盒形件毛坯外形预测中BP网络结构的确定

Determination of BP Neural Network Architecture in Predicting the Blank Shape of Square Cap

采用神经网络技术预测拉深件毛坯外形的方法中，网络的结构对预测结果的优劣有很大的影响。提出了ＢＰ网络的网络参数，并结合训练样本的数目给出了网络参数和样本数目之间的关系。应用本文提出的方法设计了毛坯外形预测的ＢＰ神经网络结构。

Architecture of a network affects the calculation results very much in determination of ideal blank shape employing ANN technology.Several network parameters are determined and the relation between network parameters and the number of training pairs is described.A BP network is designed for predicting the blank shape of a square box using the method presented in this paper.