Development of improved solution methods for deterministic and stochastic supply chain network design

In today's highly competitive business environment, only efficient supply chains that integrate decisions in various phases can survive. Two major issues in the efficient design of a supply chain network are facility location as strategic decisions, and flow of the materials as tactical decisions. Because of complexity of the integrated models due to the large size of the binary mixed integer linear programming problems, it is often decomposed into several components treated separately. However, given the importance of the interactions between these decisions, important benefits can be obtained by treating the network as a whole. Therefore, a general and flexible formulation was developed for the supply chain network design problems. The model integrated strategic and tactical decisions of supply chain planning in the deterministic, multi-commodity, and single-period context. Considering the poor convergence properties of the Benders' decomposition as an efficient approach for solving combinatorial optimization problems, an improved solution method was developed to solve the deterministic model. The method improved the solution process by reducing the required iterations and integer programming problems.Since, many real-life supply chain networks are characterized by large uncertainty in the tactical parameters; therefore, an integrated model was developed for supply chain network design problems under uncertainty. The developed model was provided as a two-stage stochastic program where the two stages in the decisionmaking process correspond to the strategic and tactical decisions. Considering the difficulties in solving the two-stage stochastic program using the sample average approximation as a well-known technique, an improved stochastic solution method was developed to solve the two-stage stochastic program. In the developed method, an improved sampling strategy was integrated with the improved Benders' decomposition approach to solve the supply chain network design problems under uncertainty. The computational experiments were performed on a set of randomly generated test problems as well as a real-life case study. The results have shown that the computational processes for solving deterministic, and stochastic problems have been expedited using the developed solution methods in comparison with the original methods by an average factor of 46%, and 69%, respectively. The reasonable computation times indicated that the re-optimization capability of the solution process increased by using the developed solution methods. The performed experiments have shown that the developed methods can be applied to solve the realistic cases of large size. Furthermore, the computational experiments have indicated that the optimality gaps to find the solution of deterministic, and stochastic problems have been improved using the developed methods in comparison with the original methods by an average factor of 1%, and 4%, respectively.

Text FK 2011 172 ir.pdf Download (10MB)

Actions (login required)

View Item