【作者】 高忠林；

【导师】 胡绳荪；

【作者基本信息】 天津大学 ， 材料加工工程， 2008， 博士

【摘要】 现代工业生产对号称“工业缝纫机”的焊接设备提出了更高的要求。提高焊接工艺性能的关键之一在于先进的焊接设备,实现先进的焊接设备在于采用先进的控制理论及针对工艺特点采取合理的控制算法。本文基于数字化焊接的概念,对数字化焊接电源,熔化极气体保护焊(GMAW)焊接过程数学模型以及焊接信号(电流、电压,焊接速度等)特征分析,焊接质量预测以及在线监控等基础问题进行了研究。首先研制了以数字信号处理器(DSP)与单片机(MCU)双处理器为控制核心的数字化弧焊电源。完成了主电路,数字控制系统电路,IGBT驱动电路,电流、电压反馈电路,人机接口电路,送丝系统电路以及保护电路的研制,并完成了电源软件系统的设计与调试。其次,进行了CO2焊接短路过渡波形控制研究,分析了CO2短路过渡可以减小飞溅和改善焊缝成型的电压、电流波形。提出了三种波形控制方法,对三种控制方案及效果进行了比较分析。再次,为了实现先进控制方法在焊接电源控制中的应用,分析了GMAW焊接过程所涉及的焊接参数及相互关系,对GMAW过程电路系统、电弧系统、熔滴上的作用力、熔滴过渡、焊丝熔化速度进行分析。建立GMAW过程数学模型;应用基于微分几何的反馈线性化方法,将GMAW过程电流及弧长模型同胚映射为等价的线性系统,使复杂的非线性控制问题转化成简单的线性系统的控制问题;将滑模变结构控制方法应用于焊接电流及弧长的控制,并运用Matlab进行仿真研究。然后,进行了有关CO2焊接电信号分析处理研究。DSP强大的数据处理能力和快速运算能力为焊接信号的实时处理分析提供了合适的平台,为形成焊接过程质量实时评价系统、形成焊接过程的实时闭环控制提供了可能。利用相关性分析、傅立叶谱、短时傅立叶、功率密度谱、小波分析等多种现代信号分析方法对CO2焊接电压、电流的时域、频域及时频域特征进行数据挖掘,从信号分析角度丰富对焊接电压、电流信号深层所蕴含信息的认识,为实现焊接过程的实时监控提供理论基础。最后,采用BP算法经样本训练对焊缝几何尺寸进行预测研究。针对普通BP算法存在的问题,采用自适应学习率及附加动量项的方法进行改进,以提高BP网络的运算速度。采用支持向量机,分别运用线性核函数,多项式核函数,RBF核函数以及ERBF核函数对焊缝尺寸进行预测,从而实现通过神经网络模型预测焊缝形貌来达到焊接质量的实时监控及焊接过程的在线控制的目的。更多还原

【Abstract】 With the development of digital technology, industrial production sets a still higher demand on welding equipment, which is called“industrial sewing machine”. Advanced welding equipment is the one of key factors to improve the welding technological property. Aiming at the technology characteristics, advanced control theories and appropriate control algorithms are used to realize the advanced welding equipment. In this paper, based on digital welding, digital welding power source, mathematical model of GMAW welding process, characteristic analysis of welding signals(current, voltage, welding speed, etc), prediction of welding quality, and on-line monitoring are studied.Based on dual processors of digital signal processor (DSP) and single chip microcomputer (MCU) a multi-function arc welding power source is made. Main circuit, digitization control circuit, driving circuit for IGBT, feedback circuits for current and voltage, circuit for man-machine interface system, circuit for wire feeding system and protection circuit are designed. In addition, system software is programmed and debugging the whole system.The study of CO2 arc welding short-circuit waveform control is done. The discussion of ideal voltage waveform and current waveform, analysis of characteristics of welding parameters, and further study of scheme of current waveform control in CO2 short circuit transition welding are taken. Three control schemes are designed and compared.For the application of advanced control methods in welding power source circuit system, arc system, forces on droplet, droplet transition and melting rate of welding wire in GMAW welding process are analyzed based on he welding parameter and its mutual relation. Mathematical model of GMAW welding process is established. By the feedback linearization method of differential geometry, Equivalent linear system is achieved by using homeomorphism mapping of models of current and arc length, which makes the complicated nonlinear control problem transformed into a simple linear control problem. Because of the decreasing of complex degree of controller design, control Effect improves. The method of sliding mode variable structure control is used in the control of welding current and arc length. And it is simulated with MATLAB. Time domain characteristics, frequency domain characteristics and time-frequency domain characteristics of the current and voltage signals are studied by using advanced signal analysis, such as correlation analysis, fourier spectrum, short-time fourier, power density spectrum, wavelet analysis and so on. The understanding of current and voltage signals is enriched.Prediction of weld line size is achieved by sample training at BP algorithm. Because of the existing problems of general BP algorithm, operation speed of BP Network is improved by using adaptive learning rate and additional item. In addition, weld size is predicted by using support vector machine(SVM),which is trained with linear kernel function, polynomial kernel function, RBF kernel function and ERBF kernel function. Real-time monitoring of welding quality and on-line control of welding process are achieved by predicting the weld appearance based on neural network model.更多还原