Eng | 

Approximation Algorithms for Stochastic Remanufacturing Inventory Systems

 

Abstract:

Closed-loop supply chain management has gained great momentum in both practice and academic research because of an increasing awareness of sustainable development. With the rising costs of energy and raw materials, more and more companies are now considering  closed-loop supply chain management a strategic activity  that can  lower production and inventory costs and improve profitability. One important operation in closed-loop supply chain management is remanufacturing, which restores returned products to like-new condition and then resell them to satisfy customer demand. Examples of remanufactured products include automotive parts, tires, electronics products, photocopiers, toner cartridges, among others.

The aim of this project is to study approximation algorithms for dynamic remanufacturing systems with stochastic product returns and customer demands. The key feature is that demands and product returns in different periods are correlated. A major domain of applications is remanufacturing systems where product returns in one period depend on demands in the previous periods, which are often observed in practice.  The prior studies on dynamic remanufacturing systems mostly assume independent demands and returns to avoid the well-known curse of dimensionality of the dynamic programming formulation, {\it i.e.}, the state space becomes so large that the optimal policy is computationally intractable. But such simplification greatly hinders the practical applicability of the results. Therefore, we will exploit novel approaches to formulate and analyze such remanufacturing systems. Specifically, we will evaluate decisions and construct approximation policies and algorithms based on alternative cost accounting schemes.  The policies aim to balance various costs such as inventory holding, demand backlogging, and remanufacturing/manufacturing costs. More importantly, these policies are easily computable and implementable and will be proved to have constant worst-case performance guarantees. 

We will firstly analyze a basic model where a remanufacturing firm determines manufacturing and remanufacturing quantities periodically to minimize its costs under correlated demands and product returns. And then we will study more complex systems with product return disposal, manufacturing/remanufacturing capacity constraints, and multi-type of product returns. Each of these extensions will require novel ideas to design cost accounting schemes and derive provably good approximation algorithms.  To further demonstrate the value of our approximation policies, we will compare their performance to the myopic policies that are widely adopted as heuristics for problems with a large state space as well as those that ignore the correlation between returns and demands. This project is the first attempt to develop computationally efficient policies with constant performance guarantees for remanufacturing systems.

 

Project In-charge:    Prof.  Sean Zhou

Duration:  Jan 2013 – Dec 2015

Sponsor(s):   Hong Kong Research Grant Council, UGC