Kernels for poroelastic solvers #1353

castelletto1 · 2021-03-11T22:48:11Z

This PR adds assembly kernels for fully-implicit poroelastic simulation with both single-phase (PoroelasticSolver) and multiphase (MultiphasePoroelasticSolver) flow. Capillary effects are neglected. Flow solvers can be based on either finite volumes (FVM) or hybrid schemes (HybridFVM). For the muliphase flow case the maximum number of components is equal to three).

The kernels assemble:

all contributions relative to the discrete linear momentum balance equation (note that an inexact linearization is used at the moment for the body force term)
contributions to residual and Jacobian matrix relative to the the accumulation term in mass balance equations

Currently the porosity equation considers a linear poroelastic material and assumes incompressible solid grains (1/N = 0, with N Biot's skeleton modulus):

$\phi_{n} = \phi_{0}+b(\epsilon_{v,n}-\epsilon_{v,0})+\frac{1}{N}(p_{n}-p_{0})\approx\phi_{0}+b(\epsilon_{v,n}-\epsilon_{v,0})$

with the subscript n denoting the discrete time level. The porosity treatment will be generalized introducing a dedicated constitutive model (see #1311).

A preliminary 2d example has been added (PoroElastic_deadoil_3ph_baker_2d.xml).

Currently all integrated tests pass, except for the usual ones on Lassen.

Some issues to be addressed in future PR / PRs:

Exact linerization for body forces
Add capillary effects
Add MGR recipes for (i) multiphase poromechanics with FVM, and (ii) single- and multiphase poromechanics with HybridFVM
Rename PoroelasticSolver to SinglePhasePoroelasticSolver (not done here to avoid new baseline in the integrated tests)
Add benchmark based on SPE10 model 1 [see Camargo et al., IJNAMG (2021)]
Remove the split operator strategy (SIM_FixedStress coupling option) from the single-phase poroelastic solver.

Requires a rebaseline:
https://github.com/GEOSX/integratedTests/pull/124

src/coreComponents/physicsSolvers/multiphysics/SinglePhasePoroelasticKernel.hpp

sytuannguyen · 2021-03-15T10:13:21Z

@castelletto1 @rrsettgast I am interested in poroelastic coupling, so read through your PR and add some comments, Hope this help.

Do you want to add a class for computing the poroelastic constants, i.e. the transformation between the poroelastic constants?

src/coreComponents/finiteElement/elementFormulations/FiniteElementBase.hpp

src/coreComponents/physicsSolvers/fluidFlow/CompositionalMultiphaseBase.cpp

src/coreComponents/physicsSolvers/multiphysics/MultiphasePoroelasticSolver.cpp

…mentBase.hpp Co-authored-by: Randolph Settgast <settgast1@llnl.gov>

…OSX into feature/singleKernelPoroelastic

CusiniM

This is great! Should also integrate very well with what I have been working on.

CusiniM · 2021-03-15T21:42:09Z

src/coreComponents/physicsSolvers/multiphysics/MultiphasePoroelasticKernel.hpp

+        stack.xLocal[ a ][ i ] = m_X[ localNodeIndex ][ i ];
+#endif
+        stack.u_local[ a ][i] = m_disp[ localNodeIndex ][i];
+        stack.uhat_local[ a ][i] = m_uhat[ localNodeIndex ][i];


I wonder if we should use disp and deltaDisp like we do for the pressure instead of u and u_hat.

I agree that the naming convention should be improved. However, it should be reflected across all solid mechanics kernels.

src/coreComponents/physicsSolvers/multiphysics/MultiphasePoroelasticSolver.cpp

CusiniM · 2021-03-15T21:47:51Z

src/coreComponents/physicsSolvers/multiphysics/MultiphasePoroelasticSolver.cpp

+  // update displacement field
+  m_solidSolver->applySystemSolution( dofManager, localSolution, scalingFactor, domain );
+  // update pressure field
+  m_flowSolver->applySystemSolution( dofManager, localSolution, -scalingFactor, domain );


at some point we should fix the signs.

src/coreComponents/physicsSolvers/multiphysics/SinglePhasePoroelasticKernel.hpp

src/coreComponents/physicsSolvers/multiphysics/MultiphasePoroelasticKernel.hpp

sytuannguyen · 2021-03-16T08:53:14Z

@castelletto1 I added some additional comments to the Multiphase class.

Also I have a naive question: if multi phase solver can do the job of single phase solver just by setting 100% saturation to a phase, do we really need single phase solver?

klevzoff

Looks good!

src/coreComponents/physicsSolvers/multiphysics/MultiphasePoroelasticSolver.hpp

src/coreComponents/physicsSolvers/multiphysics/MultiphasePoroelasticSolver.cpp

klevzoff · 2021-03-16T18:12:30Z

src/coreComponents/physicsSolvers/multiphysics/MultiphasePoroelasticKernel.hpp

+  /// Number of nodes per element...which is equal to the
+  /// numTestSupportPointPerElem and numTrialSupportPointPerElem by definition.
+  static constexpr int numNodesPerElem = Base::numTestSupportPointsPerElem;
+  static constexpr int numMaxComponentsMultiphasePoroelastic = 3;


We need to address this in a follow-up PR

src/coreComponents/physicsSolvers/multiphysics/MultiphasePoroelasticKernel.hpp

src/coreComponents/physicsSolvers/multiphysics/SinglePhasePoroelasticKernel.hpp

rrsettgast · 2021-03-17T03:42:27Z

src/coreComponents/finiteElement/elementFormulations/FiniteElementBase.hpp

+  template< int NUM_SUPPORT_POINTS,
+            typename GRADIENT_TYPE >
+  GEOSX_HOST_DEVICE
+  static real64 symmetricGradientTrace( GRADIENT_TYPE const & gradN,


Sorry I didn't already ask...but would it make more sense to name this traceSymmetricGradient?

rrsettgast · 2021-03-17T03:46:03Z

src/coreComponents/linearAlgebra/interfaces/hypre/HyprePreconditioner.cpp

@@ -402,7 +402,8 @@ void HyprePreconditioner::createMGR( DofManager const * const dofManager )
  std::vector< HYPRE_Int > mgr_level_interp_type;
  std::vector< HYPRE_Int > mgr_level_frelax_method;

-  if( ( m_parameters.mgr.strategy == "Poroelastic" ) | ( m_parameters.mgr.strategy == "Hydrofracture" ) )
+  if( ( m_parameters.mgr.strategy == LinearSolverParameters::MGR::StrategyType::singlePhasePoroelastic) |
+      ( m_parameters.mgr.strategy == LinearSolverParameters::MGR::StrategyType::hydrofracture ) )


Suggested change

( m_parameters.mgr.strategy == LinearSolverParameters::MGR::StrategyType::hydrofracture ) )

( m_parameters.mgr.strategy == LinearSolverParameters::MGR::StrategyType::singlePhaseFractureMechanicsFlow ) )

Should we stop calling it hydrofracture? I think that @CusiniM also had the inclination to rebrand it in the context of a poroelastic problem...but it has a transmissibility that varies with deformation.

What would be an appropriate name for "singlePhasePoroelastic" with deformation dependent transmissibility?

rrsettgast · 2021-03-17T04:50:36Z

src/coreComponents/physicsSolvers/fluidFlow/SinglePhaseBase.cpp

@@ -426,6 +426,7 @@ void SinglePhaseBase::assembleSystem( real64 const time_n,

  if( m_poroElasticFlag )


still can't get rid of these?

rrsettgast and others added 24 commits February 1, 2021 12:12

initial attempt at single kernel poroelastic assembly

433859c

Adding initialization for new stack variables

c3d55c8

Handling fluid density, WIP

d9bda6d

Adding mixture density

a8f30ed

Adding targerRegionIndex

eda8c9d

fixed inclusion of targetRegionIndex into the fe kernel infrastructure

254246c

Adding function for symgrad trace (volumetric strain)

5fd9e4a

Polishing SinglePhasePoroelasticKernel

c3dc8f0

Cleaning

cd3c0a5

Post merge fixes

820cf9a

WIP TwoPhasePoroelasticKernel

2a0d153

Merge branch 'develop' into feature/singleKernelPoroelastic

274709b

Post merge fixes

0390fd0

Adding contributions in TwoPhasePoroelasticKernel

f5da589

Merge branch 'develop' into feature/singleKernelPoroelastic

a39a443

Without hardcoded biot coefficient

8025f51

Without hardcoded biot coefficient

e28a247

With multiphase poroelasticity

465f92e

Completing kernel assuming constant body force. Porosity issue remains

a297604

Flow accumulation in SinglePhasePoroelasticKernel

23bb9ac

SinglePhasePoroelasticKernel complete (works with FMV and MFD)

dfbfd5b

WIP multiphase kernel

b49ad78

Merge branch 'develop' into feature/singleKernelPoroelastic

aa27d85

Adding kernel for multiphase poromechanics

3606a7a

castelletto1 marked this pull request as draft March 11, 2021 22:48

castelletto1 marked this pull request as ready for review March 11, 2021 22:51

castelletto1 added 4 commits March 11, 2021 17:39

Fixing doxygen warnings

4d2d846

Cleaning

3bda8da

Uncrustifying

338bbea

Minor changes + first xml example

717eb53

sytuannguyen reviewed Mar 15, 2021

View reviewed changes

src/coreComponents/physicsSolvers/multiphysics/SinglePhasePoroelasticKernel.hpp Show resolved Hide resolved

rrsettgast reviewed Mar 15, 2021

View reviewed changes

castelletto1 and others added 3 commits March 15, 2021 12:50

Update src/coreComponents/finiteElement/elementFormulations/FiniteEle…

6fd7f07

…mentBase.hpp Co-authored-by: Randolph Settgast <settgast1@llnl.gov>

Addressing some review comments

6c9ccd3

Merge branch 'feature/singleKernelPoroelastic' of github.com:GEOSX/GE…

0dc80f5

…OSX into feature/singleKernelPoroelastic

CusiniM approved these changes Mar 15, 2021

View reviewed changes

castelletto1 added 3 commits March 15, 2021 15:32

Addressing more review comments

9c75299

Typo

284c146

Cleaner handling of MGR strategies

3867dfc