The 9th International Conference on Matter and Radiation at Extremes (ICMRE2026) ( ICMRE2026)

Aiming at the accurate and efficient calculation of carbon dioxide non-equilibrium radiation under extreme conditions such as Mars entry and high-altitude expansion flow, this paper proposes a fast prediction model of physical parameters based on BP neural network. Considering the vibration-rotation non-equilibrium and vibration non-equilibrium at the same time, the three-temperature sample data of carbon dioxide molecules under the ground state electron transition are generated by the line-by-line calculation model. The accuracy of the model was tested in the radiation transmission of uniform and non-uniform media, and the maximum relative errors were 8.54 % and 4.54 %, respectively, compared with the reference data. In addition, the model is 40 times faster than LBL, occupies only 65 MB of memory, and has good prediction ability when degraded to thermodynamic equilibrium, which has certain application value in the calculation of radiation transmission under extreme conditions.

Neural network,Physical property parameters,Line-by-line,thermal non-equilibrium