【作者】 蔡广宇；

【导师】 吴昌林；

【作者基本信息】 华中科技大学 ， 机械设计及理论， 2009， 博士

【摘要】 弧焊机器人运动位姿精度是反映机器人性能的一个重要指标，也是确保机器人能正常工作、实现高精度焊接要求的保证。本文以南阳二机石油装备(集团)有限公司生产的实际需要为目的，研制了针对石油井架、箱体等大型结构件进行焊接的弧焊机器人，并对所研制的弧焊机器人从误差建模分析、协调运动轨迹精度评价、几何参数误差的标定以及焊缝轨迹跟踪等方面进行了系统研究，主要内容如下：(1)针对石油井架、箱体等大型结构件焊缝多、形式相对单一的特点，详细分析了采用机器人焊接所需实现的功能组合方案，优选了弧焊机器人本体设计方案，试制了机器人样机，构建了弧焊机器人运动学模型，采用“小位移摄动法”建立了弧焊机器人位姿误差模型。(2)针对弧焊机器人主要完成空间直线焊缝焊接工作的实际，分析了弧焊机器人协调直线运动的算法原理，在此基础上用“最小二乘拟合法”拟合空间直线焊缝，并采用实际轨迹的最小二乘拟合线与指令轨迹的偏差来评价机器人协调运动轨迹精度，确定了相应的评价指标，为弧焊机器人系统协调运动轨迹精度评价提供依据。采用双目视觉测量方法测得机器人各关节几何参数误差，对相应误差进行了补偿。(3)设计了两种焊缝跟踪系统：一种是接近传感器焊缝跟踪系统，该系统结构简单、操作方便，不受电弧烟尘和飞溅的影响，但对有凸台等附加构件的箱体则无法进行跟踪；为此，又设计了基于双目立体视觉的焊缝跟踪系统，为解决焊缝跟踪系统图像处理问题，提出了基于互信息的焊缝图像处理技术，并尝试使用GPU作为焊缝图形处理器以提高图像处理的速度和质量。(4)为加快互信息的计算，提出利用熵的最大化原则来分割图像，使得分割后的图像包含原图像更多的细节信息，从而保证分割后的二值图像与二值模板的互信息和原来的灰度图像与灰度模板的互信息具有相似的分布。对互信息匹配的算法进行了改进，改进后的算法减少了GPU临时变量需求，提高了图像匹配的速度和质量。更多还原

【Abstract】 The movement pose accuracy of arc welding robot is an important indicator to reflectthe performance of the robot, as well as the guarantee for the robot’s normal operation andmeeting the requirements on high-accuracy welding. With meeting the practical needs ofproduction of Nanyang RG Petro-machinery (Group) Co., Ltd. as the purpose, this paperdevelops an arc welding robot for welding of large structural parts like oil derrick, tankand etc., and performs systematic research on the arc welding robot developed fromaspects such as error modeling analysis, accuracy evaluation on coordinated movementtrack, calibration of geometric parameter error and welding seam tracking, etc. The majorefforts are as follow:(1) According to the features of large structural parts like oil derrick, tank andetc. that the welding seams are of large number and relatively unitary forms, this paperanalyzes in detail functional combinations need to be realized for welding by robot; andselected from many combinations the body design of Arc Welding Robot and makes aprototype of the robot,and builds a robot kinematic model and performs kinematicsimulation; and builds a Arc Welding Robot pose error model by using "smalldisplacement perturbation method", and deduces the error calculus formula for hand poseof the robot.(2) According to the fact that Arc Welding Robot mainly performs welding of spatialline welding seams, this paper analyzes the arithmetic principle of coordinated linemovement of arc welding robot, fits spatial line welding seams with "least squares fittingmethod" on its basis, and uses the deviation between the least squares fitting of the actualtrack and the instructive track to evaluate the accuracy of coordinated movement track ofthe robot; relative evaluation indicators are also determined to provide criterion for theaccuracy evaluation of coordinated movement track of the arc welding robot system,thebinocular stereo visual measurement method is used to measure the geometric parametererror of all articulations of the robot, and corresponding error are compensated. The resultshows that after error compensation, the accuracy of the movement track of the robotgreatly improves. (3) According to the practical situation of welding production, two welding seamtracking systems are designed. One is proximity sensor welding seam trackingsystem. This system is of simple construction and convenient operation and will not beinfluenced by arc flue dust and splash; however, it is not able to track for tanks withadditional components like protruding platform. So, another welding seam trackingsystem based on binocular stereo vision is designed, and in order to solve the problem ofimage processing of the welding seam tracking system, an image processing technologybased on mutual information is presented.(4) To accelerate the calculation of mutual information, this paper presents to use theprinciple pf maximization of entropy to subdivide images, making the subdivided imagescontain more details of the original image, consequently the mutual information betweenthe subdivided binary image and the binary template and the mutual information betweenthe original gray level image and the gray level template have similar distribution. Boththeory and welding test prove that the algorithm presented in this paper can effectivelyguarantee that the welding seam tracking is real-time and accurate.更多还原