Hi Edscott,
thanks for your comments!
Just a few words on the purpose of the initialisation: For setting a
hydrostatic pressure over depth as a Dirichlet condition, one usually
assumes a constant density and calculates the pressure from rho * g *
depth. Nonetheless, the density is also pressure dependent, so the idea
of the initialization run is to set the pressure at a boundary with
constant depth (either the bottom or the top) and then get the true
pressure distribution from running the model for a stationary state.
This pressure distribution over depth can then be used as the the
boundary condition for the actual simulation. In the problem file, the
pInit is updated via the setPressure function called at the end of the
initialisation. During the initialisation, the coupling between flow and
geomechanics is also switched off (coupled_ = false).
Obviously, there are scenarios, where this pressure-dependence of the
density can be neglected or doesn't play a role at all (e.g. Mandel's
problem).
Cheers,
Martin
On 3/4/19 10:48 PM, Ed Scott Wilson Garcia wrote:
Thank you very much, Martin.
I’ve started looking through the code for the Mandel problem and find
that what you mention with regard to the porosity change is dealt with
in elementvolumevariables.hh whenever the bool property
UseMandelStorageTerm is true.
On the other hand, I notice that you have an initialization episode
(we were experimenting without any initialization episode, which maybe
a mistake). Perhaps that is the best way to specify the Mandel problem
with mostly Dirichlet conditions?
BTW, interesting that the approach used by Phillips and Wheeler is
the most frequently applied to existing reservoir engineering
simulators, or so they say.
Best regards,
Edscott
*De:*Martin Beck [mailto:[email protected]]
*Enviado el:* lunes, 4 de marzo de 2019 02:12 a. m.
*Para:* DuMuX User Forum; Ed Scott Wilson Garcia
*CC:* Manuel Coronado Gallardo
*Asunto:* Re: [DuMuX] el2p issue almost clear
Hi Edscott,
I just committed Mandel's problem to the fully-coupled branch of the
Beck2019a module.
I used the description from Phillips and Wheeler:
https://link.springer.com/article/10.1007/s10596-007-9045-y
For unknown reasons, their storage term is slightly different from the
one we use. In details, it comes down to how the porosity change is
described. So to match the analytical solution, the change in porosity
is (co * p + alpha * div u) for Mandel's problem while all other
problem use (phi_0 + alpha * div u).
Let me know if you have further questions!
Cheers,
Martin
On 3/1/19 10:28 PM, Ed Scott Wilson Garcia wrote:
Thank you very much Martin.
When I change the boundary conditions on the 2.12 el2p test, the
second episode will not work. I’ll try it with the SuperLU solver.
The initialization episode has no apparent problem with amg.
I look forward to studing your implementation of the Mandel
problem. I tend to believe all my problems arise from my incorrect
understanding of the initial and boundary conditions.
best regards,
Edscott
*De:*Dumux [mailto:[email protected]] *En
nombre de *Beck, Martin
*Enviado el:* viernes, 1 de marzo de 2019 06:38 a. m.
*Para:* DuMuX User Forum
*CC:* Manuel Coronado Gallardo
*Asunto:* Re: [DuMuX] el2p issue almost clear
Hi Edscott,
the momentum balance for the geomechanics is formulated in an
incremental way. That meens it only considers changes with respect
to the initial state. Nonetheless, the initial stress state
matters for post-processing routines such a the evaluation of the
total principal stresses and the failure criteria.
I have another suspicion for your troubles: The standard test in
the 2.12-release uses the el2p-amgbackend, which seems to cause
problems for some cases. For that reason, I switched to a direct
solver (SuperLU) for my examples. Maybe this is reason why your
problem now works, too?
One more thing: I managed to implement Mandel's problem this week
for the fully coupled version of the code. It works fine and
reproduces the analytical solution quite well. I was trying to
implement it also for the decoupled version, but I ran into some
problems. Anyway, I will commit the fully coupled version to the
Beck2019a module on Monday (I am out of office today), so you can
look into it.
Cheers,
Martin
------------------------------------------------------------------------
*From:*Dumux <[email protected]
<mailto:[email protected]>> on behalf of Ed
Scott Wilson Garcia <[email protected] <mailto:[email protected]>>
*Sent:* Thursday, February 28, 2019 11:50:39 PM
*To:* DuMuX User Forum
*Cc:* Manuel Coronado Gallardo
*Subject:* [DuMuX] el2p issue almost clear
One further question, the force that presses down on the grid from
the top, is that not a boundary condition?
I see from the code that it is used to set the initial total
stress field (isotropic, lithostatic),
(i.e. stress[] = brineDensity_ * porosity * gravity * (depthBOR_
- globalPos[dim-1])
+ (1 - porosity) * rockDensity * gravity *
(depthBOR_ - globalPos[dim-1]);).
If I do not use the force to set a Neumann boundary condition for
uy, and instead use the values Beck2019 uses (all Neumann values
set to zero),
the initialization of the El2P_TestProblem problem runs just fine
with data from the Sangnimnuan 2018 paper (injection episode
turned off).
If I am not wrong about this, then my whole trouble was with that
misleading fact.
Code from the fullycoupled examples has been most useful, and the
decoupled is very interesting. And I thank Bernd and Beck very much.
Best regards,
Edscott
_______________________________________________
Dumux mailing list
[email protected]
<mailto:[email protected]>
https://listserv.uni-stuttgart.de/mailman/listinfo/dumux
_______________________________________________
Dumux mailing list
[email protected] <mailto:[email protected]>
https://listserv.uni-stuttgart.de/mailman/listinfo/dumux
--
*******************************************************
Martin Beck
Institute for Modelling Hydraulic and Environmental Systems
Department of Hydromechanics and Modelling of Hydrosystems
Pfaffenwaldring 61
D-70569 Stuttgart
Tel.: (+49) 711/ 685-64899
[email protected] <mailto:[email protected]>
*******************************************************
--
*******************************************************
Martin Beck
Institute for Modelling Hydraulic and Environmental Systems
Department of Hydromechanics and Modelling of Hydrosystems
Pfaffenwaldring 61
D-70569 Stuttgart
Tel.: (+49) 711/ 685-64899
[email protected]
*******************************************************
_______________________________________________
Dumux mailing list
[email protected]
https://listserv.uni-stuttgart.de/mailman/listinfo/dumux
