QSPR方法预测硝胺化合物撞击感度  

Prediction of impact sensitivity of nitramine with QSPR method

作　　者：王睿[1] 蒋军成[1] 潘勇[1] 曹洪印[1] 崔毅[1] 

机构地区：[1]南京工业大学安全工程研究所,江苏省城市与工业安全重点实验室,南京210009

出　　处：《固体火箭技术》2009年第6期667-672,共6页Journal of Solid Rocket Technology

基　　金：国家自然科学基金项目(20976081);高等学校博士学科点专项科研基金项目(200802910007);江苏省自然科学基金项目(BK2009360)

摘　　要：基于定量结构-性质相关性(QSPR)原理,研究了硝胺化合物撞击感度与其分子结构间的内在定量关系。根据分子结构计算用于反映分子结构信息的35个结构参数;应用遗传算法筛选出与撞击感度密切相关的5个参数作为分子描述符;采用多元线性回归方法对分子描述符与撞击感度之间的内在定量关系进行模拟,建立了根据分子结构预测硝胺化合物撞击感度的数学模型。分别采用内部验证及外部验证的方式对模型性能进行了验证。结果表明,模型具有较高的稳定性、预测能力及泛化性能。该方法的提出为工程上预测硝胺化合物撞击感度提供了一种新途径。A quantitative structure-property relationship （QSPR） study was used for prediction of impact sensitivity of nitramine. A set of 35 molecular descriptors were calculated to represent the molecular structures of nitramine. 5 descriptors which have significant contribution to the impact sensitivity were chosen by employing the variable selection method of genetic algorithm（GA）. The multiple linear regression（MLR） was then employed to simulate the possible quantitative relationship between these selected descriptors and impact sensitivity. The model was tested by internal and external validations, and the results show good robustness and high prediction ability as well as generalization. The method presented provides a new way for predicting the impact sensitivity of nitramine for engineering.

关 键 词：硝胺化合物 撞击感度 定量结构-性质相关(QSPR) 遗传算法 

分 类 号：TQ246.1[化学工程—有机化工] O189[理学—数学]