Concurrent assembly planning with genetic algorithms

This work investigates the application of genetic algorithm (GA)-based sear ch techniques to concurrent assembly planning, where product design and ass embly process planning are performed in parallel, and the evaluation of a d esign configuration is influenced by the performance of its related assembl y process. Several types of GAs and an exhaustive combinatorial approach ar e compared, in terms of reliability and speed in locating the global optimu m. The different algorithms are tested first on a set of artificially gener ated assembly planning problems, which are intended to represent a broad sp ectrum of combinatorial complexity; then an industrial case study is presen ted. Test problems indicate that GAs are slightly less reliable than the co mbinatorial approach in finding the global, but are capable of identifying solutions which are very close to the global optimum with consistency, soon outperforming the combinatorial approach in terms of execution times, as t he problem complexity grows. For an industrial case study of low combinator ial complexity, such as the one chosen in this work, GAs and combinatorial approach perform almost equivalently, both in terms of reliability and spee d. In summary, GAs seem a suitable choice for those planning applications w here response time is an important factor, and results which are close enou gh to the global optimum are still considered acceptable such as in concurr ent assembly planning, where response time is a key factor when assessing t he validity of a product design configuration in terms of the performance o f its assembly plan. (C) 2000 Elsevier Science Ltd. All rights reserved.