Purpose: To propose a Bayesian approach for failure time prediction of degrading components from condition data that accounts for uncertainties and incorporates prior information on degradation behavior based on posterior sampling to handle general statistical models Summary: With efficient use of monitoring data, reliability prediction models can be made more accurate and there by unnecessary repairs are eliminated and occurrence of failures is reduced. This article proposes a Bayesian approach for making reliability predictions of degrading components accounting for sampling uncertainty, update the failure time distribution with real-time condition monitoring data, and improve maintenance decisions. The goal of condition- based maintenance is to do repair or replacements before parts fail to achieve minimum breakdown and to ensure low maintenance. This is unlike the traditional maintenance policies that are based on the system age or occurrence of breakdowns. In traditional preventive maintenance, the inspection and maintenance activities are performed at regular time intervals determined based on failure time data analysis from a set of degradation experiments. In condition-based maintenance, on the other hand, measurements of a degradation variable, collected during the use of the product, are analyzed to plan maintenance activities. When a device exhibits a predictable degradation pattern, in situ sensory data are used to make predictions of future degradation levels. In situ sensor data can be combined with population-level reliability models to infer component-level degradation properties more precisely. A random coefficient growth model is utilized to model degradation in this study. The failure time distribution is obtained from a degradation experiment. The distribution is then updated with condition data from an individual component. The effect of the amount of condition data used in updating on the maintenance actions is investigated. (26 refs.)
展开▼
