Shaft Misalignment Fault Feature Extraction and Diagnosis via MCSA Utilizing Empirical Principal Component Analysis

The fault features of shaft misalignment in stator currents are often suppressed and masked by interference and noise, leading to weak fault features and affecting the accuracy of fault diagnosis. This paper proposes a feature extraction and diagnosis method for motor shaft misalignment through motor current signature analysis (MCSA) based on empirical principal element. The designed power frequency filtering technique is first applied to diminish the dominance of the power frequency in the signal spectrum, thereby improving the representation of other harmonics features. Subsequently, empirical principal component analysis (EPCA) is employed to extract fault features from the current signal indicative of shaft misalignment. The shaft misalignment faults diagnosis is achieved by comparing with the theoretical fault frequency associated with shaft misalignment. The proposed method was validated using the data from motor experimental platform, and was compared with empirical mode decomposition, high-pass filtering, and principal component analysis method. The results confirm the feasibility and effectiveness of the proposed method.