Experience with a Stochastic Replacement Model

Abstract

This paper reports on a study initiated to check the validity of an existing six-monthly block replacement policy for a set of filling valves in a canning line. After dealing with inadequacies in the data and changing to a production-based time measure, the paper shows how a renewal function measure of the expected number of valve replacements was obtained and led to the conclusion that the current block replacement policy could not be justified. The possibility of an age-based replacement policy being of value remained. In order to model such a policy, it was necessary to reverse the usual analysis of renewal-type processes in that an estimate was required of the distribution function of time between renewals, given the renewal function. A finite optimal replacement age was found which depended upon the objective being to minimize a measure of downtime due to replacements, or operating costs attributable to replacements. Calculations were further complicated by part of the plant having a finite remaining life before being upgraded.