具有可解释性的DRWT-Trans轴承故障诊断方法OA

Deep learning-based bearing fault diagnosis methods often face challenges in extracting suffi-cient features from complex noisy operational data,and suffer from a lack of interpretability in their deci-sion-making processes.To address these challenges,an interpretable DRWT-Trans method was intro-duced for bearing fault diagnosis,to overcomes challenges in feature extraction from noisy operational data and to enhance decision-making transparency.The DRWT-Trans module emploe discrete wavelet decom-position with soft thresholdto for noise reduction and feature extraction.An interpretable module was de-signed for model analysis increase the interpretability of the model decision.The method's validity was tested on the Case Western Reserve University dataset and a factory gearbox dataset.