【作者】 沈鸿源；

【导师】 陈善本；

【作者基本信息】 上海交通大学 ， 材料加工工程， 2008， 博士

【摘要】 目前服役的焊接机器人90%都是以“示教-再现”模式进行工作的,少数以轨迹规划方式工作,焊接过程中,焊枪与焊缝中心都会存在一定误差,并且,焊接过程又是一个复杂、非线性、多边化的过程,焊接工件热变形、咬边、错边,以及焊缝间隙的变化等是不可预知的,这些因素都会直接影响到焊接质量。因此,在“示教-再现”或轨迹规划应用的基础上,实时焊缝纠偏可以进一步提高跟踪精度,尤其适用于辅助工程上焊接易变形、装配复杂等自动焊难以控制的工件生产;同时,实时检测焊缝间隙和焊接熔池宽度的变化,有助于调整焊接规范,保证焊缝成形的均匀性,特别是在多道焊过程中,第一道打底焊的焊缝均匀程度直接影响盖面焊接接头的气孔缺陷的多少。因此,在本篇论文研究的基于视觉的实时焊缝跟踪和焊缝成形控制技术不仅仅在于提高弧焊机器人智能化程度,同时也具有促进弧焊机器人在实际焊接生产中应用的现实意义。本文以航天运载火箭推进剂贮箱生产为背景,针对方波交流GTAW(钨极惰性气体保护焊)焊接方法,带有焊接衬垫的焊接结构等生产中常用的焊接条件,将传统的“示教-再现”型机器人开发成具有实时焊缝跟踪和焊缝成形控制的弧焊机器人系统。本文先构建了焊接机器人系统平台,包括:1.设计了具有双层滤波功能的视觉传感器,可以分别为观察电区弧柱域和焊缝区域选择各自的滤光片和减光片,能够清晰采集到不同焊接电流值的焊接图像,解决了焊接电流变化对焊接图像质量影响较大的问题;2.利用PC机、A/D卡、D/A卡、图像卡实现PC机实时采集焊接图像,并可以实时纠偏机器人末关节,调整焊接工艺等功能;3.在Visual C++环境下,采用多线程技术,同时采集图像、处理图像、处理数据、控制算法运算、发送焊接纠偏控制电压和焊接工艺调整电压等功能,一个控制周期可以达到0.2s。本文研究了6mm厚LD10铝合金方波交流GTAW焊接过程中的熔池图像特征,提出了“双窗口”同时采集弧柱和焊缝信息的方法,利用滤波、边缘检测、适应环境光强变化的自动域值化、骨架细化、最小二乘法拟合等方法开发出具有较高适应性的图像处理算法。该算法能够适用于不同天气状况的白天和有灯光照明的夜晚等不同环境光线下的图像处理,同时也适用于焊接电流在(200～300)A范围内(该范围适合于中厚板铝合金焊接)的图像处理,图像处理算法的精度可以控制在±(0.1~0.2)mm范围内。提出了一种免于机器人标定的实时焊缝跟踪技术,该技术利用视觉传感器从熔池正前方观察焊枪与焊缝的位置偏差,实时调整焊枪位置对中焊缝中心,方便操作者使用。该跟踪方法同时提取熔池中心和焊缝中心的位置,只需要计算出二者的相对位置就可以进行实时跟踪。分析了机器人实时调整轨迹的运动机理,在保障焊接稳定性的前提下,根据偏差量的大小不同,设计了分段式自适应PID控制器,保证机器人快速而稳定的调整焊枪位置。本文设计的较大偏差的试验工件跟踪最大偏差可以控制在±0.3mm范围之内,正常示教焊接的跟踪最大偏差可以控制在±0.2mm范围之内。结合带有焊接衬垫的焊接结构,以焊接熔池填充金属量和提供金属量守恒,分析出焊缝余高与焊缝间隙、送丝速度、熔池宽度和焊接电流等参数的关系,建立了余高预估模型,该模型平均误差为0.085mm,标准差为0.265mm,具有较高的可靠性和有效性。然后,解决了多道焊的打底焊成形控制技术:检测焊缝间隙利用模糊控制建立了前馈焊接电流模糊控制器;以焊缝间隙、熔池宽度、送丝速度等量为输入量用余高预估模型预估出当前余高大小,作为反馈量,建立了送丝速度闭环反馈控制。将本文提出的实时焊缝跟踪技术和焊缝成形技术分别在直线平板对接焊缝和法兰盘曲线焊缝试件中试验,机器人实时焊缝跟踪偏差分别可以控制在±0.3mm和±0.6mm之内,焊缝余高在(-0.5,0)mm范围内的焊缝长度分别能够占整条焊缝长度的96%和93%。最后,将本文开发的基于被动视觉的具有实时焊缝跟踪和焊缝成形控制技术的弧焊机器人系统应用于航天运载火箭推进剂贮箱箱底模拟件产品的圆环拼接纵缝和法兰与瓜瓣环缝焊接中,其焊后结果经过χ光射线检测,焊缝均能够满足YS0620-97规定的Ⅰ级焊缝标准,达到了航天产品生产上的要求。更多还原

【Abstract】 At present, 90% of the welding robots applied in manufacturing are primarily“teach and playback”robots, and few work in trajectory planning mode. However, welding encounters many variables, such as the errors of pre-machining, fitting of work-piece and in-process thermal distortions, which will change the gap size and seam position so as to affect welding quality.So real-time seam tracking technology could be used to improve tracking precision for those robots working in“teach and playback”or trajectory planning mode, especially for those work-pieces with large distortions and complex assembly. Simultaneously, weld formation control technology can keep a well weld formation to adjust the welding procedures after detecting the changes of the seam gap and the width of weld pool, especially for multi-pass welding, air holes would increase in cosmetic welding if the first pass welding has poor formation. So both seam tracking and weld formation control technology cannot only push the development of robot, but significantly benefit for the production.In this paper, the research background is spaceflight manufacturing. The study is based on square-wave AC (alternating current) GTAW (gas tungsten arc welding) with backing bar. The“teach and playback”robot was rebuilt to be a welding robot system with seam tracking and weld formation control.Firstly, the arc welding robot system was constructed, as following,1. The visual sensor with double-layer filter was designed to detect the arc stream in front the weld pool. The computer can capture the clear welding image in different levels of welding current with the sensor device.2. The computer can capture the image by the image capturing card, rectify the position of robot and adjust welding procedures by A/D and D/A cards.3. A multithreading program was developed in Visual C++ computer language to realize the functions of capturing image, image processing, data processing, seam tracking and adjusting welding procedures. A control cycle is 0.2s. In this paper, image features was researched for 6mm thick LD10 aluminum alloy welding during square-wave AC GTAW process. A method of extracting simultaneously arc stream profile and welding seam edge image was proposed. The stable algorithms of image processing was developed including median filter, Roberts operator, automatic threshold value, thinning, fitting edge by least square method, etc. The algorithms can process the images in different light intensity and in different welding current (in the range of 200A~300A), which is usually used for making welds in medium plate aluminum alloy weld (in the range of 3mm~8mm). The precision of image processing is in the range of±(0.1~0.2)mm.A seam tracking technology free from calibrating robot was proposed in order to be easily applied. The visual sensor detects the offset of the torch to the seam center in front of the weld pool, and the computer sends the rectifying voltage to the robot controller in real time.The rectifying principle of robot was analyzed in order to design a sectional type adapting PID controller according to the offset change, which can assure the robot to track seam quickly and stably. The error of seam tracking is in the range of±0.3mm for testing work-piece designed with large offset, and in the range of±0.2mm for usual work-piece.The prediction mode of reinforcement was built for the GTAW welding procedure with backing bar through analyzing the change rule of reinforcement with the change of seam gap, wire feed rate, width of the pool and welding current. The experiment results show the mean error of the mode is 0.085mm and root mean square error is 0.256, so it has high reliability and feasibility.Then, the control technology of backing welding in multi-pass welding was researched. The Fuzzy current controller was built according to detecting seam gap. The closed-loop controller of wire feed rate was constructed using prediction model of reinforcement as the feedback element.The seam tracking technology and weld formation control technology were validated in different shape seam, respectively. The results show that the error of seam tracking is±0.3mm and±0.6mm for line seam and flange seam, respectively, and the length with reinforcement in range of (-0.5, 0) mm can reach 96% and 93% of the whole seam, respectively.At last, the presented arc welding robot system based on passive vision was applied in simulation production of GUABAN and flange production with seam tracking and weld formation technology. The quality of the weldments met the standard of first-order (highest quality according to standardYS010-97) welding seam in terms of dimensions and soundness as demonstrated by x-ray inspection, which fulfilled the requirement of spaceflight productions.更多还原