Optimization of systems reliability under discrete design alternatives

Abstract

For a series system, the problem of optimizing systems reliability under discrete design alternatives at each subsystem is initially formulated as a nonlinear 0-1 program with multiple-choice constraints. Different types of methods for achieving high systems reliability (parallel redundancy, standby redundancy, an increase in component reliability, etc.) can be easily handled as discrete design alternatives. In order to solve the problem efficiently, the nonlinear 0-1 programming problem is transformed into a linear 0-1 program. As a solution method this study presents a branch-and-bound technique with Lagrangian relaxation which provides exact optimal solutions. Further, a heuristic method by Chang and Tcha is also considered. Test problems are solved by both methods and their computational efficiencies are reported. For a class of complex systems, the possibility of obtaining a 0-1 linear program is shown, and an example problem is solved by APEX IV.