Research On Optimizing Model And Algorithm For Distribution In Supply Chain

Supply chain management (SCM) is a modern management skill which actualizes synthesis, plan, control and coordination of the whole chain structure composed of supplier, manufacturer, distributor, retailer and customer, as well as an advanced management mode that emphasizes horizontal integration, resource-sharing and partner's cooperation for win-win. SCM is an important research field of management science at present, and optimization problem is the key of supply chain management. Basing on the analyses and summaries of internal literatures and researches, this paper is focused on the optimal model for the multi-layer distribution system in supply chain and the hybrid multi-object genetic algorithm, the optimal model for delivery and the multi-object genetic algorithm. First, The model for integrated optimization of multi-layer distribution system's design and stock control is founded. The model is a hybrid integral multi-objective planning model. The sub model for optimizing distribution system's design uses 0-1 hybrid integral planning model. This model mainly involves the fixed construction cost of distribution system, logistic cost between nodes and the cost of manufacture. The sub model of distribution system's stock control uses uncertain planning model. We wishes to design a model in which we can deal with the uncertain demand of nodes by using the safety stock of distribution center through taking into consideration the certain of lead time and the uncertain demand of nodes. The retailer's operation cost comprises the order cost and the hold cost. The delayed-delivery cost, the order cost and the hold cost compose the distribution center's operation cost. Model uses parallel-selection method and hybrid strategy of GASA. The multi-objective genetic algorithm (MOGA) basing on parallel-selection method and the multi-objective hybrid genetic algorithm (MOGASA) basing on parallel-selection method are designed. C++ Builder is applied as the programming language tool to realize the algorithm. Second, the model for integrated optimization of VRP & VFP is founded. When founding the model of VFP, we consider the character of the goods that decides whether they can be put together, and let "compatible coefficient"denote it. We also consider that the goods belonging to the same retail trader should be, to the greatest extent, put into the same vehicle and let "same-vehicle coefficient"denote it. When founding the model of VRP, we consider the uncertain factors of the situation of roads between any nodes. The algorithm for the model is multi-objects genetic algorithm basing on stochastic weight coefficient. C++ Builder is applied as the programming language tool to realize the algorithm The main contributions of this paper are as follows. First, the hybrid integral multi-objective planning model that combines multi-layer distribution system's design and stock control is proposed. Distribution system's design and operation usually connect and influence each other, lone solution for them can't obtain best result. So we integrate the question of distribution system's design with stock control. Considering the demand uncertainty, we found the hybrid integral multi-objective planning model of this question. Second, the model for integrated optimization of VRP & VFP is founded. VRP and VFP are two close connective sub-questions of delivery. We consider influencing and restricting each other of VRP and VFP and make the model have strong practical value. In algorithm, we analyze actual conditions of proposed model in this paper. According to essential theory of multi-objective planning, genetic algorithm and simulation annealing algorithm, we design the multi-objective genetic algorithm (MOGA) basing on parallel-selection method and stochastic weight coefficient respectively. We optimally design coding/decoding method, selection operator, crossover operator and mutation operator in order to raise the efficiency of the algorithm. To enhance local search ability of genetic algorithm, we introduce simulation annealing algorithm ( SA ) to improve the genetic algorithm.