Laureano F. Escudero (Area of Statistics and Operations Research, Universidad Rey Juan Carlos) and Juan F. Monge (Center of Operations Research, Universidad Miguel Hernández de Elche)
Abstract:
A tight mixed integer linear programming modeling framework is presented for the multistage multiscale facility location multiproduct allocation network expansion planning under uncertainty. Two types of decisions are considered, namely, the strategic and the operational ones. The strategic decisions are the selection of facility locations in a network as well as the related capacity dimensioning and expansion along a time horizon. A comprehensive literature overview on the problem is performed. Two types of uncertain parameters are considered, namely, strategic and operational ones, to be represented in multistage and two-stage scenario trees, resp. By using the special structure of the facility location problem, the coherent time-consistent risk-averse measure to consider is the expected conditional second-order stochastic dominance. Given the intrinsic problem’s difficulty and the huge instances’ dimensions, it is unrealistic to seek an optimal solution. A specialization of the matheuristic algorithm SFR3 is presented to obtain a (hopefully good) feasible solution in reasonable time as well as a lower bound to assess the solution quality. The performance of the overall approach is computationally validated by considering a dynamic supply network design problem with 100 raw material, 50 products, 30 candidate facilities (10 plants and 20 distribution centers), 31 strategic scenario nodes in the time horizon, and 4 operational ones per stage.