Research on the partial fault diagnosis of RV reducer based on feature fusion
To improve the reliability and recognition accuracy of partial fault diagnosis methods for RV reducer under the simulated real working conditions,and reduce the failure rate of industrial ro-bots(IR),a method based on spectral and time-frequency spectrum feature fusion is proposed and three load influencing factors are set up.The current signals fed back by the servo system are col-lected under different modes.Then,fast fourier transform(FFT)and wavelet packet decomposition(WPD)are performed on the current signals to obtain the spectrum and time-frequency spectrum curves.The average frequency domain,center of gravity frequency,root mean square frequency and frequency standard deviation of spectrum curve are taken as frequency domain feature values.The energy values of four frequency bands from the time-frequency spectrum curve are used as time-frequency domain feature values.Besides,a fusion feature vector is constructed.Extreme learning machine(ELM)based on Sigmoid function is adopted to identify three operating modes of RV re-ducer.The results show that compared to using only spectrum features and time-frequency spec-trum features,the accuracy of the test set data is improved by 10.95％and 5％respectively after feature fusion.The method proposed in the paper can effectively improve the accuracy of fault iden-tification and provide guidance for the development of sensorless online monitoring of IR.
NETL
NSTL
万方数据
