Research on Feature Correlation Dimension Extraction Method and Its Application on Rolling Bearing Fault Diagnosis

Aiming at the nonlinear and non-stationary characteristics of rolling bearing fault vibration signals, feature correlation dimension (FCD) extraction method combining phase space reconstruction (PSR) and fractal theory is proposed. The method reconstructs high-dimension phase space from one-dimension vibration signals of different fault states of the rolling bearing so that deep data mining is achieved. Then through analysis of varying correlation dimension of phase space feature signals, FCD that corresponds to each fault state is extracted. The experiment shows that feature correlation dimension extracted by the method is effective to fault diagnosis of rolling bearing. The method can provide reliable feature information for condition monitoring and fault diagnosis of complicated rotating machinery, and has a broad application prospect.