摘要
为精确预测含能材料的5 s爆发点,解决大量新型含能材料实验测试难度大、安全数据不全等问题,基于定量构效关系(QSPR)原理,研究多硝基含能材料分子结构与5 s爆发点(ln T E)间的内在定量关系。应用集成学习算法随机森林(RF)筛选出8个对5 s爆发点具有显著影响的分子描述符;采用人工神经网络(ANN)建立90种多硝基含能材料5 s爆发点的预测模型。73种训练集的复决定系数为0.918,均方根误差为0.036,平均绝对误差为0.027。17个检验样本的复决定系数为0.903,均方根误差为0.061,平均绝对误差为0.053。对模型进行了验证以及应用域评价。结果表明:模型具备较好的预测性和泛化性能,可用于对多硝基含能材料的5 s爆发点进行精度较高的预测,有效解决现有含能材料的爆发点数据不够全面的问题,为相关产品研制与生产安全提供参考。
In order to accurately predict the 5 s explosion temperature of energetic materials and solve the problems of difficult experimental testing and incomplete safety data of a large number of new energetic materials,based on the principle of quantitative structure property relationship(QSPR),the intrinsic quantitative relationship between molecular structure of polynitro energetic materials and 5 s explosion temperature(ln T E)was studied.Ensemble learning algorithm random forest(RF)was applied to screen eight molecular descriptors that have significant impact on 5 s explosion temperature.Artificial neural networks(ANN)were used to build prediction model of 5 s explosion temperature of 90 kinds of polynitro energetic materials.The coefficient of multiple determination of 73 training sets is 0.918,the root mean square error is 0.036,and the mean absolute error is 0.027.The coefficient of multiple determination of 17 test samples is 0.903,the root mean square error is 0.061,and the mean absolute error is 0.053.The model was validated and its application domain was evaluated.Results show that the model has good predictability and generalization performance,and can be used to predict 5 s explosion temperature of polynitro energetic materials with high accuracy.It can effectively solve the problem that the explosion temperature data of existing energetic materials is not comprehensive enough,and provide reference for the development and production safety of related products.
作者
程露萍
何中其
CHENG Luping;HE Zhongqi(School of Chemistry and Chemical Engineering,Nanjing University of Science and Technology,Jiangsu Nanjing,210094)
出处
《爆破器材》
CAS
CSCD
北大核心
2023年第2期13-18,共6页
Explosive Materials
关键词
多硝基含能材料
5
s爆发点
随机森林
人工神经网络
定量构效关系
polynitro energetic material
5 s explosion temperature
random forest(RF)
artificial neural network(ANN)
quantitative structure property relationship(QSPR)
