Production and Logistics – serie

The central purpose of this book is to acquaint the reader especially with the cases of local-search-based learning, as well as to introduce methods of contraint-based reasoning, both with respect to their use in automated manufacturing. Attention is restricted to job-shop scheduling, as well as to one-machine scheduling with sequence-dependent setup times. Additionally, some design and planning issues in flexible manufacturing systems are considered. General-purpose search methods - which in particular include methods from local search, such as simulated annealing, tabu search and genetic algorithms - are the basic ingredients of the proposed intelligent knowledge-based local decision rules in order to introduce problem-specific knowledge.

This book is the result of my research perfonned at the University Kiel under the supervision of Professor Dr. Andreas Drexl. I am very grateful to him for suggesting the subject as well as for his guidance and support during the preparation. Professor Dr. Klaus Brockhoff deserves thanks for under- taking the task of the secondary assessment. The Institut fur Betriebswirtschaftslehre in Kiel and especially my col- leagues Dr. Knut Haase, Carsten Jordan, AlfKimrns, and Dr. Frank Salewski provided a stimulating environment. Particularly, I am indebted to Dr. Arno Sprecher and Andreas Schinner. The results of Chapter 2 and 3, respectively, are outcome of joint research with Arno; Andreas proof-read the entire manuscript. Furthennore, I would like to express my gratitude to Prof. Dr. Erik Demeulemeester, Prof Dr. James H. Patterson, Prof Dr. Roman Slowinski, and Prof Dr. Ramon Alvarez-Valdes for the valuable help they provided with suggestions, papers, and computer programs.I, As to the technical realisation of this book, lowe thanks to Reimer Karlsen for coding and impl~menting parts of the algorithms, to Marc Schumann for drawing all figures, to Ina Kantowski who looked after the literature, and to Uwe Penke for providing a ready-to-use computer-laboratory. The gratitude that I feel towards my family requires no further comments.

Production scheduling dictates highly constrained mathematical models with complex and often contradicting objectives. Evolutionary algorithms can be formulated almost independently of the detailed shaping of the problems under consideration. As one would expect, a weak formulation of the problem in the algorithm comes along with an inefficient search. This book discusses the suitability of genetic algorithms for production scheduling, and presents an approach which produces results comparable with those of more tailored optimization techniques.