基于NDT的射线技术对焊接内部缺陷识别技术优化  

Defect recognition technology of small sample X-ray image based on deep learning

作　　者：李钢卿 周庆祥 范瑞峰[2] 陈航 甘雨露 马兰[2] 崔健[2] 王晨[2] LI Gangqing;ZHOU Qingxiang;FAN Ruifeng;CHEN Hang;GAN Yulu;MA Lan;CUI Jian;WANG Chen(CRRC Qingdao Sifang Co.,Ltd.,Qingdao 266111,Shandong China;Peking University,Beijing 100871,China)

机构地区：[1]中车青岛四方机车车辆股份有限公司,山东青岛266111 [2]北京大学,北京100871

出　　处：《粘接》2023年第1期134-139,共6页Adhesion

摘　　要：焊接工艺在工业生产制造环节应用广泛。焊接过程中因各种原因,焊件会产生不同的缺陷;但焊缝内部的缺陷在制造过程中较难发现,因此焊缝缺陷检测在工业领域是一项非常重要的工作。传统的胶片技术无法获取数字化的X射线图像,那么利用计算机X射线影像(CR)获取在高铁制造过程中关键部件的X射线图像,并通过对原始图像进行预处理和扩增,结合深度学习技术实现在无损检测领域的自动化,减少人为原因产生误差;从而降低企业在实际运行中的成本,提升整体检测效率。The welding process is used on a large scale in industrial production and manufacturing.The welding process produces different defects in the welded parts due to various reasons,but the defects inside the weld are more difficult to detect during the manufacturing process,so the detection of weld defects is a very important task in the industrial field.While traditional film technology can t obtain digital X-ray images,computerized radiography(CR)is used to obtain X-ray images of key components in the manufacturing process of high-speed rail.By pre-processing and amplifying the original images,combined with deep learning technology,automation is achieved in the field of non-destructive testing,reducing errors due to human causes,and thus reducing the cost of actual operation and improving the overall inspection efficiency.

关 键 词：CR技术 胶片技术 X射线检测 

分 类 号：TG441.7[金属学及工艺—焊接]