Thermodynamics-Informed Input Reparameterization for Neural Prediction of Real-Fluid Thermodynamic Properties in Supercritical Combustion
The work introduces Target-Aligned Input Reparameterization (TAIR) to improve neural network predictions of real-fluid thermodynamic properties in supercritical combustion. TAIR replaces raw enthalpy with target-matched coordinates derived from ideal gas approximations, reducing computational cost and significantly improving accuracy across temperature, density, and compressibility predictions.