Fault Diagnosis Using Extended Belief Rule-Based Systems with Novel Rule Weight Calculation Method

Abstract

Fault diagnosis is a common problem during the process of complex system maintenance. The rule-based systems is particular fitted to address the diagnosing problem due to its interpretability. By transferring the measured data into belief rules, the extended belief rule-based systems(EBRBS) combines the advantage of both data driven method and the structure of rule-based systems and is proved to be effective in addressing many real word problems. Nevertheless, counterintuitive rule weight calculating procedures, that may deteriorate the performance of EBRBS, exists in the original method. In this study, we point out the existing problem and proposed a new rule weight calculation method. Afterwards, the proposed method is utilized to address real-world liquid ultrasonic flow meter diagnosis problems. The results shows the effectiveness of the proposed method.