基于小波包-KPCA特征提取的三种人工焊缝缺陷检测方法

Three Defect Detection Methods of Artificial Weld Based on Wavelet Packet-KPCA Fea⁃ture Extraction

为了减少超声波检测中人为造成的缺陷误判,实现焊缝缺陷的定量评价,采用16MnR焊接试件预制了表面裂纹、气孔和夹杂等三种人工缺陷,进行焊缝缺陷检测试验。利用超声相控阵对其进行了A扫,采用小波包变换的方式对信号进行三层分解处理,通过Matlab构造特征向量能量比例,并采用KPCA进行数据降维,选取了累积贡献率超过90%的前3个主元成分,结合GRNN实现不同缺陷类型的自动分类。研究结果表明,使用小波包-KPCA进行特征提取后,可以进一步去除噪声对焊缝缺陷检测的影响,降低计算时间,准确率可达93.3%,优于常规特征值分析。小波包-KPCA可作为超声信号提取的新手段,为今后的无损检测评价提供理论依据和重要参考。

In order to reduce the artificial defect misjudgment in ultrasonic testing,and to realize the quantitative evaluation of weld defects,the 16 MNR welded specimen is used to prepare three kinds of artificial defects,such as surface crack,pores,and inclusions,and the ultrasonic phased array is used for A scanning.The wavelet packet transform method is used to decompose the signal in three layers,the eigenvector energy ratio is constructed by Matlab,and the data dimension is reduced by KPCA.The first three principal components with a cumulative contribution rate of more than 90%are selected,and GRNN is combined to realize automatic classification of different defect types.The study results show that the wavelet packet-KPCA feature extraction can further remove the influence of noise,reduce the calculation time,and achieve an accuracy of 93.3%,which is better than conventional eigenvalue analysis.The wavelet packet-KPCA can be used as a new method for ultrasonic signal extraction,providing theoretical basis and important reference for future nondestructive testing and evaluation.