AdvectionReaction Solver in recent versions of Elmer : Initialisation of the advected variable in transient
Posted: 22 Feb 2019, 12:29
Dear Elmer users,
I notice what seems to be a bug with the AdvectionReaction solver with Disconstinuous Galerkin variables.
This bug is illustrated in a very simple TestCase that is available upon request (too big to be attached to this post). The TestCase consists in advecting vertically a gaussian distribution of a given physical quantity at a constant convection speed. It turns out that when I run this test with the most recent version of Elmer/Ice that is available on our local supercalculator, i.e. the version ELMER_REV=058434f, I do not get a proper initial distribution of the gaussian, but it gives instead an unphysical distribution of the physical quantity to advect. On the other hand, when the test is run with a much older version of Elmer/Ice, i.e. ELMER_REV=c227601, the initial distribution is correct. The problem arises even if I use the same source code for the AdvectionReaction solver that I compile locally. So, it means that the problem is not directly within the code of the solver.
Note also that I don't have that kind of problems with the version 058434f of Elmer/Ice when running steady tests, i.e. the recomputed density field by the Advection Reaction solver after a new velocity field has been calculated by the porous solver is correct in steady state but it is wrong, from the first timestep, in transient. So this suggests that the problem is related somehow to the initial conditions. By the way, the density test available in [ELMER_TRUNK]/elmerice/Tests/Density is in steady, this is probably why this bug went unnoticed until now.
Hope these information will help to solve this bug and, at least, help people having similar problems with this solver.
Best regards,
Julien
I notice what seems to be a bug with the AdvectionReaction solver with Disconstinuous Galerkin variables.
This bug is illustrated in a very simple TestCase that is available upon request (too big to be attached to this post). The TestCase consists in advecting vertically a gaussian distribution of a given physical quantity at a constant convection speed. It turns out that when I run this test with the most recent version of Elmer/Ice that is available on our local supercalculator, i.e. the version ELMER_REV=058434f, I do not get a proper initial distribution of the gaussian, but it gives instead an unphysical distribution of the physical quantity to advect. On the other hand, when the test is run with a much older version of Elmer/Ice, i.e. ELMER_REV=c227601, the initial distribution is correct. The problem arises even if I use the same source code for the AdvectionReaction solver that I compile locally. So, it means that the problem is not directly within the code of the solver.
Note also that I don't have that kind of problems with the version 058434f of Elmer/Ice when running steady tests, i.e. the recomputed density field by the Advection Reaction solver after a new velocity field has been calculated by the porous solver is correct in steady state but it is wrong, from the first timestep, in transient. So this suggests that the problem is related somehow to the initial conditions. By the way, the density test available in [ELMER_TRUNK]/elmerice/Tests/Density is in steady, this is probably why this bug went unnoticed until now.
Hope these information will help to solve this bug and, at least, help people having similar problems with this solver.
Best regards,
Julien