Abstract
One of the key objectives of an optimised supply chain is to maintain a low operation cost as well as the service quality at a satisfactory level under demand uncertainty. However, the supply chain network design and inventory control optimisation are usually conducted in a sequential manner where the supply chain network is first determined by solving a mixed-integer programing (MIP) problem, and then the supply chain system with the given network design is tested as a “what if” problem in order to evaluate and improve its performance. Over the last decade, simulation modelling is regarded as an efficient tool for evaluating the performance of a real-world supply chain under different conditions and flexible control policies and simulation-based optimisation has been widely studied for solving inventory management problem under various uncertainties. In this work a hybrid computational framework is proposed to solve both the network design problem and the associated inventory control problem simultaneously. By incorporating region-wise metamodeling method to reduce the computation load, a multi-echelon supply chain case with 13 inventory stocking nodes can be solved within 3,621 s with the proposed algorithm. As a comparison, the simulation-based problem is also solved by the genetic algorithm (GA) toolbox in MATLAB, which only returns a 31 % higher cost after 13,844 s.
