Research On Fault Diagnosis Of Rolling Bearing Based On Improved FSVM

The era of intelligence has quietly arrived,and the intelligence of various types of machinery and equipment is getting higher and higher.While we are enjoying the high efficiency and convenience brought to our life by the intellectualization of equipment,we are also facing various losses brought to us by the failure of equipment.Rolling bearing is a ubiquitous and very easy to malfunction device in mechanical equipment.According to statistics,the failure rate caused by rolling bearing damage is about 30%.Therefore,the accurate classification and prediction of various failure states of mechanical equipment rolling bearing has practical significance and engineering value that cannot be ignored.Based on this,this paper selected the rolling bearing as the research object,through the collection of vibration signal feature extraction,which carried out a series of research and analysis.The main research contents are as follows:(1)A rolling bearing signal analysis method based on Hilbert Vibration Decomposition(HVD)is used.Firstly,the idea of Empirical Mode Decomposition(EMD)is introduced.Then,for the shortcomings of EMD method,modal aliasing is easy to occur.The HVD method is introduced to decompose the fault signal.The HVD method decomposes the original fault signal into several components with different amplitudes,which has higher decomposition accuracy and effectively overcomes the disadvantages of EMD decomposition mode aliasing.However,in the decomposition process,the HVD method has to eliminate the boundary effect due to the truncation of the data.In order to solve this problem,this paper adopts a new continuation method of adaptive waveform matching.It can be seen from the comparison experiments that the improved HVD has better decomposition performance than HVD and EMD.Then,several components containing more feature information are selected and the corresponding sample entropy is obtained.The sample entropy is constructed into a machine-recognizable feature vector to prepare for the following work.(2)A membership algorithm based on density function is proposed to improve noise clustering.On the basis of analyzing several common membership degree solving methods that are greatly affected by noise,the concept of fuzzy noise clustering is introduced.This method considers noise as an independent class,so the influence of noise on the stability of the algorithm is greatly reduced.However,the initial clustering center is randomly initialized.Once the noise point is mis-selected as the initial clustering center,the clustering result will be seriously deviated.Therefore,this paper proposes to use density function instead of random assignment to initialize the clustering center,and the final experimental results well prove the effectiveness of this method.(3)The parameters of Fuzzy Support Vector Machines(FSVM)were optimized by Whale Optimization Algorithm(WOA).Firstly,the principle of whale algorithm is introduced.Then,the problems in the whale algorithm are discussed.For example,the random initialization of the population is easy to cause the loss of population diversity,the influence of the convergence factor and the weighting factor on the performance of the algorithm,and the algorithm is easy to fall into the local optimum.The problem is to make a corresponding optimization scheme to obtain the Improved Whale Optimization Algorithm(IMOA).Finally,the design comparison experiment is carried out to verify that the proposed optimization algorithm has better robustness.(4)A new improved whale algorithm is used to optimize the fuzzy support vector machine(IWOA-FSVM)fault classification model,combined with feature extraction work,and finally complete the fault diagnosis of rolling bearings.