Add inlet outlet bc #633

g-rydquist · 2024-12-05T17:28:42Z

Add an inlet boundary condition that prescribes a velocity on a sideset. Pressure and density are set such that they have zero gradient. A couple of potential features are missing as of now:

Outlet BC has not been added.
There is only functionality to assign a single inlet block to all sidesets (the first one in the .lua file). The functionality to read and store multiple inlets is there, but I wasn't sure how to, in BCFunctions::inlet(), differentiate which sideset was associated with which inlet, so I just take the first one that gets stored. However, this inlet can be assigned to multiple sidesets.

A further note: instead of removing the loop over the bclist::Keys in the MultiMat constructor and looping manually, I removed inlet from the list and just read it manually.

This change is

…into bc configuration vec

…exception

…/ FVMultiMat

…was resolved

…ot have momentum flux from these

…multi mats

adityakpandare

Thank you @g-rydquist. The inlet BC parsing code looks good. I have a few comments on the BC enforcement itself.

Reviewed 16 of 16 files at r1, all commit messages.
Reviewable status: all files reviewed, 12 unresolved discussions (waiting on @g-rydquist)

src/Inciter/Transporter.cpp line 484 at r1 (raw file):

  using bclist = ctr::bclist::Keys;
  const auto& bcs = g_inputdeck.get< tag::bc >();
  brigand::for_each< bclist >( UserBC( g_inputdeck, usedsets ) );

This brigand::for_each is neither for timedep nor inlet BCs. Let's move it above the comment, as it was before (undo the change of moving it under the comment). Same for the bclist.

src/Inciter/Transporter.cpp line 489 at r1 (raw file):

      for (auto i : b.get< tag::sideset >())
        usedsets.insert(static_cast<int>(i));
    }

Nice! I like this! At some point, I will attempt at moving all BC's to this framework.

src/PDE/MultiMat/FVMultiMat.hpp line 102 at r1 (raw file):

            v.clear();
            v.insert(v.end(), sideset.begin(), sideset.end());
            m_bc.push_back( { v, inlet, noOpGrad } );

Is the noOpGrad correct for inlet? Shouldn't we be using velocity/temperature info from the BCs to get the grad? Or are we punting on it for now (which is okay; just making sure)?

src/Control/Inciter/InputDeck/InputDeck.hpp line 139 at r1 (raw file):

      tag::materialid,   std::size_t
    > >
  >,

Just to be a little "clear", let's move the tag::inlet to the very end, adjacent to tag::timedep, as a hint that inlet is treated similar to timedep.

src/PDE/MultiMat/BCFunctions.hpp line 137 at r1 (raw file):

    auto nmat = g_inputdeck.get< tag::multimat, tag::nmat >();
    const auto& solidx = g_inputdeck.get< tag::matidxmap, tag::solidx >();
    auto& inbc = g_inputdeck.get< tag::bc >()[0].get< tag::inlet >();

This index 0 corresponds to the mesh id. This indicates that for now, our MultiMat solver BCs aren't linked to the overset mesh solver.

src/PDE/MultiMat/BCFunctions.hpp line 140 at r1 (raw file):

    // inlet velocity and material
    auto u_in = inbc[0].get< tag::velocity >();

This is where only the first entry of the tag::inlet vector is used. I see. This will need some reworking. Let's create an issue on github, pointing to this function, and mentioning that it is not compatible with multiple inlet-bcs.

src/PDE/MultiMat/BCFunctions.hpp line 158 at r1 (raw file):

    auto v1r = 2.0*u_in[0] - v1l;
    auto v2r = 2.0*u_in[1] - v2l;
    auto v3r = 2.0*u_in[2] - v3l;

I'm not sure this vector algebra works out for an unstructured mesh an general velocity. Instead, the ghost-state velocity should be set to the inlet BC values. So basically, v1r = u_in[0] etc. This is what is called a "weak" BC enforcement that's typically used in FV/DG formulations.

src/PDE/MultiMat/BCFunctions.hpp line 177 at r1 (raw file):

      // zero bulk pressure and density gradient
      ur[ncomp+pressureIdx(nmat, k)] = ur[volfracIdx(nmat, k)] * p;

This quantity is the material pressure, not the bulk pressure. So if you had different EOS for each material, this code is incorrect. If the inlet BC is supposed to be used for subsonic conditions, use the pressure from inside (one characteristic moving out from inside at subsonic conditions):

ur[ncomp+pressureIdx(nmat, k)] = ur[volfracIdx(nmat,k)] * ul[ncomp+pressureIdx(nmat,k)]/ul[volfracIdx(nmat,k)];

If the inlet BC is supposed to be used for supersonic conditions, we need to accept inlet PC pressure and use that as the pressure (all characteristics moving in at supersonic conditions):

ur[ncomp+pressureIdx(nmat, k)] = ur[volfracIdx(nmat,k)] * p_in;

src/PDE/MultiMat/BCFunctions.hpp line 178 at r1 (raw file):

      // zero bulk pressure and density gradient
      ur[ncomp+pressureIdx(nmat, k)] = ur[volfracIdx(nmat, k)] * p;
      ur[densityIdx(nmat,k)] = ur[volfracIdx(nmat,k)] * rho;

This quantity is the material density, not the bulk density. So if you had different EOS for each material, this code is incorrect. We need to use the EOS to compute material density based on the user-specified inlet BC temperature and appropriate pressure (based on sub or super sonic).

src/PDE/MultiMat/BCFunctions.hpp line 180 at r1 (raw file):

      ur[densityIdx(nmat,k)] = ur[volfracIdx(nmat,k)] * rho;
      ur[energyIdx(nmat,k)] = ur[volfracIdx(nmat,k)] *
        mat_blk[k].compute< EOS::totalenergy >(rho, v1r, v2r, v3r, p);

Material density should be used here: ur[densityIdx...]/ur[volfracIdx...].

src/PDE/MultiMat/BCFunctions.hpp line 185 at r1 (raw file):

    ur[momentumIdx(nmat, 0)] = rho * v1r;
    ur[momentumIdx(nmat, 1)] = rho * v2r;
    ur[momentumIdx(nmat, 2)] = rho * v3r;

Bulk density rho needs to be recomputed once each material density is computed from EOS as commented above.

src/PDE/MultiMat/DGMultiMat.hpp line 108 at r1 (raw file):

          }
        }
      }

This function repeats in all fv/dg implementations. Let's move it outside to DGPDE.cpp and have everyone call that function.

g-rydquist

Reviewable status: 9 of 18 files reviewed, 13 unresolved discussions (waiting on @adityakpandare)

src/Control/Inciter/InputDeck/InputDeck.hpp line 139 at r1 (raw file):