基于多信息融合理论的MIG焊机器人焊接质量控制  

MIG Welding Robot Welding Quality Control Based on Multi-information Fusion

作　　者：岳建锋[1] 张翠宣[2] 李亮玉[1] 

机构地区：[1]天津工业大学现代机电装备技术天津市重点实验室,天津300160 [2]天津渤海职业技术学院,天津300402

出　　处：《中国机械工程》2009年第9期1112-1115,共4页China Mechanical Engineering

基　　金：天津市自然科学基金重点资助项目(05YFJZ02100);高等学校博士学科点专项科研基金资助项目(200800580005)

摘　　要：在熔化极气体保护焊(metal inert-gas,MIG)焊接质量的自动控制过程中,采用多传感器信息融合理论将正面熔池宽度信息和温度信息进行一致性检验,实现了正面熔宽有效采集。通过主成分分析和经验公式确定了焊接质量控制BP神经网络模型的输入参量和隐含层节点数。对常规BP神经网络进行改进,提高了其收敛速度。在机器人MIG焊接中利用该模型实施焊接试验,获得了较好的焊接效果,这为利用焊接正面传感信息对焊接质量闭环控制提供了新的解决方案。During MIG welding quality auto--control, top molten pool width and temperature in- formation were tested on consistency by multi--sensor information fusion theory and realized top mol ten pool width gather effectively. Input parameters and hidden nodes of BP neural network model on welding quality were determined by the principal component analysis and experimental formula. Com- mon BP neural network was developed and increased convergence speed. Welding experiment was im- plemented by the modified model during robot MIG welding and gained good welding results which provide a new scheme for welding penetration closed-loop control by using welding work piece top acquiring information.

关 键 词：熔化极气体保护焊 多信息融合 BP神经网络 焊接质量 

分 类 号：TG439.4[金属学及工艺—焊接]