本文将神经网络模式识别用于金属间化合物3元填隙d88结构形成条件的判别;神经网络由3层组成,训练用后向传播算法。为了评价所得模型的行为,使用了交叉验证法。计算结果表明,当适当的键参数作为输入时,71种化合物能被正确地分类,且隐层节点数经优选后可以提高分类的准确率,减少计算时间。 In this paper, neural network pattern recognition is used to determine the formation conditions of intermetallic compound 3-element interstitial d88 structures. The neural network consists of 3 layers and is trained with backward propagation algorithm. In order to evaluate the behavior of the resulting model, a cross-validation method was used. The results show that when the appropriate key parameters are used as inputs, 71 kinds of compounds can be correctly classified, and the number of hidden layer nodes can be improved to improve the classification accuracy and reduce the calculation time.