Solving Complex Reliability Optimization Problems Using Only a Standard Mixed-Integer Programming Solver

This article shows that recent advances in mathematical-programming solvers now allow some commercially available tools to efficiently tackle complex system-reliability optimization problems. This simple, flexible and widely applicable approach eliminates the requirement for relatively sophisticated, problem-tailored algorithms or significant implementation effort, both of which may pose challenges for some of the approaches proposed in the literature. Although convergence toward the optimum may be slow in some cases, premature termination of the solver or adjustment of the solution accuracy allows for the rapid production of results that, while potentially of slightly lower quality, are still practically valuable. Experimental results on classical problems, several of which have a long history in the literature, consistently demonstrate performance that matches or exceeds previously reported results. Despite often modest improvementsunsurprising given the high quality of published solutionsthe approach is appealing for the minimal effort it requires. Detailed data and results are provided for each experiment to facilitate both the adoption of the approach by researchers and practitioners, and its extension in future research.