神经网络法研究硝基含能化合物的撞击感度

Research on the Impact Sensitivity of Nitro Energetic Compounds by Neural Network

为了研究硝基含能化合物的撞击感度与分子结构之间的关系,基于邻接矩阵和空间结构,计算了36个硝基含能化合物的分子连接性指数X和电性拓扑状态指数E,建立了这些分子的撞击感度(lg H50)与1X、2X、4X、5X、E13、E16和E28共7种分子结构参数的定量结构-性质相关性(QSPR)模型。将这7种分子结构参数作为神经网络法的输入神经元,采用7∶2∶1的神经网络结构,建立了预测精度较好的神经网络模型,总相关系数为0.975 5。利用该模型计算得到的撞击感度预测值与其实验值的相对平均误差为4.03%,优于多元回归模型的相对平均误差(9.57%)。预测值与实验值较吻合,表明硝基含能化合物的撞击感度与7种分子结构参数具有非线性关系。

In order to study the correlativity between the impact sensitivity and the molecular structure of nitro energetic compounds,the molecular connectivity index Xand electrotopological state index Eof 36 nitro energetic compounds were calculated based on conjugation matrix and molecular space structure.Thus,we developed the QSPR between the impact sensitivity(lg H50)and seven molecular structural parameters(1X,2X,4X,5X,E13,E16 and E28)of 36 nitro energetic compounds.Using the seven molecular structural parameters as input neurons of the neural network,a neural network model with better prediction accuracy was established by neural network method whose network structure was 7∶2∶1.The total correlation coefficient is 0.975 5.The average relative error between the experimental value and the predicted value(lg H50)is 4.03%,which is better than that of the multivariate regression model(9.57%).The predicted values are in good agreement with the experimental ones.The results show that there is a nonlinear relationship between the impact sensitivity and the seven molecular structural parameters.