IMPROVED QUALITY OF INPUT DATA FOR MAINTENANCE OPTIMIZATION USING EXPERT JUDGMENT

Most maintenance optimization models need an estimate of the so-called 'naked' failure rate function as input. In practice it is very diffic ult to estimate the 'naked' failure rate, because overhauls and other preventive maintenance actions tend to 'corrupt' the recorded lifeleng ths. The purpose of this paper is to stress the importance of utilizin g the knowledge of maintenance engineers, i.e., expert judgment, in ad dition to recorded equipment lifelengths, in order to get credible inp ut data. We have shown that without utilizing expert judgment, the est imated mean time to failure may be strongly biased, often by a factor of 2-3, depending on the life distribution that is assumed. We recomme nd including a simple question about the mean remaining lifelength on the work-order forms. By this approach the knowledge of maintenance en gineers may be incorporated in a simple and cost-effective way. (C) 19 98 Elsevier Science Limited.