Case study for condition-based maintenance combined with on-line monitoring
Résumé
We consider a gradually deteriorating system subject to changes in its deterioration rate and the aim is to propose a maintenance model to avoid failures and their costly consequences. The system condition is described by a scalar ageing variable which increases with the system deterioration and the failure occurs as soon as the system state crosses a known fixed threshold called failure threshold. A stochastic process (gamma process) is used to model the deterioration. The system is periodically inspected and to avoid failures preventive maintenance actions take place if the deterioration level of the system exceeds a preventive threshold. The preventive threshold is lower than the failure threshold. Based on available data through monitoring an on-line change detection algorithm is used to estimate the change time in the deterioration rate with an acceptable delay. If the deterioration parameters after the change time are unknown they should be estimated in the framework of the on-line detection method. At each inspection time based on the collected information it can be decided which kind of maintenance action (corrective, preventive or nothing) should be implemented. It is sensible to apply an on-line readjustment of the maintenance parameters according to the obtained information through the detection/estimation procedure. Therefore we propose an adaptive condition-based maintenance policy in which the maintenance parameters after the change detection time are readjusted to the estimated values. This policy is applied to a vehicle maintenance problem. The vehicle usage profi le could change for many reasons: speci cation errors, changes in the conditions of use.... It is necessary to detect the gap between the specifi ed and current vehicle usage profi le where vehicle usage profi le is de ned a priori from to the given usage parameters. Afterward the maintenance parameters are readjusted, which will be more adapted to the current usage.