Perishable Inventory Models With Multiple Demand Classes And Their Applications

Effective inventory management of perishable products is critical to business suc-cess and environmental sustainability.Related literature shows that fresh foods such as fruits,vegetables,meat,fish,dairy,and baked goods typically account for up to 40%of grocery chains' revenues,and they are also strong drivers of store traffic and customer loyalty.In the meanwhile,one-third of all food produced for human consumption is wasted with estimated associated costs equal to US$ 990 billion every year.At the re-tailing level,for example,the percentage of waste in the U.S.is around 11%for dairy products and 4%for meat products;for fruit and vegetables,waste is 8-%in the U.S.and in Switzerland,while studies in Sweden,Norway and Austria report waste in the range of 4%to 5%.For platelets,the percentage of waste of apheresis platelets and whole-blood-derived platelets in U.S.is about 10.9%and 22.2%respectively in 2006.All these figures illustrate that waste at the retail level is significant.As a results,per-ishable inventory management is an important research topic.In this thesis,we will consider the following two perishable inventory management models that cater to per-ishable products of different types.The first research problem investigates the inventory management of blood platelet.Due to the uncertain supply(and the procedure of whole blood collection),we will con-sider joint decisions for blood collection and platelet inventory control in a central blood bank.We fully characterize the structure of the optimal policy regarding whole blood collection,platelet production,and inventory issuing,rationing and disposal.We find that the optimal platelet production quantity in each period is nonincreasing in the in-ventory levels of platelets and whole blood but that interestingly,the optimal blood collection effort may increase with the on-hand platelet inventory level.We demon-strate with a real data set that joint decision making leads to significant cost savings compared with separate decision making.The benefit is mainly derived from reduced blood collection and platelet production,better utilization of the collected whole blood,and reduced platelet shortage.For practical implementation,we develop a lookahead heuristic,which is shown to be very effective by numerical experiments.The second research problem considers the management of perishable products in retailing.Due to preference of customers on freshness,we will consider the inventory control problem in perishable inventory systems with multiple demand classes,each characterized by a different lost-sale cost and the least freshness requirement.Demands of different classes in the same period could be correlated,while demands across peri-ods are independent but not necessarily identical.In each period,the firm jointly makes the decisions regarding inventory issuing,rationing,production/ordering and disposal.The objective is to minimize the expected total discounted cost over the entire plan-ning horizon including linear ordering cost,inventory holding/lost-sales cost,expiration cost and disposal cost.By establishing new properties of multimodularity,we explore some monotonicity and bounded sensitivity properties of the optimal policies.Based on the optimality analysis,we then propose a novel approximation approach,called adaptive approximation approach,which can be recursively calculated through a single-dimension dynamic program.Numerical studies demonstrate that the proposed approx-imation approach is nearly optimal with the average optimality gap around 0.30%and significantly outperforms the existing heuristics studied in the literature.In practice,we notice that neither retailers nor e-tailers incorporate customers' freshness requirements into their practical operations.However,our numerical studies show that ignoring cus-tomers' freshness requirements can lead to a significant increase in total costs,especially for online retailing.