Multi-sensor Data Fusion for Rotating Machinery Fault Detection Using Improved Cyclic Spectral Covariance Matrix and Motor Current Signal Analysis

When an abnormal situation occurs in rotating machinery, fault feature information may be scattered on multiple sensors, and fault feature extraction through a single sensor is not enough for fault detection. Moreover, fault detection techniques based on vibration signals are commonly applied to monitor the health of rotating machinery. However, the installation of vibration sensor is inconvenient, which will greatly affect collected signal and thus influence detection effect. This paper proposes a novel method with improved cyclic spectral covariance matrix (ICSCM) and motor current signal analysis, which achieves multi-sensor data fusion for rotating machinery fault detection. Firstly, an improved cyclic spectral is proposed to process multi-sensor signals collected from rotating machinery, which adaptively acquires multi-sensor mode components. Subsequently, sample entropy of acquired mode components is utilized to construct the ICSCM, which can fully preserve the interaction relationship between different sensors. Finally, ICSCM is incorporated into extreme learning machine classifier to identify different fault types for rotating machinery. The merits of the proposed method are validated using two datasets.
Analysis results demonstrate that the proposed method has achieved satisfactory results and more reliable diagnosis accuracy than other state-of-the-art algorithms in rotating machinery fault detection.