【作者】 周毅；

【导师】 章国宝；

【作者基本信息】 东南大学 ， 控制科学与工程， 2024， 硕士

【摘要】 随着建筑行业的蓬勃发展,预应力高强度混凝土(Pre-stressed High-strength Concrete,PHC)管桩凭借其优异性能被广泛应用于基础工程中的支撑结构,使用机器人对管桩进行自动焊接的需求也日益增多。然而在管桩焊接的过程中会存在大量的焊接噪声和不确定干扰,可能导致机器人在焊接过程中产生定位偏差。因此本文基于校企合作开发的PHC管桩焊接机器人平台,对机器人的自动定位方法展开了重点研究。一、介绍焊接机器人的应用场景特点和功能指标需求,给出了焊接机器人和视觉传感系统的结构设计和硬件选型。通过分析视觉传感器成像原理,构建出了图像像素坐标到空间世界坐标的转换模型。对模型中相机参数和结构光平面参数完成了标定,并进行了验证实验,结果表明系统的标定误差能够满足焊接机器人的定位需求。二、针对传统焊缝识别算法难以处理强噪声干扰的局限性,本文设计了一种基于改进主动轮廓模型的焊缝识别算法。该算法添加了轮廓初始化步骤,来解决主动轮廓模型对初始轮廓位置和形状敏感的问题,同时改进了主动轮廓模型的能量函数,针对焊缝图像的灰度值分布特征优化梯度计算,并添加了带有梯度信息的结构张量,来确保轮廓在结构光条纹和焊接噪声相交的角点处能够正确收敛。对最终轮廓采用分段直线拟合的方法提取出焊缝特征点,并进行实验验证,结果表明该算法能够准确识别焊缝特征点。三、通过坐标转换从焊缝特征点二维图像信息确定了待焊点的三维位置信息,并对焊接轨迹中插补点求解,完成了轨迹规划。为了实现焊接机器人的精确定位,对焊枪实际位姿和目标位姿之间的偏差设计了模糊控制方案,结合一、二阶模糊控制的优点进一步设计出混合模糊控制模型,并对多种控制模型进行仿真比较实验,验证了混合模糊控制具有良好的跟踪精度、响应速度和抗干扰性。四、完成了焊接机器人软件系统的开发,并进行了自动定位实验。在实验中可以有效地提取出焊缝坡口信息,并且可以将采集到的结构光条纹数据点拼接来还原焊缝的三维形貌。最终结果表明系统能够准确识别焊缝并将焊枪精确定位到待焊点,多组实验的平均定位误差为0.29mm,满足实际焊接要求。更多还原

【Abstract】 With the booming development of the construction industry,pre-stressed high-strength concrete(PHC)pipe piles are widely used in supporting structures in foundation engineering due to their excellent performance.The demand for automatic welding of pipe piles using robots is also increasing.However,during the welding process of pipe piles,there will be a large amount of welding noise and uncertain interference,which may lead to positioning deviation of the robot during the welding process.Therefore,this article focuses on the automatic positioning method of the PHC pipe pile welding robot platform developed through school enterprise cooperation.Firstly,this thesis introduces the application scenario characteristics and functional indicator requirements of welding robots,and provides the structural design and hardware selection of welding robots and visual sensing systems.A conversion model from image pixel coordinates to spatial world coordinates was constructed by analyzing the imaging principle of visual sensors.The camera parameters and structured light plane parameters in the model have been calibrated and validated through experiments.The results show that the calibration error of the system can meet the positioning requirements of the welding robot.Secondly,in response to the limitations of traditional weld seam recognition algorithms in dealing with strong noise interference,this thesis proposes a weld seam recognition algorithm based on an improved active contour model.This algorithm adds a contour initialization step to address the sensitivity of the active contour model to the initial contour position and shape.At the same time,it improves the energy function of the active contour model,optimizes gradient calculation for the grayscale value distribution characteristics of the weld seam image,and adds a structural tensor with gradient information to ensure that the contour converges correctly at the corner where the structured light stripes and welding noise intersect.The segmented straight line fitting method was used to extract weld seam feature points from the final contour,and experimental verification was conducted.The results showed that the algorithm can accurately identify weld seam feature points.Thirdly,the three-dimensional position information of the welding point was determined through coordinate transformation from the two-dimensional image information of the weld seam feature points,and the interpolation points in the welding trajectory were solved to complete the trajectory planning.In order to achieve precise positioning of welding robots,a fuzzy control scheme was designed for the deviation between the actual pose of the welding gun and the target pose.Combining the advantages of first and second order fuzzy control,a hybrid fuzzy control model was further designed.Multiple control models were simulated and compared to verify that the hybrid fuzzy control has good tracking accuracy,response speed,and anti-interference performance.Finally,the development of the welding robot software system was completed,and automatic positioning experiments were conducted.In the experiment,weld groove information can be effectively extracted,and the collected structured light stripe data points can be concatenated to restore the three-dimensional morphology of the weld.The final result shows that the system can accurately identify the weld seam and accurately position the welding gun to the desired welding point.The average positioning error of multiple experiments is 0.29mm,which meets the actual welding requirements.更多还原