Parametric dmdc heat conduction #264

andersonw1 · 2024-01-29T19:06:34Z

Adds parametric DMDc for heat conduction.

Example changing parameter in state vector:
mpirun -np 8 parametric_dmdc_heat_conduction -s 1 -offline --kappa 0.5 -rdim 6
mpirun -np 8 parametric_dmdc_heat_conduction -s 1 -offline --kappa 0.75 -rdim 6
mpirun -np 8 parametric_dmdc_heat_conduction -s 1 -offline --kappa 1.25 -rdim 6
mpirun -np 8 parametric_dmdc_heat_conduction -s 1 -offline --kappa 1.5 -rdim 6

mpirun -np 8 parametric_dmdc_heat_conduction -s 1 --kappa 1 -online -predict -visit

Output:
Relative error of DMDc temperature (u) at t_final: 0.5 is 0.0025944158
Elapsed time for solving FOM: 1.006058e+00 second
Elapsed time for training DMDc: 7.535820e-01 second
Elapsed time for predicting DMDc: 3.593333e-03 second

Example changing parameter in control vector:
mpirun -np 8 parametric_dmdc_heat_conduction -s 1 -offline -amp-in 1 -rdim 6
mpirun -np 8 parametric_dmdc_heat_conduction -s 1 -offline -amp-in 2 -rdim 6
mpirun -np 8 parametric_dmdc_heat_conduction -s 1 -offline -amp-in 4 -rdim 6
mpirun -np 8 parametric_dmdc_heat_conduction -s 1 -offline -amp-in 5 -rdim 6

mpirun -np 8 parametric_dmdc_heat_conduction -s 1 -amp-in 3 -online -predict -visit

Output:
Relative error of DMDc temperature (u) at t_final: 0.5 is 0.0028885833
Elapsed time for solving FOM: 9.007740e-01 second
Elapsed time for training DMDc: 7.236116e-01 second
Elapsed time for predicting DMDc: 5.489291e-03 second

Some things I am suspicious of:

Changing the amplitudes of the sources seems to have very little effect on the final accuracy of the ROM. For example, in the second example above, if we predict:

mpirun -np 8 parametric_dmdc_heat_conduction -s 1 --kappa 1 -amp-in 1000 -online -predict

the relative error for the final temperature is 0.013669751 despite the fact that -amp-in 1000 is far outside our trained values for the inlet amplitude.

I changed the if statements in matrixinterpolator.cpp. I think other examples that call the matrix interpolator should still work with the new logic, since we are now just checking if dimensions match. I still need to verify this.

dylan-copeland · 2024-02-12T18:58:17Z

There is a lot of file output that is hard to delete. It would be more convenient to use a subdirectory for file output, as done in examples/dmd/de_dg_advection_greedy.cpp (search for io_dir).

dylan-copeland · 2024-02-12T19:05:47Z

lib/algo/DMDc.h

@@ -204,10 +238,11 @@ class DMDc
     * @param[in] dt d_dt
     * @param[in] t_offset d_t_offset
     * @param[in] state_offset d_state_offset
+     * @param[in] basis d_basis


d_basis -> basis
For backwards compatibility, should this have default value basis = nullptr? Before this PR, the member d_basis was set by the class, not the user. In what cases should this be input, and why is not overwritten by the class? Some documentation on this would be helpful.

Good point! I changed the line of code to match your suggestion and added some documentation in the hpp file (maybe it would be more helpful elsewhere?)

The idea is that for offline DMDc we compute the basis. For the interpolated DMDc we interpolate the basis, and then we need to store it so that we can later project the controls (see line 678 in DMDc.cpp)

dylan-copeland · 2024-02-12T19:07:15Z

lib/algo/ParametricDMDc.h

+                            dmdcs[0]->d_k,dmdcs[0]->d_dt,
+                            dmdcs[0]->d_t_offset, dmdcs[0]->d_state_offset, W);
+
+//    delete W;


Decide whether these pointers should be deleted?

I keep both of them. Added a commit to delete the commented lines

examples/dmd/parametric_dmdc_heat_conduction.cpp

andersonw1 · 2024-02-21T22:42:47Z

There is a lot of file output that is hard to delete. It would be more convenient to use a subdirectory for file output, as done in examples/dmd/de_dg_advection_greedy.cpp (search for io_dir).

The new commit adds an output directory using the same "io_dir" code as in the other example.

lib/algo/DMDc.cpp

ckendrick · 2024-02-22T14:10:27Z

examples/dmd/parametric_dmdc_heat_conduction.cpp

+
+    if (offline)
+    {
+        delete dmd_u;


I think dmd_u can be deleted for the online case as well, since it is allocated in getParametricDMDc().

We should introduce smart pointers in libROM (unique_ptr, shared_ptr), which is beyond the scope of this PR. For now, let's delete pointers as needed, but I think we have too many pointer issues and need to switch to smart pointers.

You're right, dmd_u is now always deleted

@dylan-copeland I definitely agree, we should probably open an issue for switching to smart pointers so we remember to come back to it in the future.

examples/dmd/parametric_dmdc_heat_conduction.cpp

lib/algo/ParametricDMDc.h

siuwuncheung

Thanks @andersonw1! The changes look good overall. I only have some minor comments in the formats. I am also curious if we can do a fair comparison between DMD and DMDc on the same example with external source, which we should expect DMDc works better.

siuwuncheung · 2024-03-05T04:10:46Z

lib/algo/DMDc.cpp

@@ -107,7 +107,7 @@ DMDc::DMDc(std::string base_file_name)

 DMDc::DMDc(std::vector<std::complex<double>> eigs, Matrix* phi_real,
           Matrix* phi_imaginary, Matrix* B_tilde, int k,
-           double dt, double t_offset, Vector* state_offset)
+           double dt, double t_offset, Vector* state_offset,Matrix* basis)


Add a space before Matrix?

siuwuncheung · 2024-03-05T04:12:11Z

lib/algo/DMDc.cpp

@@ -561,6 +569,7 @@ DMDc::constructDMDc(const Matrix* f_snapshots,
    // Calculate the projection initial_condition onto column space of d_basis.
    project(init, f_controls);

+


Redundant line.

siuwuncheung · 2024-03-05T04:14:43Z

lib/algo/DMDc.h

+     * @param[in] parameter_points  The training parameter points.
+     * @param[in] dmdcs              The DMD objects associated with
+     *                              each training parameter point.
+     * @param[in] controls     The matrices of controls from previous runs which we use to interpolate.


Non-uniform indents

siuwuncheung · 2024-03-05T04:15:36Z

lib/algo/ParametricDMDc.h

+ * @param[in] dmdcs              The DMDc objects associated with
+ *                              each training parameter point.
+ * @param[in] controls     The matrices of controls from previous runs which we use to interpolate.
+ * @param[in] controls_interpolated    The interpolated controls.


Non-uniform indents

siuwuncheung · 2024-03-05T04:16:03Z

lib/algo/ParametricDMDc.h

+ * @param[in] dmdc_paths         The paths to the saved DMD objects associated with
+ *                              each parameter point.
+ * @param[in] desired_point     The desired point at which to create a parametric DMD.
+ * @param[in] controls     The matrices of controls from previous runs which we use to interpolate.


Non-uniform indents

andersonw1 requested review from ptranq, ckendrick, dylan-copeland, ebchin, dreamer2368, chldkdtn and siuwuncheung January 29, 2024 19:33

andersonw1 added the RFR Ready for review label Jan 31, 2024

dylan-copeland reviewed Feb 12, 2024

View reviewed changes

examples/dmd/parametric_dmdc_heat_conduction.cpp Outdated Show resolved Hide resolved

ckendrick reviewed Feb 22, 2024

View reviewed changes

ckendrick approved these changes Feb 28, 2024

View reviewed changes

dylan-copeland reviewed Feb 29, 2024

View reviewed changes

lib/algo/ParametricDMDc.h Outdated Show resolved Hide resolved

dylan-copeland approved these changes Mar 1, 2024

View reviewed changes

siuwuncheung reviewed Mar 5, 2024

View reviewed changes

siuwuncheung approved these changes Apr 1, 2024

View reviewed changes

andersonw1 added 11 commits April 2, 2024 07:42

Updated CMakeUpdated CMakeUpdated CMakeUpdated CMake

07150f2

Algo files

e6faa9c

Added file and headers

372cd2b

Changes to header files

f1dad69

DMDc online/predict phases compile and give output

7bcc1d9

Cleanup comments/print statements

67bb83d

Started adding interpolation for controls

f0f6ef0

Closer to interpolating controls

19c708a

Controls are always interpolated now

62b308c

Controls correctly interpolated

73f1daf

Remove ds.store

5f3dddd

andersonw1 added 15 commits April 2, 2024 07:43

Cleanup

8805864

Changed logic in matrixinterpolator

0ffb90f

Updated documentation at beginning

defc26e

Changed documentation at start

baead5b

Changed example at start and renamed visit output directories

0002efc

astyle

f39eb1f

documentation and formatting

3879cbc

Formatting on parametricDMDc

66373c3

Added directory for outputs

cef9beb

Deleting pointers/making sure pointers are deleted

a42e190

Deleting controls

f659671

Astyle

69444fd

Copyright year and documentation in ParametricDMDc

9d121ca

formatting changes

2f2fe2f

astyle

3eceaa3

andersonw1 force-pushed the parametric_dmdc_heat_condution branch from 6ef8179 to 3eceaa3 Compare April 2, 2024 15:54

DMDc.cpp formatting

69a0c74

andersonw1 merged commit 336ae98 into master Apr 2, 2024
4 checks passed

andersonw1 deleted the parametric_dmdc_heat_condution branch April 10, 2024 20:13

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Parametric dmdc heat conduction #264

Parametric dmdc heat conduction #264

andersonw1 commented Jan 29, 2024 •

edited

Loading

dylan-copeland commented Feb 12, 2024

dylan-copeland Feb 12, 2024

andersonw1 Feb 19, 2024 •

edited

Loading

dylan-copeland Feb 12, 2024

andersonw1 Feb 19, 2024

andersonw1 commented Feb 21, 2024 •

edited

Loading

ckendrick Feb 22, 2024

dylan-copeland Feb 22, 2024

andersonw1 Feb 26, 2024

ckendrick Feb 28, 2024

siuwuncheung left a comment

siuwuncheung Mar 5, 2024

siuwuncheung Mar 5, 2024

siuwuncheung Mar 5, 2024

siuwuncheung Mar 5, 2024

siuwuncheung Mar 5, 2024

		@@ -561,6 +569,7 @@ DMDc::constructDMDc(const Matrix* f_snapshots,
		// Calculate the projection initial_condition onto column space of d_basis.
		project(init, f_controls);

Parametric dmdc heat conduction #264

Parametric dmdc heat conduction #264

Conversation

andersonw1 commented Jan 29, 2024 • edited Loading

dylan-copeland commented Feb 12, 2024

Choose a reason for hiding this comment

andersonw1 Feb 19, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

andersonw1 commented Feb 21, 2024 • 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

siuwuncheung left a 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

andersonw1 commented Jan 29, 2024 •

edited

Loading

andersonw1 Feb 19, 2024 •

edited

Loading

andersonw1 commented Feb 21, 2024 •

edited

Loading