Strategic decisions on the location, the capacity, the layout, and the mission of production and distribution facilities are key drivers of manufacturing company's competitiveness. The aim of supply chain design models is to recommend economically efficient decisions to the company's administrator. The realism of the mathematical modeling of the aforementioned issues conditions the validity and the applicability of the prescribed solutions. The elaboration of realistic methodologies is thus one of the main success factors of decision support processes. This thesis first proposes a generic methodology to design the production-distribution network of divergent process industry companies. The approach is validated by applying it to Virtu@l-Lumber, a virtual but realistic case from the lumber industry. Second, an approach that takes into account market opportunities when designing productiondistribution networks is proposed and tested. This approach is based on a stochastic programming with recourse model. Lastly, a mathematical model combining the two previous formulations is proposed and its potential impact on the lumber industry is investigated with the Virtu@l-Lumber case.
