Research On The Closed-loop Supply Chain Management Under Uncertainty

As the environmental and resource-saving pressures are more and more intense, the issue of promoting sustainable economic and social development has become one of the most important subjects, which has leaded the closed-loop supply chain (CLSC) to come forth. Although it has been just more than ten years since the theory was formed, the CLSC has been attacting more attentions from the economic and academic communities and become a research focus in recent years. Because the concept of CLSC that the material/energy should be recycled to maximize the utilization of the limited resource, has leaded the public an environmental friendly and resource saving view to handle the used products.The thesis studies on the closed-loop supply chain management under uncertainty. Combining with the academic frontiers and using the quantitative analysis, the thesis focuses on the researches of production planning, selection of the third party reverse logistics provider, pricing and facility location in a CLSC. The main research works and achievements of this dissertation can be summarized as follows:(1) The production planning problem in a CLSC is elaborated, and a CLSC system is first addressed based on it, in which the conventional forward supply chain is combined with reverse supply chain by integrating supplier, retailer, third party recycle dealer and manufacturer. The multi-echelon inventory model is then proposed by considering the uncertain factors of customer demand, return quantity and successful reproduction rate as triangular fuzzy numbers, whose target function is to maximize the pessimistic/optimistic joint profit at a confidence level based on the methodology of constraint programming. With the model analysis and fuzzy set method, the constraint programming is transformed into crisp one. Besides, a modified multi-swarm co-evolutionary algorithm (MMCA) is proposed to compute the optimization efficiently. The experimental results show that the proposed algorithm is effective for the optimization problem.(2) A conceptual model for selecting and evaluating third-party reverse logistics providers is proposed, in which three main phases are included:performance criteria selection and comparision, weight calculation for criteria and final selection. In this model, the analysts could use triangular fuzzy numbers to describe the criteria evaluation and the related performance for each supplier. A hybrid algorithm integrating particle swarm optimization and simulated annealing is also proposed to search for the optimal solution for the weight calculation. To give sufficient consideration to each analyst's evaluations on each supplier in each phase, the self-feedback neural network is then developed to predicte the comprehend performance for each supplier. A numerical example is also presented to interpret the above methodology.(3) A Mamdani fuzzy inference system (FIS) model is developed to simulate the complex relations among the quantity, quality and price of return products. Based on the proposed FIS model, two CLSC profit models are then proposed for perfect substitute mode and non substitute mode, in which the remanufactured products could be sent to different markets. A modified niching particle swarm optimization is then proposed to search for the optima that could be proved effectively in the following experiments. Finally, the analysis of the single remanufacturing mode and manufacturing/remanufacturing mode in the same market is also taken into account by the discussion of the impacts of parameter variation for the two modes on the price, sales quantity, production type and profit, respectively.(4) Considering the existence of different uncertainties, the fuzzy random variable is applied to describe the uncertainties in the CLSC facility location. Based on the analysis of CLSC facility location problem, a fuzzy random Maximax chance constraint quadratic programming model is put forward. For one type of LR fuzzy random variable, the chance constraint programming model is transferred to a crisp one by using fuzzy and stochastic theories. To compute the optima, a hybrid particle swarm optimization with double populations (HPSODP) is also proposed and the following experiment proves its efficiency at last.