ART-2网络在超声检测信号分析中的应用

Application of ART-2 network in the signal analysis of ultrasonic testing

本文以焊缝超声检测常见的缺陷(夹渣、气孔、未焊透、未熔合)为对象,选取缺陷时域波形的上升时间、波宽、上升角度和频域特征的波动次数、底宽等特征值组成特征矢量作为ART-2的输入,来实现缺陷性质的识别,并与射线检测结果对比。实验结果表明：采用该方法得到的正确识别率达到81．5％,具有良好的应用前景。

Based on the four types of weld flaws in Ultrasonic testing, this paper presents the ART-2 method to recognize the defect pattern .Five characteristics in the time domain and frequency domain are chosen as the input vector for the ART-2 network.The results were compared with those of X-Ray testing. Experimental results show that the recognition rate can reach 81.5% due to the method and it has a future for applications.