文章编号:1000-8055(2023)04-0806-10doi:10.13224/j.cnki.jasp.20220836基于人工神经网络模型的超临界RP-3热物性计算陶凯航1,2,朱剑琴1,2,程泽源1,2(1.北京航空航天大学能源与动力工程学院,北京100191;2.北京航空航天大学航空发动机气动热力国家级重点实验室,北京100191)摘要:为准确得到超临界压力下RP-3的热物性,基于人工神经网络(ANN)方法建立超临界RP-3的密度、黏度、比定压热容和导热系数的计算模型。以广义对应态法则计算得到的RP-3热物性结果训练神经网络,并耦合了实验误差模型得到修正后的ANN模型。计算温度变化范围为300~800K,压力变化范围为3~6MPa。结果表明:ANN模型能准确地预测超临界RP-3的热物性,且计算精度比广义对应态法则计算得到的结果提高了16.3%。在压力为5MPa的工况下,ANN模型预测的密度、黏度、比定压热容和导热系数的回归系数均大于0.99,与实验结果平均相对误差分别为1.5%、4.1%、0.9%和0.7%。关键词:超临界;航空煤油RP-3;广义对应态法则;人工神经网络模型;热物性中图分类号:V231文献标志码:ACalculationofthermophysicalpropertiesofsupercriticalRP-3basedonartificialneuralnetworkmodelTAOKaihang1,2,ZHUJianqin1,2,CHENGZeyuan1,2(1.SchoolofEnergyandPowerEngineering,BeihangUniversity,Beijing100191,China;2.NationalKeyLaboratoryofScienceandTechnologyonAeroEngineAero-thermodynamics,BeihangUniversity,Beijing100191,China)Abstract:InordertoaccuratelyobtainthethermophysicalpropertiesofRP-3undersupercriticalpressure,thecalculationmodelsofdensity,viscosity,specificheatcapacityatconstantpressureandthermalconductivityofsupercriticalRP-3wereestablishedbasedonartificialneuralnetwork(ANN)method.TheRP-3thermophysicalpropertiesobtainedbytheextendedcorrespondingstatewereusedtotraintheneuralnetwork,andthemodifiedANNmodelwasobtainedbycouplingtheexperimentalerrormodel.Thecalculatedtemperaturerangewas300−800K,andthepressurerangewas3−6MPa.TheresultsshowedthattheANNmodelcanaccuratelypredictthethermophysicalpropertiesofsupercriticalRP-3,andthecalculationaccuracywas16.3%higherthanthatoftheextendedcorrespondingstate.Atthepressureof5MPa,theregressioncoefficientsofdensity,viscosity,specificheatcapacity...