Fault detection and prognostic of abnormal equipment situations using wavelet decomposition technique

Abstract

Unplanned system failure results in high cost and liabilities for system operators. The implementation of effective prognostics systems would allow identification of faults before the actual occurrence of critical failure thereby avoiding and mitigating system failures. This project focuses on identifying fault characteristics of Marathon Electric AC Induction motor. Broken Rotor Bar, Motor Bearing and Rotor Unbalance Faults would be presented in detail.

Wavelet Packet Decomposition is used to extract a windowed frequency from the vibration signal for Bearing and Unbalance fault. Broken Rotor Bar faults are diagnosed using a new approach. A system is created to train no fault signals and then tested with an unknown fault signal. Analysis is conducted to extract the characteristic fault frequencies and conduct a comparison with the no fault signal counterpart.

Bearing and Unbalance fault are successfully identified; however, Broken Rotor Bar's experiments do not tally with the findings in the existing literature. An alternative method has been utilized. The software system that is developed provides fairly good fault identification capabilities.