The Dynamic Analysis And Control Of Supply Chain System Based On Order-up-to Policy

The dynamics in the form of bullwhip effect and inventory variance amplification lead to huge inefficiency in supply chain system. They could mislead the production plan; generate too much waste in inventory investment and big reduction in profit. At the same time they could kill the customer service level; bring in transportation inefficiency and so on. Hence, it's meaningful to study systematically the dynamics of supply chain system, and to understand its operation mechanism, influencing factors, control policy and so on.This paper is under the assumptions of first order auto-regressive (AR(1)) demand process and minimum mean square error (MMSE) forecasting. The main influencing factors including net stock level, ordering policy, production or distribution lead-time, information sharing and so on, are analyzed elaborately. In the end, without the assumptions of certain demand process and ordering policy in advance, the dynamic analysis and optimized modeling is implemented for complicated supply network. The main contents and the results of the research are as following:(1) The control engineering technique is applied to supply chain system. By adjusting the key parameters the two kinds of variance amplification ratio are quantified and optimized. Under the assumptions of AR(1) demand pattern and MMSE forecasting, it proves the simple order-up-to (OUT) model is not always arouse bullwhip effect. By introducing the weight factor to variance amplification ratios of smoothing OUT model, the optimum design of parameters is found to balance bullwhip effect and the net stock variance amplification.(2) The effect of lead-time misidentification to variance ratios of simple OUT model is analyzed. The results show quantitatively the importance of conservative estimation. By applying the initial and final value theorems to net stock level, the inventory drift phenomenon is exhibited. Then the design of zero inventory drift ordering policy is obtained, which can eliminate the inherent inventory drift caused by OUT policy while keeping the variance ratios of supply chain system in a reasonable range. The obtained new model has two more valuable attributes besides zero inventory drift, one is that both order and net stock variance ratios decrease as the value of fractional controller increases, the other is that zero inventory drift model variants are more suitable to deal with the situation with random lead-times than traditional OUT model.(3) Starting with the analysis of supply chain system, the lead-time is treated as endogenous variable to fit the probability distribution. An iterative procedure is presented to solve the dependency problem between order pattern and the lead-time. By embedding the analysis of phase type distribution, queuing model and markov chain to the iterative procedure, the convergent value of the final lead-time distribution is obtained. By considering the influence of the ordering policy to the lead-time, a win-win solution is obtained to get rid of the conflict between bullwhip effect and the net stock variance amplification in supply chain system.(4) The upstream player in supply chain already has enough information of historical ordering datum to estimate the downstream player's demand during lead-time. So demand information sharing provides no remarkable value to the improvement of the supply chain performance. Subsequently, the cost fuction concerning to order and net stock standard deviation is optimized in the environment of two-stage and multi-stage supply chain. A few cooperation mechanisms are proposed based on the width and depth of information sharing. It shows that the value of information sharing is not just about the less uncertainty in demand but the management of supply chain dynamics. In other words, the performance of the entire supply chain can be optimized by sharing deep information including ordering policies, cost coefficients, motivation requirements and so on.(5) The CL-CROC model is applied to supply chain field. In the existence of all the uncertainties, CL-CROC model is able to respond to current inventory state in real time. It has no need to make assumptions about the demand pattern, ordering policy and forecasting method in advance. CL-CROC model can analyze supply chain dynamics and handle system uncertainty very well. It's a comprehensive technique to control desired net stock level and ordering rules.