为了探索新的含能材料撞击感度预测方法，分类研究了若干硝基类含能材料，发现关于对称芳香族类硝基含能材料的部分NQ/Y类描述符与lnh50呈良好线性关系。结合若干与lnh50线性相关性较高的NQ/M (M表示摩尔质量)类描述符，对35个对称芳香族硝基类含能材料线性回归建立模型1，相关系数为0.952；并以8个对称芳香族硝胺类含能材料作为训练样本建立模型2，相关系数为0.991。对模型1检验对比表明，模型1具有较高的内部稳定性及外部预测性。而模型2将特性落高(h50)与结构描述符直接关联，能够克服对数模型引起误差放大的缺点。%In order to explore new prediction methods for impact sensitivity, a number of nitro energetic materials are classified and researched. Results indicate that the sectionsNQ/Y class descriptor of symmetrical aromatic nitro energetic materials are in a good linear relationship with lnh50. The model 1 established by linear regression method for 35 symmetrical aromatic nitro energetic materials has a correlation coefficient of 0.952, by combination of someNQ/M class descriptors which have higher linear correlation with lnh50. And the model 2 has a correction coefficient of 0.991 which takes eight symmetrical aromatic nitramine energetic materials as training samples. The comparison of model tests indicates that the model 1 has a high internal stability and external prediction. The model 2 established the relationship betweenh50 and descriptors, which could overcome the shortage of error amplifier caused by logarithmic models.
展开▼
