A computational study of using preprocessing and stronger formulations to solve large general fixed charge problems

Abstract The application of traditional branch-and-bound (B&B) procedures to the standard formulation of the general fixed charge problem (GFCP) has not proven to be an acceptable approach to solving this well-known problem. It has been suggested by various authors that preprocessing the data and/or adding constraints to the GFCP might lead to improved B&B performance. In this paper, we present extensive computational results to test these hypotheses using the M PSX/MIP/370 package. The results demonstrate that while preprocessing the data is not always effective, simple constraints may be added to the problem allowing us to solve randomly generated GFCPs that are significantly larger than any previously reported in the literature. Results arc also presented which give insight into the effectiveness of a frequently used heuristic for the GFCP.