摘要
熔透信息的实时获取是实现打底焊接自动化的关键环节之一,通过熔池形状特征预测熔透状态可为熔透信息的有效提取提供参考。鉴于焊工通过视觉观察熔池正面形状特征来对熔透状态进行实时判断,建立基于GMAW的双目视觉传感焊接试验系统。在不同焊接电流及焊接速度下开展打底焊接试验,在焊接过程中实时同步采集熔池正面与背面图像。基于熔池图像特点结合成熟的图像处理算法,提取熔池正面二维与三维形状特征参数以及背面熔宽信息,作为训练样本,以熔池正面形状特征参数作为输入量,以背面熔宽作为输出量。通过BP算法对神经网络进行训练,建立熔透状态预测模型,分析熔池正面形状特征参数与背面熔宽之间的映射关系。采用Garson算法计算出每个形状特征参数对于背面熔宽的权重系数。通过GMAW打底焊试验对熔透状态预测模型进行了验证,试验结果表明,建立的BP神经网络模型可以有效地预测焊缝的熔透状态。
Real-time acquisition of the penetration information is one of the key links in the automation of backing welding. Predicting the penetration state through the shape feature parameters of weld pool can provide a reference for the effective extraction of penetration information. Since the welder estimates the penetration state in real time by observing the shape features of weld pool, a GMAW test system with binocular vision sensors is established. Backing welding tests are carried out under different welding currents and welding speeds, and the front and back images of weld pool are collected synchronously during the welding process. Based on the characteristics of weld pool images, combined with the mature image processing algorithms, the two-dimensional and three-dimensional shape feature parameters of weld pool surface and the back-bead width information are extracted, which are taken as the training samples. The feature parameters of weld pool surface are taken as the input and the back-bead width is taken as the output. The BP algorithm is used to train the neural network, and the penetration state prediction model is set up to analyse the mapping relationship between the shape feature parameters of weld pool surface and the back-bead width. The weight coefficient of each shape feature parameter to the back-bead width is calculated by the Garson algorithm. The penetration state prediction model is verified by the backing GMAW tests. The test results show that the BP neural network model can predict the penetration state of the weld effectively.
作者
黄军芬
薛龙
黄继强
邹勇
马可
焦向东
HUANG Junfen;XUE Long;HUANG Jiqiang;ZOU Yong;MA Ke;JIAO Xiangdong(College of Mechanical Engineering, Beijing Institute of Petrochemical Technology, Beijing 102617)
出处
《机械工程学报》
EI
CAS
CSCD
北大核心
2019年第17期41-47,共7页
Journal of Mechanical Engineering
基金
国家自然科学基金(51505035)
北京市自然科学基金-北京市教委联合项目(KZ201810017022)资助项目