Glass fiber-reinforced plastics (GFRP) is widely used in many industrial fields. When acoustic emission (AE) technology is applied for dynamic monitoring, the interfering signals often affect the damage evaluation results, which significantly influences industrial production safety. In this work, an effective intelligent recognition method for AE signals from the GFRP damage is proposed. Firstly, the wavelet packet analysis method is used to study the characteristic difference in frequency domain between the interfering and AE signals, which can be characterized by feature vector. Then, the model of back-propagation neural network (BPNN) is constructed. The number of nodes in the input layer is determined according to the feature vector, and the feature vectors from different types of signals are input into the BPNN for training. Finally, the wavelet packet feature vectors of the signals collected from the experiment are input into the trained BPNN for intelligent recognition. The accuracy rate of the proposed method reaches to 97.5%, which implies that the proposed method can be used for dynamic and accurate monitoring of GFRP structures.

玻璃纤维增​​强塑料（GFRP）广泛用于许多工业领域。当声发射（AE）技术应用于动态监测时，干扰信号常常会影响损伤评估结果，从而严重影响工业生产安全。在这项工作中，提出了一种有效的针对GFRP损伤的声发射信号的智能识别方法。首先，利用小波包分析方法研究了干扰信号和声发射信号在频域的特征差异，该特征可以通过特征矢量来表征。然后，构建了BP神经网络模型。根据特征向量确定输入层中的节点数，并将来自不同类型信号的特征向量输入到BPNN中进行训练。最后，实验收集的信号的小波包特征向量被输入到训练后的BPNN中进行智能识别。所提方法的准确率达到97.5％，表明所提方法可用于动态，准确地监测GFRP结构。