Dear Ranjeet,
you are right, the test uses deformation-dependent porosities, however,
you can observe rather poor Newton solver convergence behavior. That is
because of the mentioned bug which we are currently trying to fix. As I
said in my previous mails, the derivatives of the deformation-dependent
porosities are wrong in the current Dumux master.
On the branch "temp/poromech-storagederivs-hacky-fix" is a temporary,
hacky fix in case you want to use this. It requires two little changes
in your main.cc and spatialparams.hh files. On that branch I have
modified the el2p test so that you can see where to make the changes.
You can see this in this commit:
https://git.iws.uni-stuttgart.de/dumux-repositories/dumux/-/commit/6250c8bdd041f4133b5ed5a65506f8ee7b7c28a6
If you run the el2p test, you will see that instead of approximately 19
iterations, the Newton solver only needs about 3 iterations per time
step. The branch is based on the current Dumux master of today, with
only two added commits for the fix.
Regarding your other questions:
The base class you mentioned, "FVSpatialParamsPoroElastic", is a base
class for spatial parameters of poroelastic applications. In such
applications, the momentum balance equation of the porous medium is
solved using Biots law for the effective stresses. The pressure that
enters these effective stresses can either be a given distribution, or,
come from a coupled flow model. In the case of the latter, such
simulations are realized within the MultiDomain module where a
poroelastic problem is coupled to a flow problem. This is what is done
in the el2p test from which you started developing your application (if
I remember correctly).
This means, the spatialparams_poroelastic.hh file you mentioned only
defines the parameters of the poroelastic sub-problem. The momentum
balance equation of the porous medium contains the porosity in the body
force term if gravity is considered, but the permeability only shows up
in the fluid mass balances. Thus, the poroelastic spatial parameters
don't need to (or should not) define permeability distributions.
The flow model needs both porosity (for the storage term) and
permeability (for the flux term). That is why in the spatial parameters
of the two-phase flow sub-problem (spatialparams_2p.hh) you have to
define both. What is important is that the porosity you define in
spatialparams_2p.hh should be the same as in spatialparams_poroelastic.hh.
I hope this helps!
Dennis
On 07.04.20 14:15, Ranjeet kumar wrote:
Hi Dennis,
Thank you for making things clear. But I still have few doubts. I am
using dumux/test/multidomain/poromechanics/el2p example for my model.
In spatialparams_2p.hh, both deformation dependent porosity and
deformation dependent permeability are being evaluated while in
*spatialparams_poroelastic.hh*, deformation dependent porosity is
being evaluated. In the base class "FVSpatialParamsPoroElastic", there
is no method for the permeability calculation. Now I am confused, not
sure which case is being solved here. Would you please explain
implementation differences for the both cases?
Also, I really appreciate if you could point out the 3.2 branch to use
deformation-dependent porosities model.
Thanks & Regards,
Ranjeet
On Wed, Apr 1, 2020 at 7:02 PM Dennis Gläser
<[email protected]
<mailto:[email protected]>> wrote:
Hi Ranjeet,
regarding the formulation: there are no publications using the
poromechanics module of Dumux 3, yet. But, you can have a look at
the PhD thesis of Martin Beck:
https://elib.uni-stuttgart.de/bitstream/11682/10435/3/Promo_MartinBeck.pdf
and in particular the sections 3.1.1 and 3.1.2. Equation 3.6
states the momentum balance equation of the porous medium right
before the linearization. This is what is currently solved in
Dumux. This implies that you should have knowledge about your
initial stress/deformation conditions, or, solve for them
primarily. In other words, if you initialize your system with
non-zero fluid pressures, you will observe stresses/deformations
caused by the initial conditions. If gravity is neglected and
incompressible fluids are considered, you could start your
simulation with zero pressure and zero displacement.
Regarding the deformation-dependent porosities: That depends on
your application. Currently, there is an issue with the coupling
derivatives of the storage term, and thus, if the porosity is
chosen deformation-dependent, the non-linear solver convergence
deteriorates significantly. On the other hand,
deformation-dependent permeabilities work fine.
We are planning to fix this issue very soon and hope that you want
to use Dumux for you application. We will not be able to fix it
until the 3.2 release which is scheduled for a few weeks from now.
If you want, I can point you to a fix branch on which the bugfix
is done in an unsatisfactory way, but on that branch you could
develop your application for now. Once we have solved the problem
you could switch back to the release branch.
Regards,
Dennis
On 29.03.20 15:01, Ranjeet kumar wrote:
Hi All,
I want to develop a poroelastic model using el2p module in Dumux
3.1. As per Dumux2.12 class documentation, soil mechanics sign
convention (i.e. compressive stresses are negative) has been used
in el2p module. However, papers & thesis based on Dumux2 has
used opposite sign convention. Can you clarify which sign
convention is being used in Dumux3.1?
Q2. As a linearised form of momentum balance equation is being
solved in el2p model, the solution effective stress obtained from:
const auto effSigma = StressType::effectiveStressTensor(*this,
element, fvGeometry, elemVolVars, fluxVarsCache);
is the total effective stress or change in effective stress with
respect to initial effective stress?
References:
1. p-3, Volume-Based Modelling of Fault Reactivation in Porous
Media Using a Visco-Elastic Proxy Model. Martin Beck1 · Gabriele
Seitz1 · Holger Class1
2. p-3, Modelling fault reactivation with characteristic
stress-drop terms. Martin Beck and Holger Class
3. p68-69, Coupling Models of Different Complexity for the
Simulation of CO2 Storage in Deep Saline Aquifers
Thank you,
Regards,
Ranjeet
_______________________________________________
Dumux mailing list
[email protected] <mailto:[email protected]>
https://listserv.uni-stuttgart.de/mailman/listinfo/dumux
_______________________________________________
Dumux mailing list
[email protected]
https://listserv.uni-stuttgart.de/mailman/listinfo/dumux
