投影寻踪和神经网络方法研究金属离子水解规律

Study of Hydrolysis Rule of Metal Ions by Projection Pursuit and Artificial Neural Network

对金属离子水解常数pK1 的线性模型的局限性进行了讨论,并采用投影寻踪技术对pK1数据结构进行研究。非线性建模可能是更适宜的方法。采用函数连接型神经网络(FLN) ,以金属离子的电荷、半径、价电子结构、电负性及空价轨道数等作为描述变量,对60种金属离子水解常数pK1 数据进行建模,获得了满意的结果,并对1 0种金属离子的pK1 作出了预测。

Projection Pursuit (PP) was employed in studying the relationships between structural parameters of metal ions and their hydrolysis rule. It showed that non-linear model was more suitable than linear model to study the hydrolysis effect of metal ions. A functional-link net (FLN) was used to seek the accurate quantitative relationship of 60 metal ions by 10 descriptors, such as radius, electric charge of metal ions etc. The correlation coefficient(R) was 0 9933 By the results, hydrolysis constant of ten kinds of metal ions were predicted. These obtained results were acceptable and explicable.