A Stochastic Model for Indirect Condition Monitoring Using Proportional Covariate Model

This paper introduces a model to make decision on the maintenance of a mechanical component subject to condition monitoring. A stochastic model is used to determine what maintenance action should be taken at a monitoring check and the follow up inspection times. The condition of component has a stochastic relation with measurements. A new state space model is developed and used, to predict the hazard rate and condition monitoring measurements, to indirectly asses the hazard rate of the system. The Proportional Covariate Model (PCM) which was proposed by Yong Sun (2004) was also used to develop the model. The known Kalman Filter was employed to derive the probability of the conditional hazard rate, which is predicted and updated for condition monitoring. The maintenance is being performed based on the estimated hazard rate so that the desired level of reliability is achieved, in a cost effective approach. This approach is validated by using the experimental data obtained from gearboxes which ran and failed on the Mechanical Diagnostic Test Bed (MDTB) at the Penn State University Applied Research Laboratory.