紧密对接激光焊焊缝位置神经网络预测  

Prediction of weld position by neural network during micro-gap butt joint laser welding

作　　者：莫玲[1] 高向东[1] 萧振林 陈晓辉 

机构地区：[1]广东工业大学机电工程学院,广东广州510006 [2]广州番禺高勋染整设备制造有限公司,广东番禺511400

出　　处：《电焊机》2016年第3期75-78,共4页Electric Welding Machine

基　　金：国家自然科学基金资助项目(51175095);广东省自然科学基金资助(10251009001000001);广东省学科建设科技创新资助项目(2013KJCX0063)

摘　　要：激光焊接过程中,控制激光束准确对中焊缝是获得良好焊件的关键。以低碳钢板紧密对接激光焊(焊缝间隙不大于0.1 mm)作为研究对象,利用磁光传感法摄取焊接过程中焊缝区域磁光图像。分析焊缝区域图像特征,定义并提取紧密对接焊缝位置坐标,以前时刻的焊缝位置及其变化值作为输入量,当前时刻焊缝位置坐标作为输出量,应用神经网络建立焊缝位置的预测模型。试验结果表明,建立的前馈型神经网络能够较好地预测焊缝位置坐标,为激光焊缝及时纠偏和自动跟踪奠定基础。It is critical to control the laser beam focus position accurately aligning the weld seam to obtain good welding quality in the laser welding process.The micro-gap butt joint（weld gap is less than 0.1 mm） laser welding of low-carbon steel was chosen as the research object,and a magneto optical sensor was used to capture the weld region images during welding.The feature of weld images was analyzed and the coordinates of micro-gap joint weld position was defined and extracted.A weld position prediction model was established using the neural network whose inputs were the weld position coordinate value and the coordinate changing value at previous sampling time,and output was the weld position coordinate value at current sampling time.Experimental results indicate that the proposed feed forward neural network can predict the weld position accurately,which provides a foundation for real time correcting weld deviations and automatic seam tracking.

关 键 词：激光焊 焊缝位置 神经网络 预测 

分 类 号：TG456.7[金属学及工艺—焊接]