CAJ | 학술논문

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.
운용신경망락편최소이승분별여유전산법화주성분분석상결합,이함능재료적결구묘술부화폭굉성능등삼수,건립료"분자결구-폭굉성능"지간적정량관계예측모형,병대30충함능재료적밀도화이론폭속진행료예측,기상대오차균재5%이하.표명저충방법위신형함능재료분자설계화폭굉성능예고제공료신적방법화수단.
The methods have been developed for model construction of quantitative structure-detonation relationship by combining netural network partial least squares with genetic algorithms and principal component analysis respectively. The model can predict the density and detonation velocity of 30 explosives, all the relative errors are less than 5%. The result shows that it offers novel method to design molecules and estimate the detonation relationship performance for new energetic materials.