运用神经网络偏最小二乘分别与遗传算法和主成分分析相结合,以含能材料的结构描述符和爆轰性能等参数,建立了“分子结构-爆轰性能”之间的定量关系预测模型,并对30种含能材料的密度和理论爆速进行了预测,其相对误差均在5%以下。表明这种方法为新型含能材料分子设计和爆轰性能预估提供了新的方法和手段。 By using neural network partial least squares combined with genetic algorithm and principal component analysis, the quantitative relationship between “molecular structure - detonation performance ” was established by parameters such as structure descriptors and detonation performance of energetic materials The model was predicted and the densities and theoretical detonation velocities of 30 kinds of energetic materials were predicted. The relative errors were below 5%. It shows that this method provides a new method and method for the molecular design and detonation performance prediction of new energetic materials.