光学 精密工程, 2017, 25 (5): 1135, 网络出版: 2017-06-30
焊接缺陷磁光成像动态检测与识别
Dynamic detection and recognition of welded defects based on magneto-optical imaging
动态磁光成像 焊接缺陷 交变磁场 模式识别 dynamic magneto-optical imaging welded defect alternating magnetic field pattern recognition
摘要
为了实现焊接缺陷的自动检测, 研究一种交变磁场激励下焊缝表面及亚表面缺陷的磁光成像动态无损检测方法。分析了基于法拉第磁致旋光效应的焊接缺陷磁光成像机理, 并结合交变磁场原理推导出励磁变化与动态磁光成像的关系。探索低碳钢板的亚表面焊缝磁光成像特征试验, 验证了所提方法可用于检测焊缝亚表面的未熔合缺陷。最后对高强钢焊缝特征的动态磁光图像进行分析, 采用主成分分析法和支持向量机(PCA-SVM)模式识别方法建立了焊接缺陷分类模型。试验结果表明, 所提方法可以识别高强钢焊件中的焊缝特征(未熔透、裂纹、凹坑和无缺陷), 缺陷分类模型的整体识别率达到92.6%, 能够实现焊缝表面及亚表面缺陷的自动检测。
Abstract
To realize automatic inspection of welded defects, a dynamic magneto-optical imaging non-destructive detection of weld surface and subsurface defects under alternating magnetic field excitation was researched. The welded defect magneto-optical imaging mechanism based on Faraday magneto optical effect was analyzed and employed to derive the relationship between excitation variation and dynamic magneto-optical imaging by combining with alternating magnetic field principle. The subsurface weld magneto-optical imaging feature test of low-carbon steel was investigated, verifying that the proposed method could be used to detect incomplete penetration defects of weld surface. Finally, dynamic magneto-optical image of high-strength steel weld feature was analyzed and weld defect classification model was constructed through Principal Component Analysis and Support Vector Machine (PCA-SVM) mode recognition method. The result shows that the proposed method can recognize weld features (penetration, crack, sag and perfectness) in high-strength steel weldment with the entire recognition rate of defect classification model reaches to 92.6%, subsequently the automatic inspection of weld surface and subsurface defects can be realized.
高向东, 蓝重洲, 陈子琴, 游德勇, 李国华. 焊接缺陷磁光成像动态检测与识别[J]. 光学 精密工程, 2017, 25(5): 1135. GAO Xiang-dong, LAN Chong-zhou, CHEN Zi-qin, YOU De-yong, LI Guo-hua. Dynamic detection and recognition of welded defects based on magneto-optical imaging[J]. Optics and Precision Engineering, 2017, 25(5): 1135.