The Root Cause of the Numerical Overheating Problem

摘要

Previously we identified the receding flow, where two fluid streams recede from each other, as an open numerical problem, because all well-known numerical fluxes give an anomalous temperature rise, thus called the overheating problem. This phenomenon, although presented in several textbooks, and many previous publications, has scarcely been satisfactorily addressed and the root cause of the overheating problem not well understood. We found that this temperature rise was solely connected to entropy rise and proposed to use the method of characteristics to eradicate the problem. However, the root cause of the entropy production was still unclear. In the present study, we identify the cause of this problem: the entropy rise is rooted in the pressure flux in a finite volume formulation and is implanted at the first time step. It is found theoretically inevitable for all existing numerical flux schemes used in the finite volume setting, as confirmed by numerical tests. This difficulty cannot be eliminated by manipulating time step, grid size, spatial accuracy, etc, although the rate of overheating depends on the the flux scheme used. Finally, by incorporate the entropy transport equation, in place of the energy equation, to ensure preservation of entropy, the temperature anomaly is corrected, thus also numerically confirming the theoretical analysis for the root cause. Its applicability is demonstrated for some relevant ID and 2D problems. In conclusion, the present study finds that the entropy generated ab initio is the genesis of the overheating problem.