基于RBF神经网络的稀土碱性焊条焊缝耐蚀性预测

Prediction of corrosion resistance of rare earth alkaline electrode weld based on RBF neural network

通过正交试验设计9组稀土氧化物碱性焊条的药粉配方,并获取焊缝腐蚀深度的样本数据,利用Matlab的工具箱函数Newrb建立RBF神经网络预测模型,研究了稀土氧化物对焊缝耐碱蚀性能的影响和效应作用。结果表明, RBF神经网络较好地反映稀土氧化物与焊缝腐蚀深度之间的非线性关系,预测精度较高,可用于焊缝耐碱蚀性能的有效预测;适量稀土氧化物有利于提高焊缝耐碱蚀性能,其效应曲线具有抛物线特征;当添加w(La2O3)1%, w(CeO2)1%和w(Y2O3)1%焊条的焊缝腐蚀深度预测值达到最小,其腐蚀深度为0.000 8 mm/a。

The nine group of rare earth oxide alkaline electrode coating formulas were designed by the orthogonal test, and the sample data of weld corrosion depth were measured. The RBF neural network prediction model was established by the toolbox function Newrb of Matlab, the effect and role of rare earth oxide on caustic corrosion resistance of weld were studied. The results showed that RBF neural network model reflected the non-linear relationship between rare earth oxide and corrosion depth of weld, and high prediction precision of RBF neural network, and it was used to prediction of weld caustic corrosion resistance. Proper amount rare earth oxide were advantageous to the improvement caustic corrosion resistance of weld. The effector curve had the feature of parabola.Weld corrosion depth predicted value of electrode with w(La_2 O_3)1%, w(CeO_2)1%, w(Y_2 O_3)1% was reached least value, the corrosion depth was 0.000 8 mm/a.