This IAMReX repo extends the capability of original IAMR codes, aiming at simulating the multiphase incompressible flows and fluid structure interaction problems on both CPUs and GPUs with/without subcycling. The Navier-Stokes euqations are solved on an adaptive semi-staggered grid using the projection method. The gas-liquid interface is captured using the level set (LS) method. The fluid-solid interface is resolved using the multidirect forcing immersed boundary method (IBM). The particle-wall as well as the particle-particle collisions are also captured by the adaptive collision time model (ACTM).
- LS method and reinitialization schemes
- Multidirect forcing Immersed Boundary Method
- Particle Collision Algorithms with Discrete Element Method
Profiles of drop interface in the RSV problem at t/T=1 after one rotation. Black line: Analytical Solution; Red line: 64*64; Blue line: 128*128; Green line: 256*256
(a) Density profile at t/T=2.42 using LS method. (b) Density profile at t/T=2.42 using IAMR convective scheme.
Comparison of the tip locations of the falling fluid and the rising fluid.
Contours of velocity magnitude in y − z plane
Our codes rely on the following packages:
- AMReX:
git clone https://github.com/ruohai0925/amrex - AMReX-Hydro:
git clone https://github.com/ruohai0925/AMReX-Hydro - This repo:
git clone https://github.com/ruohai0925/IAMReX/tree/development
After cloning, one will have three folder in your current directory:AMReX,AMReX-Hydro,IAMReX.
We recommend using the GNU compiler to compile the program on the Linux platform. The compilation process requires preparing a make file, which you can find in the example folder under Tutorials. It is strongly recommended to use the GNUmakefile prepared in advance by the example. If you want to know more about the configuration information of the compilation parameters, you can check it in the AMReX building.
For example, if we want to compile in the FlowPastSphere , refer to the following steps:
-
cdto the FlowPastSphere directorycd IAMReX/Tutorials/FlowPastSphere -
Modify compilation parameters in
GNUmakefile.The compilation parameters depend on your computing platform. If you use
MPIto run your program, then setUSE_MPI = TRUE. If you are running the program withNvidia GPU(CUDA), then setUSE_CUDA = TRUE. When using GPU runtime, please make sure your CUDA environment is ok. -
Compile
With the above settings, you can compile the program:
make
You can add parameters after
make, such as-f your_make_fileto customize your parameter file, and-j 8to specify8threads to participate in the compilation.If the compilation is successful, an executable file will be generated. Usually the executable file is named in the following format:
amr[DIM].[Compiler manufacturers].[computing platform].[is Debug].ex. The executable file name of the Release version of the three-dimensional compiled using the GNU compiler in the MPI environment is:amr3d.GNU.MPI.ex.
You should takes an input file as its first command-line argument. The file may contain a set of parameter definitions that will overrides defaults set in the code. In the FlowPastSphere, you can find a file named inputs.3d.flow_past_sphere, run:
./amr3d.GNU.MPI.ex inputs.3d.flow_past_sphereIf you use MPI to run your program, you can type:
mpirun -np how_many_threads amr3d.GNU.MPI.ex inputs.3d.flow_past_sphereThis code typically generates subfolders in the current folder that are named plt00000, plt00010, etc, and chk00000, chk00010, etc. These are called plotfiles and checkpoint files. The plotfiles are used for visualization of derived fields; the checkpoint files are used for restarting the code.
In the Tools folder, there are some post-processing script for DIBM. The postProcess submodule is used to process the generated particle data file. The specific usage can be seen in the README.
In addition, a Fortran code for generating a particle bed is provided, You can customize the domain size and particle size you want.
! Line 10 - 15
! domain scheme
LX_DOMAIN=3.0D0*Pi
LY_DOMAIN=6.0D0*Pi
LZ_DOMAIN=3.0D0*Pi
! D
SPD=1.D0Compile the code and run it, and a file named SPC.DAT will be generated, in which each line represents the position of a particle.
gfortran -o fixBed DomainFill.F90
./fixBedTo cite IAMReX, please use
@article{10.1063/5.0236509,
author = {Li, Xuzhu (李虚竹) and Li, Chun (李春) and Li, Xiaokai (李晓凯) and Li, Wenzhuo (李文卓) and Tang, Mingze (唐铭泽) and Zeng, Yadong (曾亚东) and Zhu, Zhengping (朱正平)},
title = {An open-source, adaptive solver for particle-resolved simulations with both subcycling and non-subcycling methods},
journal = {Physics of Fluids},
volume = {36},
number = {11},
pages = {113335},
year = {2024},
month = {11},
issn = {1070-6631},
doi = {10.1063/5.0236509},
url = {https://doi.org/10.1063/5.0236509},
eprint = {https://pubs.aip.org/aip/pof/article-pdf/doi/10.1063/5.0236509/20247663/113335\_1\_5.0236509.pdf},
}
@inbook{doi:10.2514/6.2025-1865,
author = {Dewen Liu and Shuai He and Haoran Cheng and Yadong Zeng},
title = {Investigate the Efficiency of Incompressible Flow Simulations on CPUs and GPUs With BSAMR},
booktitle = {AIAA SCITECH 2025 Forum},
chapter = {},
pages = {},
doi = {10.2514/6.2025-1865},
URL = {https://arc.aiaa.org/doi/abs/10.2514/6.2025-1865},
eprint = {https://arc.aiaa.org/doi/pdf/10.2514/6.2025-1865},
}
We are grateful to Ann Almgren, Andy Nonaka, Andrew Myers, Axel Huebl, and Weiqun Zhang in the Lawrence Berkeley National Laboratory (LBNL) for their discussions related to AMReX and IAMR. Y.Z. and Z.Z. also thank Prof. Lian Shen, Prof. Ruifeng Hu, and Prof. Xiaojing Zheng during their Ph.D. studies.
If you have any question or wanna contribute to the code, please don't hesitate to contact us.