煤基喷气燃料代用组分神经网络混合构建方法

为了建立航空燃料的喷雾模型，用于高保真液雾燃烧数值模拟，提出了基于人工神经网络混合模型的煤基喷气燃料代用组分构建方法.基于这一构建方法，重点针对煤基喷气燃料的雾化特性，利用多组分混合燃料的理化性质数据库对神经网络进行训练，获得了混合燃料理化性质隐式预测模型，结合随机投点优化方法，构建出能够很好地模拟煤基喷气燃料目标理化性质的代用组分.结果表明:该代用组分包含了5种碳氢化合物成分，摩尔分数为11.46%正癸烷、23.29%正十二烷、49.87%正十四烷、6.66%异辛烷和8.72%甲基环己烷.通过雾化特性实验，验证了代用组分对真实燃料雾化性能的模拟效果.该代用组分构建方法可以较好地解决混合燃料模拟过程中的非线性问题，通过改变目标理化性质可构建出相应代用组分. 

Surrogate formulation methodology of coal-based jet fuel based on neural network mixing model

In order to build spray model of aviation fuel for the high-fidelity numerical simulation of spray combustion, a surrogate formulation methodology was proposed for coal-based jet fuel based on artificial neural network mixture model. An implicit prediction model was developed on the blended physico-chemical properties using the multi-component fuel properties data set to train the neural network. And then the surrogate of coal-based jet fuel was formulated from the neural network mixing model by the stochastic points'' optimization method, which could well simulate the target physico-chemical properties focusing on its atomization.Result shows that the surrogate is composed of 5 hydrocarbons(n-decane,n-dodecane,n-tetradecane,iso-octane and methylcyclohexane),their mole fraction are 11.46%, 23.29%,49.87%,6.66% and 8.72%, respectively. Compared with the real fuel, the atomization simulation of the surrogate was evaluated by experiments. This surrogate formulation methodology can solve the nonlinear issue in the mixing process, and formulate different surrogates for various requirements.