镍基高温合金中γ’和γ相点阵常数的简捷人工神经网络预测法

A SIMPLE AND ACCURATE NEURAL NETWORK PREDICTION METHOD FOR ESTIMATING LATTICE PARAMETERS OF γ' AND γ PHASES IN NICKEL BASE SUPERALLOYS

介绍了一种测算镍基高温合金在不同温度下γ’和γ相点阵常数的简捷人工神经网络法.基于对多种镍基高温合金在不同温度下两相成分及其点阵常数数据归一化处理的反向传播神经网络模拟,得到了可以比较准确表达两相点阵常数与相成分、温度间复杂非线性关系的权值矩阵,由此可对镍基合金在不同温度下γ’和γ相的点阵常数进行测算,并获得与X射线衍射法测定数据甚为接近的点阵常数数据,证实了该方法的可行性及准确性.

A neural network method is proposed to quickly and precisely predict lattice parameters of γ' and γ phases at different temperatures for nickel-base superalloys. Algorithm weight matrixes accurately express the complex non-linear relationships of lattice parameters of the two phases with temperatures and phase compositions are obtained based on the back-propagation neural network modeling on the practical values of temperature, phase composition and lattice parameter in various superalloys. The feasibility and accuracy of the proposed prediction method are verified through the comparison of the prediced and measured lattice parameter values.