粘接件拉伸强度的超声检测实验研究

Experimental research on tensile strength of bonding structure by ultrasonics

利用RAM-5000 SNAP超声测试系统,在水池中对穿透环氧树脂粘接件的超声信号的基频波和二次谐波进行测量,二者振幅的重复测量误差分别在1%和3%以下,实验具有较好的可重复性。利用反向传播人工神经网络算法,将穿透样品的超声波的基频波振幅、二次谐波振幅、粘接前样品表面的粗糙度作为输入参量,粘接件的拉伸强度作为输出参量进行训练,并对另一组样本的强度进行了预测,与实际破坏强度相比,平均算术误差为8.3%,这为利用超声方法预测粘接件的拉伸强度提供了一个可能有效的实验方法。

An ultrasonic test with RAM-5000 SNAP is performed in water to measure the transmission signals（including the fundamental and second harmonic）of epoxy bonded structure,the repeatability errors are below 1%and 3%,respectively.A back propagation neural network is used to predict the tensile strength with the surface roughness as well as amplitudes of the fundamental and second harmonic being input parameters;the arithmetical mean error is 8.3%.This provides a possible way to predict the tensile strength of bonding structure.