RELIABILITY OPTIMIZATION OF SYSTEMS BY A SURROGATE-CONSTRAINTS ALGORITHM

This paper presents a solution method for the problem of optimizing both redundancy (number of redundant components) and component reliability in each stage of the system under multiple constraints. It is formulated as a mixed-integer nonlinear programming problem. We apply the surrogate dual method to this problem. Surrogate problems with only one constraint, which appear in the optimization process, are solved by dynamic programming. The solution of the surrogate dual problem is not always feasible in the original problem. That is, a surrogate gap exists. We consider 2 countermeasures to surrogate gaps: 1) Modifying the original problem to tighten the constraints; this modification is continued until the solution of the surrogate dual Problem of the modified problem becomes feasible in the original problem, and 2) Decreasing component reliabilities in the vertical direction to the tangential plane of the objective function. Our solution method applies to reliability optimization problems for general systems, eg, they need not be series-parallel. This enables us to treat complex systems such as communication networks. Some computational experiences are shown along with a comparison with other approaches; they show the efficiency of our method.