The invention discloses a locomotive
traction motor bearing degradation monitoring method. The locomotive
traction motor bearing degradation monitoring method comprises the steps thata motor bearing is measured to obtain full-
life time domain vibration
signal u(i); multiple
time domain features,
multiple frequency domain features, and multiple time-
frequency domain multi-dimensional features are extracted based on the full-
life time domain vibration
signal u(i) to form a high-dimensional
feature set, and normalization
processing is conducted; from the three aspects of normal operation period,early failure, and failure
development period, 10 features of each of three types are optimally selected for the high-dimensional
feature set, an
autoencoder network is used for de-redundancy
processing correspondingly to obtain three features ofx<1>, x<2>, and x<3>, then the
Mahalanobis distance formula is used for calculating the distance between samples to obtain the
Mahalanobis distance d<ij>between each sample, similarity coefficient alpha<ij> and similarity coefficient mean value M are calculated; an adaptive neighborhood K is initially constructed and corrected, the corrected K is subjected to LLE fusion indexconstruction, a fusion index Z is obtained, processed, and exponentially fitted, exponential fit parameters are calculated, and the final fusion index is obtained; and thefinal fusion index is used for determiningdivision threshold values of four degradation stagesof the normal operation period, the early failure period, the failure
development period, and a failure period.