Elmer Discussion Forum

raback wrote: ↑24 May 2021, 13:58 Hi

Do you have mesh dense enough to resolve the boundary layers? Your picture looks very coarse. I think Kevin got good results with a fine enough mesh. Isn't this the same problem?

-Peter

raback wrote: ↑02 Jun 2021, 18:05 Hi,

I had some time to study this interesting problem. Convergence depends quite a bit on the strength of the coupling. It is difficult to get good convergence wit large magnetic fields without some relaxation.

To make the study somewhat faster I use only 2 elements in x-direction + periodic BCs. This way I can iterate a converged solution of the profile in seconds. The price is that the Ha number can only be defined after the simulation. We should add a piece of code to constrain the average velocity, now an offset of pressure is used. Still, this can be used to ensure the correctness of the coupling. I hope it helps.

Plot the output profile with

Code: Select all

load f.dat
plot(f(:,5),f(:,13))

-Peter

raback wrote: ↑02 Jun 2021, 18:21 It, may be added as a test. Just would like to add a constraint for the total flow before that. -Peter

raback wrote: ↑02 Jun 2021, 18:05 Hi,

I had some time to study this interesting problem. Convergence depends quite a bit on the strength of the coupling. It is difficult to get good convergence wit large magnetic fields without some relaxation.

To make the study somewhat faster I use only 2 elements in x-direction + periodic BCs. This way I can iterate a converged solution of the profile in seconds. The price is that the Ha number can only be defined after the simulation. We should add a piece of code to constrain the average velocity, now an offset of pressure is used. Still, this can be used to ensure the correctness of the coupling. I hope it helps.

Plot the output profile with

Code: Select all

load f.dat
plot(f(:,5),f(:,13))

-Peter

load f.dat
plot(f(:,5),f(:,13))

raback wrote: ↑15 Jun 2021, 17:31 Hi

This case inspired me to implement integral constraint for any field. So the idea is that on a given boundary you can define \int f dA = F(t,...). This comes handy in this case as we can set the average inlet velocity to one without knowing the profile. We don't even need long tubes to develop the flow. It can be quite a short piece of tube.

Here is the compute profile with different values of the nondimensional parameter.
HartmannProfile.PNG

The case was just added as a test case here:
https://github.com/ElmerCSC/elmerfem/tr ... ntHartmann

This case can be used to study the severity of the boundary layer challenge and difficulty of relaxation. The larger the external magnetic force the smaller the relaxation factor must be. These observations may be good to study before going into more complicated cases.

-Peter

PS. The code for integral constraint was just committed so you need to recompile or wait for the new builds to appear.

Applied Magnetic Field 1 = Variable Coordinate 2
     Real MATC " B1*exp(-(tx-0.01)^2/(2*(1002)^2))"

raback wrote: ↑15 Jun 2021, 17:31 Hi

This case inspired me to implement integral constraint for any field. So the idea is that on a given boundary you can define \int f dA = F(t,...). This comes handy in this case as we can set the average inlet velocity to one without knowing the profile. We don't even need long tubes to develop the flow. It can be quite a short piece of tube.

Here is the compute profile with different values of the nondimensional parameter.
HartmannProfile.PNG

The case was just added as a test case here:
https://github.com/ElmerCSC/elmerfem/tr ... ntHartmann

This case can be used to study the severity of the boundary layer challenge and difficulty of relaxation. The larger the external magnetic force the smaller the relaxation factor must be. These observations may be good to study before going into more complicated cases.

-Peter

PS. The code for integral constraint was just committed so you need to recompile or wait for the new builds to appear.