Wavelet Feature Extraction and Neural Network Pattern Recognition of Plywood Acoustic Emission Signals
Yunfei Liu · Modern Electronics Technique · 2011
To identify the different damage types of plywood,a feature extraction method of plywood acoustic emission signal based on time-frequency and proportion of energy is proposed by combining wavelet-packet time-frequency analysis with energy spectrum.The research indicates that dilatational wave and flexural wave are main modes of plywood matrix cracks signal with wide frequency spectrum,and the energy of signal is mainly concentrated in the first,second,third,fourth and seventh-band of the wavelet power spectrum.Delamination and fiber fracture signals of five-story plywood are mainly dominated by dilatational wave and flexural wave mode respectively,the former frequency is unitary and amplitude is higher,the latter energy mostly focus on the first,second band.Degumming signal waveform are composed of dilatational wave and flexural wave,and the flexural wave is dominant,whose signal energy focus on the first,second,third and fourth band of the wavelet power spectrum.An intelligent pattern classifier with BP neural network was used in recognition of those four kinds of AE signals,the recognition accuracy of flaws amounted to 92.6%.