【作者】 张铁山；

【导师】 张友良；

【作者基本信息】 南京理工大学 ， 机械制造及其自动化， 2004， 博士

【副题名】基于分形理论和神经网络的板料冲压成形接触摩擦与回弹研究

【摘要】 板料冲压成形是汽车、机械、机械电子等制造行业的重要加工方式。在板料冲压成形中，板料与模具的接触摩擦及板料的回弹是密切相关的两个问题，其关键点是摩擦及其状态、屈服条件、受力状态。而摩擦与回弹是进行工艺设计、进行工艺参数确定和模具设计要考虑的关键问题。并且，这些问题又是板料冲压成形工艺设计的难题。因此，本文对板料冲压成形中的接触摩擦和回弹问题进行深入研究。 本文从板料冲压成形的接触与摩擦理论分析入手，在此基础上，建立了基于分形数学的板料冲压成形的接触与摩擦力学模型。并将该模型用于板料沖压成形的载荷计算。而后，利用HOMMELWERKE型触针式表面粗糙度轮廓仪对研究对象进行了测试研究，证明了板料具有分形的特征，为验证板料冲压成形的接触摩擦力学模型的正确性奠定了基础。在万能材料试验机上进行了摩擦系数的测试研究，在此基础上，建立了神经网络法预测摩擦系数的方法。 在进行完接触与摩擦的研究之后，从板料纯弯曲成形回弹的现有理论分析入手，利用力学理论，推导了实际弯曲回弹的方程式，对拉深成形回弹的进行了分析。由于上述理论解法中存在许多未知数，故作者采用理论解法、模拟法、实验法相结合的研究方法，提出了基于三次样条和神经网络的回弹预测方法。即将成形零件看成由诸多截面构建而成，每个截面简化成一条曲线，诸多曲线构成的曲面即为零件的曲面。每个截面曲线看成样条曲线，由样条卸载后的回弹仿真，加上利用神经网络法预测成形中摩擦、间隙、压边力、凸模圆角半径等影响的修正量来预测每个截面的回弹。则各个截面回弹后的样条曲线构成零件回弹后的形状。 进行了接触摩擦状态下和板料冲压状态下材料的屈服条件分析，得出了适用的屈服准则。作者还进行了有限元仿真研究，提出了较厚板料冲压成形有限元仿真的解决方案。 经试验验证，可以得出结论： (1) 板料具有分形特征；作者建立的接触与摩擦力学模型，并用于板料冲压成形的载荷计算，计算结果，精度比经验公式高。 (2) 神经网络法预测摩擦系数的方法可行，预测结果可用于实际工程中。 (3) 基于三次样条和神经网络的回弹预测方法的模拟计算结果能够用于实际工程中，计算速度快。更多还原

【Abstract】 Sheet metal forming is one of the important manufacturing methods in vehicle and engineering industry. In sheet metal forming,contact and friction ,springback ,are correlative a question for discussion.Friction and statejield condition,stress condition of sheet metal are sticking point.Friction and springback are key of making process planning and process parameter and die design.And the question are difficult problem in making process planning.In the dissertation, contact and friction,springback were studied detailedly.In the paper, mechanics model of contact and friction based on fractal mathematics is set up according to the theory of contact and friction. Then the model is applied to calculate the forming load.And studing object were tested by HOMMELWERKE model roughness profiler, fractal character of sheet metal was proved.It is foundation for building correctness contact and friction mechanics model. Besides , the method using neural network to predict friction coefficient is proposed,based on friction coefficient testing by universal material tester.Then start with the spring theory of pure bend in the sheet metal forming, under contact and friction is set up, springback equation on bending and drawing was derived or analysed .As a result of many unknown parameter appeared in theory solution method, writer adopted research method of combining theory solution and simulation solution with test solution, a method based on cubic spline and neural network to predict springback is also put forward in the paper, that is, suppose the forming part is composed of a good many section, and each section is predigested to be a curve, the surface which is constituted of all curves is exactly part surface. Each section curve is regarded as a spline curve, then make use of springback simulation result after unloading of the spline and some modified value such as friction , clearance, press margin power and round radius of punch to predict each section springback . In the light of the method , each section’s spline curve structure part’s shape after springback.And yield condition of material were analysed under contact and friction, punch, applicable yield condition was confirmed. At last,writer carry out finite element simulation research and put forward the solving scheme to finite simulation of thicker sheet metal forming.By experiment validating, can obtain conclusion:(1) Fractal character of sheet metal was proved. Mechanics model of contact and friction based on fractal mathematics is set up, then the model is applied to calculate the forming load. Experiment prove the calculating result using the model is right and the precision is higher than experience formula.(2) The method using neural network to predict friction coefficient is feasible, predict friction coefficient can been used to engineering.(3) Simulation result based on cubic spline and neural network to predict springback can been used to engineering, and the simulation calculation is faster.更多还原