Petsc DAE Solver Utils #552

eslickj · 2021-10-14T00:30:43Z

Summary/Motivation:

This adds utilities mainly for the PETSc time stepping (TS) solvers. I'll attach examples for now and write some clean official examples for the examples repo.

This depends on Pyomo/pyomo#2152 (which has been merged, but you will need a recent development version of Pyomo).

Examples

car_example1.zip modified version of the car example from Pyomo DAE
chemakzo.zip from: https://archimede.dm.uniba.it/~testset/report/chemakzo.pdf, this one is a little more interesting. Some stuff happens very fast and it would be harder to solve in the fully discretized form without some careful tuning.
pid.zip IDAES model, this is the test for the PID controller solved with the PETSc solver. Full IDAES model with a few units, controller, and external function calls for steam properties.
pendulum.zip
Used to test constraint and variable scaling in DAEs.

Legal Acknowledgement

By contributing to this software project, I agree to the following terms and conditions for my contribution:

I agree my contributions are submitted under the license terms described in the LICENSE.txt file at the top level of this directory.
I represent I am authorized to make the contributions and grant the license. If my employer has rights to intellectual property that includes these contributions, I represent that I have received permission to make contributions and grant the required license on behalf of that employer.

idaes/core/solvers/petsc.py

codecov · 2021-11-08T22:00:18Z

Codecov Report

Merging #552 (c87818e) into main (8134c40) will increase coverage by 0.14%.
The diff coverage is 85.46%.

@@            Coverage Diff             @@
##             main     #552      +/-   ##
==========================================
+ Coverage   64.23%   64.37%   +0.14%     
==========================================
  Files         362      363       +1     
  Lines       58340    58744     +404     
  Branches    10660    10744      +84     
==========================================
+ Hits        37473    37816     +343     
- Misses      18975    19014      +39     
- Partials     1892     1914      +22

Impacted Files	Coverage Δ
idaes/util/download_bin.py	`47.40% <14.28%> (-1.82%)`	⬇️
idaes/core/solvers/petsc.py	`84.08% <84.08%> (ø)`
idaes/config.py	`81.65% <100.00%> (+1.25%)`	⬆️
idaes/core/solvers/features.py	`96.46% <100.00%> (+10.74%)`	⬆️
idaes/ver.py	`61.53% <0.00%> (-4.62%)`	⬇️
idaes/ui/fsvis/model_server.py	`83.78% <0.00%> (-1.63%)`	⬇️

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 8134c40...c87818e. Read the comment docs.

eslickj · 2021-11-09T17:15:56Z

I added another example based on the steam tank PID controller test (files attached to PR description under examples). So IDAES models seem to work. I'm ready to try some more, if anyone has something to try.

Robbybp · 2021-11-18T19:11:52Z

@eslickj Is the example on the steam tank part of this PR? I don't see it. I'd like an example/test I can run so I don't break anything when I update the "set_dae_suffixes" method.

eslickj · 2021-11-19T14:28:47Z

@Robbybp, it's not in the PR yet, but it's attached to the PR description (the pid.zip link https://github.com/IDAES/idaes-pse/files/7506724/pid.zip).

dallan-keylogic

Remind me what happens when a derivative variable is fixed to a nonzero value?

idaes/core/solvers/tests/test_petsc.py

eslickj · 2022-01-13T19:05:59Z

Remind me what happens when a derivative variable is fixed to a nonzero value?

It would turn the differential equation into an algebraic equation whether the derivative is 0 and any other value.

Robbybp · 2022-01-13T20:09:13Z

idaes/core/solvers/petsc.py

+                _sub_problem_scaling_suffix(m, t_block)
+                with idaeslog.solver_log(solve_log, idaeslog.INFO) as slc:
+                    res = solver_snes.solve(t_block, tee=slc.tee)
+        res_list.append(res)


Should this be indented twice more? To be inside the if len(constraints) > 0 branch?

Robbybp

I think this is good to go.

… petsc_dae

Robbybp · 2022-01-13T20:37:38Z

idaes/core/solvers/petsc.py

+            # We need to check if there are derivatives in the problem before
+            # sending this to the solver.  We'll assume that if you are using
+            # this and don't have any differential equations, you are making a
+            # mistake.
+            if len(differential_vars) < 1:
+                raise RuntimeError(
+                    "No differential equations found, you do not need a DAE solver.")


This sounds good, but should this check happen earlier? E.g. at the end of _get_derivative_differential_data_map?

I was thinking that too, but unfortunately a derivative could be fixed for one time step and not another. We could possibly write the check up front, but I'm not sure it's worth the overhead.

Wait maybe it's not that hard. I should be able to check that there is at least one derivative for each time point. I think I can improve this. I'll also add a test for this.

What happens if you send a model with no derivatives to Petsc? Does it just solve the square system?

With your comment in mind, I'm fine with the check where it is, but I'll leave it up to you.

Okay changed my mind again. It's probably not worth the trouble of checking this earlier.

What happens if you send a model with no derivatives to Petsc? Does it just solve the square system?

It should but it crashes. Most likely I messed up and assumed I had at least one derivative. I’ll sort that out in the solver part eventually. For now this check will stop it from getting there.

eslickj · 2022-01-13T23:12:59Z

I added tests for a few more cases. I have 5 tests based on @Robbybp's example. 2 work and 3 are caught with errors and raise exceptions. I think this is ready, at least really close.

Once we get this in, I'll put out new binaries with a few very minor improvements and a couple example notebooks in the examples repo.

Robbybp

Looks great! Thanks for all the work on this interface, John. This will be very useful.

Robbybp · 2022-01-14T23:32:26Z

idaes/core/solvers/tests/test_petsc.py

+def rp_example5():
+    m = pyo.ConcreteModel()
+
+    m.time = pyodae.ContinuousSet(initialize=(0.0, 10.0))
+    m.x = pyo.Var(m.time, initialize=1)
+    m.u = pyo.Var(m.time, initialize=1)
+    m.dxdt = pyodae.DerivativeVar(m.x, wrt=m.time)
+
+    def diff_eq1_rule(m, t):
+        return m.dxdt[t] == m.x[t]**2 - m.u[t]
+    m.diff_eq1 = pyo.Constraint(m.time, rule=diff_eq1_rule)
+
+    discretizer = pyo.TransformationFactory('dae.finite_difference')
+    discretizer.apply_to(m,nfe=1,scheme='BACKWARD')
+
+    m.u[0].fix(1.0)
+    m.dxdt[:].fix(2)
+
+    return m


What's the difference between rp_example5 and rp_example3?

You know, I don't see a difference and can't remember. I think I must have copied it and meant to change it to test another case, and now I cannot recall what it was for. Doing to many things at once I guess. If the tests pass I'll merge this and make another small PR to update or remove the test if I still can't remember why it's there.

eslickj · 2022-01-15T00:02:02Z

Looks great! Thanks for all the work on this interface, John. This will be very useful.

Thanks! I think now that this is squared away, we can use the same method to hook up other integrators. We can come up with a standard for explicit and implicit-explicit methods too. I have some ideas I may propose if I find time.

Add PETSc solver utils.

f48e0db

eslickj requested review from jsiirola, Robbybp, andrewlee94 and blnicho October 14, 2021 00:30

eslickj self-assigned this Oct 14, 2021

eslickj marked this pull request as draft October 14, 2021 00:33

John Eslick added 4 commits October 13, 2021 20:55

Add doc string

22e06e9

Use solver logging

09d3596

Minor fix up of solver tests.

8a4dfaa

Arg name change in Pyomo

ee947ac

Robbybp reviewed Oct 15, 2021

View reviewed changes

idaes/core/solvers/petsc.py Outdated Show resolved Hide resolved

Robbybp reviewed Oct 15, 2021

View reviewed changes

idaes/core/solvers/petsc.py Outdated Show resolved Hide resolved

ksbeattie added the Priority:Normal Normal Priority Issue or PR label Oct 15, 2021

John Eslick added 4 commits October 20, 2021 15:54

Merge branch 'main' of https://github.com/idaes/idaes-pse into petsc_dae

5d4aa4a

Initial stub of soec model and docs

5f79fc1

Merge branch 'main' of https://github.com/idaes/idaes-pse into soec

8d9a78d

Merge branch 'main' of https://github.com/idaes/idaes-pse into petsc_dae

6c6ccd9

Modify default PETSc DAE options

6f2171a

Add docs about using WSL solvers on Windows with external functions

eb58dca

Doc progress

f17449e

John Eslick added 2 commits December 1, 2021 13:42

Merge branch 'main' of https://github.com/idaes/idaes-pse into soec

33147e4

Work in progress on non-isothermal dynamic sofc/soec

9396f30

dallan-keylogic previously approved these changes Jan 13, 2022

View reviewed changes

idaes/core/solvers/tests/test_petsc.py Outdated Show resolved Hide resolved

remove extra imports

029be31

eslickj dismissed dallan-keylogic’s stale review via 029be31 January 13, 2022 19:02

Robbybp reviewed Jan 13, 2022

View reviewed changes

indent res_list.append to proper level

f31a6e3

Robbybp previously approved these changes Jan 13, 2022

View reviewed changes

John Eslick added 2 commits January 13, 2022 15:29

Add check for no differential equations

4a2216f

Merge branch 'petsc_dae' of https://github.com/eslickj/idaes-pse into…

d5bd87d

… petsc_dae

eslickj dismissed Robbybp’s stale review via d5bd87d January 13, 2022 20:31

dallan-keylogic previously approved these changes Jan 13, 2022

View reviewed changes

Robbybp reviewed Jan 13, 2022

View reviewed changes

Add a couple more tests trying to cover more edges

ce68741

eslickj dismissed dallan-keylogic’s stale review via ce68741 January 13, 2022 21:52

Add tests

705804f

report time at which no differential equations are found

c87818e

eslickj requested review from Robbybp and dallan-keylogic January 14, 2022 19:17

dallan-keylogic approved these changes Jan 14, 2022

View reviewed changes

Robbybp approved these changes Jan 14, 2022

View reviewed changes

eslickj added approved This PR has been approved CI:run-integration triggers_workflow: Integration labels Jan 15, 2022

idaes-build removed the CI:run-integration triggers_workflow: Integration label Jan 15, 2022

eslickj enabled auto-merge January 15, 2022 01:39

eslickj merged commit ef82f17 into IDAES:main Jan 15, 2022

eslickj deleted the petsc_dae branch February 2, 2022 18:54

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Petsc DAE Solver Utils #552

Petsc DAE Solver Utils #552

eslickj commented Oct 14, 2021 •

edited

Loading

codecov bot commented Nov 8, 2021 •

edited

Loading

eslickj commented Nov 9, 2021 •

edited

Loading

Robbybp commented Nov 18, 2021

eslickj commented Nov 19, 2021 •

edited

Loading

dallan-keylogic left a comment

eslickj commented Jan 13, 2022

Robbybp Jan 13, 2022

Robbybp left a comment

Robbybp Jan 13, 2022

eslickj Jan 13, 2022

eslickj Jan 13, 2022 •

edited

Loading

Robbybp Jan 13, 2022

Robbybp Jan 13, 2022

eslickj Jan 13, 2022

eslickj Jan 14, 2022

eslickj commented Jan 13, 2022

Robbybp left a comment

Robbybp Jan 14, 2022

eslickj Jan 14, 2022

eslickj commented Jan 15, 2022

Petsc DAE Solver Utils #552

Petsc DAE Solver Utils #552

Conversation

eslickj commented Oct 14, 2021 • edited Loading

Summary/Motivation:

Examples

Legal Acknowledgement

codecov bot commented Nov 8, 2021 • edited Loading

Codecov Report

eslickj commented Nov 9, 2021 • edited Loading

Robbybp commented Nov 18, 2021

eslickj commented Nov 19, 2021 • edited Loading

dallan-keylogic left a comment

Choose a reason for hiding this comment

eslickj commented Jan 13, 2022

Choose a reason for hiding this comment

Robbybp left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

eslickj Jan 13, 2022 • 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

eslickj commented Jan 13, 2022

Robbybp left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

eslickj commented Jan 15, 2022

eslickj commented Oct 14, 2021 •

edited

Loading

codecov bot commented Nov 8, 2021 •

edited

Loading

eslickj commented Nov 9, 2021 •

edited

Loading

eslickj commented Nov 19, 2021 •

edited

Loading

eslickj Jan 13, 2022 •

edited

Loading