焊缝超声无损检测中的缺陷智能识别方法

Intelligent defect recognition methods in the ultrasonic non-destructive test of welds

针对焊缝超声无损检测中的缺陷智能化定性识别一直未能很好解决的难题,提出采用小波包变换技术对缺陷回波信号进行降噪和特征值提取,采用基于距离的类别可分性判据和人工神经网络模式识别方法对其进行自动识别和归类,在此基础上开发了基于MATLAB软件的焊缝超声波探伤缺陷智能识别系统,可实现对缺陷回波信号的自动采集、处理及智能化识别和分类。实验表明采用的特征值提取方法以及缺陷定性识别方法是有效的。

Intelligent and qualitative defect recognition problem in the ultrasonic non-destructive tests of welds has been a difficult problem to be solved properly. Wavelet packet transform technology is used to perform the noise reduction and extract characteristic value of defect echo signals. Separability criterion based on the distance and pattern recognition method for artificial neural network are then applied to perform automatic recognition and classification for them. After that, the intelligent defect recognition system for ultrasonic flaw detection of welds based on the MATLAB software is developed. It can achieve automatic collection, processing as well as intelligent recognition and classification of defect echo signals. The testing results show that both the characteristic value extraction method and the qualitative defect recognition method are effective.