基于声发射和神经网络的复合材料冲击定位

Impact localization for composite based on acoustic emission and neural networks

为了提高复合材料结构冲击定位的精度和实时性,基于声发射和神经网络技术,提出了复合材料结构冲击定位两步法,以压电陶瓷(PZT)和自制信号采集系统替代商用声发射仪器,实现了一种能够高精度、实时、在线监测冲击的系统。用小波变换求出原点处冲击源传播到各传感器的波达时间差,用这组时间差修正其他位置上的冲击源到达各传感器的波达时间,利用修正后的波达时间,根据四点圆弧定位算法得到冲击源坐标,实现初步定位;将所求出的位置坐标作为神经网络的输入,训练之后的神经网络可以准确预测冲击位置,实现精确定位。在复合材料板上的试验表明:该方法能快速、准确地识别出冲击位置。

In order to improve accuracy and real-time of impact localization for composite, two-step method based on acoustic emission and neural networks is proposed and a real-time and on-line impact monitoring system for composite is developed by using piezoelectric transducers （PZT）. The self-made signal collection system, instead of commercial acoustic emission instrument. Wavelet transformation is used to estimate time-delay of the impact source in the origin propagateing to several transducers. Arrival-time of other impact sources between several transducers is corrected by the time-delay. Combining the corrected arrival-time and four-point arc localization, impact source coordinates are calculated. The calculated coordinates are used as input of neural networks, the trained neural networks can determine accurately position of impact source. Experiments on laminated composite indicate that the present methodology may identify impact location fast and accurately.