FMECA-BASED OPTIMIZATION APPROACHES UNDER AN EVIDENTIAL REASONING FRAMEWORK

Abstract

One of the major shortcomings of traditional failure modes, effects and criticality analysis is the absence of any interconnection between failure ranking and a procedure for selecting the most critical maintenance/improvement tasks to be carried out. This limits the potential of FMECA for implementation in real environments. In order to bridge this gap, three different 0-1 knapsack models have been formulated. The first aims to select the failures in order to maximise cost savings. The second enriches the selection problem by also taking into account the probabilities of solving the failures with a set of maintenance tasks. The third aims to select the maintenance tasks to maximise the expected profit. In particular, the last two models make use of an evidential reasoning framework to deal with the epistemic uncertainty related to these probabilities. A dataset from a manufacturer of lift winches has been used to validate this proposal, as well as to comment on the need for group decision support systems that are capable of converting the FMECA ranking into maintenance tasks in real environments.