Improvement Of The Distribution Performance Of Warehouse Centers Under Uncertain Environment

The adoption of new management philosophies brings new challenges for warehouse systems, including smaller lot-sizes, higher frequency, and shorter response time, particularly for mail order companies which sell products online. The put systems are becoming more and more daily practice. Under this circustmance, the basic function in a warehouse has changed that is to receive items from suppliers, put away them to storage locations according to customer orders, assemble them for shipment. Thus, the put-away efficiency is important for put systems, as it impacts the customer service level. The put systems discussed in the dissertation are characterized by that both the identities of the customers and their demands are unknown before the start of distribution. Based on five frequently used types of storage assignment, the dissertation proposes some new storage assignment policies and studies the impact of these storage assignment policies on the performance of put systems. Under the uncertainty demand of me customers, the redistribution operation will be interactive. The dissertation attempts to solve this problem. The study will enrich the theory of order picking systems. On the other hand, the study will give some new inspirations to warhouse managers and promote the development of warehousing and logistics in China.The main contributions of the paper are summarized as follows:(1) The effect of closest open location and random storage policies on the performance of put systems with no information. First, the dissertation sets up the analytical travel-distance model for put systems, describes the process of put systems and gives some assumptions. The travel distance under full turnover storage policy is treated as the benchmark. Then, the dissertation derives the expected travel distance under random storage and compares the peformance between closest open location policy and random storage policy. Lastly, the reason for that closest open location outperforms random storage is given by simulation.(2) The effect of class-based location policy on the performance of put systems with incomplete information. First, the dissertation derives the expected travel distance under class-based location policy. Through simulation, the travel distance under class-based location, closest open location and random policies are compared. Then, the dissertation gives an analytical analysis about the performance of a modified class-based location policy, which is proposed in Qin et al. (2015) that use closest open location method in subareas instead of random method. Also, he dissertation compares the travel distance under a modified class-based location policy with other polices by simulation. Lastly, the implementation of a modified class-based location policy in put systems is studied, including the customer classification, the number of storage locations in subareas and the number of class for customers.(3) The effect of storage assignment policies in distribution and redistribution area that based on demand estimation on the performance of put systems. The distribution of demand comes from A/F value. Firstly, the dissertation proposes the demand-based storage policy. The method of estimating demand and implementing demand-based storage policy is given. For a special case, we describe the simulation methods on estimating demand and implementing the demand-based storage policy. The travel distance under demand-based storage policy is compared with full turnover storage policy. Secondly, the dissertation proposes the storage assignment policy when customers are assigned multiple locations in distribution area so that the redistribution area is removed, which is compared with full turnover storage policy. We also analyze the performance of this policy by other measures. Finally, we study the storage assignment policies in redistribution area. Direct allocation policy and matching allocation policy are proposed. We define some performance measures to analyze the performance of the two policies.All the studies are based on the real transaction data from a flower auction market. The conclusions are as follows:(1) The analytical results show that closest open location policy outperforms random storage policy, which is caused by the number of empty locations when random storage is adopted. The performance of random storage policy decreases as the number of empty locations increases. The results are also confirmed in simluations. The closest open location policy generates a shorter travel distance when the randomness under random storage policy exists and the customer does not arrive randomly. Also, the shape of the distribution area has an effect on the performance of put systems.(2) From the analytical results, we can find that the class-based location policy gives a shorter travel distance than random storage, and the modified class-based location policy gives a shorter travel distance than closest open location policy. The results are also obtained from simualtions. Customers are classified by the demand volume and probability value rules, the number of storage locations in subareas is determined by the expected number of present customers, and the heuristic method extending the customer classification from 2 classes to K classes, can standardize the implementation process for the modified class-based location policy and improve the performance of the put systmes. The performace of the modified class-based location policy is increasing as the number of classes increases, but the increasing extent reduces with the the number of classes gradually.(3) Demand distribution given by A/F value will simplify the demand estimation and is easy to implement. The performance of demand-based storage policy is affected by the error of demand estimation and the probability of expected getting a discount by demand estimation. As the error of estimating demand decreases and the probability of expected getting a discount increases, the performance of demand-based storage policy will be closer to the performance of full turnover storage policy.(4) The distribution area is enlarged when the customer is assigned multiple locations based on demand estimation, which caused more than twice travel distance under full turnover storage policy. Thus, it is not a better policy to remove the redistribution operation. In redistribution area, compared with direct allocation policy, matching allocation policy can improve the performance of redistribution and raise the decision of the layout design for redistribution area.