Diagnosis of bearing faults in induction motors by vibration signals - Comparison of multiple signal processing approaches

Abstract

Early detection of faults in the bearings of electric motors is vital to reduce maintenance costs of industrial motors. Vibration signal analysis is a well-known and widely used diagnostic approach for bearing fault identification, and usually leads to good results in terms of effectiveness and detection capability. However, small defects, at an early stage of development, can be hard to find and require advanced signal processing techniques to facilitate the extraction of the fault characteristic frequencies from the noisy vibration signals. This work compares three different techniques applied to vibration signals to facilitate the extraction of the fault frequency components, namely the Teager-Kaiser operator, discrete wavelet transform and the Hilbert transform. A test bench was built and several types of defects were introduced in the motor bearings to compare vibration signals obtained with a healthy and a faulty motor. Comparative graphs of the results obtained with the three techniques are presented and the results are discussed.