Fault diagnostic system based on approximate reasoning

Guaranteeing 100% fault free products is becoming an emerging standard in many branches of manufacturing. This paper addresses the design of an end–quality diagnostic system for detection of mechanical faults in electronically commutated motors. One of the main requirements of an end–quality diagnostic system is its ability for detecting faults in their earliest stages. Main issue in detecting final products with such faults lies in the fact that the faulty product is indistinguishable from the fault–free one. In order to overcome this problem, we have performed the feature extraction method using the spectral kurtosis method. These features were used as an input to the fault localization module, which is based on approximate reasoning technique known as Transferable Belief Model (TBM). Results show that the diagnostic system comprising spectral kurtosis and transferable belief model successfully isolates the most common mechanical faults. Additionally the strength of conflict can be used as a measure of certainty of the diagnostic results. The performance of the diagnostic system was evaluated on a batch of 60 motors.