The Fault Diagnosis Of Machinery Based On Fractal And Support Vector Machines

SVM (Support Vector Machines), which is based on Statistical Learning Theory, provides a uniform method for learning subject of limited samples. This algorithm accomplishes the structural risk minimization principle commendably, and can change the problem in non-linearity space to that in the linearity space in order to reduce the algorithm complexity by using the kernel function idea. SVM have become the hotspot of machine learning because of their excellent learning performance. They also have successful applications in many fields, especially in fault diagnosis field. Fault diagnosis is a limited samples subject because the fault sample gained hardly in production, while the most predominance of SVM is proper for limited samples decision, so the foreground of SVM in fault diagnosis field is very bright. More over, the method of feature extraction on fault sample was also the important component element. Fractal approach can quantized describe the subtle change of complicated mechanical system by characteristic parameter------fractal dimension, it will be helpful to identify the mechanical equipment's state by dint of fractal dimension.Based on the research situation of SVM and fractal, this paper's main work carried out was as followed: take roll bearing as subject of research, analyzed its fault behavior and collected the type fault signal of different condition on the bearing experiment table by the way of fault simulation at first, then slaked the trend term of vibrate signal and used the way of wavelet analysis to denoise the signal . After that, the method of fractal was used for abstracting the fault character of vibration signal which had been preprocessed. Finally, we put the feature vector which got from fractal way as sample into SVM, and made use of the regression algorithm of SVM for fault diagnosis of roll bearing. In addition, the problem of parameter selection of SVM was discussed, a method of parameter selection was proposed. Research and experiment indicated that: it was feasible that combined method of fractal with SVM on feeble fault diagnosis.