基于ANN的冲裁合理间隙的预测研究

冲裁间隙是冲裁工艺中非常重要的工艺参数,其大小会影响冲裁件的表面质量、板料的后续成形和使用性能。不同材料,不同厚度板料的合理冲裁间隙均不同,现有方法是通过实验或查询已有的冲裁实验数据库获得。本课题应用人工神经网络技术和航空材料冲裁间隙数据库,建立板料厚度及冲裁零件的尺寸精度、光亮带比、毛刺高度与合理间隙之间非线性映射的神经网络模型,可以方便地获得不同厚度材料的冲裁合理间隙,用以指导实际生产。本文采用贝叶斯规则化的训练方法,训练好的BP网络较常用的训练方法具有更好的精度和泛化能力。预测数据和实验数据的比较证实,预测模型有效并具有实际推广应用价值。

Prediction of Optimum Blanking Clearance Based on Aritficial Neural Networks

Blanking clearance is an important parameter in blanking processes. It may influence the quality of a blanked surface, subsequent formation and use. The optimum blanking clearances are different for different materials of differeat thickness. The current method for obtaining the clearances is to do experi- ments or search for the existing blanking experimental databases. In this paper artificial neural network technology and the blanking clearance database for aviation materials are used to establish the nerual network model of the nonlinear mapping among thickness, dimensional precision, burr height, ratio and optimum blanking clearance. Thus it is easy to get the blanking clearances of the materials of different thick hess through this model. The paper used the Bayes regularization algorithm to train the BP network, the precision and generalization of which are better than the network that uses ordinary training algorithms. A comparison of prediction data with experimeatal data proves that the prediction model is effective and has extension and application values.