A Novel Bearing Fault Diagnosis Method Under Signal Loss Based on LSTM Feature Compensation

In bearing fault diagnosis, sensor signal acquisition can suffer severe data loss due to hardware failures or external interference, making it challenging for traditional methods to effectively handle such incomplete data. To address this issue, this paper proposes a novel bearing fault diagnosis method under signal loss based on Long Short-Term Memory network Feature Compensation (LSTM_FC). First, Convolutional Neural Network (CNN) is utilized to extract deep features from bearing signals. Subsequently, LSTM is applied to compensate for the missing features, generating a complete feature representation. Finally, Global Average Pooling (GAP) and fully connected layers are used to perform the classification task. The proposed method is validated on the Case Western Reserve University (CWRU) bearing dataset. Experimental results demonstrate that even with 60% signal loss, the method achieves a classification accuracy exceeding 80%, showcasing its strong robustness and potential for practical applications.