PrimaVera conducts regular (virtual) colloquiums with interesting talks around predictive maintenance.
Speaker: dr. Ayse Sena Eruguz
We consider a multi-component system in which a condition parameter is monitored by a single sensor. Monitoring gives the decision maker some information about the system state, but it does not reveal the exact state of the components. The decision maker infers a belief about the exact state from the current condition signal and the past data, in order to decide when to intervene for maintenance. A maintenance intervention consists of a complete and perfect inspection followed by component replacement decisions. We model this problem as a partially observable Markov decision process. We consider a deterioration process that suitably reflects the deterioration characteristics of a multi-component system and a probabilistic relation between system states and condition signals. Under reasonable conditions, we investigate the structure of the optimal maintenance intervention policy.