Navier-Stokes in 2D on DGMulti and TreeMesh #1165

andrewwinters5000 · 2022-06-07T20:44:03Z

First implementation of the compressible Naiver-Stokes equations which extends the capability added by #1156 . Should be useful to determine if the current boundary condition strategy and other storage architectures extend to equation systems.

codecov · 2022-06-07T22:21:50Z

Codecov Report

Merging #1165 (3281b49) into parabolic-treemesh (39d06de) will decrease coverage by 0.01%.
The diff coverage is 97.12%.

❗ Current head 3281b49 differs from pull request most recent head 98ea1bd. Consider uploading reports for the commit 98ea1bd to get more accurate results

@@                  Coverage Diff                   @@
##           parabolic-treemesh    #1165      +/-   ##
======================================================
- Coverage               96.79%   96.78%   -0.01%     
======================================================
  Files                     309      326      +17     
  Lines                   24258    25034     +776     
======================================================
+ Hits                    23479    24228     +749     
- Misses                    779      806      +27

Flag	Coverage Δ
unittests	`96.78% <97.12%> (-0.01%)`	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
src/solvers/dgsem_tree/dg.jl	`100.00% <ø> (ø)`
src/equations/laplace_diffusion_2d.jl	`81.25% <81.25%> (ø)`
src/equations/compressible_navier_stokes_2d.jl	`88.33% <88.33%> (ø)`
src/solvers/dgmulti/dg_parabolic.jl	`93.38% <93.38%> (ø)`
...ization/semidiscretization_hyperbolic_parabolic.jl	`96.43% <96.43%> (ø)`
src/solvers/dgsem_tree/dg_2d_parabolic.jl	`99.60% <99.60%> (ø)`
examples/dgmulti_2d/elixir_advection_diffusion.jl	`100.00% <100.00%> (ø)`
...multi_2d/elixir_advection_diffusion_nonperiodic.jl	`100.00% <100.00%> (ø)`
.../dgmulti_2d/elixir_advection_diffusion_periodic.jl	`100.00% <100.00%> (ø)`
...les/dgmulti_2d/elixir_navier_stokes_convergence.jl	`100.00% <100.00%> (ø)`
... and 16 more

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 39d06de...98ea1bd. Read the comment docs.

ranocha · 2022-06-08T15:33:39Z

I updated some of our branches and changed the base of this one to the parabolic TreeMesh stuff to make it easier to review this PR.

ranocha

Thanks a lot for this first draft, @andrewwinters5000! I am sure @sloede and @jlchan will also have some comments and ideas.

src/Trixi.jl

examples/tree_2d_dgsem/elixir_navier_stokes_source_terms.jl

ranocha · 2022-06-08T15:34:56Z

examples/tree_2d_dgsem/elixir_navier_stokes_source_terms.jl

+# I really do not like this structure but it should work for now
+equations_parabolic = CompressibleNavierStokes2D(1.4,  # gamma
+                                                 1000, # Reynolds number
+                                                 0.72, # Prandtl number
+                                                 0.5,  # free-stream Mach number
+                                                 1.0,  # thermal diffusivity
+                                                 equations)


Do we need all of these parameters? If so, should we maybe create a constructor with keywords?

I am not sure if we need all these constants. For example, the Prandtl number basically never changes för the ideal gas assumption. I threw in all the constants I thought we would need and am open to trimming it down. Also, we could possibly store mu = 1 / Reynolds for the sake of code readability.

Keywords is probably the best way to do this but I wasn't sure if having the constructor look like

function CompressibleNavierStokes2D(equations; gamma, Reynolds, Prandtl, Mach_freestream, kappa)

would be kosher to have the call structure

equations_parabolic = CompressibleNavierStokes2D(gamma=1.4, Reynolds=1000, Prandtl=0.72, Mach_freestream=0.5, kappa=1.0, # thermal diffusivity equations)

for the sake of consistency with the LaplaceEquation2D where equations comes last.

Keyword arguments need to come after positional ones, so it could be

CompressibleNavierStokes2D(equations; Reynolds, Prandtl, Mach_freestream, kappa)

where we should grab gamma from equations.

@sloede We should probably unify the argument names with the lattice Boltzmann stuff (using Ma and Re for now).

examples/tree_2d_dgsem/elixir_navier_stokes_source_terms.jl

ranocha · 2022-06-08T15:38:13Z

src/equations/compressible_navier_stokes_2d.jl

+# TODO:
+# 1) For now I save gamma and inv(gamma-1) again, but we could potentially reuse them from
+#    the Euler equations
+# 2) Add more here and probably some equations


src/equations/compressible_navier_stokes_2d.jl

ranocha · 2022-06-08T15:41:45Z

src/equations/compressible_navier_stokes_2d.jl

+# This is the flexibility a user should have to select the different gradient variable types
+# varnames(::typeof(cons2prim)   , ::CompressibleNavierStokes2D) = ("v1", "v2", "T")
+# varnames(::typeof(cons2entropy), ::CompressibleNavierStokes2D) = ("w2", "w3", "w4")
+
+varnames(variable_mapping, equations_parabolic::CompressibleNavierStokes2D) =
+  varnames(variable_mapping, equations_parabolic.equations)


I think varnames is meant to still give the variable names of the underlying conserved (or transformed) variables, not their gradients. But we should probably make this more clear in the other parts.

Ah, okay. I got confused then. In my other codes I carry around an nvars for the conservative variables and a separate nvars_grad for the number of gradient variables.

From my end it would be helpful to (somewhere) have the constructor for the parabolic equations output what are the gradient variables for the user. What I mean is that the equations constructors would report something like the following to the screen

┌──────────────────────────────────────────────────────────────────────────────────────────────────┐ │ CompressibleEulerEquations2D │ │ ════════════════════════════ │ │ #variables: ………………………………………………… 4 │ │ │ variable 1: …………………………………………… rho │ │ │ variable 2: …………………………………………… rho_v1 │ │ │ variable 3: …………………………………………… rho_v2 │ │ │ variable 4: …………………………………………… rho_e │ └──────────────────────────────────────────────────────────────────────────────────────────────────┘ ┌──────────────────────────────────────────────────────────────────────────────────────────────────┐ │ CompressibleNavierStokesEquations2D │ │ ═══════════════════════════════════ │ │ gradient variables: …………………………… primitive │ │ #variables: ………………………………………………… 3 │ │ │ variable 1: …………………………………………… v1 │ │ │ variable 2: …………………………………………… v2 │ │ │ variable 3: …………………………………………… T │ └──────────────────────────────────────────────────────────────────────────────────────────────────┘

Yeah, this sounds reasonable. Could you please add this to our issue #1147 tracking parabolic stuff?

src/equations/compressible_navier_stokes_2d.jl

andrewwinters5000 · 2022-06-09T06:31:55Z

I updated some of our branches and changed the base of this one to the parabolic TreeMesh stuff to make it easier to review this PR.

Thanks!

Co-authored-by: Hendrik Ranocha <[email protected]>

src/equations/compressible_navier_stokes_2d.jl

jlchan · 2022-07-20T00:10:16Z

I'll investigate that - the odd behavior I observe is on y_neg, so this seems consistent.

One very odd observation is that the parabolic RHS for TreeMesh is 10x smaller than that of DGMulti, up to maybe 6-7 digits. This seems much more odd to me...

EDIT: the 10x thing wasn't related to a bug, it was just a difference in the Reynolds numbers set for each example in my local branch.

jlchan · 2022-07-28T01:52:52Z

Curious - if I set Re = 1e14 to make the parabolic terms small and use a SemidiscretizationHyperbolic (ignoring parabolic terms but still using the same Navier-Stokes source term) then DGSEM with TreeMesh still crashes. Can anyone else confirm this?

jlchan · 2022-07-28T14:54:55Z

I think the issue is somehow related to slip wall BCs for TreeMesh. I made an issue with a MWE for it here #1190.

jlchan · 2022-07-28T19:31:51Z

Hopefully these recent changes fix things. I get convergence now for compressible Navier-Stokes using TreeMesh now.

I've added tests for both TreeMesh and DGMulti now. @ranocha @andrewwinters5000 @sloede - if the tests pass, is there anything else we need to add to this PR before merging into dev?

ranocha

Looks quite good already! From my point of view, we can merge this once the following comments are addressed.

examples/dgmulti_2d/elixir_navier_stokes_convergence.jl

examples/dgmulti_2d/elixir_navier_stokes_lid_driven_cavity.jl

examples/tree_2d_dgsem/elixir_advection_diffusion_nonperiodic.jl

examples/tree_2d_dgsem/elixir_navier_stokes_convergence.jl

src/equations/compressible_navier_stokes_2d.jl

Co-authored-by: Hendrik Ranocha <[email protected]>

sloede · 2022-07-29T14:43:05Z

Would it make sense for me to review this as well? If yes: Is it possible that the original branch parabolic-treemesh is currently not up-to-date with main, whereas this one is? The diff seems to show differences that are not due to Navier-Stokes being added...

jlchan · 2022-07-29T14:47:58Z

Hm...since this PR is merging into parabolic-treemesh, wouldn't that bring parabolic-treemeshup to date with main as well?

jlchan · 2022-07-29T14:48:49Z

I think that's correct; we stopped working on parabolic-treemesh after @andrewwinters5000 introduced this PR.

ranocha · 2022-07-29T14:49:23Z

Will do some cleanup and then trigger the merge. I think @sloede does not need to review this right now since we will have several other rounds of review

sloede · 2022-07-29T14:50:07Z

OK, got it 👍 Will not review.

…kes_2d

* Add prototype for advection-diffusion elixir for TreeMesh2D * Initial implementation of `calc_gradient!` for TreeMesh{2} * First complete implementation of the parabolic rhs operator for TreeMesh * Make it work * Add first elixir that works * Clean up elixir * first draft of container for Navier-Stokes constants and fluxes * remove unneeded temperature computation * draft of elixir with possible boundary condition structure * added manufactured solution and source term * fix typo in function name for MMS * update variable names for consistency. improve comments * fix dumb typos in new equation system name * actually export new equations * add comment near variable_mapping declaration. * Apply suggestions from code review Co-authored-by: Hendrik Ranocha <[email protected]> * parabolic equations now exported separately * change name to CompressibleNavierStokesEquations2D * export NS with proper name * explicitly require compressible Euler in the type parameter * name kinematic viscosity nu for consistency with Lattice-Boltzmann * add promotion in constructor * make Reynolds, Prandtl, Mach, and kappa keyword arguments * update constructor call in elixir * reduce computation by exploiting stress tensor symmetry * fix unpacking of flux * modifying parabolic cache creation in cache, we assume we take the gradient of all hyperbolic variables. since the number of parabolic variables can differ from the number of hyperbolic variables, we pass in the hyperbolic equations to `create_cache_parabolic` now * comments * comments * formatting and renaming equations to equations_hyperbolic formatting comments * fix unpacking of gradients in flux * adding CNS BCs * adding lid-driven cavity elixir * adding variable transform, editing cons2prim for CNS * add prim2cons for NS (inconsistent right now) * add draft of DGMulti Navier Stokes convergence elixir * converging solution using elixir for TreeMesh with BCs * fixing DGMulti advection diffusion elixir convergence * naming equations as parabolic/hyperbolic * generalizing transform_variables * add TODO more todos * additional checks on get_unsigned_normal * adding isothermal BC * commenting out unused CNS functions for now * fix call to transform_variables * comments and cleanup * changing default solver and Re for cavity * adding more advection diffusion tests * label tests * add gradient_variables field to SemidiscretizationHyperbolicParabolic * Revert "add gradient_variables field to SemidiscretizationHyperbolicParabolic" This reverts commit 063b602. * allowing for specialization in transform_variables adding a function `gradient_variable_transformation` which should get specialized if the gradient variables are not conservative variables * formatting and comments * reverting elixir * comments * standardizing time tol * minor fix to CNS boundary flux for convenience make it so that the density state is computed correctly, even though it's not used * formatting + comments * using primitive variables in viscous flux instead of conservative * minor formatting * add CNS tests * fix test * testing if scoping issues fix TreeMesh tests * decreasing timestep tol for TreeMesh parabolic test * enabling periodic BCs for TreeMesh + more tests * fix density BC flux (unused, but could be useful) * adding non-working TreeMesh elixirs * adding AD checks * standardizing parameters in convergence elixirs * minor cleanup * revert DGMulti CNS elixir setup back to the one used in tests * adding TreeMesh CNS convergence test * removing redundant elixir * add more tests * add more test * Apply suggestions from code review Co-authored-by: Hendrik Ranocha <[email protected]> * set version to v0.4.43 * set development version to v0.4.44-pre * add YouTube links to JuliaCon presentations (#1192) * add YouTube links to JuliaCon presentations * Apply suggestions from code review Co-authored-by: Michael Schlottke-Lakemper <[email protected]> Co-authored-by: Michael Schlottke-Lakemper <[email protected]> * improved readability of equation docstrings in compressible Euler files * Navier-Stokes in 2D on DGMulti and TreeMesh (#1165) * first draft of container for Navier-Stokes constants and fluxes * remove unneeded temperature computation * draft of elixir with possible boundary condition structure * added manufactured solution and source term * fix typo in function name for MMS * update variable names for consistency. improve comments * fix dumb typos in new equation system name * actually export new equations * add comment near variable_mapping declaration. * Apply suggestions from code review Co-authored-by: Hendrik Ranocha <[email protected]> * parabolic equations now exported separately * change name to CompressibleNavierStokesEquations2D * export NS with proper name * explicitly require compressible Euler in the type parameter * name kinematic viscosity nu for consistency with Lattice-Boltzmann * add promotion in constructor * make Reynolds, Prandtl, Mach, and kappa keyword arguments * update constructor call in elixir * reduce computation by exploiting stress tensor symmetry * fix unpacking of flux * modifying parabolic cache creation in cache, we assume we take the gradient of all hyperbolic variables. since the number of parabolic variables can differ from the number of hyperbolic variables, we pass in the hyperbolic equations to `create_cache_parabolic` now * comments * comments * formatting and renaming equations to equations_hyperbolic formatting comments * fix unpacking of gradients in flux * adding CNS BCs * adding lid-driven cavity elixir * adding variable transform, editing cons2prim for CNS * add prim2cons for NS (inconsistent right now) * add draft of DGMulti Navier Stokes convergence elixir * converging solution using elixir for TreeMesh with BCs * fixing DGMulti advection diffusion elixir convergence * naming equations as parabolic/hyperbolic * generalizing transform_variables * add TODO more todos * additional checks on get_unsigned_normal * adding isothermal BC * commenting out unused CNS functions for now * fix call to transform_variables * comments and cleanup * changing default solver and Re for cavity * adding more advection diffusion tests * label tests * add gradient_variables field to SemidiscretizationHyperbolicParabolic * Revert "add gradient_variables field to SemidiscretizationHyperbolicParabolic" This reverts commit 063b602. * allowing for specialization in transform_variables adding a function `gradient_variable_transformation` which should get specialized if the gradient variables are not conservative variables * formatting and comments * reverting elixir * comments * standardizing time tol * minor fix to CNS boundary flux for convenience make it so that the density state is computed correctly, even though it's not used * formatting + comments * using primitive variables in viscous flux instead of conservative * minor formatting * add CNS tests * fix test * testing if scoping issues fix TreeMesh tests * decreasing timestep tol for TreeMesh parabolic test * enabling periodic BCs for TreeMesh + more tests * fix density BC flux (unused, but could be useful) * adding non-working TreeMesh elixirs * adding AD checks * standardizing parameters in convergence elixirs * minor cleanup * revert DGMulti CNS elixir setup back to the one used in tests * adding TreeMesh CNS convergence test * removing redundant elixir * add more tests * add more test * Apply suggestions from code review Co-authored-by: Hendrik Ranocha <[email protected]> * add docstrings Co-authored-by: Hendrik Ranocha <[email protected]> Co-authored-by: Hendrik Ranocha <[email protected]> Co-authored-by: Jesse Chan <[email protected]> * add docstring for CNS equations * removing some addressed TODOs * adjust definition of the kappa constant * avoid overwriting u_grad with viscous_flux results * remove old TODOs * clarify TODOs for later * removing let statements * rearrange main docstring. Fix typos in math environment causing it to not parse * update TODO notes * Apply suggestions from code review Co-authored-by: Andrew Winters <[email protected]> * update TODO notes * Apply suggestions from code review Co-authored-by: Andrew Winters <[email protected]> * update TODO notes * changing diffusivity names Auto stash before merge of "parabolic-treemesh" and "origin/parabolic-treemesh" * Reynolds/Prandtl number name changes for consistency * Apply suggestions from code review Co-authored-by: Hendrik Ranocha <[email protected]> Co-authored-by: Michael Schlottke-Lakemper <[email protected]> Co-authored-by: Andrew Winters <[email protected]> * fixing bug introduced by GH code review * moving manufactured solution into the CNS equations file * fix allocation issue * get_diffusivity -> diffusivity * removing cruft code, adding comment on messy prim2cons routine * moving manufactured solution back into elixirs * Update examples/dgmulti_2d/elixir_advection_diffusion_nonperiodic.jl * TODO -> Todo, and .6 -> 0.6 d * Update src/solvers/dgsem_tree/dg_2d_parabolic.jl Co-authored-by: Hendrik Ranocha <[email protected]> * adding note to Adiabatic/Isothermal BCs * replacing CompressibleNavierStokesEquations with ...Diffusion * Update src/solvers/dgsem_tree/dg_2d_parabolic.jl Co-authored-by: Hendrik Ranocha <[email protected]> * Update src/solvers/dgsem_tree/dg_2d_parabolic.jl Co-authored-by: Hendrik Ranocha <[email protected]> * documenting Jacobian sign flip * remove extra arg to `gradient_variable_transformation` * Update src/solvers/dgsem_tree/dg_2d_parabolic.jl Co-authored-by: Michael Schlottke-Lakemper <[email protected]> * change warning to error * fix bug introduced by search/replace * dg_parabolic -> parabolic_scheme * BoundaryConditionViscousWall -> BoundaryConditionNavierStokesWall * u_grad -> gradients * fix test * renaming * update comment * converting to signature `flux(u, gradients, orientation, equations)` * using prolong2interfaces/boundaries * unpacking `gradients` and `flux_viscous` * using calc_surface_integral! * adding @muladd * Apply suggestions from code review * update comments * Rename grad_u -> gradients * Update src/equations/compressible_navier_stokes_2d.jl Co-authored-by: Michael Schlottke-Lakemper <[email protected]> * Rename remaining grad_u/u_grad -> gradients * Refactor into prolong2interfaces! * Refactor calc_volume_integral! * Refactor calc_interface_flux * Refactor prolong2boundaries! * Refactor calc_divergence! * Formatting consistency * Remove dispatch on volume integral type * Add comment on why surface_integral is passed to prolong2interfaces! * Explicitly use weak form volume integral to allow easy overriding in test * Add flux differencing test for CNS test * Remove erroneous P4estMesh{2} dispatch * Remove non-cons terms from parabolic solver Co-authored-by: Andrew Winters <[email protected]> Co-authored-by: Hendrik Ranocha <[email protected]> Co-authored-by: Hendrik Ranocha <[email protected]> Co-authored-by: Jesse Chan <[email protected]> Co-authored-by: Jesse Chan <[email protected]>

@threaded

* Parabolic terms for DGMulti (#1136) * minor formatting * adding multi-term semidiscretization * fixing bug and adding test * changing name * changing name, importing SplitODEFunction * fixing test by adding norm * trying another test fix * adding norm back * adjusting tol * try to fix test again * syntax error * removing norm frm test * adding cache to parabolic terms (enable reuse of inviscid cache) * generalize dgmulti iterators * fix typo * generalizing reset_du! for dgmulti * WIP gradient and divergence computations * fix name * more generalization * divergence + remove repeated code * adding cache * remove test * update diffusion test * update diffusion rhs * moving routines around * fix multiple includes * removing time dependence for now from viscous rhs * export ScalarDiffusion2D * fix test * working periodic advection RHS * cleanup and comments * making periodic advection-diffusion example * trim whitespace * adding t back in to parabolic rhs add t back in * refactoring boundary condition code * update elixir with BCs * change default DGMulti interface flux behavior same as in #1139 * cleanup * renaming variables * BC prepping * cleanup of elixir * more name fixing * Update src/semidiscretization/semidiscretization_hyperbolic_parabolic.jl Co-authored-by: Hendrik Ranocha <[email protected]> * adding Neumann BCs * switching to weak form weak form more weak form weak form * fixing cache variable names * simplified parabolic boundary flux treatement * updated elixir with different types of BCs * Update src/semidiscretization/semidiscretization_hyperbolic_parabolic.jl Co-authored-by: Michael Schlottke-Lakemper <[email protected]> * Update src/semidiscretization/semidiscretization_hyperbolic_parabolic.jl Co-authored-by: Michael Schlottke-Lakemper <[email protected]> * Update src/equations/equations.jl Co-authored-by: Michael Schlottke-Lakemper <[email protected]> * rhs! -> rhs_parabolic! * move Gradient/Divergence types * add "no boundary condition" BC * rename ScalarDiffusion -> LaplaceDiffusion r r * rhs -> rhs_parabolic! again * comments * moving tests around * scalarDiffusion -> LaplaceDiffusion * tuple-based constructor * updating examples * add ScalarPlotData2D docstring (#1145) Co-authored-by: Jesse Chan <[email protected]> * Apply suggestions from code review Co-authored-by: Michael Schlottke-Lakemper <[email protected]> Co-authored-by: Hendrik Ranocha <[email protected]> * renaming for consistency parabolic_equations -> equations_parabolic no_boundary_condition -> boundary_condition_do_nothing * move Neumann BCs into equations.jl * renaming, e.g., ParabolicEquations -> EquationsParabolic, etc * rename test module * generalize LaplaceDiffusion2D to any number of variables * removing saveat times and standardizing `tol` names * parabolic_cache -> cache_parabolic cache * renaming `create_cache` -> `create_cache_parabolic` avoids conflicts when using dg::DGMultiFluxDiff * Update examples/dgmulti_2d/elixir_advection_diffusion.jl Co-authored-by: Michael Schlottke-Lakemper <[email protected]> * Update src/equations/equations.jl Co-authored-by: Michael Schlottke-Lakemper <[email protected]> * Update src/semidiscretization/semidiscretization_hyperbolic_parabolic.jl Co-authored-by: Michael Schlottke-Lakemper <[email protected]> * parabolic_boundary_conditions -> boundary_conditions_parabolic similarly for hyperbolic d * Update src/semidiscretization/semidiscretization_hyperbolic_parabolic.jl Co-authored-by: Michael Schlottke-Lakemper <[email protected]> * add equations as field to LaplaceDiffusion2D * moving Neumann BCs into equations.jl and adding default BC behavior * running parabolic tests, adding integration test * adding parabolic tests to ci * missing comma * Update src/equations/laplace_diffusion_2d.jl Co-authored-by: Hendrik Ranocha <[email protected]> * fix test * fix test typo * fix tests (for real this time) * improve test coverage Co-authored-by: Jesse Chan <[email protected]> Co-authored-by: Hendrik Ranocha <[email protected]> Co-authored-by: Michael Schlottke-Lakemper <[email protected]> * Additional work on parabolic terms (#1148) * make "do nothing BC" a struct, move it to basic_types * add varnames for LaplaceDiffusion * add forgotten @threaded * update comment * Revert "add forgotten @threaded" This reverts commit 1564501. * add @threaded (without enumerate) * remove periodic advection diffusion elixir (redundant) * adding test * one more test * Update src/basic_types.jl Co-authored-by: Hendrik Ranocha <[email protected]> Co-authored-by: Jesse Chan <[email protected]> Co-authored-by: Hendrik Ranocha <[email protected]> * Reorganization of boundary conditions (#1152) * moving "default" boundary conditions back into equations file * add periodic advection_diffusion back in * updating test * update test * Apply suggestions from code review Co-authored-by: Jesse Chan <[email protected]> Co-authored-by: Hendrik Ranocha <[email protected]> Co-authored-by: Hendrik Ranocha <[email protected]> * Adding parabolic solver option (#1153) * adding parabolic solver types * incorporating parabolic solver types into dg_parabolic penalties * specializing parabolic solver options in LaplaceDiffusion2D * moving things around to fix CI * dropped arg in test * Apply suggestions from code review Co-authored-by: Hendrik Ranocha <[email protected]> * renaming files and ViscousFlux solvers renaming files rename rename * Added TODO note for penalty term * ViscousFlux___ -> ViscousFormulation___ * move function def around Co-authored-by: Jesse Chan <[email protected]> Co-authored-by: Hendrik Ranocha <[email protected]> * rearrange include order * Add support for hyperbolic-parabolic systems to TreeMesh2D (#1156) * Add prototype for advection-diffusion elixir for TreeMesh2D * Initial implementation of `calc_gradient!` for TreeMesh{2} * First complete implementation of the parabolic rhs operator for TreeMesh * Make it work * Add first elixir that works * Clean up elixir * first draft of container for Navier-Stokes constants and fluxes * remove unneeded temperature computation * draft of elixir with possible boundary condition structure * added manufactured solution and source term * fix typo in function name for MMS * update variable names for consistency. improve comments * fix dumb typos in new equation system name * actually export new equations * add comment near variable_mapping declaration. * Apply suggestions from code review Co-authored-by: Hendrik Ranocha <[email protected]> * parabolic equations now exported separately * change name to CompressibleNavierStokesEquations2D * export NS with proper name * explicitly require compressible Euler in the type parameter * name kinematic viscosity nu for consistency with Lattice-Boltzmann * add promotion in constructor * make Reynolds, Prandtl, Mach, and kappa keyword arguments * update constructor call in elixir * reduce computation by exploiting stress tensor symmetry * fix unpacking of flux * modifying parabolic cache creation in cache, we assume we take the gradient of all hyperbolic variables. since the number of parabolic variables can differ from the number of hyperbolic variables, we pass in the hyperbolic equations to `create_cache_parabolic` now * comments * comments * formatting and renaming equations to equations_hyperbolic formatting comments * fix unpacking of gradients in flux * adding CNS BCs * adding lid-driven cavity elixir * adding variable transform, editing cons2prim for CNS * add prim2cons for NS (inconsistent right now) * add draft of DGMulti Navier Stokes convergence elixir * converging solution using elixir for TreeMesh with BCs * fixing DGMulti advection diffusion elixir convergence * naming equations as parabolic/hyperbolic * generalizing transform_variables * add TODO more todos * additional checks on get_unsigned_normal * adding isothermal BC * commenting out unused CNS functions for now * fix call to transform_variables * comments and cleanup * changing default solver and Re for cavity * adding more advection diffusion tests * label tests * add gradient_variables field to SemidiscretizationHyperbolicParabolic * Revert "add gradient_variables field to SemidiscretizationHyperbolicParabolic" This reverts commit 063b602. * allowing for specialization in transform_variables adding a function `gradient_variable_transformation` which should get specialized if the gradient variables are not conservative variables * formatting and comments * reverting elixir * comments * standardizing time tol * minor fix to CNS boundary flux for convenience make it so that the density state is computed correctly, even though it's not used * formatting + comments * using primitive variables in viscous flux instead of conservative * minor formatting * add CNS tests * fix test * testing if scoping issues fix TreeMesh tests * decreasing timestep tol for TreeMesh parabolic test * enabling periodic BCs for TreeMesh + more tests * fix density BC flux (unused, but could be useful) * adding non-working TreeMesh elixirs * adding AD checks * standardizing parameters in convergence elixirs * minor cleanup * revert DGMulti CNS elixir setup back to the one used in tests * adding TreeMesh CNS convergence test * removing redundant elixir * add more tests * add more test * Apply suggestions from code review Co-authored-by: Hendrik Ranocha <[email protected]> * set version to v0.4.43 * set development version to v0.4.44-pre * add YouTube links to JuliaCon presentations (#1192) * add YouTube links to JuliaCon presentations * Apply suggestions from code review Co-authored-by: Michael Schlottke-Lakemper <[email protected]> Co-authored-by: Michael Schlottke-Lakemper <[email protected]> * improved readability of equation docstrings in compressible Euler files * Navier-Stokes in 2D on DGMulti and TreeMesh (#1165) * first draft of container for Navier-Stokes constants and fluxes * remove unneeded temperature computation * draft of elixir with possible boundary condition structure * added manufactured solution and source term * fix typo in function name for MMS * update variable names for consistency. improve comments * fix dumb typos in new equation system name * actually export new equations * add comment near variable_mapping declaration. * Apply suggestions from code review Co-authored-by: Hendrik Ranocha <[email protected]> * parabolic equations now exported separately * change name to CompressibleNavierStokesEquations2D * export NS with proper name * explicitly require compressible Euler in the type parameter * name kinematic viscosity nu for consistency with Lattice-Boltzmann * add promotion in constructor * make Reynolds, Prandtl, Mach, and kappa keyword arguments * update constructor call in elixir * reduce computation by exploiting stress tensor symmetry * fix unpacking of flux * modifying parabolic cache creation in cache, we assume we take the gradient of all hyperbolic variables. since the number of parabolic variables can differ from the number of hyperbolic variables, we pass in the hyperbolic equations to `create_cache_parabolic` now * comments * comments * formatting and renaming equations to equations_hyperbolic formatting comments * fix unpacking of gradients in flux * adding CNS BCs * adding lid-driven cavity elixir * adding variable transform, editing cons2prim for CNS * add prim2cons for NS (inconsistent right now) * add draft of DGMulti Navier Stokes convergence elixir * converging solution using elixir for TreeMesh with BCs * fixing DGMulti advection diffusion elixir convergence * naming equations as parabolic/hyperbolic * generalizing transform_variables * add TODO more todos * additional checks on get_unsigned_normal * adding isothermal BC * commenting out unused CNS functions for now * fix call to transform_variables * comments and cleanup * changing default solver and Re for cavity * adding more advection diffusion tests * label tests * add gradient_variables field to SemidiscretizationHyperbolicParabolic * Revert "add gradient_variables field to SemidiscretizationHyperbolicParabolic" This reverts commit 063b602. * allowing for specialization in transform_variables adding a function `gradient_variable_transformation` which should get specialized if the gradient variables are not conservative variables * formatting and comments * reverting elixir * comments * standardizing time tol * minor fix to CNS boundary flux for convenience make it so that the density state is computed correctly, even though it's not used * formatting + comments * using primitive variables in viscous flux instead of conservative * minor formatting * add CNS tests * fix test * testing if scoping issues fix TreeMesh tests * decreasing timestep tol for TreeMesh parabolic test * enabling periodic BCs for TreeMesh + more tests * fix density BC flux (unused, but could be useful) * adding non-working TreeMesh elixirs * adding AD checks * standardizing parameters in convergence elixirs * minor cleanup * revert DGMulti CNS elixir setup back to the one used in tests * adding TreeMesh CNS convergence test * removing redundant elixir * add more tests * add more test * Apply suggestions from code review Co-authored-by: Hendrik Ranocha <[email protected]> * add docstrings Co-authored-by: Hendrik Ranocha <[email protected]> Co-authored-by: Hendrik Ranocha <[email protected]> Co-authored-by: Jesse Chan <[email protected]> * add docstring for CNS equations * removing some addressed TODOs * adjust definition of the kappa constant * avoid overwriting u_grad with viscous_flux results * remove old TODOs * clarify TODOs for later * removing let statements * rearrange main docstring. Fix typos in math environment causing it to not parse * update TODO notes * Apply suggestions from code review Co-authored-by: Andrew Winters <[email protected]> * update TODO notes * Apply suggestions from code review Co-authored-by: Andrew Winters <[email protected]> * update TODO notes * changing diffusivity names Auto stash before merge of "parabolic-treemesh" and "origin/parabolic-treemesh" * Reynolds/Prandtl number name changes for consistency * Apply suggestions from code review Co-authored-by: Hendrik Ranocha <[email protected]> Co-authored-by: Michael Schlottke-Lakemper <[email protected]> Co-authored-by: Andrew Winters <[email protected]> * fixing bug introduced by GH code review * moving manufactured solution into the CNS equations file * fix allocation issue * get_diffusivity -> diffusivity * removing cruft code, adding comment on messy prim2cons routine * moving manufactured solution back into elixirs * Update examples/dgmulti_2d/elixir_advection_diffusion_nonperiodic.jl * TODO -> Todo, and .6 -> 0.6 d * Update src/solvers/dgsem_tree/dg_2d_parabolic.jl Co-authored-by: Hendrik Ranocha <[email protected]> * adding note to Adiabatic/Isothermal BCs * replacing CompressibleNavierStokesEquations with ...Diffusion * Update src/solvers/dgsem_tree/dg_2d_parabolic.jl Co-authored-by: Hendrik Ranocha <[email protected]> * Update src/solvers/dgsem_tree/dg_2d_parabolic.jl Co-authored-by: Hendrik Ranocha <[email protected]> * documenting Jacobian sign flip * remove extra arg to `gradient_variable_transformation` * Update src/solvers/dgsem_tree/dg_2d_parabolic.jl Co-authored-by: Michael Schlottke-Lakemper <[email protected]> * change warning to error * fix bug introduced by search/replace * dg_parabolic -> parabolic_scheme * BoundaryConditionViscousWall -> BoundaryConditionNavierStokesWall * u_grad -> gradients * fix test * renaming * update comment * converting to signature `flux(u, gradients, orientation, equations)` * using prolong2interfaces/boundaries * unpacking `gradients` and `flux_viscous` * using calc_surface_integral! * adding @muladd * Apply suggestions from code review * update comments * Rename grad_u -> gradients * Update src/equations/compressible_navier_stokes_2d.jl Co-authored-by: Michael Schlottke-Lakemper <[email protected]> * Rename remaining grad_u/u_grad -> gradients * Refactor into prolong2interfaces! * Refactor calc_volume_integral! * Refactor calc_interface_flux * Refactor prolong2boundaries! * Refactor calc_divergence! * Formatting consistency * Remove dispatch on volume integral type * Add comment on why surface_integral is passed to prolong2interfaces! * Explicitly use weak form volume integral to allow easy overriding in test * Add flux differencing test for CNS test * Remove erroneous P4estMesh{2} dispatch * Remove non-cons terms from parabolic solver Co-authored-by: Andrew Winters <[email protected]> Co-authored-by: Hendrik Ranocha <[email protected]> Co-authored-by: Hendrik Ranocha <[email protected]> Co-authored-by: Jesse Chan <[email protected]> Co-authored-by: Jesse Chan <[email protected]> * Add commentary on reuse of interfaces/boundaries data structures (#1199) * Add comments on storing flux values in `u` * Rearrange function names for consistency * Update src/solvers/dgsem_tree/dg_2d_parabolic.jl Co-authored-by: Hendrik Ranocha <[email protected]> * Update src/solvers/dgsem_tree/dg_2d_parabolic.jl Co-authored-by: Andrew Winters <[email protected]> Co-authored-by: Hendrik Ranocha <[email protected]> Co-authored-by: Andrew Winters <[email protected]> * Adding parabolic term description to NEWS.md (#1207) * Add Literate docs for advection-diffusion (#1208) * adding literate docs for advection-diffusion * Apply suggestions from code review Co-authored-by: Hendrik Ranocha <[email protected]> * rename "adding parabolic terms" -> "parabolic terms" Co-authored-by: Hendrik Ranocha <[email protected]> * Adding a tutorial on adding new parabolic terms (#1209) * Viscous terms from entropy gradients (#1202) * first draft of entropy stable gradients * update TODOs * added elixir with discontinuous solution for ES testing * forgot one TODO marker * adding GradientVariables**** type parameter * fix bug (missing type param in constructor) * add elixir with periodic CNS manufactured solution * fix gradient conversion * fix entropy variables w/BCs * updating elixirs with `gradient_variables=GradientVariablesPrimitive()` * Update src/equations/compressible_navier_stokes_2d.jl Co-authored-by: Michael Schlottke-Lakemper <[email protected]> * removing unused routines * removing specialized entropy2cons and cons2entropy routines * adding more tests * removing unused elixirs * Apply suggestions from code review Co-authored-by: Hendrik Ranocha <[email protected]> * adding docs/warning for gradient variable type * extracting w_inner[4] into a more clearly named variable Co-authored-by: Hendrik Ranocha <[email protected]> Co-authored-by: Jesse Chan <[email protected]> Co-authored-by: Jesse Chan <[email protected]> Co-authored-by: Michael Schlottke-Lakemper <[email protected]> Co-authored-by: Jesse Chan <[email protected]> Co-authored-by: Hendrik Ranocha <[email protected]> * adding some minor documentation (#1214) * fix parabolic PID (#1224) * fix parabolic PID * activate analysis_callback in Navier-Stokes convergence test * fix typo * Rescale compressible Navier-Stokes (#1230) * remove the nondimensionalization and update the docstring for CNS * update TreeMesh elixirs and test values * update the elixirs and test values for DGMulti * fix unbalanced parentheses * file renaming for consistency * update file names in the parabolic test script * remove the gas constant as a parameter and update comments * rename viscosity to dynamic_viscosity * add comment about units in the docstring * Update src/equations/compressible_navier_stokes_2d.jl Co-authored-by: Michael Schlottke-Lakemper <[email protected]> * unify notation to call the viscosity mu * fix undefined variable Co-authored-by: Michael Schlottke-Lakemper <[email protected]> * improve type stability of semidiscretize for parabolic semi (#1231) * Apply suggestions from code review Co-authored-by: Hendrik Ranocha <[email protected]> Co-authored-by: Michael Schlottke-Lakemper <[email protected]> Co-authored-by: Hendrik Ranocha <[email protected]> Co-authored-by: Andrew Winters <[email protected]>

andrewwinters5000 added 5 commits June 7, 2022 21:02

first draft of container for Navier-Stokes constants and fluxes

60cce54

remove unneeded temperature computation

46ae5fb

draft of elixir with possible boundary condition structure

4f6558f

added manufactured solution and source term

dde54d3

fix typo in function name for MMS

daaeac3

jlchan self-requested a review June 7, 2022 20:46

andrewwinters5000 added 5 commits June 8, 2022 06:59

update variable names for consistency. improve comments

4a39883

Merge branch 'main' into nav_stokes_2d

84d8454

fix dumb typos in new equation system name

3879a3f

actually export new equations

435e081

add comment near variable_mapping declaration.

7b59fef

ranocha changed the base branch from dev to parabolic-treemesh June 8, 2022 15:29

Merge branch 'parabolic-treemesh' into nav_stokes_2d

d3b7870

ranocha reviewed Jun 8, 2022

View reviewed changes

sloede reviewed Jun 8, 2022

View reviewed changes

src/equations/compressible_navier_stokes_2d.jl Outdated Show resolved Hide resolved

andrewwinters5000 and others added 4 commits June 9, 2022 08:33

Apply suggestions from code review

5dc8301

Co-authored-by: Hendrik Ranocha <[email protected]>

parabolic equations now exported separately

f316684

change name to CompressibleNavierStokesEquations2D

02c14e1

export NS with proper name

dd733bb

ranocha reviewed Jun 10, 2022

View reviewed changes

src/equations/compressible_navier_stokes_2d.jl Outdated Show resolved Hide resolved

andrewwinters5000 mentioned this pull request Jun 10, 2022

Notes on parabolic terms #1147

Open

35 tasks

andrewwinters5000 added 6 commits June 10, 2022 10:17

explicitly require compressible Euler in the type parameter

a1a48fe

name kinematic viscosity nu for consistency with Lattice-Boltzmann

d8db203

add promotion in constructor

06e69b8

make Reynolds, Prandtl, Mach, and kappa keyword arguments

eda9737

update constructor call in elixir

b469bce

reduce computation by exploiting stress tensor symmetry

fb646b0

jlchan mentioned this pull request Jul 28, 2022

Issue with slip wall BCs for CompressibleEulerEquations2D using TreeMesh #1190

Closed

ranocha mentioned this pull request Jul 28, 2022

fix boundary_condition_slip_wall for 2D TreeMesh #1191

Merged

jlchan added 5 commits July 28, 2022 14:13

Merge remote-tracking branch 'origin/main' into nav_stokes_2d

e0c3b84

standardizing parameters in convergence elixirs

bfffc00

minor cleanup

17eafbb

revert DGMulti CNS elixir setup back to the one used in tests

0c4e098

adding TreeMesh CNS convergence test

b70bbc0

jlchan marked this pull request as ready for review July 28, 2022 19:40

jlchan changed the title ~~[WIP] Navier-Stokes in 2D on TreeMesh~~ Navier-Stokes in 2D on DGMulti and TreeMesh Jul 28, 2022

jlchan added 3 commits July 28, 2022 17:18

removing redundant elixir

13e0c8f

add more tests

8cecb5f

add more test

adb8eca

ranocha requested changes Jul 29, 2022

View reviewed changes

andrewwinters5000 and others added 2 commits July 29, 2022 09:32

Apply suggestions from code review

3281b49

Co-authored-by: Hendrik Ranocha <[email protected]>

add docstrings

98ea1bd

ranocha approved these changes Jul 29, 2022

View reviewed changes

Merge remote-tracking branch 'origin/parabolic-treemesh' into nav_sto…

bf43508

…kes_2d

ranocha merged commit 6002ba6 into parabolic-treemesh Jul 29, 2022

ranocha deleted the nav_stokes_2d branch July 29, 2022 14:51

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Navier-Stokes in 2D on DGMulti and TreeMesh #1165

Navier-Stokes in 2D on DGMulti and TreeMesh #1165

andrewwinters5000 commented Jun 7, 2022

codecov bot commented Jun 7, 2022 •

edited

Loading

ranocha commented Jun 8, 2022

ranocha left a comment

ranocha Jun 8, 2022

andrewwinters5000 Jun 9, 2022

ranocha Jun 10, 2022

ranocha Jun 10, 2022

andrewwinters5000 Jun 10, 2022

ranocha Jun 8, 2022

ranocha Jun 8, 2022

andrewwinters5000 Jun 9, 2022

andrewwinters5000 Jun 9, 2022 •

edited

Loading

ranocha Jun 10, 2022

andrewwinters5000 commented Jun 9, 2022

jlchan commented Jul 20, 2022 •

edited

Loading

jlchan commented Jul 28, 2022 •

edited

Loading

jlchan commented Jul 28, 2022

jlchan commented Jul 28, 2022

ranocha left a comment •

edited

Loading

sloede commented Jul 29, 2022

jlchan commented Jul 29, 2022

jlchan commented Jul 29, 2022

ranocha commented Jul 29, 2022

sloede commented Jul 29, 2022

Navier-Stokes in 2D on DGMulti and TreeMesh #1165

Navier-Stokes in 2D on DGMulti and TreeMesh #1165

Conversation

andrewwinters5000 commented Jun 7, 2022

codecov bot commented Jun 7, 2022 • edited Loading

Codecov Report

ranocha commented Jun 8, 2022

ranocha 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

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

andrewwinters5000 Jun 9, 2022 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

andrewwinters5000 commented Jun 9, 2022

jlchan commented Jul 20, 2022 • edited Loading

jlchan commented Jul 28, 2022 • edited Loading

jlchan commented Jul 28, 2022

jlchan commented Jul 28, 2022

ranocha left a comment • edited Loading

Choose a reason for hiding this comment

sloede commented Jul 29, 2022

jlchan commented Jul 29, 2022

jlchan commented Jul 29, 2022

ranocha commented Jul 29, 2022

sloede commented Jul 29, 2022

codecov bot commented Jun 7, 2022 •

edited

Loading

andrewwinters5000 Jun 9, 2022 •

edited

Loading

jlchan commented Jul 20, 2022 •

edited

Loading

jlchan commented Jul 28, 2022 •

edited

Loading

ranocha left a comment •

edited

Loading