【作者】 王立伟；

【导师】 孙鹤旭；

【作者基本信息】 河北工业大学 ， 电气工程， 2012， 博士

【摘要】 在钢铁行业，板材在整个钢铁产品中的比重是衡量钢铁生产水平的重要依据。因此，板材生产线的连续稳定运行已是一个关键的研究课题。如何确保激光焊接的焊缝质量能够达到板材连续生产线的要求已经成为钢铁行业的关键技术之一。作为板材连续生产线的第一个工艺步骤，高功率激光焊接焊缝的焊后焊缝检测技术是判断焊缝质量是否满足连续生产线稳定连续运行的重要依据。本文将机器视觉技术融入到针对钢板连续生产线激光焊接中。从分析激光焊接的机理入手，研究激光功率、焦点位置、焊接速度等钢板连续生产激光深熔焊的主要影响因素，针对激光焊机这个机电复杂的系统，分别从机械和电气两部分着手，剖析了某生产线激光焊机的各部分组成结构和特点，通过解析焊接各环节和约束条件，研究了激光焊机的工艺过程。主要研究了激光焊接的微细焊缝的视觉特征提取方法。提出了一种区域搜索算法实现最优感兴趣区域的确定,在保留最多信息的同时，最大化地减小的计算负担。提出一种焊缝特征点鲁棒提取方法,利用最小二乘法，样条曲线拟合方法实现了结构光条纹信息的直线部分，曲线部分精确提取，进而实现了宽度、错边、余高等维系焊缝特征参数的提取提供了丰富的数据信息，以便于进一步的焊缝质量评估。针对激光焊接的微细焊缝，研究了其三维重建的关键技术。提出了一种基于显微视觉的结构光视觉测量方法，建立了视觉模型，结构光测量模型，得到了焊缝视觉特征对应的笛卡尔空间的位置信息。进而提出一种基于弦长的B样条曲线拟合方法，减弱了测量噪声的影响，提高了焊缝曲线拟合的精度，最终结合视觉测量和拟合方法完成了激光焊接微细焊缝的三维重建。分析了激光焊接焊缝缺陷种类与质量等级的基础上，重点研究了激光焊接焊缝质量监控技术，设计了一种基于结构光视觉图像的焊缝质量智能评估方法，采用一种递归神经网络分类器，根据图像处理技术所获取的焊缝表面几何特征对焊后焊缝的焊接质量进行等级划分；基于Labview虚拟仪器技术设计了一款新颖的焊缝质量监控软件，可识别焊缝的关键质量参数，实现了焊缝特征参数的计算、数据三维可视化和数据的存储管理,为提高焊接质量提供了有力的保证。根据激光焊接工艺分析和系统结构，提出钢板连续生产激光焊机的位置伺服跟踪方案，实现焊缝、焦点位置以及间隙调整等精确的速度和位置控制，为提高焊接质量提供了有力的保证。更多还原

【Abstract】 In the iron and steel industry, the ratio of metal sheets in the whole steel products is animportant index mesauring the level of steel production. Thus, the stability, operating safety ofthe metal sheets production line have always been a key research project. Therefore, how toguarantee the quality of the laser welding seam to meet the requirements on the plate productionline continuous operation has become one of core issues. As first processing step, the high powerlaser welder post weld quality detection technology is a very valuable judgment whether thewelding quality of the high power laser welder meets the requirements of continuous operationof metal sheets production lineAt First, from the analysis of the mechanism of the laser welding, the thesis studies somefactors that influence the high laser welding quality, such as laser power, the focus of theposition, the welding speed and so on. For the electromechanical complex systems of the laserwelder, all the parts of the structure and characteristics of the high laser welder are analysizedfrom the mechanical and electrical two parts. Then the laser welding process is determined by allthe parts of the structure of the high laser welde and related electrical constraint conditions.The dissertation mainly studies the visual feature extraction method of the micro laser weldseam. A kind of region search algorithm is proposed to determine the optimal region of interest,which can keep the most information, maximize reduce calculation burden, at the sametime.Then, a weld feature point robust extraction method is presented by using the least squaremethod, the spline curve fitting method, which realizes the accurate extraction of straight part,curve part in the structure light stripe. And then some weld feature parameters, such as weldwidth, misalignment, reinforcement, can be recognized, which provides abundant datainformation, so as to further weld quality evaluation.Then, a three dimension reconstructure method for the laser weld micro and the thin seam isstudied. A structure light vision measuring methods based on microscopic visual is proposed.And the corresponding visual model, structure light measurement model is established so as toobtain the weld visual characteristics corresponding position information in Cartesian space.Furthermore, And a kind of b-spline curve fitting method based on chord long is put forward,which help to reduce the influence of measurement noise and improve the weld curve fittingaccuracy. Finally the laser welding micro weld3d reconstruction is completed by combiningvisual measurement and fitting method.On the basis of the analysis of the laser welding of weld defect types and quality level, thelaser welding seam quality monitoring technology is studied. And a kind of intelligent weldquality assessment methods based on structure light vision image is designed by using arecurrent neural network classifier. Then, the seam quality can be classed by using the weldgeometry characteristics from image processing results.Based on Labview virtual instrumenttechnology, a novel weld quality monitoring softwar that can identify the key parameters of theweld quality, realize the weld characteristic parameter calculation, data3d visualization and datastorage management is accomplished, which help to improve the laser weld quality and ensurethe safe operation of the metal sheet production line. According to the laser welding process analysis and system structure, a steel continuous production laser welder position servo trackingscheme that can weld and accurate speed and position control of the focus position and gapadjustment and provide an effective scheme to improve the quality of welding is proposed.更多还原