通过动态质量损失腐蚀试验获取样本数据，利用Matlab的工具箱函数建立了合金铸铁碱腐蚀速率的RBF神经网络预测模型，并对网络模型的预测精度进行了研究。结果表明，在样本集和训练条件下，RBF神经网络模型较好地反映出腐蚀时间、合金铸铁主要合金成分与腐蚀速率之间的非线性关系，可用于合金铸铁在高温浓碱液中的动态腐蚀性能的预测; 当RBF网络的扩展系数为0.47时，动态腐蚀速率的试验值与网络预测值之间的误差最小，且耐蚀性评价准确率达到100%。

The sample data were measured by the dynamic mass loss method. The RBF neural network prediction model of alloy cast iron corrosion rate was established by the toolbox function of Matlab, and the prediction precision of network model was studied. The results show that under this sample set and training condition, RBF neural network model reflected the non-linear relationship between corrosion time and main components of alloy cast iron and corrosion rate, and it was used to predict the dynamic corrosive nature of alloy cast iron in high temperature concentrated alkaline solution. When the spread coefficient of RBF neural network was 0.47, the error between measured values of dynamic corrosion rate and predicted values of network was minimum, and the appraisal accuracy rate of corrosion resistance reached 100%.