人工神经网络在覆盖拉延件要领设计中的应用

人工神经网络在覆盖拉延件要领设计中的应用

中国机械工程

CHINA MECHANICAL ENGINEERING

1999年 第19卷第1期No.1 voel.191999

人工神经网络在覆盖拉延件要领设计中的应用*

高凯祁　胡世光

摘要　结合覆盖拉延件的特点，分析了在工艺要领设计中应用人工神经网络的必要性和可能性，同时对隐层节点数、网络权重总数及样本总数之间的关系提出了一些看法。

关键词　人工神经网络　汽车覆盖件　计算机辅助工艺设计　拉延件

中国图书资料分类法分类号　TP18　TP386

The Application of Artificial Neural Network in Technical Essential Design for Stretching Part of Automobile Panels　

　Gao Kaiqi

Hu Shiguang　

(Beijing University of Aeronautics and Astronautics,Beijing,China)　

Abstract:Based on the characteristics of technical essential design(TED) for automobile stretching parts, the necessity and possibility of applying artificial neural network (ANN) to TED are analyzed. It is pointed out that length of flown-in material(LFM), which flow into female die from underside of blank-holder during forming, is external performance of local forming resistance. ANNs are founded by actual data: from the height of automobile panel wall to LFM ; from LFM to the geometric parameters of drawbead and die corner. Moreover, there are two viewpoints about applying ANN to engineering practice to be presented: a) the number of whole weights should not be more than the number of training samples; b) the hidden node number should be between the number of input and output nodes.

Key words:ANN automobile panel CAPP stretching part technical essential design(TED)

近年来，人工神经网络的研究与应用十分活跃。在机械制造领域，BP网络是应用最广泛、最成熟的网络，其实质是实现由n维输入空间到m维输出空间的映射。与传统的最小二乘及其它回归方法相比，神经网络不存在选择基函数、求解系数的困难，因而能有效地逼近高维的复杂函数。

在覆盖拉延件要领设计中［1］，工艺专家的设计经验往往是定性的、模糊的，很难通过交谈、问卷调查等形式进行系统的提取，尤其是难以量化，因此制约了工艺设file:///E|/qk/zgjxgc/zgjx99/zgjx9901/990117.htm（第 1／6 页）2010-3-23 15:21:49