zhulianhua / dugksfoam

An OpenFOAM solver for Boltzmann model equation using discrete unified gas kinetic scheme

dugksFoam

An OpenFOAM solver for the Boltzmann model equations using discrete unified gas kinetic scheme [1,2,3].

Key features

• Solving discrete velocity Boltzmann equation;
• Based on Shakhov-BGK collision model;
• Using discrete unified gas kinetic scheme (Asymptotic preserving property);
• A standard OpenFOAM solver;
• Arbitrary unstructured meshes;
• MPI parallel computing capability;
• 1D & 2D & 3D in a single solver;
• Post processing tools of OpenFOAM are ready to use;
• Various boundary condition types.

Installation and documentation

NOTE: Tested on the following OpenFOAM versions :

• 2.4.0, gcc/intel
• 3.0.0, gcc
• 3.0.1, gcc
• 4.0, gcc
• 4.1 gcc/intel
• 6.0, gcc

of240 # change the version code according your OF installation
cd dugksFoam/src
./Allwmake

See documentation at doc/dugksFoam.pdf.

References

• [1] Z. Guo, K. Xu, R. Wang, Discrete unified gas kinetic scheme for all Knudsen number flows: low-speed isothermal case, Phys. Rev. E, 88 (2013) 033305.
• [2] Z. Guo, R. Wang, K. Xu, Discrete unified gas kinetic scheme for all Knudsen number flows. II. Thermal compressible case, Phys. Rev. E, 91(2015) 033313.
• [3] L. Zhu, Z. Guo, K. Xu, Discrete unified gas kinetic scheme on unstructured meshes, Comp. Fluids, 127(2016) 211-225
• [4] L. Zhu, S. Chen, Z. Guo, dugksFoam: An open source OpenFOAM solver for the Boltzmann model equation, Comp. Phys. Commun., 213(2017) 155-164
• [5] L. Zhu, Z. Guo, Numerical study of nonequilibrium gas flow in a microchannel with a ratchet surface, Phys. Rev. E. 95, (2017), 023113
• [6] A. Lotfian, E. Roohi, Radiometric flow in periodically patterned channels: fluid physics and improved configurations, Journal of Fluid Mechanics(2019), 860, 544-576
• [7] S. Chen, C. Zhang, L. Zhu, Z. Guo, A unified implicit scheme for kinetic model equations. Part I. Memory reduction technique, Science Bulletin, 62 (2016) 119–129

dugksfoam's Issues

Bug when cyclic boundary patches pair are not in the same parallel sub zone

When the cyclic patches are not in the same subzone (by physical space parallel decomposition), the parallel runs will crash. A simple Couette flow case will reproduce this issue. I guess this is due to the improper treatment of cyclicProcessor boundary in the dugksFoam source code.

write out distribution function by providing cell list

The writeDF(cellID) function has been implemented but we should provide a general input to call it, other than hard code it in dugksFoam.C

farfield boundary condition for external flow problems (flow around bodies)

Hi, everyone!

Problem description

I want to use dugksFoam to calculate the low-speed incompressible flow passing a micro-cylinder. I found cases like internal flow problems (only wall boundary condition is needed) and supersonic flow around a body in the user guide and published papers using dugksFoam, so I decided to try a supersonic-flow-around-a-2-D-cylinder case first. The case appears in Discrete unified gas kinetic scheme on unstructured meshes.

However, the description of the farfield BC in the user guide is a bit vague, which reads: "To specify such a boundary type, just set the boundary types as fixedValue, and provide the free-stream flow condition as the boundary values in 0/rho, 0/U and 0/T." If the velocity on the whole outer boundary is set to fixedValue, how can the flow field adjust to a steady state? I tried setting the BC as the user guide suggested, but the solving soon diverged. I changed fixedValue to farField; the solving has not diverged yet but the result is weird.

So, how should I specify the farfield BC?
Can dugksFoam be used to calculate the incompresible external flow problem?
(I post this issue in parallel-cdugksFoam at first. Hope it doesn't bother and many thanks.)

Details of the settings

fvDVMparas
{
    xiMax       xiMax [0 1 -1 0 0 0 0]    5.053155000000000e+03;
    xiMin       xiMin [0 1 -1 0 0 0 0]   -5.053155000000000e+03;
    nDV               89;       // Number of discrete velocity, shoud be 4*Z + 1 if using compound N-C quardrature
}

gasProperties
{
    R            R [0 2 -2 -1 0 0 0] 207.85; // Specific gas constant
    omega        0.81;              // VHS viscosity ~ Temperature index
    Tref         Tref [0 0 0 1 0 0 0] 273.0; // Reference temperature
    muRef        muRef [1 -1 -1 0 0 0 0] 2.117e-5;
    Pr           0.6667; // Prantl number
}

0/U

boundaryField
{
    cylinder
    {
        type            fixedValue;
        value           uniform (0 0 0);
    }
    farfield
    {
        type            farField;
        value 		uniform (1538.73 0 0);        
    }
    frontAndBack
    {
        type            empty;
    }
}

0/T

boundaryField
{
    cylinder
    {
        type            fixedValue;
        value           uniform 273;
    }
    farfield
    {
        type            farField;
        value           uniform 273;
    }
    frontAndBack
    {
        type            empty;
    }
}

0/rho

boundaryField
{
    cylinder
    {
        type            calculatedMaxwell;
        value           uniform 8.598e-5;  // dummy
    }
    farfield
    {
        type            farField;
        value           uniform 8.598e-5;
    }
    frontAndBack
    {
        type            empty;
    }
}

Result

The solution time interval is about 2e-8, and the flow fields of , , and at 3e-4

Ouput stress tensor from dugksFoam

Excuese me, I have a question when using dugksFoam. How to output stress tensor from dugksFoam for the full flow field?

Better create a MPI compilation stuff??

In the user manual you suggest to override C++ OpenFOAM rules:

Change the compier from CXX to MPICXX, because we implemented
the velocity-space parallel feature which use the standard alone MPI
instead of the Pstream library of OpenFOAM.
cp $WM_DIR/rules/$WM_ARCH$WM_COMPILER/c++ \
$WM_DIR/rules/$WM_ARCH$WM_COMPILER/c++.bak
sed -i "s/$WM_CXX/mpicxx/" \
$WM_DIR/rules/$WM_ARCH$WM_COMPILER/c++

The obvious drawback is that you are editing OpenFOAM installed rules, which may affects future compilations (bad for rockies). Instead of that, you can create alternative compilation rules:

cp -r $WM_DIR/rules/$WM_ARCH$WM_COMPILER "$WM_DIR/rules/$WM_ARCH"MPI
export WM_COMPILER=MPI
sed -i "s/$WM_CXX/mpicxx/" $WM_DIR/rules/$WM_ARCH$WM_COMPILER/c++

It's a bit more complicated, but cleaner in my opinion, since you are actually not changing nothing in the OF installation.

output U with surface value and use Uboudanry Tboudanry and rhoBoudanry as input

Sometimes, we need to output slip velocity and wall temperature, so it's better to use the standalone fields to assign boundary condition such as U and T. Just like the dsmcFoam.

That would require changing the setup of the case.

Repeating the calculation?

In the function Foam::discreteVelocity::updateGHsurf() , gSurf_ = (1.0 - relaxFactor)gSurf_ + relaxFactorgEq; has already calculated all the boundary fields (operator overloading calculates both the interior and boundary fields). Why is there a need to explicitly iterate through forAll(gSurfPatch, facei) and calculate the boundary fields again?

Ouput stress tensor from dugksFoam

Excuese me, I have a question when using dugksFoam. How to output stress tensor from dugksFoam for the full flow field?

convergence control not work

convergeTol 1e-8; in the system/contorlDict has no effect.

Why bcMap?

OpenFoam10

There are problems compiling the code on OpenFoam10

Update cell center macro by macro flux

Not fully verified, bugs still exist. Use with caution

python3 setDV.py update

setDV.py may have some error when use python3. In "def dvGH", the variable "N" can replace with "int(N)" and in "def dvNC", "for i in range(nBy4)" can replace with "for i in range(int(nBy4))".

add support of optimal high-order discrete velocity set

Used optimized discrete velocity for 2D and 3D flows, instead of using tensor product of 1D Gauss-Hermite discrete velocity to construct for 2D&3D cases. Add option to use discrete equilibrium distribution function to save computing time and improve conservation property of collision term.

disable old-style-cast warning

Change the wmake rule file C++ compiling flag