基于BP神经网络的合金铸铁腐蚀深度预测  

Corrosion Depth Prediction of Alloy Cast Iron Based on BP Neural Network

作　　者：王玉荣[1] 乌日根[1] 

机构地区：[1]包头职业技术学院,包头014030

出　　处：《腐蚀与防护》2011年第12期962-964,1012,共4页Corrosion & Protection

摘　　要：通过动态质量损失法腐蚀试验获取BP神经网络的样本数据。利用Matlab的工具箱函数建立了拓扑结构为4×15×8×1的BP神经网络,并对网络模型的预测精度和应用进行了研究。结果表明,在样本集和训练条件下,4×15×8×1型BP网络能较好地反映腐蚀时间、合金铸铁主要合金成分与腐蚀深度之间的非线性关系。可用于合金铸铁在高温浓碱液中的动态腐蚀性能的预测;当稀土和铜质量分数较低且适量时,其耐碱蚀作用较显著,而镍质量分数越高耐碱蚀作用越明显。The sample data of BP neural network were measured by the dynamic hydrometer method. The 4 × 15 × 8× 1 BP neural network model was established by the toolbox function of Matlab, and the prediction precision and application of network model were studied. The results showed that under this sample set and training condition, 4 × 15 × 8× 1 BP neural network model reflected the non-linear relationship between corrosion time and main components of alloy cast iron and corrosion depth very well, and it was used to predict dynamic corrosive nature of alloy cast iron in high temperature concentrated alkaline solution. When rare earth and copper contents were relatively low and proper, the caustic corrosion resistance function of rare earth and copper was comparatively obvious, the higher the nickel content, the obvious the caustic corrosion resistance function.

关 键 词：BP网络 合金铸铁 腐蚀深度 耐碱蚀 预测 

分 类 号：TG174.2[金属学及工艺—金属表面处理] TG143.9[金属学及工艺—金属学]