Extraction methods for numerical and graphical features of long course eccentric fault of rotating machinery

The eccentric fault is one of the typical faults of rotating machinery and it is important to accurately extract the numerical features and graphical features of eccentric fault in fault study. For this reason, some extraction methods were presented, including the fluctuation stability of sampling points was taken to describe the degree of eccentric fault; the information fusion, wavelet threshold de-noise, data compression, constructing features matrix, calculating the variance and so on were used to deal with the long course of eccentric faults data to describe the degree of eccentric fault; the low-frequency time-domain signal obtained by wavelet decomposition and two-dimensional axis trajectory were employed to explore the essence change in the fault deterioration process. The numerical features and graphical features of long course eccentric fault obtained with the three methods were simulated by a rotor experiment platform. Results show that the numerical features and graphical features change unidirectionally with fault deterioration. Therefore, the three methods can accurately track on the process and degree of eccentric fault, and all of them have a high stability robustness.