Fault Diagnosis for Robot RV Reducer Using Nonlinear Frequency Spectrum and Kernel Principal Component Analysis

Aiming at the low accuracy of fault diagnosis of RV reducer for robot, a method based on the combination of nonlinear output frequency response (NOFRF) spectrum and kernel principal component analysis (KPCA) is adopted to diagnose the fault of RV reducer. The input and output data of RV reducer in normal state and fault state are collected by the performance testing equipment. Then, the first four-order NOFRF spectrum values, which are sent to KPCA as system fault feature for compression are estimated with batch estimation algorithm. Setting the cumulative contribution rate of principal components, 400-dimension data are compressed to 5-dimension ones. The low-dimension data generated by KPCA are sent to support vector machine classifier for training. Experimental results show that compared with the method only using time domain or frequency domain of vibration signal as data sets, the fault recognition rate of proposed method achieved 96.67% with 27.50% and 8.34% increasement, respectively, the effectiveness of the proposed method in fault diagnosis of RV reducer is thus verified. © 2020, Editorial Office of Journal of Xi'an Jiaotong University. All right reserved.