Orthogonal Wavelet Transform SVDD In Fault Diagnosis

In the field of fault diagnosis, the traditional fault diagnosis methods are aimed at a smooth signal changes based on the Fourier-based signal feature extraction, and then troubleshooting experience or theory to determine the fault type. In addition the lack of training samples of the problems often are unable to reach the expected diagnostic purposes.This paper presents an orthogonal wavelet transform-support vector data description method (OWTSVDD), this method can fail under the non-stationary signal samples of a good classification, and address a limited number of samples under the conditions of the problem of fault diagnosis. Orthogonal wavelet transform is the choice of wavelet functions orthogonal wavelet transform, so that in the time domain and frequency domain characterization of signals have the capacity of local features. Support Vector Data Description is a single classification methods, can only under the condition of the normal sample fault signal classification. Orthogonal wavelet transform and support vector description of the method of combining the use of orthogonal wavelet transform to extract the details of the signal peak-peak as SVDD input parameters, the equipment can quickly identify the operation of the state. In this paper, research orthogonal wavelet transform-support vector data description based on the theory, focused on the following aspects:1. Statistical learning theory and support vector data description theory and algorithms, support vector data description in the introduction of nuclear function, results show a different feature extraction and selection of different kernel functions of the SVDD classification impact.2. The basic theory of orthogonal wavelet transform, orthogonal wavelet transform choice of different wavelet feature extraction of the signal and, through case analysis, orthogonal wavelet transform in the letter-noise separation, singular point positioning to eliminate and bearing fault diagnosis have achieved the desired effect.3. For the orthogonal wavelet transform and support vector description of the characteristics, this paper presents a wavelet transform based on orthogonal-support vector data description of the fault classification method to orthogonal wavelet decomposition of the signal peak after peak levels constitute a feature vector with the normal state of the data sample to establish the knowledge base to classify the test samples, and without a line of feature extraction SVDD compared to computing efficiency and diagnostic efficiency greatly improved.4. Orthogonal wavelet transform-support vector data description in the bearing performance degradation study, the right only under fault conditions pitting the bearing fault state conducted an assessment of pitting failure, with the details of orthogonal wavelet transform maps the trend line shows that the method Failure assessment of effectiveness.