A simulated annealing approach for the composite facility location and resource allocation problem: A study of strategic positioning of United States Air Force munitions

The US Air Force faces the difficult decision of where to strategically preposition munitions stocks in preparation for a variety of possible future wartime scenarios. This problem includes aspects of both the capacitated facility location problem and the resource allocation problem. The problem addressed is considered multi-objective in nature and cost minimization is balanced against minimizing the coverage distances that munitions must be transported to meet demands.; Typical solutions to the facility location problem and resource allocation problem do not take into account the constraints of the logistics environment. Therefore, this study incorporates transportation and facility costs, and uses actual geographic distances with adjustments made for available modes of transportation. Feasible solutions to the composite facility location and resource allocation problem are generated using a simulated annealing algorithm that explores both inventory transfers and location transfers during the course of the search. Simulated annealing is a meta-heuristic technique analogous to the physical annealing of solids and has been successfully used in many operations research problems, but has not been applied to a problem of where to position strategic inventory.; The study uses an experimental design which tests the ability of the algorithm to provide improved solutions to the problem when using different search parameter values. Different inventory transfer sizes are used in the search in order to analyze the effects of repositioning inventory in larger packages than the typical transfer size of one unit. In addition, the search algorithm periodically redirects the search based on the best coverage solution found after a number of iterations. How often to accomplish this redirection is also an experimental factor of the study.; The results of the study indicate that munitions inventories can be pre-positioned to simultaneously improve both objectives of the problem in comparison to the existing initial solution. In addition, it is shown that the cost and coverage values achieved by the model depend on the configuration and size of the problem being solved. Also, the quality of the solutions is dependent on the combination of transfer size and reset frequency used by the algorithm. Improvement in the quality of solutions is evident when using the largest transfer size, and the most improved solutions are found when the transfer size is combined with the largest reset frequency. The results of the study also provide a means for analyzing which warehouse locations should be opened from the set of potential locations and what inventories quantities should be stocked at each location.