Branch-And-Bound To Optimize Repair Part Inventory Policies Using Simheuristics

Researchers: Ruthairut Wootisarn, Xueying Wang, Liying Zhang

Many organizations manage the inventory of parts used to repair their facilities or their customers’ products. This project was initiated to determine inventory policies for the MBTA central warehouse that stores parts used by their repair garages. It utilizes a branch-and-bound algorithm and a simheuristics approach.

The simheuristic algorithm identifies a good set of periodic review inventory policies for each part in a repair kit. The parameters of interest are the reorder point and the order-up-to level for each part. The branch-and-bound algorithm attempts to find the optimal solution. This technique uses a simulation to analyze feasible solutions that may be better than previous feasible solution.

We use a deterministic model to generate an initial solution for the simulation. Then, a heuristics model is used to calculate the slope of the total cost equation that includes the setup, delay, and holding cost. The slopes are calculated for each unit change in every parameter. The minimum slope is identified and the simulation is run for the new policy. This procedure is continued until the stopping rule is encountered. 

Current efforts concentrate on creating a more robust stopping rule, embedding the branch and bound algorithm within the simulation code, and developing the mechanism for determining the quality of the branch and bound solution.