Richards Component Transport #1929

TomFischer · 2017-08-28T09:31:13Z

~~Please do not review - this PR is to test the generation of the documentation.~~

Please review!

chleh · 2017-10-04T09:32:49Z

There is a huge ParaView state file in this PR. Can it be removed, or moved to the data repo?

hydromarc · 2017-10-09T13:59:19Z

@chleh You are right, this state file should be in data... ( @TomFischer can you move this?)

endJunction · 2017-10-10T08:07:45Z

...tation/ProjectFile/prj/processes/process/RichardsComponentTransport/t_specific_body_force.md

@@ -0,0 +1,3 @@
+Specific body forces applied to fluid.
+
+It is usually used to apply gravitational forces. Given as a 2d or 3d vector.


Nit picking: it's probably copy-paste, but I'd think it also applies to 1d problems...

endJunction · 2017-10-10T08:09:06Z

ProcessLib/HT/HTFEM.h

@@ -306,6 +306,8 @@ class LocalAssemblerData : public HTLocalAssemblerInterface
            auto const& N = ip_data.N;
            auto const& dNdx = ip_data.dNdx;

+            pos.setIntegrationPoint(ip);


This can be directly be pushed to the master. Has nothing to do with Richards...

endJunction · 2017-10-10T10:08:36Z

ProcessLib/RichardsComponentTransport/RichardsComponentTransportFEM.h

+        cache.clear();
+        auto cache_vec = MathLib::createZeroedMatrix<
+            Eigen::Matrix<double, 1, Eigen::Dynamic>>(
+            cache, 1, n_integration_points);


Not sure if the mapping is needed, because the cache is accessed only once by index (of integration point). But you might want to keep it if this style already became idiomatic (then the mapping could be to a VectorXd instead of a matrix).
The access below would be cache[ip] = Sw.

endJunction · 2017-10-10T10:19:12Z

ProcessLib/RichardsComponentTransport/RichardsComponentTransportFEM.h

+        SpatialPosition pos;
+        pos.setElementID(_element_id);
+
+        auto const num_nodes = ShapeFunction::NPOINTS;


Just a comment about variable naming: shape function's number of points is not always equal to the number of nodes of an element. Maybe n_points? Feel free to ignore the suggestion.

In some other multi variable processes, e.g. HM, there are class variables X_index and X_size for each variable X (pressure, displacement, temperature,...). Then the submatrix creation looks like:

auto KCC = local_K.template block<C_size, C_size>(C_index, C_index); auto KCp = local_K.template block<C_size, p_size>(C_index, p_index);

endJunction · 2017-10-10T10:40:58Z

I'd recommend to split the Process.h into declarations and an implementation parts, where the latter Process-impl.h is only included in the Process.cpp. This could save some compilation time.
See, for example HM process again with template instantiation declarations and the explicit instantiations in the HMProcess.cpp.

Never mind. This is not a template, as Tom mentions below... Silly me ;)

TomFischer · 2017-10-12T07:45:00Z

@endJunction
I don't understand your last comment concerning the splitting of the process into declarations and implementations. The RichardsComponentTransport process isn't a template. Maybe you want to suggest to split the FEM?

wenqing · 2017-10-13T08:05:42Z

ProcessLib/RichardsComponentTransport/RichardsComponentTransportFEM-impl.h

+        NumLib::shapeFunctionInterpolate(local_x, N, C_int_pt, p_int_pt);
+
+        double const pc_int_pt = -p_int_pt;
+        double const Sw = (pc_int_pt > 0)


You can remove this ternary condition because (pc_int_pt > 0) is already checked inside

_process_data.porous_media_properties .getCapillaryPressureSaturationModel(t, pos) .getSaturation(pc_int_pt)

as

double BrooksCoreyCapillaryPressureSaturation::getSaturation( const double capillary_pressure) const { const double pc = (capillary_pressure < 0.0) ? _minor_offset : capillary_pressure; ... return MathLib::limitValueInInterval(S, _saturation_r + _minor_offset, _saturation_max - _minor_offset);

Thanks for the hint. Changed all locations.

wenqing · 2017-10-13T08:07:35Z

ProcessLib/RichardsComponentTransport/RichardsComponentTransportFEM-impl.h

+                                    .getSaturation(pc_int_pt)
+                              : 1.0;
+
+        double const dSw_dpc =


The same reason to remove this ternary condition. With the change, the benchmark results will be changed slightly if the given maximum saturation is less than 1.

wenqing · 2017-10-13T08:09:25Z

ProcessLib/RichardsComponentTransport/RichardsComponentTransportFEM-impl.h

+
+        // saturation
+        double const pc_int_pt = -p_int_pt;
+        double const Sw = (pc_int_pt > 0)


The same here

wenqing · 2017-10-13T08:10:10Z

ProcessLib/RichardsComponentTransport/RichardsComponentTransportFEM-impl.h

+
+        // saturation
+        double const pc_int_pt = -p_int_pt;
+        double const Sw = (pc_int_pt > 0)


Here, the same.

endJunction

⏩

- Copied ComponentTransport folder - Renamed ComponentTransport to RichardsComponentTransport

…tData.

Change the assembly of the local matrix for the pressure Kpp.

Change the assembly of the local mass matrix for the pressure Mpp.

Change the assembly for the pressure related entries of the local right hand side.

Summarize several times used mathematical expression into a variable.

Use variables for indices and size for accessing local matrices.

ogsbot · 2020-06-19T12:47:56Z

OpenGeoSys development has been moved to GitLab.

See this pull request on GitLab.

TomFischer added stop review 🛑 WIP 🏗️ labels Aug 28, 2017

TomFischer force-pushed the RichardsComponentTransport branch 2 times, most recently from 5977bb8 to cc043fa Compare August 31, 2017 12:20

TomFischer force-pushed the RichardsComponentTransport branch from cc043fa to 3fc308b Compare September 8, 2017 09:48

TomFischer force-pushed the RichardsComponentTransport branch 3 times, most recently from 4d5c437 to ca5f624 Compare September 19, 2017 05:33

TomFischer force-pushed the RichardsComponentTransport branch 6 times, most recently from 23b6963 to 021992f Compare October 4, 2017 04:00

TomFischer added please review and removed stop review 🛑 WIP 🏗️ labels Oct 4, 2017

endJunction reviewed Oct 10, 2017

View reviewed changes

TomFischer force-pushed the RichardsComponentTransport branch from 021992f to d28361c Compare October 10, 2017 09:56

endJunction reviewed Oct 10, 2017

View reviewed changes

wenqing reviewed Oct 13, 2017

View reviewed changes

TomFischer force-pushed the RichardsComponentTransport branch from 20a4868 to b5cddcf Compare October 16, 2017 06:01

wenqing approved these changes Oct 16, 2017

View reviewed changes

endJunction approved these changes Oct 16, 2017

View reviewed changes

TomFischer and others added 18 commits October 17, 2017 07:52

[PL] Initial RichardsComponentTransport process.

9920750

- Copied ComponentTransport folder - Renamed ComponentTransport to RichardsComponentTransport

[PL] RichardsComponentTransport: Add relative perm.

efd2dec

[PL] RichardsComponentTransport: Add p_c.

8a72e32

[PL] RichardsComponentTransport: Add saturation.

42ad38d

[PL] RichardsComponentTransport: Add mass_operator to IntegrationPoin…

399ebf6

…tData.

[PL] RichardsComponentTransport: Change Kpp.

e290db8

Change the assembly of the local matrix for the pressure Kpp.

[PL] RichardsComponentTransport: Change Mpp.

89c9964

Change the assembly of the local mass matrix for the pressure Mpp.

[PL] RichardsComponentTransport: Change Bp.

82fbe58

Change the assembly for the pressure related entries of the local right hand side.

[PL] RichardsComponentTransport: Add K_times_k_rel_over_mu.

909df2e

Summarize several times used mathematical expression into a variable.

[PL] RichardsComponentTransport: Update secondary var. computation.

55688c1

[PL] RichardsComponentTransport: Documentation.

4fcd1e9

Docu for Padilla et al., 1999

a704e00

[PL/RichardsComponentTransport] Enable Padilla tests.

c72095f

[PL/HT] Set pos to int pt in comp of Darcy vel.

49aea7e

[PL] RichardsComponentTransport: Save element id.

8dfe672

[PL/RichardsComponentTransport] Relax tolerances.

9eb7001

[PL/RichardsComponentTransport] Vars for matrix access.

0f10e94

Use variables for indices and size for accessing local matrices.

[PL/RichardsComponentTransport] Split impl. and decl. of local asm.

4d4ee32

TomFischer force-pushed the RichardsComponentTransport branch from b5cddcf to 4d4ee32 Compare October 17, 2017 06:03

TomFischer merged commit a80a987 into ufz:master Oct 17, 2017

TomFischer deleted the RichardsComponentTransport branch October 17, 2017 12:30

endJunction removed the please review label Oct 17, 2017

TomFischer restored the RichardsComponentTransport branch November 8, 2017 09:48

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Richards Component Transport #1929

Richards Component Transport #1929

TomFischer commented Aug 28, 2017 •

edited

Loading

chleh commented Oct 4, 2017

hydromarc commented Oct 9, 2017

endJunction Oct 10, 2017

endJunction Oct 10, 2017

endJunction Oct 10, 2017 •

edited

Loading

endJunction Oct 10, 2017

endJunction Oct 10, 2017

endJunction commented Oct 10, 2017 •

edited

Loading

TomFischer commented Oct 12, 2017 •

edited

Loading

wenqing Oct 13, 2017

TomFischer Oct 16, 2017

wenqing Oct 13, 2017

wenqing Oct 13, 2017

wenqing Oct 13, 2017

endJunction left a comment

ogsbot commented Jun 19, 2020

		@@ -0,0 +1,3 @@
		Specific body forces applied to fluid.

		It is usually used to apply gravitational forces. Given as a 2d or 3d vector.

Richards Component Transport #1929

Richards Component Transport #1929

Conversation

TomFischer commented Aug 28, 2017 • edited Loading

chleh commented Oct 4, 2017

hydromarc commented Oct 9, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

endJunction Oct 10, 2017 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

endJunction commented Oct 10, 2017 • edited Loading

TomFischer commented Oct 12, 2017 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

endJunction left a comment

Choose a reason for hiding this comment

ogsbot commented Jun 19, 2020

TomFischer commented Aug 28, 2017 •

edited

Loading

endJunction Oct 10, 2017 •

edited

Loading

endJunction commented Oct 10, 2017 •

edited

Loading

TomFischer commented Oct 12, 2017 •

edited

Loading