Hi Ali,
I would like to ask you a question about
intersectionToVertexBC_.boundaryTypes(values, vertex) function. I
cannot understand where, why and how it should be used.
For box model, I remove it from code without any problem, as you can
see in heterogeneousproblem.hh. I used :
void boundaryTypes(BoundaryTypes &values, const Vertex &vertex) const
to define my boundary types.
we use this function to assign the type of boundary condition depending
on a boundary Id that is defined
in the gird file. In order to access the boundary Id from the grid file,
we need the intersection objects.
- First boundaryTypes(BoundaryTypes &values, const Vertex &vertex) is
called with the vertex as an argument
- intersectionToVertexBC_.boundaryTypes(values, vertex) is then called
which calls
void boundaryTypes(BoundaryTypes &values, const Intersection
&intersection) giving an intersection objects for
the vertex. Now the boundary Id from the grid file can be obtained, and
the boundary condition types can be
defined accordingly.
This procedure is only necessary if you have a complex grid and you
would like to specify the boundary conditions
from the grid file. In all other cases using only
boundaryTypes(BoundaryTypes &values, const Vertex &vertex)
is the way to go.
As I explained, I had a problem with cell-centered models. I added two
voids in co2models that let me calculate the advective flux and
diffusive flux. I followed the explanations given by Alexander
Kissinger to write these voids.
However, it seems that they don't work for cell-centered problems. You
cans find the error message in test_ccco2.stderr. Can you please take a
look at co2model.hh and explain me how I can calculate flux when I use
cell-centered model?
We have not used this function for the CC method. But we are currently
setting up a model with the CC method, where we also want to measure the
advective fluxes. I can inform you if we have a function that works with
CC method as well.
kind regards
Alex
On 11/21/2013 08:54 AM, [email protected] wrote:
Dear Bernd,
Thanks a lot for your help and comments.
I am sorry for my late answer. We had several problems to install
dune-pdelab and to apply patches. Good news, we finally solved all our
problems.
The test_box1pwithamg and test_cc1pwithamg are running without any
problem, however, I have a warning message about SuperLU.
WARNING: You are using AMG without SuperLU! Please consider installing
SuperLU, or set the use superlu flag to false to suppress this warning.
I mentioned that SuperLU is by default included in amg.hh :
#include<dune/istl/superlu.hh>. Now, is it really necessary to install
SuperLU or I can continue my simulations without it?
I also applied the necessary modifications in heterogeneousproblem to
use AMG as the linear solver.
I ran a 100 years simulation using the box model (not cell-centered as
you mentioned in your previous mail, I have a problem with
cell-centered that I will explain it later).
I have the same warning about SuperLU, as before. Comparing the
results of this problem with results of test_box1pwithamg problem, I
found a difference that make me doubt if the AMG is really used in my
problem.
In test_box1pwithamg, each time that the linear solver couldn't
converge, I had this message :
[Newton: Caught exception: "MathError
[mgc:/rap/jda-332-aa/apps/dumux-2.4.0-2/dune-istl-2.2.1/dune/istl/paamg/amg.hh:739]:
Coarse solver did not converge"
Newton solver did not converge with dt=6026.29 seconds. Retrying with
time step of 3013.15 seconds
But in heterogeneousproblem, I had this message :
[Newton: Caught exception: "ISTLError
[apply:/rap/jda-332-aa/apps/dumux-2.4.0-2/dune-istl-2.2.1/dune/istl/solvers.hh:768]:
breakdown in BiCGSTAB - rho 0 <= EPSILON 1e-80 after 59.5 iterations"
Newton solver did not converge with dt=7.59684e+06 seconds. Retrying
with time step of 3.79842e+06 seconds
These messages should not be related and should come from two
different steps, but I would like to have your confirmation. I send
you the heterogeneousproblem.hh file (Lines 89 and 90) for your
attention.
As I explained, I had a problem with cell-centered models. I added
two voids in co2models that let me calculate the advective flux and
diffusive flux. I followed the explanations given by Alexander
Kissinger to write these voids.
However, it seems that they don't work for cell-centered problems. You
cans find the error message in test_ccco2.stderr. Can you please take
a look at co2model.hh and explain me how I can calculate flux when I
use cell-centered model?
I would like to ask you a question about
intersectionToVertexBC_.boundaryTypes(values, vertex) function. I
cannot understand where, why and how it should be used.
For box model, I remove it from code without any problem, as you can
see in heterogeneousproblem.hh. I used :
void boundaryTypes(BoundaryTypes &values, const Vertex &vertex) const
to define my boundary types.
For cell-centered model, as you can see in
heterogeneousproblem_cc_without_intersectionToVertexBC, I used :
void boundaryTypesAtPos(BoundaryTypes &values, const GlobalPosition
&globalPos) const
to define my boundary types.
Thanks a lot for your help and collaboration.
Best regards
Ali NOWAMOOZ
*From:*[email protected]
[mailto:[email protected]] *On Behalf Of *Bernd
Flemisch
*Sent:* mardi 22 octobre 2013 03:23
*To:* DuMuX User Forum
*Subject:* Re: [DuMuX] parallel computing in version 2.4
Hi Ali,
it is nice to hear that you already use Dumux for advanced problems.
What kind of linear solver do you use? From our experience, the
standard linear solvers will not work nicely for more complex problems
in parallel.
Since Dumux 2.3, we have a backend for the ISTL AMG solver. We are
using that for our parallel CO2 computations. You can have a look at
how to use it from test/implicit/1p/test_cc1pwithamg and the problem
file in there. You first need to install the DUNE module dune-pdelab,
http://www.dune-project.org/pdelab/, version 1.0 or 1.1 and to apply
patches to dune-istl and dune-pdelab. See dumux/patches/README for
details. You then need to rerun dunecontrol.
For heterogeneous parameters, it can also be benefitial to use the
cell-centered version instead of box (vertex-centered). You can have a
look at the co2 test problem file to see how this is done.
Currently, the salinity is a public static variable of the Brine
component and not intended to be varying over space. If you use the
BrineCO2FluidSystem in the standard way, the density and viscosity
values of the brine will be tabulated depending on this salinity. You
can use an untabulated version, see the documentation of
BrineCO2FluidSystem at line 745ff. Beware that this can slow down your
computation.
Once you use an untabulated Brine component, you could manipulate the
static salinity value. In a quick and dirty way, you could do this at
the very beginning of the update function in co2volumevariables.hh,
line 122, for example
if (element.geometry().center()[2] < 1000.0)
FluidSystem::Brine::salinity = 0.1;
else
FluidSystem::Brine::salinity = 0.05;
Then all the secondary variables like density would be calculated
depending on this salinity value.
Please let us know how this works out.
Kind regards
Bernd
On 10/22/2013 01:59 AM, Ali NOWAMOOZ wrote:
Dear Dumux,
First, I would like to congratulate all DuMux team for this new
version.
I have a question about the new version of DuMux :
I would like to know If any progress has been made on the parallel
computing in this version comparing to version 2.3. We could install dumux on
Clumeq (server of laval universityhttps://www.calculquebec.ca/index.php/en/)
and run a CO2 2p2c model on 96 processors (12 nodes * 8 processors). However,
we recognized a problem with the linear solvers of DUNE that couldn't converge
(actually, it converged but only with time steps lower than 500 s). I found
some explanations on the dumux website at this
addresshttp://www.mail-archive.com/[email protected]/msg00006.html.
According to this mail, the parallel version of dumux was experimental at that
moment. Plea
se let me
know if it is also the case in the new version. If you need any more
information, please let me know. I can also run a model and send you the
results.
I have also a question about CO2 model. I would like to know if it is
possible to define different salinities in the domain. I don't want to simulate
the salt transport. I only need to distinct fresh water from brine in my domain.
Best regards
Ali NOWAMOOZ
_______________________________________________
Dumux mailing list
[email protected] <mailto:[email protected]>
https://listserv.uni-stuttgart.de/mailman/listinfo/dumux
--
_______________________________________________________________
!!!! CMWR 2014: 10th - 13th June 2014 in Stuttgart !!!!
Please visitwww.cmwr14.de <http://www.cmwr14.de>
_______________________________________________________________
Bernd Flemisch phone: +49 711 685 69162
IWS, Universität Stuttgart fax: +49 711 685 60430
Pfaffenwaldring 61 email:[email protected]
<mailto:[email protected]>
D-70569 Stuttgart url:www.hydrosys.uni-stuttgart.de
<http://www.hydrosys.uni-stuttgart.de>
_______________________________________________________________
_______________________________________________
Dumux mailing list
[email protected]
https://listserv.uni-stuttgart.de/mailman/listinfo/dumux
--
*******************************************************
!!!! CMWR 2014: 10th - 13th June 2014 in Stuttgart !!!!
Please visit www.cmwr14.de
*******************************************************
Alexander Kissinger
Institut für Wasser- und Umweltsystemmodellierung
Lehrstuhl für Hydromechanik und Hydrosystemmodellierung
Pfaffenwaldring 61
D-70569 Stuttgart
Telefon: +49 (0) 711 685-64729
E-Mail: [email protected]
_______________________________________________
Dumux mailing list
[email protected]
https://listserv.uni-stuttgart.de/mailman/listinfo/dumux
