Intelligent Optimisation Techniques

This book covers four optimisation techniques loosely classified as "intelligent": genetic algorithms, tabu search, simulated annealing and neural networks. • Genetic algorithms (GAs) locate optima using processes similar to those in natural selection and genetics. • Tabu search is a heuristic procedure that employs dynamically generated constraints or tabus to guide the search for optimum solutions. • Simulated annealing finds optima in a way analogous to the reaching of minimum energy configurations in metal annealing. • Neural networks are computational models of the brain. Certain types of neural networks can be used for optimisation by exploiting their inherent ability to evolve in the direction of the negative gradient of an energy function and to reach a stable minimum of that function. Aimed at engineers, the book gives a concise introduction to the four techniques and presents a range of applications drawn from electrical, electronic, manufacturing, mechanical and systems engineering. The book contains listings of C programs implementing the main techniques described to assist readers wishing to experiment with them. The book does not assume a previous background in intelligent optl1TIlsation techniques. For readers unfamiliar with those techniques, Chapter 1 outlines the key concepts underpinning them. To provide a common framework for comparing the different techniques, the chapter describes their performances on simple benchmark numerical and combinatorial problems. More complex engineering applications are covered in the remaining four chapters of the book.

1 Introduction.- 1.1 Genetic Algorithms.- 1.1.1 Background.- 1.1.2 Representation.- 1.1.3 Creation of Initial Population.- 1.1.4 Genetic Operators.- 1.1.5 Control Parameters.- 1.1.6 Fitness Evaluation Function.- 1.2 Tabu Search.- 1.2.1 Background.- 1.2.2 Strategies.- 1.3 Simulated Annealing.- 1.3.1 Background.- 1.3.2 Basic Elements.- 1.4 Neural Networks.- 1.4.1 Basic Unit.- 1.4.2 Structural Categorisation.- 1.4.3 Learning Algorithm Categorisation.- 1.4.4 Optimisation Algorithms.- 1.4.5 Example Neural Networks.- 1.5 Performance of Different Optimisation Techniques on Benchmark Test Functions.- 1.5.1 Genetic Algorithm Used.- 1.5.2 Tabu Search Algorithm Used.- 1.5.3 Simulated Annealing Algorithm Used.- 1.5.4 Neural Network Used.- 1.5.5 Results.- 1.6 Performance of Different Optimisation Techniques on Travelling Salesman Problem.- 1.6.1 Genetic Algorithm Used.- 1.6.2 Tabu Search Algorithm Used.- 1.6.3 Simulated Annealing Algorithm Used.- 1.6.4 Neural Network Used.- 1.6.5 Results.- 1.7 Summary.- References.- 2 Genetic Algorithms.- 2.1 New Models.- 2.1.1 Hybrid Genetic Algorithm.- 2.1.2 Cross-Breeding in Genetic Optimisation.- 2.1.3 Genetic Algorithm with the Ability to Increase the Number of Alternative Solutions.- 2.1.4 Genetic Algorithms with Variable Mutation Rates.- 2.2 Engineering Applications.- 2.2.1 Design of Static Fuzzy Logic Controllers.- 2.2.2 Training Recurrent Neural Networks.- 2.2.3 Adaptive Fuzzy Logic Controller Design.- 2.2.4 Preliminary Gearbox Design.- 2.2.5 Ergonomic Workplace Layout Design.- 2.3 Summary.- References.- 3 Tabu Search.- 3.1 Optimising the Effective Side-Length Expression for the Resonant Frequency of a Triangular Microstrip Antenna.- 3.1.1 Formulation.- 3.1.2 Results and Discussion.- 3.2 Obtaining a Simple Formula for the Radiation Efficiency of a Resonant Rectangular Microstrip Antenna.- 3.2.1 Radiation Efficiency of Rectangular Microstrip Antennas.- 3.2.2 Application of Tabu Search to the Problem.- 3.2.3 Simulation Results and Discussion.- 3.3 Training Recurrent Neural Networks for System Identification.- 3.3.1 Parallel Tabu Search.- 3.3.2 Crossover Operator.- 3.3.3 Training the Elman Network.- 3.3.4 Simulation Results and Discussion.- 3.4 Designing Digital Finite-Impulse-Response Filters.- 3.4.1 FIR Filter Design Problem.- 3.4.2 Solution by Tabu Search.- 3.4.3 Simulation Results.- 3.5 Tuning PID Controller Parameters.- 3.5.1 Application of Tabu Search to the Problem.- 3.5.2 Simulation Results.- 3.6 Summary.- References.- 4 Simulated Annealing.- 4.1 Optimal Alignment of Laser Chip and Optical Fibre.- 4.1.1 Background.- 4.1.2 Experimental Setup.- 4.1.3 Initial Results.- 4.1.4 Modification of Generation Mechanism.- 4.1.5 Modification of Cooling Schedule.- 4.1.6 Starting Point.- 4.1.7 Final Modifications to the Algorithm.- 4.1.8 Results.- 4.2 Inspection Stations Allocation and Sequencing.- 4.2.1 Background.- 4.2.2 Transfer Functions Model.- 4.2.3 Problem Description.- 4.2.4 Application of Simulated Annealing.- 4.2.5 Experimentation and Results.- 4.3 Economic Lot-Size Production.- 4.3.1 Economic Lot-Size Production Model.- 4.3.2 Implementation to Economic Lot-Size Production.- 4.4 Summary.- References.- 5 Neural Networks.- 5.1 VLSI Placement using MHSO Networks.- 5.1.1 Placement System Based on Mapping Self-Organising Network.- 5.1.2 Hierarchical Neural Network for Macro Cell Placement.- 5.1.3 MHSO2 Experiments.- 5.2 Satellite Broadcast Scheduling using a Hopfield Network.- 5.2.1 Problem Definition.- 5.2.2 Neural-Network Approach.- 5.2.3 Simulation Results.- 5.3 Summary.- References.- Appendix 1 Classical Optimisation.- A1.1 Basic Definitions.- A1.2 Classification of Problems.- A1.3 Classification of Optimisation Techniques.- References.- Appendix 2 Fuzzy Logic Control.- A2.1 Fuzzy Sets.- A2.1.1 Fuzzy Set Theory.- A2.1.2 Basic Operations on Fuzzy Sets.- A2.2 Fuzzy Relations.- A2.3 Compositional Rule of Inference.- A2.4 Basic Structure of a Fuzzy Logic Controller.- A2.5 Studies in Fuzzy Logic Control.- References.- Appendix 3 Genetic Algorithm Program.- Appendix 4 Tabu Search Program.- Appendix 5 Simulated Annealing Program.- Appendix 6 Neural Network Programs.- Author Index.