Tacholess skidding evaluation and fault feature enhancement base on a two-step speed estimation method for rolling bearings

Abstract Rolling bearing is an indispensable part in rotary machinery, and its fault diagnosis and life prediction have been a hot issue in the field of research and engineering. With the development of fault diagnosis technology, researchers pay a growing attention to skidding which is easy to cause incipient bearing failure. Due to the complex structure and motion process, it is difficult to evaluate skidding degree of in-service rolling bearing. In addition, skidding will further modulate the fault impulse signal and make fault feature extraction more difficult. Especially when the key phase information is not available, the coupling modulation of speed and skidding bring difficulty to precisely evaluate skidding rate through the vibration information. To solve the above problems, a skidding evaluation and fault feature enhancement method for rolling bearings with no key-phasor is proposed in this work. Based on the short-time Fourier transform, this paper proposes a two-step speed estimation approach by introducing iterative Gaussian process regression and phase recalculation process of the impact envelope signal. This method cannot only estimate the reference shaft speed more accurately but also calculate the cage speed delay relative to the theoretical value caused by skidding. The experimental results show that the method is effective in skidding evaluation and fault feature enhancement, which lays a foundation for further study on the performance degradation assessment in the whole life of rolling bearings.