I am reading ever more often papers in high impact factor journals containing transient simulations without a proper assessment of the phase error due to time discretisation. In several of these I had good reasons to believe that the effect of these errors was enough to lead to wrong results.
Why aren't phase errors properly assessed? And, more importantly, why don't reviewers and editors make sure that their policies on numerical accuracy (which contain specific guidelines for transient simulations) are followed?
Dear Tiziano,
this is a really interesting note. Could you provide some examples? I recently worked on a steady-state model in which I considered the latent heat absorption and release. In this case, I just needed to be aware of the space discretisation. Later on, I used my results as a heat source in a transient calculation, so that the latent heat was already accounted for. I wrote about the advantage of this approach, be means of independence on the time step size.
Here some methods of heat conductions with change of phase are well presented:
H.T. Hashemi, C.M. Sliepcevich, A numerical method for solving two- dimensional problems of heat conduction with change of phase, in: Chemical Engineering Progress Symposium Series, vol. 63(79), 1967, pp. 34–41.
Best regards,
Antoni Artinov
Hi Antoni,
do you want me to cite articles that introduced a phase error due to an insufficient temporal discretisation?
Hi Tiziano,
I think Antoni was asking you just some suggestions about fluid dynamics problems where a not correct time step used in the CFD solution can give wrong results. I provide one example: the case of a pitching airfoil, at low reduced frequency and angles below static stall. In these condition, without a proper time step, CFD can produce a phase error between upstroke and downstroke phases, generating a false hysteresis.
Regards,
Francesco
- Badges
- Science topic