: sum{t} (PROF-158.85*EXCAP-39.71*t*EXCAP);
Writep(TProf,GR,SB,GRBUY,GRSELL,SBBUY,SBSELL,EXLAB,
EXCAP,HFSELL,COWS,NEWCOW,PROF);
end

2.10. Economic Planning (will09)

Problem: An economy consists of the three industries coal, steel, and trans-
port. Each industry requires inputs from its own, inputs from the other indus-
tries, and manpower. Output may also be used to build productive capacity,
which become effective two periods later. Manpower capacity is limited and
an initial stock and productive capacity is given. We wanted to know the
growth patterns for the economy pursuing the following objectives:
1. Maximizing the productive capacity at the end of the considered time

while meeting an exogenous demand
2. Maximizing total production the the last two periods ignoring the ex-
ogenous demand (set to zero)
3. Maximizing the total manpower requirement ignoring the manpower
capacity limitation while meeting the exogenous demand.


model Will09 "Economic Planning";
set o,o1 := [Coal Steel Transport] "activities";
t,t1 := [1..6] "periods";
parameter
req{o1,o} := [.1 .5 .4 , .1 .1 .2 , .2 .1 .2];
cap{o1,o} := [.0 .7 .9 , .1 .1 .2 , .2 .1 .2];
Mreq{o} := [.6 .3 .2];
Mcap{o} := [.4 .2 .1];
Stocks{o} := [ 150 80 100 ] "initial stock" ;
Capa {o} := [ 300 350 280 ] "capa. in period 0";
consum{o} := [ 60 60 30 ] "exo. demand";
consum1{o}:= consum "make a copy of consum";
lastx {o} := [166.4 105.7 92.3] "min. end output";
manCapa := 470 "labour capacity";
variable
x{o,t} [if(t=#t,lastx,0)..99999] "outputs in t";
s{o,t} "stock level at beginning of t";
y{o,t|t<#t-1} "extra productive capacity for t+1";
constraint
TotOutput{o1,t}: sum{o} (req*x+cap*y) + s <=
x[o1,t-1] + s[o1,t-1] + if(t=1,Stocks,-consum);
Manpower{t}: sum{o} (Mreq*x+Mcap*y) <= manCapa;
prodCapa{o,t|t>1}: x - sum{t1|t1<t} y <= Capa;
maximize obj: sum{o} Capa + sum{o,t} y[o,t];
Write('-- (1) maximize the productive capacity\n');
Writep(obj,y,x,s);
consum{o}:=0; /*ignore the exogenous demand*/
maximize obj1: sum{o,t|t>=#t-1} x[o,t];
Write('-- (2) maximize the final output\n');
Writep(obj1,y,x,s);
consum{o}:=consum1; /*recopy the original consum*/
maximize obj2: sum{t,o} (Mreq*x[o,t+1] + Mcap*y[o,t+2])
subject_to (Will09, ~Manpower);
Write('-- (3) maximize the manpower requirement\n');
Writep(obj2,y,x,s);
end

2.11. Decentralization (will10)

Problem: A large company wants to move 5 of its departments out of London
(to Bristol or Brighton). There are benefits (cheaper housing etc.) and costs
(communication) to be derived from doing this. Where should each depart-
ment be located so as to minimize overall yearly costs?


model Will10 "Decentralization";
set D,i,k := [A B C D E] "Departements";
C,j,l := [Bristol Brighton London] "Cities";
parameter
Benefit{D,C}:=[10 10 0 15 20 0 10 15 0 20 15 0 5 15 0];
Cost{C,C}:= [5 14 13, 14 5 9, 13 9 10] "Comm. costs";
Comm{D,D}:=[(A,C) 1.0, (A,D) 1.5, (B,C) 1.4, (B,D) 1.2,
(C,E) 2.0, (D,E) 0.7] "quant. of communications";
set CO{i,j,k,l} := k>i and Comm[i,k];
binary variable
d{D,C} "departement d is in city c";
g{CO}; -- g[CO[i,j,k,l] <-> d[i,j] and d[k,l];
constraint
-- each department must be located in exactly one city
Dept{D}: sum{C} d = 1;
-- at most 3 departments be located in the same city
City{C}: atmost(3){D} d;
La{CO[i,j,k,l]}: (g[CO]->d[i,j]) and (g[CO]->d[k,l]);
Lc{CO[i,j,k,l]}: d[i,j] and d[k,l] -> g[CO];
minimize TCOST: -(sum{D,C|C<>'London'} Benefit*d)
+ sum{CO[i,j,k,l]} Comm[i,k]*Cost[j,l]*g[CO];
Write('Total communication cost %d\n\n', -1000*TCOST);
Write{C|exist{D} d}('Locate departements %2s in %s\n',
{D|d} D , C);
end

This problem is a quadratic assignment problem and objective function in fact

would be:
minimize TCOST: -(sum(D,C|C<>'London') Benefit*d
+sum([i,j,k,l]=CO) Comm[i,k]*Cost[j,l]*d[i,k]*d[j,l];
To solve this with a standard linear solver, it is possible to reformulate the

model as a linear integer program by the following steps
1. replace xy by a new 0-1 variable z (x and y are also 0-1 variables)

2. impose the logical condition: z = 1 ↔ x = 1 ∧ y = 1 by means of the

three constraints: z − x ≤ 0 , z − y ≤ 0 , x + y − z ≤ 1.
An alternative is the objective in the model code together with the two con-
straints La and Lb .
2.12. Curve Fitting (linear) (will11)

Problem: Two quantities x and y are known to be correlated by a straight line
y = ax + b. A number of corresponding values are known.
1. Find a straight line, such that the sum of absolute deviation of all cor-
responding values are minimized.
2. Find a straight line, such that the maximum deviation is minimized.
(Note that the standard way to model this is by regression and we minimize
the sum of the quadratic deviations. In this approach different measures of
“deviations” are used.)
Modeling Steps
The problem is from [17]. For a problem description see Chapter 13.11.
Define the data: A set I of points (x, y) is given. Model this be two vectors xi
and yi , for each point i ∈ I.
1. The goal is to find a linear function: y = Ax + B where A and B are

unknown.
2. For each given point, we introduce a unknown positive P and negative
N deviation from the line y = Ax + B. Hence, the following condition
must hold:
yi = Axi + B + Pi − Ni for all i ∈ I
3. In a first round we minimize the sum of all deviations, that is:

X
min (Pi + Ni )
i
4. To minimize the maximal deviation, we introduce a new variable maxdev

and add the conditions maxdev ≥ Pi and maxdev ≥ Ni for all points
i. Now maxdev is larger than any Pi or Ni , that means it defines the
maximal deviation. Finally, we minimize maxdev.

model Will11 "Curve Fitting (linear)";
set i := 1..19; -- 19 points are known
parameter
x{i}:=[0.0 0.5 1.0 1.5 1.9 2.5 3.0 3.5 4.0 4.5
5.0 5.5 6.0 6.6 7.0 7.6 8.5 9.0 10.0];
y{i}:=[1.0 0.9 0.7 1.5 2.0 2.4 3.2 2.0 2.7 3.5,
1.0 4.0 3.6 2.7 5.7 4.6 6.0 6.8 7.3];

variable
A [-10..10]; B [-10..10]; maxdev;
P{i}; N{i};
constraint
DEV{i}: y = A*x + B + P - N;
ULA{i}: maxdev >= P;
ULB{i}: maxdev >= N;
minimize deviationSum: sum{i} (P+N);
Write('Min. deviation sum: y=%7.4f*x + %7.4f\n', A,B);
parameter a:=A; b:=B;
minimize minmax: maxdev;
Write('Min. max deviation: y=%7.4f*x + %7.4f\n', A,B);
Draw.Scale(40,-40);
x{i}:=x+1; y{i}:=y+1;
{i} Draw.Circle(x,y,.07);
Draw.Line(0,b,13,a*13+b,5);
Draw.Line(0,B,13,A*13+B,11);
end

Solution: The solution is given as follows:

Minimize deviation sum: y = 0.64*x + 0.58
Minimize maximum deviation: y = 0.63*x + -0.40
The two functions are depicted in Figure 2.1
Figure 2.1: Linear Solution
Question (Answer see 2.12)
1. Suppose now that a quadratic (not a linear) function y = Ax2 + Bx + C

has to be found. How should A, B, and C be chosen such that (1) the
sum of absolute deviation, (2) the maximum deviation is minimized?
Answer (Question see 2.12)
1. The only part in the model that changes is the constraint DEV. It must
be replaced by the following constraint:
DEV{i}: y = A*x^2 + B*x + C + P - N;
The model can be found and ran at: will11a3 .
The solution is:

Minimize deviation sum: y = 0.03*x^2 + 0.29*x + 0.98
Minimize max deviation: y = 0.13*x^2 - 0.62*x + 2.47
The two function are depicted in Figure 2.2
Figure 2.2: Quadratic Solution
3
http://lpl.virtual-optima.com/lpl/Solver.jsp?name=/will11a
2.13. Curve Fitting (quadratic) (will11a)

Problem: This model calculates the best quadratic curve fitting. For a problem
description see model will114 .


model Will11a "Curve Fitting (quadratic)";
set i := 1..19; -- 19 points are given
parameter
x{i}:=[0.0 0.5 1.0 1.5 1.9 2.5 3.0 3.5 4.0 4.5
5.0 5.5 6.0 6.6 7.0 7.6 8.5 9.0 10.0];
y{i}:=[1.0 0.9 0.7 1.5 2.0 2.4 3.2 2.0 2.7 3.5,
1.0 4.0 3.6 2.7 5.7 4.6 6.0 6.8 7.3];
variable
A [-10..10]; B [-10..10]; C [-10..10]; maxdev;
P{i}; N{i};
constraint
DEV{i}: y = A*x^2 + B*x + C + P - N;
ULA{i}: maxdev - P >= 0;
ULB{i}: maxdev - N >= 0;
minimize deviationSum: sum{i} (P+N);
Write('Min dev sum: y = %7.4f*x*x + %7.4f*x + %7.4f\n',
A,B,C);
parameter a:=A; b:=B; c:=C;
minimize minmax: maxdev;
Write('Min max dev: y = %7.4f*x*x + %7.4f*x + %7.4f\n',
A,B,C);
x{i}:=x+1; y{i}:=y+1;
Draw.Scale(20,-20);
{i} Draw.Circle(x,y,.07);
{i in 1..12}
Draw.Line(i-1,a*(i-1)^2+b*(i-1)+c,i,a*i^2+b*i+c);
{i in 1..12}
Draw.Line(i-1,A*(i-1)^2+B*(i-1)+C,i,A*i^2+B*i+C,5);
end

4
2.14. Logical Design (will12)

Problem: A logical circuit has to be designed on the base of nor-gates such
that it can perform a certain logical function. We may suppose that 7 gates
will do it. The gates are connected as specified by a InOut-table. How many
gates and which ones (of the seven) are needed to produce a given final output
– here the xor-gate ?
scription see Chapter 13.12. This problem also appeared in Williams H.P., Ex-
periments in the formulation of integer programming problems, Math. Prog.
Study, 2, pp.180-197.

model Will12 "Logical Design";
set
i,j := [Gate1 Gate2 Gate3 Gate4 Gate5 Gate6 Gate7];
J:= [1 2]; --maximal number of inputs on a gate
K:= [1 2 3 4]; --4 combi. are given by a true table
InOut{i,j} :=
[Gate1 Gate2 , Gate1 Gate3
Gate2 Gate4 , Gate2 Gate5
Gate3 Gate6 , Gate3 Gate7];
parameter
ALPHA{K,J}:= [ 0,0 0,1 1,0 1,1 ];
Output{K} := [ 0 1 1 0 ]; -- this is the XOR gate
binary variable
s{i}; -- =1 if the gate is needed, else 0
t{i,J}; -- left and right input signal at a gate
OUT{i,K}; -- output signal on the four combinations
constraint
ENOR{i,J}: s-t >= 0;
INNOR{i|exist{j} InOut[i,j]}:
sum{j|InOut[i,j]} s[j] + sum{J} t[i,J] <= 2;
LO{K,i,j|InOut[i,j]}: OUT[i,K] + OUT[j,K] <= 1;
LC{K,i,J}: ALPHA*t+OUT <= 1;
LOGC{K,i}: sum{j|InOut[i,j]} OUT[j,K]
+sum{J} ALPHA*t + OUT - s >= 0;
POUT{i,K}: s-OUT >= 0;
BOUNDS{K}: OUT[1,K] = Output[K];
BOUNDS1: s[1] >= 1; --force a gate to exist to avoid a
trivial solution
minimize NumberOfGates: sum{i} s;
Writep(s,t);
end

Explanation: The nor-gate is a device that has exactly two entries a and b with
value 0 or 1, and exactly one output c with value 0 or 1. The device simulates
the Boolean nor (not or) function: c = ¬(a ∨ b). We use also the operator | for
the nor operation, hence, it can also be written as c = a|b (Figure 2.3).
Figure 2.3: The nor-operation, and the nor gate
Every Boolean operation can be expressed by a combination of nor-operations.

For example, we have for the and-operation as:
a ∧ b ≡ (a|0)|(b|0)
The xor-operation can be modeled as:
a∨
˙ b ≡ ((a|0)|(b|0)) | (a|b)
Figure 2.4: The and/xor circuit
Note that the solution of the model exposes exactly the circuit given in Figure
2.4.
2.15. Market Sharing I (will13)

Problem: A distribution company with two division A and B has to allocate
market share between its divisions. The company supplies 23 retailers with oil
and spirit. Division A should control 40 (plus or minus 5%) of the company’s
operation with respect of the following criteria:
1. number of delivery points

2. spirit market
3. oil market in region 1
6. number of retailer with good/bad prospects
scription see Chapter 13.13. This problem also appeared in Williams H.P., Ex-
Study, 2, pp.180-197.

model Will13 "Market Sharing I";
set r := 1..23; --all retailers
r1{r}:= [1 2 3 4 5 6 7 8]; --in 1st region
r2{r}:= [9 10 11 12 13 14 15 16 17 18];--in 2nd region
r3{r}:= [19 20 21 22 23]; --in 3rd region
k := [1 2 3 4 5 6]; --six criteria
l := [delivery spirit oil prospects];
parameter T{l,r} :=
[11,47 ,44,25 ,10,26 ,26,54 ,18 ,51,20,105, 7,16,34,
100,50 ,21,11,19,14,10,11 --number of deliveries
34,411,82,157,5 ,183,14,215,102,21,54, 0 , 6,96,118,
112,535,8 ,53,28,69,65,27 --size of spirit market
9, 13,14, 17,18, 19,23, 21, 9,11,17, 18,18,17, 22,
24, 36,43, 6,15,15,25,39 --size of oil market
1, 1, 1, 0, 1, 1, 0, 0, 0, 1, 0, 0, 0, 0, 1,
0, 0, 0, 0, 1, 0, 0, 0]; --good/bad prospects
Total{k} := [730, 2395, 134, 215, 100, 8];
variable
P{k}; -- positive deviation
N{k}; -- negative deviation
MAXDEV [0..120];
binary M{r}; -- 1 if retailer is assigned to division A
constraint
DELIVER:sum{r} T[1,r] *M[r] +P[1]-N[1] = 0.4*Total[1];

SPIRITS:sum{r} T[2,r] *M[r] +P[2]-N[2] = 0.4*Total[2];
OILA: sum{r1} T[3,r1]*M[r1]+P[3]-N[3] = 0.4*Total[3];
OILB: sum{r2} T[3,r2]*M[r2]+P[4]-N[4] = 0.4*Total[4];
OILC: sum{r3} T[3,r3]*M[r3]+P[5]-N[5] = 0.4*Total[5];
GROWTH: sum{r} T[4,r] *M[r] +P[6]-N[6] = 0.4*Total[6];
DEVA{k}: MAXDEV-P >= 0;
DEVB{k}: MAXDEV-N >= 0;
BOUND1{k}: P <= 0.05*Total >= N;
--minimize SUMDEV: sum{k} Total*(P+N) + 8*(P[6]+N[6]);
--Writep(SUMDEV,MAXDEV,P,N,M);
minimize minmax: MAXDEV;
Writep(MAXDEV,P,N,M);
end

2.16. Market Sharing II (will13a)

Problem: A distribution company with two division A and B has to allocate
market share between its divisions. The company supplies 23 retailers with oil
and spirit. Division A should control 40 (plus or minus 5%) of the company’s
operation with respect of the following criteria
1. number of delivery points

2. spirit market
6. number of retailer with good prospects
7. number of retailer with good/bad prospects
scription see Chapter 12.13. This problem also appeared in WILLIAMS H.P.,
Experiments in the formulation of integer programming problems, Math. Prog.
Study, 2, pp.180-197.
Extra tighter forms of constraints OILA, OILB, OILC have been added. These
have been derived by LP using roofs and ceilings of the original constraints.

model Will13a "Market Sharing II";
set r := 1..23; --all retailers
r1{r}:= [1 2 3 4 5 6 7 8]; --in 1st region
r2{r}:= [9 10 11 12 13 14 15 16 17 18];--in 2nd region
r3{r}:= [19 20 21 22 23]; --in 3rd region
k := [1 2 3 4 5 6]; --six criteria
l := [delivery spirit oil prospects];
parameter T{l,r} :=
[11,47 ,44,25 ,10,26 ,26,54 ,18 ,51,20,105, 7,16,34,
100,50 ,21,11,19,14,10,11 --number of deliveries
34,411,82,157,5 ,183,14,215,102,21,54, 0 , 6,96,118,
112,535,8 ,53,28,69,65,27 --size of spirit market
9, 13,14, 17,18, 19,23, 21, 9,11,17, 18,18,17, 22,
24, 36,43, 6,15,15,25,39 --size of oil market
1, 1, 1, 0, 1, 1, 0, 0, 0, 1, 0, 0, 0, 0, 1,
0, 0, 0, 0, 1, 0, 0, 0]; --good/bad prospects
Total{k} := [730, 2395, 134, 215, 100, 8];
-- tighter lower and upper bounds
UOA{r}:=[5,7,8,9,10,10,13,12];
LOA{r}:=[2,3,3,4, 4, 4, 5, 5];
UOB{r}:=[. . . . . . . . 8,8,13,14,14,13,17,19,28,34];
LOB{r}:=[. . . . . . . . 8,9,14,15,15,14,18,20,30,36];
UOC{r}:=[. . . . . . . . . . . . . . . . . .
1,2,2,3,4];
LOC{r}:=[. . . . . . . . . . . . . . . . . .
1,2,2,3,5];
variable
P{k}; -- positive deviation
N{k}; -- negative deviation
MAXDEV [0..120];
binary M{r}; -- 1 if retailer is assigned to division
A
constraint
DELIVER: sum{r} T[1,r] *M[r] +P[1]-N[1]=0.4*Total[1];
SPIRITS: sum{r} T[2,r] *M[r] +P[2]-N[2]=0.4*Total[2];
OILA: sum{r1} T[3,r1]*M[r1]+P[3]-N[3]=0.4*Total[3];
OILB: sum{r2} T[3,r2]*M[r2]+P[4]-N[4]=0.4*Total[4];
OILC: sum{r3} T[3,r3]*M[r3]+P[5]-N[5]=0.4*Total[5];
GROWTH: sum{r} T[4,r] *M[r] +P[6]-N[6]=0.4*Total[6];
NUOILA: sum{r1} UOA[r1]*M[r1] <= 33;
NLOILA: sum{r1} LOA[r1]*M[r1] >= 11;
NUOILB: sum{r2} UOB[r2]*M[r2] <= 75;
NLOILB: sum{r2} LOB[r2]*M[r2] >= 63;
NUOILC: sum{r3} UOC[r3]*M[r3] <= 5;
NLOILC: sum{r3} LOC[r3]*M[r3] >= 5;
DEVA{k}: MAXDEV-P >= 0;
DEVB{k}: MAXDEV-N >= 0;
BOUND1{k}: P <= 0.05*Total >= N;
--minimize SUMDEV: sum{k} Total*(P+N) + 8*(P[6]+N[6]);
--Writep(SUMDEV,MAXDEV,P,N,M);
minimize MINMAX: MAXDEV;
Writep(MAXDEV,P,N,M);
end

2.17. Opencast Mining I (will14)

Problem:
| | | | | 16 blocks (B15-B30)
-------------------
| | | | 9 blocks (B6-B14)
-----------------
| | | 4 blocks (B2-B5)
---------------
| | 1 block (B1)
The problem is to opencast mine within a limited area. Ore body and over-
burden are divided into blocks (e.g. 50 foot cubes). Each block has a certain
revenue and a cost of extracting it. There is a permissible angle of slip (45
degree), so that a block can only be extracted if the (four) blocks above it have
also been removed. Net income must be maximized.
Model: The Model is as follows:he problem is from [17]. For a problem de-
scription see Chapter 13.14. This problem appeared also in Williams H.P., Ex-
Study, 2, pp.180-197.
We define a binary variable Di for each block i which is 1 if the block is ex-
tracted. The only restriction is that every block i laying on the top of a specific
block j must also be removed. The condition Di = 1 ↔ Dj = 1 where i is
a block above j means, that block i must be removed if and only if block j is
removed. This can be modeled as Di − Dj >= 0.

model Will14 "Opencast Mining I";
set i,j := [B1 B2 B3 B4 B5 B6 B7 B8 B9 B10 B11 B12 B13
B14 B15 B16 B17 B18 B19 B20 B21 B22 B23 B24 B25
B26 B27 B28 B29 B30] "30 Blocks on 4 levels";
LayingOn{i,j} :=
[ (B1 ,*) B2 B3 B4 B5 (B2 ,*) B6 B7 B9 B10
(B3 ,*) B7 B8 B10 B11 (B4 ,*) B9 B10 B12 B13
(B5 ,*) B10 B11 B13 B14 (B6 ,*) B15 B16 B19 B20
(B7 ,*) B16 B17 B20 B22 (B8 ,*) B17 B18 B21 B22
(B9 ,*) B19 B20 B23 B24 (B10,*) B20 B21 B24 B25
(B11,*) B21 B22 B25 B26 (B12,*) B23 B24 B27 B28
(B13,*) B24 B25 B28 B29 (B14,*) B25 B26 B29 B30 ];
parameter
-- Revenue of a block if excavated
Rev{i}:=[12 , 10 8 24 12, 4 4 1 6 6 2 8 8 4,
1.5,1.5,1,0.5,2,2,1.5,1,3,4,3,1.5,3,3,3,1.5];
-- Cost of a block if excavated
Cost{i}:=[10, 8 8 8 8, 6 6 6 6 6 6 6 6 6,
3 3 3 3,3,3, 3,3,3,3,3, 3,3,3,3, 3];
binary variable D{i} "1 if the block is extracted";
constraint CC{LayingOn[i,j]}: D[i]-D[j] <= 0;
maximize pr: sum{i} (Rev-Cost)*D;
Write('Profit=%4.1f\nExcavated blocks:%4s\n',pr,{i|D} i);
end

2.18. Opencast Mining II (will14a)

Problem: In an opencast mining 30 blocks have to be excavated potential.
Excavating one block cost an amount of money but returns also an amount of
money. The problem is to select the subset of blocks such that total profit is
maximal. The restriction is that if a block is excavated then all blocks laying
on it must also be excavated.
Modeling Steps
This is the same problem as defined in will145 . There the problem is formu-
lated as a integer program, that is, as a mathematical model using integer
variables. It has been shown that the problem can be translated into a max
flow problem in a digraph. See [27] p. 731ff, and [6] p. 49, where the prob-
lem is to find the optimal closure in a digraph. The problem is to choose an
optimal subset of “projects” where each project has a (positive or negative)
benefit. There are restrictions of the form: if a certain project is chosen then
some other projects must also be chosen. The problem is to choose the subset
with the maximal total benefit.
The problem of open mining can be formulated by creating the following di-
graph G(V, A). Let the vertices be the set of the blocks (V). Each block has a
benefit (profit) bi with i ∈ V (return-cost). There is an arc (i, j), if choosing
block i implies choosing block j, that is, block j laying on block i. The arc set
is (i, j) ∈ A.
To formulate this problem as an max-flow problem, we build another graph
G 0 (V 0 , A 0 ) as follows: Add two vertices s (source) and t (sink): (V 0 = V ∪ {s, t}).
For each vertex j ∈ V|bj > 0, add an arc (s, j) to G 0 with capacity bj . For each
vertex i ∈ V|bi < 0, add an arc (i, t) to G 0 with capacity −bi . For each arc
(i, j) ∈ E add an arc to G 0 with capacity ∞. Hence:
E 0 = E ∪ {(s, j)|j ∈ V, bj > 0} ∪ {(i, t)|j ∈ V, bi < 0}
Now calculate the max-flow from source s to sink t.

model Will14a "Opencast Mining II";
set i,j := [B1 B2 B3 B4 B5 B6 B7 B8 B9 B10 B11 B12 B13
B14 B15 B16 B17 B18 B19 B20 B21 B22 B23 B24 B25
B26 B27 B28 B29 B30] "30 Blocks on 4 levels";
LayingOn{i,j} :=
5
[ (B1 ,*) B2 B3 B4 B5 (B2 ,*) B6 B7 B9 B10

(B3 ,*) B7 B8 B10 B11 (B4 ,*) B9 B10 B12 B13
(B5 ,*) B10 B11 B13 B14 (B6 ,*) B15 B16 B19 B20
(B7 ,*) B16 B17 B20 B22 (B8 ,*) B17 B18 B21 B22
(B9 ,*) B19 B20 B23 B24 (B10,*) B20 B21 B24 B25
(B11,*) B21 B22 B25 B26 (B12,*) B23 B24 B27 B28
(B13,*) B24 B25 B28 B29 (B14,*) B25 B26 B29 B30];
parameter
Rev{i}:=[12 , 10 8 24 12, 4 4 1 6 6 2 8 8 4,
1.5,1.5,1,0.5,2,2,1.5,1,3,4,3,1.5,3,3,3,1.5];
Cost{i}:=[10, 8 8 8 8, 6 6 6 6 6 6 6 6 6,
3 3 3 3,3,3, 3,3,3,3,3, 3,3,3,3, 3];
Profit{i}:=Rev-Cost;
Addm(i,'s'); Addm(i,'t'); //add source/sink
parameter e{i,j}:=
if(LayingOn,99,
j=#j and i<#i-1 and Profit[i]<0,-Profit[i],
i=#i-1 and j<#j-1 and Profit[j]>=0,Profit[j]);;
e[#i,#i-1]:=100;
variable x{i,j|e} [0..e];
constraint A{i}: sum{j} x[i,j] - sum{j} x[j,i] = 0;
maximize obj: x[#i,#i-1];
parameter dd{i,j}:=GetValue(x,3);
set Dual{i,j|i<#i-1 and j<#j-1}:=GetValue(x,3);
set S{i}:= exist{j} Dual[j,i];
{i,j} if(S and LayingOn,S[j]:=1);
// draw the result as a graph G'
Draw.Scale(20,20);
parameter X{i}; Y{i};; --coordinates for the drawing
X[31]:=0; Y[31]:=20; --source s
X[1]:=10; Y[1]:=15; --1st layer
{i|i>1 and i<=5}(X[i]:=20, Y[i]:=3*(i-2)+10); --2nd
{i|i>5 and i<=14}(X[i]:=30, Y[i]:=3*(i-6)+5); --3rd
{i|i>14 and i<=30}(X[i]:=40, Y[i]:=2.5*(i-15)); --4th
X[32]:=50; Y[32]:=20; --sink t
{e[i,j]} Draw.Arrow(if(e=99,'',''&Round(e,-1)),
X[i],Y[i],X[j],Y[j],1);
{e[i,j]|S and j<#j-1} Draw.Arrow(if(e=99,'',''&
Round(e,-1)),X[i],Y[i],X[j],Y[j],1,3,3);
{i} Draw.Circle(''&i,X,Y,1,if(S,3,1),0);
Draw.Text('Red Blocks are excavated',0,3,30);
Draw.Text('Profit='&sum{S[i]} Profit,0,5,30);
end

Solution: The total benefit is 17.5, the same result as in will14. To find the
blocks to excavate, we inspect the dual variables from which we can recon-
struct the set of the blocks.
2.19. Opencast Mining III (will14b)

Problem:
| | | | | 16 blocks (B15-B30)
-------------------
| | | | 9 blocks (B6-B14)
-----------------
| | | 4 blocks (B2-B5)
---------------
| | 1 block (B1)
The problem is to opencast mine within a limited area. Ore body and over-
burden are divided into blocks (e.g. 50 foot cubes). Each block has a certain
revenue and a cost of extracting it. There is a permissible angle of slip (45
degree), so that a block can only be extracted if the (four) blocks above it have
also been removed. We define a binary variable D for each block which is 1 if
the block is extracted. The return on investment must be maximized.

description see Chapter 13.14. This problem is the same as will146 except that
not the net income but the return on investment must be maximized. This
problem appeared also in Williams H.P., Experiments in the formulation of
integer programming problems, Math. Prog. Study, 2, pp.180-197.
Return on investment can be expressed as the ratio given by:

P
Revi Di
Pi
i Costi Di
This function is not linear. In order to linearize it, we need to transform this
function by the following steps:
1. Introduce a new continuous variable y.

2. Equating y to the return of investment and transform it gives:
X X
Costi Di y − Revi Di = 0
i i
3. Replace the expression Di y by a new variable zi (for each block i).

6
4. Impose the condition z ↔ Di y by the following three constraints:
zi ≤ 10Di , zi ≤ y , −zi + y + 10Di ≤ 10 , for all i
To avoid of getting a 0/0 objective value by excavating nothing, we impose

that at least one block must be excavated.

model Will14b "Opencast Mining III";
set i,j "30 Blocks on 4 levels";
LayingOn{i,j} :=
[(B1 ,*) B2 B3 B4 B5 (B2 ,*) B6 B7 B9 B10
(B3 ,*) B7 B8 B10 B11 (B4 ,*) B9 B10 B12 B13
(B5 ,*) B10 B11 B13 B14 (B6 ,*) B15 B16 B19 B20
(B7 ,*) B16 B17 B20 B22 (B8 ,*) B17 B18 B21 B22
(B9 ,*) B19 B20 B23 B24 (B10,*) B20 B21 B24 B25
(B11,*) B21 B22 B25 B26 (B12,*) B23 B24 B27 B28
(B13,*) B24 B25 B28 B29 (B14,*) B25 B26 B29 B30];
parameter
Rev{i}:=[12 , 10 8 24 12, 4 4 1 6 6 2 8 8 4,
1.5,1.5,1,0.5,2,2,1.5,1,3,4,3,1.5,3,3,3,1.5];
Cost{i}:=[10, 8 8 8 8, 6 6 6 6 6 6 6 6 6,
3 3 3 3,3,3, 3,3,3,3,3, 3,3,3,3, 3];
variable y; z{i}; binary D{i};
constraint
R1: sum{i} Cost*z - sum{i} Rev*D = 0;
R2{i}: z - 10*D <= 0;
R3{i}: z - y <= 0;
R4{i}: -z + y + 10*D <= 10;
R5: sum{i} D >= 1;
CC{LayingOn[i,j]}: D[i] - D[j] <= 0;
maximize return_on_investment: y;
Writep(return_on_investment,D);
end

2.20. Tariff Rates (Power Generation) (will15)

Problem: This model concerns the scheduling of generators to meet different
demands for electricity throughout the day. There are three generators that
can be started and shut down at the beginning of five time intervalls per day.
These generators have different running and start up costs and differ also in
minimal and maximal electricity output. Minimize the cost while fullfilling
the electricity demand during a day.
scription see Chapter 13.15. The model appeared also in DAY R.E., Williams
H.P., MAGIC: The Design and Use of an Interactive Modelling Language for
Mathematical Programming, IMA Journal of Mathematics in Management
(1986) 1, pp.53-65.

model Will15 "Tariff Rates (Power Generation)";
set i:=[G1 G2 G3] "generators";
t:=[pm12_am6 am6_am9 am9_pm3 pm_pm6 pm6_pm12] "
periods";
parameter
minl{i} := [0.85 1.25 1.50];
maxl{i} := [2.00 1.75 4.00];
costph{i} := [1.00 2.60 3.00];
excost{i} := [2.00 1.30 3.00];
kctminl{i}:= [1.70 1.625 4.5];
startcost{i}:= [2 1 .5];
number{i} := [12 10 5];
demand{t} := [15 30 25 40 27];
nhours{t} := [ 6 3 6 3 6];
variable
out{i,t} "Output of a generator typ in each period";
integer num{i,t} "Number of generators working";
integer nst{i,t} "Number of generators starting";
constraint
dem{t} "Demand must be meet at each period"
: sum{i} out >= demand;
guar{t} "a 15% extra output margin must be imposed"
: sum{i} maxl*num >= 1.15*demand;
output{i,t} "Output level must lie between min/max"
: minl*num <= out <= maxl*num;
st{i,t} "The number of generators starting"
: nst >= num - num[i,(#t+t-2)%#t+1];
gennum{i,t} "the number of generators are limited"
: num <= number >= nst;
minimize cost: sum{i,t} (excost*nhours*out
+ startcost*nst + (costph-kctminl)*nhours*num);
Writep(cost,out,num,nst);
end

2.21. Hydro power (will16)

Problem: This model concerns the scheduling of generators to meet different
demands for electricity throughout the day. There are three thermal genera-
tor types and two two hydro generators that can be started and shut down at
the beginning of five time intervalls per day. These generators have different
running and start up costs and they differ also in minimal and maximal elec-
tricity output. The thermal generators can pump water back to the reservoir.
The water level of the reservoir must be 16 at the beginning of day and must
be between 15 and 20 during the day. Minimize the cost while fullfilling the
electricity demand during a day.

description see Chapter 13.16. This is an extended model of will157 . The
extension concerns the hydro generators and revervoir.

model Will16 "Hydro power";
set i:=[G1 G2 G3] "generators";
t:=[pm12_am6 am6_am9 am9_pm3 pm_pm6 pm6_pm12] "
periods";
h:= 1..2;
parameter
minl{i} := [0.85 1.25 1.50];
maxl{i} := [2.00 1.75 4.00];
costph{i} := [1.00 2.60 3.00];
excost{i} := [2.00 1.30 3.00];
kctminl{i} := [1.70 1.625 4.5];
startcost{i} := [2 1 .5];
number{i} := [12 10 5];
demand{t} := [15 30 25 40 27];
nhours{t} := [ 6 3 6 3 6];
hydrolevel{h} := [0.9,1.4];
startcosth{h} := [1.5,1.2];
costphh{h} := [0.09,0.15];
numhyd{h} := [1,1];
redheight{h} := [0.31,0.47];
variable
out{i,t} "Output of a generator";
heit{t} "reservoir depth level";
pump{t} "pumped quantity to reservoir";
integer
num{i,t} [0..number];
nst{i,t} [0..number];
numh{h,t} [0..numhyd];
7
nsth{h,t} [0..numhyd];
constraint
Dem{t}: sum{i} out +sum{h} hydrolevel*numh -pump >=
demand;
Guar{t}: sum{i} maxl*num >= 1.15*demand-sum{h}
hydrolevel;
output{i,t}: minl*num <= out[i,t] <= maxl*num;
St{i,t}: nst >= num - num[i,if(t=1,#t,t-1)];
Sth{h,t}: nsth >= numh - numh[h,if(t=1,#t,t-1)];
High{t}: heit[if(t<#t,t+1,1)] - heit - (nhours/3)*pump
+sum{h} nhours*redheight*numh = 0;
Bounds1: heit[1] = 16;
Bounds2{t|t>1}: 15 <= heit <= 20;
minimize cost: sum{i,t}(excost*nhours*out
+ startcost*nst + (costph-kctminl)*nhours*num)
+ sum{h,t} (costphh*nhours*numh + startcosth*nsth);
Writep(cost,num,nst,out);
end

2.22. T3-dimensional Noughts and Crosses (will17)

Problem: Given a 3x3x3 cube consisting of 27 cells, the problem is to fill each
cell with either a black or a white ball so as to minimize the number of ’straight
lines’ in the cube containing balls of identical colour. A straight line is a row
of 3 cells including all possible diagonals. (There are 49 such lines in all). The
cells are numbered as indicated.
/-------/
/ / |
/-------/ |
| 1 2 3 | |
| 4 5 6 | /
| 7 8 9 | /
/-------/
Model: The Model is as follows:

There exists 227 feasible integer solution, clearly too many to be explored sys-
tematically. The symmetry could be exploit to restrict the number. We in-
troduce a binary variable D, which is 1(0), if the cell contains a black(white)
ball. Furthermore we introduce a 0-1 variable G for each possible straight line,
which is 1, if the line contains only balls of the same color. The sum of these
variables G must be minimized. For each of the 49 lines (for example cells 1,
2, and 3) 2 constraints are introduced as follows:
D[1]+D[2]+D[3]-G <= 2 "at most 2 of the ball are black"
D[1]+D[2]+D[3]+G >= 1 "at least 1 ball is black"
to model the condition

((D[1]+D[2]+D[3]=3) or (D[1]+D[2]+D[3]=0)) <--> G=1
which means: G is 1, if and only if all three cells of the line are black (sum is
3) or all three cells are white (sum is 0).
The problem is from [17]. For a problem description see Chapter 13.17. This
problem appeared first in Williams H.P., Experiments in the formulation of
integer programming problems, Math. Prog. Study, 2, pp.180-197.

model Will17 "T3-dimensional Noughts and Crosses";
set p "49 possible straight lines";
c "27 cells";
Lines{p,c} := -- 49 lines containing 3 cells each
[(1,*) 1 2 3 (2,*) 4 5 6 (3,*) 7 8 9 --x

(4,*) 10 11 12 (5,*) 13 14 15 (6,*) 16 17 18
(7,*) 19 20 21 (8,*) 22 23 24 (9,*) 25 26 27
(10,*) 1 4 7 (11,*) 2 5 8 (12,*) 3 6 9 --y
(13,*) 10 13 16 (14,*) 11 14 17 (15,*) 12 15 18
(16,*) 19 22 25 (17,*) 20 23 26 (18,*) 21 24 27
(19,*) 1 10 19 (20,*) 2 11 20 (21,*) 3 12 21 --z
(22,*) 4 13 22 (23,*) 5 14 23 (24,*) 6 15 24
(25,*) 7 16 25 (26,*) 8 17 26 (27,*) 9 18 27
(28,*) 1 5 9 (29,*) 3 5 7 --xy-diagonals
(30,*) 10 14 18 (31,*) 12 14 16
(32,*) 19 23 27 (33,*) 21 23 25
(34,*) 1 13 25 (35,*) 7 13 19 --yz-diagolals
(36,*) 2 14 26 (37,*) 8 14 20
(38,*) 3 15 27 (39,*) 9 15 21
(40,*) 1 11 21 (41,*) 3 11 19 --zx-diagonals
(42,*) 4 14 24 (43,*) 6 14 22
(44,*) 7 17 27 (45,*) 9 17 25
(46,*) 1 14 27 (47,*) 3 14 25 --inner xyz diagonals
(48,*) 7 14 21 (49,*) 9 14 19 ];
binary variable
D{c}; --1 if cell has a black ball
G{p}; --1 if all balls in a line have the same color
constraint
LineA{p}: sum{c|Lines} D[c] - G <= 2;
LineB{p}: sum{c|Lines} D[c] + G >= 1;
Numb: sum{c} D = 14; -- 14 balls are black
minimize OBJ: sum{p} G;
Write(' %2d lines have the same color\n', sum {p}G);
Writep(D,G);
end

2.23. Optimizing a Constraint (will18)

Problem: In a integer programming problem with only 0-1 variables the fol-
lowing constraints occurs:
9x1 + 13x2 − 14x3 + 17x4 + 13x5 − 19x6 + 23x7 + 21x8 ≤ 37
We want to find an equivalent formulation of this constraints, but with a
smaller right-hand-side value. Furthermore, we want to find an equivalent
constraint, where the sum of the absolute values of all coefficients is mini-
mized.
Example: It can be easily checked, that the two following constraints are
equivalent, if all variables are 0-1 variables:
93x1 + 49x2 + 37x3 + 29x4 ≤ 111
is equivalent to
2x1 + x2 + x3 + x4 ≤ 2
Formulate the problem as a linear mathematical model (LP).
Modeling Steps
The problem can be formulated by a LP using the following transformation:
1. Translate the constraints in a standard form with only positive coeffi-

cients in descending order by substituting:
y1 = x7 , y2 = x8 , y3 = 1 − x6 , y4 = x4 , y5 = 1 − x3 ,
y6 = x5 , y7 = x2 , y8 = x1
this gives the standard form:
23y1 + 21y2 + 19y3 + 17y4 + 14y5 + 13y6 + 13y7 + 9y8 ≤ 70
2. We want to find another equivalent constraints of the form:

a1 y1 + a2 y2 + a3 y3 + a4 y4 + a5 y5 + a6 y6 + a7 y7 + a8 y8 ≤ a0
The coefficients ai become the variables of our problems.
3. We search for all subsets of the indices known as minimal “roof” and
maximal “ceiling”. A minimal roof is a subset of indices for which the
sum of the corresponding coefficients exceeds the right-hand side coef-
ficient and no proper subset exceeds it. A maximal ceiling is a subset of
indices for which the sum of the corresponding coefficients does not ex-
ceed the right-hand side and no subset properly containing it can also
be a ceiling.
4. All ’roofs’ and all ’ceiling’ are a constraint.

5. Furthermore we add the constraint: a1 ≥ a2 ≥ . . . ≥ a8 .
The problem is from [17]. For a problem description see Chapter 13.18.

model Will18 "Optimizing a Constraint";
set
j := [1 2 3 4 5 6 7 8]; --number of coefficients
c := [1 2 3 4 5 6]; --number of 'roofs'
r := [1 2 3 4 5 6]; --number of 'ceiling'
parameter
Ceiling{c,j}:=[1,2,3,0,0,0,0,0,
1,2,4,8,0,0,0,0,
1,2,6,7,0,0,0,0,
1,3,5,6,0,0,0,0,
2,3,4,6,0,0,0,0,
2,5,6,7,8,0,0,0];
Roofing{r,j}:=[1,2,3,8,0,0,0,0,
1,2,5,7,0,0,0,0,
1,3,4,7,0,0,0,0,
1,5,6,7,8,0,0,0,
2,3,4,5,0,0,0,0,
3,4,6,7,8,0,0,0];
variable a{j}; b;
constraint
CEiling{c}: sum{j} a[Ceiling] <= b;
ROofing{r}: sum{j} a[Roofing] >= b+1;
Order{j|j<#j}: a[j] >= a[j+1];
minimize Obja: b-a[3]-a[5];
Writep(a,b);
minimize Objb: sum{j} a;
Writep(a,b);
end

2.24. Distribution I (will19)

Problem: A company has two factories with a limited capacity and four de-
pots with a maximum throughtput capacity. It sells its products to six cus-
tomers with minimum requirements. The customers have some preferences
to be supplied from a depot or a factory. What would the distribution pat-
tern from factories to depots and customer be, if the overall cost are to be
minimized? Is it possible to meet the customers preferences?


model Will19 "Distribution I";
set f := [Liverpool,Brighton];
d := [Newcastle, Birmingham, London, Exeter];
c := [c1, c2, c3, c4, c5, c6];
parameter
FactoryCap{f}:= /Liverpool 150000, Brighton 200000/;
DepotCap{d}:= /Newcastle 70000, Birmingham 50000,
London 100000, Exeter 40000/;
CustomerReq{c} := /c1 50000, c2 10000, c3 40000,
c4 35000, c5 60000, c6 20000/;
FactoryDepotCost{f,d} :=
/Liverpool Newcastle 0.5,
Liverpool Birmingham 0.5,
Liverpool London 1.0,
Liverpool Exeter 0.2,
Brighton Birmingham 0.3,
Brighton London 0.5,
Brighton Exeter 0.2/;
FactoryCustCost{f,c} :=
/Liverpool c1 1.0,
Liverpool c3 1.5,
Liverpool c4 2.0,
Liverpool c6 1.0,
Brighton c1 2.0/;
DepotCustCost{d,c} :=
/Newcastle c2 1.5,
Newcastle c3 0.5,
Newcastle c4 1.5,
Newcastle c6 1.0,
Birmingham c1 1.0,
Birmingham c2 0.5,
Birmingham c3 0.5,
Birmingham c4 1.0,
Birmingham c5 0.5,
London c2 1.5,
London c3 2.0,
London c5 0.5,
London c6 1.5,
Exeter c3 0.2,
Exeter c4 1.5,
Exeter c5 0.5,
Exeter c6 1.5/;
--client preferences to be delivered from
PrefFtoC{f,c} := [0,1,1,1,1,1,
1,1,1,1,1,1];
--preferred depot of a customer
PrefDtoC{d,c} := [1,0,1,1,1,1,
1,1,1,1,0,1,
1,1,1,1,1,0,
1,1,1,1,1,0];
variable
FactoryDepotQty,FD{f,d|FactoryDepotCost};
FactoryCustQty,FC{f,c|FactoryCustCost};
DepotCustQty,DC{d,c|DepotCustCost};
constraint
FactoryCapa{f}: sum{d} FD + sum{c} FC <= FactoryCap;
DepotCapa{d}: sum{f} FD <= DepotCap;
DepotBalance{d}: sum{c} DC = sum{f} FD;
CustRequirement{c}: sum{f} FC + sum{d} DC =
CustomerReq;
minimize cost: sum{f,d} FactoryDepotCost*FD
+ sum{f,c}FactoryCustCost*FC + sum{d,c}DepotCustCost*
DC;
Writep(cost,FD,FC,DC);
minimize Pref: sum{f,c}PrefFtoC*FC + sum{d,c}PrefDtoC*
DC;
Writep(cost,FD,FC,DC);
end

Solution: Transportation optimum is: 198500.

2.25. Distribution II (will19a)

pots with a maximum throughtput capacity. It sells its products to six cus-
tomers with minimum requirements. The customers have some preferences
to be supplied from a depot or a factory. What would the distribution pat-
tern from factories to depots and customer be, if the overall cost are to be
minimized? Is it possible to meet the customers preferences?
scription see Chapter 13.19. This model is exactly the same as model will198 :

model Will19a "Distribution II";
set
F "factories" := [Liverpool Brighton];
D "depots" := [Newcastle Birmingham London Exeter
];
C "customers" := [c1 c2 c3 c4 c5 c6];
parameter -- distribution costs
FactCap{F} "factory capacity" := [150,200];
DepCap{D} "max. depot throughput" := [70,50,100,40];
CustReq{C}"customer requirements" :=
[50,10,40,35,60,20];
CostFtoD{F,D} := [0.5,0.5,1,0.2,
0,0.3,0.5,0.2];
CostFtoC{F,C} := [1,0,1.5,2,0,1,
2,0,0 ,0,0,0];
CostDtoC{D,C} := [0,1.5,0.5,1.5,0,1,
1,0.5,0.5,1,0.5,0,
0,1.5,2,0,0.5,1.5,
0,0,0.2,1.5,0.5,1.5];
PrefFtoC{F,C} := [0,1,1,1,1,1,
1,1,1,1,1,1];
PrefDtoC{D,C} := [1,0,1,1,1,1,
1,1,1,1,0,1,
1,1,1,1,1,0,
1,1,1,1,1,0];
variable
FtoD{F,D|CostFtoD} "quantity from factor to depot";
FtoC{F,C|CostFtoC} "quantity from factory to costomer
";
DtoC{D,C|CostDtoC} "quantity from depot to customer";
constraint
FCAP{F}: sum{D} FtoD +sum{C} FtoC <= FactCap;
8
DCAP{D}: sum{F} FtoD <= DepCap;

DCONT{D}: sum{C} DtoC = sum{F} FtoD;
CREQ{C}: sum{F} FtoC +sum{D} DtoC = CustReq;
minimize Cost: sum{F,D} CostFtoD*FtoD
+sum{F,C} CostFtoC*FtoC + sum{D,C} CostDtoC*DtoC;
Write('--model1: minimizing the costs\n');
Writep(Cost,FtoD,FtoC,DtoC);
minimize Pref: sum{F,C} PrefFtoC*FtoC +sum{D,C}
PrefDtoC*DtoC;
Write('--model2: meet customers preferences\n');
Writep(Pref,FtoD,FtoC,DtoC);
Write('%7.2f\n', sum {F,D}CostFtoD*FtoD+ sum {F,C}
CostFtoC*FtoC
+ sum {D,C}CostDtoC*DtoC);
end

2.26. Depot Location (Distribution 2) (will20)

pots (Newcastle, Birmingham, London, and Exeter) with a maximum throught-
put capacity. It sells its products to six customers with minimum require-
ments. The company has to decide to retain or close Newcastle and Exeter, to
expand Birmingham, or to build new ones at Bristol and Northampton. The
depot in London must be inchanged. In any case, there must be four depots.
There are savings as well as costs of doing so. What would the distribution
pattern from factories to depots and customer be, if the overall cost are to be
minimized, and which depots must be opened, closed or expand?

model Will20 "Depot Location (Distribution 2)";
set F "factories" := [Liverpool Brighton];
D "depots" := [Newcastle Birmingham London Exeter
Bristol Northampton];
C "customers" := [C1 C2 C3 C4 C5 C6];
parameter -- distribution costs
FactCap{F} "factory capacity" := [150,200];
DepCap{D} "max. throughput" := [70,20,100,40,30,25];
CustReq{C}"customer require." := [50,10,40,35,60,20];
CostFtoD{F,D} := [0.5,0.5,1,0.2,0.6,0.4
0,0.3,0.5,0.2,0.4,0.3];
CostFtoC{F,C} := [1,0,1.5,2,0,1, 2,0,0 ,0,0,0];
CostDtoC{D,C} := [
0,1.5,0.5,1.5,0,1, 1,0.5,0.5,1,0.5,0,
0,1.5,2,0,0.5,1.5, 0,0,0.2,1.5,0.5,1.5
1.2,0.6,0.4,0,0.3,0.8, 0,0.4,0,0.5,0.6,0.9];
FixCost{D} "cost of an action d" := [10,3,0,5,12,4];
variable
FtoD{F,D|CostFtoD} "quantity from factor to depot";
FtoC{F,C|CostFtoC} "quantity from factory to customer";
DtoC{D,C|CostDtoC} "quantity from depot to customer";
binary d{D|D<>3}; --1, if retained (Newcastle, Exeter)
--expand (Birmingham), build (Bristol, Nort.)
constraint
FCAP{F}: sum{D} FtoD +sum{C} FtoC <= FactCap;
DCAP{D}: sum{F} FtoD <= DepCap*d+if(D=2,50,D=3,DepCap);
DCONT{D}: sum{C} DtoC = sum{F} FtoD;
CREQ{C}: sum{F} FtoC +sum{D} DtoC = CustReq;
NUMB: sum{D|D<>4} d <= 2;
minimize Cost:sum{F,D}CostFtoD*FtoD+sum{F,C}CostFtoC*FtoC
+sum{D,C} CostDtoC*DtoC+sum{D} FixCost*d;
Writep(Cost,FtoD,FtoC,DtoC,d);
end

2.27. Agricultural Pricing I (will21)

Problem: The government of a country wants to fix the prices of its dairy
products milk, butter, and cheese. All these products comes from the coun-
try’s raw milk production. Parts of this production is exported or directly
comsumed. The rest is available in the form of its components: fat, dry mat-
ter, and water to produce the mentioned products. Comsumtion, prices of
previous years are known as well as the prices elasticities. The objective is
to determine what prices and resultant demand will maximize total revenue.
Politically, however, the prices should not rise to much: the cost of last year’s
consumtion should not increase. As an important result, the economic cost of
this political limitation should be quantified (through the shadow prices on
the indexing constraint). The data are as follows: Availability of fat and dry
matter are 600 0 000 and 750 0 000 tons.
Milk Butter Cheese 1 Cheese 2

Fat composition (%) 4 80 35 25
Dry M. (%) 9 2 30 40
previous year comsumption (1000t) 4820 320 210 70
previous year price (£/t) 297 720 1050 815
Price elasticity 0.4 2.7 1.1 0.4
Some cross elasticities are also given:
Cheese 1 Cheese 2
Cheese 1 0.1
Cheese 2 0.4
Recall that the elasticity of a product is defines as:
Percentage decrease in demand

E=
Percentage increase in price
And the cross elasticity of two products A and B is given as:
Percentage increase in demand for A

EAB =
Percentage increase in price for B
The problem is to fix prices in a way to maximize total revenue of the agricul-
tural sector. For a complete problem and model description see [17], Chapter
13.21.
Modeling Steps
The products are defined as a set i, j ∈ Products. The data are the percentage
content of fat (fati ), the percentage content of dry matter (dryi ), previous
year comsumption (coni ), previous year price (prii ), the elastisities (Ei ), and
the cross elastisities of cheese (E0i,j ).
1. We introduce two variables which defines the qunatity consumed xi

and the price to be fixed pi .
2. The total fat and dry matter produced limit the output:
X X
fati · xi ≤ 600000 , dryi · xi ≤ 750000
i i
3. The price index should not be higher than the previous year:
X X
coni · pi ≤ coni · prii
i i
4. The comsumtion xi is related to the prices pi through the elasticities:
dxi dpi X dpj

= −Ei + , for all i
xi pi pj
j|E0i,j
One can approximate the differntial equation (where z̄ is the known

previous year quantity – here consumption or price):
dz z − z̄
by
z z̄
Hence we have the linear equations:
xi − x̄i pi − p̄i X pj − p̄i
= −Ei + , for all i
x̄i p̄i p̄j
j|E0i,j
5. We would like to maximize the revenue for the agricultural sector:

X
max pi xi
i

model Will21 "Agricultural Pricing I";
set i,j:=[milk butter cheese1 cheese2];
parameter
fat{i} :=[4 80 35 25] "percentage composition";

dry{i} :=[9 2 30 40] "dry matter composition";
con{i} :=[4820 320 210 70] "past comsumpt. in 1000t";
pri{i} :=[297 720 1050 815] "past price of 100t";
E{i} :=[0.4 2.7 1.1 0.4] "price elastisitty";
F{i,j}:=[(cheese1,cheese2) 0.1, (cheese2,cheese1)
0.4];
variable x{i} [1..10000] "quantity comsumed";
variable p{i} [0..10000] "price";
constraint
Fat: sum{i} fat*x <= 600000;
Dry: sum{i} dry*x <= 750000;
Pri: sum{i} con*p <= sum{i} con*pri;
Ela{i}: (x-con)/con = - E*(p-pri)/pri
+ sum{j|F} F*(p[j]-pri[j])/pri[j];
maximize revenu: sum{i} p*x;
Writep(revenu,p,x);
Write('%12.6f \n', {i}x);
end

Further Comments: The model define so far is a non-linear model. Using

the elasticity relationship, we may substitute all xi variables, the remaining
model only contains variables pi . The resulting model can be formulated as a
convex quadrtaic model as follows – where Q is semi-definite positive matrix:
X
max pi Qi,j pj + Ai pi
i,j
subject to three linear constraints
linear constraints that make sure that comsumption is positive
The explicit model is given in WILL21a9 .
9
http://lpl.virtual-optima.com/lpl/Solver.jsp?name=/WILL21a
2.28. Agricultural Pricing II (will21a)

Problem: This model is a QP formulation of the model WILL2110
Model: The Model is as follows:

model Will21a "Agricultural Pricing II";
variable MILK; BUTT; CHA; CHB; //prices
constraint
QUALA: 260*MILK +960*BUTT +70.25*CHA -0.6*CHB >= 782;
QUALB: 584*MILK +24*BUTT +55.2*CHA +5.8*CHB >= 35;
INDEX: 4.82*MILK +0.32*BUTT +0.21*CHA +0.07*CHB <=
1.939;
NONNEG1: 220*CHA -26*CHB <= 420;
NONNEG2: -27*CHA +34*CHB <= 70;
NONNEG3: MILK<=1.039 and BUTT<=0.987;
maximize REVENUE: -6492*MILK*MILK -1200*BUTT*BUTT -220*
CHA*CHA
-34*CHB*CHB +53*CHA*CHB +6748*MILK +1184*BUTT +420*CHA
+70*CHB;
Write('Revenue=%d\n\nMilk = %d\nButter = %d\nCheese 1 = %
d\nChesse 2 = %d\n',
REVENUE,1000*MILK,1000*BUTT,1000*CHA,1000*CHB);
end

10
http://lpl.virtual-optima.com/lpl/Solver.jsp?name=/WILL21
2.29. Efficiency Analysis (EA) (will22)

Problem: This model shows the data envelopment analysis (DEA) approach
for measuring efficiency. A car manufacturer wants to evaluate the efficiency
of different garages, who have received a franchise to sell its cars. It does
this by weighing input against output, comparing “cost” with “returns”. In-
put and output can be measured by different criteria. In our case input is
measure by using 6 coefficients: (1) number of staff, (2) Show room space, (3)
catchment population in category I, (4) catchment population in category II,
(5) enquiries Alpha model (6) enquiries in Beta model. Output is measured
by three criteria: (1) Alpha model sales, (2) Beta model sales, (3) profit. We as-
sume proportionality, that is, a “doubling” of its inputs result in a “doubling”
of its output. Calculate which garages are the most efficient once.

description see Chapter 13.22. The approach used here is data envelopment
analysis (DEA). For a more in-depth theoretical analysis of DEA, the intersted
reader is referred to [23]. The idea is to maximize a weighted output while
nomalizing the input to 1. The unknown weights must be found for each
decision making unit (in our case a garage). This can be done by formulating
a LP (see also model dea11 . The model must be solve for each garage.

model Will22 "Efficiency Analysis (EA)";
set j,j1 := 1..28 "garages";
i := 1..6 "inputs";
k := 1..3 "outputs";
parameter
Input{j,i} :=
[7,8,10,12,8.5,4, 6,6,20,30,9,4.5,
2,3,40,40,2,1.5, 14,9,20,25,10,6,
10,9,10,10,11,5, 24,15,15,13,25,19,
6,7,50,40,8.5,3, 8,7.5,5,8,9,4,
5,5,10,10,5,2.5, 8,10,30,35,9.5,4.5,
7,8,7,8,3,2, 5,6.5,9,12,8,4.5,
6,7.5,10,10,7.5,4, 11,8,8,10,10,6,
4,5,10,10,7.5,3.5, 3,3.5,3,2,2,1.5,
5,5.5,8,10,7,3.5, 21,12,6,8,15,8,
6,5.5,2,2,8,5, 3,3.6,3,3,2.5,1.5,
30,29,120,80,35,20, 25,16,110,80,27,12,
19,10,90,12,25,13, 7,6,5,7,8.5,4.5,
12,8,7,10,12,7, 4,6,1,1,7.5,3.5,
2,2.5,1,1,2.5,1, 2,3.5,2,2,1.9,1.2];
Output{j,k} :=
11
http://lpl.virtual-optima.com/lpl/Solver.jsp?name=/dea
[2,0.6,1.5, 2.3,0.7,1.6,
0.8,0.25,0.5, 2.6,0.86,1.9,
2.4,1,2, 8,2.6,4.5,
2.5,0.9,1.6, 2.1,0.85,2,
2,0.65,0.9, 2.05,0.75,1.7,
1.9,0.7,0.5, 1.8,0.63,1.4,
1.5,0.45,1.45, 2.2,0.65,2.2,
1.8,0.62,1.6, 0.9,0.35,0.5,
1.2,0.45,1.3, 6,0.25,2.9,
1.5,0.55,1.55, 0.8,0.2,0.45,
7,2.5,8, 6.5,3.5,5.4,
5.5,3.1,4.5, 1.2,0.48,2,
4.5,2,2.3, 1.1,0.48,1.7,
0.4,0.1,0.55, 0.3,0.09,0.4];
J;
variable x{j}; w;
constraint
ilevel{i}: sum{j} Input[j,i]*x[j] <= Input[J,i];
olevel{k}: sum{j} Output[j,k]*x[j] >= Output[J,k]*w;
for{j1} do
J:=j1;
maximize weight: w;
Write('J=%2d 1/w = %5.3f %s \n',
J,1/w,if(w=1,'efficient',''));
end
end

2.30. Milk Collection (will23)

Problem: A small milk processing company is committed to collecting milk
from 20 farms and taking it back to the depot for processing. The company
has one tanker lorry with a capacity for carrying 80000 litres of milk. Eleven of
the farms are small and need a collection only every other day. The other nine
farms need a collection every day. Find the optimal route for the tanker lorry
on each day, bearing in mind that it has to (i) visit all the “every day” farms,
(ii) visit some of the “every other day” farms and work within its capacity.
On alternate days, it must again visit the “every day” farms and also visit the
“every other day” farms not visited on the previous day.
(Note: I have added the minimal necessary subtours that are needed to find a
correct feasible solution.)


model Will23 "Milk Collection";
set f,g,h := [1..21]; --farms including depot as farm 1
a{f} := f<=10; --1 to 10 visited every day
o{g} := g>10; --11 to 21 visited every other day
d := [1..2]; --alternating days
parameter X{f}:=[0 -3 1 4 -5 -5 -4 6 3 -1 0 6 2 -2 6 1 -3
-6 2 -6 5];
Y{f}:=[0 3 11 7 9 -2 -7 0 -6 -3 -6 4 5 8 10 8 1 5 9 -5
-4];
Cap{f}:=[. 5 4 3 6 7 3 4 6 5 4 7 3 4 5 6 8 5 7 6 6];
Dist{f,g|f<g}:=Round(Sqrt((X[f]-X[g])^2+(Y[f]-Y[g])^2)
,-1);
Tcap:=80 "Tanker capacity";
binary variable
x{f,g,d} "1 if go beween farms on a day";
y{f,d} "1 if farm visited on a day";
constraint
limit{d}: sum{o} Cap*y <= Tcap - sum{a} Cap "tanker
capacity";
dayvis{o[g]}: sum{d} y[g,d] = 1 "visit every other
day";
mat1{a,d}: sum{h|h>a} x[a,h,d]+sum{h|h<a} x[h,a,d] = 2
"matching for every day";
mat2{o,d}: sum{h|h>o} x[o,h,d]+sum{h|h<o} x[h,o,d] =
2*y[o,d] "match every 2nd day";
bound: y[11,1]=1 and y[11,2]=0;
--or: bound: y[11,1] and ~y[11,2];
va{o,h,d|f>o}: x[o,h,d] <= y[o,d] "tighten cuts";
vb{o,h,d|f<o}: x[h,o,d] <= y[o,d] "tighten cuts";

set s:=[1..5]; // 5 subtours elimination
S{s,f}:=[(1,*) 6 7 20, (2,*) 2 5 18, (3,*) 3 4 13 16
19 (4,*) 8 9 21, (5,*) 1 2 6 7 10 17 ];
constraint
Subtours{d,s}: sum{f,g|S[s,f] and S[s,g]} x[f,g,d] <=
sum{f|S}1 - 1;
minimize cost: sum{f,g,d|f<g} Dist[f,g]*x[f,g,d];
Draw.Scale(40,-40);
{f,g|x[f,g,1]} Draw.CLine(X[f],Y[f],X[g],Y[g],1,3,3);
{f,g|x[f,g,2]} Draw.CLine(X[f],Y[f],X[g],Y[g],-1,5);
{f} Draw.Circle(f&'',X,Y,.3,1,0);
Draw.Line(-6,11,-5,11,3,3); Draw.Text('first day tour
',-5,11);
Draw.Line(-6,10.5,-5,10.5,5,2); Draw.Text('second day
tour',-5,10.5);
end

2.31. Yield Management (will24)

Problem: An airline is selling tickets for flights to a particular destination.
The flight will depart in three weeks’ time. It can use up to six planes each
costing $50 000 to hire. Each plane has 37 First Class seats, 38 Business Class
seats, 47 Economy Class seats. Up to 10% of seats in any one category can
be transferred to an adjacent category. It wishes to decide a price for each of
these seats.


model Will24 "Yield Management";
set c := 1..3; --Classes: First,Business and
Economy
s,j,k := 1..3; --Scenes: Scenarios
p := 1..3; --Weeks from take-off
o := 1..3; --Price levels possible
parameter
Cap{c}:=[37,38,47]; -- Capacities per plane
Prob{s}:=[0.1,0.7,0.2];-- Probabilities of scenarios
Price{p,c,o} := -- Prices by Period, Class and option
[1200,1000, 950, 900, 800, 600, 500, 300, 200,
1400,1300, 1150, 1100, 900, 750, 700, 400, 350,
1500, 900, 850, 820, 800, 500, 480, 470, 450];
Dem{p,s,c,o}:= --Demand by period,Scenario,Class and
Option
[10 15 20 , 20 25 35 , 45 55 60
20 25 35 , 40 42 45 , 50 52 63
45 50 60 , 45 46 47 , 55 56 64
20 25 35 , 42 45 46 , 50 52 60
10 40 50 , 50 60 80 , 60 65 90
50 55 80 , 20 30 50 , 10 40 60
30 35 40 , 40 50 55 , 50 60 80
30 40 60 , 10 40 45 , 50 60 70
50 70 80 , 40 45 60 , 60 65 70];
Cost:=50000; -- Cost per plane
binary variable
pone{c,o}; -- set/or not price option period 1
ptwo{s,c,o}; -- set/or not price option period 2
pthr{s,j,c,o}; -- set/or not price option period 3
variable
sone{s,c,o}; -- Tickets sold in period 1
stwo{s,j,c,o}; -- Tickets sold in period 2
sthr{s,j,k,c,o}; -- Tickets sold in period 3
rone{s,c,o}; -- Revenue in period 1
rtwo{s,j,c,o}; -- Revenue in period 2

rthr{s,j,k,c,o}; -- Revenue in period 3
xess{s,j,k,c} [0.. 0.1*Cap]; -- Excess capa. allowed
xund{s,j,k,c} [0.. 0.1*Cap]; -- Capa. reduction
allowed
integer variable number [0..6]; -- Nr. of planes to send
constraint
-- Constraints forcing revenue equal to number sold
time price
r1a{s,c,o}: rone[s,c,o] <= Price[1,c,o]*sone[s,c,o];
r1b{s,c,o}: Price[1,c,o]*sone[s,c,o]-rone[s,c,o] <=
Price[1,c,o]*Dem[1,s,c,o]*(1-pone[c,o]);
r2a{s,j,c,o}: rtwo[s,j,c,o] <= Price[2,c,o]*stwo[s,j,c,
o];
r2b{s,j,c,o}: Price[2,c,o]*stwo[s,j,c,o]-rtwo[s,j,c,o]
<= Price[2,c,o]*Dem[2,j,c,o]*(1-ptwo[s,c,o]);
r3a{s,j,k,c,o}: rthr[s,j,k,c,o] <= Price[3,c,o]*sthr[s,
j,k,c,o];
r3b{s,j,k,c,o}: Price[3,c,o]*sthr[s,j,k,c,o]-rthr[s,j,k
,c,o]
<= Price[3,c,o]*Dem[3,k,c,o]*(1-pthr[s,j,c,o]);
-- Capacity limits in each class for all scenarios
Kap{s,j,k,c}: sum{o} sone[s,c,o] +sum{o} stwo[s,j,c,o]
+sum{o} sthr[s,j,k,c,o] <= (Cap[c]*number + xess[s,
j,k,c] - xund[s,j,k,c]);
-- (Cap[c]*number + xess[c] - xund[c]);
Adjust{s,j,k}: sum{c} xess[s,j,k,c] -sum{c} xund[s,j,k
,c] = 0;
-- sum{c} xess[c] -sum{c} xund[c] = 0;
-- Exactly one price option in each class
Lone{c}: sum{o} pone[c,o] = 1;
Ltwo{s,c}: sum{o} ptwo[s,c,o] = 1;
Lthr{s,j,c}: sum{o} pthr[s,j,c,o] = 1;
-- Numbers sold cannot exceed demand
l1s{s,c,o}: sone[s,c,o] <= Dem[1,s,c,o]*pone[c,o];
l2s{s,j,c,o}: stwo[s,j,c,o] <= Dem[2,j,c,o]*ptwo[s,c,o
];
l3s{s,j,k,c,o}: sthr[s,j,k,c,o] <= Dem[3,k,c,o]*pthr[s
,j,c,o];
maximize Eyield: sum{s,c,o} Prob[s]*rone[s,c,o]
+sum{s,j,c,o} Prob[s]*Prob[j]*rtwo[s,j,c,o]
+sum{s,j,k,c,o} Prob[s]*Prob[j]*Prob[k]*rthr[s,j,k,c,o]
-Cost*number;
Writep(Eyield);
end

2.32. Car Rental I (will25)

Problem: A small (“cut price”) car rental company, renting one type of car,
has depots in Glasgow, Manchester, Birmingham and Plymouth. There is an
estimated demand for each day of the week except Sunday when the com-
pany is closed. These estimates are given in Table 12.20. It is not necessary to
meet all demand. Cars can be rented for one, two or three days and returned
to either the depot from which rented or another depot at the start of the next
morning. How many cars should the company own and where should they
be located at the start of each day while maximizing the profit?


model Will25 "Car Rental I";
set i,j := 1..4; --depots: Glasgow,Manchester,
Birmingham,Plymouth
t := 1..6; --days: Monday,...,Saturday
k := 1..3; --hire: 1,2 or 3 day hire possible
parameter
demand{i,t} := [
100 150 135 83 120 230, 250 143 80 225 210 98,
95 195 242 111 70 124, 160 99 55 96 115 80];
--Proportion sent from each depot to other depots
perret{i,j} := [
0.6 0.2 0.1 0.1 , 0.15 0.55 0.25 0.05,
0.15 0.2 0.54 0.11, 0.08 0.12 0.27 0.53];
--Proportion cars hired for k days
hireper{k} := [0.55 0.20 0.25];
--Cost of transfer from depot to depot
costtran{i,j} := [0 20 30 50, 20 0 15 35,
30 15 0 25, 50 35 25 0];
repcap{i} := [0 12 20 0]; --Daily repair capacity
rcosta{k} := [50 70 120]; --Rental price return to
same depot
rcostb{k} := [70 100 150]; --return to another depot
rcostcomp{k} := [20 25 30]; --Marginal cost
variable
n; -- Total number of cars
nu{i,t}; -- Number of undamaged cars
nd{i,t}; -- Number of damaged cars
tr{i,t}; -- Number of cars rented out
eu{i,t}; -- Excess of undamaged cars (not rented out)
ed{i,t}; -- Excess of damaged cars
-- Not to be transferred or repaired
tu{i,j,t}; -- Number of undamaged cars
-- to be transferred to depot j
td{i,j,t}; -- Number of damaged cars
-- to be tranferred to depot j
rp{i,t}; -- Number repaired at depot i during day t
constraint
inudam{i,t|t>1}: sum{j,k|k<t} (0.9*perret[j,i]*hireper
[k]*(tr[j,t-k]))
+sum{j,k|k>=t} (0.9*perret[j,i]*hireper[k]*(tr[j,t-k
+6]))
+sum{j} (tu[j,i,t-1])+rp[i,t-1]+eu[i,t-1] = nu[i,t];
-- Net flow of undamaged cars into each depot on each
day>Monday
inudama{i}: sum{j,k}(0.9*perret[j,i]*hireper[k]*(tr[j
,7-k]))
+sum{j} (tu[j,i,6])+rp[i,6]+eu[i,6] = nu[i,1];
-- Net flow of undamaged cars into each depot on
Monday
indam{i,t|t>1}: sum{j,k|k<t} (0.1*perret[j,i]*hireper[
k]*(tr[j,t-k]))
+6]))
+sum{j} (td[j,i,t-1])+ed[i,t-1] = nd[i,t];
-- Net flow of damaged cars into each depot on each
day>Monday
indama{i}: sum{j,k}(0.1*perret[j,i]*hireper[k]*(tr[j
,7-k]))
+sum{j} (td[j,i,6]) +ed[i,6] = nd[i,1];
-- Net flow of damaged cars into each depot on Monday
outudam{i,t}: sum{j} (tu[i,j,t]) +tr[i,t]+eu[i,t]=nu[i
,t];
-- Net flow of undamaged cars out of depot i on day t
outdam{i,t|t>1}: sum{j} td[i,j,t] +rp[i,t-1]+ed[i,t] =
nd[i,t];
-- Net flow of damaged cars out of depot i on day t>
Monday
outdama{i}: sum{j} (td[i,j,1]) +rp[i,6]+ed[i,1] = nd[i
,1];
-- Net flow of damaged cars out of depot i on Monday
repc{i,t}: rp[i,t] <= repcap[i]; -- Repair capacity
demd{i,t}: tr[i,t] <= demand[i,t]; -- Renting out
within demand
tot: sum{i} (0.25*tr[i,1]+0.45*tr[i,2]+nu[i,3]+nd[i
,3]) = n;
-- Total number of cars owned equals those rented out
on Monday for 3 days
-- plus those rented out Tuesday more than 1 day
-- plus those in depots at beginning of Wednesday
maximize Profit: sum{i,t,k|t<>6}
-- Hired out,not Saturday, and returned to same
depot
(perret[i,i]*hireper[k]*(rcosta[k]-rcostcomp[k
]+10)*(tr[i,t]))
-- Hired out,not Saturday, and returned to
another depot
+sum{i,j,t,k|i<>j and t<> 6}
(perret[i,j]*hireper[k]*(rcostb[k]-rcostcomp[k
]+10)*(tr[i,t]))
+sum{i,t,k|t=6 and k=1}
-- Hired out on Saturday for 1 day with return to
same depot
(perret[i,i]*hireper[k]*(rcosta[k]-20-rcostcomp[k
]+10)*(tr[i,t]))
+sum{i,j,t,k|i<>j and t=6 and k=1}
-- Hired out on Saturday for 1 day with return to
another depot
(perret[i,j]*hireper[k]*(rcostb[k]-20-rcostcomp[k
]+10)*(tr[i,t]))
+sum{i,t,k|t=6 and k<>1} -- Hired out on
Saturday for more than 1 day with return to
same depot
]+10)*(tr[i,t]))
+sum{i,j,k,t|i<>j and t=6}
-- Hired out on Saturday for more than 1 day with
return to another depot
]+10)*(tr[i,t]))
-sum{i,j,t}
-- Cost of transfer to another depot
(costtran[i,j]*(tu[i,j,t]+td[i,j,t]))
-15*n; --Marginal cost to company of cars
owned
Writep(Profit,n);
end

2.33. Car Rental II (will26)

Problem: This model extends the model Will2512 . The company wants to
consider where it might be most worthwhile to expand repair capacity given
the cost structure of the repair centers.


model Will26 "Car Rental II";
set i,j := 1..4; --depots: Glasgow,Manchester,Birmingham
,Plymouth
t := 1..6; --days: Monday,...,Saturday
k := 1..3; --hire: 1,2 or 3 day hire possible
parameter
demand{i,t} := [
100 150 135 83 120 230, 250 143 80 225 210 98,
95 195 242 111 70 124, 160 99 55 96 115 80];
perret{i,j} := [
0.6 0.2 0.1 0.1 , 0.15 0.55 0.25 0.05,
0.15 0.2 0.54 0.11, 0.08 0.12 0.27 0.53];
hireper{k} := [0.55 0.20 0.25];
costtran{i,j} := [ --Cost of transfer
0 20 30 50, 20 0 15 35,
30 15 0 25, 50 35 25 0];
repcap{i} := [0 12 20 0]; --Daily repair capacity
rcosta{k} := [50 70 120]; --Rental price return to
same depot
rcostb{k} := [70 100 150]; --return to another depot
rcostcomp{k} := [20 25 30]; --Marginal cost
variable
n; -- Total number of cars
nu{i,t}; --Number of undamaged cars at depot i at
beginning of day t
nd{i,t}; --Number of damaged cars at depot i at
beginning of day t
tr{i,t}; --Number of cars rented out at depot i on
day t
eu{i,t}; --Excess of undamaged cars at depot i at
beginning of day t, Not rented out that day
ed{i,t}; --Excess of damaged cars at depot i at
beginning of day t, Not to be transferred or
repaired that day
tu{i,j,t}; --Number of undamaged cars at depot i at
beginning of day t to be transferred to depot j
12
http://lpl.virtual-optima.com/lpl/Solver.jsp?name=/Will25
td{i,j,t}; --Number of damaged cars at depot i at

beginning of day t to be tranferred to depot j
rp{i,t}; --Number repaired at depot i during day t
binary variable
db1; -- First Birmingham expansion
db2; -- Second Birmingham expansion
dm1; -- First Manchester expansion
dm2; -- Second Manchester expansion
dp; -- Plymouth expansion
constraint
inudam{i,t|t>1}: sum{j,k|k<t} (0.9*perret[j,i]*hireper
[k]*(tr[j,t-k]))
+6]))
+sum{j} (tu[j,i,t-1])+rp[i,t-1]+eu[i,t-1] = nu[i,t];
-- Net flow of undamaged cars into each depot on each
day>Monday
inudama{i}: sum{j,k}(0.9*perret[j,i]*hireper[k]*(tr[j
,7-k]))
+sum{j} (tu[j,i,6])+rp[i,6]+eu[i,6] = nu[i,1];
-- Net flow of undamaged cars into each depot on
Monday
indam{i,t|t>1}: sum{j,k|k<t} (0.1*perret[j,i]*hireper[
k]*(tr[j,t-k]))
+6]))
+sum{j} (td[j,i,t-1])+ed[i,t-1] = nd[i,t];
-- Net flow of damaged cars into each depot on each
day>Monday
indama{i}: sum{j,k}(0.1*perret[j,i]*hireper[k]*(tr[j
,7-k]))
+sum{j} (td[j,i,6]) +ed[i,6] = nd[i,1];
-- Net flow of damaged cars into each depot on Monday
outudam{i,t}: sum{j} (tu[i,j,t]) +tr[i,t]+eu[i,t] = nu
[i,t];
-- Net flow of undamaged cars out of depot i on day t
outdam{i,t|t>1}: sum{j} (td[i,j,t]) +rp[i,t-1]+ed[i,t]
= nd[i,t];
-- Net flow of damaged cars out of depot i on day t>
Monday
outdama{i}: sum{j} (td[i,j,1]) +rp[i,6]+ed[i,1] = nd[i
,1];
-- Net flow of damaged cars out of depot i on Monday
repc1{i,t|i=1}: rp[i,t] <= repcap[i]; --
Repair capacity at Glasgow
repc2{i,t|i=2}: rp[i,t] <= repcap[i]+5*db1+5*db2; --
Repair capacity at Birmingham
repc3{i,t|i=3}: rp[i,t] <= repcap[i]+5*dm1+5*dm2; --
Repair capacity at Manchester
repc4{i,t|i=4}: rp[i,t] <= repcap[i]+5*dp; --
Repair capacity at Plymouth

LI: db1>=db2 and dm1>=dm2 and db1+db2+dm1+dm2+dp <= 3;
--Limits on expansion
demd{i,t}: tr[i,t] <= demand[i,t]; -- Renting out
within demand
tot: sum{i} (0.25*tr[i,1]+0.45*tr[i,2]+nu[i,3]+nd[i
,3]) = n;
-- Total number of cars owned equals those rented out
on Monday for 3 days
-- plus those rented out Tuesday more than 1 day
-- plus those in depots at beginning of Wednesday
maximize Profit: sum{i,t,k|t<>6} -- Hired out,not
Saturday, and returned to same depot
]+10)*(tr[i,t]))
+sum{i,j,t,k|i<>j and t<>6} -- Hired out,not
Saturday, and returned to another depot
]+10)*(tr[i,t]))
+sum{i,t,k|t=6 and k=1} -- Hired out on Saturday
for 1 day with return to same depot
(perret[i,i]*hireper[k]*(rcosta[k]-20-rcostcomp[k
]+10)*(tr[i,t]))
+sum{i,j,t,k|i<>j and t=6 and k=1} -- Hired out
on Saturday for 1 day with return to another
depot
(perret[i,j]*hireper[k]*(rcostb[k]-20-rcostcomp[k
]+10)*(tr[i,t]))
+sum{i,t,k|t=6 and k<>1} -- Hired out on Saturday
for more than 1 day with return to same
depot
]+10)*(tr[i,t]))
+sum{i,j,k,t|i<>j and t=6} -- Hired out on
Saturday for more than 1 day with return to
another depot
]+10)*(tr[i,t]))
-sum{i,j,t} -- Cost of transfer to another depot
(costtran[i,j]*(tu[i,j,t]+td[i,j,t])) -15*n --
Marginal cost to company of cars owned
-18000*db1-8000*db2-20000*dm1-5000*dm2-19000*dp;
--Expansion costs
Writep(Profit,n);
end

2.34. Lost Baggage Distribution (will27)

Problem: A small company with six vans has a contract with a number of
airlines to pick up lost or delayed baggage, belonging to customers in the
London area, from Heathrow airport at 6 p.m. each evening. The contract
stipulates that each customer must have their baggage delivered by 8 p.m.
The company requires a model, which they can solve quickly each evening,
to advise them what is the minimum number of vans they need to use and to
which customers each van should deliver and in what order. There is no prac-
tical capacity limitation on each van. All baggage that needs to be delivered
in a two-hour period can be accommodated in a van.
Formulate optimisation models that will minimise the number of vans that
need to be used, and within this minimum, minimise the time taken for the
longest time delivery.


model Will27 "Lost Baggage Distribution";
set i,j := [1..15]; --locations
k := [1..6]; --vans
parameter
X{i}:=[0 1 2 5 4 11 8 7 9 3 2 2 4 6 8.5];
Y{i}:=[5 0 4 3 10 5 7 4 1 4 8 6 5 2 4];
time{i,j
}:=[0,20,25,35,65,90,85,80,86,25,35,20,44,35,82,
0,0,15,35,60,55,57,85,90,25,35,30,37,20,40,
0,0,0,30,50,70,55,50,65,10,25,15,24,20,90,
0,0,0,0,45,60,53,55,47,12,22,20,12,10,21,
0,0,0,0,0,46,15,45,75,25,11,19,15,25,25,
0,0,0,0,0,0,15,15,25,45,65,53,43,63,70,
0,0,0,0,0,0,0,17,25,41,25,33,27,45,30,
0,0,0,0,0,0,0,0,25,40,34,32,20,30,10,
0,0,0,0,0,0,0,0,0,65,70,72,61,45,13,
0,0,0,0,0,0,0,0,0,0,20,8,7,15,25,
0,0,0,0,0,0,0,0,0,0,0,5,12,45,65,
0,0,0,0,0,0,0,0,0,0,0,0,14,34,56,
0,0,0,0,0,0,0,0,0,0,0,0,0,30,40,
0,0,0,0,0,0,0,0,0,0,0,0,0,0,27,
0,0,0,0,0,0,0,0,0,0,0,0,0,0,0];
binary variable
x{i,j,k}; -- =1 iff van k goes i to j
y{i,k}; -- =1 iff van k goes to i
d{k}; -- =1 iff van k used
constraint
ind{i,k|i<>1}: y[i,k] <= d[k]; --Is van k used?

all{k}: y[1,k] >= d[k]; --All vans visit 1 (Heathrow)
limit{k}: sum{i,j|i<>j and j<>1} ((time[i,j]+time[j,i])
*x[i,j,k]) +sum{i,j|i<>j and j=1} (time[i,j]*x[i,j,
k]) <=120; -- Time limit on each van
num{i|i<>1}: sum{k} y[i,k] = 1; -- Exactly one van to
each location (except all used to 1)
into{i,k}: sum{j|j<>i} x[j,i,k]=y[i,k]; -- If van k
visits i then exactly 1 link in
out{i,k}: sum{j|j<>i} x[i,j,k]=y[i,k]; -- If van k
visits i then exactly 1 link out
D2{i,j,k}: x[i,j,k] + x[j,i,k] <= 1; --add all subtours
of length 2
set s,t := [1..30]; --additional subtours (not needed)
S{s,i}; //:=[(1,*) 4 14, (2,*) 5 7,
// (3,*) 6 8 9 15, (4,*) 1 10
//(5,*) 1 2 3 12 13 11 (6,*) 8 15 (7,*) 1 3 4];
constraint
Subtours{k,s}: sum{i,j|S[s,i] and S[s,j]} x[i,j,k] <=
sum{i|S}1 - 1;
minimize Number: sum{k} d[k]; -- Minimise number of vans
--constraint Num: sum{k} d[k]=2;
--minimize Time: max{k} ...
Writep(Number);
Draw.Scale(40,40);
{i,j,k|x} Draw.CLine(X[i],Y[i],X[j],Y[j],1,k+1,3);
{i} Draw.Circle(i&'',X,Y,.3,1,0);
end

2.35. Protein Folding (will28)

Problem: We take a protein as consisting of a chain of amino acids. For the
purpose of this problem, the amino acids come in two forms: hydrophilic
(waterloving) and hydrophobic (water hating). Such a chain naturally folds
so as to bring as many hydrophobic acids, as possible, close together.


model Will28 "Protein Folding";
set A:=[1..50]; -- Number of acids in Protein
H{A}:= -- Hydrophobic acids
[2,4,5,6,11,12,17,20,21,25,27,28,30,31,33,37,44,46];
parameter size:=8; -- Maximum linear dimension of protein
binary variable
m{H,H}; -- = 1 iff acids i and j are matched
f{A}; -- = 1 iff fold immediately after acid i
d{A,A}; -- = 1 iff fold j immediately after acid i
variable
l{A}; -- location (preceding acid number) of fold j
p{A}; -- position (right or left of acid 1) of fold j
s; -- leftmost fold
t; -- rightmost fold
constraint
mt{H[i],H[j], k in A| i+2<j and j<=#A and (j-i)%2=1
and i<=k and k<j and k<>(i+j-1)/2}: f[k]+m[i,j
]<=1;
-- Non contiguous matching requires only one fold
between and even number of acids between
mta{H[i],H[j]| i+2<j and j<=#A and (j-i)%2=1}: f[(i+j
-1)/2]=m[i,j]; -- Matching requires mid point fold
i
bd{H[i],H[j]| i+1=j}: m[i,j]=0; -- Contiguous matches
not counted again
bda{H[i],H[j]| (j-i)%2=0}: m[i,j]=0; -- Cannot match
acids with an odd number between
c{i in A, j in A}: sum{k in A| k<i} f[k] +(i-j)*d[i,j
]<=i;
ca{i in A, j in A}: sum{k in A| k<i} f[k] - j*d[i,j]
>=0; -- forces correct value oj d[i,j]
loc{i in A, j in A}: l[j]+(#A-i)*d[i,j]<=#A;
loca{i in A, j in A}: l[j]-i*d[i,j]>=0; -- forces
correct location of folds
pos0{j in A| j=1}: p[j]=l[j];
pos1{j in A| j>1 and j%2=1}: p[j]-p[j-1]+l[j-1]-l[j
]=-1;
pos2{j in A| j>1 and j%2=0}: p[j]-p[j-1]-l[j-1]+l[j

]=1; -- positions of folds
mn{j in A}: s<=p[j]; -- leftmost fold
mx{j in A}: t>=p[j]; -- rightmost fold
depth: sum{i in A} f[i]<=size;
breadth: t-s<=size;
exclude: m[33,44]=0; --to get the same folding as
Williams
maximize Number: sum{H[i],H[j]| i<j} m[i,j];
-- Maximise number of matched hyprophobic acids, not
counting those already matched by virtue of being
contiguous in chain
Writep(Number);
Write{H[i],H[j]|m[i,j]}('Acid %2s matches acid %2s\n', i,
j);
parameter YY; R; XX; X{A}; Y{A};
{i in A} ( X[i]:=XX, Y[i]:=YY ,
if(f,(YY:=YY+1, R:=1-R)),
if(~f[i], (if(R,(XX:=XX-1),(XX:=XX+1)))) );
Draw.Scale(50,50);
{i in A|i<#A} Draw.Line(X,Y,X[i+1],Y[i+1]);
{i in A,j in A|m[i,j]} Draw.Line(X[i],Y[i],X[j],Y[j],3,3)
;
{i in A} Draw.Circle(i&'',X,Y,.25,if(H,2,1),0);
end

Solution:
1 2 3
8 7 6 5 4
9 10 11 12 13 14 15 16 17 18 19 20 21 22
35 34 33 32 31 30 29 28 27 26 25 24 23
36 37 38 39 40 41 42 43 44 45 46 47 48 49 50
Figure 2.5: The Folding Structure

2.36. Protein Comparison (will29)

Problem:


model Will29 "Protein Comparison";
set a,i := [1..9];
b,j := [1..11];
parameter
A{i,a} := //edges of the first protein
[0,1,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,1,
0,0,0,1,1,0,0,0,0, 0,0,0,0,0,0,0,0,0,
0,0,0,0,0,1,0,0,0, 0,0,0,0,0,0,1,0,0,
0,0,0,0,0,0,0,0,1, 0,0,0,0,0,0,0,0,1,
0,0,0,0,0,0,0,0,0];
B{j,b} := //edges of the second protein
[0,0,0,1,0,0,0,0,0,0,0, 0,0,1,0,0,0,0,0,0,0,0,
0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,1,1,0,0,0,0,
0,0,0,0,0,1,0,0,0,0,0, 0,0,0,0,0,0,0,1,0,0,0,
0,0,0,0,0,0,0,1,0,1,0, 0,0,0,0,0,0,0,0,0,0,0,
0,0,0,0,0,0,0,0,0,1,0, 0,0,0,0,0,0,0,0,0,0,1,
0,0,0,0,0,0,0,0,0,0,0];
binary variable x{a,b}; -- =1 iff a corresponds to b
w{i,j,a,b}; /*FREE;*/ -- =1 iff x[i,k]=x[j,l]=1
constraint
Assa{a}: sum{b} x[a,b] <=1; -- At most 1 node from G2
assigned to each in G1
Assb{b}: sum{a} x[a,b] <=1; -- At most 1 node from G1
assigned to each in G2
Linka{i,j,a,b|A[i,a]=1 and B[j,b]=1 and i<a and j<b}:
w[i,j,a,b]<=x[i,j];
Linkb{i,j,a,b|A[i,a]=1 and B[j,b]=1 and i<a and j<b}:
w[i,j,a,b]<=x[a,b];
--Linkc{i,j,a,b|A[i,a]=1 and B[j,b]=1 and i<a and j<b}:
w[i,j,a,b]>=x[i,j]+x[a,b]-1;
Ncrs{i,j,a,b|i<a and b<j}: x[i,j]+x[a,b]<=1; -- No
cross overs
maximize size: sum{i,j,a,b|A[i,a]=1 and B[j,b]=1 and i<a
and j<b} w[i,j,a,b];
-- Maximise number of corresponding edges
Writep(size);
end

2.37. Product Mix I (willi005)

Problem: A factory produces five type of products by using two processes
and manpower. How much should be produced, when the capacities of the
processes and the manpower is limited and the profit is to be maximized?
Model: The Model is as follows:The problem is from [17], page 6 and page
107.

model Willi005 "Product Mix I";
variable x1; x2; x3; x4; x5;
constraint
Grinding: 12*x1+20*x2 +25*x4+15*x5 <= 288;
Drilling: 10*x1+8*x2+16*x3 <= 192;
Manpower: 20*x1+20*x2+20*x3+20*x4+20*x5 <= 384;
maximize profit: 550*x1+600*x2+350*x3+400*x4+200*x5;
Writep(profit,x1,x2,x3,x4,x5);
end

2.38. Product Mix II (willi005a)

Problem: A factory produces five type of products by using two processes
and manpower. How much should be produced, when the capacities of the
processes and the manpower is limited and the profit is to be maximized?
Model: The Model is as follows:The problem is from [17], page 6 and page
107.

model Willi005a "Product Mix II";
set product,p := [1..5];
process,r := [Grinding Drilling];
parameter
Profit{p} := [550 600 350 400 200];
Hours{r,p} := [12 20 0 25 15, 10 8 16 0 0];
AssemblyTime := 20;
DaysPerWeek := 6;
HoursPerShift := 8;
HoursPerDay := 2*HoursPerShift;
Workers := 8;
Machines{r} := [3 2];
WorkHoursAvail:= Workers*DaysPerWeek*HoursPerShift;
MachHours{r} := Machines*DaysPerWeek*HoursPerDay;
variable
Produce,P{p};
constraint
ProcessLimit{r}: sum{p} Hours*P <= MachHours;
WorkLimit: sum{p} AssemblyTime*P <= WorkHoursAvail;
maximize TotalProfit: sum{p} Profit*P;
Writep(TotalProfit,P);
end

2.39. Vegetable Oil Blending (willi008)

Problem: A food is manufactured by refining raw oils and blending them
together. How much should be taken from each raw oil in order to maximize
profit?
Model: The Model is as follows:The problem is from [17], page 8.

model Willi008 "Vegetable Oil Blending";
set oils,o := [Veg1, Veg2, Oil1, Oil2, Oil3];
VegOils,v{oils} := [Veg1, Veg2];
NonVegOils,n{o} := [Oil1, Oil2, Oil3];
parameter
MaxVegRefine := 200;
MaxNonVegRefine := 250;
Price := 150;
Cost{o} := [110, 120, 130, 110, 115];
Hardness{o} := [8.8, 6.1, 2.0, 4.2, 5.0];
MinHardness := 3;
MaxHardness := 6;
variable
RawOils{o};
Produce;
constraint
ProductWeight: Produce = sum{o} RawOils;
MaxVegRefining: sum{v} RawOils <= MaxVegRefine;
MaxNonVegRefining: sum{n} RawOils <= MaxNonVegRefine;
HardnessC: MinHardness*Produce <=
sum{o} Hardness*RawOils <= MaxHardness*Produce;
maximize Profit: Price*Produce - sum{o} Cost*RawOils;
Writep(Profit,Produce,RawOils);
end

2.40. Multi-Plant Planning (willi055)

Problem: A company has two factories and manufactures two products. Two
processes are used to make them with a limiting capacity. Calculate the max-
imal profit.

model Willi055 "Multi-Plant Planning";
set factory,f := [A, B];
product,p := [Standard, Deluxe];
process,pr := [Grinding, Polishing];
parameter
Profit{p} := [10, 15];
Capacity{f,pr} := [80, 60, 60, 75];
Time{pr,f,p} := [4, 2, 5, 3, 2, 5, 5, 6];
RawUse := 4;
RawAvail := 120;
variable Produce,P{factory,product};
constraint
RawLimit: sum{f, p} RawUse*P <= RawAvail;
ProcessLimit{f,pr}: sum{p} Time*P <= Capacity;
maximize TotalProfit: sum{f,p} Profit*P;
end

2.41. Transportation (willi082)

Problem: A number of supplier and costumer are given. The transportation
problem is to meet each costomer’s requirement, while not exceeding the ca-
pacity of any supplier and at minimum cost.

model Willi082 "Transportation";
set supplier,s := [S1 S2 S3];
customer,c := [T1 T2 T3 T4];
parameter
Capacity{s} := [135 56 93];
Requirement{c} := [62 83 39 91];
ShippingCost{s,c} :=
[132 . 97 103, 85 91 . ., 106 89 100 98];
variable Ship{s,c|ShippingCost};
constraint
MaxCapacity{s}: sum{c} Ship <= Capacity;
MinRequirement{c}: sum{s} Ship >= Requirement;
minimize TotalCost: sum{s,c} ShippingCost*Ship;
Writep(TotalCost,Ship);
end

2.42. Production Planning (willi085)

Problem: A company produces a commodity in two shifts to meet known
demands. The problem is to satisfy the demand at minimal cost.

model Willi085 "Production Planning";
set s := [Regular, Overtime];
m,m1 := [January, February, March, April];
parameter
Capacity{s,m}:=[100 150 140 160, 50 75 70 80];
ProdCost{s} := [1.00, 1.50];
Demand{m} := [80, 200, 300, 200];
StorCost := 0.30;
CombinedCost{m,s,m1} :=
if(m1>=m,ProdCost) + if(m1>m,m1-m)*StorCost;
variable
Produce,P{m,s} [0..Capacity];
Inventory,I{m};
constraint Balance{m}: I = I[m-1] + sum{s} P - Demand;
minimize Cost: sum{s,m} ProdCost*P + sum{m} StorCost*I;
Writep(Cost,P,I);
end

2.43. Minimum Cost Flow (willi090)

Problem: Find the minimum flow through a network

model Willi090 "Minimum Cost Flow";
set node,s,d := [n0 n1 n2 n3 n4 n5 n6 n7];
parameter
FlowCost{s,d}:=/n0 n2 5, n1 n3 4, n2 n3 2, n2 n4 6,
n2 n5 5, n3 n4 1, n3 n7 2, n4 n2 4, n4 n5 6,
n4 n6 3, n7 n6 4/;
Availability{node} := /n0 10, n1 15/;
Requirement{node} := /n5 9, n6 10, n7 6/;
variable Flow,F{s,d|FlowCost};
constraint FlowBalance{node}:
sum{s} F[s,node] + Availability
= sum{d} F[node,d] + Requirement;
minimize TotalCost: sum{s,d} FlowCost*F;
Writep(TotalCost,F);
end

2.44. Shortest Path Problem (willi093)

Problem: Finding the shortest path through a network

model Willi093 "Shortest Path Problem";
set node,s,d := [n0 n1 n2 n3 n4 n5 n6 n7 n8];
parameter
PathDist{s,d} :=
/ n0 n1 1, n0 n3 1, n1 n2 1, n1 n3 2, n2 n3 1,
n2 n4 1, n2 n5 3, n3 n4 2, n3 n6 4, n4 n5 3,
n4 n6 2, n4 n7 4, n5 n7 1, n6 n7 2, n6 n8 1,
n7 n8 1/;
StartPath{node} := /n0 1/;
EndPath{node} := /n8 1/;
variable Path{s,d|PathDist};
constraint
FlowBalance{node}: sum{s} Path[s,node] + StartPath
= sum{d} Path[node,d] + EndPath;
minimize TotalDistance: sum{s,d} PathDist*Path;
Writep(TotalDistance,Path);
end

2.45. Maximum Flow (willi094)

Problem: Maximizing the flow through a network

model Willi094 "Maximum Flow";
set
node ,s,d := [n0 n1 n2 n3 n4 n5 n6 n7];
SourceNodes{node} := [n0, n1];
SinkNodes{node} := [n5, n6, n7];
parameter
FlowCapacity{s,d} :=
/ n0 n2 12, n1 n3 20, n2 n3 6, n2 n4 3,
n2 n5 6, n3 n4 7, n3 n7 9, n4 n2 2,
n4 n5 5, n4 n6 8, n7 n6 4/;
variable
Flow{s,d|FlowCapacity} [0..FlowCapacity];
Sources{SourceNodes};
Sinks{SinkNodes};
constraint
FlowBalance{node}: sum{s} Flow[s,node] + Sources
= sum{d} Flow[node,d] + Sinks;
maximize TotalFlow: sum{SourceNodes} Sources;
Writep(TotalFlow,Flow);
end

2.46. Critical Path (willi095)

Problem: Finding the critical path in a network

model Willi095 "Critical Path";
set t,s := [n0 n1 n2 n3 n4 n5 n6] "Tasks";
E{t,s}:= [n0 n1, n0 n2, n0 n3, n1 n3, n2 n5, n3 n4,
n4 n2, n4 n5, n5 n6] "A precedence list";
parameter
Duration{E} := /n0 n1 4, n0 n2 12, n0 n3 7, n1 n3 2,
n2 n5 5, n3 n4 10, n4 n2 0, n4 n5 3, n5 n6 4/;
string activities{E}:=['DigFoundations', 'ObtainTiles',
'ObtainBricks', 'LayFoundations', 'Walls', 'Dummy24',
'Roofing', 'Wiring', 'Painting'];
variable S{t} "Starting time";
constraint SeqRelation{E[t,s]}: S[s]-S[t] >= Duration;
minimize FinishTime: S[#t];
Writep(FinishTime,S);
end

2.47. A Small QP (willi142)

Problem: A small QP model from [17] p.142. Geometrically, the maodel is
shown in Figure 2.6.

model Willi142 "A Small QP";
variable x; y;
constraint
A: x + y <= 4;
B: 2*x + y <= 5;
C: -x + 4*y >= 2;
minimize obj: x^2 - 4*x - 2*y;
Writep(obj,x,y);
end

Figure 2.6: The Feasible Space and the Objective

2.48. LP-relaxtion versus IP-solution (willi155)

Problem: Solve the following problem once with real variables and once with
integer variables. What is your observation? (The model is from [16], p.155.)
max x+y
subject to −2x + 2y ≥ 1
−8x + 10y ≤ 13
x, y ≤ 0
Modeling Steps
The observation is that the integer solution is not just a rounding of the real
solution. On the contrary, the two solutions are far away from each other. It is
difficult to see, how we could derive the integer solution from the lp solution.
Figure 2.7: The IP- and the LP-solution
If the quantities x and y are “the number of aeroplanes” in a company then

we need to model this by integer values, because it will be difficult to interpret
“4.5 aeroplanes”, is it 4 or 5, or – as in our example – a completely different
number?
However, if we need to model “the number of hens” in a country, then we
almost certainly can do this by real values: 3’345’678.32 hens then certainly
can be expressed by rounding this number.

model Willi155 "LP-relaxtion versus IP-solution";
integer variable x [0..100]; y [0..100];
constraint
R1: -2*x+2*y >= 1;
R2: -8*x+10*y <= 13;
maximize obj1 'relax': x+y;
Write('LP sol: (obj, x,y) = (%4.2f, %4.2f,%4.2f)\n',
obj1,x,y);
maximize obj2: x+y;
Write('IP sol: (obj, x,y) = (%-4d, %-4d,%-4d)\n', obj1,
x,y);
/* set i,j:=1..5;
Draw.Scale(70,70);
Draw.Rect(0,0,6.5,6.5,1);
Draw.Line(0,6,6,6); Draw.Line(0,6,0,0);
Draw.Line(0,5.5,5.5,0);
Draw.Line(0,4.7,5.875,0);
for{i,j} do Draw.Circle(i,j,.05); end
Draw.Circle(1,4,.08,14);
Draw.Text('(1,2) IP Optimum',1,4.25);
Draw.Circle(4,1.5,.08,14);
Draw.Text('(4,4.5) LP Optimum',4,1.75);
*/
set i,j:=[1..5];
Draw.Scale(70,70);
Draw.Rect(0,0,6.5,6.5,1);
Draw.Line(0,6,6,6); Draw.Line(0,6,0,0);
Draw.Line(0,5,5,0);
Draw.Line(0,4.8,6,0);
for{i,j} do Draw.Circle(i,j,.05); end
Draw.Circle(1,4,.08,14);
Draw.Text('(1,2) IP=LP Optimum',1,4.25);
end

Solution: The integer solution for this small model is: (x, y) = (1, 2) and the
optimal value is 3, whereas the relaxed real solution is (x, y) = (4, 4.5) with
an optimal value of 8.5.
1. What happens if the constraint R2 becomes:
R2: -9*x + 10*y <= 9;
2. Change the right hand side of constraint R1 from 1 to 2, and the right
hand side from 13 to 12. Solve the LP problem and the ILP problem
(once with real variables, once with integer variables. What do you
derive from the modified model?
1. The integer problem becomes infeasible. There is no integer point within

the feasible space. However the LP has still the same solution at (x, y) =
(4, 4.5).
2. Both models the LP and the ILP have the same optimal solution. This
come from the fact that the optimal corner point of the LP model is the
integer solution (1, 2). All we needed to to was a parallel translation of
the two lines. See Figure 2.8.
Figure 2.8: The IP- and the LP-solution are equal

2.49. LP-relaxtion versus 0-1-IP-solution (willi155z)

Problem: Solve the following problem once with real variables bounded by
[0 . . . 1] and once with 0-1 integer variables. What is your observation?
max x+y
subject to 5x + 10y ≤ 9
−5x + 10y ≥ 1
0 ≤ x, y ≤ 1 (or x, y ∈ [0, 1])
Modeling Steps
The observation is that there does not exists an integer solution (the prob-
lem is infeasible). However, there exists a solution with continuous variables
[(x, y) = (0.8, 0.5)] bounded by the interval [0 . . . 1]. In particular, one should
observe that the integer solution is not just a rounding of the real (continuous)
solution. It is difficult to see, how one could derive the fact that the integer
solution is infeasible from the continuous LP solution alone.
Figure 2.9: The IP- and the LP-solution

model Willi155 "LP-relaxtion versus 0-1-IP-solution";
binary variable x; y;
constraint
R1: 5*x + 10*y <= 9;
R2: -5*x + 10*y >= 1;
R3: x<=1 and y<=1;

maximize obj1 'relax': x+y;
Write('LP sol: (obj, x,y) = (%4.2f, %4.2f,%4.2f)\n',
obj1,x,y);
maximize obj2: x+y;
Write('IP sol: (obj, x,y) = INFEASIBLE');
set i,j:=[1..2];
Draw.Scale(150,150);
Draw.Line(0,2,3,2);
Draw.Text('5x+10y=9',2.4,1.65,0,16,-27);
Draw.Line(1,0.5,1,2.5);
Draw.Text('-5x+10y=1',1,1.75,0,16,27);
Draw.Line(0.8,1,2.8,2);
Draw.Line(0.8,2,2.8,1);
for{i,j} do
Draw.Circle(i,j,0.03);
Draw.Text('('&(i-1)&','&(j-1)&')',i+.05,3-j);
end
Draw.Circle(1.8,1.5,0.03,3);
Draw.Text('(0.8,0.5) [LP Opt.]',1.9,1.45);
end

The only way to solve the 0-1 integer problem is to “enumerate” all possible
solutions somehow. From the first constraint 5x+10y ≤ 9, one can derive that
y never can be 1. From the second constraint −5x + 10y ≥ 1, one can derive
that y never can be 0. Now we have a contradiction: y cannot be 1 neither can
it be 0. Hence, one derives that there is no integer solution for this problem.
C HAPTER 3
P ROBLEMS FROM THE

G UERET B OOK
This chapter shows an LPL implementation of all models from the book [5],
respectively [28]
107
CHAPTER 3. PROBLEMS FROM THE GUERET BOOK 108
3.1. Animal food production (guer04-2)

Problem: The problem is from [5] chap 4.2. English translation in [28] chap
6.2.

model guer04_2 "Animal food production";
set
n := [1..2] "Foods";
m := [1..3] "Raw material";
o := [1..4] "Operations";
b := [1..3] "Nutrients";
parameter
BN{b} := [9.5 2 6] "Required contents";
P{m,b}:= [13.6 7.1 7 , 4.1 2.4 3.7 , 5 0.3 25]
"Nut. contents in percent";
QMax{m} := [11900 23500 750] "Availability";
Pr{m} := [0.8 1 0.75] "Price";
C{o} := [1.5 0.3 2.5 1] "Operation cost";
D{n} := [9000 12000] "Demand";
variable
Q{n} "Quantity produced";
X{n,m} "Quantity raw material";
constraint
PRO{n}: sum{m} X = Q;
ATL{n,b|b<#b}: sum{m} P*X >= BN*Q;
MIN{n}: sum{m} P[m,#b]*X <= BN[#b]*Q;
MAX{m}: sum{n} X<=QMax;
DEM{n}: Q >= D;
minimize cost: sum{n,m} Pr*X
+ sum{n,m|m<#m} C[1]*X + sum{n,m} C[2]*X
+ sum{m} C[3]*X[1,m] + sum{m} C[4]*X[2,m];
Writep(cost,Q,X);
end

109 CHAPTER 3. PROBLEMS FROM THE GUERET BOOK
3.2. Production of alloys (guer04-3)

6.1.

model guer04_3 "Production of alloys";
set
m := [1..7] "Raw material";
e := [1..3] "Chemical elements";
parameter
C{m} := [1.2, 1.5, 0.9, 1.3, 1.45, 1.2, 1] "Cost of m";
Qmax{m} := [40 30 60 50 20 30 25] "Availability";
P{m,e} "Percent content" :=
[2.5 0 1.3 , 3 0 0.8 , 0 0.3 0 , 0 90 0 ,
0 96 4 , 0 0.4 1.2 , 0 0.6 0];
PMin{e} := [2, 0.4, 1.2] "Min. content in product";
PMax{e} := [3, 0.6, 1.65] "Max. content in product";
D := 5000 "Demand";
variable
X{m} [0..100*Qmax] "Quantity in end product";
Q "Quantity produced";
constraint
QteFabriquee: sum{m} X = Q;
Demande: Q >= D;
Pourcent{e}: PMin*Q <= sum{m} P*X <= PMax*Q;
minimize Couts: sum{m} C*X;
Writep(Couts,X,Q);
end

3.3. Refinery (guer04-4)

6.3.

model guer04_4 "Refinery";
parameter
B1:=250000 "Crude 1 in stock";
B2:=500000 "Crude 2 in stock";
variable
-- end products
B "Butane";
C "Petrol";
G "Diesel oil";
F "Heating oil";
-- Intermediary products
BC; EC; DC; GG; DG; GF; DF;
constraint
MaxButane : B+BC <= 0.03*B1+0.05*B2;
MaxEssence : EC <= 0.15*B1+0.20*B2;
MaxGazole : GG+GF <= 0.40*B1+0.35*B2;
MaxDistill : DC+DG+DF <= 0.15*B1+0.10*B2;
CompoCarb : C = EC+BC+DC;
CompoGaz : G = GG+DG;
CompoFuel : F = GF+DF;
Octane : 100*EC+120*BC+ 74*DC >= 94*C;
Vapeur : 2.6*EC+ 60*BC+4.1*DC <= 12.7*C;
Volatilite : 3*EC+105*BC+ 12*DC >= 17*C;
Soufre : 1.2*DG+0.3*GG <= 0.5*G;
Reformage : EC <= 30000;
Desulfur : GG+GF <= 40000;
Craquage : DC+DG+DF <= 50000;
-- Demand
DemButane : B >= 20000;
DemCarb : C >= 40000;
DemGaz : G >= 35000;
demFuel : F >= 50000;
minimize Cost: 250*EC+ 450*(GG+GF)+ 350*(DC+DG+DF);
Writep(Cost);
end

3.4. Cane Sugar Production (guer04-5)

6.4.

model guer04_5 "Cane Sugar Production";
parameter
N := 3 "Numberr of production lines";
DUR := 2 "Duration";
set
w := [1..11] "Lots (waggons)";
c := 1..Ceil(#w/N) "Nombres of slots";
parameter
Perte{w} := [43,26,37,28,13,54,62,49,19,28,30] "Loss";
Vie{w} := [8,8,2,8,4,8,8,8,6,8,8] "Life span";
binary variable X{w,c} "Assign";
constraint
Affect{w}: sum{c} X = 1;
LotsParSlot{c}: sum{w} X <= N;
Limite{w}: sum{c} c*X <= Vie/DUR;
minimize TotalLoss: sum{w,c} c*DUR*Perte*X;
Writep(TotalLoss,X);
end

1. What is the total loss if the life span of each wagon is infinite?
2. What is the loss on each wagon (give the expression in LPL).
3. Suppose that it is possible to accelerate the traitment. How long should
the duration be in order to push the overall loss under 1000 kg? How
does the overall loss decrease with the acceleration of the traitment?
4. How does the number of lignes influence the total loss?
5. Create and run a problem with 100 wagons.
3.5. Opencast Mining (guer04-6)

6.5.

model guer04_6 "Opencast Mining";
set
i,j := [1..18] "Number of blocks";
A{i,j} "Block j above i" :=
[(15,9)(15,10)(15,11)(16,10)(16,11)(16,12)(17,11)
(17,12)(17,13)(18,12)(18,13)(18,14)(9,1)(9,2)(9,3)
(10,2)(10,3)(10,4)(11,3)(11,4)(11,5)(12,4)(12,5)
(12,6)(13,5)(13,6)(13,7)(14,6)(14,7)(14,8)];
parameter
C{i} "Extraction cost" :=
[100, 100, 100, 100, 100, 100, 100, 100,1000,
200, 200, 200, 200,1000,1000,1000, 300,1000];
V{i} "Value of bloc" :=
[200, 0, 0, 0, 0, 0, 300, 0, 0,
500, 0, 200, 0, 0, 0, 0,1000,1200];
binary variable X{i} "Extract?";
constraint Degage{A[i,j]}: X[i] -> X[j];
maximize Profit: sum{i} (V-C)*X;
Draw.Scale(50,50);
Draw.DefFont('Verdana',10);
{i|i<=8} Draw.Rect(i&'',i,0,1,1,if(X,3,1),0);
{i|i>8 and i<15} Draw.Rect(i&'',i-7,1,1,1,if(X,3,1),0);
{i|i>=15} Draw.Rect(i&'',i-12,2,1,1,if(X,3,1),0);
Draw.Text('Extract:'&sum{i}X&' blocks',0,2.3);
Draw.Text('Total Profit:'&Profit,0,2.6);
end

3.6. Production of Electricity (guer04-7)

6.6.

model guer04_7 "Production of Electricity";
set
p := [1..5] "Time periods";
t := [1..3] "Generator types";
parameter
Larg{p} := [6 3 6 3 6] "Period length";
Dema{p} := [15000,30000,25000,40000,27000] "Demand";
PMin{t} := [ 850, 1250, 1500] "Minimal power";
PMax{t} := [2000, 1750, 4000] "Maximal power";
CDem{t} := [2000, 1000, 500] "Starting cost";
CMin{t} := [1000, 2600, 3000] "Cost at min power";
CSup{t} := [ 2, 1.3, 3] "Additional cost above min";
NDis{t} := [ 12, 10, 5] "Number of generators";
variable
NDem{t,p} "Starting generators";
PSup{t,p} "Supp. power";
integer NFon{t,p} [0..NDis] "Generator working";
constraint
R{p,t} : NDem >= NFon - if(p=1,NFon[t,#p],NFon[t,p-1]);
Lim{p,t}: PSup <= (PMax-PMin)*NFon;
SatisDeman{p}: sum{t} PMin*NFon + sum{t} PSup >= Dema;
Reserve{p}: sum{t} PMax*NFon >= 1.15*Dema;
minimize
Cost:sum{p,t}(CDem*NDem+Larg*CMin*NFon+Larg*CSup*PSup);
Writep(Cost);
end

3.7. Stadium Construction (guer05-2)

Problem: A municipality wants to construct a small stadium. A construction
company is awarded the contract an wants to complete the construction in
the shortest time possible. All tasks are listed and their duration (in weeks) is
estimated and the the order of precedence between them is established, that
is, certain tasks must be completetd before others can begin. All these data
are collected in a table (see Table 3.1.
What is the earliest possible date of completing the construction?
The model is from [5] chapter 5.2. and [28] chapter 7.1.
Task Description Duration Predecessors max reduct. add. cost

1 Installing the construction site 2 none 0
2 Terracing 16 1 3 30
3 Constructing the foundations 9 2 1 26
4 Access roads and other networks 8 2 2 12
5 Erecting the basement 10 3 2 17
6 Main floor 6 4,5 1 15
7 Dividing up the changing rooms 2 4 1 8
8 Electrifying the terraces 2 6 0
9 Constructing the roof 9 4,6 2 42
10 Lighting of the stadium 5 4 1 21
11 Installing the terraces 3 6 1 18
12 Sealing the roof 2 9 0
13 Finishing the changing rooms 1 7 0
14 Constructing the ticket office 7 2 2 22
15 Secondary access roads 4 4,14 2 12
16 Means of signalling 3 8,11,14 1 6
17 Lawn and sport accessories 9 12 3 16
18 Handing over the building 1 17 0
Table 3.1: Data for the Stadium Construction
Modeling Steps
We introduce a set for the tasks i, j ∈ I. All we need is the duration time of a
task pi and the precedence list of task tuples Preci,j . The tuple entry Preci,j
is true if and only if task i must be completed before task j. The variables are
the (earliest) starting time ti of a task i. The objective is to minimize the latest
starting time.
The model is as follows:
min max (ti + pi )

i∈I
subject to tj − ti ≥ pi , ∀(i, j) ∈ Prec
ti ≥ 0, ∀ i ∈ I
Listing 3.7: The Main Model in LPL [29]

model guer05_2 "Stadium Construction";
set i,j "tasks";
Prec{i,j} "precedences";
parameter p{i} "duration";
r{i} "max. reduction";
c{i} "add. cost per reducted week";
T "starting time of last task";
variable t{i} "start time of task i";
constraint Precedence{Prec[i,j]}: t[j] - t[i] >= p[i];
minimize lasttaskBegin: max{i} (t+p);
T:=lasttaskBegin;
Writep(T,t);
Draw.Scale(14,20);
for{i} do
Draw.Rect(t,i,p,1,i%12+2); Draw.Text((t+p)&'',t+p,i
+.5,12);
Draw.Text(i&'',t+0.1,i+.6,12);
Draw.Line(t,i+1,t,#i+2); Draw.Text(t&'',t+.1,#i
+1.8,12);
end
Draw.Rect(0,#i+2,t[#i],.1,0);
--CostModel;
EarliestLatest;
end

Further Comments: A comment on the Read statement: The first read staemetn
reads the data from an Excel sheet (the statement is outcommented). The sec-
ond Read reads the same data from the same format in a tab separated text
file. Note how similar these statements are. In the text-form we use '%1'
to indicate that the first line must be skipped, and '\t\n' indicates that the
only token separators are tabs and line feeds.
Listing 3.8: The Data Model

model data;
parameter M; string N{i}; PR{i};
--Read{i}('guer05-2.xls,[Sheet1$A2:F19]',i,N,p,PR,r,c
);
Read{i}('guer05-2.txt,;1,\t\n',i,N,p,PR,r,c);
{i,j} (M:=Strfloat(PR[i],',',j), if(M,Prec[M,i]:=1));
end
1. What is the earliest, what is the latest time that each task can begin?
2. Draw a Gantt diagram to visualiser the solution.
3. Suppose that the duration of each task could be reduced by one unit.
What is now the solution?
4. Suppose that the task number 4 can only begin at the 32 week. What
are the implication of the overall plan?
5. Suppose the municipality would like to terminate the construction ear-
lier than projected by the construction company and is ready to pay a
bonus of 30 unit of money for each reduced week. The company has
additional costs for each task ci per reduced week (last column in Table
3.1) and each task can only be reduced a maximal number of weeks,
given by ri (second to the last column in Table 3.1. How would the
construction company proceed?
6. Create and run a project with 100 tasks.
1. Replace the objective by the following code:

minimize obj: max{i} (t+p) + 0.00001*(sum{i} t);
parameter Earliest{i}:=t;
minimize obj1: max{i} (t+p) - 0.00001*(sum{i} t);
parameter Lastest{i}:=t;
The parameter Earliest gives the eariest possible starting time, and
Lastest gives the latest possible time. The Submodel EarliestLatest
shows the results.
2. See answer 1
3. Repace pi by pi − 1 the run again. The completion time is 55 weeks.
4. Add the constraint constraint C4: t[4]>=32; and run the model
again.
5. Certainly, the company would try to maximize the additional bonuses.
The additional requirements are formulated in the model CostModel.
3.8. Flow-shop Scheduling (guer05-3)

Problem: The model is from [5], Chap 5.3. and [28] chapter 7.2.

model guer05_3 "Flow-shop Scheduling";
set m,m1 := [1..3] "Machines";
j,k,k1 := [1..6] "Job j / Rank k,k1";
parameter
p{m,j}:=[3 6 3 5 5 7, 5 4 2 4 4 5, 5 2 4 6 3 6] "
processing time";
variable
E{m,k|k<#k and m>1} "Idle time on machine m between
job of rang k and k+1";
A{m,k|m<#m and k>1} "Waiting time between machine m
and m+1 of job in rang k";
binary X{j,k} "1 if job j is in rank k";
constraint
C{k}: sum{j} X = 1 "Exactly un job in each rank";
B{j}: sum{k} X = 1 "Exactly un rank for each job";
D{m,k|m<#m and k<#k}:
E[m,k] + sum{j} p[m,j]*X[j,k+1] + A[m,k+1] =
A[m,k] + sum{j} p[m+1,j]*X[j,k] + E[m+1,k]
"Relations between end of job at rang k on mach m
and begin of the following job on mach m+1";
minimize W: sum{m,j|m<#m} p*X[j,1]+sum{k|k<#k} E[#m,k]
"Total empty time on the last machine";
integer parameter
x{j}:=argmax{k|X} k;
T{m,k} := sum{m1,j|m1<m} p[m1,j]*X[j,1] +
sum{k1,j|k1<k} p[m,j]*X[j,k1] + sum{k1|k1<k} E[m,k1]
"begin on mach m of job with rank k";
t{m,j}:=T[m,x];
Write('Total wating time: %4d\n', W);
Write('Job permutation : %3d\n', {j}x);
---draw the solution
Draw.Scale(20,25);
for{m,j} do Draw.Rect(j&'',t,m,p,1,j+3); end;
set i:=[1..40];
for{i} do Draw.Line(i,#m+1,i,#m+1.3); end
for{i} do Draw.Text(i&'',i,#m+1.5,12); end
end

1. Visualize the solution with a Gantt diagram.

2. How could we force to process job 1 just before job 3 ? Add the condi-
tion for that. What is the total duration now?
3. Suppose that we must execute the job 6 as the first job. How could this
condition be added to the model? What is the consequence on the total
processing time?
4. Create and run a problem with 50 jobs.
1.
3.9. Job Shop: non-generic version (guer05-4a)

7.3.

model guer05_4a "Job Shop: non-generic version";
set
n := [1..3] "Jobs";
m := [1..3] "Machines";
parameter
P{m,n}:=[45 20 0 0 10 17 10 34 28] "Process time";
M := 1000 "Big-M";
variable
T{m,n} "Begin of task";
TEnd "Make span";
binary Y{1..5} "Disjuctions";
constraint
FinJ1 : TEnd >= T[3,1]+P[3,1];
FinJ2 : TEnd >= T[3,2]+P[3,2];
FinJ3 : TEnd >= T[3,3]+P[3,3];
-- Conjonctions : task of the same job on the machines
J1_M3M1 : T[1,1]+P[1,1] <= T[3,1];
J2_M1M3 : T[2,2]+P[2,2] <= T[1,2];
J2_M3M2 : T[1,2]+P[1,2] <= T[3,2];
J3_M2M3 : T[2,3]+P[2,3] <= T[3,3];
-- Disjonctions : precedence on the same machine
M1_J1J2_1 : T[1,1]+P[1,1] <= T[1,2]+M-M*Y[1];
M1_J1J2_2 : T[1,2]+P[1,2] <= T[1,1]+M*Y[1];
M2_J2J3_1 : T[2,2]+P[2,2] <= T[2,3]+M-M*Y[2];
M2_J2J3_2 : T[2,3]+P[2,3] <= T[2,2]+M*Y[2];
M3_J1J2_1 : T[3,1]+P[3,1] <= T[3,2]+M-M*Y[3];
M3_J1J2_2 : T[3,2]+P[3,2] <= T[3,1]+M*Y[3];
M3_J1J3_1 : T[3,1]+P[3,1] <= T[3,3]+M-M*Y[4];
M3_J1J3_2 : T[3,3]+P[3,3] <= T[3,1]+M*Y[4];
M3_J2J3_1 : T[3,2]+P[3,2] <= T[3,3]+M-M*Y[5];
M3_J2J3_2 : T[3,3]+P[3,3] <= T[3,2 ]+M*Y[5];
minimize FinOrdo: TEnd;
Writep(TEnd,T,Y);
end

3.10. Job Shop: generic version (guer05-4b)

7.3.

model guer05_4b "Job Shop: generic version";
set
m,n := [1..9] "Tasks";
U{m,n} := [1 7, 5 2, 2 8, 6 9] "m before n";
D{m,n} := [1 2, 5 6, 7 8, 7 9, 8 9] "m before n";
parameter
P{m} := [45,20,0,0,10,17,10,34,28] "Processing time";
M := sum{m} P "Big-M";
variable
T{m} "Starting time";
TEnd [0..M] "Makespan";
binary Y{D[m,n]|m<#m and n>m} "Disjonctions";
constraint
ToutesTerminees{m}: TEnd >= T+P;
Conj{U[m,n]}: T[m]+P[m] <= T[n];
Dis1{D[m,n]|m<#m and n>m}: T[m]+P[m] <= T[n]+M*Y;
Dis2{D[m,n]|m<#m and n>m}: T[n]+P[n] <= T[m]+M-M*Y;
minimize FinOrdo: TEnd;
Writep(FinOrdo,T);
end

3.11. Sequencing jobs on a bottleneck machine

(guer05-5)
7.4.

model guer05_5 "Sequencing jobs on a bottleneck machine";
set m,n := [1..7] "Job / Rank";
parameter
r{m}:=[2 5 4 0 0 8 9] "Earlist time";
p{m}:=[5 6 8 4 2 4 2] "Processing time";
d{m}:=[10 21 15 10 5 15 22] "Due date";
variable
t{n} "Starting time";
c{n} "End time";
T{n} "Tardiness";
binary u{m,n} "Job sequence";
constraint
Pos{n}: sum{m} u = 1;
Job{m}: sum{n} u = 1;
Sequence{n|n<#n}: t[n+1] >= t[n] + sum{m} p*u;
Debut{n}: t >= sum{m} r*u;
Completion{n}: c = t + sum{m} p*u;
Retard{n}: T >= c - sum{m} d*u;
minimize Cmax: c[#n];
integer parameter M{n}:=argmax{m} u;
Draw.Scale(20,20);
for{n} do Draw.Rect(M&'',t,1,c-t,1,M+1); end
Draw.Text(Cmax&' ',Cmax+1,1.1);
minimize SommeC: sum{n} c;
M{n}:=argmax{m} u;
minimize SommeT: sum{n} T;
M{n}:=argmax{m} u;
Writep(SommeC,Cmax,t);
end

1. Implement the three objectifs: (1) minimize the makespan, (2) minimize
the average processing time, (3) minimize the total tardiness.
2. Draw a Gantt diagram.
3. Create and run an model with 50 jobs.

3.12. Paint production (guer05-6)

7.5.

model guer05_6 "Paint production";
set i,j := [1..5] "Periods";
parameter
d{i} := [40,35,45,32,50] "Processing time";
c{i,j} "Setup time (cleaning time)" :=
[ 0,11, 7,13,11 , 5, 0,13,15,15
13,15, 0,23,11 , 9,13, 5, 0, 3 , 3, 7, 7, 7, 0
/* 0,7,5,11,6,0 , 6,0,9,9,7,6 , 10,8,0,6,10,10 ,
7,12,7,0,6,7 , 7,11,8,11,0,7 */ ];
variable Y{i};
binary variable X{i,j|i<>j};
constraint
A{i}: sum{j} X = 1;
B{j}: sum{i} X = 1;
C{i,j|i<>j and j>1}: Y[j] >= Y[i]+1-#i + X*#i;
minimize Cycle: sum{i,j} (d+c)*X;
Writep(Cycle);
end

3.13. Assembly line balancing (guer05-7)

7.6.

model guer05_7 "Assembly line balancing";
set t := [1..12] "Tasks";
p := [1..4] "Machines";
A{t,t} :=
[1 2, 1 3, 2 4, 2 5, 2 6, 3 6, 3 7, 4 9, 5 9,
6 8, 7 11, 8 9, 8 10, 9 12, 10 12, 11 10];
parameter D{t}:=[3,6,7,6,4,8,9,11,2,13,4,3] "Proc. time";
model gantt,g;
variable T{t};
constraint prec{A[i,j]}: T[i]+D[i] <= T[j];
minimize makespan: sum{t} T;
end
gantt;
parameter F{t}:= g.T;
variable
TT "Cycle time";
binary X{t,p} "Assign";
constraint
OneMaschPerTask{t}: sum{p} X = 1;
Sequence{A[i,j]}: sum{p} p*X[i,p] <= sum{p} p*X[j,p];
UpperBound{p}: sum{t} D*X <= TT;
minimize Cycle: TT;
Writep(TT);
parameter FF{p}:=min{t|X} F;
parameter f{t};
{p,t|X} (f[t]:=sum{t1 in t|X[t1,p] and t1<t} D[t1]+FF);
Draw.Scale(20,20);
Draw.Rect(0,0,50,10,1);
{p} Draw.Text('ligne '&p,0,p+.7);
{p} Draw.Text(' '&sum{t|X} D,2,p+.7);
{p,t|X} Draw.Rect(t&'',f+3,p,D,1,2,0,1);
end

3.14. Planning the production of bicycles (guer06-2)

8.1.

model guer06_2 "Planning the production of bicycles";
set n := [1..12] "Time periods";
parameter
D{n} "Demand" := [30000,15000,15000,25000,33000,40000,
45000,45000,26000,14000,25000,30000];
CN := 130 "Unit cost, normal time";
CS := 160 "Unit cost, over time";
CSt := 20 "Storage cost";
Capa := 30000 "Production capacilty";
SInit := 2000 "Inituial storage";
variable
XN{n} [0..Capa] "Quantity to produce, normal time";
XS{n} "Quantity to produce, overtime";
S{n} "Storage";
constraint
Periode{n}: XN + XS + if(n=1,SInit,S[n-1]) = D + S;
minimize Cost: sum{n} (CN*XN + CS*XS + CSt*S);
Writep(Cost);
end

3.15. Production of drinking glasses (guer06-3)

8.2.

model guer06_3 "Production of drinking glasses";
set
m := [1..12] "Week periods";
n := [1..6] "Products";
parameter
CAT := 390 "Workforce capacity";
CAM := 850 "Capacity machines";
CAZ := 1000 "Storage capacity";
d{n,m} "Demand" :=
[20, 22, 18, 35, 17, 19, 23, 20, 29, 30, 28, 32
17, 19, 23, 20, 11, 10, 12, 34, 21, 23, 30, 12
18, 35, 17, 10, 9, 21, 23, 15, 10, 0, 13, 17
31, 45, 24, 38, 41, 20, 19, 37, 28, 12, 30, 37
23, 20, 23, 15, 10, 22, 18, 30, 28, 7, 15, 10
22, 18, 20, 19, 18, 35, 0, 28, 12, 30, 21, 23];
C{n}:=[100 80 110 90 200 140] "Prod cost";
CS{n}:=[25 28 25 27 10 20] "Stagage cost";
SI{n}:=[50 20 0 15 0 10] "Initial storgae";
SF{n}:=[10 10 10 10 10 10] "Min. storage";
Tut{n} := [3 3 3 2 4 4] "Workforce time";
Tum{n} := [2 1 4 8 11 9] "Machine time";
Tzs{n} := [4 5 5 6 4 9] "Storage space";
variable
Q{n,m} "Quantity";
s{n,m} "Storage";
constraint
Dem{n,m}: if(m=1,SI,s[n,m-1]) + Q - s = d;
CapHom{m}: sum{n} Tut*Q <= CAT "Workforce";
CapMac{m}: sum{n} Tum*Q <= CAM "Machine";
CapSto{m}: sum{n} Tzs*s <= CAZ "Storage";
StockFin{n}: s[n,#m] >= SF "Final storage";
minimize Cost: sum{n,m} C*Q + sum{n,m} CS*s;
Writep(Cost);
end

3.16. Material Requirement Planning (guer06-4)

8.3.

model guer06_4 "Material Requirement Planning";
set n,m := [1..17] "Components";
parameter
B{n,m} "Quanlity of raw mat." :=
/4 1 2, 4 2 1, -- chassis monte
9 3 2, 9 4 2, 9 5 1, -- cabine montee
12 6 1, 12 7 2, 12 8 1, -- camion bleu
16 9 1, 16 10 1, 16 12 1, 16 13 1, 16 15 2, --
camion rouge
17 9 1, 17 11 1, 17 12 1, 17 14 1, 17 15 2/;
--...CA{n} "Raw material cost" =
-- ... modeling not finished ...
end

3.17. Production of elec. components (guer06-5)

8.4.

model guer06_5 "Production of elec. components";
set
m := [1..6] "Time periods";
n := [1..4] "Components";
parameter
CB := 0.5 "Cost to reduce production";
CA := 1.0 "Cost to augment production";
d{n,m} "Demand" :=
[1500 3000 2000 4000 2000 2500
1300 800 800 1000 1100 900
2200 1500 2900 1800 1200 2100
1400 1600 1500 1000 1100 1200];
C{n} := [20 25 10 15] "Prod cost";
CS{n}:= [.4 .5 .3 .3] "Storage cost";
SI{n}:= [10 0 50 0] "Initial stock";
SF{n}:= [50 10 30 10] "Stock minimal";
variable
Q{n,m} "Prod quantity";
s{n,m} "Storage";
B{m} "Lower production";
A{m} "Raise production";
constraint
Bal{n,m}: if(m=1,SI,s[n,m-1]) + Q = d+s;
Chg{m|m>1}: sum{n} Q + B = sum{n} Q[n,m-1] + A;
St{n}: s[n,#m] >= SF;
minimize
Cost: sum{n,m}(C*Q+CS*s) + sum{m|m>1}(CB*B+CA*A);
Writep(Cost);
end

3.18. Production of fiberglass (guer06-6)

8.5.

model guer06_6 "Production of fiberglass";
set m,n := [1..12] "Periods";
parameter
Capa{n} :=
[140,100,100,120,110,100,100,90,120,120,100,110];
Arcs{n,m} :=
/1 2 5 , 3 4 8 , 5 6 6 , 7 8 6 , 9 10 7 , 11 12 6,
2 4 0.2, 4 6 0.3, 6 8 0.2, 8 10 0.25, 10 12 0.3/;
variable x{n,m|Arcs};
constraint
Demand{m|m%2=0}: sum{n} x[n,m] = Capa + sum{n} x[m,n];
Cap{n|n%2=1}: sum{m} x <= Capa;
minimize Cost: sum{n,m|Arcs} Arcs*x;
Writep(Cost);
end

3.19. Assign prod batches to machines (guer06-7)

8.6.

model guer06_7 "Assign prod batches to machines";
set
n :=[1..10] "Batches";
m :=[1..5] "Machines";
binary variable x{m,n} "Assign";
parameter
c{m,n} "Production Cost"
:= [17 21 22 18 24 15 20 18 19 18
23 16 21 16 17 16 19 25 18 21
16 20 16 25 24 16 17 19 19 18
19 19 22 22 20 16 19 17 21 19
18 19 15 15 21 25 16 16 23 15];
v{m,n} "Processing time"
:= [ 8 15 14 23 8 16 8 25 9 17
15 7 23 22 11 11 12 10 17 16
21 20 6 22 24 10 24 9 21 14
20 11 8 14 9 5 6 19 19 7
8 13 13 13 10 20 25 16 16 17];
Ca{m} "Capacity"
:= [18 19 25 19 20];
constraint
Affectation{n}: sum{m} x = 1;
Capacite{m}: sum{n} v*x <= Ca;
minimize Cost: sum{m,n} c*x;
Writep(Cost);
end

3.20. Wagon load balancing (guer07-2)

Problem: The problem is from [5] chap 7-2. English translation in [28] chap
9.1.

model guer07_2 "Wagon load balancing";
set
n := [1..16] "Boxes";
m := [1..3] "Wagons";
parameter P{n}:=
[34,6,8,17,16,5,13,21,25,31,14,13,33,9,25,25] "Weight";
variable CMax "Max charge";
binary variable X{n,m} "Assign";
constraint
Affect{n}: sum{m} X = 1;
RegleCMax{m}: sum{n} P*X <= CMax;
--Bound: CMax >= 99; // to speed up the solver
minimize MaxCharge: CMax;
Writep(MaxCharge);
end

3.21. Barge loading (guer07-3)

9.2.

model guer07_3 "Barge loading";
set n :=[1..7] "Clients";
parameter C:=1500 "Capacity";
q{n}:=[ 12 31 20 25 50 40 60] "Av. qty";
t{n}:=[ 10 8 6 8 15 10 12] "Lot size";
p{n}:=[1000 600 600 700 1200 800 1100] "Price";
c{n}:=[ 80 70 85 80 73 70 80] "Trans. cost";
g{n}:=p-c*t "Profit";
integer variable x{n} [0..q] "Lots";
constraint Capa: sum{n} t*x <= C;
maximize Gain: sum{n} g*x;
Writep(Gain);
end

3.22. Tank loading (guer07-4)

9.3.

model guer07_4 "Tank loading";
set
q := [1..5] "Liquids";
r := [1..9] "Tanks";
parameter
Capa{r} := [500,400,400,600,600,900,800,800,800];
QIni{r} := [0 ,100, 0, 0, 0, 0,300, 0, 0];
LIni{r} := [0 , 1, 0, 0, 0, 0, 5, 0, 0];
Arrive{q} := [1200, 700,1000,450,1200];
binary variable X{q,r|QIni=0} "Assign";
constraint
PasMixer{r|QIni=0}: sum{q} X <= 1;
Suffit{q}: sum{r|QIni=0} Capa*X >=
Arrive - sum{r|LIni=q} (Capa-QIni);
minimize FreeCapa: sum{q,r|QIni=0} Capa*X;
Writep(FreeCapa);
end

3.23. Backing up files (guer07-5)

9.4.

model guer07_5 "Backing up files";
set
n := [1..16] "Files";
m := [1..3] "Disks";
parameter
A{n} := [26 35 52 77 88 94 137 164 253 364
372 388 406 432 461 851] "File size";
B := 1400 "Disk size";
variable NbU "Number of disks used";
constraint
NbUtilisees{n}: NbU >= sum{m} m*X;
ObjDansUneSeuleBoite{n}: sum{m} X = 1;
Capacite{m}: sum{n} A*X <= B;
minimize Nr: NbU;
Writep(Nr);
end

3.24. Cutting sheet metal (guer07-6)

9.5.

model guer07_6 "Cutting sheet metal";
set
m := [1..4] "Ordered sizes";
n := [1..16] "Patterns";
parameter
A{m,n} "Pattern table" :=
[1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0
2 1 0 2 1 0 3 2 1 0 5 4 3 2 1 0
0 0 0 2 2 2 1 1 1 1 0 0 0 0 0 0
0 1 3 0 1 3 0 2 3 5 0 1 3 5 6 8];
D{m}:=[8 13 5 15] "Demand";
integer variable X{n};
constraint Demande{m}: sum{n} A*X >= D;
minimize NR: sum{n} X;
Writep(NR);
end

3.25. Cutting steel bars for desk legs (guer07-7)

9.6.

model guer07_7 "Cutting steel bars for desk legs";
set
p := [1..3] "Leg types";
b := [1..2] "Bar types";
n := [1..12] "Patterns";
parameter
N := 6 "Number per bars";
A{n,p} "Pattern table" :=
[0 0 2, 0 1 1, 2 0 1, 0 2 0, 2 1 0, 3 0 0 --1.5m
0 1 2, 0 2 1, 1 0 2, 3 0 1, 0 3 0, 5 0 0]; --2m
D{p}:=[432 500 400] "Demand";
H{p}:=[40 60 70] "Height";
L{b}:=[150 200] "Length";
integer variable X{n} "Quantity";
constraint Demande{p}: sum{n} A*X >= D;
minimize Loss: sum{n} if(n<=N,L[1],L[2])*X -sum{p} D*H;
Writep(Loss);
end

3.26. Car rental (guer08-2)

10.1.

model guer08_2 "Car rental";
set n,m := [1..10] "Agencies";
parameter
C := 3 "Km unit cost";
Theo{n} := [10,06,08,11,09,07,15,07,09,12];
Reel{n} := [08,13,04,08,12,02,14,11,15,07];
X{n} := [00,04,04,06,07,07,01,01,02,00];
Y{n} := [00,04,02,02,00,05,05,02,00,03];
S{n} := Reel - Theo;
D{n,m} := (Abs(X[n]-X[m])^2 + Abs(Y[n]-Y[m])^2)^0.5;
variable Z{n,m|S[n]>0 and S[m]<0};
constraint
Surplus{n|S[n]>0}: sum{m|S[m]<0} Z = S[n];
Besoins{m|S[m]<0}: sum{n|S[n]>0} Z = -S[m];
minimize Cost: sum{n,m} C*D*Z;
Writep(Cost,Z);
end

3.27. Choosing the mode of transport (guer08-3)

10.2.

model guer08_3 "Choosing the mode of transport";
set n,k "Locations";
parameter
m{n,k} "Lower bound";
Ca{n,k} "Upper bound";
c{n,k} "Cost";
Read{n,k}('guer08-3.dat', n,k,m,Ca,c);
parameter
S := 1 "Source";
T := #n "Sink";
F := 180 "Quantity to transport";
variable p{n,k|Ca} [m,Ca];
constraint
Flow{n|n<>T and n<>S}: sum{k|Ca[k,n]} p[k,n] =
sum{k|Ca[n,k]} p[n,k];
QTY: sum{n|n<>S and Ca[S,n]} p[S,n] = F;
minimize Cost: sum{n,k|Ca} c*p;
Writep(Cost);
end

3.28. Depot location (guer08-4)

10.3.

model guer08_4 "Depot location";
set
n := [1..12] "Locations";
m := [1..12] "Costumers";
parameter
c{n,m} default 1000000 "Cost" :=
[100 80 50 50 60 100 120 90 60 70 65 110
120 90 60 70 65 110 140 110 80 80 75 130
140 110 80 80 75 130 160 125 100 100 80 150
160 125 100 100 80 150 190 150 130 . . .
190 150 130 . . . 200 180 150 . . .
200 180 150 . . . 100 80 50 50 60 100
100 80 50 50 60 100 120 90 60 70 65 110
120 90 60 70 65 110 140 110 80 80 75 130
140 110 80 80 75 130 160 125 100 100 80 150
160 125 100 100 80 150 190 150 130 . . .
190 150 130 . . . 200 180 150 . . .
200 180 150 . . . 100 80 50 50 60 100];
f{n}:=[35 90 100 40 30 90 90 30 40 100 90 35] "Cost";
d{m}:=[120 80 75 100 110 100 90 60 30 150 95 120] "
Demand";
Ca{n}:=[300 250 100 180 275 300 200 220 270 250 230
180] "Capacity";
variable x{n,m} [0,1] "Transport (%)";
binary variable y{n} "Assign";
constraint
Satisfac{m}: sum{n} x = 1;
Capacite{n}: sum{m} d*x <= Ca*y;
minimize Cost: sum{n} 100*f*y + sum{n,m} c*x;
Writep(Cost);
end

3.29. Heating oil delivery (guer08-5)

10.4.

model guer08_5 "Heating oil delivery";
set n,m := [1..7] "Clients+depot";
-- Attention, contrairement au livre, les noeuds sont ici
-- indexes dans l'ordre : Donges, Brain, Carquefou,
Guerande,
-- Haie Fouassiere, Ponts-de-Ce, Mesanger. Les fichiers
de
-- donnees sont reordonnes en consequence.
parameter
Qmax := 39000;
D{n,m} "distances" :=
[0 148 55 32 70 140 73
148 0 93 180 99 12 72
55 93 0 85 20 83 28
32 180 85 0 100 174 99
70 99 20 100 0 85 49
140 12 83 174 85 0 73
73 72 28 99 49 73 0 ];
Q{n}:=[0 14000 3000 6000 16000 15000 5000] "demand";
variable C{n};
binary variable X{n,m|n<>m};
constraint
enter{n|n>1}: sum{m} X = 1;
leave{m|m>1}: sum{n} X = 1;
First{n|n>1}: C <= Qmax-(Qmax-Q)*X[1,n];
j_Apres_i{n,m|n<>m and n>1 and m>1}: C[m] >= C[n]+Q[m]-
Qmax + Qmax*X[n,m]+(Qmax-Q[m]-Q[n])*X[m,n];
Bound{n|n>1}: Q <= C <= Qmax;
minimize kilometres: sum{n,m|n<>m} D*X;
Writep(kilometres);
end

3.30. Combining diff modes of transport (guer08-6)

10.5.

model guer08_6 "Combining diff modes of transport";
set
n := [1..5] "Locations";
m,m1 := [1..3] "Modes (rail, route, air)";
parameter
C{m,n}:=[30 40 30 60 . , 20 40 50 50 . , 40 10 60 40
.];
T{m,m1}:=[0 2 1, 2 0 1, 2 1 0] "Mode change mode";
binary variable
X{m,n|n<#n} "Assign";
Y{m,m1,n|n>1 and n<#n} "Change Mode";
constraint
UnMode{n|n<#n}: sum{m} X = 1;
UnTransf{n|n>1 and n<#n}: sum{m,m1} Y = 1;
Cons{n,m,m1|n>1 and n<#n}: X[m,n-1] + X[m1,n] >= 2*Y;
minimize Cost: sum{m,n|n<#n} C*X + sum{n,m,m1|n<#n} T*Y;
Writep(Cost);
end

3.31. Fleet planning for vans (guer08-7)

10.6.

model guer08_7 "Fleet planning for vans";
set
t := [1..3] "Contracts";
m,m1 := [1..6] "Months";
parameter
Z := 200 "Initial stock";
B{m} := [ 430, 410, 440, 390, 425, 450] "Needs";
C{t} := [1700, 2200, 2600] "Cost";
variable Y{t,m} "Number vans";
constraint Cover{m}: if(m<=2,Z) + sum{t,m1|m1 >=
Max(1,m-t-1) and m1<=Min(m,#m-t-1)} Y[t,m1] >= B;
minimize Cost: sum{t,m} C*Y;
Writep(Cost);
end

3.32. Flight connections at a hub (guer09-2)

11.1.

model guer09_2 "Flight connections at a hub";
set n,m := [1..6] "Locations";
parameter
P{n,m} "Number of passengers" :=
[ 35 12 16 38 5 2 , 25 8 9 24 6 8
12 8 11 27 3 2 , 38 15 14 30 2 9
-1000 9 8 25 10 5 , -1000 -1000 -1000 14 6 7];
variable X{n,m};
constraint
OneIn{n}: sum{m} X = 1;
OneOut{m}: sum{n} X = 1;
maximize NbPassagers: sum{n,m} P*X;
Writep(NbPassagers);
end

3.33. Composing flight crews (guer09-3)

11.2.

model guer09_3 "Composing flight crews";
set n,m,m1 "Number of pilots";
parameter
G{n,m} "Score" :=
/1 2 30, 1 4 24, 1 5 25, 2 3 27, 2 4 28, 2 6 27,
3 6 26, 3 7 30, 4 5 29, 4 6 21, 5 6 30, 5 8 30,
6 7 28, 6 8 36, 7 8 25/;
binary variable X{n,m|G};
constraint Sep{n}: sum{m,m1|m=n or m1=n} X[m,m1] <= 1;
maximize Score: sum{n,m} G*X;
Writep(Score);
end

3.34. Scheduling flight langings (guer09-4)

11.3.

model guer09_4 "Scheduling flight langings";
set n,m := [1..10] "Aeroplanes";
parameter
M := 1000 "Big-M";
inf:=999999;
Deb{n}:=[129 195 89 96 110 120 124 126 135 160]
"Begin time window";
Fin{n}:=[559 744 510 521 555 576 577 573 591 657]
"End time window";
Pre{n}:=[155 258 98 106 123 135 138 140 150 180]
"Prefered time";
Ava{n}:= [10 10 30 30 30 30 30 30 30 30]
"Penalty for too early";
Ret{n}:= [10 10 30 30 30 30 30 30 30 30]
"Penalty for too late" ;
Sep{n,m} "Delay" :=
[99999 3 15 15 15 15 15 15 15 15
3 99999 15 15 15 15 15 15 15 15
15 15 99999 8 8 8 8 8 8 8
15 15 8 99999 8 8 8 8 8 8
15 15 8 8 99999 8 8 8 8 8
15 15 8 8 8 99999 8 8 8 8
15 15 8 8 8 8 99999 8 8 8
15 15 8 8 8 8 8 99999 8 8
15 15 8 8 8 8 8 8 99999 8
15 15 8 8 8 8 8 8 8 99999];
variable
T{n} [Deb..Fin] "Langing time";
A{n} [0..Pre-Deb] "Too early";
R{n} [0..Fin-Pre] "Too late";
binary variable X{n,m|n<m};
constraint
cart1{n,m|m<n}: T + Sep <= T[m] + M*X[m,n];
Ecart2{n,m|m>n}: T + Sep <= T[m] - M*X[n,m] + M;
Avance{n}: A >= Pre - T;
Retard{n}: R >= T - Pre;
AvOuRe{n}: T = Pre -A + R;
minimize Penalty: sum{n} (Ava*A + Ret*R);
Writep(Penalty);
end

3.35. Airline hub location (guer09-5)

11.4.

model guer09_5 "Airline hub location";
set n,m,k := [1..6] "Locations";
parameter
p := 2 "Number of hubs";
Alpha := 80/100 "Cost reduction";
D{n,m} "distances" :=
[ 0, 870, 702, 4420, 4300, 4380
870, 0, 840, 3660, 3620, 3640
702, 840, 0, 4200, 4100, 4140
4420, 3660, 4200, 0, 140, 571
4300, 3620, 4100, 140, 0, 501
4380, 3640, 4140, 571, 501, 0];
Q{n,m} "transport quantity" :=
[ 0, 500, 1000, 300, 400, 1500
1500, 0, 250, 630, 360, 1140
400, 510, 0, 460, 320, 490
300, 600, 810, 0, 820, 310
400, 100, 420, 730, 0, 970
350, 1020, 260, 580, 380, 0];
C{n,m,k,n1 in n}:=D[n,k]+Alpha*D[k,n1]+D[n1,m];
binary variable
Z{n,m,k,n1 in n};
X{n};
constraint
pHubs: sum{n} X = p "Exactly p Hubs";
DeuxVilDeuxHubs{n,m} : sum{k,n1 in n} Z[n,m,k,n1] = 1;
kHub{n,m,k,n1 in n}: Z[n,m,k,n1] <= X[k];
mHub{n,m,k,n1 in n}: Z[n,m,k,n1] <= X[n1];
minimize Cost:
sum{n,m,k,n1 in n} C[n,m,k,n1]*Q[n,m]*Z[n,m,k,n1];
Writep(Cost);
end

3.36. Planning a flight tour (guer09-6)

11.5.

model guer09_6 "Planning a flight tour";
--/Bordeaux Lille Lyon Marseille Nantes Paris Toulouse
/;
parameter
d{n,m} "Distance" :=
/1 2 786, 1 3 549, 1 4 657, 1 5 331, 1 6 559,
1 7 250, 2 3 668, 2 4 979, 2 5 593, 2 6 224,
2 7 905, 3 4 316, 3 5 607, 3 6 472, 3 7 467,
4 5 890, 4 6 769, 4 7 400, 5 6 386, 5 7 559,
6 7 681/;
binary variable y{n,m};
constraint
Entree{m}: sum{n|n<>m} y = 1;
Sortie{n}: sum{m|n<>m} y = 1;
S1: y[2,6] + y[6,2] <= 1;
S2: y[1,7] + y[7,1] <= 1;
minimize Trajet: sum{n,m} if(n<m,d[n,m],d[m,n])*y;
Writep(Trajet);
end

3.37. Network reliability (guer10-2)

12.1.

model guer10_2 "Network reliability";
set
n,m := [1..11] "nodes";
G{n,m} "arcs" :=
[1 2, 1 3, 1 11, 2 3, 2 8, 2 9, 3 4, 3 9, 3 10,
3 11, 4 5, 4 6, 4 11, 5 9, 5 11, 6 7, 6 9, 6 10,
7 8, 7 10, 8 10, 9 10];;
G{n,m}:=G[n,m] or G[m,n];
parameter
s := 10 "Source";
t := 11 "Sink";
variable Phi{n,m|G[n,m] or G[m,n]};
constraint
A{n|n<>s and n<>t}: sum{m} Phi[m,n] = sum{m} Phi[n,m
];
B{n|n<>s and n<>t}: sum{m} Phi[n,m] <= 1;
C: sum{n} Phi[n,s] = 0;
maximize Debit: sum{n} Phi[s,n];
Writep(Debit);
parameter x{n} :=[2 4 2 1 2 4 6 6 4 5 1];
y{n} :=[1 1 2 4 3 4 4 1 3 2 3];
Draw.Scale(100,100);
{G[n,m]} Draw.Line(x[n],y[n],x[m],y[m],0,if(Phi[n,m
],4,1));
{n} Draw.Circle(n&'',x,y,.3,1,0);
end

3.38. Dimensioning of a mobile phone network

(guer10-3)
12.2.

model guer10_3 "Dimensioning of a mobile phone network";
set
c := [1..10] "Cells";
n := [1..5] "Hubs nodes";
parameter
K := 48 "Capacity";
C{c,n} "Cost" :=
[15 8 7 11 10, 9 11 8 5 14, 12 6 7 15 15
17 5 9 18 24, 8 22 21 19 6, 7 25 15 9 17
19 25 21 20 22, 20 9 15 18 25, 21 22 14 16 20,
25 24 13 4 11];
T{c}:=[22 12 20 12 15 25 15 14 8 22] "traffic";
D{c}:=[2 2 2 2 3 1 3 2 2 2 ];
binary variable X{c,n} "Assign";
constraint
Divers{c}: sum{n} X = D;
CapaciteAnneau: sum{c,n|n<#n} (T/D)*X <= 2*K;
minimize Cost: sum{c,n} C*X;
Writep(Cost);
end

3.39. Routing tlephone calls (guer10-4)

12.3.

model guer10_4 "Routing tlephone calls";
set
a,b := [1..6] "Nodes";
p := [1..5] "Pair numbers";
c := [1..17] "Paths";
parameter
Capa{a}:=[300,120,300,200, 80, 70] "Capacity";
Chem{c,a} "Paths table" :=
[2 0 0 0 0 0, 1 3 0 0 0 0, 1 4 6 5 0 0, 1 4 6 0 0 0,
1 3 5 0 0 0, 2 5 0 0 0 0, 2 3 4 6 0 0, 1 4 0 0 0 0,
2 3 4 0 0 0, 1 3 5 6 0 0, 2 5 6 0 0 0, 2 1 4 0 0 0,
3 4 0 0 0 0, 5 6 0 0 0 0, 4 6 0 0 0 0, 1 2 5 0 0 0,
3 5 0 0 0 0];
Pi{c}:=[1 1 1 2 2 2 2 3 3 3 3 4 4 4 5 5 5];
D{p}:=[100 80 75 100 70] "Demand";
integer variable Phi{c} "Flow";
constraint
RespD{p}: sum{c|Pi=p} Phi <= D;
RespC{b}: sum{c,a|Chem[c,a]=b} Phi[c] <= Capa[b];
maximize Flow: sum{c} Phi;
Writep(Flow);
end

3.40. Construction of a cabled networkt (guer10-5)

12.4.

model guer10_5 "Construction of a cabled networkt";
set n,m := [1..6] "nodes";
parameter
D{n,m} "Distances" :=
[ 0 120 92 265 149 194, 120 0 141 170 93 164, 92 141
0 218 103 116, 265 170 218 0 110 126, 149 93 103
110 0 72, 194 164 116 126 72 0];
variable Level{n};
binary variable X{n,m|n<>m};
constraint
NbConnexions: sum{n,m} X = #n-1;
Connexe{n,m|m<>n}: Level[m] >= Level[n]+1-#n+#n*X;
minimize Cost: sum{n,m} D*X;
Writep(Cost);
end

3.41. Scheduling of telecomm. via satellite

(guer10-6)
12.5.

model guer10_6 "Scheduling of telecomm. via satellite";
set n,m := [1..4] "stations";
parameter
T{n,m}:=[7 12 11 15, 15 8 13 9, 17 12 6 10, 6 13 15
11] "traffic";
LB;
binary variable X{n,m|T};
variable PMin;
constraint
UnRecParEme{n}: sum{m} X = 1;
UnEmeParRec{m}: sum{n} X = 1;
ReglePMin{n}: sum{m} T*X >= PMin;
parameter i:=2;
while i>0 do
maximize obj: PMin;
LB:=LB+PMin;
Writep(PMin,T,X);
T{n,m}:=if(X=1,T-PMin,T);
--if(XX,T-PMin,T);
ClearData(X);
i:=i-1;
end;
Writep(LB);
end

3.42. Localisation of GSM transmitters (guer10-7)

12.6.

model guer10_7 "Localisation of GSM transmitters";
set
n := [1..15] "Villages to deserve";
m := [1..7] "Locations of transmitters";
Cv{m,n} "Covering" :=
[1 1, 1 2, 1 4, 2 2, 2 3, 2 5, 3 4, 3 7, 3 8, 3 10,
4 5, 4 6, 4 8, 4 9, 5 8, 5 9, 5 12, 6 7, 6 10, 6 11,
6 12, 6 15, 7 12, 7 13, 7 14, 7 15];
parameter
B := 10 "Budget maximal";
p{n}:=[2,4,13,6,9,4,8,12,10,11,6,14,9,3,6]"Population";
c{m}:=[1.8, 1.3, 4.0, 3.5, 3.8, 2.6, 2.1] "Cost";
binary variable
x{m} "Transmitter there?";
y{n} "Couvered?";
constraint
CouvCom{n}: sum{m|Cv} x >= y;
Budget: sum{m} c*x <= B;
maximize MaxPop: sum{n} p*y;
Writep(MaxPop);
end

3.43. Choice of loans (guer11-2)

13.1.

model guer11_2 "Choice of loans";
set
i := [1..3] "Shops";
j := [1..3] "Banks";
parameter
n := 8 "Time horizon";
Mmax := 3000000 "Max amount to loan";
T{j,i}:=[0.05 0.065 0.061, 0.052 0.062 0.062,
0.055 0.058 0.065] "Interest";
P{i}:=[2500000,1000000,1700000] "Price of shop";
variable X{j,i};
constraint
A{i}: sum{j} X = P;
B{j}: sum{i} X <= Mmax;
minimize Annuities: sum{j,i} X*T/(1-(1+T)^(-n));
Writep(Annuities);
end

3.44. Publicity campaign (guer11-3)

13.2.

model guer11_3 "Publicity campaign";
set n "Various medias";
string parameter nN{n};
parameter
ca{n} "Potential";
c{n} "Cost";
nd{n} "Usage";
q{n} "Quality index";
B := 250000 "Max budget";
C := 100000 "Number of clients to reach";
Read{n}('guer11-3.dat', n,nN,c,ca,q,nd);
integer variable m{n} [0..nd];
constraint
BudgetMax : sum{n} c*m <= B;
CibleMini : sum{n} ca*m >= C;
maximize QualPercep: sum{n} q*m;
Writep(QualPercep,m);
end

3.45. Portfolio selection (guer11-4)

13.3.

model guer11_4 "Portfolio selection";
set
n := [Dash 'Ilog France' Eurodecision 'Ilog Usa'
'France Telecom' BNP];
Tech{n} := n<=3;
France{n}:= ~(n=1 or n=4);
parameter
C := 100000 "Capital";
r{n}:=[5.3, 6.2, 5.1, 4.9, 6.5, 3.4] "Interest";
variable v{n} [5000..40000];
constraint
Technologiques : sum{Tech} v <= 0.3*C;
Francaises : sum{France} v >= 0.5*C;
CapitalMaximum : sum{n} v <= C;
maximize RetourTotal: sum{n} r/100*v;
Writep(RetourTotal,v);
end

3.46. Finacing an early retirment scheme (guer11-5)

13.4.

model guer11_5 "Finacing an early retirment scheme";
set
n := [1..7] "Time horizon";
e := [1..3] "Bond types";
parameter
Taux := 3.2 "Annual interest rate";
F{n}:=[1000 600 640 480 760 1020 950] "Needs";
p{e}:=[1.0, 0.8, 0.5] "Priice";
t{e}:=[7.0, 7.0, 6.5] "Interestt";
T{e} := p*t/100;
TotalaPayer:=sum{n} F;
variable
CT "Capital";
I{n} "Annual investment";
integer N{e} [0..5000] "Quantity to buy";
constraint
year1: CT - sum{e} p*N - I[1] = F[1];
year{n|n>1 and n<6}:sum{e} T*N+(1+Taux/100)*I[n-1]-I=F;
year6: (p[1]+T[1])*N[1] + (p[2] + T[2])*N[2]
+ T[3]*N[3] + (1+Taux/100)*I[5] - I[6] = F[6];
year7: (p[3]+T[3])*N[3] + (1+Taux/100)*I[6] = F[7];
minimize Capital: CT;
Writep(CT,TotalaPayer,N);
parameter EE{n}:= if(n<=5,sum{e} p*N,n=6,p[3]*N[3],0);
II{n}:=I;
FF{n}:=F;
Draw.Scale(20,-5);
{n} Draw.Rect(n,0,1,EE/100,2,0,1);
{n} Draw.Rect(n,EE/100,1,II/100,3,0,1);
{n} Draw.Rect(n,EE/100+II/100,1,FF/100,4,0,1);
end

3.47. Family budget (guer11-6)

13.5.

model guer11_6 "Family budget";
set
n := [1..12] "Months";
i := [VieCourante, Loyer, Voiture, Telephone,
Impots, 'EDF/GDF'] "Expenses";
parameter
A := 900 "Public allocations";
S := 11500 "Salary";
D{i} := [3300, 3800, 2000, 800, 600, 5000] "Expenses";
m{i} := [1,1,1,2,4,6] "mensualities";
d{n,i} := if(n%m,0,D) "Expenses per month";
variable
l{n} [1000..9999999] "Leigure";
e{n} "Sauvings";
constraint Bal{n}: sum{i} d + l + e <= S + A + e[n-1];
maximize Lei: sum{n} l;
Writep(Lei,l,e);
end

3.48. Choice of expansion projects (guer11-7)

13.6.

model guer11_7 "Choice of expansion projects";
set
a := [1..5] "Time horizon";
n := [1..5] "Projects";
parameter
ca{a}:=[4.8 6.0 4.8 4.2 3.5] "Capital";
c{n,a} "Project cost" :=
[1.80 2.40 2.40 1.80 1.50
1.20 1.80 2.40 0.60 0.50
1.20 1.00 0 0.48 0
1.40 1.40 1.20 1.20 1.20
1.60 2.10 2.50 2.00 1.80];
b{n}:=[10.8 4.8 3.2 4.44 12.25] "Profit";
binary variable x{n};
constraint Invest{a}: sum{n} c*x <= ca;
maximize Profit: sum{n} b*x;
Writep(Profit,x);
end

3.49. Mean variance portfolio (guer11-8)

13.7.

model guer11_8 "Mean variance portfolio";
set
i,j := [Hardware Software Show_biz T_bills] "
Investment types";
parameter
r{i}:=[8 9 12 7] "expected return";
V{i,j} "co/variance" :=
[ 4 3 -1 0 , 3 6 1 0
-1 1 10 0 , 0 0 0 0];
Target := 8.5 "minimum target yield";
variable x{i} "Fraction of investment";
binary variable y{i};
constraint A: sum{i} x = 1;
constraint B: sum{i} r*x >= Target;
constraint C: sum{i} y <= 2;
constraint D{i}: x <= y;
minimize risk: sum{i,j} x[i]*V[i,j]*x[j];
Writep(risk,x);
end

3.50. Assigning personnel to machines (guer12-2)

14.1.

model guer12_2b "Assigning personnel to machines";
set n,m := [1..6] "Persons";
parameter
P{n,m} "Productivity" :=
[13 24 31 19 40 29 , 18 25 30 15 43 22
20 20 27 25 34 33 , 23 26 28 18 37 30
28 33 34 17 38 20 , 19 36 25 27 45 24];
constraint
UneMachParPers{n}: sum{m} X = 1;
UnePersParMach{m}: sum{n} X = 1;
maximize ProdTotale: sum{n,m} P*X;
Writep(ProdTotale);
--maximize MinProd: min{n} (sum{m} P*X);
--Writep(MinProd); //maximize a minimal prodictivity !
end

3.51. Scheduling nurses (guer12-3)

14.2.

model guer12_3b "Scheduling nurses";
set n := [1..12] "Time windows";
parameter
B{n}:=[35,40,40,35,30,30,35,30,20,15,15,15] "Demand";
D := 80 "Total nurses";
integer variable
X{n} "Number of nurses beginning";
S{n} "an overtime window";
constraint
Dispo frozen: sum{n} X <= D;
NbSup0{n}: S <= 0;
NbSup1{n} frozen: S <= X;
tr1 : S[8]+X[9]+X[10]+X[12]+X[1] >= B[1];
tr2 : S[9]+X[10]+X[11]+X[1]+X[2] >= B[2];
tr3 : S[10]+X[11]+X[12]+X[2]+X[3] >= B[3];
tr4 : S[11]+X[12]+X[1]+X[3]+X[4] >= B[4];
tr5 : S[12]+X[1]+X[2]+X[4]+X[5] >= B[5];
tr6 : S[1]+X[2]+X[3]+X[5]+X[6] >= B[6];
tr7 : S[2]+X[3]+X[4]+X[6]+X[7] >= B[7];
tr8 : S[3]+X[4]+X[5]+X[7]+X[8] >= B[8];
tr9 : S[4]+X[5]+X[6]+X[8]+X[9] >= B[9];
tr10: S[5]+X[6]+X[7]+X[9]+X[10] >= B[10];
tr11: S[6]+X[7]+X[8]+X[10]+X[11] >= B[11];
tr12: S[7]+X[8]+X[9]+X[11]+X[12] >= B[12];
minimize NbNurses: sum{n} X "Total nurses without
overtime";
Writep(NbNurses);
Unfreeze(Dispo,NbSup1); Freeze(NbSup0);
minimize TotalSup: sum{n} S "with overtime";
Writep(TotalSup);
end

3.52. Establishing a college timetable (guer12-4)

14.3.

model guer12_4 "Establishing a college timetable";
set
p := [1..9] "Teachers";
c := [1..2] "Classes";
td:= [1..20];
parameter
t := 4 "Time slots per day";
d := 5 "Days in a week";
NbC{p,c}:=[1 1, 3 3, 2 2, 0 4, 4 0, 3 3, 1 1, 1 0, 0 1]
"Number of courses";
string Prf{p}:=['a' 'b' 'c' 'd' 'e' 'f' 'g' 'h' 'i'];
variable X{p,c,td} "Assign";
constraint
TousPlanif{p,c}: sum{td} X = NbC;
Classe{c,td}: sum{p} X <= 1;
Prof{p,td}: sum{c} X <= 1;
Sport1: X[8,1,15] = 1;
Sport2: X[9,2,15] = 1;
Devoir{p,c}: X[p,c,1] = 0;
Math{td|td<=2}: X[4,2,td] = 0;
Biologie{c,td|c<=2 and td>8 and td<13}: X[2,c,td] = 0;
AuPlusUn{p,c,td|td<=d}:
sum{k in td|k>=(td-1)*t+1 and k<=td*t} X[p,c,k] <= 1;
minimize Trou: sum{p,c,td|td<=d} (X[p,c,(td-1)*t+1]
+ X[p,c,(td-1)*t+4]);
parameter Res{c,td} := argmin{p|X} p;
string parameter Table{c,td} := Prf[Res];
Writep(Trou);
end

3.53. Exam scheduling (guer12-5)

14.4.

model guer12_5a "Exam scheduling";
set
e,f := ['Analyse de donnees','Analyse numerique','C++',
'Genie logiciel','Gestion de production','Java',
'Modeles et Algorithmique','Programmation logique',
'Programmation mathematique', 'Statistiques',
'Systemes a evenements discretes'];
t := ['1:8-10h','1:10-12','1:14-16','1:16-18','2:8-10h
',
'2:10-12','2:14-16','2:16-18'];
--e,f := [1..11];
--t := [1..8];
parameter
Incomp{e,f} "Incompatibles" :=
[0 1 0 0 1 0 1 0 0 1 1 , 1 0 0 0 1 0 1 0 0 1 1
0 0 0 1 1 1 1 0 1 1 1 , 0 0 1 0 1 1 1 0 0 1 1
1 1 1 1 0 1 1 1 1 1 1 , 0 0 1 1 1 0 1 0 1 1 1
1 1 1 1 1 1 0 1 1 1 1 , 0 0 0 0 1 0 1 0 0 1 1
0 0 1 0 1 1 1 0 0 1 1 , 1 1 1 1 1 1 1 1 1 0 1
1 1 1 1 1 1 1 1 1 1 0];
binary variable X{e,t} "Assign";
constraint
AllScheduled{e}: sum{t} X = 1 "Each exam takes place";
Incomp1{e,f,t|Incomp}: X[e,t] + X[f,t] <= 1;
--T{t}: Sum{e} X <= 1;
--maximize any: X[1,1]+X[2,2];
minimize length: sum{e,t} t^2 * X;
---binary variable Y{t} <- or{e} X;
---minimize tranche_horaire: sum{t} Y;
--maximize tranche: sum{e} X[e,3];
Write(' %10s\n', {t}t);
Write{e}('%31s %10s\n', e, {t} if (X,'X',' '));
model data1 "larger data set (for testing)";
e := [1..100];
t := [1..20];
Incomp{e,f|e<>f}:=if(e<f and Rnd(0,1)<.4,1);
Incomp{e,f|e>f}:=Incomp[f,e];
end
end

3.54. Personnel assignment (guer12-6)

Problem: The model is from [5], Chapter 12.6. and [28].
Modeling Steps

model guer12_6 "Personnel assignment";
set
n "Products";
L,L1 "Production lines";
parameter
g{n} "Gain";
d{n,L} "Processing time";
m{L} "Time capacity";
T{L,L1} "Transfert possible";
tm{L} "Max hours to transfert";
variable
q{n} "Quantity";
h{L} "Work hours";
t{L,L1|T} "Hours transfered";
constraint
ChargeLigne{L}: sum{n} d*q <= m;
Travail{L}: sum{n} d*q <= h;
Equilibrage{L1}: h - sum{L|T[L,L1]} t[L,L1] + sum{L|T[
L1,L]} t[L1,L] = m;
MaxTranfert{L}: sum{L1|T=1} t <= tm;
expression Profit: sum{n} g*q;
maximize obj1: Profit subject_to ChargeLigne;
Writep(Profit,q);
maximize obj2: Profit subject_to Travail,Equilibrage,
MaxTranfert;
Writep(Profit,q);
model data; --- in the book
n := [1..4];
L := [1..5];
g{n} := [7, 8, 9, 7];
d{n,L} :=
[1.3, 0.9, 2.0, 0.3, 0.9
1.8, 1.7, 1.4, 0.6, 1.1
1.3, 1.2, 1.3, 1.0, 1.4
0.9, 1.1, 1.0, 0.9, 1.0];
m{L} := [4500, 5000, 4500, 1500, 2500];
T{L,L1} := /1 2 1, 1 3 1, 1 4 1, 2 3 1, 2 5 1,
3 1 1, 3 2 1, 3 4 1, 4 5 1, 5 1 1, 5 2 1, 5 3 1/;
tm{L} := [400 800 200 500 300];
end
model data1; --- a random data set
n := [1..20];
L := [1..5];
g{n} := Trunc(Rnd(6,10));
d{n,L} := Trunc(Rnd(10,20))/10;
m{L} := Trunc(Rnd(1500,6000));
T{L,L1|L<>L1} := if(Rnd(0,1)<.3,1,0);
tm{L} := Trunc(Rnd(200,900));
end
end

3.55. Personnel at a construction site (guer12-7)

14.6.

model guer12_7 "Personnel at a construction site";
set n := [1..6] "Time horizon";
parameter
CouArr := 100 "Arrival cost";
CouDep := 160 "Departure cost";
CouSur := 200 "Overtime cost";
CouSou := 200 "Undertime cost";
EffDeb := 3 "Initial needs";
EffFin := 3 "Final needs";
Besoin{n} := [4,6,7,4,6,2] "Needs";
integer variable
XEff{n};
XArr{n} [0..3];
XDep{n};
XSur{n};
XSou{n};
constraint
Satisf{n}: XEff - XSur + XSou = Besoin;
PremierMois: XEff[1] = EffDeb + XArr[1];
DernierMois: EffFin = XEff[#n] - XDep[#n];
Autres{n|n>1}: XEff = XEff[n-1] + XArr - XDep[n-1];
Dep{n}: XDep <= (1/3)*XEff;
Sou{n}: XSou <= (1/4)*XEff;
minimize Cost: sum{n} (CouArr*XArr + CouDep*XDep + CouSur
*XSur + CouSou*XSou);
Writep(Cost);
end

3.56. Water conveyance mamangement (guer13-2)

15.1.

model guer13_2 "Water conveyance mamangement";
set n,m := [1..12] "nodes";
parameter
s := #n-1 "Source";
t := #n "Sink";
Capa{n,m} :=
/1 3 20, 1 5 15, 1 6 12, 2 5 6, 2 6 22, 3 4 15,
3 5 10, 4 8 7, 4 9 10, 5 8 10, 5 9 15, 5 10 15,
6 7 22, 7 9 10, 7 10 10, 8 12 18, 9 12 15, 10 12 20,
11 1 35, 11 2 25/;
variable Phi{n,m|Capa} [0..Capa];
constraint FL{n|n<>s and n<>t}:
sum{m} Phi[n,m] = sum{m} Phi[m,n];
maximize Flow: sum{n} Phi[n,t];
Writep(Flow);
end

3.57. CCTV surveillance (guer13-3)

15.2.

model guer13_3 "CCTV surveillance";
set i,j:=[1..49]; -- the network
e{i,j}:=[(1 3),(1 2),(2 39),(2 41),(3 4),(4 5),(4 6),
(4 9),(6 8),(6 7),(9 10),(10 11),(11 21),(12 15),
(12 13),(13 14),(14 15),(14 18),(15 19),(15 16),
(16 20),(17 18),(18 19),(19 20),(20 21),(21 22),
(21 23),(22 25),(23 32),(24 25),(25 26),(25 30),
(26 27),(26 28),(28 29),(29 31),(30 31),(31 32),
(31 33),(32 39),(33 34),(33 37),(34 35),(35 36),
(37 38),(37 43),(38 40),(39 40),(40 41),(41 42),
(43 44),(44 49),(44 45),(45 46),(45 47),(47 48)];
parameter X{i}:= [42 170 6 32 71 38 76 49 32 56 50 6
3 4 19 29 8 16 32 43 65 69 78 40 86 92 91 110 116
93 114 124 142 130 145 138 163 170 144 177 197 212
190 186 196 158 183 182 197] "xy-coordinate";
Y{i}:= [3 55 84 73 88 64 78 34 88 100 103 98 110 133
120 116 160 155 146 138 125 132 114 188 159 167 175
172 159 144 152 132 162 185 193 214 169 150 116 130
136 184 182 194 220 225 216 230 197];;
e{i,j}:= e[i,j] or e[j,i];
binary variable x{i};
constraint A{i,j|e[i,j]}: x[i] + x[j] >= 1;
minimize Nr: sum{i} x;
Writep(Nr);
Draw.Scale(2,2);
{e[i,j]} Draw.Line(X[i],Y[i],X[j],Y[j]);
{i} Draw.Circle(i&'',X,Y,4,if(x,3,1),0);
end

3.58. Rigging elections (guer13-4)

Problem: A complete problem description and how to model it mathemati-
cally is explained in [5], Chapter 13.4.
Modeling Steps

model guer13_4 "Rigging elections";
set
q := [1..14] "Quarters";
c := [1..46] "Districts";
parameter
R := 6;
M{c}:= [0,0,1,1,0,0,0,1,0,1,1,1,1,1,0,0,0,1,1,0,
0,0,0,0,0,0,1,1,1,0,0,0,0,0,1,1,1,1,1,1,0,1,1,0,0,1];
A{c,q} := /*$I 'guer13-4.dat' */ ;

binary variable X{c};
constraint
Partition{q}: sum{c} A*X = 1;
NombreRequis: sum{c} X = R;
minimize NombreSieges: sum{c} M*X;
Writep(NombreSieges,X);
end

3.59. Gritting roads (guer13-5)

Problem: Leaving from his home post office, a postman needs to visit the
households on each block in his route, delivering and collecting letters, and
then returning to the post office. He would like to cover this route by travel-
ling the minimum possible distance. The post office is located at the intersec-
tion point 1, (see Figure 3.1).
Figure 3.1: The village streets ([28] p.324)
The problem is from [5] Chap 13.5, and from [28] Chap 15.4.
Modeling Steps
The problem can be formulated as a Min Cost Flow Problem, see mcf-a1 for
the general Min Cost Flow problem. That is, we “simulate” the traversing of
an arc as a “flow”. Flow=1 means that the arc is traversed once, flow=2 means
that the arc is traversed two times, etc. The variables are the number of flows
on each arcs (xi,j i, j ∈ V V is the set of route intersections). The flow on
each arc must be 1 (at least on traversal li,j ≤ xi,j . We do not want to limit
the number of flows on each arc, hence we can set ui,j = ∞, however on can
show that 2 is enough (ui,j = 2). So there is: li,j ≤ xi,j ≤ ui,j .
Furthermore, the number of entering a node (intersection) must be exactly the

number of leaving the node. Hence we have:
1
http://lpl.virtual-optima.com/lpl/Solver.jsp?name=/mcf-a
X X
xi,j − xj,i = 0 ∀ i∈V
j j
Hence, we have a Min Cost Flow Problem with the following specifications:
(1) All lower bounds are 1 (li,j = 1, to make sure that all streets are traversed
at least once. All upper bound must be at least 2 (ui,j ≥ 2). And for complet-
ing the formulation (of a Min Cost Flow), bi = 0 for all nodes.

model guer13_5 "Gritting roads";
set i,j := [1..12] "Intersection points";
parameter
c{i,j} "Steet distances" :=
/1 2 150, 1 5 165, 2 3 130, 2 5 230, 2 6 160,
3 2 140, 3 4 100, 4 3 100, 4 8 190, 5 1 165,
5 6 144, 6 2 170, 6 5 144, 6 7 128, 6 9 218,
6 10 174,7 3 200, 7 6 122, 7 8 109, 7 11 185,
8 4 180, 8 11 141,8 12 190,9 5 194, 9 10 148,
10 6 174,10 7 233,11 7 185,11 10 135,12 11 110/;
l{i,j}:=1;
u{i,j}:=999;
variable x{i,j|c} [l..u] "Nr of passing the street";
constraint A{i}: sum{j} x[i,j] = sum{j} x[j,i];
minimize Distance: sum{i,j} c*x;
Writep(Distance);
-- draw the graph --
parameter X{i}:=2*(i-1)%4; Y{i}:=2*Ceil(i/4);
Draw.Scale(60,60);
--{c[i,j]} Draw.CArrow('('&l&','&x&',∞)\n'&'c='&c
,
-- X[i],Y[i],X[j],Y[j],.4,.2,0);
{c[i,j]} Draw.CArrow(c&'',X[i],Y[i],X[j],Y[j],.4,.2,0);
{c[i,j]|x=2} Draw.CArrow(X[i],Y[i],X[j],Y[j],.2,.2,0);
{i} Draw.Circle(i&'',X,Y,.2,1,0);
end

Solution: The optimal traveled distance is 5990. There are 7 arcs that must
be traversed two times: the first time serving the street and the second time
without serving. The solution is shown in 3.1
Figure 3.2: The minimal traversing

3.60. Location of income tax offices (guer13-6)

15.5.

model guer13_6 "Location of income tax offices";
parameter
p := 3 "Number of offices";
D{n,m} "Distance)" :=
-- 1 2 3 4 5 6 7 8 9 10 11 12
-------------------------------------------------
[ 0, 15, 37, 55, 24, 60, 18, 33, 48, 40, 58, 67
15, 0, 22, 40, 38, 52, 33, 48, 42, 55, 61, 61
37, 22, 0, 18, 16, 30, 41, 28, 20, 58, 39, 39
55, 40, 18, 0, 34, 12, 59, 46, 24, 62, 43, 34
24, 38, 16, 34, 0, 36, 25, 12, 24, 47, 37, 43
60, 52, 30, 12, 36, 0, 57, 42, 12, 50, 31, 22
18, 33, 41, 59, 25, 57, 0, 15, 45, 22, 40, 61
33, 48, 28, 46, 12, 42, 15, 0, 30, 37, 25, 46
48, 42, 20, 24, 24, 12, 45, 30, 0, 38, 19, 19
40, 55, 58, 62, 47, 50, 22, 37, 38, 0, 19, 40
58, 61, 39, 43, 37, 31, 40, 25, 19, 19, 0, 21
67, 61, 39, 34, 43, 22, 61, 46, 19, 40, 21, 0];
H{n}:=[15,10,12,18,5,24,11,16,13,22,19,20] "Popul.";
binary variable Y{m};
variable X{n,m};
constraint
Affect{n}: sum{m} X = 1;
NbBureaux: sum{m} Y = p;
Ajust{n,m}: X <= Y;
minimize DistPondTot: sum{n,m} H*D*X;
Writep(DistPondTot);
end

3.61. Efficiency of hospitals (guer13-7)

15.6.

model guer13_7b "Efficiency of hospitals";
set
n := [1..4] "Hospitals";
K := [1..4] "Services indicators";
L := [1..3] "Resource indicators";
parameter
h := 2 "Performance measuring on 2 hospitals";
s{K,n} "service indication" :=
[30.12, 18.54, 20.88, 10.42
13.54, 14.45, 8.52, 17.74
13, 7, 8, 26
79, 55, 47, 50 ];
r{L,n} "resource indication" :=
[90, 87, 51, 66
38.89, 109.48, 40.43, 48.41
34, 33, 20, 33 ];
variable
E "Efficiency index";
c{n} "Coefficients for method DEA";
sf{K} "Fictive service indicateurs";
rf{L} "Fictive resource indicateurs";
constraint
DEACoefs: sum{n} c = 1;
ServFict{K} : sf = sum{n} s*c;
RessFict{L} : rf = sum{n} r*c;
ServSup{K} : sf >= s[K,h];
RessInf{L}: rf <= r[L,h]*E;
minimize Efficiency: E;
Writep(Efficiency);
end

3.62. MasterMind (guer14-2)

Problem: The problem is from [5] chap 14.2.

model guer14_2 "MasterMind";
set
c := [Blanc,Bleu,Jaune,Noir,Rouge,Vert];
p := [1..4] "Nombre de positions";
binary variable X{p,c} "=1 si la couleur j est a la
position i";
constraint
-- Une couleur doit etre affectee a chaque position
UneCoulParPos{p} : sum{c} X = 1;
--Essai 1: 1 couleur bien placee, 1 couleur mal placee
E1_bp: X[1,'Rouge']+X[2,'Jaune']+X[3,'Blanc']+X[4,'Vert
'] = 1;
E1_mp: sum{p|p<>1} X[p,'Rouge']+sum{p|p<>2} X[p,'Jaune
'] +
sum{p|p<>3} X[p,'Blanc']+sum{p|p<>4} X[p,'Vert'] = 1;
--Essai 2: 1 couleur bien placee
E2_bp: X[1,'Bleu']+X[2,'Vert']+X[3,'Blanc']+X[4,'Bleu']
= 1;
E2_mp: sum{p|p<>1 and p<>4} X[p,'Bleu']+sum{p|p<>2} X[p
,'Vert'] +
sum{p|p<>3} X[p,'Blanc'] = 0;
--Essai 3: 1 couleur bien placee, 1 couleur mal placee
E3_bp: X[1,'Jaune']+X[2,'Noir']+X[3,'Blanc']+X[4,'Noir
'] = 1;
E3_mp: sum{p|p<>1} X[p,'Jaune']+sum{p|p<>2 and p<>4} X[
p,'Noir'] +
sum{p|p<>3} X[p,'Blanc'] = 1;
--Essai 4: 4 couleurs mal placees
E4_bp: X[1,'Bleu']+X[2,'Jaune']+X[3,'Rouge']+X[4,'Noir
'] = 0;
E4_mp: sum{p|p<>1} X[p,'Bleu'] + sum{p|p<>2} X[p,'
Jaune'] +
sum{p|p<>3} X[p,'Rouge']+sum{p|p<>4} X[p,'Noir']=4;
solve;
Writep(X);
end

3.63. Probleme de logique sur le marathon de Paris

(guer14-3)
Problem: The model is from [5], Chapter 14.3. and [28].

model guer14_3 "Probleme de logique sur le marathon de
Paris";
set
n := [1..6] "places";
c := [Dominique,Ignace,Naren,Olivier,Philippe,Pascal] "
Coureurs";
variable X{c,n};
constraint
UnePlaceParCoureur{c}: sum{n} X = 1;
UnCoureurParPlace{n}: sum{c} X = 1;
-- Information a)
a: X['Olivier',6] = 0;
-- Information b)
b: sum{n|n>4} (X['Dominique',n] + X['Ignace',n] + X['
Pascal',n]) = 0;
-- Information c)
c1: X['Dominique',1] + X['Dominique',2] = 1;
-- Information d)
d: sum{n|n<=4} X['Philippe',n] = 1;
-- Information e)
e: X['Ignace',2] + X['Ignace',3] = 0;
-- Information f)
f1: sum{n|n<=3} X['Naren',n] = 0;
f2: sum{n|n>=4} X['Pascal',n] = 0;
-- Information g)
g: X['Ignace',4] + X['Dominique',4] = 0;
maximize any: X[1];
Writep(X);
end

3.64. Chercheurs dans un Congres (guer14-4)

Problem: Probleme de logique consistant a retrouver la nationalite et la date
d’expose de 5 chercheurs participant a un congres. The model is from [5],
Chapter 14.4.

model guer14_4 "Chercheurs dans un Congres";
set
Chercheurs,c:=[Arabinda,Eric,Hitoshi,Michael,Zhicheng];
n := [Americain,Chinois,Francais,Indien,Japonais];
Dates,d := ['7','8','9','10','11'];
variable X{c,n,d};
constraint
Chercheur {c}: sum{n,d} X = 1;
Nationalite{n}: sum{c,d} X = 1;
Dat{d}: sum{c,n} X = 1;
-- Information a)
a: sum{d} X['Michael','Japonais',d] = 0;
-- Information b)
b: sum{d|d<=3} X['Eric','Francais',d] = 1;
-- Information c)
c1: sum{n} X['Arabinda',n,'9'] = 1;
-- Information d)
d_1: sum{d} (X['Hitoshi','Chinois',d] + X['Michael','
Chinois',d]) = 0;
d_2: sum{c} X[c,'Chinois','8'] = 1;
d_3: sum{n,d|d>2} X['Michael',n,d] = 1;
-- Information e)
e_1: sum{n} (X['Hitoshi',n,'7'] + X['Hitoshi',n,'11'])
= 0;
e_2: sum{c,d|d>=4} X[c,'Indien',d] = 0;
e_3: sum{c,d|d<=2} X[c,'Americain',c] = 0;
e_4: sum{d} (X['Hitoshi','Americain',d] + X['Hitoshi','
Indien',d]) = 0;
maximize any: X[1];
Writep(X);
end

3.65. Grille et jetons (guer14-5)

Problem: The model is from [5], Chapter 14.5.

model guer14_5 "Grille et jetons";
set n,m := [1..4];
parameter P := 10;
binary variable x{n,m} "(n,m) est occupee par un jeton";
integer NL{n} [0..2] "nombre de jetons en ligne par 2";
integer NC{n} [0..2] "nombre de jetons en colonne par 2";
constraint
Total: sum{n,m} x = P;
Lig{n}: sum{m} x = 2*NL[n];
Col1{m}: sum{n} x = 2*NC[m];
maximize any: x[1];
Writep(x);
end

3.66. Partage de tonneaux entre des neveux

(guer14-6)
Problem: The model is from [5], Chapter 14.6.

model guer14_6 "Partage de tonneaux entre des neveux";
set
n := [1..5] "Nombre de neveux";
T := [Plein '3/4' Moite '1/4' Vide]
"Nombre de type de tonneaux differents (en contenu)";
parameter
r{T} := [1.00, 0.75, 0.50, 0.25, 0.00]
"remplissage des tonneaux, plein, 3/4, 1/2, 1/4 et
vide";
c{T} := [10000, 1000, 100, 10, 1]
"Coefficients pour la contrainte tous differents";
variable p{n} "Valeur du produit scalaire c(_).x(i,_)";
integer variable x{n,T} [1..5]
"Nombre de tonneaux de type T donnes au neveu n";
constraint
Neveux{n} : sum{T} x = 9;
Tonneaux{T} : sum{n} x = 9;
Equitable{n}: sum{T} r*x = 4.5;
TousDiff{n} : sum{T} c*x = p;
Ordre{n|n>1}: p - p[n-1] >= 1;
minimize any: x[1];
Writep(p,x);
end

3.67. Le probleme des n reines (guer14-7)

Problem: The model is from [5], Chapter 14.7..

model guer14_7 "Le probleme des n reines";
set n,m := [1..100];
binary variable c{n,m} "une case de l'echiquier vaut 1
si une reine est sur cette case, 0 sinon.";
constraint
Ligne{n} : sum{m} c <= 1;
Colonne{m} : sum{n} c <= 1;
DiagDL{m} : sum{n|n>=m} c[n-m+1,n] <= 1;
DiagDC{n|n>1}: sum{m|m>=n} c[m,m-n+1] <= 1;
DiagGL{m} : sum{n|n<=m} c[n,m-n+1] <= 1;
DiagGC{n|n>1}: sum{m|m>=n} c[m,#n-m+n] <= 1;
maximize NbReines: sum{n,m} c "le nombre de reines";
Writep(NbReines,c);
end

C HAPTER 4
VARIOUS T EXTBOOKS
4.1. Forestry Production Planning (forestry)

Problem: Wood-D Industries has just signed contracts for the clear felling of
three large forest tracts of second-growth radiata pines. The harvested trees
will supply the firm’s sawmill and chipboard plant. Some of the logging out-
put is also available for export. All trees harvested are cut on location into
sections 20 feet long, referred to as first cuts, second cuts, etc. On the basis of
a detailed survey of each forest, the firm’s chief forester estimated the average
composition of each forest’s total output as shown below:
Log Cuts Forest A Forest B Forest C
---------------------------------------------------------
First and Second 42% 46% 45%
Third and Fourth 40% 41% 45%
Fifth and over 18% 13% 10%
---------------------------------------------------------
The average daily output is 128 HC for forest A, 192 HC for forest B, and 200
HC for forest C (1 HC = 100 cubic feet = 2.832 cubic meter). The log cuts
are sorted onto logging trucks for transportation to either the sawmill or the
chipboard plant, the two facilities being at different locations. Transportation
costs per HC are given in the following table:
Transportation Sawmill Chipboard
-------------------------------------------
Forest A $4 $7
Forest B $3 $5
Forest C $6 $3
-------------------------------------------
Handling costs per HC is given in the table below:

Handling Sawmill Chipboard
-------------------------------------------
First and Second $2.5 $1.2
Third and Fourth $3.5 $1.5
183
CHAPTER 4. VARIOUS TEXTBOOKS 184
Fifth and over $5.0 $2.0

-------------------------------------------
At the sawmill, logs are sawn into three grades of finished products: clear
grade, dressing grade, and construction grade. A substantial fraction of the
incoming volume of wood ends up as scrap and sawdust. The table below
shows the average log conversion factors at the sawmill, as well as the average
processing rates. Excluding breakdowns, the productive capacity at the mill
averages 360 minutes per day.
Sawn Timber Scraps Sawdust Processing
LogCuts MBF/HC HC/HC HC/HC Time per HC
----------------------------------------------------
1-2 0.72 0.26 0.14 1.8
3-4 0.66 0.30 0.15 2.6
5- 0.54 0.39 0.16 3.9
----------------------------------------------------
From sample logs processed at the sawmill, average yields for each grade of
sawn timber were determined. They are summarized in the following table.
Clear Dressing Construction
----------------------------------------
LogCuts A B C A B C A B C
/Forest
--------------------------------------------------
1-2 35% 28% 20% 48% 42% 30% 17% 30% 50%
3-4 10% 3% 25% 18% 9% 30% 72% 88% 45%
5- 0 0 5% 5% 0 30% 95% 100% 65%
--------------------------------------------------
The ex-mill wholesale price per MBF is $150 for clear grades, $110 for dressing
grades and $80 for construction grades. Scraps at the sawmill are transferred
by truck to the chipboard plant for chipping. The transportation cost is $4
per HC. Sawdust is used as fuel in the mill’s drying kiln and saves $12 in
other fuel costs per HC. At the chipboard plants, logs and scraps are chipped.
The chips are then mixed with additive glues, filled into 4’ by 8’ forms and
then compressed into boards of various thicknesses. The whole process is
highly automated. Each HC of wood yields 0.76M3/4 of chipboard (1M3/4
= 1000 square feet of 3.4 thickness = 1.77 cubic meters). The plant can pro-
duce upto 112M3/4 of chipboard per day. Chipboard prices ex-factory are
$105 per M3/4. In light of predicted demand and desired stock levels, cer-
tain minimum daily output rates of finished product are set up by Wood-D
management for given planning period. These are 31 MBF of clear grades, 35
MBF of dressing grade, 48 MBF of construction grade, and 96 M3/4 of chip-
board. Export prices valid during the same planning period are $95 per HC
for first and second cuts, and $88 per HC for third and fourth cuts. Fifth or
185 CHAPTER 4. VARIOUS TEXTBOOKS
higher cuts are not exported. The problem is to determine the optimal daily
operating policy during the planning period in question.
Project Formulate the production planning problem as a linear program using
MPL. Summarize the optimal daily operating policy for the planning prob-
lem.
Reference: Prof. Ariela Sofer, Operations Research Dept., George Mason Uni-
versity Based on a problem from the book ’Introduction to Operations Re-
search Techniques’ by Hans G. Daellenbach, John A. George, and Donald C.
McNickle, 1983
4.2. Wyndor Glass Company (hill03-1-1)

Problem:
A complete problem description and how to model it mathematically is clearly

explained in [12], Example 3.1-1, Page 25. A mathematical model formulation
is:

model hill03_1_1 "Wyndor Glass Company";
set
p := [Door, Window] "Products";
f := [1..3] "Plants";
parameter
TimeAvail{f} := [4, 12, 18];
ProdTime{f,p} := [1, 0, 0, 2, 3, 2];
Profit{p} := [3.00, 5.00];
variable Produce{p};
constraint
TimeCapacity{f}: sum{p} ProdTime*Produce <= TimeAvail
;
maximize TotalProfit: sum{p} Profit*Produce;
Writep(TotalProfit,Produce);
end

4.3. Mary’s Radiation Therapy (hill03-4-1)

Problem:

explained in [12] , Example 3.4-1, Page 44 A mathematical model formulation
is:

model hill03_4_1 "Mary's Radiation Therapy";
set
b := [Beam1, Beam2] "Beam";
a := [CriticalTissues, TumorRegion, CenterTumor] "
BodyArea";
parameter
HealthyAnatomy{b} := [0.4, 0.5];
DosageAbsorbed{b,a} := [0.3, 0.5, 0.6, 0.1, 0.5,
0.4];
Restriction{a} := [2.7, 6.0, 6.0];
K{a} := [1 2 3];
variable D{b} "Dosage";
constraint
Dosage1{a|K=1}: sum{b} DosageAbsorbed*D <=
Restriction;
Dosage2{a|K=2}: sum{b} DosageAbsorbed*D =
Restriction;
Dosage3{a|K=3}: sum{b} DosageAbsorbed*D >=
Restriction;
minimize TotalHealthyDosage: sum{b} HealthyAnatomy*D;
Writep(TotalHealthyDosage,D);
end

4.4. Kibbutzim Crop Allocation (hill03-4-2)

Problem:

is:

model hill03_4_2 "Kibbutzim Crop Allocation";
set
k := [1..3] "Kibbutz";
c := [SugarBeets, Cotton, Sorghum] "Crops";
parameter
UsableLand{k} := [400, 600, 300]; -- {Acres}
WaterAlloc{k} := [600, 800, 375]; -- {
AcreFeet}
WaterUsed{c} := [3, 2, 1]; -- {
AcreFeet/Acre}
NetReturn{c} := [1000, 750, 250]; -- {$/Acre}
MaxQuota{c} := [600, 500, 325]; -- {Acres}
variable AllocAcre,A{c,k};
constraint
MaxLandUse{k}: sum{c} A <= UsableLand;
MaxWaterAlloc{k}: sum{c} WaterUsed*A <= WaterAlloc;
TotalAcreage{c}: sum{k} A <= MaxQuota;
EqProp{k}: sum{c} A/UsableLand = sum{c} A[c,k%#k+1]/
UsableLand[k%#k+1];
maximize TotalProfit: sum{c,k} NetReturn*A;
Writep(TotalProfit,A);
end

4.5. Nori/Leets Air Pollution (hill03-4-3)

Problem:

is:

model hill03_4_3 "Nori/Leets Air Pollution";
set
pollutant,p := [Particulates, SulfurOxides,
Hydrocarbons];
furnace,f := [Blast, OpenHearth];
abatementa,a := [TallerSmokestacks, Filters,
BetterFuels];
parameter
RequiredReduction{p} := [60, 150, 125];
ReducedEmissionRate,rer{p,a,f} :=
[12, 9, 25, 20, 17, 13,
35, 42, 18, 31, 56, 49,
37, 53, 28, 24, 29, 20];
AnnualCost{a,f} := [8, 10, 7, 6, 11, 9];
variable U{a,f} [0..1] "Used Method";
constraint
EmissionReduction{p}: sum{f,a} rer*U >=
RequiredReduction;
minimize TotalCost: sum{a,f} AnnualCost*U;
Writep(TotalCost,U);
end

4.6. Save-It Company (hill03-4-4)

Problem:

is:

model hill03_4_4 "Save-It Company";
set
g := [A, B, C] "Grade";
m := [1..4] "Material";
GradeMat{g,m} := [A 1, A 2, A 3, A 4, B 1, B 2, B 4,
C 1];
parameter
MinMixSpecs{g,m} :=
/A 1 0, A 2 0.40, A 3 0 , A 4 0.2,
B 1 0, B 2 0.1, B 4 0.1, C 1 0/;
MaxMixSpecs{g,m} :=
/A 1 0.3, A 2 0, A 3 0.5, A 4 0.2,
B 1 0.5, B 2 0, B 4 0.1, C 1 0.7/;
AmalgamationCost{g} := [3.00, 2.50, 2.00];
SellingPrice{g} := [8.50, 7.00, 5.50];
MaterialAvail{m} := [3000, 2000, 4000, 1000];
TreatmentCost{m} := [3.00, 6.00, 4.00, 5.00];
MinimumTreated := 0.5;
TreatmentCashAvail := 30000;
variable M{g,m} "MixMaterial";
constraint
MinMixtureSpecs{g,m|MinMixSpecs>0}: M >= MinMixSpecs
*(sum{m} M);
MaxMixtureSpecs{g,m|MaxMixSpecs>0}: M <= MaxMixSpecs
*(sum{m} M);
MaterialLimit{m}: sum{g} M <= MaterialAvail;
ReqAmountTreated{m}: sum{g} M >= MinimumTreated*
MaterialAvail;
TreatCostRestrict: sum{g,m} TreatmentCost*M =
TreatmentCashAvail;
maximize TotalProfit: sum{g,m} SellingPrice*M - sum{g,m
} AmalgamationCost*M;
Writep(TotalProfit,M);
end

4.7. Save-It Company (hill03-4-4b)

Problem: lternative formulation for the same model, adds summary vari-
ables.

explained in [12] , Example 3.4-4b, Page 53 A mathematical model formula-
tion is:

model hill03_4_4b "Save-It Company";
set grade,g := [A, B, C];
material,m := [1..4];
GradeMat{g,m}:=[A 1, A 2, A 3, A 4, B 1, B 2,B 4,C
1];
parameter
MinMixSpecs{g,m} :=
/A 1 0, A 2 0.40, A 3 0, A 4 0.20,
B 1 0, B 2 0.10, B 4 0.10, C 1 0/;
MaxMixSpecs{g,m} :=
/A 1 0.3, A 2 0, A 3 0.50, A 4 0.20,
B 1 0.50, B 2 0, B 4 0.10, C 1 0.70/;
AmalgamationCost{g} := [3.00, 2.50, 2.00];
SellingPrice{g} := [8.50, 7.00, 5.50];
MaterialAvail{m} := [3000, 2000, 4000, 1000];
TreatmentCost{m} := [3.00, 6.00, 4.00, 5.00];
MinimumTreated := 0.5;
TreatmentCashAvail := 30000;
variable
G{g} "ProduceGrade";
U{m} "MaterialUsed";
M{g,m} "MixMaterial";
constraint
GradeProd{g}: G = sum{m} M ;
MatUse{m} : U = sum{g} M ;
MinMixtureSpecs{g,m|MinMixSpecs>0}:M>=MinMixSpecs*G;
MaxMixtureSpecs{g,m|MaxMixSpecs>0}:M<=MaxMixSpecs*G;
MaterialLimit{m}: U <= MaterialAvail;
ReqAmountTreated{m}: U >= MinimumTreated*
MaterialAvail;
TreatCostRestrict: sum{m} TreatmentCost*U =
TreatmentCashAvail;
maximize TotalProfit: sum{g} SellingPrice*G - sum{g}
AmalgamationCost*G;
Writep(TotalProfit,G,U,M);
end

4.8. Union Airways Personnel (hill03-4-5)

Problem:

is:

model hill03_4_5 "Union Airways Personnel";
set
s := [1..5] "Shifts";
p := [AM_6_8, AM_8_10, AM_10_12,
PM_12_2, PM_2_4, PM_4_6, PM_6_8,
PM_8_10, PM_10_12, AM_12_6] "Periods";
ShiftSchedule{s,p} :=
[1 AM_6_8, 1 AM_8_10, 1 AM_10_12, 1 PM_12_2,
2 AM_8_10, 2 AM_10_12, 2 PM_12_2, 2 PM_2_4,
3 PM_12_2, 3 PM_2_4, 3 PM_4_6, 3 PM_6_8,
4 PM_4_6, 4 PM_6_8, 4 PM_8_10, 4 PM_10_12,
5 PM_10_12, 5 AM_12_6];
parameter
MinAgentsNeeded{p} := [48, 79, 65, 87, 64, 73, 82,
43, 52, 15];
DailyCostPerAgent{s} := [170, 160, 175, 180, 195];
variable A{s} "Assign agents to s";
constraint
MeetRequirements{p}: sum{s|ShiftSchedule} A >=
MinAgentsNeeded;
minimize TotalCost: sum{s} DailyCostPerAgent*A;
Writep(TotalCost,A);
end

4.9. Distribution Unlimited (hill03-4-6)

Problem:

is:

model hill03_4_6 "Distribution Unlimited";
set i,j := [F1, F2, DC, W1, W2] "Locations";
parameter
ShipCost{i,j} :=
/F1 F2 200, F1 DC 400, F1 W1 900, F2 DC 300,
DC W2 100, W1 W2 300, W2 W1 200/;
Supply{i} := /F1 50, F2 40/;
Demand{i} := /W1 30, W2 60/;
MaxFlow{i,j} := /F1 F2 10, DC W2 80/;
variable Ship{i,j|ShipCost};
constraint
NetFlow{i}: sum{j} Ship[j,i] + Supply
= sum{j} Ship[i,j] + Demand;
MaxShipment{i,j|MaxFlow}: Ship <= MaxFlow;
minimize TotalCost: sum{i,j} ShipCost*Ship;
end

4.10. Wyndor Glass (hill07-1-1)

Problem:

explained in [12] , Example 7.1-1, Page 311 A mathematical model formula-
tion is:

model hill07_1_1 "Wyndor Glass";
set
p := [Door, Window] "Products";
f := [1..3] "Plants";
parameter
TimeAvail{f} := [4, 12, 18];
ProdTime{f,p} := [1, 0, 0, 2, 3, 2];
Profit{p} := [3.00, 5.00];
variable T{f} "Assign agents to f";
constraint ProduceDual{p}: sum{f} ProdTime*T >= Profit;
minimize TProfit: sum{f} TimeAvail*T;
Writep(TProfit,T);
end

4.11. Upper Bound (hill07-3-1)

Problem:

explained in [12] „ Example 7.3-1, Page 319 A mathematical model formula-
tion is:

model hill07_3_1 "Upper Bound";
set
activity1,a := [Activity1,Activity2,Activity3];
constraint1,c := [constraint1,constraint2];
parameter
ResourceAvail{c} := [12,4];
ResourceUse{c,a} := [4,1,0, -2,0,1];
UnitProfit{a} := [2,1,2];
Upper{a} := [4,15,6];
variable Produce{a} [0..Upper];
constraint
TimeCapacity{c}: sum{a} ResourceUse*Produce <=
ResourceAvail;
maximize TotalProfit: sum{a} UnitProfit*Produce;
end

4.12. Goal Programming (hill07-5-1)

Problem:

tion is:

model hill07_5_1 "Goal Programming";
set
p := [1..3] "Products";
f := [Profit, Employment, Investment] "Factors";
d := ['Over', 'Under'] "Deviations";
parameter
Goal{f} := [125, 40, 55];
PenaltyWeight{d,f} := [0, 2, 3, 5, 4, 0];
UnitContrib{f, p} := [12, 9,15, 5, 3, 4, 5, 7,
8];
variable
ProdRate{p};
Amount{f,d|PenaltyWeight>0};
constraint
Calc{f}: sum{p} ProdRate*UnitContrib
= Goal + Amount[f,'Over'] - Amount[f,'Under'];
minimize WeightedSum: sum{d,f} PenaltyWeight*Amount;
Writep(WeightedSum,Amount);
end

4.13. P-T Company (hill08-1-1)

Problem:

tion is:

model hill08_1_1 "P-T Company";
set
cannery,c := [C1, C2, C3];
warehouse,w := [W1, W2, W3, W4];
parameter
ShipCost{c,w} := [464, 513, 654, 867,
352, 416, 690, 791,
995, 682, 388, 685];
Supply{c} := [75,125,100];
Demand{w} := [80, 65, 70, 85];
variable Ship{c,w};
constraint
Output{c}: sum{w} Ship = Supply;
Allocation{w}: sum{c} Ship = Demand;
minimize TotalCost: sum{c,w} ShipCost*Ship;
end

4.14. Northern Airplane (hill08-1-2)

Problem:

explained in [12] „ Example 8.1-2, Page 359 A mathematical model formula-
tion is:

model hill08_1_2 "Northern Airplane";
set
produce,p := [1..4];
install,i := [1..5];
parameter
M := 999;
DistCost{p,i} := [1.080, 1.095, 1.110, 1.125, 0,
999, 1.110, 1.125, 1.140, 0,
999, 999, 1.100, 1.115, 0,
999, 999,999, 1.130, 0];
Supply{p} := [25, 35, 30, 10];
Demand{i} := [10, 15, 25, 20, 30];
variable Engines{p,i};
constraint
Production{p}: sum{i} Engines = Supply;
Installations{i}: sum{p} Engines = Demand;
minimize TotalCost: sum{p,i} DistCost*Engines;
Writep(TotalCost,Engines);
end

4.15. Metro Water (hill08-1-3)

Problem:

tion is:

model hill08_1_3 "Metro Water";
set
r := [Colombo, Sacron, Calorie, Dummy] "Rivers";
c := [Berdoo_min, Berdoo_extra, LosDevils, SanGo,
Hollyglass] "Cities";
parameter
M := 999;
DistCost{r, c} := [16, 16, 13, 22, 17,
14, 14, 13, 19, 15,
19, 19, 20, 23, 999,
999, 0,999, 0, 0];
Supply{r} := [50, 60, 50, 50];
Demand{c} := [30, 20, 70, 30, 60];
variable Distribute{r,c};
constraint
Source{r}: sum{c} Distribute = Supply;
Destination{c}: sum{r} Distribute = Demand;
minimize TotalCost: sum{r,c} 10000000*DistCost*
Distribute;
Writep(TotalCost,Distribute);
end

4.16. Job Shop Company (hill08-3-1)

Problem:

tion is:

model hill08_3_1 "Job Shop Company";
set
m := [m1, m2, m3, Dummy] "Machines";
p := [1..4] "Locations";
parameter
M := 999;
DistCost{m, p} := [13, 16, 12, 11,
15,999, 13, 20,
5, 7, 10, 6,
0, 0, 0, 0];
Supply{m} := [1, 1, 1, 1];
Demand{p} := [1, 1, 1, 1];
variable Assign{m,p};
constraint
Source{m}: sum{p} Assign = Supply;
Destination{p}: sum{m} Assign = Demand;
minimize TotalCost: sum{m,p} DistCost*Assign;
Writep(TotalCost,Assign);
end

4.17. Better Products Company (hill08-3-2a)

Problem:

explained in [12] , Example 8.3-2a, Page 386 A mathematical model formula-
tion is:

model hill08_3_2a "Better Products Company";
set
f := [1..3] "Plants";
p := [1..5] "Products";
parameter
M := 999;
DistCost{f, p} := [41, 27, 28, 24, 0,
40, 29,999, 23, 0,
37, 30, 27, 21, 0];
Supply{f} := [75, 75, 45];
Demand{p} := [20, 30, 30, 40, 75];
variable Assign{f,p};
constraint
Source{f}: sum{p} Assign = Supply;
Destination{p}: sum{f} Assign = Demand;
minimize TotalCost: sum{f, p} DistCost*Assign;
end

4.18. Better Products Company (hill08-3-2b)

Problem:

explained in [12] , Example 8.3-2b, Page 386 A mathematical model formula-
tion is:

model hill08_3_2b "Better Products Company";
set
f := [p1a, p1b, p2a, p2b, p3] "Plants";
p := [1..5] "Products";
parameter
M := 99999;
Cost{f, p} := [820, 810, 840, 960, 0,
820, 810, 840, 960, 0,
800, 870,99999, 920, 0,
800, 870,99999, 920, 0,
740, 900, 810, 840,99999];
binary variable Assign{f, p};
constraint
Assignee{f}: sum{p} Assign = 1;
Task{p}: sum{f} Assign = 1;
minimize TotalCost: sum{f, p} Cost*Assign;
end

4.19. Shortest Path (hill09-3-1)

Problem:

tion is:

model hill09_3_1 "Shortest Path";
set i,j := [O, A, B, C, D, E, T];
parameter
Distance{i,j} :=
/O A 2, O B 5, O C 4, A B 2, A D 7, B C 1,
B D 4, B E 3, C E 4, D E 1, D T 5, E T 7/;
variable Path{i,j|Distance>0};
constraint
FlowBalance{i}: if(i='O',1) + sum{j} Path[j,i]
= if(i='T',1) + sum{j} Path[i,j];
minimize TotalDistance: sum{i,j} Distance*Path;
Writep(TotalDistance,Path);
end

4.20. Maximum Flow (hill09-5-1)

Problem:

tion is:

model hill09_5_1 "Maximum Flow";
set
node,i,j := [O, A, B, C, D, E, T];
SourceNode{i} := [O];
DestNode{i} := [T];
parameter
Capacity{i,j} :=
/O A 5, O B 7, O C 4, A B 1, A D 3, B C 2,
B D 4, B E 5, C E 4, D T 9, E D 1, E T 6/;
variable
Flow{i,j|Capacity>0} [0..Capacity];
Entrance{node|node='O'};
Destination{node|node='E'};
constraint
FlowBal{node}: Entrance + sum{i} Flow[i,node]
= Destination + sum{j} Flow[node,j];
maximize TotalFlow: Entrance['O'];
Writep(TotalFlow,Flow);
end

4.21. Minimum Cost (hill09-6-1)

Problem:

tion is:

model hill09_6_1 "Minimum Cost";
set
node,i,j := [A, B, C, D, E];
parameter
UnitCost{i,j} :=
/A B 2, A C 4, A D 9, B C 3, C E 1, D E 3, E D
2/;
Capacity{i,j} := /A B 10, C E 80/;
SupplyDemand{node} := [50, 40, 0, -30, -60];
variable Ship{i,j|UnitCost>0};
constraint
FlowBalance{node}: sum{i} Ship[i,node] + SupplyDemand
= sum{j} Ship[node,j];
MaxCapacity{i,j|Capacity>0}: Ship <= Capacity;
minimize TotalCost: sum{i,j} UnitCost*Ship;
end

4.22. Critical Path (hill10-1-1)

Problem: RT/PERT Method

tion is:

model hill10_1_1 "Critical Path";
set
n,m := [A,B,C,D,E,F,G,H,I,J,K,L,M,N,Finish];
Path{n,m} :=
[ A B, B C, C D, C E, C I, D G, E F, E H, F J,
G H, H M, I J, J K, J L, K N, L N, M Finish, N
Finish];
parameter Duration{n} :=
[2,4,10,6,4,5,7,9,7,8,4,5,2,6,0];
variable S{n} "StartTime";
constraint SeqRelation{Path[n,m]}: S[n] + Duration[n]
<= S[m];
minimize FinishTime: S['Finish'];
Writep(FinishTime,S);
end

4.23. Crashing (hill10-5-1)

Problem: ritical Path

explained in [12] , Example 10.5-1, Hillier and Lieberman, Page 494 A mathe-
matical model formulation is:

model hill10_5_1 "Crashing";
set
n,m := [A,B,C,D,E,F,G,H,I,J,K,L,M,N];
Path{n,m} :=
[A B, B C, C D, C E, C I, D G, E F, E H, F J,
G H, H M, I J, J K, J L, K N, L N];
parameter
NormalTime{n} := [2,4,10,6,4,5,7,9,7,8,4,5,2,6];
CrashTime{n} := [1,2,7,4,3,3,4,6,5,6,3,3,1,3];
NormalCost{n} := [180, 320, 620, 260, 410, 180, 900,
200, 210, 430, 160, 250, 100, 330];
CrashCost{n} := [280, 420, 860, 340, 570, 260, 1020,
380, 270, 490, 200, 350, 200, 510];
MaxReduce{n} := NormalTime - CrashTime;
CostWeekSaved{n} := CrashCost/MaxReduce - NormalCost/
MaxReduce;
variable
S{n} "StartTime";
F [0..40] "Finished";
R{n} [0..MaxReduce] "ReducedTime";
constraint
SeqRelation{Path[n,m]}: S[n] + NormalTime[n] - R[n]
<= S[m];
M: S['M'] + NormalTime['M'] - R['M'] <= F;
N: S['N'] + NormalTime['N'] - R['N'] <= F;
minimize TotalCost: sum{n} NormalCost + sum{n}
CostWeekSaved*R;
Writep(TotalCost,S,F,R);
end

4.24. Calif. Manufacturing (hill12-1-1)

Problem: acility Location

tion is:

model hill12_1_1 "Calif. Manufacturing";
set
location,lo:= [WLA, FLA, WSF, FSF];
parameter
NPV{lo} := [6, 9, 4, 5]; -- In millions
Capital{lo} := [5, 6, 2, 3]; -- In millions
binary variable x{lo};
constraint
Budget: sum{lo} Capital*x <= 10;
Warehouse: sum{lo|lo<#lo} x <= 1;
Cnt{lo|lo>1}: -x + x[lo+2] <= 0;
maximize TotalNPV: sum{lo} NPV*x;
Writep(TotalNPV,x);
end

4.25. Production Rates (hill12-4-1)

Problem:

explained in [12] , Example 12.4-1, , Page 592 A mathematical model formu-
lation is:

model hill12_4_1 "Production Rates";
set
plant ,f := [1..2];
product,p := [1..3];
parameter
Hours{f,p} := [3, 4, 2, 4, 6, 2];
Profit{p} := [5, 7, 3]; -- Per unit
Sales{p} := [7, 5, 9];
M := 999;
variable rate{p} [0..Sales];
binary variable
y{p}; -- Equals 1 if product i is produced
P; -- Equals 1 if we produce in Plant 2
constraint
Produce{p}: rate <= M*y;
NumPrd: sum{p} y <=2;
Plnt{f|f=1}: sum{p} Hours*rate <= if(f=1,30+M*P, 40+M
*(1-P));
maximize TotProfit: sum{p} Profit*rate ;
Writep(TotProfit,y,P);
end

4.26. TV Spots (hill12-4-2a)

Problem:

explained in [12] , Example 12.4-2a, Page 594 A mathematical model formu-
lation is:

model hill12_4_2a "TV Spots";
set
spots ,s := [0..3];
parameter
Profit{s,p}:= [0 0 0, 1 0 -1, 3 2 2, 3 3 4];
J{s} := [1, 2, 3, 0];
binary variable
y{s,p|s>0}; // = 1 if product i fills j spots
constraint
NumProd{p}: sum{s} y <= 1;
NumSpots: sum{p,s} J*y = 5;
maximize TotProfit: sum{p,s} Profit*y;
Writep(TotProfit,y);
end

4.27. TV Spots (hill12-4-2b)

Problem:

explained in [12] , Example 12.4-2b, Hillier and Lieberman, Page 595 A math-
ematical model formulation is:

model hill12_4_2b "TV Spots";
set
spots ,s := [0..3];
parameter
Slope{p,s} := [0 1 2 0, 0 0 2 1, 0 -1 3 2];
binary variable y{p,s};
constraint
Cnst {p,s|s>1}: y - y[p,s-1] <=0;
Tot: sum{p,s|s>0} y = 5;
maximize TotProfit: sum{p,s} Slope*y;
Writep(TotProfit,y);
end

4.28. Crew Assignments (hill12-4-3)

Problem:

tion is:

model hill12_4_3 "Crew Assignments";
set
j := [1..12];
flights := [1..11];
parameter
Cost{j} := [2, 3, 4, 6, 7, 5, 7, 8, 9, 9, 8, 9];
Feas{flights,j} :=
[1, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0,
0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0,
0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 1,
0, 0, 0, 1, 0, 0, 1, 0, 1, 1, 0, 1,
1, 0, 0, 0, 0, 1, 0, 0, 0, 1, 1, 0,
0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0, 0,
0, 0, 0, 0, 0, 0, 1, 1, 0, 1, 1, 1,
0, 1, 0, 1, 1, 0, 0, 0, 1, 0, 0, 0,
0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0,
0, 0, 1, 0, 0, 0, 1, 1, 0, 0, 0, 1,
0, 0, 0, 0, 0, 1, 0, 0, 1, 1, 1, 1];
binary variable x{j};
constraint
Sched{flights}: sum{j} Feas*x >= 1;
Numb: sum{j} x = 3;
minimize TotCost: sum{j} Cost*x ;
Writep(TotCost,x);
end

4.29. Nonlinear constraint (hill13-2-1)

Problem:

tion is:

model hill13_2_1 "Nonlinear constraint";
set
product,p := [Door, Window];
parameter
TimeAvail := 4;
ProdTime{p} := [1, 0];
Profit{p} := [3, 5];
variable P{p};
constraint
TimeCapacity: sum{p} ProdTime*P <= TimeAvail;
NLConstr: 9*P['Door']^2 + 5*P['Window']^2 <= 216;
end

4.30. Nonlinear Objective (hill13-2-2)

Problem:

tion is:

model hill13_2_2 "Nonlinear Objective";
set
product,p := [Door, Window];
plant ,f := [1..3];
parameter
TimeAvail{f} := [4, 12, 18];
ProdTime{f,p} := [1, 0, 0, 2, 3, 2];
variable P{p} "Produce";
constraint
TimeCapacity{f}: sum{p} ProdTime*P <= TimeAvail;
maximize TotalProfit: 126*P['Door'] - 9*P['Door']^2
+182*P['Window'] - 13*P['Window']^2;
end

4.31. Nonlinear Objective 2 (hill13-2-3)

Problem:

tion is:

model hill13_2_3 "Nonlinear Objective 2";
set
product := [Door, Window];
plant := [1..3];
parameter
TimeAvail{plant} := [4, 12, 18];
ProdTime{plant, product} := [1, 0, 0, 2, 3, 2];
variable Produce{product};
constraint
TimeCapacity{plant}: sum{product} ProdTime*Produce <=
TimeAvail;
maximize TotalProfit: 54*Produce['Door'] - 9*Produce['
Door']^2
+ 78*Produce['Window'] - 13*Produce['
Window']^2;
end

4.32. Convex Programming (hill13-9-1)

Problem:

explained in [12] :, Example 13.9-1, Page 698 A mathematical model formula-
tion is:

model hill13_9_1 "Convex Programming";
set i := [1..2];
parameter c{i} := [3, 2];
variable x{i};
constraint Constr: sum{i} c*x <= 6;
maximize Z: 5*x[1] - x[1]^2 + 8*x[2] - 2*x[2]^2;
Writep(Z,x);
end

4.33. Odds And Even (hill14-1-1)

Problem:

tion is:

model hill14_1_1 "Odds And Even";
set strategy,s1,s2:= [1..2];
parameter
Payoff{s1, s2} := [ 1, -1, -1, 1];
variable
P{s1} "Probabilities";
ValueOfGame;
constraint
TotalProbabilities: sum{s1} P = 1;
ExpectedPayoff{s2}: sum{s1} Payoff*P >= ValueOfGame;
maximize GameValue: ValueOfGame;
Writep(GameValue,P,ValueOfGame);
end

4.34. Political Variaton 1 (hill14-2-1)

Problem: ame Theory

tion is:

model hill14_2_1 "Political Variaton 1";
set strategy,s1,s2 := [1..3];
parameter Payoff{s1,s2} := [1,1,4,1,0,5,0,1,-1];
variable
Probabilities,P{s1};
ValueOfGame;
constraint
Writep(GameValue,P);
end

4.35. Political Variation 2 (hill14-2-2)

Problem: ame theory

tion is:

model hill14_2_2 "Political Variation 2";
set strategy1,s1,s2 := [1..3];
parameter Payoff{s1,s2}:=[-3,-2,6,2,0,2,5,-2,-4];
variable
ValueOfGame;
constraint
end

4.36. Political Variation 3 (hill14-2-3)

Problem:

tion is:

model hill14_2_3 "Political Variation 3";
set strategy,s1,s2 := [1..3];
parameter Payoff{s1,s2}:=[0,-2,2,5,4,-3,2,3,-4];
variable
ValueOfGame;
constraint
end

4.37. School Board Scheduling (hillC3-1)

Problem:

explained in [12] , Chapter 3, Case 1, Page 79 A mathematical model formu-
lation is:

model hillC3_1 "School Board Scheduling";
set
g := [6,7,8] "Grade";
s := [1,2,3] "School";
a := [1..6] "Area";
parameter
NrStudents{a} := [450, 600, 550, 350, 500, 450];
PctGrade{a,g} := [0.32 0.38 0.30, 0.37 0.28 0.35,
0.30 0.32 0.38, 0.28 0.40 0.32,
0.39 0.34 0.27, 0.34 0.28 0.38];
StudentGrade{a,g} := PctGrade*NrStudents;
MinStudGradePct := 0.30;
MaxStudGradePct := 0.35;
BusingCost{a,s} := [300 0 700, -1 400 500,
600 300 200, 200 500 -1,
0 -1 400, 500 300 0];
SchoolCapacity{s} := [900 1100 1000];
variable
AssignStudent,A{a,s,g|BusingCost>0};
constraint
StudentDistr{a,g}: sum{s} A = PctGrade*NrStudents;
MaxSchool{s}: sum{a,g} A <= SchoolCapacity;
MinGrade{s,g}: sum{a} A >= MinStudGradePct*(sum{a,g}
A);
MaxGrade{s,g}: sum{a} A <= MaxStudGradePct*(sum{a,g}
A);
minimize TotalCost: sum{a,s,g} BusingCost*A;
Writep(TotalCost,A);
end

4.38. Manufacturing (hillP3-1-3)

Problem:

explained in [12] , Chapter 3, Case 1, Page 79 A mathematical model formu-
lation is:

model hillP3_1_3 "Manufacturing";
set
p := [ 1..3] "Products";
m := [Milling, Lathe, Grinder] "Machines";
parameter
TimeAvail{m} := [500, 350, 150]; --hours
Productivity{m,p} := [9, 3, 5, 5, 4, 0, 3, 0, 2];
--hours
Profit{p} := [50, 20, 25]; --dollars
MaxSales := 20;
constraint
ProdCap{m}: sum{p} Productivity*Produce <= TimeAvail;
SalesCap: Produce[3] <= MaxSales;
maximize TotalProfit: sum{p} Profit*Produce ;
end

4.39. TV Manufacturing (hillP3-1-4)

Problem:

explained in [12] , Chapter 3.1, Problem 4, Page 69 A mathematical model for-
mulation is:

model hillP3_1_4 "TV Manufacturing";
set tvset := [i27, i20];
parameter
MaxDemand{tvset} := [40, 10];
WorkHoursAvail := 500;
WorkReq{tvset} := [20, 10];
Profit{tvset} := [120, 80];
variable Produce{tvset} [0..MaxDemand];
constraint
WorkLimit: sum{tvset} WorkReq*Produce <=
WorkHoursAvail;
maximize TotalProfit: sum{tvset} Profit*Produce ;
end

4.40. Resource PQ (hillP3-1-5)

Problem:

mulation is:

model hillP3_1_5 "Resource PQ";
set
product := [1..2];
resource := [P, Q];
parameter
ResourceReq{product,resource} := [1, 2, 3, 2];
Profit{product} := [1, 2];
ResourceAvail{resource} := [200, 300];
MaxProduce2 := 60;
constraint
ResourceLimit{resource}: sum{product} ResourceReq*
Produce <= ResourceAvail;
ProduceLimit: Produce[2] <= MaxProduce2;
maximize TotalProfit: sum{product} Profit*Produce ;
end

4.41. Insurance (hillP3-1-6)

Problem:

mulation is:

model hillP3_1_6 "Insurance";
set
p := [SpecialRisk, Mortage] "product";
d := [Underwriting, Administration, Claims] "
department";
parameter
Profit{p} := [5, 2];
HoursAvail{d} := [2400, 800, 1200];
HoursUnit{p, d} := [3, 0, 2, 2, 1, 0];
constraint WorkReq{d}: sum{p} HoursUnit*Produce <=
HoursAvail;
end

4.42. Auto Spare Parts (hillP3-1-7)

Problem:

mulation is:

model hillP3_1_7 "Auto Spare Parts";
set
m := [1, 2] "Machines";
p := [A, B, C] "Parts";
parameter
ProcessTime{m,p} := [0.02, 0.03, 0.05, 0.05, 0.02,
0.04];
TimeAvail{m} := [40, 40];
Profit{p} := [50, 40, 30];
constraint Capacity{m}: sum{p} ProcessTime*Produce <=
TimeAvail;
end

4.43. Resource QRS (hillP3-2-1)

Problem:

mulation is:

model hillP3_2_1 "Resource QRS";
set
p := [A, B] "Products";
r := [Q, R, S] "Resources";
parameter
Profit{p} := [3, 2];
ResAvail{r} := [2, 2, 4];
ResUsed{p,r}:= [2, 1, 3, 1, 2, 3];
constraint Capacity{r}: sum{p} ResUsed*Produce <=
ResAvail;
end

4.44. Invest Venture (hillP3-2-3)

Problem:

mulation is:

model hillP3_2_3 "Invest Venture";
set v := [1..2] "Ventures";
parameter
CashOnHand := 6000;
InvestCash{v} := [5000, 4000];
WorkReq{v} := [400, 500];
Profit{v} := [4500, 4500];
WorkHoursAvail:= 600;
variable InvestShare{v} [0..1];
constraint
CashLimit: sum{v} InvestCash*InvestShare <=
CashOnHand;
WorkLimit: sum{v} WorkReq*InvestShare <=
WorkHoursAvail;
maximize TotalProfit: sum{v} Profit*InvestShare ;
Writep(TotalProfit,InvestShare);
end

4.45. Investment (hillP3-4-10)

Problem:

explained in [12] , Chapter 3.4, Problem 10, Page 76 A mathematical model
formulation is:

model hillP3_4_10 "Investment";
set
a := [1, 2, 3] "assets";
y := [5, 10, 20] "years";
parameter
AssetCost := 100;
AssetIncome{a,y}:= [200, 50, 0, 100, 50, 150,
50, 100, 200];
MinCashFlowReq{y} := [4000, 0, 3000];
variable Invest{a};
constraint Requirement{y}: sum{a} AssetIncome*Invest >=
MinCashFlowReq;
minimize TotalAmount: sum{a} AssetCost*Invest ;
Writep(TotalAmount,Invest);
end

4.46. Investor ABCD (hillP3-4-11)

Problem:

formulation is:

model hillP3_4_11 "Investor ABCD";
set
i := [A, B, C, D] "investments";
y :=[ 1..5] "years";
InvestAvail{i,y} :=
[A 1, A 2, A 3, A 4, B 1, B 2, B 3, C 2, D 5];
parameter
TotalInvest := 60000;
InvestReturn{i} := [1.40, 1.70, 1.90, 1.30];
variable
MakeInvest,I{i,y|InvestAvail};
MoneyAvailEnd,M{y};
constraint
MoneyBalance{y}: M = if(y=1,TotalInvest,M[y-1])
+ sum{i} InvestReturn*I[(#y-2+y)%#y+1] - sum{i} I;
MoneyLimit{y}: sum{i} I <= if(y=1,TotalInvest,M[y-1])
;
maximize MoneyAccumulated: MoneyAvailEnd[#y];
Writep(MoneyAccumulated,I,M);
end

4.47. Alloy Blending (hillP3-4-12)

Problem:

formulation is:

model hillP3_4_12 "Alloy Blending";
set
a := [1..5] "alloy";
e := [Tin, Zinc, Lead] "elements";
parameter
NewAlloyReq{e} := [40, 35, 25];
ElementPct{e,a}:= [60, 25, 45, 20, 50,
10, 15, 45, 50, 40,
30, 60, 10, 30, 10];
AlloyCost{a} := [22, 20, 25, 24, 27];
variable Blend{a};
constraint BlendReq{e}: sum{a} ElementPct*Blend >=
NewAlloyReq;
minimize TotalCost: sum{a} AlloyCost*Blend ;
Writep(TotalCost,Blend);
end

4.48. Computer Fac Oper Assignment (hillP3-4-13)

Problem:

formulation is:

model hillP3_4_13 "Computer Fac Oper Assignment";
set
w := [Mon, Tues, Wed, Thurs, Fri] "Workdays";
o := [KC, DH, HB, SC, KS, NK] "Operators";
Male{o} := [KC, DH, HB, SC] "Males";
Female{o} := [KS, NK] "Females";
parameter
WageRate{o} := [10.00, 10.10, 9.90, 9.80, 10.80,
11.30];
HoursAvail{o,w} := [6, 0, 6, 0, 6,
0, 6, 0, 6, 0,
4, 8, 4, 0, 4,
5, 5, 5, 0, 5,
3, 0, 3, 8, 0,
0, 0, 0, 6, 2];
variable AssignHours{o,w|HoursAvail>0} [0..HoursAvail];
constraint
AssignDay{w}: sum{o} AssignHours >= 14;
MinHours{Male}: sum{w} AssignHours >= if(Male,8,7);
minimize TotalCost: sum{o,w} WageRate*AssignHours;
Writep(TotalCost,AssignHours);
end

4.49. Warehouse Storage (hillP3-4-14)

Problem:

formulation is:

model hillP3_4_14 "Warehouse Storage";
set
m,m1 := [ 1..5] "Months";
p := [ 1..5] "Periods";
parameter
SpaceReq{m} :=[30 20 40 10 50]; -- 1000sf
LeaseCost{p}:=[650 1000 1350 1600 1900]; -- $00/1000
sf
variable SpaceLeased{m,p|p<=6-m};
constraint MinSpace{m1}: sum{m,p|m<=m1 and p>m1-m}
SpaceLeased >= SpaceReq;
minimize TotalCost: sum{m,p} LeaseCost*SpaceLeased;
Writep(TotalCost,SpaceLeased);
end

4.50. Paper Manufacturing (hillP3-4-15)

Problem:

formulation is:

model hillP3_4_15 "Paper Manufacturing";
set
mill,i := [p1,p2,p3,p4];
machine,m := [m1,m2,m3];
rawmat,r := [r1,r2,r3,r4];
papertype,p := [t1,t2,t3,t4,t5];
customer,c := [c1,c2,c3,c4,c5,c6,c7,c8];
parameter
PaperDemand{c,p} := Rnd(1,20);
RawMatUsed{p,m,r} := Rnd(1,4);
RawMatAvail{i,r} := Rnd(100,1000);
MachineUse{p,m} := Rnd(1,4);
MachineAvail{i,m} := Rnd(100,1000);
ProduceCost{i,p,m} := Rnd(1,10);
TransportCost{i,c,p} := Rnd(1,10);
variable
Produce,P{i,m,p|ProduceCost};
Ship,S{i,c,p|TransportCost};
constraint
RawMat{i,r}: sum{m,p} RawMatUsed*P <= RawMatAvail;
ProdCap{i,m}: sum{p} MachineUse*P <= MachineAvail;
MillBalance{i,p}: sum{m} P = sum{c} S;
MeetDemand{c,p}: sum{i} S >= PaperDemand;
minimize TotalCost: sum{i,m,p} ProduceCost*P + sum{i,c,
p} TransportCost*S;
Writep(TotalCost,P,S);
end

4.51. Steak-Potato Diet (hillP3-4-6)

Problem:

mulation is:

model hillP3_4_6 "Steak-Potato Diet";
set
food ,f := [Steak, Potatoes];
ingredient,i := [Carbohydrates, Protein, Fat];
parameter
IngredPerServing{i,f} := [ 5, 15, 20, 5, 15, 2];
MinDailyReq{i} := [50, 40, 0];
MaxDailyReq{i} := [ 0, 0, 60];
CostPerServing{f} := [4, 2];
variable Servings{f};
constraint
MinRequirement{i|MinDailyReq}: sum{f}
IngredPerServing*Servings >= MinDailyReq;
MaxRequirement{i|MaxDailyReq}: sum{f}
IngredPerServing*Servings <= MaxDailyReq;
minimize TotalCost: sum{f} CostPerServing*Servings ;
Writep(TotalCost,Servings);
end

4.52. Pig Feed Blending (hillP3-4-7)

Problem:

mulation is:

model hillP3_4_7 "Pig Feed Blending";
set
f := [Corn, Tankage, Alfalfa] "feed";
i := [Carbohydrates, Protein, Vitamins] "ingredient";
parameter
Ingredient{f, i}:=[90 30 10, 20 80 20, 40 60 60];
MinDailyReq{i} :=[200, 180, 150];
Cost{f} :=[0.84, 0.72, 0.60];
variable FeedUse,F{f};
constraint Requirement{i}: sum{f} Ingredient*F >=
MinDailyReq;
minimize TotalCost: sum{f} Cost*F;
Writep(TotalCost,F);
end

4.53. Plant Production Planning (hillP3-4-8)

Problem:

mulation is:

model hillP3_4_8 "Plant Production Planning";
set
plant ,p := [1..3];
size ,s := [Large, Medium, Small];
parameter
Profit{s} := [420, 360, 300];
PlantCapacity{p} := [750, 900, 450];
SpaceAvail{p} := [13000, 12000, 5000];
SpaceReq{s} := [20, 15, 12];
Demand{s} := [900, 1200, 750];
variable Produce{p,s};
constraint
CapacityLimit{p}: sum{s} Produce <= PlantCapacity;
SpaceLimit{p}: sum{s} SpaceReq*Produce <= SpaceAvail;
MaxDemand{s}: sum{p} Produce <= Demand;
EqualCapUse{p|p<#p}: 1/PlantCapacity*(sum{s} Produce)
= 1/PlantCapacity[p+1]*(sum{s} Produce[p+1,s]);
maximize TotalProfit: sum{p,s} Profit*Produce ;
end

4.54. Cargo Plane Planning (hillP3-4-9)

Problem:

mulation is:

model hillP3_4_9 "Cargo Plane Planning";
set
c := [Front, Center, Back] "compartment";
g :=[ 1..4] "cargo";
parameter
WeightCapacity{c} := [12, 18, 10];
SpaceCapacity{c} := [7000, 9000, 5000];
CargoAvail{g} := [20, 16, 25, 13];
Volume{g} := [500, 700, 600, 400];
Profit{g} := [320, 400, 360, 290];
variable AcceptCargo{c,g};
constraint
CargoLimit{g}: sum{c} AcceptCargo <= CargoAvail;
WeightLimit{c}: sum{g} AcceptCargo <= WeightCapacity;
SpaceLimit{c}: sum{g} Volume*AcceptCargo <=
SpaceCapacity;
maximize TotalProfit: sum{c,g} Profit*AcceptCargo;
Writep(TotalProfit,AcceptCargo);
end

4.55. MidWest Grain Elevator (midwest)

Problem:
Midwest Grain Elevator: Case Summary Statement: Midwest Grain Elevator
performs storage and handling services for buyers and sellers of grain on a
service fee basis. To achieve maximum profit Midwest must maximize the
amount of grain it handles. In order to accomplish this objective, Midwest
must build customer loyalty on the part of both buyers and sellers of grain.
Thus, Midwest’s preferred policy is to encourage repeat business by satisfying
buyers’ orders at minimum cost. This policy benefits sellers as well as by
increasing the number of orders for high quality grain that can be filled.
Currently Midwest is storing 507,900 bushels of corn for Saw Mill River Feed
& Grain Company. This corn has been classified into 26 different types, which
differ by moisture content, weight per bushel, amount of damage, and amount
of foreign material. Saw Mill has limited quantities of each of the 26 types of
corn. The ability to meet an order is obviously restricted by this supply. The
quantities and cost of each type of corn is detailed in the table shown below.
Corn Cost and Supply Table
Corn Type Supply Cost

--------------------------------------
1 15000 1.50
2 30000 1.45
3 45000 1.44
4 25000 1.45
5 40000 1.42
6 20000 1.38
7 30000 1.37
8 75000 1.37
9 15000 1.39
10 50000 1.39
11 16000 1.27
12 20000 1.28
13 10000 1.17
14 12000 1.22
15 6000 1.12
16 2000 1.18
17 5000 1.42
18 4000 1.42
19 5000 1.42
20 6000 1.36
21 1300 1.29
22 29000 1.29
23 1900 1.42
24 33000 1.40
25 6700 1.22
26 5000 1.30
--------------------------------------
Saw Mill has now received an order for 6 different grades of corn totaling
1. bushels which it intends to fill from the grain stored with Midwest. The
Saw Mill must meet the stated demand for each of the 6 grades. Exceed-
ing demand would be ill-advised, since there is no guarantee that the
next order filled will require the same characteristics. The demand for
each type of grade is shown in the table below.
Corn Grade Demand
-------------------------
Gr1 40000
Gr2 32000
Gr3 50000
Gr4 20000
Gr5 30000
Gr6 100000
-------------------------
Midwest’s handling fee will be based on supplying the 272,000 bushels
of corn, regardless of the manner in which the various corns are blended
to meet the specifications. However, in keeping with the policy of pro-
viding quality service to other buyers and sellers, the problem Mid-
west faces is that of determining the least cost blend of available grains
that will satisfy the specified criteria of the current order. Each of the
6 grades of corn demanded must meet specified criteria for moisture
content, minimum weight per bushel, percentage damaged, and per-
centage of foreign material, which are detailed in the table below. Corn
Grade Characteristics:
Grade MaxMoisture MinWeight MaxDamage MaxForeign
---------------------------------------------------------
Gr1 13.0 56.0 2.0 2.0
Gr2 15.5 54.0 5.0 3.0
Gr3 15.0 56.0 2.0 4.0
Gr4 16.0 54.0 9.0 4.0
Gr5 23.0 54.0 10.0 6.0
Gr6 20.0 54.0 9.0 4.0
---------------------------------------------------------
For each of the 6 grades the four quality requirements are as follows:
The weighted average moisture content cannot exceed the maximum
acceptable moisture level for that grade. The weighted average weight
per bushel of the corns used in the blend of the grades must meet or
exceed the minimum level. The weighted average percent damaged

of the 26 types of corn used in the blend cannot exceed the maximum
level. The weighted average percentage of foreign material that could
be used in the blend cannot exceed the maximum requirement. The
quality characteristics for each of the 26 corn types is given in the table
below. Corn Quality Characteristics:
Corn Type Moisture Weight Damage Foreign
------------------------------------------------------
1 11.0 58.0 3.0 1.0
2 12.0 57.0 2.0 1.5
3 15.0 57.0 2.0 1.0
4 12.0 58.0 3.0 3.0
5 13.0 56.0 4.0 2.0
6 15.0 54.0 4.0 2.0
7 15.0 55.0 5.0 3.0
8 18.0 57.0 5.0 1.0
9 14.0 58.0 2.0 4.0
10 15.0 55.0 3.0 2.0
11 17.0 53.0 7.0 5.0
12 15.0 55.0 8.0 3.0
13 22.0 56.0 8.0 5.0
14 18.0 54.0 13.0 5.0
15 17.0 55.0 20.0 8.0
16 13.5 57.0 30.0 2.0
17 13.5 57.0 3.0 2.0
18 13.5 57.5 3.0 2.0
19 13.5 57.0 5.0 3.0
20 15.0 56.5 10.0 4.0
21 13.5 58.0 15.0 4.0
22 13.0 57.0 15.0 5.0
23 13.0 58.0 3.0 2.0
24 13.0 56.0 5.0 3.0
25 14.0 56.0 20.0 5.0
26 15.0 56.5 10.0 4.0
-------------------------------------------------------
The objective for this problem is to determine the number of bushels used
in the blending for each of the 26 types of corn so the the total cost of the 6
desired grades of corn is minimized.
Reference: Prof. Ramesh Sharda, Operations Research Dept., Oklahoma State
University
Modeling Steps

model midwest "MidWest Grain Elevator";
set
type := [1..26];
grade := [Gr1, Gr2, Gr3, Gr4, Gr5, Gr6];
parameter
-- Supply characteristics
Supply{type} := [15000, 30000, 45000, 25000, 40000,
20000, 30000, 75000, 15000,
50000, 16000, 20000, 10000, 12000, 6000, 2000,
5000, 4000, 5000, 6000, 1300, 29000, 1900,
33000, 6700, 5000];
Cost{type} := [1.50, 1.45, 1.44, 1.45, 1.42, 1.38,
1.37, 1.37, 1.39
1.39, 1.27, 1.28, 1.17, 1.22, 1.12, 1.18, 1.42, 1.42
1.42, 1.36, 1.29, 1.29, 1.42, 1.40, 1.22, 1.30 ];
Moisture{type} := [11, 12, 15, 12, 13, 15, 15, 18,
14,
15, 17, 15, 22, 18, 17, 13.5, 13.5 13.5,
13.5, 15, 13.5, 13, 13, 13, 14, 15];
Weight{type} := [58.0, 57.0, 57.0, 58.0, 56.0, 54.0,
55.0, 57.0, 58.0,
55.0, 53.0, 55.0, 56.0, 54.0, 55.0, 57.0, 57.0,
57.5,
57.0, 56.5, 58.0, 57.0, 58.0, 56.0, 56.0, 56.5];
Damage{type} := [ 3, 2, 2, 3, 4, 4, 5, 5, 2,
3, 7, 8, 8, 13, 20, 30, 3, 3,
5, 10, 15, 15, 3, 5, 20, 10];
Foreign {type} := [1, 1.5, 1, 3, 2, 2, 3, 1, 4,
2, 5, 3, 5, 5, 8, 2, 2, 2,
3, 4, 4, 5, 2, 3, 5, 4];
-- Demand Characteristics
Demand{grade} := [40000, 32000, 50000, 20000, 30000,
100000];
MaxMoisture{grade} := [13, 115, 15, 16, 23, 20];
MinWeight{grade} := [56, 54, 56, 54, 54, 54];
MaxDamage{grade} := [2, 5, 2, 9, 10, 9];
MaxForeign{grade} := [2, 3, 4, 4, 6, 4];
variable Bushels{type,grade};
constraint
QuantitySupplied{type}: sum{grade} Bushels <= Supply;
QuantityDemanded{grade}: sum{type} Bushels = Demand;
MoistureContent{grade}: sum{type} Moisture*Bushels <=
MaxMoisture*(sum{type} Bushels);
WeightPerBushel{grade}: sum{type} Weight*Bushels >=
MinWeight*(sum{type} Bushels);
DamagePercent{grade}: sum{type} Damage*Bushels <=
MaxDamage*(sum{type} Bushels);
ForeignMaterial{grade}: sum{type} Foreign*Bushels <=

MaxForeign*(sum{type} Bushels);
minimize totalcost: sum{type,grade} Cost*Bushels;
Writep(totalcost,Bushels);
end

Solution:
opt: 366863;
4.56. Fertilizer (murty2-1)

Problem:

explained in [19], Example 2.1, Page 23 A mathematical model formulation is:

model murty2_1 "Fertilizer";
set
fertilizer,f := ['Hi-ph', 'Lo-ph'];
material ,m := [RM1, RM2, RM3];
parameter
NetProfit{f} := [15, 10];
RawMatReq{f,m} := [2, 1, 1, 1, 1, 0];
RawMatAvail{m} := [1500, 1200, 500];
variable Produce{f};
constraint Req{m}: sum{f} RawMatReq*Produce <=
RawMatAvail;
maximize TotalProfit: sum{f} NetProfit*Produce;
end

4.57. Gasoline Blending (murty2-2)

Problem:


model murty2_2 "Gasoline Blending";
set
rawgas,r := [1..4];
fueltype,f := [1..3];
parameter
OctaneRating{r} := [68, 86, 91, 99];
RawGasAvail{r} := [4000, 5050, 7100, 4300];
RawGasCost{r} := [31.02, 33.15, 36.35, 38.75];
RawGasPrice{r} := [36.85, 36.85, 38.95, 38.95];
MinOctaneFuel{f} := [95, 90, 85];
FuelPrice{f} := [45.15, 42.95, 40.99];
MinimumDemand{f} := [0, 0, 15000];
MaximumDemand{f} := [10000, 0, 0];
variable
FuelComposition,FC{f,r};
RawGasSold,RG{r};
expression
TotalFuelRevenue : sum{r,f} FuelPrice*FC;
TotalRawGasRevenue: sum{r} RawGasPrice*RG;
TotalRawGasCost : sum{r,f} RawGasCost*FC + sum{r}
RawGasCost*RG;
constraint
RawGasLimit{r}: sum{f} FC + RG <= RawGasAvail;
OctaneReq{f}: sum{r} OctaneRating*FC >= MinOctaneFuel
*(sum{r} FC);
MinDemand{f|MinimumDemand}: MinimumDemand <= sum{r}
FC;
MaxDemand{f|MaximumDemand}: sum{r} FC <=
MaximumDemand;
maximize TotalProfit: TotalFuelRevenue +
TotalRawGasRevenue - TotalRawGasCost;
Writep(TotalProfit,FC,RG);
end

4.58. Diet Problem (murty2-3)

Problem:


model murty2_3 "Diet Problem";
set
grain := [1..2];
nutri := [Starch, Protein, Vitamins];
parameter
NutriUnits{nutri,grain} := [5, 7, 4, 2, 2, 1];
MinDailyReq{nutri} := [8, 15, 3];
GrainCost{grain} := [0.60, 0.35];
variable Amount{grain};
constraint
Requirement{nutri}: sum{grain} NutriUnits*Amount >=
MinDailyReq;
minimize TotalCost: sum{grain} GrainCost*Amount;
Writep(TotalCost,Amount);
end

4.59. Transportation (murty2-4)

Problem:

explained in [19],Example 2.4, Page 33 A mathematical model formulation is:

model murty2_4 "Transportation";
set
mine := [1..2];
plant := [1..3];
parameter
Availability{mine} := [800, 300];
Requirement{plant} := [400, 500, 200];
ShipCost{mine,plant} := [11, 8, 2, 7, 5, 4];
variable Ship{mine,plant};
constraint
OreLimit{mine}: sum{plant} Ship <= Availability;
OreReq{plant}: sum{mine} Ship >= Requirement;
minimize TotalCost: sum{mine,plant} ShipCost*Ship;
end

4.60. Marriage Problem (murty2-5)

Problem:

explained in [19],Example 2.5, Page 36 A mathematical model formulation is:

model murty2_5 "Marriage Problem";
set
man := [1..5];
woman := [1..5];
parameter
Happyness{man,woman} := [ 78, -16, 19, 25, 83,
99, 98, 87, 16, 92,
86, 19, 39, 88, 17,
-20, 99, 88, 79, 65,
67, 98, 90, 48, 60];
variable TimeTogether{man,woman};
constraint
MenMonogamous{man}: sum{woman} TimeTogether = 1;
WomenMonogamous{woman}: sum{man} TimeTogether = 1;
maximize TotalHappyness: sum{man,woman} Happyness*
TimeTogether;
Writep(TotalHappyness,TimeTogether);
end

4.61. Multi-Period Planning (murty2-6)

Problem:


model murty2_6 "Multi-Period Planning";
set period,p := [1..6];
parameter
ProdCost{p} := [20, 25, 30, 40, 50, 60];
ProdCapacity{p}:= [1500 2000 2200 3000 2700 2500];
Demand{p} := [1100 1500 1800 1600 2300 2500];
SellingPrice{p}:= [180, 180, 250, 270, 300, 320];
InventoryCost := 2;
InitialStock := 500;
FinalStock := 500;
variable
Produce,P{p} [0..ProdCapacity];
Inventory,I{p};
Sales,S{p} [0..Demand];
expression
TotalRevenue : sum{p} SellingPrice*S;
TotalProdCost: sum{p} ProdCost*P;
TotalInvtCost: sum{p} InventoryCost*I;
TotalCost : TotalProdCost + TotalInvtCost;
constraint
InventoryBalance{p}: I = if(p=1,InitialStock,I[p-1])
+ P - S;
F: I[6] = FinalStock;
maximize NetProfit: TotalRevenue - TotalCost;
Writep(NetProfit,P,I,S);
end

4.62. Breck and Dapper (shapiro1-1)

Problem:

explained in [25], Example 1.1, Page 12. A mathematical model formulation
is:

model shapiro1_1 "Breck and Dapper";
set drill := [1..4];
parameter
PlastReq{drill} := [0.82, 0.62, 1.42, 2.03]; -- lb
CopperReq{drill} := [0.43, 0.69, 0.33, 0.20]; -- lb
WireReq{drill} := [15, 16, 9, 9]; -- yd
Contrib{drill} := [12.5 11.3 17.2 19.9]; -- dollars
PlastAvail := 16000; -- lb
CopperAvail := 5000; -- lb
WireOnHand := 8000; -- yd
ProdWireCap := 80000; -- yd
ProdWireRate := 3.6/100; -- lb/yd
WireProdCost := 0.14;
WirePurchaseCost := 0.29;
variable
Produce,x{drill};
WirePurchase,Wp;
WireProduce,Wm [0..ProdWireCap];
constraint
PlastLimit: sum{drill} PlastReq*x <= PlastAvail;
CopperLimit: sum{drill} CopperReq*x + ProdWireRate*Wm
<= CopperAvail;
WireLimit: sum{drill} WireReq*x <= WireOnHand + Wm +
Wp;
Marketing: x[1] + x[2] >= x[3] + x[4];
maximize TotalProfit: sum{drill} Contrib*x -
WireProdCost*Wm - WirePurchaseCost*Wp;
Writep(TotalProfit,x,Wp,Wm);
end

4.63. DowPont Chemical (shapiro1-2)

Problem:

is:

model shapiro1_2 "DowPont Chemical";
set
chemical,c := [R,A,B,C,D];
process,p := [1..3];
parameter
RawMatCost := 1;
ProcessCapacity{p} := [56, 25, 46];
MarketPrice{c} := [0.00, 2.40, 3.20, 6.40, 12.20];
ProcessCost{p,c} := /1 R 1.50, 2 A 3.50, 3 B 4.20, 3
C 4.20/;
ProcessOutput{p,c} := /1 A 0.45, 1 B 0.55, 2 C 0.80,
3 D 0.75/;
variable
ProdSale,S{c};
ProdUse,U{c};
constraint
r1: U['A']+S['A'] = ProcessOutput['1','A'] * U['R'];
r2: U['B']+S['B'] = ProcessOutput['1','B'] * U['R'];
r3: U['C']+S['C'] = ProcessOutput['2','C'] * U['A'];
r4: S['D'] = ProcessOutput['3','D']*(U['B'] + U['C'])
;
r5: U['B'] = 1.5*U['C'];
r6: U['R'] <= ProcessCapacity['1'];
r7: U['A'] <= ProcessCapacity['2'];
r8: U['B'] + U['C'] <= ProcessCapacity['3'];
maximize TotalProfit:
sum{c} MarketPrice*S - RawMatCost*U['R'] - sum{c,p}
ProcessCost*U;
Writep(TotalProfit,S,U);
end

4.64. Portfolio Selection (shapiro1-3)

Problem:

is:

model shapiro1_3 "Portfolio Selection";
set
bond := [A, B, C, D, E];
type := [Municipal, Agency, Government];
BondType{bond,type} := [A Municipal,
B Agency,
C Government,
D Government,
E Municipal];
parameter
TotalAvailForInvest := 10;
QualityBank{bond} := [2, 2, 1, 1, 5];
YearsToMaturity{bond} := [9, 15, 4, 3, 2];
Yield{bond} := [0.086, 0.108, 0.1, 0.088,
0.09];
YieldAfterTax{bond} := [0.086, 0.054, 0.05, 0.044,
0.09];
MaxMunicipal := 3;
MaxAvgQuality := 1.4;
MaxAvgMaturity := 5;
TaxRate := 0.5;
variable Invest,x{bond};
constraint
AmountInvested: sum{bond} x <= TotalAvailForInvest;
MaxInvestType{type|type='Municipal'}: sum{bond|
BondType} x <= MaxMunicipal;
QualityReq: sum{bond} QualityBank*x <= MaxAvgQuality
*(sum{bond} x);
MaturityReq: sum{bond} YearsToMaturity*x <=
MaxAvgMaturity*(sum{bond} x);
maximize AfterTaxEarnings: sum{bond} YieldAfterTax*x;
Writep(AfterTaxEarnings,x);
end

4.65. Portfolio Selection (shapiro1-3b)

Problem:

explained in [25], Example 1.3b, Page 21. A mathematical model formulation
is:

model shapiro1_3b "Portfolio Selection";
set bond := [A, B, C, D, E];
type := [Municipal, Agency, Government];
BondType{bond,type} := [A Municipal,
B Agency,
C Government,
D Government,
E Municipal];
parameter
TotalAvailForInvest := 10;
QualityBank{bond} := [2, 2, 1, 1, 5];
YearsToMaturity{bond} := [9, 15, 4, 3, 2];
Yield{bond} := [0.086 0.108 0.1 0.088 0.09];
YieldAfterTax{bond} := [0.086 0.054 0.05 0.044 0.09];
MaxMunicipal := 3;
MaxAvgQuality := 1.4;
MaxAvgMaturity := 5;
TaxRate := 0.5;
BorrowInterestRate := 0.11;
variable
Invest,x{bond};
Borrow,y;
constraint
AmountInvested: sum{bond} x <= TotalAvailForInvest+y;
MaxInvestType{type|type='Municipal'}: sum{bond|
BondType} x <= MaxMunicipal;
QualityReq: sum{bond} QualityBank*x <= MaxAvgQuality
*(sum{bond} x);
MaturityReq: sum{bond} YearsToMaturity*x <=
MaxAvgMaturity*(sum{bond} x);
MaxBorrow: Borrow <= 1;
maximize AfterTaxEarnings: sum{bond} YieldAfterTax*x -
BorrowInterestRate*TaxRate*y;
Writep(AfterTaxEarnings,x,y);
end

4.66. Transportation (shapiro1-4)

Problem:

is:

model shapiro1_4 "Transportation";
set
mill := [1..2];
market := [Northburg, Midburg, Southburg];
parameter
WheatAvail := 50000;
MillingCost{mill} := [0.13, 0.16];
MarketDemand{market} := [10000, 15000, 25000];
ShipCostMill{mill} := [0.31, 0.33];
ShipCostMarket{market, mill} := [0.18, 0.25, 0.22,
0.23, 0.27, 0.19];
variable
ShipMill{mill};
ShipMarket{mill, market};
constraint
WheatToMills: sum{mill} ShipMill = WheatAvail;
MillBalance{mill}: ShipMill = sum{market} ShipMarket;
MarketReq{market}: sum{mill} ShipMarket =
MarketDemand;
maximize TotalCost:
sum{mill} MillingCost*ShipMill
+ sum{mill} ShipCostMill*ShipMill
+ sum{mill,market} ShipCostMarket*ShipMarket;
Writep(TotalCost,ShipMill,ShipMarket);
end

4.67. Multi-Period Scheduling (shapiro1-5)

Problem:

is:

model shapiro1_5 "Multi-Period Scheduling";
set m,month := [1..12];
parameter
Demand{m} := [3800, 4100, 4700, 5900, 7200, 7900,
6700, 5100, 5300, 6400, 5800, 4800];
ProductionCost := 100; InventoryCost := 7;
LaborCost := 1600; HiringCost := 300;
LayoffCost := 500; StockOutCost := 500;
RecordersPerDay := 8; AssemblyCapacity:= 7000;
MaxNewWorkers := 50; MaxLayoffPercent:= 0.1;
InitInventory := 750; InitWorkforce := 500;
variable
Produce,P{m};
Inventory,I{m}; Ie "final Inventory";
Workers,W{m}; Hired,H{m};
LaidOff,L{m}; StockOut,S{m};
expression
TotalProductionCost: sum{m} ProductionCost*P ;
TotalInventoryCost : sum{m} InventoryCost*1/2*(I+if(m
=#m,Ie,I[m+1]));
TotalLaborCost : sum{m} LaborCost*W ;
TotalHiringCost : sum{m} HiringCost*H ;
TotalLayoffCost : sum{m} LayoffCost*L ;
TotalStockOutCost : sum{m} StockOutCost*S ;
constraint
InvBalance{m}: P+I-if(m=#m,Ie,I[m+1])+S = Demand;
WorkforceLimit{m}: P <= RecordersPerDay*W;
AssemblyLimit{m}: P <= AssemblyCapacity;
WorkforceBalance{m}: W = if(m=1,InitWorkforce,W[m-1])
+ H - L;
MaxHired{m}: H <= MaxNewWorkers;
MaxLaidOff{m}: L <= MaxLayoffPercent * if(m=1,
InitWorkforce,W[m-1]);
II: I[1] = InitInventory;
minimize TotalCost: TotalProductionCost +
TotalInventoryCost
+ TotalLaborCost + TotalHiringCost + TotalLayoffCost
+ TotalStockOutCost;
Writep(TotalCost,P,I,W,H,L,S);
end

4.68. Giapetto Wood Carving (winst3-1-1)

Problem: “Giapetto’s Woodcarving, Inc., manufactures two types of wooden
toys: soldiers and trains. A soldier sells for $27 and uses $10 worth of raw ma-
terials. Each soldier that is manufactured increases Giapetto’s variable labor
and overhead costs by $14. A train sells for $21 and uses $9 worth of raw ma-
terials. Each train built increases Giapetto’s variable labor and overhead costs
by $10. The manufacture of wooden soldiers and trains requires two types of
skilled labor: carpentry and finishing. A soldier requires 2 hours of finishing
labor and 1 hour of carpentry labor. A train requires 1 hour of finishing and
1 hour of carpentry labor. Each week, Giapetto can obtain all the needed raw
material but only 100 finishing hours and 80 carpentry hours. Demand for
trains is unlimited, but at most 40 soldiers are bought each week. Giapetto
wants to maximize weekly profit (revenues - costs). Formulate a mathemat-
ical model of Giapetto’s situation that can be used to maximize Giapetto’s
weekly profit.”This problem is from [31] , Example 3.1-1, Page 45.
Modeling Steps

model winst3_1_1 "Giapetto Wood Carving";
set
t := [Soldier, Train] "Toys";
w := [Carpentry, Finishing] "Labor";
parameter
Price{t} := [27, 21];
VariableCost{t} := [14, 10];
RawMatCost{t} := [10, 9];
Profit{t} := Price - VariableCost - RawMatCost;
LaborHours{t,w} := [1, 2, 1, 1];
LaborAvail{w} := [80, 100];
MaxSoldier := 40;
variable Produce,P{t};
constraint
LaborCapacity{w} : sum{t} LaborHours*P <= LaborAvail;
MaxDemand: P['Soldier'] <= MaxSoldier;
maximize TotalProfit : sum{t} Profit*P;
end

4.69. Giapetto Wood Carving II (winst3-1-1s)

Problem: “Giapetto’s Woodcarving, Inc., manufactures two types of wooden
toys: soldiers and trains. A soldier sells for $27 and uses $10 worth of raw ma-
terials. Each soldier that is manufactured increases Giapetto’s variable labor
and overhead costs by $14. A train sells for $21 and uses $9 worth of raw ma-
terials. Each train built increases Giapetto’s variable labor and overhead costs
by $10. The manufacture of wooden soldiers and trains requires two types of
skilled labor: carpentry and finishing. A soldier requires 2 hours of finishing
labor and 1 hour of carpentry labor. A train requires 1 hour of finishing and
1 hour of carpentry labor. Each week, Giapetto can obtain all the needed raw
material but only 100 finishing hours and 80 carpentry hours. Demand for
trains is unlimited, but at most 40 soldiers are bought each week. Giapetto
wants to maximize weekly profit (revenues - costs). Formulate a mathemat-
ical model of Giapetto’s situation that can be used to maximize Giapetto’s
weekly profit.”This problem is from [31] , Example 3.1-1, Page 45.
Modeling Steps

model winst3_1_1s "Giapetto Wood Carving II";
variable soldier; train;
constraint
carpentry: 1*soldier + 1*train <= 80;
finishing: 2*soldier + 1*train <= 100;
maxsoldier: soldier <=40;
maximize Profit: 27*soldier + 21*train - 24*soldier -
19*train;
Writep(Profit);
end

4.70. Dorian Advertising (winst3-2-2)

Problem: ”Dorian Auto manufactures luxury cars and trucks. The company
believes that its most likely customers are high-income women and men. To
reach these groups, Dorian Auto has embarked on an ambitious TV advertis-
ing campaign and has decided to purchase 1-minute commercial spots on two
types of programs: comedy shows and football games. Each comedy com-
mercial is seen by 7 million high-income women and 2 million high-income
men. Each football commercial is seen by 2 million high-income women and
12 million high-income men. A 1-minute comedy ad costs $50,000, and a 1-
minute football ad costs $100,000. Dorian would like the commercials to be
seen by at least 28 million high-income women and 24 million high-income
men. Use linear programming to determine how Dorian Auto can meet its
advertising requirements at minimum cost.” This probelm is from [31] , Ex-
ample 3.2-2, Page 57.
Modeling Steps

model winst3_2_2 "Dorian Advertising";
set
a := [Comedy, Football] "adtype";
c := [HIW, HIM] "customers";
parameter
MinuteCost{a} := [50, 100];
SeenBy{a,c} := [7, 2, 2, 12];
TargetReach{c} := [28, 24];
variable Purchase{a};
constraint
MustReach{c}: sum{a} SeenBy*Purchase >= TargetReach;
minimize TotalCost: sum{a} MinuteCost*Purchase ;
Writep(TotalCost,Purchase);
end

4.71. Dorian Advertising (winst3-2-2s)

Problem: ”Dorian Auto manufactures luxury cars and trucks. The company
believes that its most likely customers are high-income women and men. To
reach these groups, Dorian Auto has embarked on an ambitious TV advertis-
ing campaign and has decided to purchase 1-minute commercial spots on two
types of programs: comedy shows and football games. Each comedy com-
mercial is seen by 7 million high-income women and 2 million high-income
men. Each football commercial is seen by 2 million high-income women and
12 million high-income men. A 1-minute comedy ad costs $50,000, and a 1-
minute football ad costs $100,000. Dorian would like the commercials to be
seen by at least 28 million high-income women and 24 million high-income
men. Use linear programming to determine how Dorian Auto can meet its
advertising requirements at minimum cost.” This probelm is from [31] , Ex-
ample 3.2-2, Page 57.
Modeling Steps

model winst3_2_2s "Dorian Advertising";
variable Comedy; Football;
constraint
Women: 7*Comedy + 2*Football >= 28;
Men: 2*Comedy + 12*Football >= 24;
minimize Cost: 50*Comedy + 100*Football;
Writep(Cost);
end

4.72. Auto Company (winst3-3-3)

Problem: ”An auto company manufactures cars and trucks. Each vehicle
must be processed in the paint shop and body assembly shop. If the paint
shop were only painting trucks, 40 per day could be painted. If the paint shop
were only painting cars, 60 per day could be painted. If the body shop were
only producing cars, it could process 50 per day. If the body shop were only
producing trucks, it could process 50 per day. Each truck contributes $300
to profit, and each car contributes $200 to profit. Use linear programming
to determine a daily production schedule that will maximize the company’s
profits.” This probelm is from [31] , Example 3.3-3, Page 61.
Modeling Steps

model winst3_3_3 "Auto Company";
set
vehicle := [Cars, Trucks];
shop := [Paint, BodyAssembly];
parameter
Profit{vehicle} := [200, 300];
CapPerDay{shop, vehicle} := [60, 40, 50, 50];
variable Produce{vehicle};
constraint
ProdCapacity{shop}: sum{vehicle} 1/CapPerDay*Produce
<= 1;
maximize TotalProfit: sum{vehicle} Profit*Produce ;
end

4.73. Auto Company (winst3-3-3s)

Problem: ”An auto company manufactures cars and trucks. Each vehicle
must be processed in the paint shop and body assembly shop. If the paint
shop were only painting trucks, 40 per day could be painted. If the paint shop
were only painting cars, 60 per day could be painted. If the body shop were
only producing cars, it could process 50 per day. If the body shop were only
producing trucks, it could process 50 per day. Each truck contributes $300
to profit, and each car contributes $200 to profit. Use linear programming
to determine a daily production schedule that will maximize the company’s
profits.” This probelm is from [31] , Example 3.3-3, Page 61.
Modeling Steps

model winst3_3_3s "Auto Company";
variable Cars; Trucks;
constraint
PaintShop: 1/60*Cars + 1/40*Trucks <= 1;
BodyAssembly: 1/50*Cars + 1/50*Trucks <= 1;
maximize Profit: 300*Trucks + 200*Cars;
Writep(Profit);
end

4.74. Auto Company (infeasible) (winst3-3-4)

Problem: ”Suppose that auto dealers requier that the auto company in Exam-
ple 3 produce at least 30 trucks and 20 cars. Find the optimal solution to the
new LP.” This probelm is from [31] , Example 3.3-4, Page 63.
Modeling Steps

model winst3_3_4 "Auto Company (infeasible)";
set
vehicle := [Cars, Trucks];
shop := [Paint, BodyAssembly];
parameter
Profit{vehicle} := [200, 300];
CapPerDay{shop, vehicle} := [60, 40, 50, 50];
MinProd{vehicle} := [20, 30];
variable Produce{vehicle} [MinProd..99999999];
constraint
ProdCapacity{shop} : sum{vehicle} 1/CapPerDay*Produce
<= 1;
maximize TotalProfit: sum{vehicle} Profit*Produce;
Write('The model is infeasible\n');
end

4.75. Unbounded Problem (winst3-3-5)

Problem: “Graphically solve the following LP”. This probelm is from [31] ,
Example 3.3-5, Page 64.
max z = 2x1 − x2
subject to x1 − x2 ≤ 1
2x1 + x2 ≥ 6
x1 , x2 ≥ 0
Modeling Steps

model winst3_3_5 "Unbounded Problem";
variable x1; x2;
constraint r1: x1-x2<=1 and 2*x1+x2>=6;
maximize z: 2*x1-x2;
Writep(z);
end

4.76. Diet Problem (winst3-4-6)

Problem: “My diet requires that all the food I eat come from one of the four
"basic food groups" (chocolate cake, ice cream, soda, and cheesecake). At
present, the following four foods are available for consumption: brownies,
chocolate ice cream, cola, and pineapple cheesecake. Each brownie costs 50
cents, each scoop of chocolate ice cream costs 20 cents, each bottle of cola
costs 30 cents, and each piece of pineapple cheesecake costs 80 cents. Each
day, I must ingest at least 500 calories, 6 oz of chocolate, 10 oz of sugar, and
8 oz of fat. The nutritional content per unit of each food is shown in Table
2. Formulate a linear programming model that can be used to satisfy my
daily nutritional requirements at minimum cost.” This probelm is from [31] ,
Modeling Steps

model winst3_4_6 "Diet Problem";
set
foods := [Brownie, Icecream, Cola, Cheesecake];
nutrients := [Calories, Chocolate, Sugar, Fat];
parameter
FoodCost{foods} := [0.50, 0.20, 0.30, 0.80];
NutriReq{nutrients} := [500, 6, 10, 8];
NutriValues{foods,nutrients} := [400, 3, 2, 2,
200, 2, 2, 4,
150, 0, 4, 1,
500, 0, 4, 5];
variable Consume{foods};
constraint
Required{nutrients}: sum{foods} NutriValues*Consume
>= NutriReq;
minimize Cost: sum{foods} FoodCost*Consume;
Writep(Cost,Consume);
end

4.77. Diet Problem (winst3-4-6s)

Problem: “My diet requires that all the food I eat come from one of the four
"basic food groups" (chocolate cake, ice cream, soda, and cheesecake). At
present, the following four foods are available for consumption: brownies,
chocolate ice cream, cola, and pineapple cheesecake. Each brownie costs 50
cents, each scoop of chocolate ice cream costs 20 cents, each bottle of cola
costs 30 cents, and each piece of pineapple cheesecake costs 80 cents. Each
day, I must ingest at least 500 calories, 6 oz of chocolate, 10 oz of sugar, and
8 oz of fat. The nutritional content per unit of each food is shown in Table
2. Formulate a linear programming model that can be used to satisfy my
daily nutritional requirements at minimum cost.” This problem is from [31] ,
Modeling Steps

model winst3_4_6s "Diet Problem";
variable Brownie,B; Icecream,I; Cola,C; Cheesecake,CC;
constraint
Calories: 400*B + 200*I + 150*C + 500*CC >= 500;
Chocolate: 3 *B + 2* I >= 6;
Sugar: 2 *B + 2* I + 4* C + 4*CC >= 10;
Fat: 2 *B + 4* I + C + 5*CC >= 8;
minimize Cost: 0.50*B + 0.20*I + 0.30*C + 0.80*CC;
Writep(Cost);
end

4.78. Star Oil (winst3-6-8)

Problem: “Star Oil Company is considering five different investment oppor-
tunities. The cash outflows and net present values (in millions of dollars) are
given in Table 4.22. Star Oil has $40 million available for investment at the
present time (time 0); it estimates that one year from now (time 1) $20 million
will be available for investment. Star Oil may purchase any fraction of each
investment. In this case, the cash outflows and NPV are adjusted accordingly.
For example, if Star Oil purchases one fifth of investment 3, then a cash out-
flow of 5 (5) - $1 million would be required at time 0, and a cash outflow of
-Is– (5) $1 million would be required at time 1. The one-fifth share of invest-
ment 3 would yield an NPV of I (16) $3.7 million. Star Oil wants to maximize
the NPV that can be obtained by investing in investments 1-5. Formulate an
LP that will help achieve this goal. Assume that any funds left over at time 0
cannot be used at time 1. ” This probelm is from [31] , Example 3.6-8, Page 74.
Inv.1 Inv.2 Inv.3 Inv.4 Inv.5

Time 0 cash outflow $11 $53 $5 $5 $29
Time 1 cash outflow $3 $6 $5 $1 $34
NPV $13 $16 $16 $14 $39
Table 4.1:
Modeling Steps

model winst3_6_8 "Star Oil";
set
invest := [1..5];
time := [0..1];
parameter
CashAvail{time} := [40, 20];
CashOutflow{time,invest} := [11, 53, 5, 5, 29,
3, 6, 5, 1, 34];
NPV{invest} := [13, 16, 16, 14, 39];
variable Purchase,P{invest} [0..1];
constraint
InvestLimit{time}: sum{invest} CashOutflow*P <=
CashAvail;
maximize TotalNetPresentValue: sum{invest} NPV*P;
Writep(TotalNetPresentValue,P);
end

4.79. Semicond Eletronics (winst3-7-9)

Problem: “Semicond is a small electronics company that manufactures tape
recorders and radios. The per-unit labor costs, raw material costs, and selling
price of each product aregiven in Table 4.22. On December 1, 1998, Semicond
has available raw material that is sufficient to manufacture 100 tape recorders
and 100 radios. On the same date, the company’s balance sheet is as shown
in Table 4.17, and Semicond’s asset/liability ratio (called the current ratio) is
20,000/10,000 = 2.
Semicond must determine how many tape recorders and radios should be
produced during December. Demand is large enough to ensure that all goods
produced will be sold. All sales are on credit, however, and payment for
goods produced in December will not be received until February 1, 1999.
During December, Semicond will collect $2000 in accounts receivable, and
Semicond must pay off $1000 of the outstanding loan and a monthly rent of
$1000. On January 1, 1999, Semicond will receive a shipment of raw material
worth $2000, which will be paid for on February 1, 1999. Semicond’s man-
agement has decided that the cash balance on January 1, 1999 must be at least
$4000. Also, Semicond’s bank requires that the current ratio at the beginning
of January be at least 2. To maximize the contribution to profit from Decem-
ber production, (revenues to be received) - (variable production costs), what
should Semicond produce during December?” This probelm is from [31] ,
Tape Recorder Radio

Selling Price $100 $90
Labor cost $ 50 $35
Raw material cost $ 30 $40
Table 4.2:
Assets Liabilities
Cash $10,000
Accounts receivable* $ 3,000
Inventory outstanding** $ 7,000
Bank loan $ 10,000
Table 4.3:
* Accounts receivable is money owed to Semicond by customers who have

previously purchased Semicond products.
** Value of December 1, 1998 inventory = 30(100)+ 40(100) = $7000
Modeling Steps

model winst3_7_9 "Semicond Eletronics";
set p := [TapeRec, Radio] "products";
parameter
SellingPrice{p} := [100, 90];
LaborCost{p} := [50, 35];
RawMatCost{p} := [30, 40];
ContribProfit{p} := SellingPrice - LaborCost -
RawMatCost;
RawMatAvail{p} := [100, 100];
CashOnHandDec1 := 10000;
AccountsRecDec1 := 3000;
InventoryValueDec1 := 7000;
LiabilitiesDec1 := 10000;
RawMatReceivedJan1 := 2000;
AccRecCollectDec := 2000;
LoanPaymentDec := 1000;
MonthlyRentDec := 1000;
MinCashOnHandJan := 4000;
MinAssetLiabRatio := 2;
variable ProduceDec,P{p};
expression
CashPositionJan1: CashOnHandDec1 + AccRecCollectDec -
LoanPaymentDec
- MonthlyRentDec - sum{p} LaborCost * P;
AccountsRecJan1: AccountsRecDec1 + sum{p}
SellingPrice*P - AccRecCollectDec;
InventoryValueJan1: InventoryValueDec1 - sum{p}
RawMatCost*P + RawMatReceivedJan1;
LiabilitiesJan1: LiabilitiesDec1 - LoanPaymentDec +
RawMatReceivedJan1;
AssetsJan1: CashPositionJan1 + AccountsRecJan1 +
InventoryValueJan1;
constraint
RawMatLimit{p}: P <= RawMatAvail;
CashBalance: CashPositionJan1 >= MinCashOnHandJan;
AssetLiabRatio: AssetsJan1 >= MinAssetLiabRatio*
LiabilitiesJan1;
maximize TotalProfit: sum{p} ContribProfit*ProduceDec ;
end

4.80. Rylon Corporation (winst3-9-11)

Problem: “Rylon Corporation manufactures Brute and Chanelle perfumes.
The raw material needed to manufacture each type of perfume can be pur-
chased for $3 per pound. Processing 1 lb of raw material requires 1 hour of
laboratory time. Each pound of processed raw material yields 3 oz of Regu-
lar Brute Perfume and 4 oz of Regular Chanelle Perfume. Regular Brute can
be sold for $7/oz and Regular Chanelle for $6/oz. Rylon also has the option
of further processing Regular Brute and Regular Chanelle to produce Luxury
Brute, sold at $18/oz, and Luxury Chanelle, sold at $14/oz. Each ounce of
Regular Brute processed further requires an additional 3 hours of laboratory
time and $4 processing cost and yields 1 oz of Luxury Brute. Each ounce of
Regular Chanelle processed further requires an additional 2 hours of labora-
tory time and $4 processing cost and yields 1 oz of Luxury Chanelle. Each
year, Rylon has 6000 hours of laboratory time available and can purchase up
to 4000 lb of raw material. Formulate an LP that can be used to determine how
Rylon can maximize profits. Assume that the cost of the laboratory hours is a
fixed cost.” This probelm is from [31] , Example 3.9-11, Page 90.
Modeling Steps

model winst3_9_11 "Rylon Corporation";
set p := [Brute, Chanele] "perfumes";
t := [Regular, Luxury] "types";
parameter
RawMatCost := 3.00; -- $/lb
RawMatYield{p} := [3, 4]; -- oz/lb
LabTimeRaw := 1; -- Hour/lb
LabTimeLux{p} := [3, 2]; -- Hour/oz
Price{p, t} := [7, 18, 6, 14]; -- $/oz
ProcessCost{p} := [4, 4]; -- $/oz
LabTimeAvail := 6000; RawMatAvail := 4000;
variable
Sales,S{p, t}; -- oz
RawMatPurchased,R; -- lb
expression
TotalRevenue : sum{p, t} Price*S;
TotalProcessCost: sum{p,t|t='Luxury'} ProcessCost*S;
TotalRawMatCost : RawMatCost*R;
constraint
RawMatLimit: R <= RawMatAvail;
LabTimeLimit: LabTimeRaw*R
+ sum{p,t|t='Luxury'} LabTimeLux*S <= LabTimeAvail;
MaterialBalance{p}: sum{t} S = RawMatYield*R;
maximize TotalProfit: TotalRevenue - TotalProcessCost -

TotalRawMatCost;
Writep(TotalProfit,S,R);
end

4.81. Sailco (winst3-10-12)

Problem: “Sailco Corporation must determine how many sailboats should be
produced during each of the next four quarters (one quarter = three months).
The demand during each of the next four quarters is as follows: first quarter,
40 sailboats; second quarter, 60 sailboats; third quarter, 75 sailboats; fourth
quarter, 25 sailboats. Sailco must meet demands on time. At the beginning
of the first quarter, Sailco has an inventory of 10 sailboats. At the beginning
of each quarter, Sailco must decide how many sailboats should be produced
during that quarter. For simplicity , we assume that sailboats manufactured
during a quarter can be used to meet demand for that quarter. During each
quarter, Sailco can produce up to 40 sailboats with regular-time labor at a
total cost of $400 per sailboat. By having employees work overtime during a
quarter, Sailco can produce additional sailboats with overtime labor at a total
cost of $450 per sailboat.
At the end of each quarter (after production has occurred and the current
quarter’s demand has been satisfied), a carrying or holding cost of $20 per
sailboat is incurred. Use linear programming to determine a production sched-
ule to minimize the sum of production and inventory costs during the next
four quarters.” This probelm is from [31] , Example 3.10-12, Page 95.
Modeling Steps

model winst3_10_12 "Sailco";
set
quarter,q := [1, 2, 3, 4];
labor,a := [reg, OT];
parameter
Demand{q} := [40, 60, 75, 25];
LaborCost{a} := [400.00, 450.00];
HoldCost{q} := [20.00, 20.00, 20.00, 20.00];
StartInven := 10;
variable
Produce,P{q,a};
Inventory,I{q};
constraint
ProdCapacity{q}: sum{a} P + if(q=1,StartInven,I[q-1])
= I + Demand;
up{q}: Produce[q,'reg'] <= 40;
minimize TotalCost: sum{q,a} LaborCost*P + sum{q}
HoldCost*I;
Writep(TotalCost);
end

4.82. Farmer Jones (winstP3-1-1)

Problem: “Farmer Jones must detertnme how many acres of corn and wheat
to plant this year. An acre of wheat yields bushels of wheat and requires 10
hours of labor per week. An acre of corn yields 10 bushels of com and requires
4 hours of labor per week. All wheat can be sold at $4 a bushel, and all corn
can be sold at $3 a bushel. Seven acres of land and 40 hours per week of labor
are available. Government regu-lations require that at least 30 bushels of corn
be produced during the current year. Let x1 = number of acres of corn planted,
and x2 = number of acres of wheat planted. Using these decision variables,
formulate an LP whose solution will tell Farmer Jones how to maximize the
total revenue from wheat and corn.” This probelm is from [31] , Problem 3.1-1,
Page 52.
Modeling Steps

model winstP3_1_1 "Farmer Jones";
set grain := [Corn, Wheat];
parameter
YieldBushel{grain} := [10, 25]; -- bushels/acre
ReqLabor{grain} := [4, 10]; -- hours/acres/week
Price{grain} := [3, 4]; -- $/bushel
LandAvail := 7; -- acres
LaborAvail := 40; -- hours/week
MinBushelsCorn := 30; -- bushels
variable PlantAcres {grain}; -- acres
constraint
GovReq{grain|grain='Corn'}: YieldBushel*PlantAcres >=
MinBushelsCorn;
Acres: sum{grain} PlantAcres <= LandAvail;
HoursPerWeek: sum{grain} ReqLabor*PlantAcres <=
LaborAvail;
maximize Revenue: sum{grain} Price*YieldBushel*
PlantAcres;
Writep(Revenue);
end

4.83. Farmer Jones (winstP3-1-1s)

Problem: “Farmer Jones must detertnme how many acres of corn and wheat
to plant this year. An acre of wheat yields bushels of wheat and requires 10
hours of labor per week. An acre of corn yields 10 bushels of com and requires
4 hours of labor per week. All wheat can be sold at $4 a bushel, and all corn
can be sold at $3 a bushel. Seven acres of land and 40 hours per week of labor
are available. Government regu-lations require that at least 30 bushels of corn
be produced during the current year. Let x1 = number of acres of corn planted,
and x2 = number of acres of wheat planted. Using these decision variables,
formulate an LP whose solution will tell Farmer Jones how to maximize the
total revenue from wheat and corn.” This probelm is from [31] , Problem 3.1-1,
Page 52.
Modeling Steps

model winstP3_1_1s "Farmer Jones";
variable Corn; Wheat; --acres
constraint
Acres: Corn + Wheat <= 7;
HoursPerWeek: 4*Corn + 10*Wheat <= 40;
GovReq: 10*Corn >= 30;
maximize Revenue: 3*10*Corn + 4*25*Wheat;
Writep(Revenue);
end

4.84. Truck Corporation (winstP3-1-4)

Problem: “Truckco manufactures two types of trucks: 1 and 2. Each truck
must go through the painting shop and assembly shop. If the painting shop
were completely devoted to painting type 1 trucks, 800 per day could be
painted, whereas if the painting shop were completely devoted to painting
type 2 trucks, 700 per day could be painted. If the assembly shop were com-
pletely devoted to assembling truck 1 engines, 1500 per day could be assem-
bled, and if the assembly shop were completely devoted to assembling truck
2 engines, 1200 per day could be assembled. Each type 1 truck contributes
$300 to profit; each type 2 truck contributes $500. Formulate an LP that will
maximize Truckco’s profit.” This probelm is from [31] , Problem 3.1-4, Page
52.
Modeling Steps

model winstP3_1_4 "Truck Corporation";
set
truck := [1, 2];
shop := [Painting, Assembly];
parameter
Profit{truck} := [300, 500];
CapPerDay{shop, truck} := [ 800, 700, 1500, 1200];
variable Produce{truck};
constraint ProdCapacity{shop}: sum{truck} 1/CapPerDay*
Produce <= 1;
maximize TotalProfit: sum{truck} Profit*Produce ;
end

4.85. Product-Mix (Truck Co) (winstP3-1-4s)

Problem: “Truckco manufactures two types of trucks: 1 and 2. Each truck
must go through the painting shop and assembly shop. If the painting shop
were completely devoted to painting type 1 trucks, 800 per day could be
painted, whereas if the painting shop were completely devoted to painting
type 2 trucks, 700 per day could be painted. If the assembly shop were com-
pletely devoted to assembling truck 1 engines, 1500 per day could be assem-
bled, and if the assembly shop were completely devoted to assembling truck
2 engines, 1200 per day could be assembled. Each type 1 truck contributes
$300 to profit; each type 2 truck contributes $500. Formulate an LP that will
maximize Truckco’s profit.” This probelm is from [31] , Problem 3.1-4, Page
52.
Modeling Steps

model winstP3_1_4s "Product-Mix (Truck Co)";
variable Truck1; Truck2;
constraint
Painting: 1/800*Truck1 + 1/700*Truck2 <= 1;
Body: 1/1500*Truck1 + 1/1200*Truck2 <= 1;
maximize Profit: 300*Truck1 + 500*Truck2;
Writep(Profit);
end

4.86. Leary Chemical (winstP3-2-3)

Problem: “Leary Chemical manufactures three chemicals: A, B, and C. These
chemicals are produced via two production processes: l and 2. Running pro-
cess for an hour costs S4 and yields 3 units of A, 1 of B, and 1 of C. Running
process 2 for an hour costs S1 and produces 1 unit of A and 1 of B. To meet
customer demands, at least 10 units of A, 5 of B, and 3 of C must be produced
daily. Graphically determine a daily production plan that minimizes the cost
of meeting Leary Chemical’s daily demands.” This probelm is from [31] ,
Problem 3.2-3, Page 60.
Modeling Steps

model winstP3_2_3 "Leary Chemical";
set
process := [P1, P2];
chemical := [A, B, C];
parameter
ProcessCost{process} := [4.00, 1.00];
ChemYield{chemical, process} := -- Units/hour
[3, 1, 1, 1, 1, 0];
DailyDemand{chemical} := [10, 5, 3]; -- Units
variable RunHours{process}; -- hours
constraint
MeetDemand{chemical}: sum{process} ChemYield*RunHours
>= DailyDemand;
minimize TotalCost: sum{process} ProcessCost*RunHours;
Writep(TotalCost);
end

4.87. Chemical Product Processes (winstP3-2-3s)

Problem: “Leary Chemical manufactures three chemicals: A, B, and C. These
chemicals are produced via two production processes: l and 2. Running pro-
cess for an hour costs S4 and yields 3 units of A, 1 of B, and 1 of C. Running
process 2 for an hour costs S1 and produces 1 unit of A and 1 of B. To meet
customer demands, at least 10 units of A, 5 of B, and 3 of C must be produced
daily. Graphically determine a daily production plan that minimizes the cost
of meeting Leary Chemical’s daily demands.” This probelm is from [31] ,
Modeling Steps

model winstP3_2_3s "Chemical Product Processes";
variable Process1; Process2;
constraint
ChemicalA: 3*Process1 + Process2 >= 10;
ChemicalB: Process1 + Process2 >= 5;
ChemicalC: Process1 >= 3;
minimize Cost: 4*Process1 + 1*Process2;
Writep(Cost);
end

4.88. Furniture Corporation (winstP3-2-5)

Problem: “Furnco manufactures desks and chairs. Each desk uses 4 units of
wood, and each chair uses 3. A desk contributes $40 to profit, and a chair
contributes $25. Marketing restntions require that the number of chairs pro-
duced be at least twice the number of desks produced. If 20 units of wood
are available, formulate an LP to maximize Furnco’s profit. Then graphically
solve the LP.” This probelm is from [31] , Problem 3.2-5, Page 60.
Modeling Steps

model winstP3_2_5 "Furniture Corporation";
set furniture := [Desks, Chairs];
parameter
Profit{furniture} := [40, 25];
WoodUse{furniture} := [4, 3];
WoodAvail := 20;
variable Produce{furniture};
constraint
WoodLimit: sum{furniture} WoodUse*Produce <=
WoodAvail;
Marketing: 2*Produce['Chairs'] >= Produce['Desks'];
maximize TotalProfit: sum{furniture} Profit*Produce ;
end

4.89. Product-Mix (Furniture Corporation)

(winstP3-2-5s)
Problem: “Furnco manufactures desks and chairs. Each desk uses 4 units of
wood, and each chair uses 3. A desk contributes $40 to profit, and a chair
contributes $25. Marketing restntions require that the number of chairs pro-
duced be at least twice the number of desks produced. If 20 units of wood
are available, formulate an LP to maximize Furnco’s profit. Then graphically
solve the LP.” This probelm is from [31] , Problem 3.2-5, Page 60.
Modeling Steps

model winstP3_2_5s "Product-Mix (Furniture Corporation)";
variable Desks; Chairs;
constraint
WoodAvail: 4*Desks + 3*Chairs <= 20;
Marketing: 2*Chairs >= Desks;
maximize Profit: 40*Desks + 25*Chairs;
Writep(Profit);
end

4.90. Money Manager (winstP3-3-10s)

Problem: “Money manager Boris Milkem deals with French cur-rency (the
franc) and American currency (the dollar). At 12 midnight, he can buy francs
by paying .25 dollars per franc and dollars by paying three francs per dol-
lar. Let xi = num-ber of dollars bought (by paying francs) and x2 = number
of francs bought (by paying dollars). Assume that both types of transactions
take place simultaneously, and the only constraint is that at 12:01 A.M., Boris
must have a nonnegative number of francs and dollars.
a Formulate an LP that enables Boris to maximize the number of dollars he

has after all transactions are completed.
b Graphically solve the LP and comment on the answer.”
This probelm is from [31] , Problem 3.3-10, Page 66.
Modeling Steps

model winstP3_3_10s "Money Manager";
variable Dollars; Francs;
constraint
MinDollars: Dollars + 0.25*Francs >= 0;
MinFrancs: 3*Dollars + Francs >= 0;
maximize TotalDollars: Dollars - 0.25*Francs;
Writep(TotalDollars,Dollars,Francs);
Write('The model is infeasible\n');
end

4.91. Momiss River Pollution (winstP3-4-1)

Problem: “There are three factories on the Morriss River (1, 2, and 3). Each
emits two types of pollutants and 2) into the river. If the waste from each
factory is processed, the pollution in the river can be reduced. It costs S15 to
process a ton of factory 1 waste, and each ton processed reduces the amount
of pollutant 1 by 0.10 ton and the amount of pollutant 2 by 0.45 ton. It costs
$10 to process a ton of factory 2 waste, and each ton processed will reduce the
amount of pollutant 1 by 0.20 ton and the amount of pollutant 2 by 0.25 ton.
It costs S20 to process a ton of factory 3 waste, and each ton processed will
reduce the amount of pollutant 1 by 0.40 ton and the amount of pollutant 2
by 0.30 ton. The state wants to reduce the amount of pollutant 1 in the river
by at least 30 tons and the amount of pollutant 2 in the river by at least 40
tons. Formulate an LP that will minimize the cost of reducing pollution by
the desired amounts. Do you think that the LP assumptions (Proportionality,
Additivity, Divisibility, and Certainty) are reasonable for this problem?” This
probelm is from [31] , Problem 3.4-1, Page 69.
Modeling Steps

model winstP3_4_1 "Momiss River Pollution";
set
factory :=[ 1..3];
pollutant :=[ 1..2];
parameter
ProcessCost{factory} := [15.00, 10.00,
20.00];
ReduceRate{factory,pollutant} := [0.10, 0.45, 0.20,
0.25, 0.40, 0.30];
MinReduce{pollutant} := [30, 40];
variable ProcessWaste,W{factory};
constraint RequiredReduce{pollutant}: sum{factory}
ReduceRate*W >= MinReduce;
minimize TotalCost: sum{factory} ProcessCost*W;
Writep(TotalCost);
end

4.92. US Lab (winstP3-4-2)

Problem: “U.S. Labs manufactures mechanical heart valves from the heart
valves of pigs. Different heart operations require valves of different sizes.
U.S. Labs purchases pig valves from three different suppliers. The cost and
size mix of the valves purchased from each supplier are given in Table 4.22.
Each month, U.S. Labs places one order with each supplier. At least 590Jarge,
300 medium, and 300 small valves must be purchased each month. Because
of limited availability of pig valves, at most 500 valves per month can be pur-
chased from each supplier. Formulate an LP that can be used to minimize
the cost of acquiring the needed valves.” This problem is from [31] , Problem
3.4-2, Page 69.
Cost Percent Percent Percent

Per Value Large Medium Small
Supplier 1 $5 40 40 20
Supplier 2 $4 30 35 35
Supplier 3 $3 20 20 60
Table 4.4:
Modeling Steps

model winstP3_4_2 "US Lab";
set supplier :=[ 1..3];
valve := [Large, Medium, Small];
parameter CostValve{supplier} := [5.00, 4.00, 3.00];
PurchaseMix{supplier, valve} := [0.40, 0.40, 0.20,
0.30, 0.35, 0.35, 0.20, 0.20, 0.60];
--MaxPurchase{supplier} := [500, 500, 500]; --not
feasibe with this data
MaxPurchase{supplier} := [600, 600, 600];
ReqPurchase{valve} := [500, 300, 300];
variable Purchase,P{supplier} [0..MaxPurchase];
constraint
RequiredPurchase{valve}: sum{supplier} PurchaseMix*P
>= ReqPurchase;
minimize TotalCost: sum{supplier,valve} CostValve*P;
Writep(TotalCost);
end

4.93. Diet Problem (winstP3-4-3)

Problem: “Peg and Al Fundy have a limited food budget, so eg is trying to
feed the family as cheaply as possible. However, she still wants to make sure
her family members meet their daily nutritional requirements. Peg can buy
two foods. Food 1 sells for $7 per pound, and each pound contains 3 units of
vitamin A and 1 unit of vitamin C. Food 2 sells for $1 per pound, and each
pound contains 1 unit of each vitamin. Each day, the family needs at least 12
units of vitamin A and 6 units of vitamin C.
a) Verify that Peg should purchase 12 units of food 2 each day and thus over-
satisfy the vitamin C requirement by 6 units.
b) Al has put his foot down and demanded that Peg fulfill the family’s daily
nutritional requirement exactly by obtaining precisely 12 units of vitamin A
and 6 units of vitamin C. The optimal solution to the new problem will in-
volve ingesting less vitamin C, but it will be more expensive. Why?”
This problem is from [31] , Problem 3.4-3, Page 69.
Modeling Steps

model winstP3_4_3 "Diet Problem";
set
f :=[ 1..2] "Food";
v := [A, C] "Vitamin";
parameter
FoodCost{f} := [7, 1];
AmountVitamin{f, v} := [3, 1, 1, 1];
MinDailyReq{v} := [12, 6];
variable Purchase{f};
constraint
Requirement{v}: sum{f} AmountVitamin*Purchase >=
MinDailyReq;
minimize TotalCost: sum{f} FoodCost*Purchase;
Writep(TotalCost);
end

4.94. Gold Mining (winstP3-4-4)

Problem: “Goldilocks needs to find at least 12 lb of gold and at least 18 lb of
silver to pay the monthly rent. There are two mines in which Goldilocks can
find gold and silver. Each day that Goldilocks spends in mine 1, she finds
2 lb of gold and 2 lb of silver. Each day that Goldilocks spends in mine 2,
she finds 1 lb of gold and 3 lb of silver. Formulate an LP to help Goldilocks
meet her requirements while spending as little time as possible in the mines.
Graphically solve the LP.” This probelm is from [31] , Problem 3.4-4, Page 69.
Modeling Steps

model winstP3_4_4 "Gold Mining";
set
mine :=[ 1..2];
metals := [gold, silver];
parameter
Discover{mine, metals} := [2, 2, 1, 3];
MetalReq{metals} := [12, 18];
variable DaysSpent{mine};
constraint Requirement{metals}: sum{mine} Discover*
DaysSpent >= MetalReq;
minimize TotalDays: sum{mine} DaysSpent;
Writep(TotalDays);
end

4.95. Capital Budget (winstP3-6-2)

Problem: “Two investments with varying cash flows (in thousands of dol-
lars) are available, as shown in Table 4.22. At time 0, $10,000 is available for
investment, and at time 1, $7000 is available. Assuming that r = 0.10, set up
an LP whose so-lution maximizes the NPV obtained from these investments.
Graphically find the optimal solution to the LP. (Assume that any fraction of
an investment may be purchased).” This probelm is from [31] , Problem 3.6-2,
Page 76.
Cash Flow (in thousands) at Time

0 1 2 3
Investment 1 -$6 -$5 $7 $9
Investment 2 -$8 -$3 $9 $7
Table 4.5:
Modeling Steps

model winstP3_6_2 "Capital Budget";
set
time := [0..3];
invest := [1..2];
parameter
CashAvail{time} := [10, 7, 0, 0];
CashFlow{invest,time} := [-6 -5 7 9 -8 -3 9 7];
InterestRate := 0.1;
RateMult := 1.0 + InterestRate;
NPV{invest} := sum{time} CashFlow/RateMult^time;
variable MakeInvest{invest} [0..1];
constraint InvestLimit{time}: sum{invest} - CashFlow*
MakeInvest <= CashAvail;
maximize TotalProfit: sum{invest} NPV*MakeInvest ;
Writep(TotalProfit,MakeInvest);
end

4.96. Blending Candy (winstP3-8-1)

Problem: “You have decided to enter the candy business. You are considering
producing two types of candies: Slugger Candy and Easy Out Candy, both of
which consist solely of sugar, nuts, and chocolate. At present, you have in
stock 100 oz of sugar, 20 oz of nuts, and 30 oz of chocolate. The mixture
used to make Easy Out Candy must contain at least 20mixture used to make
Slugger Candy must contain at least 10can be sold for 250, and each ounce
of Slugger Candy for 200. Formulate an LP that will enable you to maximize
your revenue from candy sales.” This probelm is from [31] , Problem 3.8-1,
Page 87.
Modeling Steps

model winstP3_8_1 "Blending Candy";
set
candy := [Slugger, EasyOut];
ingred := [Sugar, Nuts, Chocolate];
parameter
Price{candy} := [0.20, 0.25];
AvailIngred{ingred} := [100, 20, 30];
MinPercent{candy,ingred}:=[0, 0.1, 0.1, 0, 0.2, 0];
variable SweetMix{candy,ingred};
constraint
MinIngred{candy,ingred|MinPercent>0}: SweetMix >=
MinPercent*(sum{ingred} SweetMix);
MaxIngread{ingred}: sum{candy} SweetMix <=
AvailIngred;
maximize TotalProfit: sum{candy,ingred} Price*SweetMix;
end

4.97. Blending Oranges (winstP3-8-2)

Problem: “O.J. Juice Company sells bags of oranges and cartons of orange
juice. O.J. grades oranges on a scale of 1 (poor) to 10 (excellent). At present,
Q.J. has on hand 100,000 lb of grade 9 oranges and 120,000 lb of grade 6 or-
anges. The average quality of oranges sold in bags must be at least 7, and the
average quality of the oranges used to produce orange juice must be at least
8. Each pound of oranges that is used for juice yields a revenue of $1.50 and
incurs a variable cost (consisting of labor costs, variable overhead costs, in-
ventory costs, and so on) of $1.05. Each pound of oranges sold in bags yields
a revenue of 50 cents and incurs a variable cost of 200. Formulate an LP to
help O.J. maximize profit.” This probelm is from [31] , Problem 3.8-2, Page 87.
Modeling Steps

model winstP3_8_2 "Blending Oranges";
set
grade := [g6, g9];
product := [Juice, Bags];
parameter
GradeValue{grade} := [6, 9];
OrangesOnHand{grade} := [100000, 120000];
MinAvgQuality{product} := [8, 7];
Revenue{product} := [1.50, 0.50];
VarCost{product} := [1.05, 0.20];
Profit{product} := Revenue - VarCost;
variable Produce{grade,product};
constraint
LimitOranges{grade}: sum{product} Produce <=
OrangesOnHand;
ReqQuality{product}: sum{grade} GradeValue*Produce >=
MinAvgQuality*(sum{grade} Produce);
maximize TotalProfit: sum{grade,product} Profit*Produce
;
end

4.98. Blending Portfolio (winstP3-8-3)

Problem: “A bank is attempting to determine where its assets should be in-
vested during the current year. At present, $500,000 is available for invest-
ment in bonds, home loans, auto loans, and personal loans. The annual rate
of return on each type of investment is known to be: bonds, 10auto loans,
13bank’s portfolio is not too risky, the bank’s investment manager has placed
the following three restrictions on the bank’s portfolio:
a) The amount invested in personal loans cannot exceed the amount invested
in bonds.
b) The amount invested in home loans cannot exceed the amount invested in
auto loans.
c) No more than 25be in personal loans.
The bank’s objective is to maximize the annual return on its investment port-
folio. Formulate an LP that will enable the bank to meet this goal.” This
problem is from [31] , Problem 3.8-3, Page 87.
Modeling Steps

model winstP3_8_3 "Blending Portfolio";
set portfolio := [Bonds, Home, Auto, Personal];
parameter
CashAvail := 500000;
ReturnRate{portfolio} := [0.10, 0.16, 0.13, 0.20];
PercentPersonal := 0.25;
variable Invest{portfolio};
constraint
MaxInvestment: sum{portfolio} Invest <= CashAvail;
PersonalLomit: Invest['Personal'] <= Invest['Bonds'];
HomeLimit: Invest['Home'] <= Invest['Auto'];
MaxPersonal: Invest['Personal'] <= PercentPersonal*(
sum{portfolio} Invest);
maximize TotalProfit: sum{portfolio} ReturnRate*Invest;
end

4.99. Blending Investments (winstP3-8-4)

Problem: “Young MBA Erica Cudahy may invest up to $1000. She can invest
her money in stocks and loans. Each dollar invested in stocks yields 10 cents
profit, and each dollar invested in a loan yields 15 cents profit. At least 30%
of all money invested must be in stocks, and at least $400 must be in loans.
Formulate any that can be used to maximize total profit earned from Erica’s
investment. Then graphically solve the LP.” This probelm is from [31] , Prob-
lem 3.8-4, Page 87.
Modeling Steps

model winstP3_8_4 "Blending Investments";
set invest := [Stocks, Loans];
parameter
MaxAmountAvail := 1000;
MinStockPercent := 0.30;
MinLoanAmount := 400.00;
ProfitPerDollar{invest} := [0.10, 0.15];
variable Amount{invest};
constraint
TotalInvestment: sum{invest} Amount<=MaxAmountAvail;
MinStocks: Amount['Stocks'] >= MinStockPercent*(sum{
invest} Amount);
MinLoans: Amount['Loans'] >= MinLoanAmount;
maximize TotalProfit: sum{invest} ProfitPerDollar*
Amount;
end

4.100. Blending Oils (winstP3-8-5)

Problem: “Chandler Oil Company has 5000 barrels of oil 1 and 3.8 Blending
Problems 87 10,000 barrels of oil 2. The company sells two products: gasoline
and heating oil. Both products are produced by combining oil 1 and oil 2. The
quality level of each oil is as follows: oil 1, 10; oil 2, 5. Gasoline must have
an average quality level of at least 8 and heating oil, at least 6. Demand for
each product must be created by advertising. Each dollar spent advertising
gasoline creates 5 barrels of demand and each spent on heating oil creates
10 barrels of demand. Gaso- line is sold for $25 per barrel, heating oil for
$20. Formulate an LP to help Chandler maximize profit. Assume that no oil
of either type can be purchased.” This probelm is from [31] , Problem 3.8-5,
Page 87.
Modeling Steps

model winstP3_8_5 "Blending Oils";
set
oiltype,o := [Oil1, Oil2];
product,p := [Gasoline, HeatingOil];
parameter
BarrelsAvail{o} := [5000, 10000];
QualityLevel{o} := [10, 5];
QualityReq{p} := [8, 6];
Price{p} := [25.00, 20.00];
DemandPerAdDollar{p} := [5, 10];
variable
Blend,B{p,o};
AdDollars,A{p};
constraint
MinAvgQuality{p}: sum{o} QualityLevel*B >= QualityReq
*(sum{o} B);
MaxOilAvail{o}: sum{p} B <= BarrelsAvail;
AdSpending{p}: sum{o} B = DemandPerAdDollar*A;
maximize TotalProfit: sum{p,o} Price*B - sum{p} A;
end

4.101. Blending Oils (winstP3-8-5b)

Problem: “Chandler Oil Company has 5000 barrels of oil 1 and 3.8 Blending
Problems 87 10,000 barrels of oil 2. The company sells two products: gasoline
and heating oil. Both products are produced by combining oil 1 and oil 2. The
quality level of each oil is as follows: oil 1, 10; oil 2, 5. Gasoline must have
an average quality level of at least 8 and heating oil, at least 6. Demand for
each product must be created by advertising. Each dollar spent advertising
gasoline creates 5 barrels of demand and each spent on heating oil creates
10 barrels of demand. Gaso- line is sold for $25 per barrel, heating oil for
$20. Formulate an LP to help Chandler maximize profit. Assume that no oil
of either type can be purchased.” This probelm is from [31] , Problem 3.8-5,
Page 87.
Modeling Steps

model winstP3_8_5b "Blending Oils";
set
oiltype,o := [Oil1, Oil2];
product,p := [Gasoline, HeatingOil];
parameter
BarrelsAvail{o} := [5000, 10000];
QualityLevel{o} := [10, 5];
QualityReq{p} := [8, 6];
Price{p} := [25.00, 20.00];
DemandPerAdDollar{p} := [5, 10];
variable
Blend,B{p,o};
Produce,P{p};
AdDollars,A{p};
constraint
BlendOil{p}: P = sum{o} B;
MinAvgQuality{p}: sum{o} QualityLevel*B >= QualityReq
*P;
MaxOilAvail{o}: sum{p} B <= BarrelsAvail;
AdSpending{p}: P = DemandPerAdDollar*A;
maximize TotalProfit: sum{p} Price*P - sum{p} A;
end

4.102. Blending fertilizer (winstP3-8-6)

Problem: “Bullco blends silicon and nitrogen to produce two types of fertiliz-
ers. Fertilizer 1 must be at least 40sells for $70/lb. Fertilizer 2 must be at least
70sells for $40/lb. Bullco can purchase up to 80 lb of nitrogen at $15/1b and
up to 100 lb of silicon at $10/1b. Assuming that all fertilizer produced can be
sold, formulate an LP to help Bullco maximize profits. ” This probelm is from
[31] , Problem 3.8-6, Page 87.
Modeling Steps

model winstP3_8_6 "Blending fertilizer";
set
fertilizer := [f1, f2];
component := [Silicon, Nitrogen];
parameter
Profit{fertilizer} := [70.00, 40.00];
Percent{fertilizer, component} := [0, 0.40, 0.70, 0];
Cost{component} := [15.00, 10.00];
MaxPurchase{component} := [80, 100];
variable Blend{fertilizer,component};
constraint
Require{fertilizer,component|Percent>0}: Blend >=
Percent*(sum{component} Blend);
MaxComponent{component}: sum{fertilizer} Blend <=
MaxPurchase;
maximize TotalProfit: sum{fertilizer,component} (Profit
*Blend - Cost*Blend);
Writep(TotalProfit,Blend);
end

4.103. Blending chemicals (winstP3-8-7)

Problem: “Eli Daisy uses chemicals 1 and 2 to produce two drugs. Drug 1
must be at least 70at least 60at $6 per oz; up to 30 oz of drug 2 can be sold at
$5 per oz. Up to 45 oz of chemical 1 can be purchased at $6 per oz, and up to
40 oz of chemical 2 can be purchased at $4 per oz. Formulate an LP that can
be used to maximize Daisy’s profits. ” This probelm is from [31] , Problem
3.8-7, Page 87.
Modeling Steps

model winstP3_8_7 "Blending chemicals";
set
chemical,c := [c1, c2];
drug ,d := [d1, d2];
parameter
MinChemPercent{d,c} := [0.70, 0 0, 0.60];
MaxDemand{d} := [40, 30];
Price{d} := [6.00, 5.00];
MaxChemAvail{c} := [45, 40];
Cost{c} := [6.00, 4.00];
variable MixChem{c,d};
constraint
Requirement{d,c|MinChemPercent>0}: MixChem >=
MinChemPercent*(sum{c} MixChem);
LimitDemand{d}: sum{c} MixChem <= MaxDemand;
LimitChemical{c}: sum{d} MixChem <= MaxChemAvail;
maximize TotalProfit: sum{d} Price*(sum{c} MixChem) -
sum{c} Cost*(sum{d} MixChem);
end

4.104. Highland TV and Radio (winstP3-8-8)

Problem: “Highland’s TV-Radio Store must determine how many TVs and
radios to keep in stock. A TV requires 10 sq ft of floorspace, whereas a radio
requires 4 sq ft; 200 sq ft of floorspace is available. A TV will earn Highland
$60 in profits, and a radio will earn $20. The store stocks only TVs and radios.
Marketing requirements dictate that at least 60of all appliances in stock be
radios. Finally, a TV ties up $200 in capital, and a radio, $50. Highland wants
to have at most $3000 worth of capital tied up at any time. Formulate an LP
that can be used to maximize Highland’s profit.” This probelm is from [31] ,
Modeling Steps

model winstP3_8_8 "Highland TV and Radio";
set p := [TV, Radio] "products";
parameter
ReqSpace{p} := [10, 4];
SpaceAvail := 200;
Profit{p} := [60.00, 20.00];
MinStockRadioPct := 0.60;
CapitalTied{p} := [200.00, 50.00];
CapitalAvail := 3000;
variable InStock{p};
constraint
LimitSpace: sum{p} ReqSpace*InStock <= SpaceAvail;
MinimumStock{p|p='Radio'}: InStock >=
MinStockRadioPct*(sum{p} InStock);
LimitCapital: sum{p} CapitalTied*InStock <=
CapitalAvail;
maximize TotalProfit: sum{p} Profit*InStock ;
end

4.105. Production Process (Sunco Oil) (winstP3-9-1)

Problem: “Sunco Oil has three different processes that can be used to manu-
facture various types of gasoline. Each process in-volves blending oils in the
company’s catalytic cracker. Run-ning process 1 for an hour costs $5 and re-
quires 2 barrels of crude oil 1 and 3 barrels of crude oil 2. The output from
running process 1 for an hour is 2 barrels of gas 1 and 1 barrel of gas 2. Run-
ning process 2 for an hour costs $4 and requires 1 barrel of crude 1 and 3
barrels of crude 2. The output from running process 2 for an hour is 3 bar-
rels of gas 2. Running process 3 for all hour costs $1 and requires 2 barrels
of crude 2 and 3 barrels of gas 2. The output from running process 3 for an
hour is 2 barrels of gas 3. Each week , 200 barrels of crude 1, at $2/barrel,
and 300 barrels of crude 2, at $3/bar-rel, may be purchased. All gas produced
can be sold at the following per-barrel prices: gas 1, $9; gas 2, $10; gas 3, $24.
Formulate an LP whose solution will maximize revenues less costs. Assume
that only 100 hours of time on the catalytic cracker are available each week.”
This probelm is from [31] , Problem 3.9-1, Page 93.
Modeling Steps

model winstP3_9_1 "Production Process (Sunco Oil)";
set process := [1..3];
crude := [1..2]; gas := [1..3];
parameter RunCost{process} := [5, 4, 1];
CrudeReq{process,crude}:= [2, 3, 1, 3, 0, 2];
GasReq{process, gas} := [0 0 0 0 0 0 0 3 0];
OutputGas{process,gas} := [2 1 0 0 3 0 0 0 2];
MaxPurchase{crude} := [200, 300];
CrudeCost{crude} := [2.00, 3.00];
GasPrice{gas} := [9.00, 10.00, 24.00];
MaxCrackerHours := 100;
variable
SalesGas{gas}; RunProcess{process};
PurchaseCrude{crude};
constraint
CrackerLimit: sum{process} RunProcess <=
MaxCrackerHours;
CrudeBalance{crude}: PurchaseCrude = sum{process}
CrudeReq*RunProcess;
PurchaseLimit{crude}: PurchaseCrude <= MaxPurchase;
GasBalance{gas}: sum{process} OutputGas*RunProcess
= sum{process} GasReq*RunProcess + SalesGas;
maximize TotalProfit: sum{gas} GasPrice*SalesGas
- sum{process} RunCost*RunProcess
- sum{crude} CrudeCost*PurchaseCrude;
end

4.106. Production Process (winstP3-9-2)

Problem: “Furnco manufactures tables and chairs. A table requires 40 board
ft of wood, and a chair requires 30 board ft of wood. Wood may be purchased
at a cost of $1 per board ft, and 40,000 board ft of wood are available for pur-
chase. It takes 2 hours of skilled labor to manufacture an unfinished table or
an unfinished chair. Three more hours of skilled labor will turn an unfinished
table into a finished table, and 2 more hours of skilled labor will turn an un-
finished chair into a finished chair. A total of 6000 hours of skilled labor are
available (and have already been paid for). All furniture produced can be sold
at the following unit prices: unfinished table, $70; finished table, $140; unfin-
ished chair, $60; finished chair, $110. Formulate an LP that will maximize the
contribution to profit from manufacturing tables and chairs. ” This probelm
is from [31] , Problem 3.9-2, Page 93.
Modeling Steps

model winstP3_9_2 "Production Process";
set
f := [Desks, Chairs] "furniture";
s := [Unfinished, Finished] "status";
parameter
WoodUse{f} := [40, 30];
WoodCost := 1;
WoodAvail := 40000;
LaborNeeded{f,s} := [2, 5, 2, 4];
HoursAvail := 6000;
Price{f,s} := [70, 140, 60, 110];
variable
Produce{f,s};
WoodPurchase [0..WoodAvail];
constraint
WoodLimit: sum{f, s} WoodUse*Produce <= WoodPurchase;
LaborLimit: sum{f, s} LaborNeeded*Produce <=
HoursAvail;
maximize TotalProfit: sum{f,s} Price*Produce
- WoodCost*WoodPurchase;
end

4.107. Production Process (Rylon Corp.)

(winstP3-9-3)
Problem: “Rylon Corporation manufactures Brute and Chanelle perfumes.
The raw material needed to manufacture each type of perfume can be pur-
chased for $3 per pound. Processing 1 lb of raw material requires 1 hour of
laboratory time. Each pound of processed raw material yields 3 oz of Regu-
lar Brute Perfume and 4 oz of Regular Chanelle Perfume. Regular Brute can
be sold for $7/oz and Regular Chanelle for $6/oz. Rylon also has the option
of further processing Regular Brute and Regular Chanelle to produce Luxury
Brute, sold at $18/oz, and Luxury Chanelle, sold at $14/oz. Each ounce of
Regular Brute processed further requires an additional 3 hours of laboratory
time and $4 processing cost and yields 1 oz of Luxury Brute. Each ounce of
Regular Chanelle processed further requires an additional 2 hours of labora-
tory time and $4 processing cost and yields 1 oz of Luxury Chanelle. Each
year, Rylon has 6000 hours of laboratory time available and can purchase up
to 4000 lb of raw material. Formulate an LP that can be used to determine how
Rylon can maximize profits. Assume that the cost of the laboratory hours is a
fixed cost.
Suppose that in Example (3.9-11), 1 lb of raw material could be used to pro-
duce either 3 oz of Brute or 4 oz of Chanelle. How would this change the
formulation? ” This probelm is from [31] , Problem 3.9-3, Page 93.
Modeling Steps

model winstP3_9_3 "Production Process (Rylon Corp.)";
set
p := [Brute, Chanele] "perfumes";
t := [Regular, Luxury] "types";
parameter
RawMatCost := 3.00; --$/lb
RawMatYield{p} := [3, 4]; --oz/lb
LabTimeRaw := 1; --Hours/lb
LabTimeLux{p} := [3, 2]; --Hours/oz
Price{p, t} := [7, 18,
6, 14]; --$/oz
ProcessCost{p} := [4, 4]; --$/oz
LabTimeAvail := 6000;
RawMatAvail := 4000;
variable
Sales{p, t}; --oz
RawMatPurchased{p}; --lb
expression
TotalRevenue : sum{p,t} Price*Sales;

TotalProcessCost: sum{p,t|t='Luxury'} ProcessCost*
Sales;
TotalRawMatCost: RawMatCost*(sum{p} RawMatPurchased);
constraint
RawMatLimit: sum{p} RawMatPurchased <= RawMatAvail;
LabTimeLimit: LabTimeRaw*(sum{p} RawMatPurchased)
+ sum{p,t|t='Luxury'} LabTimeLux*Sales <=
LabTimeAvail;
MaterialBalance{p}: sum{t} Sales = RawMatYield*
RawMatPurchased;
maximize TotalProfit: TotalRevenue - TotalProcessCost -
TotalRawMatCost;
end

4.108. Production Process (winstP3-9-5)

Problem: “A company produces A, B, and C and can sell these products in
unlimited quantities at the following unit prices: A, $10; B, $56; C, $100. Pro-
ducing a unit of A requires 1 hour of labor; a unit of B, 2 hours of labor plus
2 units of A; and a unit of C, 3 hours of labor plus 1 unit of B. Any A that is
used to produce B cannot be sold. Similarly, any B that is used to produce C
cannot be sold. A total of 40 hours of labor are available. Formulate an LP
to maximize the company’s revenues. ” This probelm is from [31] , Problem
3.9-5, Page 93.
Modeling Steps

model winstP3_9_5 "Production Process";
set product,p,p1 := [A, B, C];
parameter
Price{p} := [10, 56, 100];
ProdHours{p} := [1, 2, 3];
BillOfMat{p,p1} := /A B 2, B C 1/;
HoursAvail := 40;
variable
Produce{p};
Sales{p};
constraint
LimitProduction: sum{p} ProdHours*Produce <=
HoursAvail;
ProdBalance{p1}: Produce = sum{p} BillOfMat*Produce +
Sales;
maximize Revenues: sum{p} Price*Sales ;
Writep(Revenues);
end

4.109. Production Process (Daisy Drug)

(winstP3-9-6)
Problem: “Daisy Drugs manufactures two drugs: 1 and 2. The dru2s are pro-
duced by blending two chemicals: 1 and 2. By weigh-t, drug 1 must contain
at least 65contain at least 55drug 2 sells for $4/oz. Chemicals 1 and 2 can be
produced by one of two production processes. Running process 1 for an hour
requires 3 oz of raw material and 2 hours skilled labor and yields 3 oz of each
chemical. Running process 2 for an hour requires 2 oz of raw material and
3 hours of skilled labor and yields 3 oz of chemical 1 and 1 oz of chemical 2.
A total of 120 hours of skilled labor and 100 oz of raw material are available.
Formulate an LP that can be used to maximize Daisy’s sales revenues. ” This
probelm is from [31] , Problem 3.9-6, Page 93.
Modeling Steps

model winstP3_9_6 "Production Process (Daisy Drug)";
set
drug :=[ 1..2];
chemical :=[ 1..2];
process :=[ 1..2];
parameter
PercentReq{drug, chemical} := [0.65, 0, 0, 0.55];
Price{drug} := [6, 4];
LaborUse{process} := [2, 3];
RawMatUse{process} := [3, 2];
ChemicalYield{process, chemical} := [3, 3, 3, 1];
LaborAvail := 120;
RawMatAvail := 100;
variable
RunProcess,R{process};
Blend,B{chemical,drug};
constraint
LaborLimit: sum{process} LaborUse*R <= LaborAvail;
RawMatLimit: sum{process} RawMatUse*R <= RawMatAvail;
ChemBalance{chemical}: sum{process} ChemicalYield*R
>= sum{drug} B;
RequireChem{drug,chemical|PercentReq>0}: B >=
PercentReq*(sum{chemical} B);
maximize TotalRevenue: sum{chemical,drug} Price*B;
Writep(TotalRevenue);
end

4.110. Production Process (Lizzies Dairy)

(winstP3-9-7)
Problem: “ Lizzie’s Dairy produces cream cheese and cottage cheese. Milk
and cream are blended to produce these two products. Both high-fat and
low-fat milk can be used to produce cream cheese and cottage cheese. High-
fat milk is 60% fat; low-fit milk is 30% fat. The milk used to produce cream
cheese must average at least 50% fat and that for cottage cheese, at least 35%
fat. At least 40% (by weight) of the inputs to cream cheese and at least 20%
(by weight) of the inputs to cottage cheese must be cream. Both cottage cheese
and cream cheese are produced by putting milk and cream through the cheese
machine. It costs 40c to process 1 lb of inputs into a pound of cream cheese. It
costs 40 cents to produce 1 lb of cottage cheese, but every pound of input for
cottage cheese yields 0.9 lb of cottage cheese and 0.1 lb of waste. Cream can be
produced by evaporating high-tat and low-fatmilk. It costs 40 cents to evap-
orate I lb of high-fat milk. Each, pound of high-fat milk that is evaporated
yields 0.6 lb of cream. It costs 40 cents to evaporate 1 lb of low-fat milk. Each
pound of low-fat milk that is evaporated yields 0.3 lb of cream. Each day, up
to 3000 lb of input may be sent through the cheese machine. Each day, at least
1000 lb of cottage cheese and 1000 lb of cream cheese must be produced. Up
to 1500 lb of cream cheese and 2000 lb of cottage cheese can be sold each day.
Cottage cheese is sold for $1.20/1b and cream cheese for $1.50/1b. High-fat
milk is purchased for 80 cents/lb and low-fat milk for 40 cents/lb. The evap-
orator can process at most 2000 lb of milk daily. Formulate an LP that can he
used to maximize Lizzie ’s daily profit.” This probelm is from [31] , Problem
3.9-7, Page 93.
Modeling Steps

model winstP3_9_7 "Production Process (Lizzies Dairy)";
set
cheese := [CreamCheese, CottageCheese];
milk := [HiFatMilk, LowFatMilk];
parameter
FatContent{milk} := [0.60, 0.30];
MinFatPct{cheese} := [0.50, 0.35];
MinCreamPct{cheese} := [0.40, 0.20];
EvapYield{milk} := [0.6, 0.3];
EvapCost{milk} := [0.40, 0.40];
EvapCapacity := 2000;
ProcessYield{cheese} := [1.0, 0.9];
ProcessCost{cheese} := [0.40, 0.40];
ProcessCapacity := 3000;
MinProduce{cheese} := [1000, 1000];

MaxProduce{cheese} := [1500, 2000];
MilkCost{milk} := [0.80, 0.40];
CheesePrice{cheese} := [1.50, 1.20];
variable
ProduceCheese,P{cheese} [MinProduce..MaxProduce];
PurchaseMilk,Bm{milk};
EvaporateMilk,Em{milk};
ProcessMilk,Pm{milk,cheese};
ProcessCream,Pc{cheese};
expression
TotalRevenue : sum{cheese} CheesePrice*P;
TotalMilkCost : sum{milk} MilkCost*Bm;
TotalEvapCost : sum{milk} EvapCost*Em;
TotalProcessCost: sum{cheese} ProcessCost*P;
constraint
PurchaseMilkBal{milk}: Bm = Em + sum{cheese} Pm;
EvapBalance: sum{milk} EvapYield*Em = sum{cheese} Pc;
EvaporateLimit: sum{milk} Em <= EvapCapacity;
CreamRequirements{cheese}: Pc >= MinCreamPct*(sum{
milk} Pm + Pc);
FatRequirements{cheese}: sum{milk} FatContent*Pm >=
MinFatPct*(sum{milk} Pm);
ProcessBalance{cheese}: P = ProcessYield*(sum{milk}
Pm + Pc);
ProcessLimit: sum{cheese} (sum{milk} Pm + Pc) <=
ProcessCapacity;
maximize TotalProfit: TotalRevenue-TotalMilkCost-
TotalEvapCost-TotalProcessCost;
Writep(TotalProfit,P,Bm,Em,Pm,Pc);
end

4.111. Production Process (Lizzies Dairy)

(winstP3-9-7b)
Problem: “ Lizzie’s Dairy produces cream cheese and cottage cheese. Milk
and cream are blended to produce these two products. Both high-fat and
low-fat milk can be used to produce cream cheese and cottage cheese. High-
fat milk is 60% fat; low-fit milk is 30% fat. The milk used to produce cream
cheese must average at least 50% fat and that for cottage cheese, at least 35%
fat. At least 40% (by weight) of the inputs to cream cheese and at least 20%
(by weight) of the inputs to cottage cheese must be cream. Both cottage cheese
and cream cheese are produced by putting milk and cream through the cheese
machine. It costs 40c to process 1 lb of inputs into a pound of cream cheese. It
costs 40 cents to produce 1 lb of cottage cheese, but every pound of input for
cottage cheese yields 0.9 lb of cottage cheese and 0.1 lb of waste. Cream can be
produced by evaporating high-tat and low-fatmilk. It costs 40 cents to evap-
orate I lb of high-fat milk. Each, pound of high-fat milk that is evaporated
yields 0.6 lb of cream. It costs 40 cents to evaporate 1 lb of low-fat milk. Each
pound of low-fat milk that is evaporated yields 0.3 lb of cream. Each day, up
to 3000 lb of input may be sent through the cheese machine. Each day, at least
1000 lb of cottage cheese and 1000 lb of cream cheese must be produced. Up
to 1500 lb of cream cheese and 2000 lb of cottage cheese can be sold each day.
Cottage cheese is sold for $1.20/1b and cream cheese for $1.50/1b. High-fat
milk is purchased for 80 cents/lb and low-fat milk for 40 cents/lb. The evap-
orator can process at most 2000 lb of milk daily. Formulate an LP that can he
used to maximize Lizzie ’s daily profit.” This probelm is from [31] , Problem
3.9-7, Page 93.
Modeling Steps

model winstP3_9_7b "Production Process (Lizzies Dairy)";
set
cheese := [CreamCheese, CottageCheese];
milk := [HiFatMilk, LowFatMilk];
parameter
FatContent{milk} := [0.60, 0.30];
MinFatPct{cheese} := [0.50, 0.35];
MinCreamPct{cheese} := [0.40, 0.20];
EvapYield{milk} := [0.6, 0.3];
EvapCost{milk} := [0.40, 0.40];
EvapCapacity := 2000;
ProcessYield{cheese} := [1.0, 0.9];
ProcessCost{cheese} := [0.40, 0.40];
ProcessCapacity := 3000;
MinProduce{cheese} := [1000, 1000];

MaxProduce{cheese} := [1500, 2000];
MilkCost{milk} := [0.80, 0.40];
CheesePrice{cheese} := [1.50, 1.20];
variable
ProduceCheese,P{cheese} [MinProduce..MaxProduce];
PurchaseMilk,Bm{milk};
EvaporateMilk,Em{milk};
ProcessMilk,Pm{milk,cheese};
ProcessCream,Pc{cheese};
ProcessInput,Pi{cheese};
expression
TotalRevenue : sum{cheese} CheesePrice*P;
TotalMilkCost : sum{milk} MilkCost*Bm;
TotalEvapCost : sum{milk} EvapCost*Em;
TotalProcessCost: sum{cheese} ProcessCost*P;
constraint
PurchaseMilkBal{milk}: Bm = Em + sum{cheese} Pm;
EvapBalance: sum{milk} EvapYield*Em = sum{cheese} Pc;
EvaporateLimit: sum{milk} Em <= EvapCapacity;
CreamRequirements{cheese}: Pc >= MinCreamPct*Pi;
FatRequirements{cheese}: sum{milk} FatContent*Pm >=
MinFatPct*(sum{milk} Pm);
ProcessBalance{cheese}: Pi = sum{milk} Pm + Pc;
OutputCheese{cheese}: P = ProcessYield*Pi;
ProcessLimit: sum{cheese} Pi <= ProcessCapacity;
maximize TotalProfit: TotalRevenue-TotalMilkCost-
TotalEvapCost-TotalProcessCost;
end

4.112. Product-Mix (Bloomington Breweries)

(winstR3-01)
Problem: “Bloomington Breweries produces beer and ale. Beer sells for $5 per
barrel, and ale sells for $2 per barrel. Producing a barrel of beer requires 5 lb
of corn and 2 lb of hops. Producing a barrel of ale requires 2 lb of corn and 1
lb of hops. Sixty pounds of corn and 25 lb of hops are available. Formulate
an LP that can be used to maximize revenue. Solve the LP graphically. ” This
probelm is from [31] , Chapter 3 Review Problem 1, page 108
Modeling Steps

model winstR3_01 "Product-Mix (Bloomington Breweries)";
set
product := [Beer, Ale];
ingred := [Corn, Hops];
parameter
Price{product} := [5.00, 2.00];
IngredUsed{product, ingred} := [5, 2, 2, 1];
IngredAvail{ingred} := [60, 25];
constraint IngredReq{ingred}: sum{product} IngredUsed*
Produce <= IngredAvail;
maximize TotalProfit: sum{product} Price*Produce;
end

4.113. Product-mix (Farmer Jones Cake)

(winstR3-02)
Problem: “Farmer Jones bakes two types of cake (chocolate and vanilla) to
supplement his income. Each chocolate cake can be sold for $1, and each
vanilla cake can be sold for 500. Each chocolate cake requires 20 minutes of
baking time and uses 4 eggs. Each vanilla cake requires 40 minutes of baking
time and uses 1 egg. Eight hours of baking time and 30 eggs are available.
Formulate an LP to maximize farmer Jones’s revenue. Then graphically solve
the LP. (A fractional number of cakes is okay.) ” This probelm is from [31] ,
Chapter 3 Review Problem 2, page 108.
Modeling Steps

model winstR3_02 "Product-mix (Farmer Jones Cake)";
set cake := [Chocolate, Vanilla];
parameter
Price{cake} := [1.00, 0.50];
BakingTime{cake} := [20, 40];
EggsUsed{cake} := [4, 1];
BakingTimeAvail := 8*60;
EggsAvail := 30;
variable BakeCakes{cake};
constraint
BakingTimeLimit: sum{cake} BakingTime*BakeCakes <=
BakingTimeAvail;
EggLimit: sum{cake} EggsUsed*BakeCakes <= EggsAvail;
maximize TotalRevenue: sum{cake} Price*BakeCakes;
Writep(TotalRevenue,BakeCakes);
end

4.114. Investment (winstR3-03)

Problem: “I now have $100. The following investments are available during
the next three years:
Investment A: Every dollar invested now yields $0.10 a year from now and
$1.30 three years from now.
Investment B: Every dollar invested now yields $0.20 a year from now and
$1.10 two years from now.
Investment C: Every dollar invested a year from now yields $1.50 three years
from now.
During each year, uninvested cash can be placed in money market funds,
which yield 6may be placed in each of investments A, B, and C. Formulate
an LP to maximize my cash on hand three years from now.” This problem is
from [31] , Chapter 3 Review Problem 3, page 108.
Modeling Steps

model winstR3_03 "Investment";
set invest := [A, B, C];
year :=[ 0..3];
parameter CashAvail{year} := [100 0 0 0];
CashFlow{invest,year} := [1.00, -0.10, 0, -1.30,
1.00, -0.20, -1.10, 0,
0, 1.00, 0, -1.50];
MaxInvestment := 50;
MoneyInterestRate := 0.06;
MoneyRateMult := 1.0 + MoneyInterestRate;
variable
Investment{invest} [0..MaxInvestment];
MoneyMarket{year};
constraint
CashBalance{year}: CashAvail + MoneyRateMult*
MoneyMarket[year-1]
= sum{invest} CashFlow*Investment + MoneyMarket;
maximize CashAtEndOfYear3: MoneyMarket[#year-1];
Writep(CashAtEndOfYear3);
end

4.115. Process Oil (Sunco Oil) (winstR3-04)

Problem: “Sunco processes oil into aviation fuel and heating oil. It costs $40
to purchase each 1000 barrels of oil, which is then distilled and yields 500 bar-
rels of aviation fuel and 500 barrels of heating oil. Output from the distillation
may be sold directly or processed in the catalytic cracker. If sold after distil-
lation without further processing, aviation fuel sells for $60 per 1000 barrels,
and heating oil sells for $40 per 1000 barrels. It takes 1 hour to process 1000
barrels of aviation fuel in the catalytic cracker, and these 1000 barrels can be
sold for $130. It takes 45 minutes to process 1000 barrels of heating oil in the
cracker, and these 1000 barrels can be sold for $90. Each day, at most 20,000
barrels of oil can be purchased, and 8 hours of cracker time are available.
Formulate an LP to maximize Sunco’s profits.” This probelm is from [31] ,
Modeling Steps

model winstR3_04 "Process Oil (Sunco Oil)";
set
product,p := [AvaiationFuel, HeatingOil];
process,c := [Distill, Cracker];
parameter
RawOilCost := 40; -- $/1000 barrels
DistillYield{p} := [0.5, 0.5]; -- 1000 barrels
Price{c,p} := [ 60, 40, -- $/1000 barrels
130, 90];
CrackerTime{p} := [60, 45]; -- min/1000 barrels
CrackerTimeAvail := 8*60;
RawOilAvail := 20; -- 1000 barrels
variable
Produce{c, p};
RawOilPurchase [0..RawOilAvail];
constraint
DistillBalance{p}: DistillYield*RawOilPurchase = sum{
c} Produce;
CrackerLimit: sum{p} CrackerTime*Produce['Cracker',p]
<= CrackerTimeAvail;
maximize TotalProfit: sum{p,c} Price*Produce
- RawOilCost*RawOilPurchase;
end

4.116. Investment (Finco) (winstR3-05)

Problem: “Finco has the following investments available:
Investment A For each dollar invested at time 0, we re ceive $0.10 at time 1

and $1.30 at time 2. (Time 0 = now; time 1 = one year from now; and so on.)
Investment B For each dollar invested at time 1, we re ceive $1.60 at time 2.
Investment C For each dollar invested at time 2, we re ceive $1.20 at time 3.
At any time, leftover cash may be invested in T-bills, which pay 10be invested
in each of investments A, B, and C. Formulate an LP that can be used to max-
imize Finco’s cash on hand at time 3.”
This probelm is from [31] , Chapter 3 Review Problem 5, page 109.
Modeling Steps

model winstR3_05 "Investment (Finco)";
set invest := [A, B, C];
year := [0..3];
parameter CashAvail{year} := [100, 0, 0, 0];
CashFlow{invest,year} := [1.00, -0.10, -1.30, 0,
0, 1.00, -1.60, 0,
0, 0, 1.00, -1.20];
MaxInvestment := 50;
TbillInterestRate := 0.10;
TbillRateMult := 1.0 + TbillInterestRate;
variable
Investment{invest} [0..MaxInvestment];
Tbill{year|year<#year-1};
constraint
CashBalance{year|year<#year-1}: CashAvail +
TbillRateMult*Tbill[year-1]
= sum{invest} CashFlow*Investment + Tbill;
maximize CashAtEndOfYear3: TbillRateMult*Tbill[2]
- sum{invest} CashFlow[invest,3]*Investment;
Writep(CashAtEndOfYear3);
end

4.117. Blending Steel (winstR3-06)

Problem: “All steel manufactured by Steelco must meet the following re-
quirements: 3.2-3.5% carbon; 1.8-2.5% silicon; 0.9-1.2% nickel; tensile strength
of at least 45,000 pounds per square inch (psi). Steelco manufactures steel by
combining two alloys. The cost and properties of each alloy are given in
Table 4.22. Assume that the tensile strength of a mixture of the two alloys can
be determined by averaging that of the alloys that are mixed together. For
example, a one-ton mixture that is 40% alloy 1 and 60% alloy 2 has a tensile
strength of 0.4(42,000) + 0.6(50,000). Use linear programming to determine
how to minimize the cost ofproducing a ton of steel.” This probelm is from
[31] , Chapter 3 Review Problem 6, page 109.
Alloy 1 Alloy 2
Cost per ton $190 $200
Percent silicon 2% 2.5%
Percent nickel 1% 1.5%
Percent Carbon 3% 4%
Tensile strenght 42,000 psi 50,000 psi
Table 4.6:
Modeling Steps

model winstR3_06 "Blending Steel";
set
elements := [Carbon, Silicon, Nickel];
alloy :=[ 1..2];
parameter
MinReq{elements} := [0.032, 0.018, 0.009];
MaxReq{elements} := [0.035, 0.025, 0.012];
MinTensilStrength := 45000;
Cost{alloy} := [190, 200];
Properties{elements, alloy} := [0.030, 0.040, 0.020,
0.025, 0.010, 0.015];
TensilStrength{alloy} := [42000, 50000];
variable Produce{alloy};
constraint
ElementLimits{elements}: MinReq <= sum{alloy}
Properties*Produce <= MaxReq;
MinTensil: sum{alloy} TensilStrength*Produce >=

MinTensilStrength;
TotalProduce: sum{alloy} Produce = 1;
minimize TotalCost: sum{alloy} Cost*Produce;
Writep(TotalCost);
end

4.118. Production Planning (winstR3-07)

Problem: “Steelco manufactures two types of steel at three different steel
mills. During a given month, each steel mill has 200 hours of blast furnace
time available. Because of differences in the furnaces at each mill, the time
and cost to produce a ton of steel differs for each mill. The time and cost for
each mill are shown in Table 4.22. Each month, Steelco must manufacture
at least 500 tons of steel 1 and 600 tons of steel 2. Formulate an LP to mini-
mize the cost of manufacturing the desired steel.” This probelem is from [31]
, Chapter 3 Review Problem 7, page 109.
Steel 1 Steel 2
Time Time
Cost (minutes) Cost (minutes)
Mill 1 $10 20 $11 22
Mill 2 $12 24 $9 18
Mill 3 $14 28 $10 30
Table 4.7:
Modeling Steps

model winstR3_07 "Production Planning";
set mill := [1..3];
steel := [1..2];
parameter
TimeAvail{mill} :=[200, 200, 200]; -- Hours/Month
ProdCost{mill, steel}:=[10 11 12 9 14 10]; -- $/Ton
ProdTime{mill,steel}:=[20 22 24 18 28 30]; -- Minutes
/Ton
MinutesPerHour := 60; -- Minutes/Hour
MinProd{steel} := [500, 600]; -- Tons/Month
variable Produce{mill,steel}; -- Tons/Month
constraint
TimeLimit{mill}: sum{steel} ProdTime*Produce <=
TimeAvail*MinutesPerHour;
MinProduction{steel}: sum{mill} Produce >= MinProd;
minimize TotalCost: sum{mill, steel} ProdCost*Produce;
Writep(TotalCost);
end

4.119. Production Planning with Diet (winstR3-08)

Problem: “Walnut Orchard has two farms that grow wheat and corn. Because
of differing soil conditions, there are differences in the yields and costs of
growing crops on the two farms. The yields and costs are shown in Table
4.22. Each farm has 100 acres available for cultivation; 11,000 bushels of wheat
and 7000 bushels of corn must be grown. Determine a planting plan that will
minimize the cost of meeting these demands. How could an extension of this
model be used to allocate crop production efficiently throughout a nation?”
This probelm is from [31] , Chapter 3 Review Problem 8, page 109.
Farm 1 Farm 2
Corn yield/acre 500 bushels 650 bushels
Cost/acre of corn $100 $120
Wheat yield/acre 400 bushels 350 bushels
Cost/acre of wheat $90 $80
Table 4.8:
Modeling Steps

model winstR3_08 "Production Planning with Diet";
set
farm :=[ 1..2];
crop := [Wheat, Corn];
parameter
CropCost{farm,crop} := [90, 100, 80, 120]; -- $/acre
CropYield{farm,crop}:= [400, 500, 350, 650]; --
bushels/acre
Demand{crop} := [11000, 7000]; -- bushels
LandAvail{farm} := [100, 100]; -- acres
variable Grow{farm,crop}; -- acres
constraint
MinCrop{crop}: sum{farm} CropYield*Grow >= Demand;
LandCapacity{farm}: sum{crop} Grow <= LandAvail;
minimize TotalCost: sum{farm, crop} CropCost*Grow;
Writep(TotalCost);
end

4.120. Process Scheduling (winstR3-09)

Problem: “Candy Kane Cosmetics (CKC) produces Leslie Per- fume, which
requires chemicals and labor. Two production processes are available: Process
1 transforms 1 unit of labor and 2 units of chemicals into 3 oz of perfume.
Process 2 transforms 2 units of labor and 3 units of chemicals into 5 oz of
perfume. It costs CKC $3 to purchase a unit of labor and $2 to purchase a
unit of chemicals. Each year, up to 20,000 units of labor and 35,000 units of
chemicals can be purchased. In the absence of advertising, CKC believes it
can sell 1000 oz of perfume. To stimulate demand for Leslie, CKC can hire the
lovely model Jenny Nelson. Jenny is paid $100/hour. Each hour Jenny works
for the company is estimated to increase the demand for Leslie Perfume by
200 oz. Each ounce of Leslie Perfume sells for $5. Use linear programming
to determine how CKC can maximize profits.” This probelm is from [31] ,
Modeling Steps

model winstR3_09 "Process Scheduling";
set process := [1..2];
parameter
LaborUnits{process} := [1, 2];
ChemicalUnits{process} := [2, 3];
PerfumeYield{process} := [3, 5];
LaborCost := 3.00;
ChemicalCost := 2.00;
LaborAvail := 20000;
ChemicalAvail := 35000;
BasicDemand := 1000;
Cost := 100;
DemandPerHour := 200;
PricePerfume := 5;
variable
ProcessUnits{process};
Hours;
expression
TotalRevenue : sum{process} PricePerfume*
PerfumeYield*ProcessUnits ;
TotalLaborCost : sum{process} LaborCost*LaborUnits*
ProcessUnits ;
TotalChemicalCost: sum{process} ChemicalCost*
ChemicalUnits*ProcessUnits ;
TotalCost : Cost*Hours;
TotalCost1 : TotalLaborCost + TotalChemicalCost

+ TotalCost;
constraint
LaborLimit: sum{process} LaborUnits*ProcessUnits <=
LaborAvail;
ChemicalLimit: sum{process} ChemicalUnits*
ProcessUnits <= ChemicalAvail;
DemandLimit: sum{process} PerfumeYield*ProcessUnits
<= BasicDemand + DemandPerHour*Hours;
maximize TotalProfit: TotalRevenue - TotalCost1;
end

4.121. Product-mix (Carco Advertising) (winstR3-10)
Problem: “Carco has a $150,000 advertising budget. To increase automobile

sales, the firm is considering advertising in news-papers and on television.
The more Carco uses a particular medium, the less effective is each additional
ad. Table 44 shows the number of new customers reached by each ad. Each
newspaper ad costs $1000, and each television ad costs $10,000. At most, 30
newspaper ads and 15 television ads can be placed. How can Carco maximize
the number of new customers created by advertising?” This problem is from
Number New
of Ads Customers
Newspaper 1-10 900
11-20 600
21-30 300
Television 1-5 10,000
6-10 5000
11-15 2000
Table 4.9:
Modeling Steps

model winstR3_10 "Product-mix (Carco Advertising)";
set media := ['NP1-10', 'NP11-20', 'NP21-30', 'TV1-5',
'TV6-10', 'TV11-15'];
parameter
AdvertBudget := 150000;
AdvertCost{media}:=[1000 1000 1000 10000 10000
10000];
MaxPlacement{media}:= [10, 10, 10, 5, 5, 5];
NewCustomers{media}:=[900 600 300 10000 5000 2000];
variable PlaceAds{media} [0..MaxPlacement];
constraint MaxBudget: sum{media} AdvertCost*PlaceAds <=
AdvertBudget;
maximize TotalCust: sum{media} NewCustomers*PlaceAds;
Writep(TotalCust);
end

4.122. Process Oil (winstR3-11)

Problem: “Sunco Oil has refineries in Los Angeles and Chicago. The Los
Angeles refinery can refine up to 2 million barrels of oil per year, and the
Chicago refinery up to 3 million. Once refined, oil is shipped to two distri-
bution points: Houston and New York City. Sunco estimates that each distri-
bution point can sell up to 5 million barrels per year. Because of dif-ferences
in shipping and refining costs, the profit earned (in dollars) per million bar-
rels of oil shipped depends on where the oil was refined and on the point of
distribution (see Table 4.22). Sunco is considering expanding the capacity of
each re-finery. Each million barrels of annual refining capacity that is added
will cost $120,000 for the Los Angeles refinery and $150,000 for the Chicago
refinery. Use linear program-ming to determine how Sunco can maximize its
profits less expansion costs over a ten-year period.” This probelm is from [31]
Profitt Per
Million Barrels
To Houston To New York
From Los Angeles $20,000 $15,000
From Chicago $18,000 $17,000
Table 4.10:
Modeling Steps

model winstR3_11 "Process Oil";
set refinery := [LosAngeles, Chicago];
distrpoint := [Houston, NewYork];
parameter
RefineCap{refinery} := [2, 3]; --million barrels
MaxSales{distrpoint} := [5, 5]; --million barrels
Profit{refinery,distrpoint}:=[20 15 18 17]; -- $1000/
million barrels
AddCapacityCost{refinery} := [120, 150]; -- $1000/
million barrels
PlanningYears := 10;
variable
Refine{refinery,distrpoint}; -- million barrels/
year
AddCapacity{refinery}; -- million barrels
expression
TotalProfitPerYear: sum{refinery, distrpoint} Profit*
Refine;
TotalAddCapCost : sum{refinery} AddCapacityCost*
AddCapacity;
constraint
RefineryLimit{refinery}: sum{distrpoint} Refine <=
RefineCap + AddCapacity;
MaxDemand{distrpoint}: sum{refinery} Refine <=
MaxSales;
maximize TotalProfit: PlanningYears*TotalProfitPerYear
- TotalAddCapCost;
end

4.123. Telephone Survey (winstR3-12)

Problem: “For a telephone survey, a marketing research group needs to con-
tact at least 150 wives, 120 husbands, 100 single adult males, and 110 single
adult females. It costs $2 to make a daytime call and (because of higher labor
costs) $5 to make an evening call. Table 4.22 lists the results. Because of lim-
ited staff, at most half of all phone calls can be evening calls. Formulate an
LP to minimize the cost of completing the survey.” This probelm is from [31]
Percent of Percent of
Person Daytime Evening
Responding Calls Calls
Wife 30 30
Husband 10 30
Single male 10 15
Single female 10 20
None 40 5
Table 4.11:
Modeling Steps

model winstR3_12 "Telephone Survey";
set time := [Daytime, Evening];
people := [Wife, Husband, SingleMale, SingleFemale];
parameter
RespondPercent{people,time}:=[0.3 0.3 0.1 0.3
0.1 .15 0.1 0.2];
MustReach{people} := [150, 120, 100, 110];
CallCost{time} := [2, 5];
variable Calls{time};
constraint
ContactReq{people}: sum{time} RespondPercent*Calls >=
MustReach;
StaffLimit: Calls['Evening']<=1/2*(sum{time} Calls);
minimize TotalCost: sum{time} CallCost*Calls;
Writep(TotalCost);
end

4.124. Feed Blending (winstR3-13)

Problem: “Feedco produces two types of cattle feed, both consist-ing totally
of wheat and alfalfa. Feed 1 must contain at least 80% wheat, and feed 2 must
contain at least 60% alfalfa. Feed 1 sells for $1.50/1b, and feed 2 sells for
$1.30/1b. Feedco can purchase up to 1000 lb of wheat at 500/1b and up to 800
lb of alfalfa at 400/lb. Demand for each type of feed is unlimited. Formulate
an LP to maximize Feedco’s profit.” This probelm is from [31] , Chapter 3
Review Problem 13, page 110.
Modeling Steps

model winstR3_13 "Feed Blending";
set
feed := [f1, f2];
grain := [wheat, alfalfa];
parameter
MinGrainPct{feed, grain} := [0.80, 0, 0, 0.60];
Price{feed} := [1.50, 1.30];
GrainAvail{grain} := [1000, 800];
variable Produce,P{feed, grain};
constraint
MinGrain{feed,grain|MinGrainPct>0}: P >= MinGrainPct
*(sum{grain} P);
MaxGrain{grain}: sum{feed} P <= GrainAvail;
maximize TotalProfit: sum{feed,grain} Price*P - sum{
feed,grain} GrainCost*P;
end

4.125. Feed Blending (winstR3-14)

Problem: “Feedco (see Problem 13) has decided to give its customer (assume
it has only one customer) a quantity discount. If the customer purchases over
300 lb of feed 1, each pound over the first 300 lb will sell for only $1.25/1b.
Similarly, if the customer purchases more than 300 pounds of feed 2, each
pound over the first 300 lb will sell for $1.00/1b. Modify the LP of Problem
13 to account for the presence of quantity discounts. (Hint: Define variables
for the feed sold at each price.)” This probelm is from [31] , Chapter 3 Review
Problem 14, page 110.
Modeling Steps

model winstR3_14 "Feed Blending";
set
feed := [f1, f2];
grain := [Wheat, AlfAlfa];
price := [Standard, Discount];
parameter
MinGrainPct{feed, grain} := [0.80, 0, 0 , 0.60];
FeedPrice{feed,price} := [1.50, 1.25, 1.30, 1.00];
DiscountCutoff := 300;
variable
Produce,P{feed,grain};
Sales,S{feed,price};
constraint
MinGrain{feed,grain|MinGrainPct>0}: P >= MinGrainPct
*(sum{grain} P);
MaxGrain{grain}: sum{feed} P <= GrainAvail;
FeedSales{feed}: sum{grain} P >= sum{price} S;
SBound: S['Standard'] <= DiscountCutoff;
maximize TotalProfit: sum{feed,price} FeedPrice*S - sum
{feed,grain} GrainCost*P;
Writep(TotalProfit,P,S);
end

4.126. Feed Blending (winstR3-14b)

Problem: “Feedco (see Problem 13) has decided to give its customer (assume
it has only one customer) a quantity discount. If the customer purchases over
300 lb of feed 1, each pound over the first 300 lb will sell for only $1.25/1b.
Similarly, if the customer purchases more than 300 pounds of feed 2, each
pound over the first 300 lb will sell for $1.00/1b. Modify the LP of Problem
13 to account for the presence of quantity discounts. (Hint: Define variables
for the feed sold at each price.)” This probelm is from [31] , Chapter 3 Review
Modeling Steps

model winstR3_14b "Feed Blending";
set
feed := [F1, F2];
grain := [Wheat, AlfAlfa];
price := [Standard, Discount];
parameter
MinGrainPct{feed, grain} := [0.80, 0, 0 , 0.60];
FeedPrice{feed,price} := [1.50, 1.25, 1.30, 1.00];
DiscountCutoff := 300;
variable Produce,P{feed, grain, price};
constraint
MinGrain{feed,grain|MinGrainPct>0}: sum{price} P >=
MinGrainPct*(sum{grain,price} P);
MaxGrain{grain}: sum{feed,price} P <= GrainAvail;
MaxStandard{feed}: sum{grain,price|price='Standard'}
P <= DiscountCutoff;
maximize TotalProfit: sum{feed,grain,price} (FeedPrice*
P - GrainCost*P);
end

4.127. Production Planning (winstR3-15)

Problem: “Chemco produces two chemicals: A and B. These chemicals are
produced via two manufacturing processes. Process 1 requires 2 hours of
labor and 1 lb of raw material to produce 2 oz of A and 1 oz of B. Process 2
requires 3 hours of labor and 2 lb of raw material to produce 3 oz of A and 2
oz of B. Sixty hours of labor and 40 lb of raw material are available. Demand
for A is unlimited, but only 20 oz of B can be sold. A sells for $16/oz, and B
sells for $14/oz. Any B that is unsold must be disposed of at a cost of $2/oz.
Formulate an LP to maximize Chemco’s revenue less disposal costs.” This
probelm is from [31] , Chapter 3 Review Problem 15, page 110.
Modeling Steps

model winstR3_15 "Production Planning";
set
chemical := [A, B];
process := [1, 2];
parameter
LaborReq{process} := [2, 3];
RawMatReq{process} := [1, 2];
ChemYield{process,chemical} := [2, 1,
3, 2];
LaborAvail := 60;
RawMatAvail := 40;
Price{chemical} := [16.00, 14.00];
Demand{chemical} := [-1, 20];
DisposeCost{chemical} := [-1, 2.00];
variable
Produce{process};
Sales{chemical};
Dispose{chemical|chemical='B'};
constraint
LaborLimit: sum{process} LaborReq*Produce <=
LaborAvail;
RawMatLimit: sum{process} RawMatReq*Produce <=
RawMatAvail;
ProcessBalance{chemical}: sum{process} ChemYield*
Produce = Sales + Dispose;
MaxDemand: Sales['B'] <= Demand['B'];
maximize TotalProfit: sum{chemical} Price*Sales
- sum{chemical} DisposeCost*Dispose;
Writep(TotalProfit,Produce,Sales,Dispose);
end

4.128. Product Mix (winstR3-17)

Problem: “Furnco manufactures tables and chairs. Each table and chair must
be made entirely out of oak or entirely out of pine. A total of 150 board ft of
oak and 210 board ft of pine are available. A table requires either 17 board
ft of oak or 30 board ft of pine, and a chair requires either 5 board ft of oak
or 13 board ft of pine. Each table can be sold for $40, and each chair for $15.
Formulate an LP that can be used to maximize revenue.” This probelm is
from [31] , Chapter 3 Review Problem 17, page 110.
Modeling Steps

model winstR3_17 "Product Mix";
set
furniture := [desk, chair];
materials := [oak, pine];
parameter
MatAvail{materials} := [150, 210];
MatUsed{furniture, materials} := [17, 30, 5, 13];
Price{furniture} := [40.00, 15.00];
variable Produce{furniture};
constraint MatReq{materials}: sum{furniture} MatUsed*
Produce <= MatAvail;
maximize TotalProfit: sum{furniture} Price*Produce ;
end

4.129. Bus-ville SchoolDistricts (winstR3-18)

Problem: “The city of Busville contains three school districts. The number of.
minority and nonminority students in each district is given in Table 4.22. Of
all students, 25% (200/800) are minority students. The local court has decided
that each of the town’s two high schools (Cooley High and Walt Whitman
High) must have approximately the same percentage of minority students
(within +5%) as the entire town. The distances (in miles) between the school
districts and the high schools are given in Table 4.17. Each high school must
have an enrollment of 300-500 students. Use linear programming to deter-
mine an assignment of students to schools that minimizes the total distance
students must travel to school.” This probelm is from [31] , Chapter 3 Review
Minority Nonminority
District Students Students
1 50 200
2 50 250
3 100 150
Table 4.12:
Cooley Walt Whitman

District High High
1 1 2
2 2 1
3 1 1
Table 4.13:
Modeling Steps

model winstR3_18 "Bus-ville SchoolDistricts";
set
district :=[ 1..3];
status := [Minority, NonMinority];
schools := [CooleyHigh, WaltWitmanHigh];
parameter
StudentCount{district,status} := [ 50, 200, 50, 250,

100, 150];
Distance{district,schools} := [1, 2, 2, 1, 1, 1];
MinorityPercent := 0.25;
MinMinority := MinorityPercent - 0.05;
MaxMinority := MinorityPercent + 0.05;
MinStudents := 300;
MaxStudents := 500;
variable AssignStudents,A{district,schools,status};
constraint
StudentAssignment{district,status}: sum{schools} A =
StudentCount;
StudentLimit{schools}: MinStudents <= sum{district,
status} A <= MaxStudents;
MinorityLowerLimit{schools}: sum{district,status|
status='Minority'} A
>= MinMinority*(sum{district,status} A);
MinorityUpperLimit{schools}: sum{district,status|
status='Minority'} A
<= MaxMinority*(sum{district,status} A);
minimize TotalDistance: sum{district,schools,status}
Distance*A;
Writep(TotalDistance);
end

4.130. Brady (winstR3-19)

Problem: “Brady Corporation produces cabinets. Each week, they require
90,000 cu ft of processed lumber. They may obtain lumber in two ways. First,
they may purchase lumber from an outside supplier and then dry it at their
kiln. Sec-ond, they may chop down logs on their land, cut them into lumber
at their sawmill, and finally dry the lumber at their kiln. Brady can purchase
grade 1 or grade 2 lumber. Grade l lumber costs $3 per cu ft and when dried
yields 0.7 cu ft of useful lumber. Grade 2 lumber costs $7 per cubic foot and
when dried yields 0.9 cu ft of useful lumber. It costs the Company $3 to chop
down a log. After being cut and dried, a log yields 0.8 cu ft of lumber. Brady
incurs costs of $4 per cu ft of lumber dried. It costs $2.50 per cu ft of logs sent
through the sawmill. Each week, the sawmill can process up to 35,000 cu ft
of lumber. Each week, up to 40,000 cu ft of grade 1 lumber and up to 60,000
cu ft of grade 2 lumber can be purchased. Each week, 40 hours of time are
available for drying lumber. The time it takes to dry 1 cu ft of grade 1 lumber,
grade 2 lumber, or logs is as follows: grade 1-2 seconds; grade 2-0.8 second;
log 1.3 seconds. Formulate a.11 LP to help Brady minimize the weekly cost
of meeting the demand for processed lumber.” This probelm is from [31] ,
Modeling Steps

model winstR3_19 "Brady";
set
lumber := [outside, own, log];
grade := [g1, g2];
parameter
Price{grade} := [3.00, 7.00];
Yield{lumber} := [0.7, 0.9, 0.8];
ChopCost := 3.00;
DryCost := 4.00;
-- model not finished
Write('The model is not complete\n');
end

4.131. Canadian Parks Commission (winstR3-20)

Problem: “The Canadian Parks Commission controls two tracts of land. Tract
1 consists of 300 acres and tract 2, 100 acres. Each acre of tract 1 can be used
for spruce trees or hunting, or both. Each acre of tract 2 can be used for spruce
trees or camping, or both. The capital (in hundreds of dollars) and labor (in
man-days) required to maintain one acre of each tract, and the profit (in thou-
sands of dollars) per acre for each possible use of land are given in Table 4.22.
Capital of $150,000 and 200 man-days of labor are available. How should the
land be allocated to various uses to maximize profit received from the two
tracts?” This probelm is from [31] , Chapter 3 Review Problem 20, page 111.
Capital Labor Profit

Tract 1 Spruce 3 0.1 0.2
Tract 1 Hunting 3 0.2 0.4
Tract 1 Both 4 0.2 0.5
Tract 2 Spruce 1 0.05 0.06
Tract 2 Camping 30 5 0.09
Tract 2 Both 10 1.01 1.1
Table 4.14:
Modeling Steps

model winstR3_20 "Canadian Parks Commission";
set Land := [Tract1, Tract2];
UsedFor:=[Spruce Hunting SpruceHunt Camping
SpruceCamp];
parameter
AcresAvail{Land} := [300, 100];
/* CapitalUse{Land,UsedFor} :=
[300 300 400 0 0 100 0 0 3000 1000];
LaborUse{Land,UsedFor} :=
[.1 .2 .2 0 0 .05 0 0 5 1.01];
Profit{Land,UsedFor} :=
[200 400 500 0 0 60 0 0 90 1100]; */
CapitalUse{Land,UsedFor}:= /Tract1 Spruce 300,
Tract1 Hunting 300, Tract1 SpruceHunt 400,
Tract2 Spruce 100, Tract2 Camping 3000,
Tract2 SpruceCamp 1000/;
LaborUse{Land,UsedFor}:= /Tract1 Spruce 0.1,
Tract1 Hunting 0.2, Tract1 SpruceHunt 0.2,

Tract2 Spruce 0.05, Tract2 Camping 5.0,
Tract2 SpruceCamp 1.01/;
Profit{Land,UsedFor}:= /Tract1 Spruce 200,
Tract1 Hunting 400, Tract1 SpruceHunt 500,
Tract2 Spruce 60, Tract2 Camping 90,
Tract2 SpruceCamp 1100/;
CapitalAvail := 150000; LaborAvail := 200;
variable Allocate{Land,UsedFor};
constraint
AcresLimit{Land}: sum{UsedFor} Allocate<=AcresAvail;
CapitalLimit: sum{Land, UsedFor} CapitalUse*Allocate
<= CapitalAvail;
LaborLimit: sum{Land, UsedFor} LaborUse*Allocate <=
LaborAvail;
maximize TotalProfit:sum{Land,UsedFor} Profit*Allocate;
end

4.132. Chandler Enterprises (winstR3-21)

Problem: “Chandler Enterprises produces two competing prod-ucts: A and B.
The company wants to sell these products to two groups of customers: group
1 and group 2. The value each customer places on a unit of A and B is as
shown in Table 4.22. Each customer will buy either product A or product B,
but not both. A customer is willing to buy product A if she believes that Value
of product A - price of product A ≥ Value of product B - price of product B
and Value of product A - price of product A ≥ 0. A customer is willing to buy
product B if she believes that Value of product B - price of product B ≥ value
of product A - price of product A and Value of product B - price of product B
≥ 0.
Group 1 has 1000 members, and group 2 has 1500 members. Chandler wants
to set prices for each product that ensure that group 1 members purchase
product A and group 2 members purchase product B. Formulate an LP that
will help Chandler maximize revenues.” This problem is from [31] , Chapter
3 Review Problem 21, page 111.
Modeling Steps
Group 1 Group 2
Customer Customer
Value of A to $10 $12
Value of B to $8 $15
Table 4.15:

model winstR3_21 "Chandler Enterprises";
set product := [A, B];
custgroup := [1..2];
parameter Value{product,custgroup}:=[10 12 8 15];
Members{custgroup} := [1000, 1500];
variable Price{product};
constraint
a: Value['A','1'] - Price['A'] >= Value['B','1'] -
Price['B'];
b: Value['B','2'] - Price['B'] >= Value['A','2'] -
Price['A'];
c: Value['A','1'] - Price['A'] >= 0;
d: Value['B','2'] - Price['B'] >= 0;
maximize TotalRevenues: Members['1']*Price['A'] +

Members['2']*Price['B'];
Writep(TotalRevenues);
end

4.133. Alden Enterprices (winstR3-22)

Problem: “Alden Enterprises produces two products. Each prod-uct can be
produced on one of two machines. The length of time needed to produce each
product (in hours) on each machine is as shown in Table 4.22. Each month,
500 hours of time are available on each machine. Each month, customers are
willing to buy up to the quantities of each product at the prices given in Table
4.17. The company’s goal is to maximize the revenue obtained from selling
units during the next two months. Formulate an LP to help meet this goal.”
This problem is from [31] , Chapter 3 Review Problem 22, page 112.
Machine 1 Machine 2
Product 1 4 3
Product 2 7 4
Table 4.16:
Demand Prices
Month 1 Month 2 Month 1 Month 2
Product 1 100 190 $55 $12
Product 2 140 130 $65 $32
Table 4.17:
Modeling Steps

model winstR3_22 "Alden Enterprices";
set product := [p1,p2];
machine := [m1,m2];
month := [Jan,Feb];
parameter TimeAvail{machine} := [500, 500];
MachineTime{product,machine} := [4, 3, 7, 4];
Price{product, month} := [55 12 65 32];
Demand{product, month} := [100, 190, 140, 130];
variable Produce{product, machine, month};
constraint
TimeCapacity{machine}: sum{product,month} MachineTime
*Produce <= TimeAvail;
MaxDemand{product,month}: sum{machine} Produce <=
Demand;
maximize TotalRevenue: sum{product,machine,month} Price

*Produce;
Writep(TotalRevenue);
end

4.134. Kiriakis (winstR3-23)

Problem: “Kiriakis Electronics produces three products. Each product must
be processed on each of three types of ma-chines. When a machine is in
use, it must be manned by a worker. The time (in hours) required to process
each product on each machine and the profit associated with each product are
shown in Table 4.22. At present, five type 1 machines, three type 2 machines,
and four type 3 machines are available. The company has ten workers avail-
able and must determine how many workers to assign to each machine. The
plant is open 40 hours per week, and each worker works 35 hours per week.
Formulate an LP that will enable Kiriakis to assign workers to machines in
a way that maximizes weekly profits. (Note: A worker need not spend the
entire work week manning a single machine.)” This probelm is from [31] ,
Product 1 Product 2 Product 3

Machine 1 2 3 4
Machine 2 3 5 6
Machine 3 4 7 9
Profit $6 $8 $10
Table 4.18:
Modeling Steps

model winstR3_23 "Kiriakis";
set p := [1..3] "Products";
m := [1..3] "Machines";
w := [1..10] "Workers";
parameter ProcessTime{m,p}:=[2 3 4 3 5 6 4 7 9];
Profit{p} := [6, 8, 10];
MachineAvail{m} := [5, 3, 4];
PlantOpenHours := 40;
WorkerHours := 35;
variable
Produce,P{p};
MachineHours,M{m};
constraint
ProcessLimit{m}: sum{p} ProcessTime*P<=MachineHours;
MachineLimit{m}: M <= PlantOpenHours*MachineAvail;
WorkerLimit: sum{m} M <= 10*WorkerHours;

end

4.135. Multi-Period Planning (winstR3-40)

Problem: “Alexis Cornby makes her living buying and selling corn. On Jan-
uary 1, she has 50 tons of corn and $1000. On the first day of each month
Alexis can buy corn at the following prices per ton: January, $300; February,
$350; March, $400; April, $500. On the last day of each month Alexis can sell
corn at the following prices per ton: January, $250; February, $400; March,
$350; April, $550. Alexis stores her corn in a warehouse that can hold at most
100 tons of corn. She must be able to pay cash for all corn the time of pur-
chase. Use linear programming to determine how Alexis can maximize her
cash on hand at the end of April. ” This probelm is from [31] , Chapter 3
Modeling Steps

model winstR3_40 "Multi-Period Planning";
set m := [Jan, Feb, Mar, Apr] "Months";
parameter
StartCorn := 50; -- tons of corm
StartCash := 1000; -- dollars
PurchasePrice{m} := [300, 350, 400, 500];
SellPrice{m} := [250, 400, 350, 550];
StorageAvail := 100; -- tons of corn
variable
PurchaseCorn,B{m};
SellCorn{m} [0..StorageAvail];
StoreCorn{m};
CashOnHand,C{m};
constraint
CashBalance{m}: C = if(m=1,StartCash,C[m-1]) +
SellPrice*SellCorn - PurchasePrice*B;
PurchaseLimit{m}: PurchasePrice*B <= if(m=1,StartCash
,C[m-1]);
SellLimit{m}: SellCorn <= StoreCorn;
CornBalance{m}: StoreCorn = if(m=1,StartCorn,
StoreCorn[m-1]) + B - SellCorn[m-1];
maximize CashOnHandApril: C[#m];
Writep(CashOnHandApril,B,SellCorn,StoreCorn,C);
end

4.136. Production Planning (winstR3-40b)

Problem: “Alexis Cornby makes her living buying and selling corn. On Jan-
uary 1, she has 50 tons of corn and $1000. On the first day of each month
Alexis can buy corn at the following prices per ton: January, $300; February,
$350; March, $400; April, $500. On the last day of each month Alexis can sell
corn at the following prices per ton: January, $250; February, $400; March,
$350; April, $550. Alexis stores her corn in a warehouse that can hold at most
100 tons of corn. She must be able to pay cash for all corn the time of pur-
chase. Use linear programming to determine how Alexis can maximize her
cash on hand at the end of April. ” This probelm is from [31] , Chapter 3
Modeling Steps

model winstR3_40b "Production Planning";
set m := [Jan, Feb, Mar, Apr];
parameter
StartCorn := 50;
-- tons of corm
StartCash := 1000;
-- dollars
PurchasePrice{m} := [300, 350, 400, 500];
SellPrice{m} := [250, 400, 350, 550];
StorageAvail := 100;
-- tons of corn
variable
PurchaseCorn,B{m};
SellCorn{m};
StoreCorn{m} [0..StorageAvail];
CashOnHand,C{m};
constraint
CashBalance{m}: C = if(m=1,StartCash,C[m-1]) +
SellPrice*SellCorn - PurchasePrice*B;
PurchaseLimit{m}: PurchasePrice*B <= if(m=1,StartCash
,C[m-1]);
SellLimit{m}: SellCorn <= StoreCorn;
CornBalance{m}: StoreCorn = if(m=1,StartCorn,
StoreCorn[m-1]) + B - SellCorn[m-1];
maximize CashOnHandApril: C[#m];
Writep(CashOnHandApril,B,SellCorn,StoreCorn,C);
end

4.137. Multi-Period Finance Planning (winstR3-41)

Problem: “At the beginning of month 1, Finco has $400 in cash. At the be-
ginning of months 1, 2, 3, and 4. Finco rcceives certain revenues, after which
it pays hills (see Table 4.22). Any money left over may be invested for one
month at the interest rate of 0.1% per month; for two months at 0.5% per
month; for three months at 1 % per month; or for four months at 2% per
month. Use linear programming to determine an invest ment strategy that
maximizes cash on hand at the beinning of month 5.” This probelm is from
Revenue Bills
Month 1 $400 $600
Month 2 $800 $500
Month 3 $300 $500
Month 4 $300 $250
Table 4.19:
Modeling Steps

model winstR3_41 "Multi-Period Finance Planning";
set period ,p :=[ 1..4];
parameter
StartCash := 400;
Interest{p} := [0.001, 0.005, 0.01, 0.02];
Revenues{p} := [400, 800, 300, 300];
Bills{p} := [600, 500, 500, 250];
variable
Invest{p};
Cash{p};
constraint Balance{p}: Cash + Bills + Invest
= if(p=1,StartCash,Cash[p-1]) + Revenues + (1+
Interest)*Invest[p-1];
maximize EndCash: Cash[#p];
Writep(EndCash,Invest,Cash);
end

4.138. Transporation: Waste disposal (winstR3-42)

Problem: “City 1 produces 500 tons of waste per day, and city produces 400
tons of waste per day. Waste must be incinerated at incinerator 1 or 2, and
each incinerator can process up to 500 tons of waste per day. The cost to
incinerate waste is $40/ton at incinerator 1 and $30/ton at 2. Incineration
reduces each ton of waste to 0.2 tons of debris, which must be dumped at
one of two landfills. Each landfill can receive at most 200 tons of debris per
day. It costs $3 per mile to trans-port a ton of material (either debris or waste).
Distances (in miles) between locations are shown in Table 4.22. Formulate an
LP that can be used to minimize the total cost of disposing of the waste of
both cities.” This probelm is from [31] , Chapter 3 Review Problem 42, page
114.
Incinerator 1 Incinerator 2
City 1 30 5
City 2 36 42
Landfill 1 Landfill 2
Incinerator 1 5 8
Incinerator 2 9 6
Table 4.20:
Modeling Steps

model winstR3_42 "Transporation: Waste disposal";
set
city := [1..2];
incin := [1..2];
landfill := [1..2];
parameter
WasteProduced{city} := [500, 400];
BurnCapacity{incin} := [500, 500];
BurnCost{incin} := [40, 30];
ResultingDebris := 0.20;
LandfillCapacity{landfill} := [200, 200];
CostPerMile := 3.00;
DistToIncin{city,incin} := [30, 5, 36, 42];
DistToLandfill{incin,landfill} := [5, 8, 9, 6];
variable
WasteToIncin,W{city,incin};
DebrisToLandfill,D{incin,landfill};
expression
TotalBurnCost : sum{city,incin} BurnCost*W;
IncinTranspCost : CostPerMile*(sum{city,incin}
DistToIncin*W);
LandfillTranspCost: CostPerMile*(sum{incin,landfill}
DistToLandfill*D);
constraint
BurnWaste{city}: sum{incin} W = WasteProduced;
IncinLimit{incin}: sum{city} W <= BurnCapacity;
IncinBalance{incin}: sum{city} W = 1/ResultingDebris
*(sum{landfill} D);
LandfillLimit{landfill}: sum{incin} D <=
LandfillCapacity;
minimize TotalCost: TotalBurnCost + IncinTranspCost +
LandfillTranspCost;
Writep(TotalCost);
end

4.139. Product-mix with blending (winstR3-45)

Problem: “Turkeyco produces two types of turkey cutlets for sale to fast food
restaurants. Each type of cutlet consists of white meat and dark meat. Cutlet
1 sells for $4/lb and must consist of at least 70% white meat. Cutlet 2 sells for
$3/lb and must consist of at least 60% white meat. At most 50 lb of cutlet 1
and 30 lb of cutlet 2 can be sold. The two types of turkey used to manufacture
the cutlets are purchased from the Gob-bleGobble Turkey Farm. Each type 1
turkey costs $10 and yields 5 lb of white meat and 2 lb of dark meat. Each
type 2 turkey costs $8 and yields 3 lb of white meat and 3 lb of dark meat.
Formulate an LP to maximize Turkeyco’s profit.” This probelm is from [31] ,
Modeling Steps

model winstR3_45 "Product-mix with blending";
set
c := [1..2] "Cutlet";
m := [White, Dark] "Meat";
t := [1..2] "Turkey";
parameter
Price{c} := [4.00, 3.00];
MinWhitePercent{c} := [0.70, 0.60];
MaxDemand{c} := [50, 30];
TurkeyCost{t} := [10.00, 8.00];
TurkeyMeat{t, m} := [5, 2, 3, 3];
variable
PurchaseTurkey{t};
CutletMeat{c,m};
expression
TotalRevenue: sum{c,m} Price*CutletMeat;
TotalCost : sum{t} TurkeyCost*PurchaseTurkey;
constraint
MinWhiteMeat{c}: sum{m|m='White'} CutletMeat >=
MinWhitePercent*(sum{m} CutletMeat);
MaximumDemand{c}: sum{m} CutletMeat <= MaxDemand;
MeatLimit{m}: sum{c} CutletMeat <= sum{t} TurkeyMeat*
PurchaseTurkey;
maximize TotalProfit: TotalRevenue - TotalCost;
end

4.140. Review Prob (Priceler) (winstR3-46)

Problem: “Priceler manufactures sedans and wagons. The number of vehicles
that can be sold each of the next three months are listed in Table 4.22. Each
sedan sells for $8000, and each wagon sells for $9000. It costs $6000 to produce
a sedan and $7500 to produce a wagon. To hold a vehicle in inventory for one
month costs $150 per sedan and $200 per wagon. During each month, at
most 1500 vehicles can be produced. Production line restrictions dictate that
during month 1, at least two thirds of all cars produced must be sedans. At the
beginning of month 1, 200 sedans and 100 wagons are available. Formulate
an LP that can be used to maximize Priceler’s profit during the next three
months.” This probelm is from [31] , Chapter 3 Review Problem 46, page 116.
Sedans Wagons
Month 1 1100 600
Month 2 1500 700
Month 3 1200 500
Table 4.21:
Modeling Steps

model winstR3_46 "Review Prob (Priceler)";
set v := [Sedan, Wagon] "Vehicles";
m :=[ 1..3] "Months";
parameter Price{v} := [8000, 9000];
ProdCost{v} := [6000, 7500];
InvtCost{v} := [150, 200];
InvtCapacity := 1500;
StartInvt{v} := [200, 100];
Demand{m, v} := [1100, 600, 1500, 700, 1200, 500];
variable
Produce,P{v,m};
Inventory,I{v,m};
Sales,S{v,m} [0..Demand];
expression
TotalRevenue : sum{v,m} Price*S;
TotalProdCost: sum{v,m} ProdCost*P;
TotalInvtCost: sum{v,m} InvtCost*I;
TotalCost : TotalProdCost+TotalInvtCost;
constraint
InventoryBalance{v,m}:I=if(m=1,StartInvt,I[m-1])+P-S;
ProductionLine{m|m=1}: P['Sedan',1]>=2/3*(sum{v} P);

MaxInventory{m}: sum{v} I <= InvtCapacity;
maximize TotalProfit: TotalRevenue-TotalCost;
Writep(TotalProfit,P,I,S);
end

4.141. EJ-Korvair DeptStore (winstR3-53)

Problem: “E.J. Korvair Department Store has $1000 in available cash. At the
beginning of each of the next six months, E.J. will receive revenues and pay
bills as shown in Table 4.22. It is clear that E.J. will have a short-term cash flow
problem until the store receives revenues from the Christmas shopping sea-
son. To solve this problem, E.J. must borrow money. At the beginning of July,
E.J. may take out a six-month loan. Any money borrowed for a six-month pe-
riod must be paid back at the end of December along with 9% interest (early
payback does not reduce the interest cost of the loan). E.J. may also meet
cash needs through month-to-month bor- rowing. Any money borrowed for
a one-month period incurs an interest cost of 4% per month. Use linear pro-
gramming to determine how E.J. can minimize the cost of paying its bills on
time.” This probelm is from [31] , Chapter 3 Review Problem 53, page 117.
Revenues Bills
July $1000 $5000
August $2000 $5000
September $2000 $6000
October $4000 $2000
November $7000 $2000
December $9000 $1000
Table 4.22:
Modeling Steps

model winstR3_53 "EJ-Korvair DeptStore";
set m := [July, August, September, October, November,
December] "Months";
parameter
CashAvail := 1000;
Revenues{m} := [1000, 2000, 2000, 4000, 7000, 9000];
Bills{m} := [5000, 5000, 6000, 2000, 2000, 1000];
LoanInterest := 0.09;
BorrowInterest := 0.04;
LoanRate := 1.0 + LoanInterest;
BorrowRate := 1.0 + BorrowInterest;
variable
CashLevel,C{m};
SixMonthLoan,L;
BorrowMonth,S{m};
constraint
CashBalance{m|m=1}:
C = CashAvail + Revenues - Bills + S + L;
CashBalance1{m|m>1 and m<#m}:
C = C[m-1] + Revenues - Bills + S - BorrowRate*S[
m-1];
CashBalance2{m|m=#m}:
C = C[m-1] + Revenues - Bills - BorrowRate*S[m-1]
- LoanRate*L;
minimize TotalCost: LoanInterest*L + sum{m}
BorrowInterest*S;
Writep(TotalCost);
end

C HAPTER 5
M ODELS FROM THE GAMS

L IBRARY
5.1. Linear Quadratic Control Problem (abel)

Problem: Linear Quadratic Riccati Equations are solved as a General Nonlin-
ear Programming Problem instead of the usual Matrix Recursion.
This model is from the GAMS model library (GAMS SEQ=64). See also [18]
Modeling Steps

model ABEL "Linear Quadratic Control Problem";
set n,np "states" :=[consumpt, invest];
m,mp "controls":=['gov-expend', money];
k "horizon" :=['1964-i' '1964-ii' '1964-iii'
'1964-iv' '1965-i' '1965-ii' '1965-iii' '1965-iv'];
parameter
a{n,np} "state vector matrix" := [0.914 -0.016,
0.097 0.424];
b{n,m} "control vector matrix" := [0.305 0.424,
-0.101 1.459];
wk{n,np}"penalty matrix for states - input" :=
[0.0625 ., . 1];
lambda{m,mp} "penalty matrix for comtrols" := [1 .,
. 0.444];
c{n} "constant term" := [-59.4, -184.7 ];
xinit{n} "initial value" := [387.9, 85.3 ];
uinit{m} "initial controls" := [110.5, 147.1];
xtilde{n,k} "desired path for x" := xinit*1.0075^(k
-1);
utilde{m,k} "desired path for u" := uinit*1.0075^(k
-1);
349
CHAPTER 5. MODELS FROM THE GAMS LIBRARY 350
w{n,np,k} "penalty matrix on states" := if(k<#k,wk[

n,np],100*wk[n,np]);
variable x{n,k} "state variable" := xinit;
u{m,k} "control variable" := uinit;
constraint
stateq{n,k|k<#k} "state constraint": x[n,k+1] = sum{
np} a*x[np,k] + sum{m} b*u + c;
bound{n}: x[n,1] = xinit;
minimize criterion: 0.5*(sum{k,n,np}(x-xtilde)*w[n,np,k
]*(x[np,k]-xtilde[np,k]))
+ 0.5*(sum{k,m,mp|k<#k}(u-utilde)*lambda*(u[mp,k]-
utilde[mp,k]));
Writep(criterion,x,u);
end

351 CHAPTER 5. MODELS FROM THE GAMS LIBRARY
5.2. Agreste Farm Level Model of NE Brazil

(agreste)
Problem: This is a farm level model of the north east region of brazil. There
is only one farm type - medium sized. There are 3 types of land and risk on
revenue is considered.
This model is from teh GAMS model library (GAMS SEQ=88). See also [21]
Modeling Steps

model AGRESTE "Agreste Farm Level Model of NE Brazil";
set c "crops" := [
cotton_h banana sugar_cane beans_arr
beans_cor oranges manioc corn sisal];
tm "months" := [ jan, feb, mar, apr, may, jun, jul,
aug, sep, oct, nov, dec ];
p "cropping activities" := [
crop_02 -- cotton_herbaceo
crop_05 -- banana
crop_10 -- sugar cane
crop_15 -- oranges
crop_16 -- manioc
crop_17 -- corn
crop_19 -- sisal
crop_25 -- beans-de-corda corn
crop_29 -- cotton_herbaceo beans-de-
arranca corn
crop_30 -- cotton_herbaceo beans-de-
corda corn
crop_33 -- beans-de-arranca corn
crop_36 -- beans-de-arranca manioc corn
];
s,sp "land types" := [
good --flat and near water or humid and low lying
medium -- hilly or more arid than good
pasture -- cultivated pasture or tree crops
];
sc{s} "crop lands" := [ good , medium ];
r "livestock feeding alternatives" := [
rec_1 rec_2 rec_3 ];
ty "year" := [ 1960..1969 ];
dr "family consumption bundle alternatives
" := [ one , two , three ];
km "technology characteristics" := [
equipment, fertilizer, seeds, sprouts,

itech ];
parameter
landc{s} "land data (ha)" := / good 8.775, medium
11.64, pasture 21.92/;
rations{r} "lvstk rations feeding alternative (cr
per head)":= / rec_1 5.141 , rec_2 21.646 ,
rec_3 49.845/;
xcropl{p,s} "limits on cropping activities" := /
crop_19 good 0.036, crop_19 medium 0.298/;
ldp{s,s} "land downgrading possibilities" := /
:good medium:
good 1 .
medium -1 1
pasture . -1 /;
/* land req for lvstk seems high but ok - pg 65 para 3 */
lio{s,r} "land requirements for lvstk feed
alternatives (ha per head)" := /
:rec_1 rec_2 rec_3:
medium 1.407 0.611 0.631
pasture 0.209 2.03 0.9 /;
labor{p,tm} "labor requirements for cropping (mandays
per ha)" := /
:jan feb mar apr may jun jul
aug sep oct nov dec:
crop_02 4.79 10.89 18.70 9.83 8.03 20.95 15.67
15.48 11.35 8.38 8.23 4.54
crop_05 8.24 0.89 5.43 3.31 11.13 10.33 3.27
2.65 5.47 3.67 3.18 1.44
crop_10 5.78 3.37 4.03 7.09 4.64 5.91 7.67
9.23 5.93 9.05 7.59 6.37
crop_15 2.00 2.36 4.13 2.13 7.18 5.88 3.21
9.20 6.89 3.30 9.70 8.87
crop_16 5.71 7.60 7.28 8.15 8.48 10.71 9.34
12.76 8.93 10.22 6.18 8.34
crop_17 9.19 14.42 10.59 12.01 3.47 10.75 5.88
7.30 4.14 4.25 3.49 2.11
crop_19 1.22 1.11 0.56 0.78 1.19 0.44 1.89
3.04 2.93 4.99 7.73 9.44
crop_25 12.46 11.35 24.38 22.03 16.91 9.47 10.97
9.70 4.93 6.41 0.53 0.89
crop_29 5.54 10.30 9.24 8.97 15.82 13.00 14.74
10.64 5.64 4.90 4.73 4.92
crop_30 9.18 3.94 16.15 23.69 28.12 20.58 17.73
15.06 4.58 9.76 7.39 2.67
crop_33 6.87 10.19 10.61 9.14 18.38 11.50 9.22
13.36 3.85 3.87 2.65 1.09
crop_36 7.44 10.21 9.63 16.18 20.11 16.86 14.86
14.00 8.04 6.69 5.49 5.26 /;
llab{tm,r} "labor requirement for lvstk feed (mandays

per head)" := /
:rec_1 rec_2 rec_3:
jan 4.261 1.873 1.933
feb 2.854 1.262 1.302
mar 0.040 0.040 0.040
apr 0.040 0.040 0.040
may 0.040 0.040 0.040
jun 0.040 0.040 0.040
jul 0.040 0.040 0.040
aug 0.040 0.040 0.040
sep 0.040 0.040 0.040
oct 0.040 0.040 0.040
nov 7.075 3.095 3.195
dec 7.075 3.095 3.195/;
cbndl{c,dr} "consumption bundles (tons per bundle)" :=
/
:one two three:
beans_arr 0.225 0.152 0.15
manioc 0.965 2.64 0.935
corn 0.235 0.232 0.581 /;
crev{c,ty} "crop revenue time series (cr per ha)" := /
:1960 1961 1962 1963 1964 1965 1966 1967
1968 1969:
banana 6399 8193 9215 8581 7988 7228 5923 6738
7221 7842
sugar_cane 1667 1899 1898 2230 2265 2067 1499 1903
1901 1929
beans_cor 546 585 1091 651 470 621 605 518
519 772
manioc 986 1543 2658 1639 1146 1380 1357 1702
1560 1580
corn 267 378 544 329 320 351 310 322
294 349 /;
price{c} "crop reference prices (cr per kg)" := /
cotton_h 1.9 banana 4 sugar_cane 35
beans_arr 2.5 beans_cor 1.5 oranges 10
manioc 0.17 corn 0.15 sisal 1 /;
/* wage data: pg 159, working capital: is this correct
: =6.4*1113, family labor available: correct? */
fwage "family reservation wage rate (cr/man-month)"
:= 75;
twage "temporary labor wage rate (cr/man-month)" :=
250;
pwage "permanent labor wage rate (cr/man-year)"
:=2054;
vsc "value of self-consumption (cr/bundle)" :=
934;
wcbar "working capital (cr)"
:=7123.2;
famlab "family labor man equivalent wrks)" :=

70.5;
lprice "livestock price (cr per head)" :=
211;
vetpr "cost of vetinanry servic per head)" :=
1;
dpm "man days per man month" :=
25;
/* some of these yields look odd: medium land has
higher land than good */
yield{p,c,s} "crop yields (kg per ha)" := /
:good medium:
crop_02 cotton_h 848 569
crop_05 banana 221 174
crop_10 sugar_cane 45 30
crop_15 oranges 92 .
crop_16 manioc 4456 3964
crop_17 corn 725 563
crop_19 sisal 2244 1666
crop_25 beans_cor 251 211
crop_29 beans_arr 285 221
crop_30 beans_cor 115 352
crop_36 manioc 3408 1031
crop_36 corn 503 328 /;
set ps{p,s} "process-land possibilities" := (sum{c}
yield > 0);
parameter
ravg{c} "average crop revenues (cr per ha)" := (sum{
ty} crev)/#ty;
prdev{c,ty|ravg} "price deviations for crops (cr per
ton)" := 1000*price*(crev/ravg - 1);
phi "risk factor" := 1;
techc{p,km} "cropping technology rquirements (cr per
ha)" := /
:equipment fertilizer seeds sprouts:
crop_02 . 31 91 .
crop_05 . . . 45
crop_10 . 36 . .
crop_15 . 106 1 184
crop_16 . 20 . .
crop_17 . 42 55 .
crop_29 4 22 . 19
crop_30 . . 27 .
crop_33 11 42 . .
crop_36 . 98 6 1 /;
pcost{p} "cropping technology costs (cr per ha)" :=
sum{km} techc;
a{p} "cropping constants" := if (sum{s} yield[p
,'cotton_h',s]=0,1);
variable
xcrop{p,s} "cropping activities (ha)";
xliver{r} "livestock activity defined on feed
techniques (head)";
lswitch{s} "land downgrading (ha)";
sales{c} "crop sales (ton)";
cons{dr} "on-farm consumption (ton)";
flab{tm} [0..famlab]"family labor (man-days)";
tlab{tm} "temporary labor (man-days)";
plab "permanent labor (workers)";
pdev{ty} "positive price deviations (cr)";
ndev{ty} "negative price deviations (cr)";
xlive "livestock production (head)";
xprod{c} "crop production (ton)";
rationr "livestock ration requirements (cr)";
revenue "from crop and livestock sales (cr)";
cropcost "accounting: cropping activities cost (
cr)";
labcost "accounting: labor costs - including
family (cr)";
vetcost "accounting: vetinary services cost (cr)
";
constraint
landb{s}: sum{p|ps and sc} a*xcrop + sum{sp} ldp[sp]*
lswitch[sp] + sum{r} lio*xliver <= landc "land
balance";
lbal : xlive = sum{r} xliver "livestock balance";
rliv : rationr = sum{r} rations*xliver "livestock
ration requirements definition";
mbalc{c}: sum{s,p} yield*xcrop/1000 >= sales + sum{dr
} cbndl*cons "material balance: crops";
dprod{c}: xprod = sum{p,s} yield*xcrop/1000 "crop
production definition";
labc{tm}: sum{p,s|ps} labor*xcrop + sum{r} llab*
xliver <= flab + tlab + dpm*plab "labor supply-
demand relation";
cond : sum{dr} cons = 1 "on farm consumption
definition";
ddev{ty}: sum{c} prdev*sales = pdev-ndev "crop price
deviation definition";
arev : revenue = lprice*xlive + 1000*(sum{c} price
*sales) "accounting: revenue definition";
acrop : cropcost = sum{p,s|ps} pcost*xcrop "

accounting: cropping cost definition";
alab : labcost = (fwage*(sum{tm} flab) + twage*(
sum{tm} tlab))/dpm + pwage*plab "accounting:
labor cost definition";
awcc : cropcost + rationr + vetcost + twage/dpm*(
sum{tm} tlab) + pwage*plab <= wcbar "accounting:
working capital requirements";
avet : vetcost = vetpr*xlive "accounting:
vetinary costs";
bound{p,s|xcropl}: xcrop <= xcropl;
income frozen: revenue + vsc*(sum{dr} cons)-labcost-
rationr-vetcost
-cropcost-phi*(sum{ty} (pdev+ndev)/#ty) "farm
income definition";
maximize obj1: income;
Writep(income);
phi := 0;
maximize obj2: income;
Writep(income);
end

5.3. Ajax Paper Company Production Schedule

(ajax)
Problem: This sample model is taken from the cybernet pds/apex sample
library of models. A paper manufacturer can produce four different types
of paper on three different machines. Given a fixed demand schedule the
objective is to find a production plan that maximizes monthly profits.
This model is from the GAMS model library (GAMS SEQ=60). See also CDC,
PDS/APEX Sample Model Library, 1977. Control Data Corporation.
Modeling Steps

model AJAX "Ajax Paper Company Production Schedule";
set m "machines at mills" := [ machine_1,
machine_2, machine_3 ];
g "paper grades" := [ '20-bond-wt', '25-bond-
wt', 'c-bond-ext', 'tissue-wrp' ];
parameter prate{g,m} "production rate (tons/hour)" := [
53 52 49 , 51 49 44 , 52 45 47 , 42 44 40];
pcost{g,m} "production cost ($ per ton)" := [
76 75 73 , 82 80 78 , 96 95 92 , 72 71 70];
demand{g} := [30000 20000 12000 8000];
price{g} := [77 81 99 105];
avail{m} "available machine time (hours/month)" := [672
600 480];
variable outp{g,m} "production (tons/month)";
constraint
Cap{m}: sum{g} outp/prate <= avail;
Dem{g}: sum{m} outp = demand;
maximize profit: sum{g} demand*price - sum{g,m} pcost*
outp;
Writep(profit);
Write('machine time report\n avail-h used-h
unused-h marginal\n');
Write{m}('%10s %7.1f %6.1f %8.2f %8.2f\n', m,avail,
Cap,avail-Cap,- GetValue(Cap,3));
Write('production allocation report\n
demand sold unsold marginal\n');
Write{g}('%10s %7.1f %6.1f %8.2f %8.2f\n', g,demand
,Dem,demand-Dem, GetValue(Dem,3));
end

5.4. Sample Problem (AMPL) (ampl)

Problem: A sample problem to demonstrate the power of modeling systems
(compares GAMS with AMPL).
This model is from teh GAMS model library (GAMS SEQ=74). See also [11].
Modeling Steps

model AMPL "Sample Problem (AMPL)";
set p "products" := [ nuts, bolts, washers ];
r "raw materials" := [ iron, nickel ];
t "periods" := [ 1..5 ];
parameter
b{r} "initial stock" := /iron 35.8 , nickel 7.32/;
d{r} "storage cost" := /iron 0.03, nickel 0.025/;
f{r} "residual value" :=/iron 0.02, nickel -0.01/;
m "maximum production" := 123;
a{r,p} "raw material inputs to produce a unit of
product" := [
0.79 0.83 0.92
0.21 0.17 0.08];
c{p,t} "profit" := [
1.73 1.8 1.6 2.2 .
1.82 1.9 1.7 0.95 .
1.05 1.1 0.95 1.33 .];
variable
x{p,t|t<#t} "production level";
s{r,t} "storage at beginning of period";
constraint
limit{t} "capacity constraint": sum{p} x <= m;
balance{r,t|t>1} "raw material balance": s = s[t-1]
- sum{p} a*x[p,t-1];
sUpper{r}: s[r,1] <= b;
maximize profit: sum{p,t}c*x + sum{r,t}if(t=#t,f,-d)*s;
Writep(profit);
end

5.5. Bid Evaluation (bid)

Problem: A company obtains a number of bids from vendors for the supply
of a specified number of units of an item. Most of the submitted bids have
prices that depend on the volume of business.
This model is from the GAMS model library (GAMS SEQ=19). See also [3],
pp. 28–36.
Modeling Steps

model BID "Bid Evaluation";
set v "vendors" := [ a, b, c, d, e ];
s "segments" := [ 1..5 ];
vs{v,s} "vendor bit possibilities";
cl "column labels" :=[setup, price, q_min, q_max];
parameter req "requirements" := 239600.48;
bid{v,s,cl} "bid data";
Read{v,s,cl}('bid.txt', v,s,cl,bid);
/* get minimum domains and ripple total cost up the
segments */
{v,s}(vs:=bid[v,s,'q_max']),
for{vs[v,s]|s>1} (bid[v,s,'setup'] := bid[v,s-1,'
setup']
+ bid[v,s-1,'q_max']*(bid[v,s-1,'price']
-bid[v,s,'price']));
variable pl{v,s} "purchase level";
binary plb{v,s} "purchase decision";
constraint
demand: req = sum{vs[v,s]} pl "Demand constraint";
mpl{vs[v,s]}:bid[v,s,'q_min']*plb <= pl <= bid[v,s,'
q_max']*plb;
oneonly{v} : sum{s|vs} plb <= 1 "At most one deal";
minimize cost:
sum{vs[v,s]} (bid[v,s,'price']*pl + bid[v,s,'setup']*
plb);
Writep(cost,pl,plb);
end

5.6. Bid Evaluation (bid1)

Problem: A company obtains a number of bids from vendors for the supply
of a specified number of units of an item. Most of the submitted bids have
prices that depend on the volume of business.
This model is from the GAMS model library (GAMS SEQ=19). See also [3],
pp.28–36.
Modeling Steps

model BID "Bid Evaluation";
set v "vendors" := [ a, b, c, d, e ];
s "segments" := [ 1..5 ];
vs{v,s} "vendor bit possibilities";
cl "column labels" :=[setup, price, q_min, q_max];
parameter req "requirements" := 239600.48;
bid{v,s,cl} "bid data" := /
:setup price q_min q_max:
a 1 3855.84 61.150 . 33000
b 1 125804.84 68.099 22000 70000
b 2 . 66.049 70000 100000
b 3 . 64.099 100000 150000
b 4 . 62.119 150000 160000
c 1 13456.00 62.190 . 165600
d 1 6583.98 72.488 . 12000
e 1 . 70.150 . 42000
e 2 . 68.150 42000
77000/;
/* get minimum domains and ripple total cost up the
segments */
{v,s}(vs:=bid[v,s,'q_max']),
for{vs[v,s]|s>1} (bid[v,s,'setup'] := bid[v,s-1,'
setup']
+ bid[v,s-1,'q_max']*(bid[v,s-1,'price']
-bid[v,s,'price']));
variable
pl{v,s} "purchase level";
binary plb{v,s} "purchase decision";
constraint
demand "demand constraint": req = sum{vs[v,s]} pl;
minpl{vs[v,s]} "min purchase": pl >= bid[vs,'q_min']*
plb;
maxpl{vs[v,s]} "max purchase": pl <= bid[vs,'q_max']*
plb;
oneonly{v} "at most one deal": sum{s|vs} plb <= 1;

minimize cost:
sum{vs[v,s]} (bid[vs,'price']*pl + bid[vs,'setup']*
plb);
Writep(cost,pl,plb);
end

5.7. Blending Problem I (blend)

Problem: The problem is to blend a new alloy from other purchased alloys.
The model is from The GAMS model library (GAMS SEQ=2). See also [14],
chap 3.4.
Modeling Steps

model BLEND "Blending Problem I";
set alloy :=[a b c d e f g h i] "products on the
market";
elem :=[lead zinc tin] "required elements";
parameter compdat{elem,alloy} := [
10 10 40 60 30 30 30 50 20
10 30 50 30 30 40 20 40 30
80 60 10 10 40 30 50 10 50];
price{alloy}:=[4.1 4.3 5.8 6.0 7.6 7.5 7.3 6.9 7.3];
rb{elem}:=/lead 30 zinc 30 tin 40/ "required blend";
ce{alloy}:=sum{elem} compdat-100 "composition error (
pct-100)";
variable v{alloy} "purchase of alloy (pounds)";
constraint
pc{elem}: sum{alloy} compdat*v = rb "purchase
constraint";
mb frozen: sum{alloy} v = 1 "material balance";
ac frozen: sum{alloy} price*v "accounting: total cost
";
minimize cost1: ac;
Writep(ac,v);
Unfreeze(mb);
minimize cost2: ac;
Writep(ac,v);
end

5.8. Hanging Chain COPS 2.0 3 (chain)

Problem: Find the chain (of uniform density) of length L suspended between
two points with minimal potential energy. This model is from the COPS
benchmarking suite. See http://www-unix.mcs.anl.gov/ more/cops/. COPS
performance tests have been reported for nh = 50, 100, 200, 400.
Taken from the GAMS model library (GAMS SEQ=231). See also [8] and [22].
Modeling Steps

model CHAIN "Hanging Chain COPS 2.0 3";
//this model is not correctly formulated
set nh ,i "number of intervals for the discretization"
:= [1..50];
parameter
L "length of the suspended chain" := 4;
a "height of the chain at t=0 (left)" := 1;
b "height of the chain at t=1 (left)" := 3;
tf "odes defined in [0 tf]" := 1;
n "number of subintervals" := #nh-1;
h "uniform interval length" := tf/n;
tmin :=if(b>a,0.25,0.75);
variable
x{i} "height of the chain" :=
4*Abs(b-a)*((i-1)/n)*(0.5*((i-1)/n) - tmin) + a;
u{i} "derivative of x" := 4*Abs(b-a)*(((i-1)/n) -
tmin);
constraint
x_eqn{i|i+1}: x = x[i-1] + 0.5*h*(u[i-1]+u);
length_eqn:
0.5*h*(sum{i|i>1} (Sqrt(1+u[i-1]*u[i-1]) + Sqrt(1+u
*u))) = L;
xfx1: x[1] = a;
xfx2: x[#nh] = b;
minimize energy "potential energy" :
0.5*h*(sum{i|i>1} (x[i-1]*Sqrt(1+u[i-1]*u[i-1]) + x*
Sqrt(1+u*u)));
Writep(energy,x,u);
end

5.9. Chance Constraint Feed Mix Problem (chance)

Problem: formulated, a deterministic model and the chance constraint ver-
sion deterministic equivalent.
Chap 9, pp.94–100.
Modeling Steps

model CHANCE "Chance Constraint Feed Mix Problem";
set f "feeds" := [ barley, oats, sesame, grnd_meal ];
n "nutrients" := [ protein, fats ];
parameter price{f} "feed prices (fgld per ton)" :=
/barley 24.55 oats 26.75 sesame 39.00 grnd_meal
40.50/;
req{n} "requirements (pct)" :=/protein 21 fats 5/;
mean{n,f} "feed characteristics (pct)" := [
12.0 11.9 41.8 52.1
2.3 5.6 11.1 1.3 ];
variance{f} "for protein" :=[0.28 0.19 20.5 0.62];
variable x{f} "feed mix (pct)";
expression cost: sum{f} price*x "total cost per ton";
constraint
mc "mix constraint": sum{f} x = 1;
nbal{n} "nutrient balance": sum{f} mean*x >= req;
cc{n} "chance constraint":
sum{f} mean*x - 1.645*Sqrt(sum{f} variance*x^2) >=
req;
minimize obj2: cost subject_to mc,cc;
Writep(cost,x);
minimize obj1: cost subject_to mc,nbal;
Writep(cost,x);
end

5.10. Organic Fertilizer Use in Intensive Farming

(china)
Problem: Models the use of organic and chemical fertilizers for triple-cropping
in the southern Jiangsu province. Detailed attention is paid to removal and
replenishment of plant nutrients and humus.
Modeling Steps

model CHINA "Organic Fertilizer Use in Intensive Farming"
;
set ca "all commodities" := [
barley
wheat
e_rice -- early rice
m_rice -- middle rice
l_rice -- late rice
l_sc_rice -- late single crop rice
g_manure -- green manure
gm_seeds -- green manure seeds
rapeseed
azolla
azolla_e -- early azolla
fodder
g_feed -- grain feed
straw
hyacinth
silt
nightsoil
amm_water -- ammonia water
amm_bi -- ammonium bicarbonate
ssp -- single superphosphate
pigs
shoats
c_straw -- composted straw
c_gm -- composted green manure
c_hyacinth -- composted hyacinth
pig_m -- pig-manure
straw_b -- straw bedding
rapes_c -- composted rapeseed
vegetables ];
c{ca} "crops" := [ barley, wheat, e_rice, m_rice,
l_rice, l_sc_rice, g_manure
gm_seeds, rapeseed, azolla, azolla_e, fodder ];

g{ca} "grains" := [ barley, wheat, e_rice, m_rice
, l_rice, l_sc_rice ];
cu{ca} "upland crops" := [ fodder, vegetables ];
cp{ca} "commodities purchased" := [ hyacinth, pig_m
, silt, nightsoil, amm_water, amm_bi, ssp ];
cs{ca} "commodities sold" := [ barley, wheat,
e_rice, m_rice, l_rice, l_sc_rice
rapeseed, pigs, shoats, straw, vegetables ];
cf{ca} "fertilizers" := [ c_straw, c_gm, c_hyacinth
, pig_m, straw_b, nightsoil
azolla, rapes_c, amm_water, amm_bi, ssp ];
s "crop sequence" := [ bar_r_r, bar_sr, bar_r,
wh_sr, wh_r, gm_r_r, gm_sr, gm_s_sr, gm_r,
rape_sr, fallow_sr ];
sh{s} "sequences with higher fertilizer
application" := [ bar_r_r, gm_r_r, wh_r, bar_r
];
nh "nutrients and humus" := [ n, p2o5, k2o,
humus ];
n{nh} "nutrients" := [ n, p2o5, k2o ];
en "effective nutrients" := [ n_imm, n_tot,
p2o5_eff, k2o_exch, humus ];
f "fertilization intensity" := [ normal, high
];
p "pig raising activities" := [ r_shoats,
r_pigs ];
ss{s,f} "sequence possibilties" := if(f='normal'
or sh,1);
t "time periods" := [ nov_1, nov_mar, mar_2,
apr_1, apr_2, may_1, may_jun, jun_2, jul_1,
jul_2
aug_1, aug_2, sep, oct_1, oct_2 ];
s1; s2; s3; s4;
parameter
paddy "paddy land available (mu)" := 268.6;
upland "upland avaialable (mu)" := 7.8;
muperha "conversion of mu to hectars (mu per hectar)
" := 15;
jinperkg "conversion of jin to kg (jin per kg)" :=
2.0;
yperd "exchange rate (yuan per dollar)" := 1.8;
mcp{s,c} "multi cropping patterns for paddyland (mu)"
:= /
:barley wheat e_rice m_rice l_rice
l_sc_rice g_manure gm_seeds rapeseed
azolla azolla_e fodder:
bar_r_r 0.704 0.193 0.807 . 1.0
. . . . .
. .
gm_r_r . 0.193 .807 . 1.0

. 0.704 . . .
. .
bar_sr 0.859 . . . .
1.0 0.141 . . 1.0
0.859 .
wh_sr . 0.859 . . .
1.0 0.141 . . 1.0
. .
gm_sr . . . . .
1.0 1.0 . . 1.0
0.859 .
rape_sr . . . . .
1.0 0.141 . 0.859 1.0
. .
fallow_sr . . . . .
1.0 0.141 . . .
. 0.859
gm_s_sr . . . . .
1.0 0.1 0.9 . .
. .
bar_r 0.9 . . 1.0 .
. 0.1 . . 1.0
. .
wh_r . 0.9 . 1.0 .
. 0.1 . . .
. .
gm_r . . . 1.0 .
. 1.0 . . 1.0
. . /;
cdata{ca,s1} "crop data" := /
/* yield : paddy land yield in jin per mu
cash_cost : yuan per mu
straw_y : kg per kg output
proc_price: yuan per ton
quota_sale: tons */
:yield cash_cost straw_y proc_price
quota_sale:
barley 427.6 11.0 1.0 226.0
.
wheat 530.3 11.0 1.1 316.0
.
e_rice 690.3 14.9 0.8 232.0
.
m_rice 962.7 23.6 1.0 238.0
.
l_rice 558.2 18.9 0.8 272.0
.
l_sc_rice 857.7 25.2 1.0 233.0
.
g_manure 4968.1 2.3 . .

.
rapeseed 221.1 9.8 1.5 720.0
4
azolla 1149.0 . . .
.
azolla_e 1149.0 . . .
.
fodder 1200 . . .
.
straw . . . 44.0
1.2
pigs . . . 1200.0
4
shoats . . . 1800.0
.
vegetables . . . 375.0
. /;
yield{ca} "paddy land crop yields (ton per mu)" :=
cdata[ca,'yield']/jinperkg/1000;
yieldu{ca} "upland yields (ton per mu)" := / fodder
0.3, vegetables 0.24 /;
cxcrop{s} "cash cost by sequence (yuan per mu)" :=
sum{c} cdata[c,'cash_cost']*mcp;
aqsprice{ca} "above quota sales price (yuan per ton)"
:= if(ca within c,1.5,1.1)*cdata[ca,'proc_price'];
grainquota "grain sales quota (tons)" := 100;
purdata{ca,s2} "purchase prices and limits" := /
:price quantity:
/* (yuan per ton) (tons) */
amm_water 30 .
amm_bi 123 .
nightsoil 20 7
ssp 100 500
pig_m 12 10000
hyacinth . 15 /;
pigio{ca,p} "pig raising input output relations" := /
:r_shoats r_pigs:
shoats 1 -0.12
pigs . 1
straw_b 8 8
g_feed -3.3 -2.6
straw -8.8 -8.8
fodder -0.55 -0.50 /;
cxpig{p} "pig raising cashcost (yuan per ton)" := /
r_shoats 90, r_pigs 60 /;
gio{ca,ca1 in ca|g[ca1]} "grain feed mixing recipes (
kg per kg)" := if(ca=ca1,-1, ca='g_feed',1, 0);
/* labor use and supply */
lu{t,c} "labor use for crops (gongs per mu per period

)" := /
:barley wheat e_rice m_rice l_rice l_sc_rice
g_manure gm_seeds rapeseed:
nov_1 8.15 8.15 . . 3.17 3.00
. . 14.90
nov_mar 7.45 7.45 7.75 5.57 6.33 13.07
1.00 1.00 15.15
mar_2 1.10 1.10 2.03 0.74 . .74
. . 1.85
apr_1 0.4 0.4 3.25 0.70 . .70
. . .
apr_2 0.9 0.9 3.36 0.70 . .70
. . 1.0
may_1 0.2 0.2 8.20 2.92 . 4.42
3.00 . 0.9
may_jun 10.80 8.80 13.58 12.12 1.50 5.27
. 6.70 8.40
jun_2 . 2.00 2.35 4.40 3.00 8.99
. . .
jul_1 . . 2.48 5.41 3.00 5.00
. . .
jul_2 . . 0.70 5.59 3.00 4.50
. . .
aug_1 . . 8.42 1.00 9.35 1.42
. . .
aug_2 . . 1.88 3.65 4.92 4.54
. . .
sep 1.30 1.30 . 3.20 5.09 4.15
. . 1.00
oct_1 1.30 1.30 . 6.13 1.52 .
1.25 1.25 0.50
oct_2 1.25 1.25 . 4.37 0.12 .
1.0 1.00 0.30 /;
labj{t} "labor adjustment coefficient (proportion)"
:= / aug_1 0.12 /;
days{t} "number of days in periods (days)" := /
nov_1 22, nov_mar 120, mar_2 16, apr_1 15 ,apr_2
15, may_1 14, may_jun 32, jun_2 15
jul_1 15, jul_2 12, aug_1 14, aug_2 21, sep 30,
oct_1 14, oct_2 10 /;
tday "total number of days in all periods (days)" :=
sum{t} days;
lsup "labor supply (gongs per day)" := 137;
dlab{t,c} "daily labor requiremnts for single crop (
gongs per day per mu)" := lu/days;
lab{t,s} "labor requirements for crop sequences (
gongs per day per mu)" := sum{c} dlab*mcp;
/* fertilizer requiremets and supply */
nc{cf,nh} "nutrient composition (pct weight)" := /
:n p2o5 k2o humus:

c_straw 0.63 0.11 0.85 21.4
c_gm 0.40 0.11 0.35 2.34
c_hyacinth 0.24 0.07 0.11 2.64
pig_m 0.35 0.32 0.75 .
straw_b 0.60 0.20 0.80 15.9
azolla 0.34 0.02 0.12 2.58
nightsoil 0.65 0.30 0.25 .
rapes_c 4.60 2.50 1.40 .
amm_water 3.75 . . .
amm_bi 17.00 . . .
ssp . 15.00 . . /;
nu{cf,nh,en} "fertilizer utilization rates (pct)" :=
/
[c_straw,*,*] n n_imm 10 n n_tot 76 p2o5
p2o5_eff 20 k2o k2o_exch 14 humus humus 100
[c_gm,*,*] n n_imm 10 n n_tot 76 p2o5
[c_hyacinth,*,*] n n_imm 10 n n_tot 76 p2o5
[pig_m,*,*] n n_imm 9 n n_tot 76 p2o5
[straw_b,*,*] n n_imm 9 n n_tot 76 p2o5
[nightsoil,*,*] n n_imm 9 n n_tot 76 p2o5
[azolla,*,*] n n_imm 20 n n_tot 76 p2o5
[rapes_c,*,*] n n_imm 10 n n_tot 76 p2o5
p2o5_eff 20 k2o k2o_exch 14
[amm_water,*,*] n n_imm 20 n n_tot 40
[amm_bi,*,*] n n_imm 22 n n_tot 45
[ssp,*,*] p2o5 p2o5_eff 25 /;
pno{c,n} "plant nutrient offtake (pct weight of
output)" := /
:n p2o5 k2o:
barley 3.0 1.0 2.0
wheat 2.75 1.2 2.4
e_rice 1.9 0.8 2.7
m_rice 2.1 1.0 2.9
l_rice 2.05 0.9 2.75
l_sc_rice 1.9 0.9 3.0
g_manure . 0.11 0.35
rapeseed 7.0 3.6 7.1/;
nup{cu,en} "upland cropping nutrient requirement (kg
per mu)" := /
:n_imm n_tot p2o5_eff k2o_exch humus:
fodder 2.0 11.0 1.0 2.5 .
vegetables 4.58 21.35 2.57 3.93 . /;
soil{c,s3} "soil data - supplies of p2o5 k2o nitrogen

requirement and irrigation water" := /
:p2o5_eff k2o_exch inr iwr:
/* p r o p o r t i o n s) (tons per ha) */
barley 0.6 0.3 0.47 4125
wheat 0.6 0.3 0.45 4125
e_rice -0.1 0.25 0.45 4875
m_rice 0.2 0.35 0.45 6000
l_rice 0.6 0.75 0.3 4875
l_sc_rice 0.05 0.25 0.24 8250
g_manure 0.9 1 . .
rapeseed 0.5 0.4 0.58 ./;
cno{c,n} "crop nutrient offtake (ton per mu)" := pno*
yield/100;
enc{cf,en} "effective nutrient content (pct weight)"
:= sum{nh} nu*nc/100;
kc "k2o content of irrigation water (ppm)" := 3.5;
hdr "humus decomposition rate (kg per kg soil n)":= 9;
cxfert{cf} "fertilization cash cost (yuan per ton)"
:= / azolla 1.3, rapes_c 200 /;
freq{c,en} "effective nutrient requirements (ton per
mu)" := if(
en='n_imm', cno[c,'n']*soil[c,'inr'],
en='n_tot', cno[c,'n'],
en='p2o5_eff', cno[c,'p2o5']*(1-soil[c,'p2o5_eff']),
en='k2o_exch', cno[c,'k2o']*(1-soil[c,'k2o_exch']) -
soil[c,'iwr']*kc/muperha/1e6,
en='humus', hdr*cno[c,'n']*(1-soil[c,'inr']) );
mult{en} "nutrient req factor" := / n_imm 1.1, n_tot
1.05, p2o5_eff 1.25, k2o_exch 1.1 , humus 1.0 /;
hyield{c} "yield for high fertilizer applications (
tons per mu)" := yield*1.07;
hreq{c,en} "nutrient requirements for high
application rates (tons per mu)" := freq*mult;
sreq{s,en,f} "crop sequence nutrient requiremets (ton
per mu)";
syield{ca,s,f} "yield of crop sequence (ton per mu)";
sys{s,f} "straw yield of crop sequences (ton per mu)";
DoAssign "a dummy name" :=
{s,en} (sreq[s,en,'normal'] := sum{c} freq*mcp,
sreq[s,en,'high'] := sum{c} hreq*mcp),
{c,s} (syield[c,s,'normal'] := yield*mcp,
syield[c,s,'high'] := hyield*mcp),
{s,f} (sys := sum{c} cdata[c,'straw_y']*syield,
---- e_rice straw is plowed under immediately,
adjust straw yields and fertilizer requirement
*/
sys := sys - syield['e_rice',s,f]*cdata['
e_rice','straw_y']),
{s,en,f}(sreq := sreq - cdata['e_rice','straw_y']*

syield['e_rice',s,f]*enc['c_straw',en]/100*mcp[s
,'e_rice']),
{s,f} (syield['straw',s,f] := sys,
syield['rapes_c',s,f] := 0.8*syield['rapeseed
',s,f]);
crec{ca,cf} "composting and fertilizing recipes" := /
:c_straw c_gm c_hyacinth azolla:
straw 0.12 . . .
g_manure . 0.12 . .
hyacinth . . 0.12 .
pig_m 0.05 0.05 0.05 .
silt 0.83 0.83 0.83 .
azolla . . . 1.0
azolla_e . . . 0.2 /;
DoAssign1 := {cf[ca1]|~(sum{ca} crec[ca,ca1])} (crec[
ca,ca1]:= 1);
chemnall{ca,s4} "chemical nitrogen (ammonium
bicarbonate) allocations" := /
/* crops: kg ammonium bicarbonate per mu of crop
planted
qsa : quota sales allocation for pigs (kg per kg)
aqsa : above quota sales allocation (kg per kg) */
:crops qsa aqsa:
barley 49 . 0.17
wheat 49 . 0.17
e_rice 53 . 0.17
m_rice 45 . 0.17
l_rice 45 . 0.17
l_sc_rice 45 . 0.17
rapeseed 45 . 0.45
g_manure 4 . .
pigs . 0.225 0.225/;
schem{s} "chemical fertilizer allocation for crop
sequences (tons amm_bi per mu)" := 0.001*(sum{c}
chemnall[c,'crops']*mcp);
/* model definition */
variable
xcrop{s,f} "paddy land cropping activities (mu)";
xupland{ca} "upland cropping activities (mu)";
xpig{p} "pig raising activities (ton)";
xfeed{g} "grain feed mixing (ton)";
xfert{ca} "fertilization activities (kg)";
purchase{ca} "purchasing activities (ton)";
sales{ca} "quota sales (ton)";
aqsales{ca} "above quota sales (ton)";
ccost "cash cost (yuan)";
constraint
mb{ca} "material balance (ton)": sum{s,f|ss} syield*
xcrop + sum{p} pigio*xpig + if(cp,purchase)
+if(cu,yieldu*xupland) + sum{g[ca1]} gio[ca,ca1]*

xfeed[ca1] >= if(cs,sales+aqsales) + sum{cf[ca1
]} crec[ca,ca1]*xfert[ca1];
labor{t} "labor balance (gong per day)": sum{s,f|ss}
lab*xcrop <= (1+labj)*lsup;
fert{en} "nutrient and humus balance (ton)": sum{s,f|
ss} sreq*xcrop + 0.001*(sum{cu} nup*xupland)
<= 0.01*(sum{cf} enc*xfert);
chemn "chemical nitrogen allocation (ton)": sum{s,f|ss
} schem*xcrop + sum{cs} (chemnall[cs,'qsa']*sales
+ chemnall[cs,'aqsa']*aqsales) >= purchase['amm_bi
'];
landp "paddy land constraint (mu)": sum{s,f|ss} xcrop
<= paddy;
landu "upland land constraint (mu)": sum{cu} xupland
<= upland;
gmseed "green manure seed requirements (mu)": sum{s,f|
ss} ((1-0.16)*mcp[s,'gm_seeds'] - 0.16*mcp[s,'
g_manure'])*xcrop >= 0;
grainq "grain quota sales (ton)": sum{g} sales >=
grainquota;
cdef "cash cost definition (yuan)": ccost = sum{s,f|ss
} cxcrop*xcrop + sum{p} cxpig*xpig
+ sum{cf} cxfert*xfert + sum{cp} purdata[cp,'price
']*purchase;
saleslower{cs}: sales >= cdata[cs,'quota_sale'];
purchaseupper{cp|purdata[cp,'quantity']}: purchase <=
purdata[cp,'quantity'];
maximize income "brigade income (yuan)": sum{cs} (cdata[
cs,'proc_price']*sales + aqsprice*aqsales) - ccost;
-- solution report
set su := ['<demand>', '<supply>', '<s_price>'] "
fertilization demand and supply summary";
parameter frep{en,su}:=if(su=1, 0.01*(sum{cf} enc*xfert)
+ GetValue(fert,5),
su=2, 0.01*(sum{cf} enc*xfert)
,
su=3, -GetValue(fert,3));
Writep(income,frep);
end

5.11. Financial Optimization: Financial Engineering

(cmo)
Problem: Collateralized Mortgage Obligations (CMO) are used to restructure
the cashflows from underlying mortgage collateral into a set of high quality
bonds with different maturities. In the following mixed-integer programming
is used to restructure a mortgage pool into six normal tranches and on zero
tranch. The model is written in a general form to allow experimentation with
different periodicity. The following version is annual.
Modeling Steps

model CMO "Financial Optimization: Financial Engineering"
;
set
i,j "tranches" := [ n1 n2 n3 n4 n5 n6 , m ];
n{i} "normal tranches" := [ n1 n2 n3 n4 n5 n6 ];
m{i} "m tranches" := [m];
tp "time periods" := [ 0 1 2 3 4 5 6 7 8 9 10 ];
ts{tp} "settlement date" := [ 0 ];
t{tp} "payment periods" := [ 1 2 3 4 5 6 7 8 9 10 ];
tl{tp} "last payment" := [ 10 ];
/* note: all rates are monthly rates x 12. For
aggregation we need to compute equivalent rates.
(1+r/12)**12 == (1+rd)**d */
parameter
cg "collateral annual gross coupon" := 0.10;
nom "nominal value of collateral" := 100;
s "servicing rate" := 0.005;
psa "pricing speed" := 115;
d "periodicity" := 1;
tmax "term to maturity of collateral" := #t;
age "age of the collateral in years" := (360 -
tmax*12/d)/12;
minn "minimum number of normal tranches" := 3;
minm "minimum number of m tranches" := 1;
--tranches data
trcoupon{i}:= [0.085 0.085 0.085 0.0875 0.0875
0.0875 /*0.0875 0.0875*/ 0.0995];
tryield{i} := [0.099 0.0999 0.0999 0.1006 0.1009
0.1017 /*0.1018 0.1025*/ 0.105];
trlowWal{n}:= [1.00 2.00 3.00 4.00 5.00 7.00 /*10.00
15.00*/];
trupWal{n} := [1.24 2.44 3.44 4.44 5.65 7.94 /*10.94

20.00*/];
tryr := 0.1285;
rev{tp} "reverse order addresses" := #tp-2*tp + 1;
cd "periodic gross coupon" := (1+cg/12)^(12/d
)-1;
sd "periodic servicing rate" := (1+s/12)^(12/d
)-1;
po{tp} "outstanding collateral principal at zero
prepayment" := nom*(1-1/(1+cd)^(tmax+1-tp))
/(1-1/(1+cd)^tmax);
a "zero prepayment annuity" := nom*cd
/(1-1/(1+cd)^tmax);
b{tp} "prepayment rate" := if(tp>1, 0.06*psa
/100*Min((tp+age*d)/(d*2.5) , 1));
coupon{i} "periodic coupon by tranche" := (1+trcoupon
/12)^(12/d)-1;
yield{i} "periodic yield by tranche" := (1+tryield
/12)^(12/d)-1;
yr "periodic residual yield" := (1+tryr/12)
^(12/d)-1;
cmax "maximum cmo coupon" := max{i} coupon;
wallo{i} "lower weighted average life in periodicity
" := trlowWal*d;
walup{i} "upper weighted average life in periodicity
" := trupWal*d;
bv{tp} "temporary factor for bvf calculation";;
bv[#tp]:=0, {tp} (bv[tp+rev-1]:=(bv[tp+rev]+a-sd*po[tp+
rev-1])/(1+cmax));
parameter
bvf{tp} "bond value factor" := if(po,Min(bv/po,1),
#tp,1);
pe{tp} "expected outstanding collateral principal"
;
dum1 := pe[1]:=nom, for{tp|tp>1} (pe:=pe[tp-1]*(1-b)
^(1/d)*po/po[tp-1]);
cflow{tp} "expected collateral payments" := if(tp
>1,(1+cd-sd)*pe[tp-1] - pe[tp]);
prin{tp} "structuring principal payments from
collateral" := if(tp>1,pe[tp-1]*bvf[tp-1] - pe*
bvf);
sump "sum of prin" := sum{tp} prin;
psum{tp} "sum of principal payments on collateral
till tp";
tpsum{tp} "sum of time principal payment products";
dum2 := tpsum[1]:=0, psum[1]:=0, for{tp|tp>1} (psum:=
psum[tp-1]+prin, tpsum:=tpsum[tp-1]+(tp-1)*prin);
walp{tp} "wal on principal payments from collateral"
:= if(tp>1,tpsum/psum);
bign "big number":= sump*0.7;
smalln "small number" := sump*0.03;

set
zpos{i,tp} "possible payment periods by tranche" :=
if(walp>walup and i within n,0,1);
variable
x{i,tp} "outstanding principal in each tranche";
c{i,tp} "cashflow in each tranche";
y{i,tp} "upper triangular structure";
r{tp} "residual payments";
pv{i} "def pv tranches";
p{i,tp} [-1e18..1e18] "principal payments on tranches
";
tpp{i} [-1e18..1e18] "product of time and principal
payment";
pvres [-1e18..1e18] "def pv residuals";
binary z{i,tp} "tranche utilization variable (1=usage
)";
tin{i} "tranche in the structure";
constraint
defpv{i} "def pv tranches": pv = sum{t|zpos} 1/(1+
yield)^(t-1)*c;
defpvres "def pv residuals": pvres = sum{t} 1/(1+yr)
^(t-1)*r;
pdef{zpos[i,t]|t>1} "definition of principal payments
": p = x[i,t-1] - x;
cdef{zpos[i,t]|t>1} "cashflow accounting": c =
coupon*x[i,t-1] + p;
retiren1{t} "retirement of normal tranches": sum{n|
zpos} p = prin + sum{m} (x[m,tp-1]*coupon-c);
retire{zpos[i,t]|i within n} "retirement of tranches"
: p <= cflow*z;
retirem{m,t} "retirement of m tranche": c <= cflow*z
;
retirem1{m,t} "retirement of m tranche": p <= prin*z;
cbal{t} "cashflow balance": sum{i|zpos} c + r =
cflow;
tppdef{n} "definition of tpp": tpp = sum{t|zpos}
(t-1)*p;
lowal{n} "lower bound on weighted average life":
wallo*(sum{ts} x) <= tpp;
upwal{n} "upper bound on weighted average life":
walup*(sum{ts} x) >= tpp;
seq1{t} "sequencing constraint 1": sum{i|zpos}
z = 1;
seq2{zpos[i,t]|t>1} "sequencing constraint 2": y >= y
[i,t+1];
ydef{zpos[i,t]|t>1} "definition of y": y = y[i-1,t] +
z;
tindef1{i} "tranche in definition": sum{ts} x <=
tin*bign;
tindef2{i} "tranche in definition": sum{ts} x >=

tin*smalln;
ncon "constraint on number of tranches": sum{n} tin
>= minn;
mcon "constraint on number of m tranches": sum{m} tin
>= minm;
pLow{n,tp}: p>=0;
xfx1{i,tl}: x=0;
xfx2{i,t|~zpos[i,t+1]}: x=0;
zfx1{i,tp|~zpos}: z=0;
zfx2{m,tp|~(sum{j in n} zpos[j,tp])}: z=1;
maximize proceeds: sum{i} pv + pvres;
Writep(proceeds,z,x,c,r);
end

5.12. Peacefully Coexisting Armies of Queens (coex)
Problem: Two armies of queens (balck and white) peacefully coexist on a

chessboard when they are placed on the board in such a way that no two
queens from opposing armies can attack each other. The problem is to find
the maximum two equal-sized armies.
This model is from the GAMS model library (GAMS SEQ=219). See also [24].
Modeling Steps

model COEX "Peacefully Coexisting Armies of Queens";
set i ,j,ii "size of chess board" := [ 1..8 ];
m{i,j,ii,jj in i} "shared positions on the board" :=
(i=ii) or j=jj or Abs(i-ii)=Abs(j-jj);
binary variable b{i,j} "square occupied by a black
queen";
w{i,j} "square occupied by a white queen";
variable tot "total queens in each army";
constraint
eq1{m[i,j,ii,jj]} "keeps armies at peace": b[i,j] + w
[ii,jj] <= 1;
eq2 "add up all the black queens": tot = sum{i,j} b;
eq3 "add up all the white queens": tot = sum{i,j} w;
maximize obj: tot;
end

Solution:
opt: solution reported in OPTIMA
b.fx('6','2') = 1;
b.fx('7','2') = 1;
b.fx('8','2') = 1;
b.fx('7','1') = 1;
b.fx('8','1') = 1;
b.fx('7','6') = 1;
b.fx('8','6') = 1;
b.fx('7','7') = 1;
b.fx('8','7') = 1;
5.13. Peacefully Coexisting Armies of Queens -

tight (coexx)
Problem: We have set the size of the board to 5 in order to find solutions
quickly. In addition we fix the position of one queen. Two armies of queens
(balck and white) peacefully coexist on a chessboard when they are placed on
the board in such a way that no two queens from opposing armies can attack
each other. The problem is to find the maximum two equal-sized armies.
This model is from the GAMS model library (GAMS SEQ=218). See also [24].
Modeling Steps

model COEXX "Peacefully Coexisting Armies of Queens -
tight";
set i ,j "size of chess board" := [1..8];
s "diagonal offset" :=[1..13]; --2i-3 diagonals
parameter sh{s} "shift values for diagonals" :=s-#i+1;
rev{s,i} "reverse shift order" := #i+1-2*i+ sh;
binary variable xw{i,j} "has a white queen";
xb{i,j} "has a black queen";
wa{i} "white in row i";
wb{i} "white in column j";
wc{s} "white in diagonal s";
wd{s} "white in backward diagonal s";
variable tot;
constraint
aw{i,j} "white in row i": wa[i] >= xw;
bw{j,i} "white in column j": wb[j] >= xw[i,j];
cw{s,i} "white in diagonal s": wc[s] >= xw[i,i+sh];
dw{s,i} "white in backward diagonal s": wd[s] >= xw[i
,i+rev];
ab{i,j} "black in row i": 1-wa[i] >= xb;
bb{j,i} "black in column j": 1-wb[j] >= xb[i,j];
cb{s,i} "black in diagonal s": 1-wc[s] >= xb[i,i+sh];
db{s,i} "black in backward diagonal s": 1-wd[s] >= xb
[i,i+rev];
eb "total black": tot = sum{i,j} xb;
ew "total white": tot = sum{i,j} xw;
fx11: xb[1,1] = 1; --fix one position in the NW
corner
maximize obj: tot;
Writep(tot);
end

5.14. Alcuin’s River Crossing (cross)

Problem: A farmer carrying a bushel of corn and accompanied by a goose and
a wolf came to a river. He found a boat capable of transporting himself plus
one of his possessions - corn, goose, or wolf - but no more. Now, he couldn’t
leave the corn alone with the goose, nor the goose alone with the wolf, else
one would consume the other. Nevertheless, he succeeded in getting himself
and his goods across the river safely.
This model is from the GAMS model library (GAMS SEQ=191) (Contributed
by Soren Nielsen, Institute for Mathematical Sciences University of Copen-
hagen). See also [2].
Modeling Steps

model CROSS "Alcuin's River Crossing";
set i "items" := [goose, wolf, corn];
t "time" := [1..10];
parameter dir{t} "crossing - near to far is +1 - far
to near -1" := -1^(t-1);
variable
cross{i,t} "crossing the river";
done{t} "all items in far side";
binary y{i,t} "1 iff the item is on the far side at
time t";
constraint
DefCross{i,t|t>1} "crossing": y = y[i,t-1] + dir[t
-1]*cross[i,t-1];
DefDone{i,t} "everything on far side": done <= y;
limCross{t|t>1}: sum{i} cross[i,t-1] <= 1;
EatNone1{t}: dir*(y['goose',t] + y['wolf',t] - 1) <=
done;
EatNone2{t}: dir*(y['goose',t] + y['corn',t] - 1) <=
done;
yFx{i}: y[i,1] = 0;
maximize nocross "number of non crossing periods": sum{t
} done;
Writep(nocross,y,cross);
end

5.15. 3-dimensional Noughts and Crosses (cube)

Problem: White and black balls are to be arranged in a cube one to a cell in
such a way as to minimize the number of lines with balls of equal color. For
this example the length of the cube is three. a total of 49 lines exist in a cube of
3x3x3. 27 lines by changing only one coordinate, 18 diagonals within a plane,
and 4 diagonals going across planes. (see Will16.lpl)
and te model WILL16.
Modeling Steps

model CUBE "3-dimensional Noughts and Crosses";
set s,sp,spp "domain for line" := [a b c incr decr];
x{s} "coordinate labels" := [ a, b, c ];
y{sp} "coordinate labels" := [ a, b, c ];
z{spp} "coordinate labels" := [ a, b, c ];
d{s} "directions" := [ incr, decr ];
dp{sp} "directions" := [ incr, decr ];
b "bounds" := [ low, high ];
ld{s,sp,spp} "line definition" :=
(s='incr' and sp within x and spp within x or
s within x and sp='incr' and spp within x or
s within x and sp within x and spp='incr' or
s='incr' and sp within d and spp within x or
s within x and sp='incr' and spp within d or
s within d and sp within x and spp='incr' or
s='incr' and sp within d and spp within d);
parameter
ls{b} "sign for line definitions" :=/low 1 high -1/;
lr{b} "rhs for line definitions" :=/low 2 high -1/;
df{x,s} "line definition function" := if(s='decr',1+#
x-2*x, s='incr',0, s-x);
variable
line{s,sp,spp} "line identification";
binary core{x,y,z} "placement of balls (white 0
black 1)";
constraint
nbb "total number of balls definition": sum{x,y,z}
core = Floor(#x^3/2);
ldef{ld[s,sp,spp],b} "line definitions": ls*(sum{v in
x} core[v+df[v,s],v+df[v,sp],v+df[v,spp]]) <=
line+lr;
minimize num "number of lines of equal color": sum{ld[s

,sp,spp]} line;
Writep(num,core,line);
end

5.16. Simple Farm Level Model (demo1)

Problem: This is the first in a series of single farm models . This simplest
version has only 7 principal crops and 2 basic inputs, land and labor, which
are specified on a monthly basis.
Modeling Steps

model DEMO1 "Simple Farm Level Model";
set c "crops" := [ wheat, clover, beans, onions, cotton
, maize, tomato ];
t "period" := [ jan, feb, mar, apr, may, jun, jul,
aug, sep, oct, nov, dec ];
parameter landreq{t,c} "months of land occupation by
crop (hectares)" := [
1.0 1.0 1.0 1.0 . . .
1.0 1.0 1.0 1.0 . . .
1.0 0.5 1.0 1.0 0.5 . .
1.0 . 1.0 1.0 1.0 . .
1.0 . . 0.25 1.0 0.25 .
. . . . 1.0 1.0 .
. . . . 1.0 1.0 0.75
. . . . 1.0 1.0 1.0
. . . . 1.0 1.0 1.0
. . . . 1.0 .5 1.0
0.5 0.25 0.25 0.5 0.75 . 0.75
1.0 1.0 1.0 1.0 . . . ];
laborreq{t,c} "crop labor requirements (man days per
hectare)" := [
1.72 4.5 0.75 5.16 . . .
0.5 1.0 0.75 5.0 . . .
1.0 8.0 0.75 5.0 5.0 . .
1.0 . 16. 19.58 5.0 . .
17.16 . . 2.42 9.0 4.3 .
2.34 . . . 2.0 5.04 .
. . . . 1.5 7.16 17.0
. . . . 2.0 7.97 15.0
. . . . 1.0 4.41 12.0
. . . . 26.0 1.12 7.0
2.43 2.5 7.5 11.16 12.0 . 6.0
1.35 7.5 0.75 4.68 . . . ];
yield{c} "crop yield (tons per hectare)" := /
wheat 1.5, clover 6.5, beans 1, onions 3
cotton 1.5, maize 2 , tomato 3 /;

price{c} "crop prices (dollars per ton)" := /
wheat 100, beans 200, onions 125
cotton 350, maize 70, tomato 120 /;
miscost{c} "misc cash costs (dollars per hectare)" := /
wheat 10, beans 5, onions 50
cotton 80, maize 5, tomato 50 /;
/* farm data, size labor availability etc. */
land "farmsize (hectares)" := 4.;
famlab "family labor available (days per month)" := 25;
owage "hire-out wage rate (dollars per day)" := 3.;
twage "temporary labor wage (dollars per day)" := 4;
dpm "number of working days per month" := 25;
/* endogenous variables and equations */
variable
xcrop{c} "cropping activity (hectares)";
flab{t} "family labor use (days)";
fout{t} "hiring out (days)";
tlab{t} "temporary labor (days)";
revenue "value of production (dollars)";
mcost "misc cash cost (dollars)";
labcost "labor cost (dollars)";
labearn "labor income (dollars)";
constraint
landbal{t} "land balance (hectares)": sum{c} xcrop*
landreq <= land;
laborbal{t} "labor balance (days)": sum{c} xcrop*
laborreq <= flab + tlab;
flabor{t} "family labor balance (days)": famlab =
flab + fout;
arev "revenue accounting (dollars)": revenue = sum{c}
xcrop*yield*price;
acost "cash cost accounting (dollars)": mcost = sum
{c} xcrop*miscost;
alab "labor cost accounting (dollars)": labcost = sum
{t} tlab*twage;
aout "labor income accounting (dollars)": labearn =
sum{t} fout*owage;
maximize yfarm "farm income": revenue+labearn-labcost-
mcost;
/* report on solution */
set crep := [ landuse, output, revenue ] "crop report
summary";
lrep := [ demand, family, temporary unused,
hire_out ] "labor report summary(days)";
parameter
croprep{crep,c}:=if(crep=1, xcrop,
crep=2, xcrop*yield,
crep=3, croprep['output',c]*price
);
labrep{t,lrep}:= if(lrep=1, sum{c} xcrop*laborreq,

lrep=2, flab,
lrep=3, tlab,
lrep=4, -GetValue(laborbal,5),
lrep=5, fout);
Writep(yfarm,croprep,labrep);
end

5.17. Non-transitive Dice Design (dice)

Problem: Probabilistic dice - an example of a non-transitive relation. We want
to design a set of dice with an integer number on each face such that on av-
erage dice1 beats dice2, and dice2 on average beats dice3 etc, but diceN has
to beat dice1. MIP codes behave very erratic on such a problem and slight
reformulations can result in dramatic changes in performance. Also note the
face value will be integers automatically.
This model is from the GAMS library (GAMS SEQ=176). See [4].
Modeling Steps

model DICE "Non-transitive Dice Design";
set f ,fp "faces on a dice" := [1..6];
dice ,d "number of dice" := [dice1 dice2 dice3];
parameter
flo "lowest face value" := 1;
fup "highest face value" := #dice*#f;
wn "wins needed - possible bound" := Floor(0.5*#f
^2)+1;
variable
wnx "number of wins";
fval{d,f} [flo..fup] "value of dice - will be
integer";
binary comp{d,f,fp} "one if f beats fp";
constraint
ffx: fval['dice1','1'] = flo;
equ1{d} "count the wins": sum{f,fp} comp = wnx;
eq3{d,f,fp} "definition of non-transitive relation":
fval+(fup-flo)*(1-comp) >= fval[d%#d+1,fp]+1;
eq4{d,f|f<#f} "different face values for a single
dice": fval+1 <= fval[d,f+1];
maximize obj: wnx;
Writep(wnx);
end

5.18. Stigler’s Nutrition model (diet1)

Problem: This model determines a least cost diet which meets the daily al-
lowances of nutrients for a moderately active man weighing 154 lbs.
This model is from the GAMS library (GAMS SEQ=7) and from [14], Chap
27.1.
Modeling Steps

model DIET "Stigler's Nutrition model";
set
n "nutrients" := [ calorie /*thousands*/, protein /*
grams*/ , calcium /*grams*/, iron /*milligrams*/
vitamin_a /*thousand ius*/, vitamin_b1 /*
milligrams*/, vitamin_b2 /*milligrams*/, niacin
/*milligrams*/, vitamin_c /*milligrams*/ ];
f "foods" := [ wheat, cornmeal, cannedmilk, margarine
, cheese, peanut_b, lard liver, porkroast, salmon
, greenbeans, cabbage, onions, potatoes spinach,
sweet_pot, peaches, prunes, limabeans, navybeans
];
parameter b{n} "required daily allowances of nutrients"
:=
/ calorie 3, protein 70, calcium 0.8, iron 12
vitamin_a 5, vitamin_b1 1.8, vitamin_b2 2.7,
niacin 18, vitamin_c 75 /;
a{f,n} "nutritive value of foods (per dollar spent)"
:= [
/*(1000) (g) (g) (mg)(1000iu)(mg) (mg) (mg) (mg)
*/
44.7 1411 2.0 365 . 55.4 33.3 441 .
36 897 1.7 99 30.9 17.4 7.9 106 .
8.4 422 15.1 9 26 3 23.5 11 60
20.6 17 0.6 6 55.8 0.2 . . .
7.4 448 16.4 19 28.1 0.8 10.3 4 .
15.7 661 1 48 . 9.6 8.1 471 .
41.7 . . . 0.2 . 0.5 5 .
2.2 333 0.2 139 169.2 6.4 50.8 316 525
4.4 249 0.3 37 . 18.2 3.6 79 .
5.8 705 6.8 45 3.5 1 4.9 209 .
2.4 138 3.7 80 69 4.3 5.8 37 862
2.6 125 4 36 7.2 9 4.5 26 5369
5.8 166 3.8 59 16.6 4.7 5.9 21 1184
14.3 336 1.8 118 6.7 29.4 7.1 198 2522
1.1 106 . 138 918.4 5.7 13.8 33 2755

9.6 138 2.7 54 290.7 8.4 5.4 83 1912
8.5 87 1.7 173 86.8 1.2 4.3 55 57
12.8 99 2.5 154 85.7 3.9 4.3 65 257
17.4 1055 3.7 459 5.1 26.9 38.2 93 .
26.9 1691 11.4 792 . 38.4 24.6 217 .
];
variable x{f} "dollars of food f to be purchased daily
(dollars)";
constraint nb{n} "nutrient balance (units)": sum{f} (a
[f,n]*x) >= b;
minimize cost "total food bill (dollars)": sum{f} x;
Writep(cost,x);
end

5.19. Fertilizer Production (egypt)

Problem: This is a one-period model of the Egyptian fertilizer industry devel-
oped at the World Bank. (for a GAMS representation see also: BISSCHOP J.,
MEERAUS A., On the Development of a General Algebraic Modeling System
in a Strategic Planning Environment. Development Research Department,
Mathematical Programming Study 20, pp.1-29, 1982. and CHOKSI A.M.,
MEERAUS A., STOUTJESDIJK, The Planning of Investment Programs in the
Fertilizer Industry, Johns Hopkins University Press, Batimore, 1980.
Cited also in: FOURER R., A Modeling Language for Mathematical Program-
ming, 1987
A model formulation is as follows:
The Model
function Egypt
set Center (Ce) , P (i, j) “Plants”
set Port (Po) , Region (R) , Units (Un) , Process (Pr) , Nutr (N) ,
Commod (C, c) , cRaw (r) , cInter , cShip (s) , cFinal (f)
param exch := 0.4
param utilPct := 0.85
set pExceptP,Pr
param cf75R,cFinal , fncFinal,N , roadR,Ce , railHalfP,P , impdBargP
, impdRoadP , ioC,Pr , utilUn,Pr , pImpC , pRcRaw , pPrP,cRaw ,
dcapP,Un , raili,j := if(railHalfi,j , railHalfi,j , railHalfj,i )
param tranFinalP,R := 0.5 + 0.0144 · roadR,PwithinCe
param tranImportR,Po := 0.5 + 0.0144 · roadR,PowithinCe
param tranInteri,j := 3.5 + 0.03 · raili,j
param tranRawP := if(impdBargP , 1+0.003·impdBargP )+if(impdRoadP , 0.5+
0.144 · impdRoadP )
param pDomP,cRaw := if(pRcRaw , pRcRaw , pPrP,cRaw )
param icapUn,P := 0.33 · dcapP,Un
set mPosP,Un := icapUn,P > 0
set pCapP,Pr := ∀Un|utilUn,Pr >0 mPosP,Un
P P,Pr ∧ ¬pExceptP,Pr
set pPosP,Pr := pCap
set cpPosC,P := io >0
P Pr|pPosP,Pr C,Pr
set ccPosC,P := Pr|pPosP,Pr io C,Pr < 0
set cPosC,P := cpPosC,P ∨ ccPosC,P
var ZP,Pr|pPosP,Pr , XfcFinal,P,R|cpPoscFinalwithinC,P , Xis,i,j|cpPosswithinC,i ∧ccPosswithinC,j
, VfcFinal,R,Po , VrcRaw,P|ccPoscRawwithinC,P , UcRaw,P|ccPoscRawwithinC,P ∧pDomP,cRaw >
, Psip , Psil , Psii
P P P
s.t.PmbdN,R : cFinal,Po fncFinal,N ·( Po VfcFinal,R,Po + P XfcFinal,P,R ) ≥
cFinal fncFinal,N · cf75 P
R,cFinal P
s.t. mbdbcFinal,R|cf75R,cFinal >0 : Po VfcFinal,R,Po + P XfcFinal,P,R ≥
cf75R,cFinal
P P P
s.t. mbc,i : PrPioc,Pr ·Zi,Pr + j Xicwithins,i,j + j Xicwithins,j,i +if(pImpc , Vrcwithinr,i )+
Ucwithinr,i ≥ R XfP cwithinf,i,R
s.t. ccP,Un|mPosP,Un
P : Pr utilUn,Pr · ZP,Pr ≤ utilPct · icapUn,P
s.t. ap : Psip =P cRaw,P pDomP,cRaw · UcRaw,P P
s.t. al : Psil = cFinal,P,R
P tranFinalP,R ·XfcFinal,P,R
P + cFinal,Po,R tranImportR,Po ·
P tranInteri,j ·Xis,i,j + cRaw,P tranRaw
VfcFinal,R,Po + s,i,j P P ·VrcRaw,P
s.t. ai : Psii/exch = f,R,Po pImpfwithinC · Vff,R,Po + r,P pImprwithinC ·
Vrr,P
minimize Psi := Psip + Psil + Psii
Writep(Psi, Z, Xf, Xi, Vf, Vr, U, Psip, Psil, Psii)
Writep(mPos, pCap, pExcept, pPos, cpPos, ccPos, cPos)
end
5.20. House Plan Design (house)

Problem: This problem designs the size of an L shaped house meeting limita-
tions imposed by city codes and aesthetic considerations.
This model is from the GAMS library (GAMS SEQ=99) and from Borland,
Eureka: The Solver. Tech. rep., Borland International, 1987.
Modeling Steps

model HOUSE "House Plan Design";
variable
x := 30 "width of front wing (ft)";
y "length of front wing (ft)";
z "length of second story (ft)";
b [40..68] := 68 "total house width (ft)";
a "length of back wing (ft)";
l [56..100] "total house length (ft)";
a1 [0..3000] "area of first floor (sq ft)";
a2 "area of the second floor (sq ft)";
constraint
defa1 "definition of first floor area (sq ft)": a1 =
x*y+a*b;
defa2 "definition of second floor area (sq ft)": a2 =
x*z;
maxw "front wing maximum width (ft)": b/3<=x<=b/2;
minp "minimum pool area (sq ft)": y*(b-x) >= 1500;
defl "definition of total house length(ft)": l=y+a;
balk "balkoney size restriction (ft)": z = a+y/2;
prop "wing proportionality (ft)": a >= y/2;
maximize obj "total area of the house (sq ft)": a1+a2;
Writep(obj);
Draw.Scale(7,7);
Draw.Line(0,0,b,0); Draw.Text('b='&b,b/3,1);
Draw.Line(0,0,0,l); Draw.Text('l='&l,1,l/2);
Draw.Line(b,0,b,a); Draw.Text('a='&a,b,a/2);
Draw.Line(0,l,x,l); Draw.Text('x='&x,x/3,l-2);
Draw.Line(x,l-y,x,l);Draw.Text('y='&y,x/2,a+(l-y)/1.5);
Draw.Line(x,l-y,b,a);
Draw.Line(0,l-y/2,x,l-y/2); Draw.Text('Balcony',2,l-y
/2+2);
Draw.Text('Swimming Pool',x+2,l-y/2+2);
end

5.21. Tanglewood Chair Manufacturing (tangle)

Model from : MATURANA S. V., Survey and Analysis of LINGO, Mgt 298D
Course Project; John E. Anderson Graduate School of Management, Univer-
sity of California, Los Angeles, CA 90024, p.17f
Problem: The Tanglewood Chair Manu. Co. has four plants located around
the country. The fabrication data are the following:
Plant Cost per chair Max production Min production

P1 5 500 0
P2 7 750 400
P3 3 1000 500
P4 4 250 250
The wood required to make each chair is twenty pounds and is obtained from
two suppliers. The company purchase of at least 8 tons of wood per month
from each supplier. The cost of wood is 0.10/lb from supplier 1 and 0.075/lb
from supplier 2. The shipping cost in dollar per pound of wood from each
supplier to each plant is shown below:
P1 P2 P3 P4
sup1 0.01 0.02 0.04 0.04
sup2 0.04 0.03 0.02 0.02
The chairs are sold in New York, Houston, San Francisco and Chicago. The
transportation costs are shown below:
NY H SF C
P1 1 1 2 0
P2 3 6 7 3
P3 3 1 5 3
P4 8 2 1 4
Following data are also available:
City Selling price maximum demand minimum demand

NY 20 2000 500
H 15 400 100
SF 20 1500 500
C 18 1500 500
Where should each plant buy its raw materials? How many chairs should be
made at each plant? How many chairs should be sold at each city? Where
should each plant ship its product?
5.22. A Travel Optimization Problem (travel)

SCICON Ltd. (or Spectrum, 1984, 6, p.125-130 References
5.23. A Transportation Problem (trnsport)

Problem: This problem finds a least cost shipping schedule that meets re-
quirements at markets and supplies at factories. From Gams model library
SEQ=1. [14], chap 3.3, and [26], chap 2.

model trnsport "A Transportation Problem";
set
i "canning plants" := [ seattle, 'san-diego' ];
j "markets" := [ 'new-york', chicago, topeka ];
parameter
a{i} "capacity of plant i in cases" :=
/ seattle 350, 'san-diego' 600 /;
b{j} "demand at market j in cases" :=
/ 'new-york' 325, chicago 300, topeka 275 /;
d{i,j} "distance in thousands of miles" := [
2.5 1.7 1.8
2.5 1.8 1.4 ];
f "freight in dollars per case per thousand miles" :=
90;
c{i,j} := f*d/1000 "transport cost in thousands of
dollars per case";
variable x{i,j} "shipment quantities in cases";
constraint
supply{i}: sum{j} x <= a "observe supply limit at
plant i";
demand{j}: sum{i} x >= b "satisfy demand at market j
";
minimize cost: sum{i,j} c*x "total transportation costs
in thousands of dollars";
Writep(cost,x);
end

C HAPTER 6
B ILLONNET
6.1. A MIQP problem (Bill018)

Problem: Model and solve the following iQP-problem (problem with a quadratic
objective function and integer variables).
X
min (xi − ai )2
X
i∈I
subject to bi xi ≤ K
i∈I
xi ∈ Z for all i ∈ I = {1 . . . 5}
with K = 50, a = {30, 20, 4, 8, 16}, and b = {3, 2, 5, 7, 1} (see [1] p.18).
Modeling Steps

model p9 "A MIQP problem";
set i := [1..5];
parameter a{i} := [30 20 4 8 16];
b{i} := [ 3 2 5 7 1];
integer variable x{i} [-50..50];
constraint C1: sum{i} b*x <= 50;
minimize obj: sum{i} (x-a)^2;
Writep(obj,x);
end

395
CHAPTER 6. BILLONNET 396
6.2. Localisation Optimal (Bill023)

Problem: A company wants to build m new factories in a regions located
at m <= n of n potential sites to serve its new p costumers with the new
product. The demand qj of each customer are given as well as the capacity
Nk in each factory, with j ∈ J = {1 . . . p} and k ∈ K = {1P . . . m}. Since
P we will
not generate stock the following equation must hold: j∈J qj = k∈K Nk ,
that is, the capacity is chosen in a way to just cover all the demand.
Furthermore, the cost are given as ck,i for building the factory k at site i, with
k ∈ K = {1 . . . m} and i ∈ I = {1 . . . n}, as well as the distances (in km) between
the site i and the costumer j, which is di,j . The unit cost for one km is C = 6
Euros.
At which sites the factories should be built in order to minimize overall costs
(building costs and transportation costs)? (see [1] p.22).
Modeling Steps
We introduce a binary variable xk,i which is 1 if the factory k is built at site

i and 0 otherwise. Furthermore, we need a variable yi,j ≥ 0 for the quan-
tity transported from site i to costumer j in order to be able to formulate the
transportation costs.
1. Now we can express the total costs: (1) the buildung costs are:
X
ck,i xk,i
k∈K,i∈I
(note that xk,i is one if a factory k is built at site i). (2) the transportation
costs are: X
Cdi,j yi,j
i∈I,j∈J
The total should be minimal.

2. A factory k can only be built at exactly one site i, hence we have con-
straint X
xk,i = 1 forall k ∈ K
i
3. At each site i we only build at most one factory k, hence constraint

X
xk,i ≤ 1 forall i ∈ I
k
397 CHAPTER 6. BILLONNET
4. The demand of each customer must be exactly fulfilled:

X
yi,j = qj forall j ∈ J
i
5. At each site i the quantity produced must be the quantity transported

to the different customers j (no stock):
X X
yi,j = Nk xk,i forall i ∈ I
j k

model p8 "Localisation Optimal";
set i := [1..5] "locations";
j := [1..9] "clients";
k := [1..3] "factories";
parameter c{k,i} "costs" := [
2000 2500 3000 2000 2800
1500 2000 1800 2100 1800
2300 3000 2500 1800 2600 ];
d{i,j} "distances" := [
250 250 200 450 650 700 500 600 600
300 200 550 750 750 600 350 350 200
500 300 300 300 350 350 350 450 550
650 400 350 200 200 350 450 550 700
700 400 700 700 450 200 150 150 400 ];
N{k} "capacity" := [1800 1200 2200];
q{j} "demand" := [550 370 540 660 500 480 750 650
700];
C "unit costs" := 6;
binary variable x{k,i} "factory k is built at i?";
variable y{i,j} "quantity of product";
constraint
C1{k}: sum{i} x = 1;
C2{i}: sum{k} x <= 1;
C3{j}: sum{i} y = q;
C4{i}: sum{j} y = sum{k} N*x;
minimize cost: sum{k,i} 1000*c*x + sum{i,j} C*d*y;
Writep(cost,y);
end

6.3. Optimal Portefolio (Bill026)

Problem: We want to invest a certain amount of money into a subset of a
given number n of different titles. The set of titles is I = {1 . . . n}. Let ri be the
expected earnings in percent and vi the variance of these earnings. (The vari-
ance is a measure of risk for a title: the larger the variance the larger the risk
that the realized earning will deviate substantially from the expected earn-
ings). We decide to invest in exactly K < n titles. We suppose that the earn-
ings of the different titles are independent from each others. Furthermore, we
desire to have a yield of at least ρ. In addition, there is a minimal loi and
maximal part upi of the whole investment that can be placed into one single
title.
What is the part of each title in the portfolio such that the total risk is minimal?
(see [1] p.25ff).
Modeling Steps
We introduce a variable xi which is the part (percentage of the whole amount)

that is to be placed on title i. We also introduce a binary variable yi saying
whether the title i is used/not used in the final portfolio.
1. The sum of all invested parts must be one, this is formulated by the
constraint C1.
2. The yield must be at least ρ, The earning on a title is the expected rate
times the invested part of this title. The sum must be ρ. This is con-
straint C2.
3. The number of titles in the final portfolio is K. This gives constraint C3.
4. The parts must be between the minimal and maximal requirements.
loi yi is the minimal requested amount to be placed in a title i: if yi = 0
(the part of i is zero), then loi yi = 0, if yi = 1 (there is a positive part
of i in the portfolio), then loi yi is loi . upi yi is the maximal requested
amount to be placed in a title i: if yi = 0, then upi yi = 0, if yi = 1 then
upi yi is upi . Hence xi should be between loi yi and upi yi : if yi = 0
then xi = 0 anyway, if yi = 1 then xi is between loi and upi . That is
exactly what is requested. The constraint is C4.
5. Finally, we want to minimize the risk. The total risk is measured by
the sum of the (positive and negative) deviations from the expected
eanrings on each title. The deviation of a single title i can be measured
by vi x2i . The quadratic term makes sure that the risk increased with an
increasing (positive or negative deviation. Hence the sum of these risks
is minimized.

model p11 "Optimal Portefolio";
set i := [1..10] "titles";
parameter r{i} "Expected earning" := [
5 9 3 12 5 4 8 2 4 9 ];
v{i} "Variance" := [
30 145 7 180 34 10 100 6 10 120 ];
rho := 3 "Requested interest yield" ;
K := 4 "Number of different titles in the
portfolio";
lo{i}:= 0.01 "minimal proportion of a title";
up{i}:= 1 "maximal proportion of a title";
variable x{i} "Part attributed in the portfolio";
binary variable y{i} "Title used/not used in the
portfolio";
constraint
C1: sum{i} x = 1;
C2: sum{i} r*x >= rho;
C3: sum{i} y = K;
C4{i}: lo*y <= x <= up*y;
minimize risk: sum{i} v*x^2;
Writep(risk,y,x);
end

6.4. Optimal Positioning on the Market (Bill028)

Problem: Consider a number m of products and a number of potential cus-
tomers n. Each product has a given number of attributes out of all p attributes.
Let the set of products, customers, and attributes be J = {1 . . . m}, I = {1 . . . n},
and K = {1 . . . p}. We measure the “value” of an attribute k of product j by a
positive number dj,k . Each customer has a given concept of an “ideal” prod-
uct in mind, which is expressed by the values of its attributes. These data
are stored in the table zi,k and the costumer also assigns a given weight wi,k
to each attribute. We suppose that each customers chooses the product that
corresponds the best to his ideal. The distance
qP between a product j and the
“ideal” product for customer i is given by 2
k∈K wi,k (dj,k − zjk ) . We de-
note the minimum over all products for a customer i as Ri . Hence
sX
Ri = wi,k (dj,k − zjk )2
k∈K
A company now wants to design a new product with the attributes such that
it maximises the profit by attracting as many customer as possible supposing
that the company gets a revenu ci if the customer is attracted by the new
product.
Furthermore, we suppose that the introduced product with characteristic val-
ues of xk with k ∈ K costs f(x1 , x2 , . . . , xp ). Let this function be given by the
quadratic expression as follows (p = 5):
f(x1 , x2 , . . . , xp ) = 0.6x21 − 0.9x2 − 0.5x3 + 0.1x24 + x5
In addition the characteristcs of the product are only possible under the fol-
lowing linear constraints Φ:
x1 − x2 + x3 + x4 + x5 ≤ 10
0.6x1 − 0.9x2 − 0.5x3 + 0.1x4 + x5 ≤ −0.64
x1 − x2 + x3 − x4 + x5 ≥ 0.69
0.157x1 + 0.05x2 ≤ 1.5
0.25x2 + 1.05x4 − 0.3x5 ≥ 4.5
2 ≤ x1 ≤ 4
0 ≤ x2 ≤ 8
3 ≤ x3 ≤ 9
0 ≤ x4 ≤ 5
4 ≤ x5 ≤ 10
What characteristics should this product contain and which customers will be
attracted by it? (see [1] p.27ff)
Modeling Steps
We introduce a variable xk with the characteristic “value” for attribute k and

a binary variable yi saying that the customer i is/is not attracted by the prod-
uct.
1. A customer i is attracted by the product, if the product fulfills the fol-

lowing condition: X
wi,k (xk − zi,k )2 ≤ Ri
k∈K
We can formulated this condition for each customerPi by the follow-

ing constraints (where h is a large number, such that k∈K wi,k (dj,k −
zjk )2 − h ≤ Ri ):
X
wi,k (xk − zi,k )2 − h(1 − yi ) ≤ Ri
k∈K
2. The other constraints are given by the possible charactreistics Φ.

P
3. We need to maximize the profit, which is the revenue (= i ci yi ) and
the costs given by f(x1 , x2 , . . . , xp ).

model p11 "Optimal Positioning on the Market";
set i := [1..25] "customers";
set j := [1..10] "products";
set k := [1..5] "attributes";
parameter d{j,k} := [
0.62 5.06 7.82 0.22 4.42
5.21 2.66 9.54 5.03 8.01
5.27 7.72 7.97 3.31 6.56
1.02 8.89 8.77 3.10 6.66
1.26 6.80 2.30 1.75 6.65
3.74 9.06 9.80 3.01 9.52
4.64 7.99 6.69 5.88 8.23
8.35 3.79 1.19 1.96 5.88
6.44 0.17 9.93 6.80 9.75
6.49 1.92 0.05 4.89 6.43 ];
z{i,k} := [
2.26 5.15 4.03 1.74 4.74
5.51 9.01 3.84 1.47 9.92
4.06 1.80 0.71 9.09 8.13
6.30 0.11 4.08 7.29 4.24
2.81 1.65 8.08 3.99 3.51
4.29 9.49 2.24 9.78 1.52
9.76 3.64 6.62 3.66 9.08
1.37 6.99 7.19 3.03 3.39
8.89 8.29 6.05 7.48 4.09

7.42 4.60 0.30 0.97 8.77
1.54 7.06 0.01 1.23 3.11
7.74 4.40 7.93 5.95 4.88
9.94 5.21 8.58 0.13 4.57
9.54 1.57 9.66 5.24 7.90
7.46 8.81 1.67 6.47 1.81
0.56 8.10 0.19 6.11 6.40
3.86 6.68 6.42 7.29 4.66
2.98 2.98 3.03 0.02 0.67
3.61 7.62 1.79 7.80 9.81
5.68 4.24 4.17 6.75 1.08
5.48 3.74 3.34 6.22 7.94
8.13 8.72 3.93 8.80 8.56
1.37 0.54 1.55 5.56 5.85
8.79 5.04 4.83 6.94 0.38
2.66 4.19 6.49 8.04 1.66 ];
w{i,k} := [
9.57 2.74 9.75 3.96 8.67
8.38 3.93 5.18 5.20 7.82
9.81 0.04 4.21 7.38 4.11
7.41 6.08 5.46 4.86 1.48
9.96 9.13 2.95 8.25 3.58
9.39 4.27 5.09 1.81 7.58
1.88 7.20 6.65 1.74 2.86
4.01 2.67 4.86 2.55 6.91
4.18 1.92 2.60 7.15 2.86
7.81 2.14 9.63 7.61 9.17
8.96 3.47 5.49 4.73 9.43
9.94 1.63 1.23 4.33 7.08
0.31 5.00 0.16 2.52 3.08
6.02 0.92 7.47 9.74 1.76
5.06 4.52 1.89 1.22 9.05
5.92 2.56 7.74 6.96 5.18
6.45 1.52 0.06 5.34 8.47
1.04 1.36 5.99 8.10 5.22
1.40 1.35 0.59 8.58 1.21
6.68 9.48 1.60 6.74 8.92
1.95 0.46 2.90 1.79 0.99
5.18 5.10 8.81 3.27 9.63
1.47 5.71 6.95 1.42 3.49
5.40 3.12 5.37 6.10 3.71
6.32 0.81 6.12 6.73 7.93 ];
c{i} := [1 .2 1 .2 .9 .9 .1 .8 1 .4 1 .3
.1 .3 .5 .9 .8 .1 .9 1 1 1 .2 .7 .7];
R{i} := min{j} (sum{k} w*(d-z)^2);
h := 2000;
lo{k} := [2 0 3 0 4];
up{k} := [4 8 9 5 10];
variable x{k} [lo..up] "Characteristic value of
attribute k";
binary variable y{i} "Customer attracted or not";
constraint
C1{i}: sum{k} w*(x-z)^2 - h*(1-y) <= R;
C2: x[1] - x[2] + x[3] + x[4] + x[5] <= 10;
C3: 0.6*x[1] -0.9*x[2] - 0.5*x[3] + 0.1*x[4] + x[5]
<= -0.64;
C4: x[1] - x[2] + x[3] - x[4] + x[5] >= 0.69;
C5: 0.157*x[1] + 0.05*x[2] <= 1.5;
C6: 0.25*x[2] + 1.05*x[4] - 0.3*x[5] >= 4.5;
expression Fx: 0.6*x[1]^2 - 0.9*x[2] - 0.5*x[3] + 0.1*x
[4]^2 + x[5];
maximize profit: sum{i} c*y - Fx;
Writep(profit,y,x);
end

6.5. An iQP Problem: Clustering (Bill035)

Problem: Formulate the following problem as a convexe problem and solve
it, where n > 0 ci,j ∈ R, Ax = b expresses a system of linear equations.
X
n X
n
min ci,j · xi · xj
i j
subject to Ax = b
xi ∈ {0, 1} forall i ∈ I = {1 . . . n}
(see [1] p.18).
A concrete application is clustering: Given n locations with distances of ci,j

(i, j ∈ {1 . . . n}) between them. Find a subset of size k < n for which the total
distance between all in the subset is minimal.
This problem is in general not convex, depending on matrix c. However, it can
be reformulated as a convex problem and then be solved by standard solvers
(gurobi, cplex). This is done by replacing (with λ sufficiently large):
X
n X
n X
n X
n X
n
ci,j · xi · xj with ci,j · xi · xj + λ (x2 − x)
i j i j i
Note that the two expressions are identical for x ∈ {0, 1}, since in this case we
have: x2i = xi .
Modeling Steps
As noted the application can be formulated as a model with a quadratic ob-

jective. The interpretation of the variable xi is a s follows: i is in the subset if
xi = 1 otherwise it is not. Hence,
X
n X
n
ci,j · xi · xj
i j
is an expression of the total distance within the subset, which is to be mini-

mized. The single constraint in the clustering application is
X
n
xi = k
i
For a concrete problem, let n = 6, k = 3, and

 
0 1 7 3 9 3
1 0 5 2 1 3
 
7 5 0 1 7 8
ci,j =
3

 2 1 0 5 1

9 1 7 5 0 2
3 3 8 1 2 0
with λ = 15 the objective function becomes convex.

model p19 "An iQP Problem: Clustering";
set i,j := [1..6];
integer parameter c{i,j} := [
0 1 7 3 9 3
1 0 5 2 1 3
7 5 0 1 7 8
3 2 1 0 5 1
9 1 7 5 0 2
3 3 8 1 2 0 ];
lambda := 15;
k := 3;
binary variable x{i};
constraint C1{i}: sum{i} x = k;
minimize obj: sum{i,j} c*x[i]*x[j] + lambda*(sum{i}(x
^2-x));
Writep(obj,x);
end

6.6. Fraction Objective (Bill046)
Problem: The fractional LP is a model where the constraints are linear and the
objective function consists of a fraction with the nominator and the denomi-
nator are both linear functions. The general model is as follows:
Pn
b0 + bj xj
max Pj=1
n
c0 + j=1 cj xj
X
n
subject to ai,j xj ≤ di forall i ∈ I = {1 . . . m}
j=1
xj ≥ 0 forall j ∈ J = {1 . . . n}
We suppose that c0 is strictly positive and b0 , bj , and ai,j are non-negative.

Transform this problem into a linear problem.
As an example we have:
3 + 2x1 + 5x2 + x3
max
1 + 3x1 + x2 + 4x3
subject to 5x1 + x2 + 6x3 ≤ 10
x1 + 2x2 + x3 ≤ 2
xj ≥ 0 forall j ∈ {1, 2, 3}
Solve this problem. (see [1] p.46).
Modeling Steps
We define the following two expressions:
1
z= Pn
c0 + j=1 cj xi
x
yj = Pnj
c0 + j=1 cj xi
now we substitute them into the model to obtain the following linear model:
X
n
max b0 z + bj y j
j=1
Xn
subject to c0 z + cj yj = 1
j=1
X
n
ai,j yj ≤ di z forall i ∈ I = {1 . . . m}
j=1
yj ≥ 0 forall j ∈ J = {1 . . . n}
z≥0
Applications are: maximize a ratio, for example, maximize the productivity,

that may be expressed by the ratio of output (a linear expression) to input
(also a linear expression). Another is to maximize the investment of return,
a.o.

model FractLp "Fraction Objective";
variable y1; y2; y3; z;
constraint
A: z + 3*y1 + y2 + 4*y3 = 1;
B: 5*y1 + y2 + 6*y3 <= 10*z;
C: y1 + 2*y2 +y3 <= 2*z;
maximize obj: 3*z + 2*y1 + 5*y2 + y3;
Write('obj: %3.1f \n x1 = %3.1f , x2 = %3.1f , x3 =
%3.1f\n', obj,y1/z,y2/z,y3/z);
end

Further Comments: In general, we can translate a linear-fractional program

(LFP) in an equivalent linear program (LP), as follows. The LFP is as follows:
cT x + d
min
eT x + f
subject to eT x + f > 0
Gx ≤ h
The equivalent LP is:
min cT y + dz
subject to Gy − hz ≤ 0
eT y + fz = 1
z≥0
If x∗ , and y∗ , z∗ are the optimal points of these two problems then
x∗ = y∗ + z∗ , y∗ = x∗ /(eT x∗ + f) , z∗ = 1/(eT x∗ + f)
6.7. Optimal Portefolio (Bill053)

Problem: We want to invest a certain amount of money into a subset of a
given numberr n of different titles. The set of titles is I = {1 . . . n}. Let ri
be the expected earnings in percent and vi,j the covariance of these earnings.
(The covariance is a measure of risk for a title: the larger the covariance the
larger the risk that the realized earning will deviate substantially from the
expected earnings). Furthermore, we desire to have a yield of at least ρ. What
is the part of each title in the portfolio such that the total risk is minimal? (see
[1] p.54ff). Compare this model with the model Bill26. What is the difference?
Modeling Steps
We introduce a variable xi which is the part (percentage of the whole amount)

that is t be placed on title i. We also introduce a binary variable yi saying
whether the title i is used/not used in the final portfolio.
1. The sum of all invested parts must be one, this is formulated by the
constraint C1.
2. The yield must be at least ρ, The earning on a title is the expected rate
times the invested part of this title. The sum must be ρ. This is con-
straint C2.
3. Finally, we want to minimize the risk. The total risk is measured by
the sum of the (positive and negative) deviations from the expected
eanrings on each title. The deviation of a single title i can be measured
by vi x2i . The quadratic term makes sure that the risk increased with an
increasing (positive or negative deviation. Hence the sum of these risks
is minimized.

model p11 "Optimal Portefolio";
set i,j := [1..5] "titles";
parameter r{i} "Expected earning" := [
4 2 3 3 4 ];
v{i,j} "Variance" := [
21.5 -17.25 -2.75 2.75 -14.75
-17.25 56.5 -0.25 0.25 -3.5
-2.75 -0.25 0.5 -0.5 3.25
2.75 0.25 -.5 0.5 -3.25
-14.75 -3.5 3.25 -3.25 49.5 ];
rho := 3.75 "Requested interest yield" ;
variable x{i} [0..10000] "Part attributed in the
portfolio";
constraint
C1: sum{i} x = 1;
C2: sum{i} r*x >= rho;
minimize risk: sum{i,j} v[i,j]*x[i]*x[j];
Writep(risk,rho,x);
end

6.8. Piecewise Linear Function (Bill227)

Problem: This is a small model to show the syntax of piecewise linear function
in LPL. Let the following small problem be given:
min f(y) + z1 − 2z2 + z3 − 2y

subject to 4.5y + z1 ≤ 27.5
y + 2f(y) + 2z1 + 2z2 ≤ 24
1 ≤ y ≤ 7 , z1 ≤ 6 , z2 ≤ 4
z1 , z2 , z3 ≥ 0
where f(y) is a non-linear function that can be approximated by a piecewise

linear function defined in the interval [1, 7] and given by the 4 points [y, f(y)]
as follows: [1, 3], [3, 5], [5, 3], [7, 5]. Solve the problem by transforming it into a
integer linear model. The model is from [1] p.227).
Modeling Steps
Let f(y) be a non-linear function approximated by k = {1 . . . K} points as fol-

lows:
[ak , f(ak )] with 0 ≤ a1 ≤ . . . ≤ aK
To translate the piecewise linear function into a linear model the following
steps have to be processed:
1. Add a variable e ≥ 0 and replace each occurrence of f(y) by e.

2. Add a variable vector:
λk ≥ 0, k = {1 . . . K}
3. Add the following four constraints to the model:

X
e= λk f(ak )
Xk
y= λk a k
X k
λk = 1
k
Sos2(λk ) forall k ∈ K
Sos2(< variable − list >) is a special constraint saying that exactly 2 consec-
utive variables in the list are different from zero, all others must be zero. (For
more theoretical background see [16], Chapter 9.3.)
A Sos2(λk , ∀ k = {1 . . . K}) can be transformed into linear constraints as fol-
lows:
1. Add a binary variable vector:
xk ≥ 0, k = {1 . . . K − 1} with xk ∈ {0, 1}
2. Add the following constraints:
λ
X k ≤ xk + xk+1 , ∀ k = {1 . . . K − 1}
xk = 1
k
In LPL, the piecewise linear function is added using the << ... >> notation.
LPL then translates the model autoamtically into an integer linear model:

model Bill4_18 "Piecewise Linear Function";
set k:=[1..4] "pieces";
parameter a{k}:=[1,3,5,7];
b{k}:=[3,5,3,5];
variable y [1..7]; z1 [0..6]; z2 [0..4]; z3; e;
constraint
A: 4.5*y + z1 - z3 <= 27.5;
B: y + 2*e + 2*z1 + 2*z2 <= 24;
C: e = <<y, {k}(a,b)>>;
minimize obj: e + z1 - 2*z2 + z3 -2*y;
Writep(obj);
end

The constraint defining the piecewise linear function in LPL is:

C: e = <<y, {k}(a,b)>>;
This notation could also be explicitly added to the model by replacing the
constraint with the following piece of code in LPL:
-- the following is equivalent to the <<piecewise-
linear-code>>
variable la{k};
constraint
C1: e = sum{k} la*b;
C2: y = sum{k} la*a;
C3: sum{k} la = 1;
C4: Sos2({k} la);
The Sos() constraint (C4) can also be replaced by the linear constraints as
follows:
-- the following three instructions are equivalent to

the Sos2() constraint
binary variable x{k|k<#k};
constraint C41{k}: la <= x[k] + x[k+1];
constraint C42 : sum{k} x = 1;
A more involving model using piecewise linear functions is given by the

model Bill3201 .
1
http://lpl.virtual-optima.com/lpl/Solver.jsp?name=/Bill320
6.9. A Cost Flow Problem (bill320)

Problem: Consider a graph G = (V, A) with a set of vertices V = {1 . . . n} and
a set of arcs A ⊆ V × V. The vertex 1 ∈ V is called source and the vertex n ∈ V
is called terminal (or sink). Each arc (i, j) ∈ A is equiped with a capacity ui,j
and an unknown flow xi,j ∈ [0, . . . , ui,j ]. The cost of the flow from i to j is
given by functions fi,j (xi,j ), and the total cost of flow is defined as
X
Γ (x) = fi,j (xi,j ) · xi,j
(i,j)∈A
.
Let fi,j (xi,j ) be piecewise linear fonctions (for each (i, j) ∈ A there is one piece-
wise linear function) defined by points as follows:
(xi,j , fi,j (xi,j ))
The p points with k = {1 . . . p} for each piecewise linear function (i, j) ∈ A are
given by the data as follows:
{(bi,j,k , di,j,k) )} forall k = {1 . . . p}
Find a flow of size ω from the source (node 1) to the sink (node n) through
the network such that the total costs are minimal. (This problem instance is
from [1] p.320, Problem 23).
Modeling Steps
The problem can be formulated as follows
X
min fi,j (xi,j ) · xi,j
(i,j)∈A
subject to 0 ≤
X xi,j ≤ ui,j X forall (i, j) ∈ A
xj,i = xi,j forall i ∈ V \ {1, n}
X
j|(i,j)∈A j|(i,j)∈A
x1,j = ω
j|(1,j)∈A
In the objective function we minimize the total flow costs. The flow on each
arc (i, j) should be in the range [0 . . . ui,j ]. For each arc (except the source
and the sink) the sum of the incoming flows must be the sum of the outgoing
flows. And finally, the sum of the flows outcoming from the source is fixed to
be ω.

model p23_2 "A Cost Flow Problem";
set i,j "Nodes";
k :=[1..4] "piecewise points";
e{i,j} "Arcs";
parameter u{i,j}; flow;
b{i,j,k}; d{i,j,k}; // piecewise points
variable x{i,j|e} [0..u];
constraint
A{i|i>1 and i<#i}: sum{j} x[i,j] = sum{j} x[j,i];
B: sum{j} x[1,j] = flow;
minimize cost: sum{e[i,j]} <<x,{k}(b,d)>>;
/* --S3.3.1. page 201 -- the piecewise linear ::
explicit formulation
variable lambda{e[i,j],k};
constraint C1{e[i,j]}: sum{k} lambda = 1;
constraint C2{e[i,j]}: x = sum{k} b*lambda;
constraint C3{e[i,j]}: Sos2({k} lambda);
/* --the Sos2 constraint :: explicit
binary variable y{e[i,j],k|k<#k};
constraint C31{e[i,j],k}: sum{k|k<#k} y = 1;
constraint C32{e[i,j],k}: lambda <= y[i,j,k-1] + y;
*/
--minimize cost: sum{e[i,j]} (sum{k} d*lambda);
* /
/* --S3.3.3 page 201 :: here is still another piece of
correct code
parameter T{e[i,j],k|k<#k} := (d[i,j,k+1]-d) / (b[i,j,k
+1]-b);
variable lambda{e[i,j],k|k<#k};
constraint C1{e[i,j],k|k<#k}: b*y <= lambda <= b[i,j,k
+1]*y;
constraint C2{e[i,j]}: x = sum{k|k<#k} lambda;
constraint C3{e[i,j]}: sum{k|k<#k} y = 1;
minimize cost: sum{e[i,j]} (sum{k|k<#k} (T*lambda + (d-
T*b)*y));
*/
/*
// This is still another way to formulate the complete
model as a MIP model:
model p23_2 "Minimal Cost Flow"; --- ok..That works
everything is explicit
set i,j "Nodes";
k :=[1..3] "piecewise intervals";
e{i,j} "Arcs";
parameter b{i,j,k}; d{i,j,k}; u{i,j};
flow := 50;
Read{i,j}('Bill320.dat,%1:Table', i,j,e, {k}b[i,j,k+1])
;
Read{i,j}('%2:Table', i,j, {k}d[i,j,k+1]);
u{i,j}:=sum{k} b;
variable t{i,j,k};
binary variable y{i,j,k};
constraint
C1{e[i,j]}: x[i,j] = sum{k} t[i,j,k];
C2{e[i,j],k|k<#k}: t[i,j,k] >= b[i,j,k]*y[i,j,k+1];
C3{e[i,j],k}: t[i,j,k] <= b[i,j,k]*y[i,j,k];
minimize cost: sum{e[i,j],k} d[i,j,k]*t[i,j,k];
write cost,x;
end
*/
end

Solution: Consider a network with 8 nodes (n = 8) (see Figure 6.1).
Figure 6.1: A Small Example Graph
The data and the solution are given in the Table 6.1. Further
Pdata are given as
follows: the flow is ω = 50, the capacity is given as ui,j = k bi,j,k .
As an example, the piece-wise linear function for the arc (7, 8) is given in
Figure 6.2.
Further Comments: The implementation of piece-wise linear function in LPL

is now explained. The formulation of a piecewise linear function in LPL is
straightforward. Suppose, we need to implement the following piecewise lin-
ear function as given be Figure [6.3]. The function is piecewise linear between
Arc (xi,j , f(xi,j ) (k = 4) ui,j xi,j

(1,2) (0,0) , (12,-108) , (15,-105) , (18,-78) 18 13
(1,3) (0,0) , (10,40) , (22,124) , (27,79) 27 27
(1,4) (0,0) , (3,6) , (10,62) , (14,78) 14 10
(2,5) (0,0) , (5,10) , (15,90) , (21,96) 21 21
(3,2) (0,0) , (8,-64) , (12,-60) , (24,48) 24 8
(3,4) (0,0) , (7,-49) , (17,-19) , (26,8) 26 11
(3,5) (0,0) , (11,77) , (20,140) , (31,195) 31 8
(3,6) (0,0) , (4,28) , (13,-62) , (16,-50) 16 0
(4,6) (0,0) , (12,36) , (18,-18) , (24,0) 24 20
(4,7) (0,0) , (8,72) , (17,72) , (24,51) 24 1
(5,8) (0,0) , (8,-48) , (17,-111) , (29,-219) 29 29
(6,8) (0,0) , (4,4) , (10,16) , (20,-84) 20 20
(7,6) (0,0) , (9,81) , (19,1) , (29,101) 29 0
(7,8) (0,0) , (11,44) , (20,80) , (25,40) 25 1
Table 6.1: Data and solution for the model
the following 6 given points:
{(0, 3) , (1, 6) , (2, 9) , (3, 4) , (4, 2) , (5, 3)}
In LPL we only need to specify the set of these points as follows:

set i := [1..6];
parameter a{i} := [0 1 2 3 4 5];
b{i} := [3 6 9 4 2 3];
Futhermore we defined a variable x – representing the horizontal axis.

variable x;
To use the piecewise linear function f(x) in an expression – e.g. in a constraint

– in LPL, we only need to specify the variable x and the points of f(x) in within
<< and >> as follows:
constraint A: ... + <<x,{i}(a,b)>> + ... ;
(For a simpler example of piecewise linear functions see model Bill2272 .)
LPL translates the piecewise linear construct as follows:

2
http://lpl.virtual-optima.com/lpl/Solver.jsp?name=/Bill227
Figure 6.2: A piecewise linear function of arc (7, 8)
1. The piecewise linear construct {i} <<x,{k}(a,b)>> is substituted

by the expression {i} (sum{k} la*b)
2. A new variable variable la{i,k} is added to the model.
3. Further constraints constraint A{i}: x = sum{k} la*a; are added.
4. Another set of constraint is added: constraint B{i}: sum{k} la = 1;
5. Finally, Sos2 constraints are added: C{i}: Sos2({k} la);
Hence in the previous LPL model the following minimizing function:

minimize cost: sum{e[i,j]} <<x,{k}(b,d)>>;
could simply be replaced by the following piece of LPL code – it’s exactly the
same:
constraint C1{e[i,j]}: sum{k} lambda = 1;
constraint C2{e[i,j]}: x = sum{k} b*lambda;
minimize cost: sum{e[i,j]} (sum{k} d*lambda);
To solve the problem, the Sos2 constraints must be interpreted by the solver.
LPL sends them directly to the solver, if it is capable to interpret them (this
is the case for Gurobi and Cplex). If the solver does not understand these con-
straints then they must be further linearized. In the previous model we must
replace the following constraint:
with the following code :

Figure 6.3: A piecewise linear function

constraint C31{e[i,j],k}: sum{k|k<#k} y = 1;
constraint C32{e[i,j],k}: lambda <= y[i,j,k-1] + y;
So, basically, we need to specify additional binary variables y that selects the
linear interval within the piecewise linear function and two additional con-
straint types. The complete mixed integer model of the flow model would
then be as follows – a model that is much more involving from the formula-
tion point of view3 :
model p23_2 "A Cost Flow Problem";
set i,j "vertices";
k :=[1..4] "piecewise points";
e{i,j} "Arcs";
parameter flow; u{i,j};
d{i,j,k}; b{i,j,k}; // piecewise points
3
The mathematical background of these linearization can be found in [1] p. 200ff,
227, and 320ff

constraint
C1{e[i,j]}: sum{k} lambda = 1;
C2{e[i,j]}: x = sum{k} b*lambda;
C31{e[i,j],k}: sum{k|k<#k} y = 1;
C32{e[i,j],k}: lambda <= y[i,j,k-1] + y;
minimize cost: sum{e[i,j]} (sum{k} d*lambda);
end
C HAPTER 7
M ISCELLANEOUS
7.1. A simple blending problem (alloy)

Problem: A small blending problem

model ALLOYING "A simple blending problem";
set
metals,m := [ A,B,C1,C2,D ];
ores, o := [ ore1 ore2 ore3 ore4 ore5 ore6 ];
parameter
bound{o} "availability of ore"
:= [1000 3000 2000 800 2600 1700];
boundu{o} "availability of ore"
:=0;
price {o} := [25 10 20 18 22 12 ];
purity{o} := [ 0.8 0.4 0.7 0.6 0.7 0.4];
alloy{m} := [ 0.18 -0.30 -0.40 0.60 0.02];
obtain{o,m} := [
0.20 -0.20 -0.40 0.40 0.00
0.15 -0.00 -0.20 0.20 0.05
0.00 -0.40 -0.30 0.30 0.00
0.10 -0.20 -0.30 0.30 0.00
0.10 -0.25 -0.25 0.25 0.10
0.05 -0.08 -0.17 0.17 0.10];
quantity,Q := 3000;
variable AMOUNT,X{o} [boundu..bound];
constraint
quareq: sum{o} purity*X = Q "Quantity to produce";
metreq{m}: sum{o}(alloy*purity-obtain)*X >= 0 "Metal
requirements";
minimize cost: sum{o} price*X "Minimize costs";
Writep(cost,X);
end

421
CHAPTER 7. MISCELLANEOUS 422
7.2. Assign Consulters (assign1)

Problem: A consulter company plans to distibute its consulter with different
knowledge into the different projects over the periods. Each project needs
various knowledge in the different time periods.
Modeling Steps

model assign1 "Assign Consulters";
set c "Consulters";
k "knowledge";
t "time periods";
p "projects";
CK{c,k} "c has the knowledge k";
PK{p,k,t} "project uses knowledge in time";
parameter level{c};
binary variable x{c,p,t};
constraint
R{c,t}: sum{p} x <=1 "In each period, a consulter can
only be with one project";
S{p,t,k|PK}: sum{c|CK} x>=1 "The project must be
adequatly accompanied";
minimize obj: sum{c,p,t} level*x "minimize the pool of
consulter";
model data;
c := [1..5];
k := [1..3];
t := [1..20];
p := [1..3];
CK{c,k} := [1 1, 1 2, 2 2, 2 3, 3 3, 4 1, 5 2, 5 1];
PK{p,k,t} := if(Rnd(0,1)>0.2,Trunc(Rnd(0,2)),0);
level{c} := [1 2 3 4 5];
end
model output;
Writep(PK,x);
end
end

423 CHAPTER 7. MISCELLANEOUS
7.3. Agriculture Production (crop)

Problem: see Katz S. (IBM Systems Journal 19:4,1980, p.505-520, Appendix B)

model CROP "Agriculture Production";
set
Month,m := [ May June July ];
Crop,c := [ Cotton Onion Pear Avoccado ];
Field,f := [ Cotton Onion ];
parameter
WaterBND{m} := [200000 260000 270000];
Labor{c} := [2.9 2.7 1.0 1.5];
Profit{c} := [6453 6110 4814 8813];
Ceils{c}:= [2000 250 500 800];
Land:=2700; FieldLand:=1850; LaborTot:=5850;
Water{c,m}:=[65 80 90 . 60 . . 53 64 . 75 85];
variable CropVar{c} [0..Ceils];
constraint
r1: sum{c} CropVar <= Land;
r2: sum{f} CropVar[f within Crop] <= FieldLand;
wat{m}: sum{c} Water*CropVar <= WaterBND;
r3: sum{c} Labor*CropVar <= LaborTot;
maximize prof: sum{c} Profit*CropVar;
Writep(prof,CropVar);
end

7.4. Import and Export Goods (export)

Problem: A country has four industries: Steel, Automotive, Electronics and
Plastics. There unit prices are: 500, 1500, 300 and 1200. The total available
Manpower is 830,000 year-hours. No more than 650,000 units of Automotive
and 60,000 units of Plastic can be produced. Production of 1 unit of Steel re-
quires 0.02 unit of Automotive, 0.01 units of Plastics, 250 currency units of
imported raw matrials and 0.5 manpower unit. Production of 1 unit of Au-
tomotive reqires 0.8 unit of steel, 0.15 units of electronics 0.11 units of plastic,
300 units of imported materials and 1 manpower unit. Production of 1 unit
of Electronics reqires 0.01 unit of steel, 0.01 units of automotive, 0.05 units of
plastic half a man years of labor and 50 units of imported materials. Produc-
tion of 1 unit of plastics requires .03 units of automotive production,
1. units of steel, 0.05 units of electronics, 2 man years of labor and

2. units of imported materials. The economic minister would like to max-
imize the exceeding export value.
Schrage L.,LINDO, 1986, p.116f.";

model EXPORT "Import and Export Goods";
set i,j := [ steel automotive electronic plastics];
parameter prices{i}:=[500 1500 300 1200];
ImportCosts{i}:=[250 300 50 300];
ManPowerUsed{i}:=[.5 1 .5 2];
Limits{i}:=[. 650000 . 60000];
ManPower := 830000;
Inputs{i,j} :=
[. .8 .01 .2 , .02 . .01 .03 , . .15 . .05 , .01
.11 .05 . ];
variable P{i} "PRODUCTION"; E{i} "EXPORTS";
constraint
A{i}: P = E + sum{j} Inputs[i,j]*P[j];
B: sum{i} ManPowerUsed*P = ManPower;
C{i|Limits}: P <= Limits;
maximize obj: sum{i} ( prices*E - ImportCosts*P );
Writep(obj,P,E);
end

7.5. A Small Portfolio Model (lin193)

Mode from SCHRAGE, Lindo , page 193.

model LIN193 "A Small Portfolio Model";
variable x; y; z; budget; yield; xfrac; yfrac; zfrac;
constraint
R1: 6*x+2*y-z + budget-1.3*yield+xfrac >= 0;
R2: 2*x+4*y-0.8*z + budget-1.2*yield+yfrac >= 0;
R3: -x-0.8*y+2*z + budget-1.08*yield+zfrac >= 0;
R4: x+y+z=1;
R5: 1.3*x+1.2*y+1.08*z >= 1.12;
R6: x <= 0.75;
R7: y <= 0.75;
R8: z <= 0.75;
minimize obj: x+y+z+budget+yield+xfrac+yfrac+zfrac;
Writep(obj,x,y,z,budget,yield,xfrac,yfrac,zfrac);
end

7.6. Exposures in Media Vehicles (media)

Problem: An advertising agency plans to select the media mix that will max-
imize the number of effective exposures subject to a set of constraints. There
are three media vehicles A, B and C. Vehicle A gives 3,100 (in thousands) ef-
fective exposures per issue , vehicle B gives 2,000, and vehicle C gives 2,400.
The media planner has an advertising budget of $500,000, which cannot be
exceeded. Vehicle A costs $15,000 per issue, vehicle B, $4,000, and vehicle C,
$5,000. Furthermore, the media planner wants to spend at least $250,000 on
vehicle A. Vehicle A puts out 52 issues a year, vehicle B, eight issues, and ve-
hicle C, twelve issues. The media planner wants to buy at least one issue of
vehicle B and six issues of vehicle C.

model MEDIA "Exposures in Media Vehicles";
set media := [A B C ];
parameter ExposureCost{media} := [ 15000 4000 5000 ];
Issues{media} := [ 3100 2000 2400 ];
PutsOutPerYear{media} := [ 52 8 12 ];
MinimalExposures{media} := [ . 1 6 ];
Budget := 500000;
MinimalBudgetForA := 250000;
variable NumberOfExposures{media} [MinimalExposures..
PutsOutPerYear];
constraint
r1: sum{media} ExposureCost*NumberOfExposures <=
Budget;
r2: ExposureCost['A']*NumberOfExposures['A'] >=
MinimalBudgetForA;
maximize number: sum{media} Issues*NumberOfExposures;
Writep(number,NumberOfExposures);
end

7.7. Mexican Steel Industry (mexican)

Model from MATURANA S. V., Survey and Analysis of LINGO, Mgt 298D

model Mexican "Mexican Steel Industry";
set
plants := [AHMSA FUNDIDOR SICARTSA HYLSA HYLSAP];
markets := [MexicoDF Monterey Guadalaj];
cf := [STEEL];
ci := [ SPONGE PIGIRON ];
cr := [PELLETS COKE NATGAS ELECTRIC SCRAP];
process :=[pigironp spongep steeloh steelel steelbof];
produnit := [blastfrn ofenhrth bof directed elecarc];
parameter
rd2{plants}:=[739 521 . 521 315];
dd{markets}:=[55 30 15];
rd3{markets}:=[428 521 300];
pv{cf}:=[150]; pe{cf}:=[140]; eb{cf}:=[1];
pd{cr}:=[18.7 52.17 14 24 105];
parameter
DT := 5.209;
rse := 40;
k{produnit,plants} := [ 3.25 1.40 1.1 0 0
1.5 0.85 0 0 0
2.07 1.5 1.3 0 0
0 0 0 0.98 1
0 0 0 1.13 0.56 ];
rd1{plants,markets} :=
[ 1204 218 1125, 1017 . 1030,
819 1305 704, 1017 . 1030,
185 1085 760 ];
b{produnit,process} :=
[ 1 0 0 0 0,0 0 1 0 0,0 0 0 0 1,0 1 0 0 0,0 0 0 1 0
];
a1{cr,process} := [ -1.58 -1.38 0 0 0
-0.63 0 0 0 0
0 -0.57 0 0 0
0 0 0 -0.58 0
0 0 -0.33 0 -0.12 ];
a2{ci,process} := [ 1.00 0 -0.77 0 -0.95
0 1.00 0 -1.09 0 ];
a3{cf,process} := [ 0 0 1.0 1.0 1.0 ];
variable
phipsi; philam; phipi; phieps;
z{process,plants};
u{cr,plants};
e{cf,plants};
d{cf,markets};
v{cf,markets};
x{cf,plants,markets};
constraint
r1{cf,markets}: d = DT*(1+rse/100)*dd/100;
mbf{cf,plants}: sum{process} a3*z >= sum{markets} x +e;
mbi{ci,plants}: sum{process} a2*z >= 0;
mbr{cr,plants}: sum{process} a1*z + u >= 0;
cc{produnit,plants}: sum{process} b*z <= k;
mr{cf,markets}: sum{plants} x+v >= d;
me{cf}: sum{plants} e <= eb;
r2: phipsi = sum{cr,plants} pd*u;
r3: philam = sum{cf,plants,markets} if(rd1 , (0.0084*
rd1+2.48)*x)
+ sum{cf,plants} if(rd2,(0.0084*rd2+2.48)*e)
+ sum{cf,markets} if(rd3,(0.0084*rd3+2.48)*v);
r4: phipi = sum{cf,markets} pv*v;
r5: phieps = sum{cf,plants} pe*e;
minimize phi: phipsi + philam + phipi - phieps;
Writep(phi);
end

7.8. A Production Planning Model (omp)

Problem: An application of aggregate multiperiod production planning: A
company has 5 products to manufacture on 5 machine types. How much
should be manufactured, stored, and sold in each period over the next 6 peri-
ods (6 months) in order to maximize the contribution. The production capac-
ity on each machine is limited, as well as the sales. The prices of the products
are given. At the end of the six periods the company wants a stock of 50
units of each product in its inventory. The inventory is empty at the begin-
ning of the first period. The only costs are the inventory: each unit of each
products costs 0.5 (money-unit) in each period. (The model is from: O.M.P.
Optimisation, an Application of the Model Generator to Aggregate Produc-
tion Planning, A brochure of Beyers and Partners, pp. 3-11.)
Modeling Steps
The data of the problem are given as follows.
1. There are three sets: the time periods (t), the products (p), and the ma-
chines (m).
2. The prices for each product is Pricep .
3. Timem,p is the time (in hours) that a machine m takes to manufacture
one unit of product p.
4. Limitt,p is the upper limit of sales of a product p in period t.
5. Capacityt,m is the quantity of any product a machine m can manufac-
ture in period t.
6. We introduce three variables: the quantity of products p that are manu-
factured in period t: Manufp,t ; the quantity sold of product p in period
t: Salesp,t , and the quantity stored of product p at the end of time pe-
riod t: Stockp,t .
7. The inventory at the end of the last period must be 50:
Stockp,JUN = 50 forall Period p
8. The proction should not exceed the capacities of the machines:

X
Timem,p · Manufp,t ≤ Capacityt,m forall (m, t)
p
9. The product balance over the periods must hold:
Manufp,t + Stockp,t−1 = Stockp,t + Salesp,t forall (p, t)

10. The contribution (to be maximized) is:

X
(Pricep · Salesp,t − 0.5 · Stockp,t )
p,t

model OMP "A Production Planning Model";
set
t := [ JAN FEB MAR APR MAY JUN ] "Time periods";
p := [ Prod1 Prod2 Prod3 Prod4 Prod5 ] "Products";
m := [ MACHA MACHB MACHC MACHD MACHE ] "machines";
parameter
Price{p} := [ 10 6 8 4 11 ];
Time {m,p} "Production time unit of a product at a
machine" :=
[ 0.5 0.7 . . 0.3
0.1 0.2 . 0.3 .
0.2 . 0.8 . .
0.05 0.03 . 0.07 0.1
. . 0.01 . 0.05 ];
Limit{t,p} "Upper bound on sales of each product in
period" :=
[ 500 1000 300 300 800
600 500 200 100 400
300 600 200 200 500
200 300 400 500 200
400 100 500 100 1000
500 500 100 300 1100 ];
Capacity{t,m} "Available capacity (in hours) on a
machine" :=
[ 1152 768 1152 384 384
1536 768 384 384 384
1536 768 1152 0 384
1536 384 1152 384 384
1152 768 768 384 384
1536 768 1152 384 0 ];
variable
Manuf{p,t} "Quantity manufactured";
Stock{p,t} [0..100] "Quantity in stock (at the end of
period)";
Sales{p,t} [0..Limit] "Quantity to sold";
constraint
tjune{p}: Stock[p,'JUN'] = 50;
c1{m,t}: sum{p} Time*Manuf <= Capacity;
c2{p,t}: Manuf + Stock[p,t-1] = Stock + Sales;
maximize z: sum{p,t} (Price*Sales - 0.50*Stock);
Write('Total contribution is : %9.2f\n', z);
for{p} do
Write(' Product: %5s

/-------------------------------------------------/
/ Period | Manufact. | Stock | Sales /
/-------------------------------------------------/\n
',p);
Write{t}(' / %3s | %9.2f | %8.2f | %8.1f
/\n',
t,Manuf,Stock,Sales) , ' ');
Write(' /--
-----------------------------------------------/\
n');
end
parameter A{m,p,t}:=Time*Manuf;
end

1. Suppose that machine MACHE has broken down at the end of the first
period and cannot be replaced. What is the main consequence for the
production? What is the new contribution?
2. The model did not consider any production costs (only inventory cost
were considered). How could we take into account production cost, say,
to manufacture one unit of a product costs 5 (for all products). Model
it!
1. Product 3 and five are not manufactured anymore after period 1. Why?
The contribution is: 54592.14. To achieve this, the capacities of MACHE
must be set to zero after the first period. Just set:
{t|t>1} (Capacity[t,'MACHE']:=0);
2. All we need is to modify the objective function as follows:

maximize z: sum{p,t} (Price*Sales - 5*Manuf -
0.50*Stock);
7.9. One Machine Scheduling (ordon)

ref Groeflin H. Course DSS II

model ordon "One Machine Scheduling";
set i,j := [1..7] "jobs";
parameter a{i} := [10 15 8 20 30 0 30] "earliest time";
d{i} := [5 6 7 4 3 6 2] "duration";
q{i} := [7 26 24 21 6 17 0] "tail time";
variable t{i} [a..1000] "starting time";
constraint A{i,j|i<j}: t[j]-t[i] >= d[i] or t[i]-t[j]
>= d[j];
minimize obj: max{i} (t+d+q);
Draw.Scale(10,10);
{i} Draw.Rect(i&'',t,1,d,2);
{i|t+d+q=obj} Draw.Line(t+d,2.7,t+d+q,2.7);
Draw.Text(obj&'',obj,1);
end

7.10. A Transport Model Fragment (pam)

Model from Creegan J.B.jr., PAM Overview";

model PAM "A Transport Model Fragment";
set
source,s := [ SS HD ];
destination,d := [ BA PH WA RI ];
product,p := [ P1 P2 P3 ]; --not specified in
the example
parameter
trcost{s,d} := [ 3.52 9.47 0.38 8.63
2.04 6.61 7.22 9.97 ];
prcost{s,p} := [ 4.22 5.05 4.60 , 1.45 2.45 2.03
];
prCap{s} := [ 5000 6000 ];
orders{d} := [ 120 100 30 234 ];
variable
Produce,PR{p,s};
Ship,SH{p,s,d};
constraint
Capacity{s} : sum{p} PR <= prCap;
Balance{p,s}: PR = sum{d} SH;
Demand{p,d} : sum{s} SH >= orders;
minimize costs: sum{p,s}(prcost*PR + sum{d} trcost*SH);
Writep(costs,PR,SH);
end

7.11. The Beer Game (beergame)

The Beer Game perhaps is one of the best known demonstrations in manage-
ment. It was designed by John Sterman (Teaching Takes Off, Flight Simulators
for Management Education, or/MS Today, October 1992, pp. 40-44).
The purpose of the Beer Game is to introduce the concepts of system dynam-
ics in the context of supply chain management. The supply chain contains
five roles:
1. Customer Demand (from the model)
2. Retailer
3. Wholesaler
4. Distributor
5. Factory
Figure 7.1: The Supply Chain
Problem: This is a version of the famous beer game problem: Consider a sim-
plified beer supply chain, consisting of a single retailer, a single wholesaler
which supplies the retailer, a single distributor which supplies the wholesaler,
and a single factory with unlimited raw materials which makes (brews) the
beer and supplies the distributor. Each component in the supply chain has
unlimited storage capacity, and there is a fixed supply lead time and order
delay time between each component.
Each week, each component in the supply chain tries to meet the demand of
the downstream component. Any orders which cannot be met are recorded
as backorders, and are met as soon as possible. No orders will be ignored,
and all orders must eventually be met. At each period, each component in the
supply chain is charged a (1.00) shortage cost per backordered item. Also, at
each period, each location is charged (0.50) inventory holding cost per inven-
tory item that it owns. Each component owns the inventory at that facility. In
addition, the wholesaler owns inventory in transit to the retailer; the distrib-
utor owns inventory in transit to the wholesaler; the factory owns both items
being manufactured and items in transit to the distributor.
Each supply chain member orders some amount from its upstream supplier.
It takes one week for this order to arrive at the supplier. Once the order ar-
rives, the supplier attempts to fill it with available inventory, and there is an
additional two week transportation delay before the material being shipped
by the supplier arrives at the customer who placed the order.
The goal of the retailer, wholesaler, distributor, and factory, is to minimize to-
tal cost, either individually, or for the system. (see: http://beergame.mit.edu).
Modeling Steps
1. First we identify the players of the system and fix the time horizon:
There are four players: Retailer WholeSaler Distributor Factory, in this
order defining the down- and up-flow stream of placing orders and
delivery of the products. The time horizon defines the length of the
"game".
2. In each period the costumer demand an amount of the product (demand).
3. Then we define a minimal stock at which an order should be placed, if
the Inventory fall beneath that point (s) and an amount of stock that is
desirable (S).
4. Further data are the initial stock at the first period for each player (Init).
5. The main data the be calculated are: the Order issues in each period
from each player the up-stream (O) and the resulting Inventory level in
each period for each player (I).
6. The Orders are decisions to take at each period: How much should I
order to minimize my costs? If each player knows the end demand
from the customer, then he can take the politic to order the average
demand in each period. This would probably minimize the Inventory
overall and for each player.
7. However, normally the players use their own strategies to fill the stock.
One such strategy could be the s-S-rule: "If the stock is below the amount
of s then issue an order that fill it up to S, with s ≤ S.
8. In our case, we also consider back log orders, that is orders that cannot
be fulfilled in a period must be fulfilled at a later time period (at heavy
costs). (we can model this fact as negative Inventories.)
9. The rule implemented in the model is: use the s-S rule, however look at
the order issues in the previous period and deduce it from the present
order (this avoids to have to much stock after the time lag. That is,
Op,i = ⇐⇒ (Ip,i + Op,i−1 ≤ s, S − Ip,i − Op,i−1 ).
10. The Inventory then is calculated as follows: Ip,i = ⇐⇒ (i = 1, Initp , Ip,i−1 )+
Op,i−3 −demandi for the retailer and Ip,i = ⇐⇒ (i = 1, Initp , Ip,i−1 )+
Op,i−3 + Op−1,i−1 where we think of p − 1 the next player up-stream.

model BeerGame "The Beer Game";
set p := [Retailer WholeSaler Distributor Factory ];
i := [1..100]; k:=[Order Inv];
parameter demand{i} "Market Demand at a period i";
s{p} "Place order if Inventory is below"
;
S{p} "Order to get that of Inventory";
Init{p} "Stock at the first period";
O{p,i} "Order issues";
I{p,i} "Inventory of p at period i";
T{p,k,i} "Collect all data in one table";
dem "Average demand in a period";;
data;
{i,p}(I[p,i] := if(i=1,Init,I[p,i-1]) + O[p,i-3] - if
(p=1,demand,O[p-1,i-1]) ,
O[p,i] := if(I+O[p,i-1]<s,S-I-O[p,i-1]));
--O[p,i] := 2); -- alternative strategy
{p,i} (T[p,'Order',i] := O,
T[p,'Inv' ,i] := I);
dem:= sum{i}demand/#i "Average demand in a period";
model output;
Draw.Scale(8,-3);
Draw.Line(1,0,#i,0,0);
Draw.Text('Inventory Level',1,80,20);
{i,p|i<#i} Draw.Line(i,I[p,i],i+1,I[p,i+1],p+2);
{p} Draw.Text(p&' ',#i,if(p>2,5)+I[p,#i],12,p+2);
{i|i<#i} Draw.Line(i,demand,i+1,demand[i+1],0);
Draw.Text('demand',#i,5+demand[#i],12);
end
end

Solution: The solutions to the two order strategies are given in the following
figures.
Figure 7.2: Local s-S strategy of ordering

Figure 7.3: Demand estimated strategy of ordering
Further Comments: The beer game show very well that a supply chain is
not inherently stable, as one can see from the results. Demand variability in-
creases as one moves up the supply chain away from the retail customer, and
small changes in consumer demand can result in large variations in orders
placed upstream. Eventually, the network can oscillate in very large swings
as each organization in the supply chain seeks to solve the problem from its
own perspective. This phenomenon is known as the bullwhip effect and has
been observed across most industries, resulting in increased cost and poorer
service.
The following reasons contribute to the bullwhip effect:
1. Overreaction to backlogs
2. No communication nor coordination up and down the supply chain
3. Local decision in ordering in order to reduce inventory
4. Order batching, that is, larger orders in an effort to reduce ordering
costs
5. Demand forecast inaccuracies.
6. Larger delay in delivery down the chain, an delay placing orders up the
chain.
One way to reduce the bullwhip effect is through better information, either in
the form of improved communication along the supply chain or (presumably)
better forecasts. Because managers realize that end-user demand is more pre-
dictable than the demand experienced by factories, they attempt to ignore sig-
nals being sent through the supply chain and instead focus on the end-user
demand. This approach ignores day-to-day fluctuations in favor of running
level. (In the present model, for example, it would substantially reduce the
inventories for all players, if each player only orders up-stream the mean de-
mand from costumers, independently what the successors or follower order.
Try: Op,i = 2.)
Another solution is to reduce or eliminate the delays along the supply chain.
In both real supply chains and simulations of supply chains, cutting order-to-
delivery time by half can cut supply chain fluctuations by 80reduced inven-
tory carry costs, operating costs also decline because less capacity is needed
to handle extreme demand fluctuations.
7.12. Production Planning (aggrplan)

Problem: Various products have to be manufactured over several time peri-
ods. The demand of each product in each time period is given. The company
has to decide in which time period it is more cost-effective to produce under
overtime regime or to store the product. (see [15], p.153).
Modeling Steps
1. We have two sets: products and time periods p and t.

2. The unknown quantity to produce is Xp,t , the unknown quantity to
store is Lp,t , and the unknown overtime amount is Ut in each period.
3. The data are: demand dp,t , technical and workforce coefficients ap , bp ,
Storage and overtime cost lop , ut , the initial product storage L0p , and
the different maximal capacities for overtime, workpower and techical
capacity.
4. We minimize the overtime and the storage costs.
5. The balance in product quantity between two period must hold.
6. We have also technical and workforce capacity.

model aggrplan "Production Planning";
set p "A number of products";
t "A number of time periods";
parameter
lo{p} "Storage cost";
a{p} "Production coefficient workforce";
b{p} "Production coefficient material";
u{t} "Cost of overtime workforce";
L0{p} "Initial stock of product at period 0";
d{p,t} "Demand of product p at time period t";
Nmax{t} "Max. workforce capacity in time period t";
Umax{t} "Max. workforce overtime capacity";
Cmax{t} "Maximal technical capacity";
variable X{p,t} "Product quantity to produce";
L{p,t} "Product quantity to store";
U{t} [0..Umax] "Used overtime workforce in time
period t";
minimize costs: sum{p,t} lo*L + sum{t} u*U "Minimize
storage costs and overtime workforce";
constraint
Balance{p,t}: X + if(t=1,L0,L[p,t-1]) - L = d;
TecCapacity{t}: sum{p} b*X <= Cmax;

ManCapacity{t}: sum{p} a*X - U <= Nmax;
Writep(costs,X,U);
end

A small instance with two products and 4 time periods is given in the data
model:
Listing 7.13: The Data Model

model data "A Production plan";
p := [A B];
t := [1..4];
Nmax{t} := [160 160 160 160];
Cmax{t} := [200 200 200 200];
Umax{t} := [100 100 100 100];
lo{p} := [4 4];
L0{p} := [ 36 220];
a{p} := [1.0 0.5];
b{p} := [0.5 1.0];
u{t} := [5 5 5 5];
d{p,t} := [100 90 110 100 , 200 190 210 200];
d{p,t} := 1.1*d[p,t];
end
Question (Answer see ??)
1. How does the solution change if the technical capacity is unlimited?

2. Demand is uncertain, how does the production plan change if demand
augements/drops by 10%? Model these requirements.
1. Remove the constraints TecCapacity and the costs drop from 1467.50
to 615.
2. Add a new last instruction in the data model:
d{p,t} := 1.1*d[p,t]; (d{p,t} := 0.9*d[p,t];)
Note that at an addition 10% of the demand the problem becomes in-
feasible, because the technical capacity is too small to generate the de-
mand.
7.13. Production Planning (planning)

Further Comments: Compare with MPL

model PLANNING "Production Planning";
set product,p := [1..3];
month,m := [January,February,March,April,May,June,
July,
August,September,October,November,
December];
parameter
price{p} := [105.09, 234.00, 800.00];
Demand {m,p} := [ -- insert them here instead of
extra file
50 24 5, 60 30 6, 70 36 7, 80 42 8, 90 48 9,
100 52 10,110 50 11,120 48 12,120 44 13,120 40 14,
110 36 14,100 32 13 ];
ProductionCapacity{p} := [10000, 42000, 14000 ] ;
ProductionCost{p} := [64.30, 188.10, 653.20] ;
InventoryCost := 8.8 ;
variable
Inventory,I{p,m};
Production,P{p,m};
Sales,S{p,m};
constraint
InventoryBalance{p,m}: I = I[p,m-1] + P - S;
Bn1{p,m}: S <= 1000*Demand ;
Bn2{p,m}: P <= ProductionCapacity ;
MaxInventory{p,m | m<>#m}: I <= 90000 ;
CloseInventory{p}: I[p,'December'] = 20000 ;
maximize Profit: sum{p,m} (price*S-InventoryCost*I-
ProductionCost*P);
Writep(Profit,I,P,S);
end

7.14. Multiperiod Production (prod)

This model is a multiperiod production model from [Fourer al. 1990]. The for-
mulation was motivated by the experiences of a large producer in the United
States. There exists an AMPL formulation and the reader may compare both,
the LPL and the AMPL formulation. This model, with three products and 13
periods, consists of 235 variables, constraints and 896 non-zero entries.
Problem: This model determines a series of workforce levels that will most
economically meet demands and inventory requirements over time, while
minimizing overall costs. There are wage costs (regular and overtime), hiring
costs, layoff costs, demand shortage costs, and inventory costs. How many
workers should be employed, on a regular or overtime basis, hired or fired at
each period?
Modeling Steps

model Production "Multiperiod Production";
set
t,q "Periods" := [ p0 p1 p2 p3 p4 p5 p6 p7 p8 p9 p10
p11 p12 p13 p14];
t1{t} := [ p1 p2 p3 p4 p5 p6 p7 p8 p9 p10
p11 p12 p13 ];
t2{q} := [ p1 p2 p3 p4 p5 p6 p7 p8 p9 p10
p11 p12 p13 ];
p "Products" := [reg18 reg24 pro24];
life,l := [1..2];
parameter
dpp{t1} "working days" := [0 19.5 19 20 19 19.5 19 19
20 19 20 20 18 18];
ol{t1} "overtime limit" := 96;
cmin{t1}"min crew size" := 0;
cmax{t1}"max crew size" := 8;
hc{t1} "hiring costs" := [0 7500 7500 7500 7500 15000
15000 15000 15000 15000 15000 7500 7500 7500];
lc{t1} "layoff costs per period" := hc;
pt{p}:= [1.194 1.509 1.509] "production time";
pc{p}:= [2304 2920 2910] "costs";
cri{p}:= [0.015 0.015 0.015] "inventory costs";
crs{p}:= [1.1 1.1 1.1] "shortage costs";
iinv{p}:= [82 792.2 0] "initial inventory";
rtr "wage per hour in regular time" := 16;
otr "wage per hour in overtime" := 43.85;
rir "regular inventory ratio" := 0.75;
pir "promotional inventory ratio" := 0.8;
sl "regular-time hours per shift" := 8;

cs "crew size" := 18;
iw "initial workforce" := 8;
dem{t,p} "demand" := /
: reg18 reg24 pro24 :
p1 63.8 1212 .
p2 76 306.2 .
p3 88.4 319 .
p4 913.8 208.4 .
p5 115 298 .
p6 133.8 328.2 .
p7 79.6 959.6 .
p8 111 257.6 .
p9 121.6 335.6 .
p10 470 118 1102
p11 78.4 284.8 .
p12 99.4 970 .
p13 140.4 343.8 .
p14 63.8 1212 . /;
pro{t,p} "promoted products in a period (=1)" :=
/ p1 reg24 1
p4 reg18 1
p7 reg24 1
p10 reg18 1
p10 pro24 1
p13 reg24 1
p14 reg24 1 /;
iil{t1,p}:=Max(iinv - sum{t2|t2<=t1} dem[t2,p],0) "
initial inventory left";
minv{t in t1,p} := dem[t+1,p]*if(pro[t+1,p],pir,rir) "
minimum inventory";
variable
Crews{t|t<#t} "Average number of crews employed";
Hire{t1} "Crews hired from previous to current
period";
Layoff{t1} "Crews laid off from previous to current
period";
Rprd{t1,p} "Regular production";
Oprd{t1,p} "Overtime production";
Inv{t1,p,life|life<t1} "Inventory";
Shortage{t1,p} "Accumulated unsatisfied demand";
constraint
rlim{t1}: sum{p} pt*Rprd <= sl*dpp[t1]*Crews[t1] "
Regular time limits";
olim{t1}: sum{p} pt*Oprd <= ol[t1] "Overtime limits";
empl0: Crews['p0'] = iw "initial crew level";
empl{t1}: Crews[t1] = Crews[t1-1] + Hire[t1] - Layoff[
t1] "Crew levels";
Bounds{t1}: cmin[t1] <= Crews[t1] <= cmax[t1] "Crew
limits";
dreq{t1,p}: Rprd + Oprd + Shortage[t1,p] - Shortage[t1

-1,p]
+ sum{l} ( Inv[t1-1,p,l] - Inv[t1,p,l] ) = dem[t1,p]-
Max(iil[t1,p],0)
"Demand requirement";
ireq{t1,p}: sum{life} Inv + iil >= minv "Inventory
requirements";
ilim1{t1,p}: Inv[t1,p,1] <= Rprd + Oprd "New-
inventory limits";
alim{t1,p,l | t1>1 and l>1}: Inv <= Inv[t1-1,p,l-1] "
Inventory limits";
minimize cost: sum{t1} (rtr*sl*dpp[t1]*cs*Crews[t1] + hc[
t1]*Hire[t1]
+ lc[t1]*Layoff[t1]) + sum{t1,p} (otr*cs*pt*Oprd +
crs*pc*Shortage)
+ sum{t1,p,life} cri*pc*Inv
"Regular wages+hired and layoff costs+overtime
wages+shortage and inventory costs";
Write('\nperiod Crews Hire Layoff\n');
Write{t1}(' %5s %3d %3d %4d\n', t1,
Crews[t1],Hire[t1],Layoff[t1]);
Writep(cost,Rprd,Oprd,Shortage,Inv);
end

Model from: FOURER R. and GAY D.M. and KERNIGHAN B.W., [1990], A
Modeling Languagef for Mathematical Programming, in: Management Sci-
ence, Vol. 36, p. 519ff.
7.15. Production and Planning (prodplan)

Problem: A small proction problem: ...
Modeling Steps

model PRODPLAN "Production and Planning";
set
month ,m := [Jan,Feb,Mar,Apr,May,Jun];
machine,r := [Grind,Vdrill,Hdrill,Boring,Planing];
parameter
price {p} := [ 10.00 6.00 8.00 4.00 11.00 ];
time {r,p} := [
50 70 0 0 30
10 20 0 30 0
20 0 80 0 0
5 3 0 7 10
0 0 1 0 5];
MaxSales {m,p} := [
500 1000 300 300 800
600 500 200 100 400
300 600 200 200 500
200 300 400 500 200
400 100 500 100 1000
500 500 100 300 1100];
DaysPerMonth := 6*4;
HoursPerDay := 2*8;
HoursPerMonth := HoursPerDay * DaysPerMonth ;
MachineHours {m,r} := [ 3, 2, 3, 3, 3,
4, 2, 1, 3, 3,
4, 2, 3, 2, 3,
4, 1, 3, 3, 3,
3, 1, 3, 3, 3,
4, 2, 2, 3, 2 ];
InventoryCost := 0.50;
variable
Sale {p,m};
Produ {p,m};
Invt {p,m};
constraint
Capacity{m,r}: sum{p} time/100*Produ <= HoursPerMonth*
MachineHours;
Balance{p,m}: Invt = Invt[p,m-1] + Produ - Sale;
Bound {p,m} : Sale <= MaxSales;
InvtCapacity {p,m} : if(m='Jun' , Invt=50 , Invt<=100);

maximize Contribution: sum{p,m} (price*Sale-InventoryCost
*Invt);
Writep(Contribution,Sale,Produ,Invt);
end

7.16. Production and Shipment Planning (prodship)
Problem: There are a number of plants, each manufacturing several prod-

ucts. The products are shipped to various warehouses from where they are
distributed to demand-centers or customers. Each demand-center is supplied
from only one warehouse. The question is how much, and at which plants, to
produce a given commodity in order to meet the demand requirements at the
demand-centers while minimizing production and transportation costs. Fig-
ure 7.4 gives a overview of the overall product flow, (The model was designed
by: [9]).
Figure 7.4: Product Flow Description
Modeling Steps
1. ...

model ProdShip "Production and Shipment Planning";

set
plants,f "A set of plants (factories)";
products,p "A set of products";
warehouses,w "A set of warehouses";
centers,c "A set of demand-centers";
parameter
CAPACITY{f,p} "The quantity that can maximally be
produced";
PCOST {f,p} "The production cost";
SCOST{f,w} "The shipping cost (plant to warehouse)";
TCOST{w,c} "The transshipping cost (warehouse to
centers)";
DEMAND{c,p} "Demands at the centers";
variable
PRODUCE{f,p|PCOST} [0..CAPACITY] "The quantity to
produce (cannot exceed the capacity)";
SHIP{f,w,p|PCOST and SCOST} "The quantity to be
shipped";
binary ASSIGN{w,c|TCOST} "=1, if the center is
assigned to the warehouse, otherwise 0";
constraint
PRODROW{f,p|PCOST} :sum{w} SHIP - PRODUCE = 0
"All produced quantities must be shipped to the
warehouses";
SHIPROW{f,w,p|SCOST}: sum{c} DEMAND*ASSIGN - sum{f}
SHIP = 0
"The demands must be satisfied (nothing must be
stored)";
CENTROW{c}: sum{w} ASSIGN = 1
"A center can be supplied from only one warehouse";
minimize COST: sum{f,p} PCOST*PRODUCE + sum{f,w,p}
SCOST*SHIP
+ sum{w,c} TCOST*(sum{p} DEMAND)*ASSIGN
"The overall cost, i.e. production, shipping and
transshipping cost";
----- data sets
model data1;
/* data1: OPT=3290215.695, Var:720, Const:601,Ints
:169,
NonZeroes:12751, Density:2.9467% */
f := [1..5];
p := [1..11];
w := [1..10];
c := [1..20];
CAPACITY{f,p} := if(Rnd(0,1)<0.95 , Rnd(500,2200));
PCOST{f,p} := if(CAPACITY , Rnd(200,300));
SCOST{f,w} := if(Rnd(0,1)<0.95 , Rnd(1,20));
TCOST{w,c} := if(Rnd(0,1)<0.90 , Rnd(15,70));
DEMAND{c,p} := Rnd(20,100);
end
model data2;
/* data2: OPT= , Var:1613, Const:1521, Ints:181,
NonZeroes: 44071, density: 1.7963% */
f := [1..13];
p := [1..11];
w := [1..10];
c := [1..20];
/* different parameters
DEMAND{c,p} := Rnd(20,100); */
end
model data3;
f := [1..13];
p := [1..20];
w := [1..10];
c := [1..20];
end
model data4;
f := [ topeka ny ];
p := [ chips nachos ];
w := [ topeka ny ];
c := [ east south west ];
CAPACITY{f,p} := /
topeka chips 200
topeka nachos 800
ny chips 600/;
PCOST{f,p}:=/
topeka chips 230
topeka nachos 280
ny chips 255 /;
SCOST{f,w} := /
topeka topeka 1
topeka ny 45
ny ny 2
ny topeka 45 /;
TCOST{w,c} := /
topeka east 60
topeka south 30
topeka west 40
ny east 10
ny south 30
ny west 80 /;
DEMAND{c,p} := /
east chips 200
east nachos 50
south chips 250
south nachos 80
west chips 150
west nachos 300 /;
end
model data;
f := [1..10];
p := [1..7];
w := [1..20];
c := [1..20];
end
model output;
Write('Total costs are: %15.2f\n', COST);
Write('Assignment:\n warehouse assigned centers\
n\
-------------------------------------\n');
Write{w}(' %6s %5s\n', w, {c|ASSIGN}c);
Writep(PRODUCE,SHIP,ASSIGN);
end
end

7.17. A Production Model (produ)

Problem:
Model from: FOURER R., A Modeling Language for Mathematical Program-
ming, 1987
Modeling Steps

model PRODU "A Production Model";
set
p "products";
r "raw materials";
t "number of storage periods";
parameter
initStock{r} "maximum initial stock";
cost{r} "raw material cost";
value{r} "estimated residual value or a disposal
cost (if < 0)";
maxPrd "maximum unit of production";
units{r,p} "quantity of raw material needed to
produce a product";
profit{p,t|t<>#t} "estimated profit or disposal cost
(if < 0) of a unit of product in each period
except the last one";
variable
X{p,t | t<>#t} "number of units of product
manufactured in each period except the last one";
S{r,t} "number of units of raw material in
storage at the beginning of each period";
constraint
limit{t | t<>#t}: sum{p} X <= maxPrd
"total production in each period is limited";
start{r}: S[r,1] <= initStock
"storage on raw materials in first period is
limited";
balance{r,t | t<>#t}: S[r,t+1] = S - sum{p} units*X
"storage of each raw material in a period must
equal storage in the last period plus the used
units for production";
maximize totalProfit: sum{t|t<>#t} (sum{p} profit*X -
sum{r} cost*S) + sum{r} value*S[r,#t]
"maximize the total over all periods of the
estimated profit minus the storage costs plus
the value of remaining raw materials after the
last period";
Writep(totalProfit,X,S);
model data "an model instance";
t := [1..5];
p := [nuts bolts washers];
r := [nickel iron];
initStock{r}:=[7.32 35.8];
value{r}:=[-0.01 0.025];
cost{r}:=[0.025 0.03];
maxPrd := 123.7;
profit{p,t} :=[1.73 1.8 1.6 2.2 .
1.82 1.9 1.7 2.5 .
1.05 1.1 0.95 1.33 .];
units{r,p} := [0.21 0.17 0.08
0.79 0.83 0.92];
end
end

7.18. Multiperodic Production Planning (product)

Problem: A company manufactures k ∈ K products (say three: P1, P2, P3)
and has at its disposal h ∈ H machines (say three: M1, M2, M3) to produce
them. It can undertake normal and overtime production and needs to plan
for two time periods (say winter and summer). The products can be stored
between periods, but the capacity is limited for each product. Any product
left over after the last period has very little resale value. Machine capacities,
production time of a unit on a machine, selling prices, demand for each prod-
uct are given. How should the company produce each product in order to
fulfil all constraints? (This problem is from: [10]).
Modeling Steps
There are different ways of attacking the problem. One way is to begin with
the data.
1. Resale value r and storage cost s: they depend on the product types,
hence we have rk and sk for each product type k.
2. Storage capacity u: supposing that each product type has its limits, then
we have uk for each product type k.
3. Demand d and selling price: Suppose we have a different demand and
price for each product type k in each time period i (summer and win-
ter), then this gives two tables di,k and pi,k .
4. The different machines h have different availability time a in the vari-
ous production modes j (normal and overtime) and in the different time
periods i. Hence the data are given in a three-dimensional table ai,j,h .
5. Each unit of product k manufatured at each machine h in each period i
and under the different production modes j has different costs c. This
is specified in a four-dimensional table ci,j,k,h .
6. Finally, each machine h needs a different time t to manufacture a unit of
a product k in each time period i under the different production modes
j. Hence ti,j,k,h . (Of course, we might suppose that this time t does not
depend on production modes j neither on the time period i. In this case
we would have tk,h . The modeller has the liberty here to specify the
data, but it depends on the context. We choose a finer grid in the data.)
7. Now we might ask: what are the variables, the unknown quantities,
to be calculated. Certainly we need to specify the quantity sold Z and
stored Y of each product type k in each time period i. Hence we get Zi,k
and Yi,k .
8. Furthermore, we need the quatity produced X for each product k un-

der the different modes j in each time period i manufactured by the
different machines h. Hence Xi,j,k,h .
9. We want to maximize profit, that is the value produced minus storage
costs minus the final resale value.
10. The value produced is (unit price P minus unit cost) times quantity over
all products in all periods, that is: i,j,k,h (pi,k − ci,j,k,h ) · Xi,j,k,h
11. The storage costs are: unit storage cost times Pstorage quantity of all
products in all period except the last. That is, i,k|i<#i si,k · Yi,k .
P
12. The final resale value is k (rk − p#i,k ) · Y#i,k .
13. The constraints are as follows. The demand must be fulfilled: Zi,k ≥
di,k . for each product type k in each time period i.
14. The storage in each period i and for each product is limited: Yi,k ≤ uk .
P
15. The machine availability is limited: k ti,j,k,h · Xi,j,k,h ≤ ai,j,h . for
each machine h in each time period i and in each production mode j.
16. Finally, there is a balance of quantities for each product k in each time
period i: produced quantity plus stock at the beginning of the period
is equalP to sold quantity plus stored quantity at the end of a period.
Hence: j,h Xi,j,k,h +Yi−1,k = Yi,k +Zi,k . The formulation is complete.
The pieces are put together and the complete model follows.

model Production "Multiperodic Production Planning";
set
i := [summer winter] "Time periods";
j := [normal overtime] "Productionmodes";
k := [p1 p2 p3] "Product types";
h := [m1 m2 m3] "Machines";
parameter
t{i,j,k,h} "Machine hours used to produce a unit"
:= [4 7 3 5 6 . 6 6 .
3 6 2 4 5 . 5 5 .
5 8 4 6 7 . 7 7 .
4 7 3 5 6 . 5 6 .];
c{i,j,k,h} "The production cost"
:= [2 4 1 3 3 . 4 2 .
3 5 2 4 4 . 5 3 .
3 5 2 4 4 . 5 3 .
4 6 3 5 5 . 6 4 .];
a{i,j,h} "Machine availability"
:= [100 100 40 80 90 30
110 110 50 90 100 40];
p{i,k} "Selling price" := [10 10 9 11 11 10];

d{i,k} "Demand" := [25 30 30 30 25 25];
s{k} "Storage cost" := [ 1 1 1 ];
u{k} "Storage capacity" := [20 20 . ];
r{k} "Final resale value" := [ 2 2 1 ];
variable
X{i,j,h,k|t} "Quantity of product to be produced";
Y{i,k} "Quantity of product to be stored at end of i"
;
Z{i,k} "Quantity of product k sold";
constraint
MACHINES{i,j,h}: sum{k} t*X <= a
"Machine capacity in hours";
STOCK{k,i}: sum{j,h} X + Y[i-1,k] = Y + Z
"Product balances between periods";
DEMAND{i,k}: Z >= d
"Demands to be satisfied";
BOUND{i,k|i<#i}: Y <= u
"Upper bound on storage";
maximize Profit: sum{i,j,k,h}(p-c)*X - sum{k,i|i<#i}s*Y
+ sum{k}(-r+p[#i,k])*Y[#i,k]
"Revenue minus production costs minus storage costs";
parameter Sx{i,k}:=sum{j,h}X;
Writep(Sx,Y,Z);
end

Further Comments: This formulation is a mathematical system of 27 equa-

tions and inequations and 40 variables with a maximizing function. We call
this a linear program (LP). It can be solved efficiently with the well-known
simplex method. The method finds the best values for the unknowns X, Y,
and Z.
Solution: The solution is written as tables of sum{j,h} X, Y, and Z. The three

resulting tables are as follows:
Sx{i,k}
p1 p2 p3
summer 45.00 32.17 30.00
winter 51.79 22.83 25.00
Y{i,k}
p1 p2
summer 20.00 2.17
winter 41.79
Z{i,k}
p1 p2 p3
summer 25.00 30.00 30.00

winter 30.00 25.00 25.00
The result says that 45 and 51.79 units are produced of product p1 in two
periods where 20 are stored after the first period and 41.79 after the second
period. 25 units are sold in the first period and 30 in the second. Hence 41.79
are left to be sold at a discount price at the end. Only 2.16 units are stored of
the product p2 after the first period and nothing after the second period, and
nothing is stored of product p3.
Question (Answer see ??)
1. Run the model. What is the profit?

2. What happens if the storage has unlimited capacity?
1. Click on the link at the beginning of the model specification. Then send
the model to the Internet solver. It returns 1842.7714.
2. Remove the storage bounding constraint BOUND and solve again. There
is almost no change in the result.
7.19. Multiperiod Production Scheduling (schedule)
Problem: Solve the following production planning problem by specifying

how many units of product should be manufactured each period in order
to minimize the sum of inventory holding c and production costs. .. There are
20 units of product in inventory at the start of period 1. Smythe W.R.,Johnson
L.A..,1966,p.202

model SCHEDULE "Multiperiod Production Scheduling";
set
t := [1..5]; -- periods
c := [ regular,forallTime,inventory ];
parameter
rc{t} := [120 100 90 140 90]; -- Regular Time
Capacity, Units
ov{t} := [ 30 40 20 40 60]; -- Overtime
Capacity, Units
pr{t} := [110 160 100 100 150]; -- Production
Requirements, Units
co{c,t} :=
[12 12 16 10 11 -- Regular Time Cost per
Unit
15 16 18 14 14 -- Overtime Cost per Unit
1 1 1 1 1]; -- Inventory costs from
period i to i+1
variable
PRODUCT{c,t};
constraint
init: PRODUCT['inventory',1] = 20 "initial inventory";
prodcapa{t}: PRODUCT['regular',t] <= rc "production
capacity";
capa{t}: PRODUCT['forallTime',t] <= ov "foralltime
capacity";
const{t}: sum{c} PRODUCT - PRODUCT['inventory',t+1] =
pr;
minimize costs: sum{c,t} co * PRODUCT;
Writep(costs,PRODUCT,co);
end

7.20. Multiperiod Production Scheduling (schedule)
Problem: Solve the following production planning problem by specifying

how many units of product should be manufactured each period in order
to minimize the sum of inventory holding c and production costs. .. There are
20 units of product in inventory at the start of period 1. Smythe W.R.,Johnson
References
L.A..,1966,p.202

model SCHEDULE "Multiperiod Production Scheduling";
set
t := [1..5]; -- periods
c := [ regular,forallTime,inventory ];
parameter
rc{t} := [120 100 90 140 90]; -- Regular Time
Capacity, Units
ov{t} := [ 30 40 20 40 60]; -- Overtime
Capacity, Units
pr{t} := [110 160 100 100 150]; -- Production
Requirements, Units
co{c,t} :=
[12 12 16 10 11 -- Regular Time Cost per
Unit
15 16 18 14 14 -- Overtime Cost per Unit
1 1 1 1 1]; -- Inventory costs from
period i to i+1
variable
PRODUCT{c,t};
constraint
init: PRODUCT['inventory',1] = 20 "initial inventory";
prodcapa{t}: PRODUCT['regular',t] <= rc "production
capacity";
capa{t}: PRODUCT['forallTime',t] <= ov "foralltime
capacity";
const{t}: sum{c} PRODUCT - PRODUCT['inventory',t+1] =
pr;
minimize costs: sum{c,t} co * PRODUCT;
Writep(costs,PRODUCT,co);
end

7.21. Tanglewood Chair Manufacturing (tangle)

Model from : MATURANA S. V., Survey and Analysis of LINGO, Mgt 298D
Problem: The Tanglewood Chair Manu. Co. has four plants located around
the country. The fabrication data are the following:
Plant Cost per chair Max production Min production

P1 5 500 0
P2 7 750 400
P3 3 1000 500
P4 4 250 250
The wood required to make each chair is twenty pounds and is obtained from
two suppliers. The company purchase of at least 8 tons of wood per month
from each supplier. The cost of wood is 0.10/lb from supplier 1 and 0.075/lb
from supplier 2. The shipping cost in dollar per pound of wood from each
supplier to each plant is shown below:
P1 P2 P3 P4
sup1 0.01 0.02 0.04 0.04
sup2 0.04 0.03 0.02 0.02
The chairs are sold in New York, Houston, San Francisco and Chicago. The
transportation costs are shown below:
NY H SF C
P1 1 1 2 0
P2 3 6 7 3
P3 3 1 5 3
P4 8 2 1 4
Following data are also available:
City Selling price maximum demand minimum demand

NY 20 2000 500
H 15 400 100
SF 20 1500 500
C 18 1500 500
Where should each plant buy its raw materials? How many chairs should be
made at each plant? How many chairs should be sold at each city? Where
should each plant ship its product?
7.22. A Travel Optimization Problem (travel)

SCICON Ltd. (or Spectrum, 1984, 6, p.125-130 References
7.23. Multi-period Blending Problem (xpress)

Problem: The following is a multi-period blending model. Production of a
metal smelter is to be planned over each of the 30 weeks. Unit production
costs increase with production rate. The smelted metal must satisfy certain
quality requirements and metal produced in one week may be stored for sale
in subsequent weeks. Objective: find the cheapest blend of ores to smelt and
plan production to trade off production and storage cost.
Modeling Steps

model xpress "Multi-period Blending Problem";
set
i,nweeks := [1..30];
nores := [1.. 5];
nlin := [1..10];
nqual := [1..10];
variable
metal{nweeks};
diff{nweeks};
ore{nores,nweeks};
lin{nlin,nweeks};
stock{nweeks};
parameter -- choose random data
demand{nweeks} := Rnd(1,10);
Myield{nores} := Rnd(2,2000);
qual{nqual,nores} := Rnd(1000,3000);
lowerq{nqual} := Rnd(20,30);
upperq{nqual} := Rnd(1000,5000);
orecst{nores} := Rnd(1,90);
orebnd{nores} := Rnd(1000,2000);
prdlvl{nlin} := [0,10,20,30,40,50, 70,100,200, 300];
prdcst{nlin} := [0, 5,15,30,50,80,160,295,795,1345];
D{nweeks} := / 1 1 /;
maxstock := 1000;
maxdiff := 300;
curstock := 10;
curmetal := 100;
constraint
yield{nweeks}: sum{nores} Myield*ore - metal = 0;
lqual{nqual,nweeks}: sum{nores} (qual-lowerq)*ore >=
0;
uqual{nqual,nweeks}: sum{nores} (qual-upperq)*ore <=
0;
produ{nweeks}: sum{nlin} prdlvl*lin - metal = 0;
convex{nweeks}: sum{nlin} lin = 1;

smooth{nweeks }:
metal - sum{i} metal[i-1] + diff = D*curmetal;
malbal{nweeks}:
-stock + metal + sum{i} stock[i-1]
>= demand - D*curstock;
StockBound{nweeks}: stock <= maxstock;
DiffBounds{nweeks}: max(0,-maxdiff) <= diff <=
maxdiff;
OreBound{nores,nweeks}: ore <= orebnd;
minimize TotalCost: sum{nweeks,nores} orecst/1000*ore
+ sum{nweeks,nlin} prdcst/1000*lin;
Writep(TotalCost,metal,diff,stock,ore,lin);
end

Further Comments: The XPRESS-LP formulation: (copied from: "XPRESS-LP,

a prospectus of Dash Associates").
. Multiperiod production/inventory furnace blending problem

.
constraint smelter . gives the model a name
.
LET nweeks = 30 . number of periods
LET nores = 5 . number of ores can buy
LET nqual = 10 . number of ore qualities
LET nlin = 10 . number of linearisation points
.
VARIABLES . define decision variables
metal(nweeks) . final metal production
diff(nweeks) . change in prod level one week to next
ore(nores,nweeks) . raw ores to buy
lin(nlin,nweeks) . production linearisation variables
stock(nweeks) . stocks of smelted metal
.
TABLES . define main data tables
demand(nweeks) . weekly demand
qual(nqual,nores) . ore qualities (first is metal yield)
lowerq(nqual) . lower limits on metal qualities
upperq(nqual) . upper limits on metal qualities
orecst(nores) . ore cost
prdlvl(nlin) . metal production level
prdcst(nlin) . production cost at prdlvl
orebnd(nores) . upper bounds on ore purchases
maxstock . maximum metal stock level
D(nweeks) . switch array
.
DISKDATA . input data to major tables

demand(1) = xpress.dem
qual(1,1) = xpress.qul
lowerq(1) = xpress.lqu
upperq(1) = xpress.uqu
orecst(1) = xpress.ocs
orebnd(1) = xpress.orb
DATA . data assigned explicitly
prdlvl(1) = 0,10,20,30,40,50, 70,100,200, 300
prdcst(1) = 0, 5,15,30,50,80,160,295,795,1345
D(1) = 1 . switch - all other D(t) are zero
maxstock = 100 . maximum stock level
maxdiff = 30 . maximum change in production level
curstock = 10 . current stock level
curmetal = 100 . current metal production level
.
CONSTRAINTS . total cost and problem constraints
.
. total cost (to be minimized)
.
totalcst: sum(t=1:nweeks,j=1:nores) &
(orecst(j)/1000*ore(j,t) &
+ sum(t=1:nweeks,i=1:nlin) (prdcst(i)/1000*lin(i,t) $
.
. metal yield
.
yield(t=1:nweeks): sum(j=1:nores) qual(1,j)*ore(j,t) &
- metal(t) = 0
.
. lower quality limits (where applicable)
.
qual(k=2:nqual,t=1:nweeks | lowerq(k).ne.0): &
sum(j=1:nores) (qual(k,j)-lowerq(k))*ore(j,t) > 0
. upper quality limits (where applicable)
.
uqual(k=2:nqual,t=1:nweeks) | upperq(k).ne.0): &
sum(j=1:nores) (qual(k,j)-upperq(k))*ore(j,t) < 0
.
. production level (so can have piecewise linear cost)
.
prod(t=1:nweeks): sum(i=1:nlin) prdlvl(i)*lin(i,t) &
- metal(t) = 0
.
convex(t=1:nweeks): sum(i=1:nlin) lin(i,t) = 1
.
. production smoothing
.
smooth(t=1:nweeks): &
metal(t) - sum(|t.ne.1) metal(t-1) + diff(t) &
= D(t)*curmetal
.
. material balance and satisfy demand
.
matbal(t=1:nweeks): &
-stock(t) + metal(t) + sum(|t.ne.1) stock(t-1) &
> demand(t) - D(t)*curstock
.
BOUNDS . simple bounds
.
stock(t=1:nweeks) < maxstock . stock level
diff(t=1:nweeks) < maxdiff . decrease in production
diff(t=1:nweeks) > -maxdiff . increase in production
ore(j=1:nores,t=1:nweeks) < ore purchases
.
GENERATE . generate the matrix
. end of XPRESS-LP formulation
Model from: DASH ASS. (XPRESS-LP is a trademark of DASH ASSOCIATES,

Northants, UK).
B IBLIOGRAPHY
[1] Billionnet A. Optimisation discrète, De la modélisation à la résolution par des

logiciels de programmation mathématique. Dunod, Paris, 2007.
[2] R. Borndoerfer, M. Groetschel, and A. Loebel. Alcuin’s Transportation
Problem and Integer Programming. Konrad Zuse Zentrum for Informa-
tionstechnik, Berlin, 1995.
[3] J. Bracken and G.P. McCormick. Selected Applications of Nonlinear Pro-
gramming. John Wiley and Sons, New York, 1968.
[4] A. Brooke, D. Kendrick, and A. Meeraus. GAMS: A User’s Guide. The
Scientific Press, Redwood City, California, 1988.
[5] Guéret Ch., Prins Ch., and M. Sevaux. Programmation linéaire, 65 prob-
lèmes d’optimisation modélisés et résolus avec Visual Xpress. Eyrolles, 2000.
[6] Pulleyblank W.R. Schrijver A. Cook W.J., Cunningham W.H. Combinato-
rial Optimization. John Wiley and Sons, Inc., 1998.
[7] H. Dahl, A. Meeraus, and S.A. Zenios. Some Financial Optimization
Models: Risk Management. in: Zenios, S.A., ed, Financial Optimization.
Cambridge University Press, New York, 1993.
[8] E.D. Dolan and J.J. More. Benchmarking Optimization Software with
COPS., Tech. Report. Mathematics and Computer Science Division„
2000.
[9] Johnson E.L. Modeling and Strong Linear Programs for Mixed Integer Pro-
gramming, pages 1–43. NATO ASI Series F, Vol. 51, Springer, 1989.
[10] E.F.D. Ellison and G. Mitra. UIMP: User Interface for Mathematical Pro-
gramming. ACM Transaction on Mathematical Software, 8:3:229–255, 1982.
465
Bibliography 466
[11] R. Fourer, D.M. Gay, and B.W. Kernighan. AMPL: A Modelling Language
for Mathematical Programming. Duxbury Press/Brooks/Cole Publ. Co.,
second edition, 2003.
[12] Hillier F.S. and Liebermann G.J. Introduction to Operations Research.
McGraw-Hill, 6rd ed., 1999.
[13] Polya G. How To Solve It. Penguin Books 1990, London, 1957.
[14] Dantzig G.B. Linear Programming and Extensions. Princeton University
Press, Princeton, New Jersey, 1963.
[15] H-O. Günther and H. Tempelmeier. Produktion und Logistik. Springer,
Berlin, 2002.
[16] Williams H.P. Model Building in Mathematical Programming. John Wiley,
Third Edition Revisted, 1993.
[17] Williams H.P. Model Building in Mathematical Programming. John Wiley,
Fifth Edition, 2013.
[18] D. Kendrick. Caution and Probing in a Macroeconomic Model. Journal
of Economic Dynamics and Control 4, 2, 1982.
[19] Murty K.G. Operations Research: Deterministic Optimization Models. Pren-
tice Hall, 1994.
[20] G.P. Kutcher, A. Meeraus, and G.T. O’Mara. Agriculture Sector and Pol-
icy Models. The World Bank, 1988.
[21] G.P. Kutcher and P.L. Scandizzo. The Agricultural Economy of North-
east Brazil. The Johns Hopkins University Press, Baltimore and London,
1981.
[22] Cesari L. Optimization – Theory and Applications. Springer Verlag, 1983.
[23] Seiford L.M. and Thall R.M. Recent developments in DEA: the mathe-
matical programming approach to frontier analysis. Journal of Economet-
rics, Vol 46, p. 7–38, 1990.
[24] Bosch R. Mind Sharpener. OPTIMA MPS Newsletter, January 2000.
[25] Shapiro RD. Optimization models for planning and allocation. John Wiley
Sons, 1984.
[26] Rosenthal R.E. Tutorial of GAMS in: GAMS: A User’s Guide. The Scientific
Press, Redwood City, California, 1988.
[27] Ahuja R.K., Magnanti T.L., and Orlin J.B. Network Flows, Theory, Algo-
rithms, and Applications. Prentice Hall, 1993.
[28] Heipcke S. Applications of optimization with Xpress. dash optimization,
2000,
467 Bibliography
www.maths.ed.ac.uk/hall/Xpress/FICO_Docs/
Xpress-booka4.pdf
A translation of [5].
[29] Hürlimann T. Reference Manual for the LPL Modelling Language, most
recent version. lpl.virtual-optima.com/lpl/doc/manual.pdf.
[30] Wiens T.B. The Economics of High-yield Agriculture, in: China: Triple-
Cropping at the Baimano Peoples’ Commune. The World Bank, Wash-
ington DC, 1985.
[31] Winston W.L. Introduction to mathematical programming, applications and
algorithms. Duxbury Press, second edition, 1995.

Textbooks

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Textbooks

Uploaded by

Copyright:

Available Formats

Mathematical Modeling

2 Problems from William's Book 19

3 Problems from the Gueret Book 107

4 Various Textbooks 181

5 Models from the GAMS Library 347

1.1 The Feasible Space . . . . . . . . . . . . . . . . . . . . . . . . 13

2.1 Linear Solution . . . . . . . . . . . . . . . . . . . . . . . . . . 41

3.1 The village streets ([28] p.324) . . . . . . . . . . . . . . . . . . 170

6.1 A Small Example Graph . . . . . . . . . . . . . . . . . . . . . 414

7.1 The Supply Chain . . . . . . . . . . . . . . . . . . . . . . . . 432

1.1 Logical Operators . . . . . . . . . . . . . . . . . . . . . . . . 18

3.1 Data for the Stadium Construction . . . . . . . . . . . . . . . 114

6.1 Data and solution for the model . . . . . . . . . . . . . . . . . 415

This text is a collection of many mathematical models formulated in the LPL

After a short introduction to LPL in Chapter 1, the next Chapter 2 contains

Chapter 4 contains small model from exercises in several known operations

Chapter 5 implements several models from the GAMS Library.

Availability of the LPL-Modeling-System

textbooks home page (lpl.virtual-optima.com/lpl/mainothers.html)

“The essence of mathematics is not to make simple things com-

“Ich betrachte diese einleitenden theoretischen Betrachtungen als

1.1. Model and Related Concepts

Models, and especially mathematical models, are capable of giving deeper

Science is modeling – and mathematics as a research of patterns is at the very

Modeling is a powerful means to solve problems, however, it is all but easy.

1.2. The Model Building Process

Building a mathematical model means to conceptualize, to invent, to contrive

right for everything. In short, modeling is a heuristic process.

That’s the model! We are done. To find a solution, we do some algebraic

1.3. The Modeling Language LPL

model and end are keywords and <modelname> is a user-defined identi-

A set declaration is similar to the definition of a mathematical set. For ex-

ai,j with i, j ∈ I (that is, a 3 × 3 data matrix)

If the variables are integer or binary we write:

The constraint is written in LPL as follows:

These basic syntax elements allows us to write complete mathematical models

Figure 1.1: The Feasible Space

max 300x + 200y

lated as follows (see model tutor013 ):

As expected, solving the model gives x = 40 and y = 30.

In LPL the general model can be formulated as follows:

could be written in LPL as:

However, sometimes we need to format the output. To write each value of xj

ai,j = cj + bi for all i ∈ I, j ∈ J

would be formulated in LPL as follows:

The loop can contain a sequence of any instructions.

1.4. Mathematical Notation

1.4.1 Set-Names versus Index-Names

1.4.2 Logic-mathematical Notation

The first expression is common in all kind of programming languages and

1.4.3 Logical Operators

Meaning Math. notation LPL syntax

Table 1.1: Logical Operators

P ROBLEMS FROM W ILLIAM ’ S

“No mathematician can be a complete mathematician unless he

2.1. Food Manufacture I (will01)

Model: The Model is as follows:The problem is from [17] Chapter 13.1.

Listing 2.1: The Complete Model in LPL [29]

2.2. Food Manufacture II (will02)

Listing 2.2: The Complete Model in LPL [29]

Bal{r,p}: if(p=1,Sto,sto[p-1,r])+buy = use+sto;

2.3. Factory planning I (will03)

Model: The Model is as follows:The problem is from [17] Chapter 13.3.

Listing 2.3: The Complete Model in LPL [29]

DELIVER:sum{r} T[1,r] M[r] +P[1]-N[1] = 0.4Total[1];