Research On Health Condition Assessment Method Of A Rolling Bearing Based On Adaptive Increment LLE-SVM Model

Rolling bearing is regarded as one of the most important vulnerable parts of the rotating machinery,and its health condition directly affects the safety and the reliability of the operation of the integrated mechanical equipment.Once the breaking down of a bearing can not be monitored and eliminated in time,which may result in the failure of the entire mechanical equipment,even lead to huge disastrous consequences.Therefore,it is very important to accurately assess the health conditions of a rolling bearing and realize intelligent quantitative assessment,convert scheduled or unscheduled maintenance to condition maintenance,and finally active maintenance of a rolling bearing can be achieved.Establishing health conditions assessment method of the rolling bearing is regarded as the research center line,and the research is expanded from several aspects of feature extraction,dimension reduction,health condition assessment model and optimization algorithm.For feature extraction,in order to efficiently use various features with significant category differences information of time domain,frequency domain and time-frequency domain,and characterize the health condition of rolling bearing more completely and accurately.The time-domain and frequency-domain features of the actual rolling bearing vibration signal are extracted,and the time-frequency features are obtained by using ensemble empirical mode decomposition(EEMD)combined with singular value decomposition(SVD).On this basis,high dimensional multi-domain feature sets are constructed.For dimension reduction,locally linear interpolation(LLE)algorithm is studied deeply.The features of the vibration signal have the characteristics of thenon-uniform distribution of data manifold local geometric structure,at the same time,in order to avoid the shortcoming of the repeated calculation when added new samples,an adaptive incremental improvement of locally linear embedding algorithm is improved.The effectiveness of the improved locally linear embedding algorithm is proved through the comparison experiments.For optimization algorithm,due to the blindness of artificially selecting the parameters of support vector machine(SVM),and the global optimal solution is can not be found quickly using the traditional parameter optimization algorithms.Therefore,the chicken swarm optimization(CSO)algorithm is studied deeply,and the chaos theory is introduced into the chicken swarm optimization algorithm,which is named as chaos CSO(C-CSO)algorithm.The performance of each optimization algorithm is analyzed in detail through basic function optimization comparison experiments.For health condition assessment,the parameters of SVM are optimized by chaos CSO algorithm,and the adaptive incremental locally linear embedding-support vector machine classification model is constructed.And a health condition assessment index is proposed to achieve effective assessment of the rolling bearing health condition based on consistency relative compensation distance.