The Study On Fuzzy Supply Chain Network Design Problem

A supply chain (SC) is a network of suppliers, manufacturing plants, warehousesand distribution channels organized to acquire raw materials, convert these rawmaterials to finished products, and distribute these products to customers. Thesupply chain network design (SCND) problem is to make the decisions to satisfy thedemands of customers and minimize the sum of strategic and tactical costs. It isone of the biggest strategic decision problems which are used for efcient long termoperations in the whole supply chain. In realistic SCND problem, the transportationcosts and customer demands usually have fuzziness, so we should employ a fuzzytheory that deal with fuzziness to study SCND problem.This thesis first presents a fuzzy two-stage supply chain mean-semivariancemodel by the risk coefcient method. We introduce the value of fuzzy solutions forfuzzy mean-semivariance model, and illustrate the necessary of solving the fuzzymodel in theory. We use approximation approach (AA) to reformulate the originalproblem as a mixed-integer problem. We design an hybrid memetic algorithm (MA)to solve the proposed model. Furthermore, we consider the costs and the risk astwo objective functions, and design a fuzzy two-stage supply chain mean-deviationmodel by multi-objective method. In theoretical aspect, we define the concept aboutthe value of fuzzy solution for the mean-deviation problem to discuss its properties.We apply AA to transform the mean-deviation model into a deterministic multi-objective model, then we design a multi-objective biogeography-based optimization(MO-BBO) algorithm to solve this model. Finally, numerial examples are providedto illustrate the modeling ideas and efectiveness of the proposed hybrid algorithms.The major new results of this thesis include the following four aspects:(i)Two classes of fuzzy two-stage SCND problems are proposed;(ii) We introduce twokinds of value of fuzzy solutions, and illustrate the necessary of the models in theory;(iii) In the cases that transportation costs and customer demands are multivariatenormal fuzzy vectors, the two proposed models can be turned into their deterministicapproximating ones;(iv) An improved hybrid MA algorithm and a MO-BBO are designed to solve the proposed models, and two realistic problems are provided toillustrate the modeling ideas and the efectiveness of the proposed hybrid algorithm.