Research On Two-echelon Stochastic Inventory Model With Partial Lost Sales

Supply chain management (SCM) came from classical logistic problems. With the trend of global economic and third part logistics (TPL) developing coming into being, the relation of enterprise becomes much closer. Therefore, if you do not think from SCM view and adopt appropriate concepts, methods and tools, you can not achieve success in global market nowadays. The inventory cost is a primary component of the supply chain cost, and the method for multi-echelon inventory optimization and control must be adopted for the sake of the entire supply chain optimization and control, t present, papers on multi-echelon inventory systems have solved some problems, but they still have limitations. Firstly, they assumed that leadtime are constants. In practice, as the unknowable incidents exist in production, transportation and distribution processes, leadtime usually are uncertain. Secondly, there exist the hypotheses of totally back orders or completely lost sales under the condition of stockout in most literatures. Thirdly, the existing two-echelon inventory approximate model optimum always fixes one variable in the inventory strategy, and then seeks the optimal in the remained parameters, which makes the result can not be the best outcome.This research is based on Supply Chain Management theory, optimization theory, decision-making theory, and inventory management theory. The research methods include model analysis, numeral analysis, comparative analysis and so on. This paper at first discusses a two-echelon distribution inventory system with a central warehouse and multiple retailers, where the retailers face stochastic compound Poisson demand process and normal transportation time. Partial lost sales would be happened when retailers out of stock. And the rate of lost sales is a fuzzy variable. The system is controlled by continuous review installation stock (R, Q) policies. This paper presents an improved approximated method that the leadtime demand of warehouse and retailers are approximated by the normal distribution. Then the fuzzy of the rate of the lost sales is be considered to construct the fuzzy and approximate mode of the stochastic distribution inventory system.In the end, a genetic algorithm is designed to seek the approximate optimization solutions of order policy between warehouse centers and retailers. The final results and the sensitivities of parameters show that our proposed model and optimization approaches can quickly acquire the inventory policy of distribution system. Besides, a conclusion that both the approximate optimization solutions of reorder and cost of units for the retailers are increased with the increasing of means and variances of the corresponding leadtime are arrived at, which furtherly proves that the changing of the random leadtime remarkably influences the enterprises' order policies and the performance of the whole inventory system.