A hybrid model for solving single source capacitated facility location problem

Facility location problems are often encountered in application areas such as material distribution, transportation, supply chain management, and telecommunication networks. This paper deals with a facility location problem where each customer is served by a single source chosen from a preset number of facilities. The locations of facilities will be determined in a continuous Euclidean space, while the allocation of a facility to a customer will be decided as well. This location-allocation problem can be seen as a mixed integer nonlinear programming (MINLP) problem. Traditional methods for solving such a NP-hard problem include the Branch & Bound and Alternate Location-Allocation approaches. In this paper, we propose a novel approach combining genetic algorithm and integer programming to find a near-optimal global solution for the MINLP problem. Experimental results compare favorably with a well-established Internet based optimizer (the NEOS server). This methodology can be easily extended to other MINLP problems.