Hybrid De-warping of Non-stationary Vibrations for Bearing Fault Diagnosis under Variable Speed Conditions

Rotating machinery often operates under non-stationary conditions, posing particular challenges for the fault diagnosis of bearings. A common approach for processing bearing vibrations at variable rotational speeds involves resampling the signals and converting them into corresponding vibrations at a constant speed. This process is also known as de-warping. However, the current de-warping methods do not account for the different warping mechanisms between pulse trains and other components in fault-bearing vibrations. Therefore, there is a certain degree of error in the de-warped waveform. This paper proposes a hybrid de-warping method for processing non-stationary bearing vibration signals. The core idea is to segment the vibrations into pulse areas and other waveform-distortion areas: a global shift is performed in the pulse-based areas; while a scale transform is implemented in the remaining parts. The scale transform is achieved through the warping path calculated by the proposed unilateral dynamic time warping (UDTW) algorithm. The proposed method exhibits superior de-warping accuracy and effectively restores the regular statistical properties of the vibration. The effectiveness of the proposed method is validated through numerical simulations and experimental case studies.