A military fleet mix problem for high-valued defense assets: A simulation-based optimization approach

The military fleet mix problem is traditionally used when several modernization operations are required to transition a force from an outdated fleet to a more modern one over a planning period. The main objective of such a transition is to enhance the capability of a military fleet to deploy its forces for missions at any moment by continually improving its operational readiness with minimum cost and time. The complexity of the problem is significantly increased with the increase in the number of assets to be modernized, sub-divisions (Brigades) and the length of the planning period. In this paper, to effectively address the fleet mix problem in the military context, a simulation-optimization approach is introduced with the aim of providing decision -makers with a fleet modernization schedule that enables the military to achieve maximum deployment of its available assets/forces while minimizing the total cost and full deployment time. The proposed approach uses a combination of (1) an enhanced genetic algorithm (GA) to generate initial solutions, explore and exploit the generated solutions till achieving the optimal one, and (2) a capability simulation model to evaluate each solution and determine its quality. Furthermore, the proposed GA comprises several enhanced components to further improve its performance and hence find the optimal solution in lesser computational time. (a) A heuristic repairing method, to provide feasible solutions through repairing a certain number of solutions from the initial population which can significantly guide GA search towards (near-)optimal solutions, (b) a new solution representation, that better represents the fleet schedule decision variables and characteristics, (c) improved designs of crossover and mutations, that can help GA to reach optimum solutions in a faster manner, and (d) new sorting and selection functions, that use many objectives values to allow the better solutions to survive through generations, are all examples of such enhanced GA's components. The proposed approach has been validated by solving a case study that addresses recent fleet modernization strategies of the Australian Army to recapitalize its forces over the next decade and in a continual process. Two scenarios of the case study with different scales have been used. The experimental results show that the proposed approach can effectively provide efficient modernization fleet schedules for both scenarios which allow the maximum deployment with the minimum cost and time. Compared with traditional GA and Differential Evolution (DE), the proposed approach yields 2.37% and 7.17% more deployability, and 99.59% and 98.81% less total cost than DE and GA, respectively. A surrogate model is also presented to replace the computationally expensive simulation model for fitness evaluation. The results show that the surrogate model can significantly reduce the computational time but its accuracy for expecting the optimal objectives still need more enhancements.