【作者】 李湘文；

【导师】 洪波；

【作者基本信息】 湘潭大学 ， 材料加工工程， 2012， 博士

【摘要】 中厚板焊接结构广泛应用于造船、大型桥梁、石油化工、锅炉容器、重型机械等工业部门，亦是未来钢结构焊接的主流趋势。中厚板结构焊接一般采用多层多道焊，焊接过程中会产生焊接线能量无法控制、后层焊道对前层焊道的再熔化热处理而改变焊道走向、以及多层焊的热影响致使母材变形等诸多问题，导致中厚板结构焊接的自动化程度低下。因此，实现中厚板焊接的自动跟踪显得尤为迫切。1、首先设计出一款适用于中厚板焊接的旋转电弧传感器。针对多层多道焊对传感器智能化焊接的需求，结合目前旋转电弧传感器的各种旋转方案，设计出无级偏心调节的旋转电弧传感器，实现焊接过程中旋转半径的智能化控制。同时，通过ADAMS仿真分析，确定了设计参数和电机型号，机构设计的合理性得以验证；再根据中厚板焊接发热量大等问题，设计了一种水冷结构，并通过ANSYS热分析以及实验，优化了冷却装置，为该机构在中厚板焊接中的连续作业提供了保证。2、建立了多层多道焊的弧长模型，实现不同焊道下的焊缝偏差及旋转半径的求解。基于多层多道焊的坡口模型，建立了旋转电弧传感器的焊接电弧信号模型，实现了不同焊道的电流模拟输出，并与实际电流信号进行了对比分析。考虑实际焊接电流信号的干扰严重性，采用硬件滤波对信号进行预处理。随后对预处理的数据进行主成分分析，建立回归方程求解出焊缝偏差，根据获得的偏差采用拟合与积分相结合的方法优化旋转半径的数值。3、提出摆动旋转电弧传感器方案，建立焊枪的空间姿态弧长模型，实现了焊枪的姿态识别。根据目前应用的电弧传感器信号单一等特点，提出摆动旋转电弧传感器方案，并根据摆动旋转电弧传感器的运动模型建立不同焊枪姿态的弧长模型。由于姿态识别算法信号的特征值要求较高，因此对采样数据采用了Gabor滤波后处理方案，并应用主元法对信号数据进行特征提取与降维，并利用分类器与最速下降法获得当前焊枪的姿态，最终达到识别焊枪姿态的目的。4、开发出多自由度焊接小车为硬件平台的多层多道焊焊缝跟踪系统。根据焊枪姿态的特点，提出六自由度焊接小车方案，建立其运动学模型，通过所获得的最优旋转半径以及焊缝姿态等信息，开发出以工控机为控制核心，多自由度焊接小车为硬件执行平台的多层多道焊焊缝跟踪系统，并实现了对多种焊道、多种轨迹焊缝的自动跟踪。5、为提高系统的跟踪精度，针对不同控制方案提出不同跟踪精度模型的计算方法。通过简化多自由度焊接小车的模型，实现对不同控制方案下的焊缝跟踪精度建立数学模型，仿真和试验验证了这种跟踪精度模型的正确与有效性，为日后采用优秀的控制算法提供了理论指导作用。更多还原

【Abstract】 Plate welding structure is widely used in industrial sectors, such as ship-building, largebridges, petrochemical, boiler containers, heavy machinery and so on. It is also themainstream trend for steel welding. Nowadays, the multi-pass welding is commonly used inwelding plate structure. But some problems leading to the low degree of automation appearedduring plate structure welding, for example: the welding energy can not be controlled duringwelding process, the direction of welding bead may be changed by the re-melting heattreatment effected from the posterior to the anterior layer, as well as the deformation of theparent material occurred by the heat-affection of multi-pass welding etc. Therefore, thetechnology to realize the automatic of seam tracking in plate welding is particularly urgentneeded.1. A rotating arc sensor that is suitable for the plate welding is first designed.Considering the requirements, the multi-pass welding for the sensor intelligent welding,and a rotating arc sensor with stepless eccentric adjustment is designed based on the usingrotating methods of rotating arc sensor. During the welding process, the intelligent control ofrotating radius can be realized as well. After the simulation of ADAMS, the optimal designparameters and motor sizes are determined, as well as the institutional rationality of designverified. Then, for the problems such as large quantity of heat in the plate welding, awater-cooled structure is designed. By the thermal analysis of ANSYS and experimental, thewater-cooled structure is further optimized. This mechanism has provided a guarantee forlong-time welding in the factory.2. The arc length model of multi-pass welding is established, as the seam deviation androtating radius under different welding pass have been solved out.Based on the groove model of multi-pass welding, welding arc signal model of rotatingarc sensor is built. It can achieve current analog output of different welding. And thecomparative analysis between this output and practical current signal is taken in this paper. Inorder to eliminate the serious signal interference of practical welding current, the hardwarefilter is used to preprocess that signal. In the following, the principal component method istaken to analysis the preprocessed data, and the regression equation is built to solve out theseam deviation values. As using the combination method of fitting and integral, the value ofrotating radius is optimized by the obtained deviation.3. With the program of oscillating rotating arc sensor first put forward, and the spaceposture arc length model of welding torch built, the torch gesture recognition has been realized.According to the single feature of arc sensor signal used at present, the program ofoscillating rotating arc sensor is put forward. And the arc length model of different posture inspace is also built based on the motion model of oscillating rotating arc sensor. Because thatthe eigenvalue of signal in gesture recognition algorithm requires higher, the processingprogram of Gabor filter is used for the sampled data. Meanwhile, the principal componentmethod is taken to extract the feature and to reduce dimension of signal data. Finally, the aimof welding torch gesture recognition is reached, after the current torch gesture obtained by theclassifier and the steepest descent method.4. The seam tracking system of multi-pass welding, based on the hardware platform ofmultiple freedom welding car is developed.On the basis of torch posture characteristics, the program of welding car with6-DOF isproposed, and the motion model is built then. Through the obtained information such as theoptimal rotating radius and the welding posture, the multi-pass welding seam tracking systemis developed. For this system, the control core is the industrial control computer, the hardwareexecution platform is the6-DOF welding car. And the automatic seam tracking for severalwelding beads or trajectory welding seams can be realized by this system.5. In order to improve the tracking accuracy of system, the different calculation methodsof tracking accuracy models based on different control schemes are finally put forward.By simplifying the model of multiple freedom welding car, the mathematical models ofseam tracking accuracy under different control options are established. Also, the correctionand efficiency of this tracking accuracy model are verified by the simulations and tests.Doubtlessly, these models can give theoretical guidance to the use of excellent controls in thefuture.更多还原