Effects of partitioning strategies on network reliability bounds

Abstract

A difficulty with network reliability algorithms is failure to complete on problems of practical size. Hence algorithms which give improved convergence on upper and lower bounds for the reliability midway through the problem are desirable. This paper introduces a previously unreported idea for partitioning in network reliability algorithms and compares experimental results on convergence rates of the reliability bounds using several strategies. The results indicate that as much as an order of magnitude difference in convergence rates exist between best and worst strategies midway through the problem.<>