| The breakthroughs in artificial intelligence,the Internet of Things,and big data are transforming traditional manufacturing into smart manufacturing.As an important part of industrial robots,robot servo motors play a vital role in intelligent manufacturing.According to statistics,motor bearing faults account for the largest proportion of motor faults.Once the motor bearing fails,it will affect the normal operation of other components of the motor,which may cause the entire servo motor to malfunction or even affect the entire industrial robot using the servo motor.In order to avoid the fatal failure of the robot servo motor,research on performance degradation assessment and remaining life prediction of the motor bearing should be carried out.This article takes the robot servo motor bearing(hereinafter referred to as bearing)as the research object.The main research content has the following three aspects:Aiming at the difficulty of extracting the most sensitive features to bearing faults and the difficulty of effective feature compression in the construction of bearing degradation feature space,a construction method of bearing degradation feature space based on stacked sparse autoencoder is proposed.The algorithm first uses the time-domain and time-frequency domain characteristics of the bearing vibration signal as the original feature set,and then uses SSAE to perform deep feature extraction and feature compression on the original feature set to construct the final degenerated feature space.The feature vector is a more abstract representation of the original feature.Finally,an experimental verification was carried out on the bearing data set.The experimental results show that the feature space constructed by the algorithm can comprehensively and effectively represent the degradation process of the bearing,and can be used in the subsequent evaluation of bearing performance degradation and remaining life prediction.Aiming at the problem that traditional SVM algorithm parameters are often set based on experience,it is difficult to establish an optimal performance degradation evaluation model to accurately detect early bearing faults.A bearing performance degradation evaluation method based on multi-objective and acoustic search optimization SVM is proposed.The algorithm first uses SSAE to extract the degradation feature vector of the bearing vibration signal,and then uses the trend and monotonicity of the bearing degradation index curve as the fitness function,and uses the MOHS algorithm to optimize the SVM penalty parameters and Gaussian kernel function parameters to establish the optimal Performance degradation evaluation model to get the performance degradation index of the bearing.Finally,an experimental verification was carried out on the bearing data set.The experimental results show that the algorithm can accurately detect the early failure of the bearing,and the degradation index curve obtained has good trend and monotonicity.Aiming at the problem that the existing bearing life prediction model often considers each time point independently and discards a large amount of useful information in the historical data,which leads to a large prediction error,a bearing remaining life prediction based on a deep gated recurrent unit network is proposed.First,SSAE is used to extract the degraded feature vector of the bearing vibration signal,and then the remaining service life prediction model is established using the DGRU network,which directly maps the bearing feature vector to the remaining service life of the bearing.Finally,an experimental verification was carried out on the bearing data set.The experimental results show that the algorithm can effectively predict the remaining service life of the bearing,which has higher accuracy than the traditional bearing life prediction algorithm. |
