Research On SVR Rolling Bearing Performance Decline Prediction Based On Krill Population Algorithm

Rolling as one of the key components in large rotating machinery,to ensure the normal operation of machinery and equipment plays a decisive role.It is very important to predict maintenance in advance.Therefore,how to establish the correct bearing degradation evaluation index is related to the accuracy of the prediction results.If it can take appropriate measures to deal with the impending bearing failure to avoid the occurrence of accidents in advance plays a vital role in ensuring the normal operation of equipment.In this thesis,based on the existing vibration signal feature extraction method of rolling bearing,In view of the shortcomings of the existing methods,a feature extraction method combining CEEMD and wavelet packet semi-soft threshold is proposed.This method is different from the traditional time domain,frequency domain and time-frequency domain characteristics.Based on the integrity of the original signal,the noise in the high frequency vibration signal is filtered out.The experimental results show that the proposed method is more effective than other time-frequency methods in feature extraction.Aiming at the above improved feature extraction method,this thesis reduces the dimensionality of the high-dimensional feature set based on the multiple feature parameters,in order to solve the shortcomings of PCA and KPCA,a method combining LLE with fuzzy C-means is proposed.After clustering by LLE and then clustering by fuzzy C-means,the clustering effect of different methods was compared by experiment.It can effectively classify the degradation degree of inner ring of rolling bearing.According to the deficiency of the traditional SVR,a multivariate support vector regression prediction method based on krill population algorithm isproposed.By using the krill feeding principle,the optimal parameters in the SVR are selected,and the genetic algorithm and krill population algorithm are compared.The degradation degree of the inner ring of rolling bearing is predicted,which is very important for the health evaluation of the rolling bearing.Finally,the multi-feature extraction of the rolling bearing life cycle test data of the University of Cincinnati is used to classify the degenerated phases.The forecasting trend and precision of the rolling bearing are evaluated more effectively and accurately by three different prediction input methods.The results show that the proposed method has high accuracy and more comprehensive information,which is significant for the research of rolling bearing performance degradation prediction.