深層学習による超臨界非理想熱・輸送物性モデルの計算高速化

抄録

Fluids under supercritical conditions exhibit unique physical properties that differ from those under subcritical conditions or standard temperature and pressure conditions. For this reason, when performing numerical simulations of fluids under supercritical conditions, the fluid cannot be treated as an ideal gas, and it is necessary to consider the non-ideal property of the fluid. However, if a simulation is performed considering the non-ideal property of the fluid, more computation time is required compared to the case where the fluid is treated as an ideal gas. To reduce the computational cost if non-ideal properties, we developed a mathematical model that can estimate fluid properties considering non-idealities using deep learning, and by replacing it with a conventional method, we confirmed that the calculation speed can be accelerated while maintaining the calculation accuracy of the conventional method in a multi-component flow field under supercritical conditions.