摘要
超薄板焊接对接头处的任一细微变化,均要求实时调节焊接热输入和焊枪对中,以保证焊接质量。因此,对超薄板对接接头微尺度信息的检测和识别尤为重要。本研究基于超薄板焊接对接接头的视觉传感检测,以焊接接头图像的像素点RGB值为试验对象,通过K-means聚类模型进行深度学习,训练后的二值化图像采用Canny算子提取轮廓,实现了100μm超薄板焊接对接接头不同微尺度信息的检测和识别,并计算出了焊接接头间隙宽度的细微变化和接头中心位置的偏移程度。试验结果表明,采用对像素点RGB值进行K-means聚类的方法,能更有效地分割焊接接头图像。进行一定次数的机器学习后,可以精确识别100μm厚对接接头的细微变化,通过Canny算子可以可靠提取由机器学习训练出的焊接接头二值化图像的细小毛刺、间隙直线度和间隙尺寸的细微变化,计算出的接头间隙宽度和中心位置偏移量,可为精确实时调节焊接热输入和焊枪对中提供有力支撑。
Heat input and torch alignment must be adjusted in real time with any slight change in the welded joint of ultrathin plates for sound weld quality.It is very important to detect and identify micro-scale information about welded joints of ultra-thin plates.Based on visually sensing detection,a deep learning method is proposed in this paper.The proposed method incorporates K-means clustering on the pixel RGB value of acquired joint images and Canny operator treatment on the trained binary images to extract joint contour to make sure different micro-scale information about welded butt joints of 100 μm ultra-thin plates can be detected and identified.In addition,tiny changes in joint gap width and center position offset are calculated.The results show that welded joint images could be segmented efficiently using K-means clustering on RGB values.The subtle changes in 100 μm thick welded butt joints could be precisely identified.Tiny edge burrs and slight variations in joint gap straightness and gap width in binary images trained with machine learning are reliably extracted using Canny operator.The joint gap width change and center position offsets could be calculated,providing a strong support for precisely real-time adjustment of heat input and torch alignment.
作者
潘雪航
何建萍
PAN Xuehang;HE Jianping(School of Materials Science and Engineering,Shanghai University of Engineering Science,Shanghai 201620,china)
出处
《智能计算机与应用》
2023年第10期103-107,共5页
Intelligent Computer and Applications
基金
国家自然科学基金(51775327)
上海市科委基础研究重点项目(14JC1402700)。
关键词
超薄板
焊接接头
K-MEANS聚类
轮廓提取
微尺度信息
ultrathin plates
welded joint
K-means clustering
contour extraction
micro scale information
作者简介
潘雪航(1997-),女,硕士研究生,主要研究方向:微束等离子弧焊、自动控制、设备通讯等;通讯作者:何建萍(1964-),女,博士,教授,主要研究方向:精密焊接过程的机电光焊一体化复杂系统的智能化控制。Email:janejphe@sues.edu.cn。
