Using proportional-intensities models to schedule preventive-maintenance intervals

This paper is concerned with developing realistic models to determine preventive-maintenance (PM) schedules for complex systems when data have been collected on failure times and PM interventions along with down times and man-hours expended on each. The aim is to develop procedures for analysing such data in order to identify relevant cost-availability models for optimally scheduling PM, while allowing for system deterioration or improvement. We propose the use of a proportional-intensities model and consider the addition of Bayesian methods to incorporate prior knowledge about regression coefficients. The paper discusses further aspects of this approach, which include extending the model to cover time-dependent explanatory variables, and presents an application of the proportional-intensities model to actual reliability data from major industrial plants.