Fault diagnosis of variable speed rolling bearings based on COT-EWT-LSTM

Under complex variable-speed conditions, rolling bearing fault signals are highly susceptible to noise interference, which often results in weakened feature representations and spectral aliasing. These effects significantly hinder fault identifiability and reduce the effectiveness of traditional classification methods. To address these challenges, a fault classification methodology combining Computed Order Tracking, Empirical Wavelet Transform, and Long Short-Term Memory (COT-EWT-LSTM) is proposed. In the initial stage, Computed Order Tracking converts the original time-domain signal into a smooth angular-domain representation, mitigating the influence of rotational speed fluctuations. The signal is then decomposed using the Empirical Wavelet Transform, and the frequency band exhibiting the highest kurtosis is selected to extract the most informative time-frequency features. Subsequently, these features are fed into a Long Short-Term Memory network for fault classification. Experimental results demonstrate that the proposed method achieves an overall classification accuracy of 93%, with particularly improved identification of ball faults, confirming its effectiveness and superior performance under variable-speed conditions.