基于显式函数和神经网络的喷气燃料混合模型的研究及应用  

Mixing Model of Jet Fuel Based on Explicit Equations and Neural Networks and Its Application

作　　者：许全宏[1,2] 刘振涛[1] 张弛[1,2] 林宇震[1,2] 

机构地区：[1]北京航空航天大学能源与动力工程学院航空发动机气动热力国家级重点试验室,北京100191 [2]北京航空航天大学先进航空发动机协同创新中心,北京100191

出　　处：《燃烧科学与技术》2017年第5期391-397,共7页Journal of Combustion Science and Technology

基　　金：国家自然科学基金资助项目(51306010);北京市自然科学基金资助项目(3152020)

摘　　要：针对喷气燃料复杂的理化性质,采取显式函数和人工神经网络相结合的策略,以氢碳比、黏度和蒸馏曲线等作为目标性质,发展了喷气燃料代用组分的构建方法,对混合燃料的理化性质具有很好的预测精度.并将此混合模型应用于煤基喷气燃料的代用组分构建,所获得的代用组分模型与真实燃料之间的理化性质差异符合模拟要求.Surrogate model of jet fuel is one of the most important premises to high-fidelity numerical simulations of spray combustion in aero-engine.For complex physico-chemical properties of jet fuel,a strategy was proposed based on a mixing model combining explicit equations and implicit neural networks.Hydrogen-carbon ratio,viscosity,and distillation curve were chosen as the target properties,and the method of jet fuel surrogate formulation was developed,which could well simulate the target properties of mixture.The mixing model was used to build the surrogate of coal-derived synthetic jet fuel,and the difference of physico-chemical properties between surrogate components and real fuel could meet the simulating demands.

关 键 词：喷气燃料 代用组分 显式函数 神经网络 混合模型 

分 类 号：V312.1[航空宇航科学与技术—航空宇航推进理论与工程]