Managing Capacitated Multi-Echelon Supply Chains
- Yimin YU (Principal Investigator / Project Coordinator)Department of Management Sciences
- Saif BENJAAFAR (Co-Investigator)
DescriptionThe proposed project is partially motivated by recent experiences with some manufacturing firms in Chinaand some recent advances in academics. The coordination of production, storage, and delivery of goodsamong different members in supply chains is of critical importance in supply chain management. It is wellknown that a simple echelon base stock policy is optimal for coordinating serial supply chains with unlimitedproduction capacity. An echelon is a tier in a supply chain, and echelon inventory is the sum of all of thedownstream inventory and inventory in transit to the echelon. However, in practice, different stages in asupply chain are usually subject to (tight) production capacities. With capacity constraints, a supply chain isless responsive to demand spikes and cannot deplete overstock quickly, which can cause excessive fulfillmentdelays and inventory cost if ill-managed. However, our understanding of how to coordinate material flows insupply chains with capacity constraints is quite limited. This project will aim to propose new techniques toanalyze multi-echelon supply chains with capacity constraints to gain insights into how to achieve efficientmaterial flows in these business processes.We will consider a single-item periodic-review capacitated multi-echelon inventory system in which the rawmaterials must go through multiple production stages before being converted into the final product(s). Wewill consider the following supply chain structures: serial systems, assembly systems and possibly distributionsystems. Each production stage is subject to a capacity constraint, which is determined by the productioncapacity of the machines and manpower involved. In addition, there is a fixed production/transportationleadtime between two consecutive stages. Demand for the final products occurs randomly in each period. Afirm's objective is to minimize its expected total discounted cost, which includes production/replenishment,inventory holding and backordering costs, by optimizing its inventory replenishment/production policy.In the extensive literature related to multi-echelon inventory management, only a few reported resultsfor problems with capacity constraints have emerged. These studies have focused on either characterizingthe structure of the optimal policies for two-echelon systems with the leadtime at stage 2 not exceedingtwo periods or proposing heuristics for general serial systems. The structure of an optimal inventory con-trol/production policy for a general capacitated multi-echelon system, e.g., an N-stage serial inventory systemwith fixed leadtimes or an assembly system, is an open area of research due to the intrinsic challenges. Thefundamental question is as follows: what is the structure of the optimal policy for such a system? A fullcharacterization of the optimal policy can potentially reduce the computational complexity. However, anoptimal policy structure may not be easy to implement, even if we have a full characterization. The secondyet important question relates to whether any effective simple heuristics exist.To tackle these challenging issues, this project will propose novel mathematical techniques to the problem.It will then use optimization tools and the new mathematical techniques to analyze and solve the multi-echelon inventory problem with capacity constraints to obtain a full characterization of the optimal policystructure. Finally, it will provide bounds for the optimal policy and propose simple and effective algorithmswith low computational complexity. The performance properties of the proposed heuristics will be analyzedand evaluated based on data collected from real-world supply chains.Preliminary analytical results are presented in this proposal along with the tasks necessary to achieve ourstated research goals. Our preliminary results for related problems and the techniques developed duringthose experiences have encouraged us to take up the challenge described here.
|Effective start/end date||1/11/15 → 31/01/20|
- dynamic programming,multi-echelon,capacitated,optimal policy,