Thermodynamics-Informed Input Reparameterization for Neural Prediction of Real-Fluid Thermodynamic Properties in Supercritical Combustion

2026-07-21

AI Summary

This work addresses the high computational cost of accurately evaluating real-fluid thermodynamic properties in supercritical combustion simulations and the challenges of directly regressing temperature, density, and compressibility factor from enthalpy, pressure, and composition using neural networks due to complex nonlinearities. To overcome this, the authors propose Target-Aligned Input Reparameterization (TAIR), which leverages the ideal-gas assumption to convert enthalpy into initial estimates of temperature or density as network inputs. This physically aligns the inputs with the prediction targets, guiding the model to learn deviations of real-fluid behavior from the ideal-gas baseline. TAIR requires only variables already available in solvers and fundamental property constants, introducing no additional computational overhead. Evaluated on methane–oxygen supercritical flame data, TAIR significantly improves extrapolation performance, reducing RMSE for temperature, density, and compressibility factor by 1.5×, 2.0×, and 7.5×, respectively, compared to baseline methods.

Abstract

Real-fluid thermodynamic property evaluation is a major computational cost in supercritical combustion simulations. In the enthalpy-based pressure-correction formulation, the closure evaluates temperature T, density $ρ$, and compressibility coefficient $ψ$ from the solver state (h,p,Y) through enthalpy-temperature inversion and repeated real-fluid equation-of-state evaluations. Neural-network surrogates offer fixed-cost inference, but direct mapping from (h,p,Y) to $(T,ρ,ψ)$ must capture the enthalpy-temperature relation and non-ideal equation-of-state response, resulting in a complex regression problem. This work introduces a thermodynamics-informed input reparameterization strategy, termed target-aligned input reparameterization (TAIR). TAIR replaces the raw enthalpy coordinate of each property network with a target-matched thermodynamic coordinate: the temperature network uses a temperature estimate obtained by inverting a constant-$c_p$ ideal-gas mixture enthalpy approximation, whereas the density and compressibility networks use an ideal-gas density estimate. These algebraic transformations use only solver-available variables and species constants, guiding the networks to learn real-fluid departures from ideal-gas baselines rather than reconstructing the full closure from raw enthalpy. The method is assessed using supercritical methane-oxygen counterflow flame data against a raw-input baseline and target-inconsistent cross-reparameterization controls. TAIR reduces held-out RMSE by factors of about 1.5, 2.0, and 7.5 for T, $ρ$, and $ψ$, respectively. For an unseen strain-rate flame within the augmented thermodynamic envelope, the corresponding factors are 3.6, 14.5, and 6.0. The target-inconsistent controls perform worse, indicating that the gains arise from thermodynamically matched input design rather than generic preprocessing.

Problem

• real-fluid thermodynamics
• supercritical combustion
• neural surrogate modeling
• enthalpy-temperature inversion
• equation of state

Innovation

• thermodynamics-informed reparameterization
• neural surrogate modeling
• real-fluid thermodynamics
• supercritical combustion
• target-aligned input