论文部分内容阅读
为了研究AB5型储氢合金初始放电容量与合金成分间的关系,设计了径向基函数型人工神经网络模型。用“留一法”训练了模型,然后用训练好的神经网络模型预测了5个样本的初始放电容量,预测值和实验值在散点图中沿45°线分布,统计学指标为:均方误差(MSE)为6.063,相对均方误差(MSRE)为0.0262%,拟合分值(VOF)为1.934 5,说明人工神经网络预测的结果是准确、可靠的。最后用神经网络分析了AB5型储氢合金的合金成分对其初始放电容量的定量影响,结果表明:La、Nd含量对初始放电容量影响呈抛物线关系,初始放电容量存在一个极小值;Ce含量对初始放电容量影响较大,随Ce含量的增加而增大,且增幅较大;Pr含量的影响不大,随Pr含量的增加初始放电容量有小幅增大,最后趋于平稳。
In order to study the relationship between initial discharge capacity and alloy composition of AB5 hydrogen storage alloy, a radial basis function artificial neural network model was designed. The model was trained with “leaving one method ”, and then the initial discharge capacity of five samples was predicted with the trained neural network model. The predicted value and the experimental value were distributed along the 45 degree line in the scatter plot with the statistical index of The mean square error (MSE) was 6.063, the relative mean square error (MSRE) was 0.0262%, and the fitting score (VOF) was 1.934 5, indicating that the artificial neural network prediction is accurate and reliable. The results show that the influence of La and Nd content on the initial discharge capacity is parabolic, and the initial discharge capacity has a minimum value. The Ce content The initial discharge capacity has a greater impact, with the increase of Ce content increases, and a larger increase; Pr content has little effect, with the Pr content of the initial discharge capacity increased slightly, and finally tends to be stable.