| The producing process of petroleum drilling is a continuous process, if it goes wrong, the economic loss will be very serious. The drilling is a large-scale equipment with many transmission mechanisms, and its rolling bearings easily break down. Thus, developing the fault diagnosis of drilling rolling bearings has great practical significance.In this paper, the fault mechanism and the vibration characteristics of roller bearings are summarized, and the traditional vibration signal analysis methods are introduced. However, excellent effects can not be obtained by using regular Fourior to analyze the non-stationary vibration signal produced by the fault rolling bearings. The wavelet analysis method, which can analyze signal both in time and frequency domains, is suitable to the fault diagnosis of rolling bearings. After introducing the theory of wavelet analysis, this paper uses the principle of wavelet denoise to preprocess the vibration signal. The noise disturbance of the vibration signal is effectively eliminated and the signal-to-noise ratio is improved. Then the wavelet packet analysis is used to extract energy eigenvector of frequency domain. The results show that the real fault features of the vibration signal could be extracted very well by this method.On the basis of feature extraction, this paper makes some discusses on RBF neural network(RBFNN). It is practicable that RBFNN is applied to diagnosis of rolling bearings through the network analysis results. Aiming at the problem that it is difficult to obtain enough samples to train neural networks due to the limitation of data transmission bandwidth and memory, causing shortage of neural network train and inaccurate diagnosis results in practical application, support vector machine (SVM) is proposed in this paper and a multifault SVM classifier based on classification is developed. The results of experiment indicate that this method is effective and has higher diagnosis precision than RBFNN in case of fewer samples.
|
PDF Full Text Request