Assembly planning with an ordering genetic algorithm

The goal of assembly planning consists in generating feasible sequences to assemble a product and selecting an efficient assembly sequence from amongst these. This paper describes an original ordering genetic algorithm (OGA) developed to solve this problem. The developed approach permits the generation of assembly trees for a mechanical product. The algorithm is based on three main ideas. First, a mapping transforms any studied assembly plan into a valid one using ‘precedence values’ changing through the sequence, so that an invalid sequence will never be proposed. Secondly, to identify subsets, trace is kept all along the sequence of the components membership to a set of parts. Finally, the individuals of the OGA are compared with each other using a multi-criteria decision aided method called PROMETHEE II. The use of this method avoids aggregating several technical criteria into a unique fitness value. The proposed algorithm, illustrated through the simple example of a mouse device, has been applied on an industrial signalling relay made of 34 parts.