Experts' Knowledge Renewal and Maintenance Actions Effectiveness in High-Mix Low-Volume Industries, Using Bayesian Approach
Résumé
Increasing demand diversity have resulted in high-mix low-volume production where success depends on our ability to quickly design and develop new products. This requires sustainable production capacities and efficient equipment utilization which is ensured through appropriate maintenance strategies. At present, these are derived from experts' knowledge, capitalized in FMECA (Failure Mode, Effect and Criticality Analysis) and/or maintenance procedures. (Abu-Samah et al. 2015) found increasing unscheduled breakdowns, failure durations and number of repair actions in each failure as the key challenges while sustaining production capacities in complex production environment. This is an evidence that maintenance based on the historical knowledge is not always effective to cope up with an evolving nature of equipment failure behaviors. Therefore, in this paper, we present an operational methodology based on Bayesian approach and an extended FMECA method to support experts' knowledge renewal and maintenance actions effectiveness. In the proposed methodology, we capitalize and model experts' existing knowledge from FMECA files as an operational Bayesian network (O-BN) to provide real time feedback on poorly executed maintenance actions. The accuracy of O-BN is monitored through drift in maintenance performance measurement (MPM) indicators that results in learning an unsupervised Bayesian network (U-BN) to discover new causal relations from historical data. The structural difference between O-BN and U-BN highlights potential new knowledge which is validated by experts prior to modify existing FMECA and associated maintenance procedures. The proposed methodology is evaluated in a well reputed high-mix low-volume semiconductor production line to demonstrate its ability to dynamically renew experts' knowledge and improve maintenance actions effectiveness.