A New Crossover Operator for Assembly Sequence Planning by Genetic Algorithms

Assembly Sequence Planning (ASP) is the generation of the best or optimal sequence to assemble a certain product, given its design files. Although many planners were introduced in research to solve this problem automatically, it is still solved manually in many advanced assembly firms. The reason behind this is that most introduced planners are very sensitive to large increases in product parts. In fact, most of these planners seek the exact solution, while performing a part basis decision process. As a result, they are trapped in tedious and exhaustive search procedures, which make them inefficient and sometimes obsolete. To overcome these difficulties, Sebaaly and Fujimoto (1996) introduced a new concept of ASP based on Genetic Algorithms application, where the search procedure is performed on a sequence population basis rather than a part basis, and a best sequence is generated without searching the complete set of potential candidates. This paper addresses the problem of improving the GA performance for assembly application, by introducing a new crossover operator. The genetic material can be divided and classified as ‘good’ or ‘bad’. The new crossover insures the maximum transmission of ‘good’ features from one generation to another. This results in a faster GA convergence. The performance of the new algorithm is compared with that of the ordinary matrix crossover for a modified industrial example, where it proved to be faster and more efficient.