使用Levenberg-Marquardt训练函数的神经网络算法在复合材料板上进行声发射定位,采用声波到达时间作为输入向量,训练得到了到达各个传感器的声波触发时间与板上坐标的良好的映射关系。考虑到波形衰减,用传统的固定门槛提取的到达时间容易与实际情况偏差较大,讨论了一种相对准确的适合工程使用的到达时间提取方法。 Using the neural network algorithm of Levenberg-Marquardt training function to locate the acoustic emission on the composite board, using the arrival time of the acoustic wave as the input vector, the good mapping relationship between the acoustic trigger time and the on-board coordinates is achieved. Taking into account the waveform attenuation, the arrival time of traditional fixed thresholds is easily deviated from the actual situation, and a relatively accurate method of arrival time extraction suitable for engineering use is discussed.