【作者】 陈华斌；

【导师】 陈善本；

【作者基本信息】 上海交通大学 ， 材料加工工程， 2009， 博士

【摘要】 焊接自动化是焊接技术发展的一个趋势,由于现有的“示教再现型”焊接机器人在焊接过程中缺少对外部信息传感反馈和实时调节的功能,不能满足航天高技术产品复杂焊缝精密焊接的要求。焊接过程中的热变形、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缝成形质量。本文运用综合分析的手段,找到影响焊接质量的关键因素以及这些因素之间的相互关系并进行量化,对产品的焊接质量进行重点、定量的控制。从焊接变形预测与工艺优化、系统设计、焊接动态过程建模以及焊缝成形实时控制等多个角度实现五通连接器精度焊接成形质量控制。以非线性有限元技术为基础,运用MSC.Marc软件对运载火箭动力系统五通连接器焊接过程进行了数值模拟,通过调整焊接顺序、改变焊接热输入量获得了五通连接器焊接变形影响规律。在此基础上,进一步优化了焊接工艺和夹具设计。为了保证焊接过程中的稳定性,必须精确控制焊接热输入量、根据焊接热变形、间隙情况实时调整焊接规范参数。为此,建立了基于视觉传感的弧焊机器人在线质量监控系统,使其能够完成从起弧、工艺参数设定、熔池图像自动采集及图像尺寸计算、焊接规范实时调节以及自动熄弧的完整焊接工作过程。焊接过程视觉传感是为了实时准确地提取表征熔池形状和大小的特征信息,定义了焊缝正、反面熔宽,焊缝余高、间隙及焊缝位置对熔池形状及运动方向进行了描述。运用DT_CWT+BivaShrink小波降噪、改进的约束最小二乘方法进行对焊缝噪声图像进行降噪恢复和复原,提出了一种适合在焊缝图像中寻找目标区域的高精度自适应阈值分割算法。针对熔池图像特点,采用边界表示与描述算法来获取熔池图像的形状参数,并借助分段多项式拟合的方法对熔池边缘进行了恢复,提取到焊缝正面熔宽尺寸;而对焊缝间隙图像,直接采用面积滤波及Hough变换处理方法获取间隙尺寸信息。最后,运用上述算法对平焊法兰、五通连接器等实际工件的焊接图像进行处理,进一步验证了算法的鲁棒性和实时性。弧焊焊接过程是被焊金属在电弧热源热输入的作用下,产生局部熔化,形成熔池,最后液态金属凝固之后形成焊缝。根据焊接热过程的这一特点,本文引入非线性Hammerstein模型描述焊接热过程,并建立了IpVf-WbHt之间的关系模型。通过实际焊接过程观测数据与模型实际输出进行了比较,模型精度满足焊接过程控制的要求。针对实际焊接过程中背面熔宽无法实时检测,本文还建立了焊接规范参数同熔池正面参数联合预测熔池反面宽度的RPROP网络动态模型,该算法具有更好的学习效率与泛化能力,预测模型的准确度高于传统分析方法。焊接过程是一个复杂的、时变的、不确定的过程,本文基于大量工艺试验基础上,提出了送丝速度、焊接电流同焊缝间隙变化量的定量法则。在此基础上,设计了基于Hammerstein模型的Ip-Wb非线性自校正控制器非线性自校正控制器和基于参数预置前馈的复合智能控制器,并进行了仿真和控制器有效性验证试验。焊接峰值电流作为单变量的非线性自校正控制器,能够较好地克服外界干扰、保证熔池反面宽度比较均匀一致。然而,利用送丝速度、焊接电流多变量参数预置前馈复合智能控制器即使在变间隙和变错边的干扰下都能得到焊缝正、反面成形均匀一致的理想焊缝。通过对平焊法兰、螺旋管的工艺试验,进一步验证了基于参数预置前馈的复合智能控制器的可靠性和稳定性,统计计算结果表明,平焊法兰和螺旋管焊缝背面熔宽和正面余高剩余标准差分别为:(0.84 mm,0.27 mm)和(0.61 mm,0.30mm),符合航天标准要求,并确定了该控制器允许的焊接间隙大小变化范围(0,1.8 mm)。最后,基于本文研究成果构建的焊接机器人GTAW焊接质量控制系统,对运载火箭动力系统五通连接器进行了焊接。焊后X,Y,Z方向上的变形位移量分别为UX=0.6mm, UY=0.8mm, UZ=-0.3mm,其中Z方向焊接变形较传统手工TIG方法降低了73%。所有焊缝均能满足YS010-97规定的Ⅰ级焊缝标准,产品焊接成形质量得到了极大改进,为进一步工程应用奠定了基础。更多还原

【Abstract】 Welding automation is a trend of welding technical development. Since the current“teaching and playback type”robot lack external information sensor feedback and the function of real-time adjusting, it could not meet the demands for the precision welding of complicated seam in aerospace high-tech product. Considering the uncertain factors of welding process, such as the welding distortion, alternate edge in the welded seams and the variable quantity of the gap, and they would affect the weld appearance quality directly.In this paper, several influence factors were analyzed synthetically by systematic method to emphasize in quantitative control of welding quality. Analysis and discussion from different angles, such as the prediction of the weld distortion, welding process optimization, system design, welding dynamic process modeling and weld appearance control, was studied. It is demonstrated that systematic method can meet the precision welding demands.Based on the non-linear finite element method, the numerical simulation of five-port connector was carried out using MSC.Marc software. The complex models are presented to normalize mechanical boundaries, thermal boundaries and heat source model. Based on the normalized finite element model, the above aerospace structure is calculated again, respectively studying the effects of the welding sequence and welding heat input quantity to welding distortion. On this basis, the welding fixture of the five-port connector was presented for the optimization design.In order to ensure the welding stability, welding heat input should be controlled precisely. And then the welding parameters could be adjusted according to the welding deformation and the dimension of the weld gap. Therefore, online quality control system for arc welding robot was established. The main functions of this system are listed as follows: arc start operation, welding parameters setting, the welding pool information acquisition and processing, real time adjusting and controlling the main parameters, and arc off operation.In order to extract the characteristic information of the weld pool accurately, several parameters about weld pool geometry and welding direction are defined. We make de-noising weld image by combining the Dual Tree Complex Transform Wavelet (DT_CWT) with the Biva-Shrink method. A new threshold segmentation algorithm of welding image with self-adaptive capacity and high precision was introduced. Aimed at the characteristic of the degradation weld pool, the shape parameter of the weld pool was obtained using boundary representation and description algorithm. Then, the restoration and geometry of the weld pool was extracted through the piecewise polynomial fitting method. For top-side weld pool image, the dimension information of the weld gap was extracted directly by the area filtering and Hough transform method. At last, plane flange weld and five port connector weld image was processed using above algorithms, and the results were reliable and stable.Arc welding procedure is the process, in which the welded material under the heat input started melting locally and then form weld pool. According to the characteristic of the weld process, non-linear Hammerstein model was led into the welding process. On this basis, IpVf-WbHt relation model was established. The results show that the model precision can meet the demands of the actual control. Meanwhile, the direct measurement of back-side width of the weld pool is very difficult, so RPROP dynamic prediction model was built to predict the Wb through procedure parameters and top-side information of the weld pool. The result shows that the RPROP algorithm provides both higher learning efficiency and stronger generalization capacity versus traditional method.The welding process is a complex, time-variant and uncertain system. In this paper, a quantitative rule (about wire feeding rate, welding current and the variable quantity of the gap) is introduced. On this basis, a nonlinear self-corrected controller and the compound intelligent controller with parameter preset was designed and its validation is conformed by Matlab Simulink toolbox. Butt welding experiments were conducted on unequal thickness test plate with varied gap. The results show that nonlinear self-corrected controller with weld current control variable could get better controlled performance.For the requirement of stabilizing backside width and weld reinforcement simultaneously, the compound intelligent controller with parameter preset can stabilize the shape of the weld pool under the conditions of the varied gap and alternate edge.In order to testify the reliability and stability of the compound intelligent controller in farther, the experiments were conducted on the plane flange and the spiral tube.It has been shown that the residual standard deviation of the back-side weld width and weld reinforcement are 0.84 mm,0.27 mm , 0.61 mm and 0.30mm, and it meets the demands of the space flight standard. Meanwhile, the maxium dimension of the gap is 1.8mm.In the end, based on the above development of robotic GTAW welding quality control system, the experiments were conducted on the five-port connector. After welding process, the distortion of the X,Y,Z direction are UX=0.6mm, UY=0.8mm, UZ=-0.3mm. The distortion of the Z direction was decreased by 73% than the traditional TIG process. The quality of the weldments met the standard of first-order (according to standard YS010-97).In general, the production quality was inproved increasingly. The work has laid the foundation for the engineering application in further.更多还原