Volumetric source terms #2220

TomFischer · 2018-09-25T07:19:15Z

First version of volumetric source terms. Source term function has to be defined on the entire domain.

chleh

Only minor things.

chleh · 2018-09-25T07:48:19Z

ProcessLib/SourceTerms/NodalSourceTerm.cpp

 {
-    DBUG("Assemble NodalSourceTerm.");
+    DBUG("Integrate NodalSourceTerm.");


Hmm... A nodal source term is not really integrated...

chleh · 2018-09-25T07:51:46Z

ProcessLib/SourceTerms/SourceTerm.h

+class SourceTerm
+{
+public:
+    SourceTerm(const NumLib::LocalToGlobalIndexMap& source_term_dof_table)


Minor: explicit. ✅

some parts are lost, it seems... But when you add mesh and variable's id to it, it's not needed then...

explicit still missing...

chleh · 2018-09-25T07:53:50Z

ProcessLib/SourceTerms/VolumetricSourceTerm.cpp

+
+#include "VolumetricSourceTerm.h"
+
+#include <cassert>


not needed I guess.:white_check_mark:

chleh · 2018-09-25T07:55:07Z

ProcessLib/SourceTerms/VolumetricSourceTermData.h

+        : volumetric_source_term(volumetric_source_term_)
+    {}
+
+    VolumetricSourceTermData(VolumetricSourceTermData&& other)


Maybe now we don't have to specify that explicitly anymore due to newer compilers. ✅

chleh · 2018-09-25T07:55:41Z

ProcessLib/SourceTerms/VolumetricSourceTermData.h

+
+struct VolumetricSourceTermData final
+{
+    VolumetricSourceTermData(


Minor: explicit ✅

chleh · 2018-09-25T08:04:04Z

ProcessLib/GroundwaterFlow/Tests.cmake

+    TESTER vtkdiff
+    REQUIREMENTS NOT (OGS_USE_LIS OR OGS_USE_MPI)
+    DIFF_DATA
+    square_1x1_quad_1e2_volumetricsourceterm_analytical_solution.vtu square_1e2_volumetricsourceterm_pcs_0_ts_1_t_1.000000.vtu analytical_solution pressure 1e-14 1e-16


Is the solution a linear slope or a constant? OGS is extremely close to the analytical result.

The analytical solution is h(x,y) = - 1/2 (x^2 + x) + 2.

chleh · 2018-09-25T08:05:50Z

...e/prj/process_variables/process_variable/source_terms/source_term/Volumetric/c_Volumetric.md

@@ -0,0 +1 @@
+Declares a source term to be associated to the entire domain.


Maybe Declares a source term that is defined on the entire domain. ✅

chleh · 2018-09-25T08:10:17Z

...le/prj/process_variables/process_variable/source_terms/source_term/Volumetric/t_parameter.md

@@ -0,0 +1,2 @@
+The name of the parameter that defines value that should be used for the source
+term.


the value.

Maybe a comment that nodal parameters are not supported would be appropriate. And maybe there should also be an addendum that the user should carefully check which physical quantity that parameter is. It's not necessarily the same as \dot{primary_variable}.

I do not understand what is meant by 'nodal parameter'. Can you explain, please?

Suggestion to keep the comment as is, and bake another PR handling this case.

chleh · 2018-09-25T08:13:36Z

web/content/docs/benchmarks/elliptic/groundwater-flow-dirichlet-volumetric-source-term.pandoc

+
+The solution of this problem is
+$$
+h(x,y) = - 1 \over 2 (x^2 + x) + 2.


the fraction seems to be set wrongly. Btw. in LaTex there is \frac{...}{...}. ✅

chleh · 2018-09-25T08:15:51Z

web/content/docs/benchmarks/elliptic/groundwater-flow-dirichlet-volumetric-source-term.pandoc

+  [menu.benchmarks]
+    parent = "elliptic"
+
+++


The test is OK and nicely documented, but for an actual user it might be more instructive to have some physical quantities with some actual units.

You are right. I'll ask @WaltherM or @Thomas-TK for a more practical test.

endJunction · 2018-09-25T11:14:16Z

ProcessLib/SourceTerms/CreateVolumetricSourceTerm.h

+#pragma once
+
+#include <memory>
+#include "ProcessLib/Process.h"


Use forward decls like in CreateBoundaryCondition.h ... ✅

endJunction · 2018-09-25T11:30:22Z

ProcessLib/SourceTerms/NodalSourceTerm.h


 private:
-    NumLib::LocalToGlobalIndexMap const& _dof_table;
    std::size_t const _bulk_mesh_id;
    MeshLib::Mesh const& _st_mesh;
    int const _variable_id;


I'd move both, the variable's id and the mesh to the base class. The mesh is the domain of definition.
The variable's id is also part of domain of definition.

I also thought about this. The VolumetricSourceTerm is also derived from the class SourceTerm and it doesn't need the mesh and the variable id as attributes. In the VolumetricSourceTerm implementation the mesh and the variable id are only used in the constructor.

OK, got it. Keep it as is.

endJunction · 2018-09-25T12:13:46Z

ProcessLib/SourceTerms/VolumetricSourceTerm.cpp

+            source_term_dof_table.getNumberOfVariableComponents(variable_id))
+    {
+        OGS_FATAL(
+            "Variable id or component id too high. Actual values: (%d, %d), "


Could be split in two errors...

endJunction · 2018-09-25T12:15:30Z

ProcessLib/SourceTerms/VolumetricSourceTerm.h

+
+#include "SourceTerm.h"
+#include "VolumetricSourceTermFEM.h"
+#include "VolumetricSourceTermData.h"


Some of the includes could be replaced with fwd decls. ✅

endJunction · 2018-09-25T12:18:01Z

ProcessLib/SourceTerms/VolumetricSourceTermData.h

+    VolumetricSourceTermData(VolumetricSourceTermData&&) = default;
+
+    //! Copies are forbidden.
+    VolumetricSourceTermData(VolumetricSourceTermData const&) = delete;


Question: If the base class has a deleted assignment op or ctor, is a default assignment op or ctor created implicitly in the derived class?

cppreference:
A defaulted copy assignment operator for class T is defined as deleted if T has a virtual base class that cannot be copy-assigned (selects a deleted or inaccessible function).

I'm unsure if this is an answer to your question?

endJunction · 2018-09-25T12:26:22Z

web/content/docs/benchmarks/elliptic/groundwater-flow-dirichlet-volumetric-source-term.pandoc

+### Comparison of the analytical solution and the computed solution
+
+{{< img src="../square_1e2_volumetricsourceterm_pcs_0_ts_1_t_1.000000_h_AnalyticalSolution_VolumetricSourceTerm.png" >}}
+{{< img src="../square_1e2_volumetricsourceterm_pcs_0_ts_1_t_1.000000_Diff_h_AnalyticalSolution_VolumetricSourceTerm.png" >}}


Description of both figures is missing. The analytical solution line is not visible. In the second figure h should be pressure, I guess.

In the formulae above the primary variable was 'h'. I renamed h -> p to make it consistent.

Preparation for the SourceTerm interface class.

endJunction · 2018-09-26T15:53:39Z

ProcessLib/SourceTerms/VolumetricSourceTerm.h

+#include <vector>
+
+#include "SourceTerm.h"
+#include "VolumetricSourceTerm.h"


self include? Or am I missing the difference? ✅

endJunction · 2018-09-26T15:55:27Z

ProcessLib/SourceTerms/SourceTerm.h

+    virtual void integrate(const double t, GlobalVector& b) const = 0;
+
+    virtual ~SourceTerm() = default;
+protected:


Why protected? Could be private as far as I see. It can only be set in the ctor...

Protected because the attribute should be used in the derived classes.

endJunction · 2018-09-26T15:57:30Z

ProcessLib/SourceTerms/VolumetricSourceTerm.h

+
+private:
+    Parameter<double> const& _volumetric_source_term;
+    std::vector<std::unique_ptr<VolumetricSourceTermLocalAssemblerInterface>>


I'd think a fwd-decl to the local asm interface would be sufficient...

The fwd-decl leads to:

error: invalid application of 'sizeof' to an incomplete type 'ProcessLib::VolumetricSourceTermLocalAssemblerInterface'

endJunction · 2018-09-26T15:58:05Z

ProcessLib/SourceTerms/VolumetricSourceTermFEM.h

+
+template <typename ShapeFunction, typename IntegrationMethod,
+          unsigned GlobalDim>
+class VolumetricSourceTermLocalAssembler


endJunction · 2018-09-26T16:00:39Z

ProcessLib/SourceTerms/VolumetricSourceTermFEM.h

+            auto const st_val = _volumetric_source_term(t, pos)[0];
+
+            _local_rhs.noalias() +=
+                sm.N * st_val * sm.detJ * sm.integralMeasure * wp.getWeight();


It would be much better to store only the N vector in the ip data (not all of the shape matrices) and precompute the integration_weight = detJ * intMeasure * w_ip.

Introduced class IntegrationPointData, see commit 3ba3927.

endJunction

Looks good in general but still few things either are not yet uploaded, or forgotten.

First approach use the original bulk mesh as integration domain.

ogsbot · 2020-06-19T12:54:53Z

OpenGeoSys development has been moved to GitLab.

See this pull request on GitLab.

chleh approved these changes Sep 25, 2018

View reviewed changes

endJunction reviewed Sep 25, 2018

View reviewed changes

TomFischer added 4 commits September 26, 2018 11:40

[PL/NodalSourceTerm] Renames for common interface.

584fe4a

Preparation for the SourceTerm interface class.

[PL/SourceTerms] Introduce source term interface.

e73b277

[PL] Remove unused include.

9377147

[PL] Use SourceTerm instead of NodalSourceTerm.

fc4ea3d

TomFischer force-pushed the VolumetricSourceTerms branch from 63c3fab to 3e91419 Compare September 26, 2018 10:20

endJunction reviewed Sep 26, 2018

View reviewed changes

endJunction approved these changes Sep 26, 2018

View reviewed changes

TomFischer force-pushed the VolumetricSourceTerms branch from 3e91419 to 3ba3927 Compare September 27, 2018 05:33

TomFischer added 7 commits September 27, 2018 07:45

[PL] Impl. of volumetric source terms.

9b10fde

First approach use the original bulk mesh as integration domain.

[PL] Include VolumetricSourceTerm creation.

031edd6

[T] Test volumetric source terms on quad_1e2 mesh.

def7f5b

[T] Test volumetric source terms on quad_1e3 mesh.

3608235

[Docu/web] Add volumetric source term docu.

dde3e3c

[Docu] Prj. file params for volumetric source term.

7f028aa

[PL] IntegrationPointData for frequently used data.

79c3e71

TomFischer force-pushed the VolumetricSourceTerms branch from 3ba3927 to 79c3e71 Compare September 27, 2018 05:46

endJunction merged commit 6ec6371 into ufz:master Sep 27, 2018

TomFischer deleted the VolumetricSourceTerms branch September 27, 2018 11:42

		@@ -0,0 +1 @@
		Declares a source term to be associated to the entire domain.

		@@ -0,0 +1,2 @@
		The name of the parameter that defines value that should be used for the source
		term.


		#include "VolumetricSourceTerm.h"

		#include <cassert>

Volumetric source terms #2220

Volumetric source terms #2220

Conversation

TomFischer commented Sep 25, 2018

chleh left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

chleh Sep 25, 2018 • edited by TomFischer Loading

Choose a reason for hiding this comment

endJunction Sep 25, 2018 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

chleh Sep 25, 2018 • edited by TomFischer Loading

Choose a reason for hiding this comment

chleh Sep 25, 2018 • edited by TomFischer Loading

Choose a reason for hiding this comment

chleh Sep 25, 2018 • edited by TomFischer Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

chleh Sep 25, 2018 • edited by TomFischer 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

chleh Sep 25, 2018 • edited by TomFischer Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

endJunction Sep 25, 2018 • edited by TomFischer 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 Sep 25, 2018 • edited by TomFischer 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 Sep 26, 2018 • edited by TomFischer 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 Sep 26, 2018 • edited by TomFischer Loading

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

chleh Sep 25, 2018 •

edited by TomFischer

Loading

endJunction Sep 25, 2018 •

edited

Loading

chleh Sep 25, 2018 •

edited by TomFischer

Loading

chleh Sep 25, 2018 •

edited by TomFischer

Loading

chleh Sep 25, 2018 •

edited by TomFischer

Loading

chleh Sep 25, 2018 •

edited by TomFischer

Loading

chleh Sep 25, 2018 •

edited by TomFischer

Loading

endJunction Sep 25, 2018 •

edited by TomFischer

Loading

endJunction Sep 25, 2018 •

edited by TomFischer

Loading

endJunction Sep 26, 2018 •

edited by TomFischer

Loading

endJunction Sep 26, 2018 •

edited by TomFischer

Loading