Determining bearing faults using wavelet and approximate entropy

Bearing faults appear to be one of the main factors that cause the interruption of automation processes. In this paper "Wavelet Analysis" and "Approximate Entropy" was applied to vibration data taken from a shaft-bearing setup to predict the presence and development of bearing faults. The purpose is to distinguish the normal and the defective bearing sorted by the degree of defectiveness, using wavelet packet analysis and approximate entropy with high frequency demodulated raw vibration data. Normal bearing with low amplitude but numerous frequency components, and defective bearing with a characteristic of high amplitudes and certain frequency components can be distinguished by approximate entropy. Wavelet packet analysis can be applied to the derived data to get which frequency components are more essential for getting better results.