Development of Surrogate Models for Thermophysical Properties of Petroleum-Based Rocket Kerosene Using BP-ANN Algorithm

Abstract

Simplified surrogate models have gained significant attention for effectively reproducing thermophysical properties of chemically complex rocket kerosene, which is widely used in liquid rocket engines. We used the Helmholtz-type equation of state and the extended corresponding state model to calculate the thermophysical properties of the surrogate model. To optimize the composition ratio of the candidate components, we employed the BP artificial neural network algorithm. As a result, we established four surrogate models, namely the 3-species component, 4-species component, 7-species component, and 9 species component models, which can effectively represent the thermophysical properties of petroleum-based rocket kerosene. The H/C ratio, molecular weight, density, isobaric heat capacity, viscosity, and thermal conductivity were selected as the performance indexes of the surrogate model. A test system designed for measuring the thermodynamic and thermal transport properties of rocket kerosene was used to test and report, for the first time, the thermophysical properties of petroleum-based rocket kerosene. This study was the first to obtain four thermophysical properties data of petroleum-based rocket kerosene at temperatures and pressures ranging from 298 K to 500 K and 1 MPa to 30 MPa, respectively. Upon averaging the average deviations of the four thermophysical properties for the four models under all test conditions, the data analysis revealed that the C9 model, which consisted of nine species, was the most suitable choice for calculating thermophysical properties. The average deviation in the thermodynamic properties between the C9 model and petroleum-based kerosene was 2.53 %.