Feature Enhancement and Extraction Based on Amplitude Modulation for Bearing Fault

Kurtogram have made significant contributions to signal decomposition and fault information screening, but the fault decomposition process does not consider the enhancement of weak features. This paper proposes an amplitude modulation-assisted spectral segmentation and enhancement diagram (AMHarmogram) to extract and enhance fault information in bearing vibration data. A trend spectrum-based modal separation method is proposed to expand the flexibility of fixed frames, and a novel diagram is presented to display and separate possible fault features. The modulation factor was designed to exponentially enhance the amplitude of the spectrum within the frequency band based on the Fourier transform characteristics, thereby amplifying the fault characteristics. Filtering and reconstruction of each segmented frequency band rely on a filter bank constructed using Meyer wavelets. A well-developed theory and the existence of transition bands ensure the rationality of the reconstruction. To improve the accuracy of feature screening results, the Spectral negentropy index, which is more sensitive to cyclostationary characteristics, is used in the diagram. The effectiveness of the algorithm was verified using a high-speed bearing test rig at Mie University in Japan. The proposed method offers improvements in signal-to-noise ratio and is capable of detecting faults in both the inner and outer races of bearings.