A Bayesian Networks Approach to Fleet Availability Analysis Considering Managerial and Complex Causal Factors

Availability analysis of a fleet of assets requires modelling uncertainty sources that affect equipment reliability and maintainability. These uncertainties include complex, managerial causalities and risks which have been seldom examined in the asset management literature. The objective of this study is to measure the reliability, maintainability and availability of a fleet, considering the effect of common causal factors and extremely rare or previously unobserved events. We develop a fully probabilistic availability analysis model using hybrid Bayesian networks (BNs), to capture managerial, organisational and environmental causal factors that influence failure or repair rate, as well as those that affect both failure and repair rates simultaneously. The proposed methodology has been found more accurate in forecasting failure rate, repair rate, and average availability level of a fleet of assets, providing asset managers with an inference mechanism to not only measure the performance of the assets based on common causal factors, but also learn the actual level of such factors and thereby identify improvement areas. We have demonstrated the application of the model using a fleet of excavators located in Toronto, Ontario. The prediction accuracy of the proposed model is evaluated by use of a measure of prediction error. [Received: 19 March 2019; Accepted: 3 September 2019]