Methods for Solving a Mixed Integer Program for Semiconductor Supply Chain Optimization at IBM

IBM Systems and Technology Group uses Operations Research models and methodology extensively for solving large-scale supply chain optimization (SCO) problems for planning of its extended enterprise semiconductor supply chain. Centralized supply chain planning systems need to simultaneously consider macro level decisions, such as sourcing among multiple plants, and worldwide shipment logistics, and micro level decisions about production plans for individual plants. The large-scale nature of these problems necessitates the use of computationally efficient solution methods. However the complexity of the models makes development of robust solution methods a challenge. In this article we describe our experiences in developing a mixed-integer-programming (MIP) model and supporting heuristics for optimizing IBM’s semiconductor supply chain. We describe aspects of supply chain planning specific to the semiconductor industry and discuss the MIP model in detail. We present three heuristics we have developed, driven by their practical application, for capturing the discrete aspects of the MIP. The model and methodology we describe could be adapted for use in other industries with similar characteristics.