基于CAE的薄板冲压毛坯反算技术与应用

【作者】 周军；

【导师】 李光耀； 钟志华；

【作者基本信息】 湖南大学 ， 车辆工程， 2002， 博士

【摘要】 确定合理的毛坯形状和尺寸是薄板冲压成型工艺中的一个关键，也是一个难点。合理的毛坯形状设计将使得冲压过程中材料的流动更加合理，从而明显地减少起皱、拉裂的发生，提高成型件的品质，并尽可能节省材料。 本文以有限元仿真技术为基础，提出了一种新的毛坯反算方法—有限元网格映射方法，以解决薄板冲压工艺设计中毛坯反算的一些关键问题。有限元仿真能很好地模拟材料的流动，具体体现在有限元网格中节点的移动，在不同变形时刻网格具有不变的单元节点拓扑关系和不同的节点坐标值。本方法先将与产品边界轮廓相关的目标曲线映射到板料成型终了网格上，保持目标曲线与网格之间位置关系不变，然后在板料成型初始网格上反算出毛坯形状和尺寸，即能获知初始毛坯中哪些部分参与了零件成型。通过反复几次计算，最终获得优化的毛坯形状和尺寸。本文给出了几个复杂的冲压实例，验证了本方法的正确性和快速性。 有限元网格映射方法能考虑实际冲压中的工艺余量问题，如为拉延筋、辅助平面、后面的翻边工序预留位置等。本方法不仅能获得大致的，而且通过控制毛坯面积之间的误差来获取最优的毛坯形状，尤其能考虑带孔洞的零件。在实际生产中，精确确定毛坯形状和尺寸可以将落料、冲孔工序合二为一，这将大大地降低生产成本，提高生产效率。 为了方便在实际生产中的使用，本文采用神经网络来快速获得毛坯的形状和尺寸，并提出了正交和均匀设计相结合的方法来确定样本。确定样本的新方法不仅能保证选取的样本具有取点均衡、整齐可比和在整个样本空间内均匀散布等特点，而且提高了神经网络的预测精度，并大大降低了试验次数。 本文对毛坯形状预测“傻瓜”系统进行了初步探讨，为汽车覆盖件冲压成型工艺设计参数预测系统的创建提供了有力的支持，并给出了具体的实施方案。更多还原

【Abstract】 The determination of the optimum blank is one of the key and difficult problems in sheet metal forming. Reasonable design of the blank will decrease the possibility of wrinkle and fracture, and increase the formability obviously.A Mesh Mapping Method based on the FEM results has been proposed to predict the optimum blank in sheet metal forming in this thesis. The method assumed that the adaptive technique was not been used during the finite element simulation. The basic idea of the presented method is to map the target contour to the blank wire at final configuration after the punch was released completely and the optimal blank can be obtained according to the blank wire at the initial configuration. After the iteration loops subjected to the shape error constant, the size and shape of the blank can be determinated very accurately as long as the FEM results were accurate enough. Several examples are applied to confirm the validity, effectivity and versatility of the presented method.It is possible for the presented method to consider the problems about the technique allowance, for example, for drawbed, aided plane and crimp. Not only the rough profile but also the exact optimum blank can be obtained by controlling of the shape error constant. Especially, the presented method can deal with the parts with the holes. Due to specifying the exact optimum blank, some subsequent techniques will be taken out and the production cost will been decreased.The blank prediction system of U-shape has been developed using the Radial Basis Function Network to get the optimal blank rapidly for practical purpose. The Orthogonal table mixing with the Uniform table was proposed to specify the specimens for neural network, which improved the precision of prediction and decreased the test numbers.The big ’fool’ prediction system has been introduced and some viable methods have been given. The system will include the prediction of the blank shape, and some other design parameters in automotive sheet metal forming.更多还原