We proposed a nonlinear mixed integer model for developing an acquisition policy with uncertain demand, which combines vender selection with inventory management and maximizes the benefit of the manufacturer. An iterative solution algorithm is proposed to find how much to purchase from which supplier, and what quantities to order in each cycle.

This paper presents an approach to take into account market opportunities when designing production-distribution networks for the lumber industry. The deployment of the production-distribution network is formulated as a two-stage stochastic mathematical program. An application to a realistic lumber industry case, Virtu@l-Lumber, is proposed and analysed.

One particular difficulty is to design supply chains with a very large number of potential customers, which makes the use of mixed-integer programming prohibitive. In this presentation, we describe the modeling and computational benefits of our approach which is based on the continuous approximation of customer demand for such problems. In contrast with the use of mixed-integer programming which prioritizes the location decisions, our approach gives priority to the determination of service regions (each served by a single facility) given a predetermined market area. We show that by deciding on customer allocations (i.e. service regions), rather than the facility locations, we can accelerate the exact solution procedure.

We present recent improvements of our metaheuristic for facility layouts. By coupling ant system with an optimization model, we improve the quality of the solution generated by the ants and guarantee optimality for the ‘layout code’ considered. Empirical results for classical cases will be presented.