Identifying Of Structures' States Based On Wavelet Packet And Support Vector Machine

With the development of the computer technology, the testing technology and the signal analysis technology, the method of identifying structures'states by the artificial experience was displaced by the computer pattern recognition which was more effective. The pattern recognition technology has already been applied in artificial intelligence, robotics, system control, remote sensing data analysis, bio-medical engineering, military target recognition and so on. It has already had profound influence in the national economy, the national defense and the social development.This paper analyzed the technique of structures'states identifying and the application of the acoustics resonance spectroscopy examination technology. It also analyzed comparison of hammering method and acoustics resonance spectroscopy method on structures'states identifying. The tests showed acoustics resonance spectroscopy method was more efficient than hammering method in subtle structures.This paper introduced the application of the feature extraction method of acoustics resonance spectroscopy signal based on wavelet packet transforming. According to characteristics of acoustics resonance spectroscopy signal, this paper introduced two methods based on wavelet packet transforming. The first method was taking the energy information of the wavelet transform coefficients. The second method was taking the singular value decomposition of the wavelet transform coefficients. The energy information and the singular value were characteristic vectors of the structure.Finally it discussed the principle of the support vector machine and its application in classification. It also discussed the structure of BP-neural network and its theory. The results showed that the SVM method achieved good recognition if we extracted suitable eigenvector according to the ARS signals. The method of identifying of structures'states based on wavelet packet and support vector machine was effectively.