Intelligent design of manufacturing systems
The design of a manufacturing system is normally performed in two distinct stages, i.e. steady state design and dynamic state design. Within each system design stage a variety of decisions need to be made of which essential ones are the determination of the product range to be manufactured, the layout of equipment on the shopfloor, allocation of work tasks to workstations, planning of aggregate capacity requirements and determining the lot sizes to be processed. This research work has examined the individual problem areas listed above in order to identify the efficiency of current solution techniques and to determine the problems experienced with their use. It has been identified that for each design problem. although there are an assortment of solution techniques available, the majority of these techniques are unable to generate optimal or near optimal solutions to problems of a practical size. In addition, a variety of limitations have been identified that restrict the use of existing techniques. For example, existing methods are limited with respect to the external conditions over which they are applicable and/or cannot enable qualitative or subjective judgements of experienced personnel to influence solution outcomes. An investigation of optimization techniques has been carried out which indicated that genetic algorithms offer great potential in solving the variety of problem areas involved in manufacturing systems design. This research has, therefore, concentrated on testing the use of genetic algorithms to make individual manufacturing design decisions. In particular, the ability of genetic algorithms to generate better solutions than existing techniques has been examined and their ability to overcome the range of limitations that exist with current solution techniques. IIFor each problem area, a typical solution has been coded in terms of a genetic algorithm structure, a suitable objective function constructed and experiments performed to identify the most suitable operators and operator parameter values to use. The best solution generated using these parameters has then been compared with the solution derived using a traditional solution technique. In addition, from the range of experiments undertaken the underlying relationships have been identified between problem characteristics and optimality of operator types and parameter values. The results of the research have identified that genetic algorithms could provide an improved solution technique for all manufacturing design decision areas investigated. In most areas genetic algorithms identified lower cost solutions and overcame many of the limitations of existing techniques.