用人工神经网络法预估高氮化合物的生成焓

摘要

With the molecular structure describers, the enthalpy of formation for high nitrogen compounds are predicted by using an artificial neural network (ANN) approach.The influences of neural network parameters and molecular structure describers (MSD) on the standard enthalpy of formation are studied.It was found that the calculated enthalpy of formation agrees well with the lilerature ones with the correlation coefficient of 0.99823.The relative errors are around 10％.The ANN approach is an effective and simple method for predicting the enthalpy of formation of high nitrogen compounds.%采用误差反向传播学习(BP)人工神经网络算法,以分子结构中不同基团作为描述码,对高氮化合物的标准生成焓进行预估,研究了网络参数及分子结构描述码对标准生成焓的影响,计算结果与文献值符合得较好,其回归方程相关系数为0.99823,相对误差在10%左右.人工神经网络(ANN)法是一种简单有效的预测高氮化合物生成焓的方法.