Optimization And Simulation Of Multi-echelon Inventory In Supply Chain Under Stochastic Demand Based On Control Theory

Multi-echelon inventory optimization in supply chain system has always been a hot topic both in theoretical research and practical applications.With the increasingly fierce market competition,customer demand for different commodities has become more diversified and unstable.It is particularly important to reduce inventory levels throughout the supply chain when the stochastic customer demands are met.In order to improve the stability of the multi-stage supply chain inventory system and the comprehensive competitiveness of enterprises,it is necessary to control the input and output of the inventory system.In this thesis,the research achievements and shortcomings of the previous studies were discussed,and a multi-echelon inventory feedback control systemwith stochastic demand based on nonlinear autoregressive neural network with information distortion compensator was proposed and built.This system was also tested both in simulation and comparison researches on Matlab.In this thesis,discussions on the logical model for single-stage inventory systems with dynamic market demand were carried out to clearly point out the purchase-sell-stock process of an inventory system.Firstly,a dynamic mathematic model for inventory was proposed based on the Forward difference approximation and the Taylor expansion.Secondly,the dynamic mathematic model was transferred to a double-input feedback control model with stochastic demand as disturbance.Based on that,a single-stage inventory feedback control system was carried out.Thirdly,simulation researches on SIMULINK were conducted and the exponential smoothing forecasting method was replaced by a non-linear autoregressive(NAR)neural network to better forecast the stochastic demand.A comparison of the simulated results was also made to indicate the superiority of the NAR neural network.With the same hypothesis that the demand is stochastic,this thesis analyzed the actual product flow,the information flow and the capital flow in the supply chain.Firstly,a group of dynamic mathematic models for inventory were proposed with similar hypothesis.Secondly,the mathematic models were transferred to feedback control models and simulated on the SIMULINK platform.In the meanwhile,the problem ofdemand amplification,also known as the bullwhip effect was spotted in the simulation results.Thirdly,three optimization methods were conducted trying to solve the bullwhip problem.The three methods were information sharing,implication of the NAR network and information distortion compensator.The proposed information distortion compensator can narrow the fluctuation of the bullwhip while maintain the mean of the order quantity.Fourthly,A group of comparisons based on the simulation results were conducted which includes the average inventory,the frequency of stock out,the average stock out,the bullwhip effect of ordering and the bullwhip effect of inventory.The results show that the multi-echelon inventory feedback control system with stochastic demand based on nonlinear autoregressive neural network with information distortion compensator can optimize the complex multi-stage supply chain inventory system in a significant manner.The strategy of information sharing ensures the validity and authenticity of the information transfer,the NAR network helps improve the prediction accuracy and the information distortion compensator narrows the fluctuations of the bullwhip effect.The proposed system can help solve a multi-echelon inventory optimization problem,ease the bullwhip effect in supply chain and reduce overall inventory levels.It can put a certain theoretical guidance and reference on the problem of multi-echelon inventory in supply chain under stochastic demand.