将铸造生产中缺陷产生的主要原因抽象为神经元,构建了灰铁铸件缺陷诊断的BP神经网络模型.采用非线线性变换,确定了样本的变化范围,用遗传算法学习网络模型的权值,开发出了基于C++的铸件缺陷诊断及防止的应用软件,并通过建立铸件缺陷数据库,以克服专家系统的“瓶颈”问题,从而有效地降低铸件的废品率. The main reason of defects in foundry production was abstracted as neurons and a BP neural network model was built to diagnose the defects of gray iron castings.The range of samples was determined by nonlinear linear transformation.The genetic algorithm was used to study the weights of network models, The application software of diagnosis and prevention of casting defects based on C ++ is developed. By establishing the database of casting defects, the “bottleneck” problem of the expert system can be overcome, which can effectively reduce the rejection rate of castings.