Re: [Nek5000-users] Nek5000 and Non-Newtonian Fluids
Paul, I switched this to the nek5000-users channel in case we discuss some thing relevant to others. For the Bingham model, it seems that all I need is to calculate the second invariant of the strain rate tensor, which is then used to calculate the time-varying apparent viscosity. I suspect the routine building-blocks for generating this tensor are already in the nek distribution. If so, could you point me in that direction? Once those routines are in order, all the calculation will be directed out of uservp, correct? Thanks. --Mike On Jan 14, 2011, at 7:27 AM, Paul Fischer wrote:
Hi Mike,
[...] The "peris" case uses the stress formulation, as do the "fs" cases. These have the correct SiZE files (resp. for 3D and 2D) and flags in the .rea file, save that you would set "IFMVBD" to false, since you don't want moving boundaries.
[...] The code was used for a lot of non-Newtonian stuff by Lee Ho and Einar, however, so it has been exercised significantly in the past. [...]
Hi Mike, Probably the best point to start is in the lambda2 routine, to be found in postpro.f, which reads in part as: do ie=1,nelv ! Compute velocity gradient tensor call comp_gije(gije,vx(1,1,1,ie),vy(1,1,1,ie),vz(1,1,1,ie),ie) do l=1,nxyz ! decompose into symm. and antisymm. part do j=1,ndim do i=1,ndim ss(i,j) = 0.5*(gije(l,i,j)+gije(l,j,i)) oo(i,j) = 0.5*(gije(l,i,j)-gije(l,j,i)) enddo enddo etc. I think from their you can see how to compute what you need. (Let me know if you have questions...) If you get such a routine (e.g., "compute_q") and are happy with it after a few trials, I'd be happy to insert it into the repo. Cheers, Paul On Mon, 17 Jan 2011, [email protected] wrote:
Paul,
I switched this to the nek5000-users channel in case we discuss some thing relevant to others.
For the Bingham model, it seems that all I need is to calculate the second invariant of the strain rate tensor, which is then used to calculate the time-varying apparent viscosity. I suspect the routine building-blocks for generating this tensor are already in the nek distribution. If so, could you point me in that direction?
Once those routines are in order, all the calculation will be directed out of uservp, correct?
Thanks. --Mike
On Jan 14, 2011, at 7:27 AM, Paul Fischer wrote:
Hi Mike,
[...] The "peris" case uses the stress formulation, as do the "fs" cases. These have the correct SiZE files (resp. for 3D and 2D) and flags in the .rea file, save that you would set "IFMVBD" to false, since you don't want moving boundaries.
[...] The code was used for a lot of non-Newtonian stuff by Lee Ho and Einar, however, so it has been exercised significantly in the past. [...]
_______________________________________________ Nek5000-users mailing list [email protected] https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users
Paul, That's got what I need, thanks. Also, are there any examples in the distribution that use the stress formulation and variable viscosity? Thanks. --Mike On Jan 18, 2011, at 6:22 AM, <[email protected]> wrote:
Hi Mike,
Probably the best point to start is in the lambda2 routine, to be found in postpro.f, which reads in part as:
do ie=1,nelv ! Compute velocity gradient tensor call comp_gije(gije,vx(1,1,1,ie),vy(1,1,1,ie),vz(1,1,1,ie),ie)
do l=1,nxyz ! decompose into symm. and antisymm. part do j=1,ndim do i=1,ndim ss(i,j) = 0.5*(gije(l,i,j)+gije(l,j,i)) oo(i,j) = 0.5*(gije(l,i,j)-gije(l,j,i)) enddo enddo
etc.
I think from their you can see how to compute what you need. (Let me know if you have questions...)
If you get such a routine (e.g., "compute_q") and are happy with it after a few trials, I'd be happy to insert it into the repo.
Cheers,
Paul
On Mon, 17 Jan 2011, [email protected] wrote:
Paul,
I switched this to the nek5000-users channel in case we discuss some thing relevant to others.
For the Bingham model, it seems that all I need is to calculate the second invariant of the strain rate tensor, which is then used to calculate the time-varying apparent viscosity. I suspect the routine building-blocks for generating this tensor are already in the nek distribution. If so, could you point me in that direction?
Once those routines are in order, all the calculation will be directed out of uservp, correct?
Thanks. --Mike
On Jan 14, 2011, at 7:27 AM, Paul Fischer wrote:
Hi Mike,
[...] The "peris" case uses the stress formulation, as do the "fs" cases. These have the correct SiZE files (resp. for 3D and 2D) and flags in the .rea file, save that you would set "IFMVBD" to false, since you don't want moving boundaries.
[...] The code was used for a lot of non-Newtonian stuff by Lee Ho and Einar, however, so it has been exercised significantly in the past. [...]
_______________________________________________ Nek5000-users mailing list [email protected] https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users
_______________________________________________ Nek5000-users mailing list [email protected] https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users
Hi Mike, Just added "var_vis" to /examples --- it's a test case that I've tested in the past (no exact solution). So far, all exact solns I've found don't properly exercise the full stress tensor, whereas this particular example does - but I don't know the exact solution. In this case, I rely on the fact that the code is correct and convergent and thus trust that we are getting the right answer. If you happen to think of a case that is a better test (I'm certain they exist), please let me know. I believe that if you cd to /examples and say "svn update" that you will find this case deposited into your repo. Let me know if otherwise. Cheers, Paul On Tue, 18 Jan 2011, [email protected] wrote:
Paul,
That's got what I need, thanks.
Also, are there any examples in the distribution that use the stress formulation and variable viscosity?
Thanks. --Mike
On Jan 18, 2011, at 6:22 AM, <[email protected]> wrote:
Hi Mike,
Probably the best point to start is in the lambda2 routine, to be found in postpro.f, which reads in part as:
do ie=1,nelv ! Compute velocity gradient tensor call comp_gije(gije,vx(1,1,1,ie),vy(1,1,1,ie),vz(1,1,1,ie),ie)
do l=1,nxyz ! decompose into symm. and antisymm. part do j=1,ndim do i=1,ndim ss(i,j) = 0.5*(gije(l,i,j)+gije(l,j,i)) oo(i,j) = 0.5*(gije(l,i,j)-gije(l,j,i)) enddo enddo
etc.
I think from their you can see how to compute what you need. (Let me know if you have questions...)
If you get such a routine (e.g., "compute_q") and are happy with it after a few trials, I'd be happy to insert it into the repo.
Cheers,
Paul
On Mon, 17 Jan 2011, [email protected] wrote:
Paul,
I switched this to the nek5000-users channel in case we discuss some thing relevant to others.
For the Bingham model, it seems that all I need is to calculate the second invariant of the strain rate tensor, which is then used to calculate the time-varying apparent viscosity. I suspect the routine building-blocks for generating this tensor are already in the nek distribution. If so, could you point me in that direction?
Once those routines are in order, all the calculation will be directed out of uservp, correct?
Thanks. --Mike
On Jan 14, 2011, at 7:27 AM, Paul Fischer wrote:
Hi Mike,
[...] The "peris" case uses the stress formulation, as do the "fs" cases. These have the correct SiZE files (resp. for 3D and 2D) and flags in the .rea file, save that you would set "IFMVBD" to false, since you don't want moving boundaries.
[...] The code was used for a lot of non-Newtonian stuff by Lee Ho and Einar, however, so it has been exercised significantly in the past. [...]
_______________________________________________ Nek5000-users mailing list [email protected] https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users
_______________________________________________ Nek5000-users mailing list [email protected] https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users
_______________________________________________ Nek5000-users mailing list [email protected] https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users
participants (1)
-
nek5000-users＠lists.mcs.anl.gov