examples/fluids: add IJacobian for SUPG

I think this is exact for Galerkin on channel.yaml and blasius.yaml.
Specifically, SNES converges almost perfectly in this test than when
-snes_mf_operator is

examples/fluids: add IJacobian for SUPG

I think this is exact for Galerkin on channel.yaml and blasius.yaml.
Specifically, SNES converges almost perfectly in this test than when
-snes_mf_operator is added.

-options_file examples/fluids/channel.yaml -snes_monitor -ksp_converged_reason -stab none -ts_max_steps 3 -newtonian_unit_tests -ksp_rtol 1e-10 -order 2 -ts_dt 3e-6 -pc_type lu

I say "think" because something is fishy with the viscous energy term
velocity*stress. If I disable that term, then convergence is clearly
better with the analytic Jacobian (makes sense because it's accuracy is
machine epsilon) than with -snes_mf_operator. This is unexpected, but
the difference is near the limit of numerical stability so I'm not
confident it's wrong. Also, the term itself is so simple.

The Jacobian is inexact for SUPG in that Tau_d is frozen.

WIP: examples/fluids: support MatShell and Blasius_Inflow_Jacobian

show more ...

Merge pull request #958 from CEED/jed/fluids-cleaning

examples/fluids cleaning

examples/fluids: make setup contexts independent

These are now private to each example and can be developed independently.

examples/fluids: consolidate two-step setup

Contexts are created in one place now and we have removed duplicate
struct definitions for physics.

examples/fluids: create struct in Problem for each qfunction + related

This is a cosmetic step toward better modularity.

Cleaner density inflow (#953)

* density weakly set on the inflow....also seems to have a commit intended for ken/primitiveSUPGterm OOOP

Merged onto main this density ONLY inflow step...fixed E_k

Cleaner density inflow (#953)

* density weakly set on the inflow....also seems to have a commit intended for ken/primitiveSUPGterm OOOP

Merged onto main this density ONLY inflow step...fixed E_kinitec bug along the way which used interior velocity.

* first cut at choose your weak inflow rho or T

* cleanup and a missing gamma

* dropped time step to 2e-6, added time limit of 0.001, added suggested parameters for tau constants for linear vs. quadratic. Note this is far from certain, expecially quadratic. Also left commented the required settings to forces SNES to do a fixed number of NL iterations. The current settings do 2 NL iterations for a while then drop to 1 which is ok for steady state but might not be good for SRS.

* potentially better Ctau_{C,M,E} that have much lower oscillations in the free stream even at large time steps like 3.2e-5

* examples/fluids: fix output_freq

* examples/fluids: use bool for weakT

Suggested-by: James Wright <james@jameswright.xyz>

Co-authored-by: Jed Brown <jed@jedbrown.org>

show more ...

feat: Add blasius boundary layer and channel flow examples to examples/fluids (#942)

* examples/fluids: Upload current case for collaboration

* examples/fluids: Replace developing BL with duct

feat: Add blasius boundary layer and channel flow examples to examples/fluids (#942)

* examples/fluids: Upload current case for collaboration

* examples/fluids: Replace developing BL with duct

* examples/fluids: Allow density to "float" at inflow

* fix: Add u_normal

* examples/fluids: bug fix in density flux

* examples/fluids: implement "floating" pressure; prescribed u, T

* examples/fluids: fixing initialization

We need has_neumann=true to get the weak boundary integrals called. It
is initialized in problem->setup_ctx and used in SetupLibceed(). This is
a hack to always apply.

* examples/fluids: include kinetic energy in Blasius IC

* examples/fluids: Blasius quasi-2D (slip in z planes)

* feat: Add Exact_Channel function

* feat: Use Exact_Channel for IC and Inflow

* examples/fluids: Add channel example

* examples/fluids: Define mu in blasius.c, cleanup

* examples/fluid: Add Blasius solution calculator

* examples/fluids: Add blasius IC and BCs

* Convert to implicit TS

* examples/fluids: Update blasius.yaml

* Move to ChannelContext, refactor DC and newtonian

- Added a dedicated ChannelContext, which allows for user setting of
flow parameters and consistent sharing of parameters between
QFunctions (instead of hardcoding)
- Moved density current (DC) specific settings from newtonian.c
densitycurrent.c

* Make gravity into vector quantity

* Add in body force for channel

* examples/fluids: remove explicitly setting coordinate field

This is created automatically now by DMPlex and DMProjectCoordinates
breaks (localized) periodicity, resulting in a tangled mesh.

* examples/fluids: transpose flux Jacobian in SUPG stabilization

* examples/fluids: update docs commensurate with flux Jacobian transpose fix

* examples/fluids: work on docs for tau

* examples/fluids: Increase blasius Re, implement Xi to SUPG

- Increased the Uinf and mu by 4x for the blasius problem
- Implemented the calculation of Xi for the SUPG term in newtonian.h

* examples/fluids: Remove flux jacobian transpose

- Removed for both eulervortex.h and newtonian.h

* examples/fluids: Reset Newtonian unit scaling to neutral

- Note that this resets the scaling for the densitycurrent problem,
thus the examples (and possibly tests) will probably need to be
changed

* examples/fluids: fix PetscOptionsBegin for new PETSc

* feat: Add slanted domain top surface

* feat: Add graded mesh option for blasius

* fix: Correct inflow boundary condition handling

* feat: Move to blasius_context, add CLI options

* feat: Move Blasius to physical air properties

* fix: Blasius profile and mesh generation

- Add the correct float->int rounding
- Fix eta to a value after exceeding the table's bounds

fix: Add fix eta after exceeding table

* examples/fluids: Move to physical mu default, fix tests

* doc: Add minimum documentation for channel and blasius

* examples/fluids: Add primitive jacobian and tau

- Also add misc comments

* examples/fluids: propose alternative flux Jacobian via prim-to-conservative

* examples/fluids: Add PHASTA diagonal tau

* fix: Add back in timestep obtaining

* examples/fluids: Add analytic tractions to inflow and outflow

- This is a stop-gap until viscous flux can be computed properly on the
boundary from current solution.
- Tau constants in a state of flux still but this produced a decent v at
inflow and outflow, removing the steep dive caused by forcing
dv/dx=-du/dy from a zero traction of omission.

* examples/fluids: CLI options for diagonal c_tau's

* examples/fluids: Update dt via context labels

* feat: Add more blasius table points

* examples/fluids: Cleanup stab and adjust Ctau_E

* examples/fluids: Replace spatial tau with diagonal

* examples/fluids: Fix stab in Euler

- Initialize `stab` to zero, add sum over loop

* examples/fluids: Document diagonal tau formulation

* examples/fluids: Fix compiler warnings

* examples/fluids: check error codes and elide unused header

* examples/fluids: Correct blasius example CLI

- Also add mention of the `blasius.yaml` file

* examples/fluids: Add supg to default blasius.yaml

* examples/fluids: Use yaml for channel and blasius

- Instead of long CLI options
- Also cleaned up the yaml formatting
- Added Cartesian directions to the faceMarker names

* examples/fluids: Show example yamls in README

Co-authored-by: Jed Brown <jed@jedbrown.org>
Co-authored-by: Kenneth E. Jansen <Kenneth.Jansen@colorado.edu>

show more ...