Determining an Optimal Policy for Skipping Scheduled Maintenance

In this paper we summarize the study that was carried out for a system (machine) that is subject to random failures. We assume here that failures result in a renewal of the system. Most typically such systems exhibit a rising failure intensity and hence there is an incentive to perform a preventive replacement (PR) and thus avoid a potentially costly emergency renewal (ER). Two forms of PR policies seem common in industry and appear in the literature. One approach requires determination of the optimal interval between PR's. The idea here is to schedule PRs on a fixed cycle (regardless whether or not failures have occurred in the intervai) From a maintenance scheduling point of view, such an approach is easy to implement and offers considerable economies if major resources need to be organized in order to carry out the PR. Utilizing straightforward renewal arguments, an optimal interval is easily determined. [I, pg48]. A serious drawback of this approach is the possibility of performing a PR shortly after a failure replacement (an ER) Partly to circumvent this, a second approach is to find the optimal replacement age. Here a system that does not fail is left running to a predetermined age at which time the PR is carried out. [1,pg61), [2,pg87). The drawback in this approach is that because the next PR is based on age since last event (ER or PR), the time between PR's becomes random. In reality, most of the research on age replacement models has tended to assume that emergency failures are minimally repaired More recently considerable attention has focused on the concept of imperfect repairs, in which the repair partially resets the failure intensity of the system . Our concern is to focus on the first of these approaches and to propose an improved policy for fixed cycle time maintenance planning. It seems clear that if a firm wished to follow a fixed cycle time PR model, they would not like to schedule a PR if the system had only recently failed and received an ER. Assuming the system operates with an increasing failure intensity, the ER would have refreshed the intensity of the system. Hence there is little to be gained in shutting down at the end of the cycle time to incur the cost of again resetting its intensity. A better approach is to skip the imminent PR and plan on having the PR at the next cycle time The important question then is to determine a length of time before the imminent PR in which you should decide to skip the upcoming PR and plan on performing the next PR one cycle out? Thus the approach is to identify a cutoff time, beyond which, if a failure (and thus an ER) occurs, it is optimal to skip the imminently scheduled PR and wait until the end of the next cycle Conversely, before this cutoff point, it is still optimal to carry out the PR at its scheduled time Thus in this way the wasted expenditure of a PR just after one or more ER's can be avoidcd In this study, a discrete semi-Markov model

Determining an Optimal Policy for Skipping Scheduled Maintenance Keywords:

Preventive Maintenance , Emergency Maintenance , Fixed Cycle Times