Fault Tree Analysis Using Bayesian Optimization: A Reliable and Effective Fault Diagnosis Approaches

Abstract

Fault tree analysis (FTA) is a powerful technique that can be used to analyze the faults related to a large-scale complicated system. In the conventional FTA, it can be used to assess the effects of combinations of failures on system behavior but the ambiguities and uncertainties of basic events cannot be handled effectively. Therefore, employing Bayesian theory helps probabilistic estimation of basic events and subsequently the top event. This study presents an integrated approach to Bayesian theory and FTA for handling uncertainty in the fault diagnosis of heavy CNC machines. In this context, making use of a directed graph to describe the fault, and then expressed as reachability matrix, the fault tree and fault information are obtained. In order to set the determinate conditions at every node of fault tree combining FTA with rule reasoning, the minimal cut sets are determined by representing the fault tree as a top-down logical equation. Finally, Bayesian method is integrated into the fault tree diagnostic method to calculate the posterior probability triggered by each fault tree to locate where on the fault tree and ensure high efficiency of fault diagnosis. The proposed methodology is applied to the fault diagnosis of fuzzy probabilistic analysis in the Φ 160 CNC boring and milling machine. The results indicate that the proposed approach is very reliable and effective in fault diagnosis considering uncertainty reduction or handling.