Fault Diagnosis Research Of Rolling Bearing Based On The Analysis Of Vibration Signal

Rolling bearings are widely used in industrial production equipment, is an important part of a large rotating machinery equipment necessary, its running state directly related to the entire production system will be safe and stable operation. In order to avoid accidents and economic losses, therefore the rolling bearing condition monitoring and fault diagnosis has the extremely vital significance. Rolling bearing is running, the vibration signals contain the abundant running status information, by collecting the runs of rolling bearing vibration signal processing and analysis, and can complete State monitoring of rolling bearing. In this paper,the vibration signal of rolling bearing as the analysis object, the application of wavelet packet analysis, Hilbert-Huang transform and support vector machine(SVM) classification of technology to achieve the fault diagnosis of rolling bearing.Hilbert-Huang transform itself has adaptability, completeness and orthogonality, and it is suitable for processing non-linear and non-stationary vibration signals of rotating machinery. However in the vibration signals were collected in the industrial field, often contain a lot of noise, the signal-to-noise ratio of the signal itself is not high, making the Hilbert-Huang transform is very difficult to achieve accurate results. thus, this paper the Hilbert-Huang transform analysis of the original signal is improved by wavelet packet drop noise processing, the signal which the wavelet packet drop noise is decomposed by the EMD, using the kurtosis value and relationship with each other for all filtered decomposition intrinsic mode function, after screening the IMF to Hilbert envelope analysis, found in the envelope spectrum of the fault characteristic frequency, thus found the fault type.To the fault diagnosis of rolling bearing with intelligent, according to feature that the signal energy will change in the some spectrum when rolling bearing occur different fault types or damage degree. In order to complete intelligent fault diagnosis on rolling bearing, this article puts forward combine the wavelet packet energy extraction technology and the support vector machine(SVM) classification techniques, to improve the support vector machine(SVM) classification accuracy. In this paper, feature vectors which normalized are disposed by principal component analysis(PCA), make the original 8-dimensional vector feature vectors to 3-dimensional reduce the input parameters of support vector machine(SVM), which improves the accuracy and generalization ability of support vector machine(SVM) classification.In this paper, the two methods are simulated by using the MATLAB software and LIBSVM toolbox.The experimental results show that: The representative of the early fault of intrinsic mode functions can be screened by wavelet packet drop noise and kurtosis value and relations. Fault characteristic frequency can be obviously found in its package spectrum by the method of envelope analysis. However, this method cannot judge fault damage degree; By combining wavelet packet energy extraction, principal component analysis and support vector machine(SVM) technology, not only can realize the different categories of fault classification, also can realize the recognition of different damage degree.