Research On Fault Diagnosis Method For Rolling Bearings Based On Entropy

Rolling bearings,as the key component of supporting rotating parts and maintaining rotation accuracy in rotating machinery,are widely used in metallurgy,power,machinery,agriculture and defense departments.Fault detection and diagnosis for rolling bearings,accurately identify the fault location and fault degree,can effectively ensure the safe and stable operation of equipment,avoid production accidents and increase maintenance costs.The vibration signal of rolling bearings obtained in engineering application often have the characteristics of non-stationary,non-linear,and impact.The traditional methods for rolling bearings fault diagnosis have certain limitations.Entropy analyzes the nonlinear behavior of time series to characterize complexity and detect dynamic changes,which has excellent application potential in the field of mechanical fault diagnosis.Therefore,this paper takes the rolling bearings in rotating machinery as the research object,after summarizing the existing fault detection,weak fault diagnosis,feature extraction and identification methods,the fault diagnosis method for rolling bearings based on entropy is deeply carried out.The main research contents of this paper are as follows:(1)The dispersion pattern relative entropy is proposed to integrate dispersion pattern analysis and relative entropy,which is used to quantitatively analyze the dispersion pattern distribution difference between the current working status and the normal status for the vibration signal of rolling bearings.On this basis,in order to solve the problem that it is difficult to obtain rolling bearings fault data for establishing fault detection model,a fault detection method based on dispersion pattern relative entropy is proposed.The method determines the upper and lower control limits to realize fault detection by combining the dispersion pattern relative entropy and chebyshev inequality,the establishment of fault detection model only relies on normal data.Through experimental data analysis,the results verify the effectiveness of the proposed method in rolling bearings fault detection.(2)In order to extract the weak fault characteristic of rolling bearings to realize the fault diagnosis,a fault diagnosis method based on adaptive multiscale combined difference morphological filtering(AMCDMF)is proposed.Firstly,based on average combined difference morphological filter,and inspired by the idea of multiscale morphological analysis,a multiscale combined difference morphological transformation is defined.Secondly,in order to assign the weight coefficients of multiscale structural elements,the ratio of fluctuation dispersion entropy and envelope spectral sparsity is proposed as the evaluation standard for filtering signal optimization,the particle swarm optimization method is used to adaptively optimize the weight coefficients of each scale structural element.Finally,fault frequency of filtered signal is extracted through the envelope demodulation analysis and realizes the fault diagnosis of the rolling bearings.Through simulation and experimental data analysis,the results show that the AMCDMF method can effectively extract weak fault features,and realize the fault diagnosis of rolling bearings.(3)Focused on insufficient ability of multiscale permutation entropy(MPE)in extracting fault feature of rolling bearings,on the basis of MPE,by increasing the weight and drawing on the composite multiscale analysis theory,the composite multiscale weighted permutation entropy(CMWPE)is proposed.In order to realize the intelligent recognition of the rolling bearing fault location,a feature extraction and recognition method based on CMWPE-JMI is proposed by combining the CMWPE,joint mutual information(JMI)feature selection and extreme learning machine.Through analyzing the experimental data of rolling bearings,the results verify the effectiveness of the proposed method in fault location feature extraction and recognition.(4)Multiscale fluctuation dispersion entropy(MFDE)is proposed by combining FDE and multiscale analysis theory,which evaluates the complexity of time series on multiple scales.In order to suppress the entropy estimation stability decrease in coarse graining process of MFDE,the composite multiscale fluctuation dispersion entropy(CMFDE)is proposed.By analyzing the white noise and pink noise,CMFDE can measure the complexity of time series under different scales,and has higher entropy estimation stability than MFDE.On this basis,according to the hierarchical analysis procedure,the hierarchical composite multiscale fluctuation dispersion entropy(HCMFDE)is further proposed,and HCMFDE is used to extract multilevel and multiscale fault features of rolling bearings.In order to realize the intelligent recognition of the rolling bearing fault severity degree,a feature extraction and recognition method based on HCMFDE-m RMR is proposed by combining the HCMFDE,maximum correlation minimum redundancy(m RMR)feature selection and mahalanobis distance classifier.The experimental results show that,compared with MFDE-m RMR and CMFDE-m RMR,the HCMFDE-m RMR method has higher recognition accuracy.(5)By building an experiment platform,taking rolling bearings as the research object,the effectiveness of the proposed method is further verified by using the fault data obtained from the experiment.Firstly,the experimental scheme of rolling bearings simulation is explained in detail.Then,the experimental data are analyzed by using the proposed fault detection method and weak fault diagnosis method to validate the effectiveness.Finally,the proposed feature extraction and recognition methods are analyzed and compared through the collected experimental data of rolling bearings.