Feature Denoising-based Fault Diagnosis for Rotating machinery

Machinery health condition identification is crucial to reduce machine downtime and ensure its normal and continuous operation. This study proposes a feature denoising-based fault diagnosis method for rotating machinery. In the proposed method, raw signals are firstly decomposed into several subsignals of different frequency bands using the empirical mode decomposition. Based on these subsignals, multiple fault features are extracted. Then, variational mode decomposition (VMD)-based feature denoising technique is used to process the obtained features. Finally, a random forest classifier is applied to identify different machinery faults. Experimental results show that the VMD-based feature denoising approach can effectively remove the noise data and largely improve the classification performance.