【作者】 苏春建；

【导师】 赵军； 官英平；

【作者基本信息】 燕山大学 ， 材料加工工程， 2008， 博士

【摘要】 板材弯曲在汽车、农机、航空航天、仪表等领域中应用极为广泛。弯曲成形中最突出的问题是弯曲回弹难以精确控制,尤其是形状复杂的弯曲件。影响弯曲回弹的因素很多,即使是同牌号、同批次的板材,由于材料性能及工艺条件的波动也会引起弯曲件回弹量的较大变化。当前制造业市场的发展趋势是对高精度、多品种、小批量的弯曲件的需求日益增加,而原材料成本与最终产品价格的差额逐渐减小,并且竞争日趋激烈,迫使制造业寻求更为有效的方法,降低生产成本,提高生产率。采用智能化控制技术来实现弯曲回弹的精确控制,是解决上述问题的一种有效途径。板材弯曲智能化是将每个坯料看成性能各异的个体,在弯曲成形过程中,实时识别出每个个体的性能参数及工况参数,进而预测出最优的工艺参数,并以最优的工艺参数完成弯曲成形及其卸载的过程,从而达到更高的精度。本文在板材拉深成形和V形件自由弯曲成形智能化控制的研究成果基础上,分析了板材帽形件弯曲智能化需要解决的关键技术,在理论解析、参数识别及最优工艺参数的预测等方面的相关问题展开了系统研究。在板材弯曲智能化控制所要求的四个基本要素中,参数识别模型和最优工艺参数预测模型的建立是基于对弯曲成形规律认识的基础上的。利用板材成形塑性理论,根据弯曲变形特点,对多角单次弯曲成形及回弹机理进行了深入的理论研究,建立了能够描述多角单次弯曲成形及其回弹过程的U形件自由弯曲模型和帽形件弯曲模型。得出了几何参数和力学参数对弯曲回弹的影响规律,为帽形件弯曲智能化控制提供了理论依据。利用理论解析、有限元模拟和实验等研究手段,分析了对U形件自由弯曲和帽形件弯曲成形及卸载后回弹有影响的主要因素,从而确定了帽形件弯曲智能化控制过程中的参数识别及最优工艺参数预测模型。网络拓扑结构选择前向神经网络结构,Levenberg-Marquarat算法作为网络优化算法,并利用Matlab编程语言编写了神经网络算法程序。通过数值模拟和实验得到数据样本,进行网络训练,使得识别收敛精度达到1‰,且识别的泛化精度较高。此外,研究了样本数据和隐层节点数目对网络模型的效率、精度和泛化能力的影响规律。采用虚拟仪器控制软件LabVIEW、6062E数据采集卡及相关模块,建立了便携式数据采集系统。在该系统上开发了信号采集和传感器标定等程序,获得了令人满意的信号检测与控制的速度和精度。开发了信号采集程序与识别和预测模型之间的接口程序,从而实现了帽形件弯曲过程中材料参数的实时识别与最优工艺参数的实时预测及控制。建立了帽形件弯曲智能化控制系统,通过控制压边力,实现了弯曲回弹的精确控制,从而实现了帽形件弯曲成形的智能化控制。四种板材的帽形件弯曲智能化实验结果表明,按前述理论和实验研究结果建立的智能化弯曲系统是成功的。更多还原

【Abstract】 Sheet metal bending is widely used in the industry field of automobile, agriculture machine, aviation, instrument and so on. The most prominent problem in bending process of sheet metal is that it is difficult to control bending springback accurately, especially complex shape bending workpiece. There are many factors that effect the springback of bending, even if is the same trademark、batch plate, because the material performance and the craft condition undulation also can cause the springback comparatively big change. The current manufacturing industry market trend of development is to the high accuracy, the multi-varieties, the small batch bending workpiece demand increases day by day, but the difference between raw material cost and the final product price reduces gradually, and the competition in market change intensely day by day, so seeks a more effective method is needed to reduce the production cost and to enhance the productivity. Uses the intellectualization to realize the springback accurately controlled, is one effective way to solve above question.The intellectualization of sheet metal bending regards each blank as individual with different performance, identifies in real-time the performance parameter of each individual, consequently confirms the optimal craft parameter, and accomplishes the process of bending forming and unloading with the optimization craft parameter, accordingly achieves the high precision.Based on the research achievements of intelligent control of deep drawing and v bending, the key technologies of intelligent bending are analyzed, and the relevant issues to the theoretic analysis, the identification of parameters, the prediction of optimal technological parameter and the establish of control system is studied in this paper.Among the four basic elements required by intelligent bending of sheet metal, the establishment of the identification model of parameters and the prediction model of optical technological parameter is dependent upon the level of understanding the forming law for bending. Based on sheet metal plastic forming theory, according to the bending characteristic, has completed the thorough fundamental research to the U bending forming and the mechanism of springback, has established the U free bending model and cap-bending model which can describe precisely the forming of U bending, has obtained the influence rule of the geometry parameter and mechanics parameter to the bending springback, has provided the basic theory for the real-time identification and real-time prediction of bending intellectualized control.The dominant factors influenced on U bending forming and springback after unloading has been analyzed by using theoretic analysis、finite element method simulation and experiments, and then the identification model of parameters and the prediction model of optimal technological parameter are determined. The Levenberg-Marquarat is chosen as optimal algorithm of neural network whose topology structure is feed forward network, and Matlab language has been used to compile the neural network arithmetic program. The network has been trained using the date sample obtained by experiments and numerical simulation. Convergence precision of the identification network reaches 1‰, and generalization precision of identification is high. Furthermore, the influence of sample date and hidden nodes on network convergence efficiency, precision as well as generalization has been studied.A portable DAQ system is established by applying virtual instrument control software LabVIEW、6062E card and other modules. Based on portable DAQ system, the signal acquisition and sensor calibration are developed, and get the satisfied speed and precision to monitoring and control of signals. Interface programs between DAQ and the model of identification and prediction have been developed, accordingly has realized the real-time identification and the optimal craft parameter real-time predication and control. The intellectualization control system of cap bending is established, the bending springback is controlled precisely through control pressure accordingly the intelligent control of sheet metal bending forming is realized. The results of four kind of plate in intelligent bending experiment indicated that the system is successful.更多还原