%0 Journal Article
%A CHEN Yao
%A HE Cai-hou
%A LI Qiu-feng
%A LONG Sheng-rong
%A QI Tian-tian
%T A Wavelet Packet Neural Network Feature Recognition Method for Damage Acoustic Emission Signals
%D 2021
%R 10.13190/j.jbupt.2020-118
%J Journal of Beijing University of Posts and Telecommunications
%P 124-130
%V 44
%N 1
%X In the testing and evaluation of material damage, in order to identify effective acoustic emission (AE) signals among a large number of received signals, a neural network recognition method based on wavelet packet feature extraction is proposed. Firstly,the advantage of wavelet packet global decomposition is used to accurately extract the feature information from non-stationary signals,so, the corresponding feature vectors are established to characterize effective AE signals and interfering noise signals. Then,according to feature vectors and recognition output requirements,a three-layer back propagation neural network is established to analyze and identify signals,which could filter out noise signals and retain effective AE signals. Finally,400 sets of AE signals are collected in the experiment of glass fiber reinforced plastics to verify the method. The collected AE signals are identified with an accuracy of 97.5%,which can meet requirements of engineering.
%U https://journal.bupt.edu.cn/EN/10.13190/j.jbupt.2020-118