In a context of product family design, industrials are led to design and manufacture a large variety of products to answer di erent customer's needs and speci c constraints linked to the supply chain. Modular design is a strategy that allows to generate various bills of materials of products from a limitted set of modules. As weel as the de nition of the bills of materials, the choice of production sites for each module is also necessary. Often, the adopted strategies are extreme, and propose a production-for-stock to anticipate customer's needs or a production-to-order to minimize the storage costs. The objective of this thesis is to study an intermediate strategy said "assembly to order" allowing to balance between the storage costs and the manufacturing delays. The contribution of this thesis concerns the proposition of a model to support a global approach for the design of a product familly with a large diversity. The main contribution lies from one part in the integration of the data relative to the supply chain in the design operation, and in the other part in the proposition of an algorithm to determine a good solution in a rather reasonable computational time. Besides, the advantages of certain strategies of assembly of nished products such as the standardization are studied.
