[WIP] Aero-Structural Discrete Adjoint Sensitivities and Python Wrapper Developments #1750

patelha57 · 2022-09-05T13:11:46Z

Motivation

Fundamental bottlenecks exist for industrial adoption of high-fidelity physics codes with fully coupled discrete adjoint sensitivity analysis. These bottlenecks include a lack of flexibility, modularity, and robustness of the computational tools, as well as the potential startup development costs needed to implement and verify the MDAO features. The purpose of this work is to facilitate the coupling of SU2 with external structural codes (e.g. NASTRAN, TACS, Airbus structural suite Lagrange) for gradient-based aerodynamic shape and structural sizing optimization using dedicated frameworks (e.g. OpenMDAO).

Startup development costs include the creation of modular tools that are designed to be driven by another framework like OpenMDAO, rather than to drive execution themselves. However, leveraging those MDAO frameworks assumes that software codes and modules to be coupled exist, have appropriate data structures, execution and query APIs, and are wrapped in Python for flexibility and ease of use. The proposed changes were motivated with the goal of making SU2 more modular and flexible, particularly to facilitate its integration into large-scale MDAO frameworks. For more details on the motivation, methodology, and verification & validation results, please refer to our paper from AIAA Aviation 2022.

Proposed Changes

@aa-g and I propose the following code updates:

Implementation of residual-based discrete adjoint solver as CDiscAdjResidualSolver
a. New config option: KIND_DISC_ADJ
Enhancements and standardization of SU2 Python API
a. Add pysu2/pysu2ad methods
b. Standardize Python API and function names
c. Overloaded getter/setter methods to make data handling more flexible

Related Work

These efforts are related to Issue #1262, Pull Request #1300, and Discussion #1325.

PR Checklist

I am submitting my contribution to the develop branch.
My contribution generates no new compiler warnings (try with --warnlevel=3 when using meson).
My contribution is commented and consistent with SU2 style (https://su2code.github.io/docs_v7/Style-Guide/).
I used the pre-commit hook to prevent dirty commits and used pre-commit run --all to format old commits.
I have added a test case that demonstrates my contribution, if necessary.
I have updated appropriate documentation (Tutorials, Docs Page, config_template.cpp), if necessary.

…eanup

…rs/setters

…ith CDriver

patelha57 · 2023-10-30T19:36:39Z

Sounds good, with that it will be easier to reason about when the recording types are created and used, right now someone would have to sit down and reverse-engineer the process. My intuition is that we can use some of the machinery introduced for multizone (partial tape evaluation) to simplify the recording management. And just to be clear I am very interested in having this feature in the code for comparison with the FP approach.

That's exactly right. We started investigating the multizone driver a while back but we didn't get very far... It's been a while so i will need a refresher.

Sure, there's broader interest in comparison with FP. Ole, Nico, and I talked about doing a detailed characterization last summer and we briefly brought it up in Varenna last week

SU2_CFD/src/solvers/CDiscAdjResidualSolver.cpp

SU2_CFD/src/drivers/CDiscAdjSinglezoneDriver.cpp

SU2_CFD/include/drivers/CDiscAdjSinglezoneDriver.hpp

Co-authored-by: Pedro Gomes <38071223+pcarruscag@users.noreply.github.com>

patelha57 · 2024-02-13T19:31:18Z

SU2_CFD/include/drivers/CDiscAdjSinglezoneDriver.hpp

+  CPreconditioner<Scalar>* PrimalPreconditioner = nullptr;
+  CSysMatrixVectorProduct<Scalar>* PrimalJacobian = nullptr;
+
+  class LinOperator : public CMatrixVectorProduct<Scalar> {


A new compiler warning is generated...

[738/745] Generating 'SU2_PY/pySU2/_pysu2.so.p/pySU2.cxx'. ../SU2_PY/pySU2/../../SU2_CFD/include/drivers/CDiscAdjSinglezoneDriver.hpp:91: Warning 325: Nested class not currently supported (LinOperator ignored) ../SU2_PY/pySU2/../../SU2_CFD/include/drivers/CDiscAdjSinglezoneDriver.hpp:102: Warning 325: Nested class not currently supported (LinPreconditioner ignored) [744/745] Installing files.

SU2_CFD/src/drivers/CDiscAdjSinglezoneDriver.cpp

+    }
+
+    CopiedJacobian.TransposeInPlace();
+    PrimalJacobian = new CSysMatrixVectorProduct<Scalar>(CopiedJacobian, geometry, config);


patelha57 · 2024-02-14T00:33:06Z

All the regression and unit tests are passing now!

patelha57 · 2024-02-27T19:22:26Z

The following functions in python_wrapper_structure.cpp could be implemented as CPyWrapperMatrixView:

Adjoints

- GetMarkerAdjointForces()

Sensitivities

- GetCoordinatesCoordinatesSensitivities()
- GetMarkerCoordinatesDisplacementsSensitivities()
- GetObjectiveFarfieldVariablesSensitivities()
- GetResidualsFarfieldVariablesSensitivities()
- GetObjectiveStatesSensitivities()
- GetResidualsStatesSensitivities()
- GetForcesStatesSensitivities()
- GetObjectiveCoordinatesSensitivities()
- GetResidualsCoordinatesSensitivities()
- GetForcesCoordinatesSensitivities()
- GetMarkerObjectiveDisplacementsSensitivities()
- GetMarkerResidualsDisplacementsSensitivities()
- GetMarkerForcesDisplacementsSensitivities()
- GetMarkerForceSensitivities()

Flow solution

- GetHeatFluxes()
- GetMarkerHeatFluxes()
- GetNonequilibriumMassFractions()
- GetVibrationalTemperatures()

pcarruscag · 2024-02-27T19:49:02Z

SU2_CFD/include/drivers/CDiscAdjSinglezoneDriver.hpp

+  /*!
+   * \brief Update the primal geometry (does not include mesh deformation).
+   */
+  void UpdateGeometry(void);


Remove the (voids) please.

Suggested change

void UpdateGeometry(void);

void UpdateGeometry();

SU2_CFD/include/drivers/CDriver.hpp

pcarruscag · 2024-02-27T19:50:50Z

SU2_CFD/include/drivers/CDriver.hpp

+  vector<passivedouble> GetHeatFluxes(unsigned long iPoint) const;
+
+  /*!
+   * \brief Get the heat fluxes at a marker vertex.
+   * \param[in] iMarker - Marker index.
+   * \param[in] iVertex - Marker vertex index.
+   * \return Vertex heat fluxes (nVertex, nDim).
+   */
+  vector<passivedouble> GetMarkerHeatFluxes(unsigned short iMarker, unsigned long iVertex) const;
+
+  /*!
+   * \brief Get the normal heat fluxes at a marker vertex.
+   * \param[in] iMarker - Marker index.
+   * \param[in] iVertex - Marker vertex index.
+   * \return Vertex normal heat fluxes (nVertex, nDim).
+   */
+  passivedouble GetMarkerNormalHeatFluxes(unsigned short iMarker, unsigned long iVertex) const;


Aren't these duplicating an existing function?

pcarruscag · 2024-02-27T19:51:18Z

SU2_CFD/include/drivers/CDriver.hpp

+   * \param[in] iVertex - Marker vertex index.
+   * \param[in] values - Vertex normal heat flux.
+   */
+  void SetMarkerNormalHeatFluxes(unsigned short iMarker, unsigned long iVertex, passivedouble values);


Duplication?

pcarruscag · 2024-02-27T19:51:57Z

SU2_CFD/include/drivers/CDriver.hpp

+  unsigned long GetNumberNonequilibriumSpecies() const;
+
+  /*!
+   * \brief Get the number of nonequilibrium conservative state variables.
+   * \return Number of nonequilibrium conservative state variables.
+   */
+  unsigned long GetNumberNonequilibriumStateVariables() const;
+
+  /*!
+   * \brief Get the number of nonequilibrium primitive state variables.
+   * \return Number of nonequilibrium primitive state variables.
+   */
+  unsigned short GetNumberNonequilibriumPrimitiveVariables() const;
+
+  /*!
+   * \brief Get nonequilibrium chemical mass fractions.
+   * \param[in] iPoint - Point index.
+   * \return Nonequilibrium chemical mass fractions (nSpecies).
+   */
+  vector<passivedouble> GetNonequilibriumMassFractions(unsigned long iPoint) const;
+
+  /*!
+   * \brief Get nonequilibrium vibrational electronic temperatures.
+   * \return Nonequilibrium vibrational electronic temperatures (nPoint).
+   */
+  vector<passivedouble> GetVibrationalTemperatures() const;


Is this not available via the access to primitive variables?

pcarruscag · 2024-02-27T19:53:23Z

SU2_CFD/src/python_wrapper_structure.cpp

@@ -58,24 +59,599 @@ void CDriver::PreprocessPythonInterface(CConfig** config, CGeometry**** geometry
  }
 }

+/////////////////////////////////////////////////////////////////////////////
+/* Functions related to the far-field flow variables.                      */


Looks like you have weird symbols here

pcarruscag · 2024-02-27T19:53:54Z

SU2_CFD/src/python_wrapper_structure.cpp

+    SU2_MPI::Error("Vertex index exceeds mesh size.", CURRENT_FUNCTION);
+  }
+
+  return SU2_TYPE::GetValue(solver_container[ZONE_0][INST_0][MESH_0][FLOW_SOL]->GetNodes()->GetSoundSpeed(iPoint));


Isn't this part of the primitives too?

pcarruscag · 2024-02-27T19:54:53Z

SU2_CFD/src/python_wrapper_structure.cpp

+  return values;
+}
+
+void CDriver::SetMarkerAdjointForces(unsigned short iMarker, unsigned long iVertex, vector<passivedouble> values) {


These also look duplicated

WallyMaier and others added 30 commits May 7, 2020 14:20

adding GetStates() function

9491e2b

fixing output + adding GetAIP

92ae46e

adding in area extraction

5146f8d

clearing ninja/meson stuff

e9408c7

adding HYBRID option for AIP averaging

85ba068

adding in Average normal velocity for DA

d2db580

fixing function name

42d9d78

Add methods to set AoA and AoS

68eb015

Add method to get conservative states

1c02df0

Add method to get surface coordinates

b984e67

Add method to get fluid volume coordinates

2d37896

Add method to get flow residuals

8b43265

Merge code from feature_QUAD and resolve conflicts

7b4fcf0

Fix compilation errors, comment out GetAIP()

4239d0c

Add method to get aero forces

1eb61de

Generalize GetConservativeStates() and GetResiduals() for 2D cases

6167cf1

Add method to set conservative states

363f288

Add methods to get flow states and flow residuals at marker, minor cl…

588fa38

…eanup

Add methods to get aero moment coefficients

1d1c062

Implement residual-based DA solver, add pysu2 methods for sensitivites

9842113

Merge small changes

7b3999c

[WIP] Implement dxadua sensitivity

da2933b

Add pysu2 methods for dfa/dua and dxv/dua sens

eb64f51

Add new methods to CDeformationDriver

17c4cd7

[WIP] Consolidate and extend implementation of Python interface gette…

5cc8301

…rs/setters

[WIP] Match implementation of getters/setters in CDeformationDriver w…

440710f

…ith CDriver

Implement element ID and connectivities in Python API

98440d5

[WIP] Corrections to the previous commit

099651d

Fix bugs introduced in previous commits

0828142

Reintroduce sensitivities w.r.t. mesh coordinates

b2f6e1a

patelha57 added 2 commits October 26, 2023 23:19

Update function names and fix ENUM_DISC_ADJ_TYPE usage

add0fe4

Add GetMarkerForceSensitivities() to CDriver.hpp

af113be

patelha57 added 4 commits October 30, 2023 20:47

Small documentation fix for CDriver.hpp

dbfd39f

Add name to AUTHORS.md

1467aab

Update documentation in CDriver.hpp

918181f

Revert 'Add matrix view for Residuals and MarkerResiduals'

62f2969

github-advanced-security bot found potential problems Nov 3, 2023

View reviewed changes

patelha57 added 5 commits January 17, 2024 17:18

Merge 'develop' branch into 'feature_pysu2_fsi_adjoint'

60f3328

Cleanup CDiscAdjSinglezoneDriver.cpp

1b52055

Fix uninitialized variable error in regression tests

ca15ae5

Fix small errors with regression tests

193d27a

Merge 'develop' branch into 'feature_pysu2_fsi_adjoint'

7f448c7

patelha57 commented Feb 7, 2024

View reviewed changes

SU2_CFD/src/drivers/CDiscAdjSinglezoneDriver.cpp Show resolved Hide resolved

pcarruscag reviewed Feb 10, 2024

View reviewed changes

SU2_CFD/include/drivers/CDiscAdjSinglezoneDriver.hpp Outdated Show resolved Hide resolved

patelha57 and others added 3 commits February 13, 2024 09:03

Fix type conversion error in CDiscAdjSinglezoneDriver.hpp

fe7f82c

Co-authored-by: Pedro Gomes <38071223+pcarruscag@users.noreply.github.com>

Apply clang format

99f8def

Fix function names in Python regression tests

8d59b45

patelha57 commented Feb 13, 2024

View reviewed changes

patelha57 added 3 commits February 13, 2024 13:07

Fix UpdateBoundaryCondition() in Python regression test

b78601c

Remove multiple spaces after operator in Python testcases

f50e216

Fix angle of attack sensitivity in regression tests

96e45c7

github-advanced-security bot found potential problems Feb 14, 2024

View reviewed changes

Merge 'develop' branch into 'feature_pysu2_fsi_adjoint'

d7c398e

pcarruscag reviewed Feb 27, 2024

View reviewed changes

patelha57 and others added 3 commits March 20, 2024 10:55

Merge 'develop' branch into 'feature_pysu2_fsi_adjoint'

57423b7

Merge 'develop' branch into 'feature_pysu2_fsi_adjoint'

c446d44

Merge branch 'develop' into feature_pysu2_fsi_adjoint

c087e23

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[WIP] Aero-Structural Discrete Adjoint Sensitivities and Python Wrapper Developments #1750

[WIP] Aero-Structural Discrete Adjoint Sensitivities and Python Wrapper Developments #1750

patelha57 commented Sep 5, 2022 •

edited

patelha57 commented Oct 30, 2023

patelha57 Feb 13, 2024

patelha57 commented Feb 14, 2024

patelha57 commented Feb 27, 2024

pcarruscag Feb 27, 2024

pcarruscag Feb 27, 2024

pcarruscag Feb 27, 2024

pcarruscag Feb 27, 2024

pcarruscag Feb 27, 2024

pcarruscag Feb 27, 2024

pcarruscag Feb 27, 2024

[WIP] Aero-Structural Discrete Adjoint Sensitivities and Python Wrapper Developments #1750

Are you sure you want to change the base?

[WIP] Aero-Structural Discrete Adjoint Sensitivities and Python Wrapper Developments #1750

Conversation

patelha57 commented Sep 5, 2022 • edited

Motivation

Proposed Changes

Related Work

PR Checklist

patelha57 commented Oct 30, 2023

Choose a reason for hiding this comment

patelha57 commented Feb 14, 2024

patelha57 commented Feb 27, 2024

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

Choose a reason for hiding this comment

Choose a reason for hiding this comment

patelha57 commented Sep 5, 2022 •

edited