Heuristics for Determining Module Stocks in Assemble-to-order Systems

In an assemble-to-order context, the choice of a modular design is of paramount importance. The selection of the semi-finished products that would be kept in stock, indeed impacts the whole supply chain (storage, transport, production, etc.). In this article, this lever is used to minimize, for a given number of stocked modules, the mean final assembly time. The choice of a stock composition has to be done according to the information about the demand; the major issue is to address this problem when the demand is not totally known. Here, an original modelling that enables to integrate this aspect is proposed. This combinatorial optimization problem is addressed using heuristics. The performances of the proposed heuristics are evaluated trough computational tests with different repartition of demands. Furthermore, the heuristics are studied when dealing with partial information about final demand. Results show that the heuristics are pertinent; they also stress the impact on performances of information acquisition.