基于宽动态视觉传感的GMAW机器人焊接偏差实时识别及电弧监测研究

发布时间：2018-05-21 21:43

本文选题：视觉传感 + 焊接偏差　；参考：《华南理工大学》2016年博士论文

【摘要】：焊接自动化和智能化是焊接技术的发展趋势。为了克服示教再现型焊接机器人的不足,在焊接过程中利用传感器和相关算法对焊接偏差进行实时识别的同时对电弧进行实时监测,将大大提升焊接机器人的自动化水平和智能化水平。本课题提出了不使用外部光源或滤光片,仅使用宽动态范围(Wide Dynamic Range,WDR)相机采集焊接图像的视觉方案。在此基础上构建了熔化极气体保护焊(Gas Metal Arc Welding,GMAW)机器人系统,并开展了以下几个方面的研究:(1)提出一种基于熔池中心的CO_2焊焊接偏差识别算法。利用基于快速混合重建的形态学算法自动提取感兴趣区域(Region of Interest,ROI)。开发了改进的Canny算法得到坡口边缘,并结合Hough变换和先验知识提取出坡口边缘。为了拟合出熔池前端边缘,依据熔池前端的外形类似椭圆的特点,提出一种基于条件约束的自适应差分进化(Differential Evolution,DE)算法。求取熔池边缘的两个特征点并将其作为DE算法寻优迭代的约束条件。然后,初始化其他三个未知点的种群并重新定义了评价函数和终止条件。变异因子可在算法进化中自适应地确定。在此基础上计算得出焊接偏差。最后,通过试验对算法的鲁棒性和精度进行了验证并阐述了其实时性。(2)提出一种基于电弧顶端中心的MAG焊焊接偏差识别算法。开发了基本全局阈值和最佳全局阈值相结合的双阈值方法分割出精确电弧区域。根据MAG焊射流过渡时电弧和焊丝末端的衔接关系,提出了焊丝中心线提取方法。在改进Canny算法基础上,结合MAG焊焊接图像的特点,对Canny算法进行了进一步的优化、改进和完善,提出一种基于Canny理论的坡口边缘检测算法。ROI经高斯滤波后,分别计算x、y方向的梯度幅值。对x方向梯度幅值图像进行坡口边缘提取而y方向梯度幅值图像产生掩模,两者做逻辑与操作后得到单像素坡口边缘二值图像。针对基于Canny理论的坡口边缘检测算法,提出了高斯滤波标准差的自调整方法。试验表明,基于电弧顶端中心的焊接偏差识别算法的精度满足MAG焊焊缝跟踪的要求。(3)阐述了导电嘴与工件之间的距离(Contact Tip to Work Distance,CTWD)与焊丝伸出长度变化的基本原理。运用CTWD的静态数学模型和焊丝伸出长度模型分析了CTWD与焊丝伸出长度之间的关系,并设计CTWD阶跃变化试验进行了验证。分析了喷嘴边缘在焊接图像中的特征并提出了焊丝伸出长度检测算法。试验表明,算法可有效检测焊接图像中的焊丝伸出长度,精度可满足GMAW机器人焊接过程中保持CTWD基本恒定的要求。(4)在射流过渡、射滴过渡和旋转射流过渡三种不同的熔滴过渡过程中分别采集了包含圆锥形、钟罩形和扭曲束形三种不同电弧形态的焊接图像。在运用改进Canny算法提取非闭合电弧边缘的基础上,提出一种改进的快速行进法(Fast Marching Method,FMM),重新定义了FMM的速度函数和停止准则,从而获得了完整的电弧形态。提出一种基于标记图的识别方法对三种电弧形态进行识别。试验表明,电弧形态提取算法和电弧形态识别算法能够在MAG焊过程中有效提取并识别电弧形态,可应用于焊接过程的电弧实时监测。最后,通过卡尔曼滤波算法利用前一帧的偏差预测值和当前帧的偏差识别值来估计最优焊接偏差值。结果表明,焊接偏差的最优估计值更接近焊接偏差预设值。以GMAW机器人系统为平台,分别针对CO_2焊和MAG焊设计了二维、三维焊接轨迹和电弧监测试验,验证了焊接偏差实时识别、CTWD调节和电弧监测在自动化焊接过程中的有效性。
[Abstract]:Automation and intelligentization of welding is the development trend of welding technology. In order to overcome the shortage of teaching and reproducing welding robot, the real-time recognition of welding error and real-time monitoring of welding arc by sensors and related algorithms in the process of welding will greatly improve the automation level and intelligent level of the welding robot. The project proposes a visual scheme that uses a wide dynamic range (Wide Dynamic Range, WDR) camera to collect welding images without using an external light source or filter. On this basis, a melted polar gas shielded arc welding (Gas Metal Arc Welding, GMAW) robot system is constructed, and the following studies are carried out: (1) a kind is proposed based on the molten pool. The CO_2 welding error recognition algorithm of the heart is used to automatically extract the region of interest (Region of Interest, ROI) based on the fast hybrid reconstruction algorithm. The improved Canny algorithm is developed to get the edge of the slope, and the edge of the slope is extracted with the Hough transformation and prior knowledge. An adaptive differential evolution (Differential Evolution, DE) algorithm based on conditional constraints is proposed. The two feature points of the edge of the pool are obtained and the constraint conditions for the optimization of the DE algorithm are obtained. Then, the population of the other three unknown points is initialized and the evaluation function and the terminating condition are redefined. The difference factor can be adaptively determined in the evolution of the algorithm. On this basis, the welding deviation is calculated. Finally, the robustness and accuracy of the algorithm are verified by the experiment and its practicality is expounded. (2) a method of identifying the MAG welding deviation based on the center of the arc is proposed. The basic global threshold and the best global threshold are developed. The precise arc region is segmented by the combined double threshold method. Based on the connection between the welding wire and the end of the welding wire, the method of extracting the center line of the welding wire is proposed. On the basis of the improved Canny algorithm and the characteristics of the welding image of the MAG welding, the Canny algorithm is optimized, improved and perfected, and a kind of Can based on the Can is proposed. The slope edge detection algorithm of NY theory.ROI calculates the gradient amplitude in the direction of X and y after Gauss filtering. The masks are generated for the gradient edge of the X direction gradient image and the gradient amplitude image of the Y direction. Both of them get the two value image on the edge of the single pixel slope after the logic and operation. For the edge detection based on the Canny theory, the slope edge detection based on the Canny theory is detected. The self adjustment method of the Gauss filter standard deviation is proposed. The experiment shows that the accuracy of the welding deviation recognition algorithm based on the center of the arc top satisfies the requirements of the MAG welding seam tracking. (3) the basic principle of the distance between the guide nozzle and the workpiece (Contact Tip to Work Distance, CTWD) and the elongation of the welding wire is described. The application of CTWD The relationship between CTWD and the elongation length of welding wire is analyzed by the static mathematical model and the length model of welding wire. The CTWD step change test is designed and verified. The characteristics of the nozzle edge in the welding image are analyzed and the detection algorithm of the length of the welding wire is put forward. The experiment shows that the method can effectively detect the welding wire in the welding image. The length and precision can satisfy the requirement of maintaining the basic CTWD constant during the welding process of the GMAW robot. (4) in the three different droplet transition processes of the jet transition, the droplet transition and the rotating jet transition, the welding images containing three different arc shapes including conical, bell shaped and twisted beam are collected. The improved Canny algorithm is used to extract the welding images. On the basis of the edge of unclosed arc, an improved fast moving method (Fast Marching Method, FMM) is proposed. The velocity function and stopping criterion of FMM are redefined, and the complete arc shape is obtained. A recognition method based on the label graph is proposed to identify three kinds of electric arc states. The arc shape recognition algorithm can effectively extract and identify the arc shape during the MAG welding process, and can be applied to the arc real-time monitoring of the welding process. Finally, the optimal welding deviation value is estimated by using the Calman filtering algorithm and the deviation recognition value of the previous frame. The results show that the optimal estimation value of the welding deviation is shown. It is closer to the preset of welding deviation. Based on the GMAW robot system, the two-dimensional, three-dimensional welding trajectory and the arc monitoring test are designed for the CO_2 welding and MAG welding respectively. The validity of the real-time identification of the welding deviation, the CTWD regulation and the arc monitoring in the automatic welding process are verified.
【学位授予单位】：华南理工大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP391.41;TP242

【参考文献】