Research On Fault Diagnosis Method Of Bearing Acoustic Emission Signal Based On HMM-SVM

As an important part of modern mechanical equipment,rolling bearing plays an important role in industrial production,but it is also one of the vulnerable parts.With the development of high-speed,automation and systematization of modern mechanical equipment,the working environment of rolling bearing becomes more complex,which further tests its ability to ensure the normal operation of mechanical equipment and even industrial production lines.If the failure of rolling bearing is not found in time,it will bring great hidden danger to the safety of mechanical work and serious production accident.Therefore,timely detection of bearing faults,especially early faults,can be carried out in advance for maintenance of mechanical equipment,to avoid downtime maintenance and unnecessary losses.In view of the traditional vibration signal is not sensitive to bearing early fault,this paper uses acoustic emission technology to obtain bearing acoustic emission signal,carries out fault diagnosis method research,and develops supporting online monitoring software.According to the characteristics of the acoustic emission signal of bearing,which is nonlinear and easy to be interfered by noise,an effective feature extraction and evaluation method is studied.This paper discusses the acoustic emission detection technology and signal processing method,and applies the wavelet transform method to reduce the noise of acoustic emission signal.Two signal decomposition algorithms,empirical mode decomposition and variational mode decomposition,are compared and analyzed,and the better one is chosen as signal decomposition.On this basis,the singular value,energy value and entropy of envelope arrangement are calculated,and the feature set is established by combining the center of gravity frequency of the original signal envelope.The compensation distance evaluation technology is used to evaluate the feature set,and the sensitive features are selected according to the size of the sensitive factors.Aiming at the bearing fault diagnosis method,the fault diagnosis model based on hmm-svm series is adopted.In this paper,the Hidden Markov Model(HMM)and Support Vector Machine(SVM)are studied in depth.The theory of the two algorithms and their advantages and disadvantages are discussed and analyzed.Combined with the advantages of HMM which is good at identifying continuous dynamic signal state and SVM which is suitable for small sample classification,a series fault diagnosis model of hmm-svm is established.In the HMM model of bearing,the Gaussian mixture model is used to describe the continuous eigenvector of bearing.The grid search method is used to optimize the penalty factor c and kernel function parameters g of SVM.HMM,SVM and HMM-SVM are used to identify different bearing states.The results show that HMM-SVM has better recognition ability.According to the requirement of online monitoring of bearing fault diagnosis,a bearing acoustic emission signal fault diagnosis software based on Lab VIEW and MATLAB platform is developed.The diagnosis software developed in this paper includes three parts: data acquisition and storage,data processing and analysis,and intelligent fault diagnosis.The simulation experiments are carried out on the mechanical fault simulation experimental platform with different rotating speed and different fault types of bearings,and the acoustic emission signals are collected,stored and analyzed by the software system developed in this paper.The experimental data show that the fault diagnosis software can identify the running state of the bearing in real time and accurately.