A Python framework for developing parallelized Computational Fluid Dynamics software to solve the hyperbolic 2D Euler equations on distributed, multi-block structured grids.

Last update: Nov 22, 2022

Overview

pyHype: Computational Fluid Dynamics in Python

pyHype is a Python framework for developing parallelized Computational Fluid Dynamics software to solve the hyperbolic 2D Euler equations on distributed, multi-block structured grids. It can be used as a solver to generate numerical predictions of 2D inviscid flow fields, or as a platform for developing new CFD techniques and methods. Contributions are welcome! pyHype is in early stages of development, I will be updating it regularly, along with its documentation.

The core idea behind pyHype is flexibility and modularity. pyHype offers a plug-n-play approach to CFD software, where every component of the CFD pipeline is modelled as a class with a set interface that allows it to communicate and interact with other components. This enables easy development of new components, since the developer does not have to worry about interfacing with other components. For example, if a developer is interested in developing a new approximate riemann solver technique, they only need to provide the implementation of the FluxFunction abstract class, without having to worry about how the rest of the code works in detail.

NEW: Geometry not alligned with the cartesian axes is now supported!
NEW: 60% efficiency improvement!
COMING UP: Examples of simulations on various airfoil geometries, and a presentation of the newly added mesh optimization techniques.
COMING UP: Examples of simulations on multi-block meshes.

Explosion Simulation

Here is an example of an explosion simulation performed on one block. The simulation was performed with the following:

600 x 1200 cartesian grid
Roe approximate riemann solver
Venkatakrishnan flux limiter
Piecewise-Linear second order reconstruction
Green-Gauss gradient method
RK4 time stepping with CFL=0.8
Reflection boundary conditions

The example in given in the file examples/explosion.py. The file is as follows:

from pyHype.solvers import Euler2D

# Solver settings
settings = {'problem_type':             'explosion',
            'interface_interpolation':  'arithmetic_average',
            'reconstruction_type':      'conservative',
            'upwind_mode':              'primitive',
            'write_solution':           False,
            'write_solution_mode':      'every_n_timesteps',
            'write_solution_name':      'nozzle',
            'write_every_n_timesteps':  40,
            'CFL':                      0.8,
            't_final':                  0.07,
            'realplot':                 False,
            'profile':                  True,
            'gamma':                    1.4,
            'rho_inf':                  1.0,
            'a_inf':                    343.0,
            'R':                        287.0,
            'nx':                       600,
            'ny':                       1200,
            'nghost':                   1,
            'mesh_name':                'chamber'
            }

# Create solver
exp = Euler2D(fvm='SecondOrderPWL',
              gradient='GreenGauss',
              flux_function='Roe',
              limiter='Venkatakrishnan',
              integrator='RK4',
              settings=settings)

# Solve
exp.solve()

Double Mach Reflection (DMR)

Here is an example of a Mach 10 DMR simulation performed on five blocks. The simulation was performed with the following:

500 x 500 cells per block
HLLL flux function
Venkatakrishnan flux limiter
Piecewise-Linear second order reconstruction
Green-Gauss gradient method
Strong-Stability-Preserving (SSP)-RK2 time stepping with CFL=0.4

The example in given in the file examples/dmr/dmr.py. The file is as follows:

from pyHype.solvers import Euler2D

# Solver settings
settings = {'problem_type':             'mach_reflection',
            'interface_interpolation':  'arithmetic_average',
            'reconstruction_type':      'conservative',
            'upwind_mode':              'conservative',
            'write_solution':           False,
            'write_solution_mode':      'every_n_timesteps',
            'write_solution_name':      'machref',
            'write_every_n_timesteps':  20,
            'plot_every':               10,
            'CFL':                      0.4,
            't_final':                  0.25,
            'realplot':                 True,
            'profile':                  False,
            'gamma':                    1.4,
            'rho_inf':                  1.0,
            'a_inf':                    1.0,
            'R':                        287.0,
            'nx':                       50,
            'ny':                       50,
            'nghost':                   1,
            'mesh_name':                'wedge_35_four_block',
            'BC_inlet_west_rho':        8.0,
            'BC_inlet_west_u':          8.25,
            'BC_inlet_west_v':          0.0,
            'BC_inlet_west_p':          116.5,
            }

# Create solver
exp = Euler2D(fvm='SecondOrderPWL',
              gradient='GreenGauss',
              flux_function='HLLL',
              limiter='Venkatakrishnan',
              integrator='RK2',
              settings=settings)

# Solve
exp.solve()

High Speed Jet

Here is an example of high-speed jet simulation performed on 5 blocks. The simulation was performed with the following:

Mach 2 flow
100 x 1000 cell blocks
HLLL flux function
Venkatakrishnan flux limiter
Piecewise-Linear second order reconstruction
Green-Gauss gradient method
RK2 time stepping with CFL=0.4

The example in given in the file examples/jet/jet.py. The file is as follows:

from pyHype.solvers import Euler2D

# Solver settings
settings = {'problem_type':             'subsonic_rest',
            'interface_interpolation':  'arithmetic_average',
            'reconstruction_type':      'primitive',
            'upwind_mode':              'conservative',
            'write_solution':           True,
            'write_solution_mode':      'every_n_timesteps',
            'write_solution_name':      'kvi',
            'write_every_n_timesteps':  20,
            'plot_every':               10,
            'CFL':                      0.4,
            't_final':                  25.0,
            'realplot':                 False,
            'profile':                  False,
            'gamma':                    1.4,
            'rho_inf':                  1.0,
            'a_inf':                    1.0,
            'R':                        287.0,
            'nx':                       1000,
            'ny':                       100,
            'nghost':                   1,
            'mesh_name':                'jet',
            'BC_inlet_west_rho':        1.0,
            'BC_inlet_west_u':          0.25,
            'BC_inlet_west_v':          0.0,
            'BC_inlet_west_p':          2.0 / 1.4,
            }

# Create solver
exp = Euler2D(fvm='SecondOrderPWL',
              gradient='GreenGauss',
              flux_function='HLLL',
              limiter='Venkatakrishnan',
              integrator='RK2',
              settings=settings)

# Solve
exp.solve()

Mach Number:

Density:

Current work

Integrate airfoil meshing and mesh optimization using elliptic PDEs
Compile gradient and reconstruction calculations with numba
Integrate PyTecPlot to use for writing solution files and plotting
Implement riemann-invariant-based boundary conditions
Implement subsonic and supersonic inlet and outlet boundary conditions
Implement connectivity algorithms for calculating block connectivity and neighbor-finding
Create a fully documented simple example to explain usage
Documentation!!

Major future work

Use MPI to distrubute computation to multiple processors
Adaptive mesh refinement (maybe with Machine Learning :))
Interactive gui for mesh design
Advanced interactive plotting

A Python framework for developing parallelized Computational Fluid Dynamics software to solve the hyperbolic 2D Euler equations on distributed, multi-block structured grids.

Related tags

Overview

pyHype: Computational Fluid Dynamics in Python

Explosion Simulation

Double Mach Reflection (DMR)

High Speed Jet

Current work

Major future work

Owner

Mohamed Khalil

Code for the paper "Reinforced Active Learning for Image Segmentation"

The official start-up code for paper "FFA-IR: Towards an Explainable and Reliable Medical Report Generation Benchmark."

Inkscape extensions for figure resizing and editing

Pairwise learning neural link prediction for ogb link prediction

Code for the Weighted, Accelerated and Restarted Primal-dual algorithm. This algorithm achieves stable linear convergence for reconstruction from undersampled noisy measurements under an approximate sharpness condition. See the paper for details.

Convolutional Neural Networks

Source code for the NeurIPS 2021 paper "On the Second-order Convergence Properties of Random Search Methods"

CrossNorm and SelfNorm for Generalization under Distribution Shifts (ICCV 2021)

Official repository for Fourier model that can generate periodic signals

Code implementation from my Medium blog post: [Transformers from Scratch in PyTorch]

Listing arxiv - Personalized list of today's articles from ArXiv

This repository contains the database and code used in the paper Embedding Arithmetic for Text-driven Image Transformation

This porject is intented to build the most accurate model for predicting the porbability of loan default

A simple algorithm for extracting tree height in sparse scene from point cloud data.

PyTorch reimplementation of the paper Involution: Inverting the Inherence of Convolution for Visual Recognition [CVPR 2021].

A large-scale benchmark for co-optimizing the design and control of soft robots, as seen in NeurIPS 2021.

Self-Supervised Pre-Training for Transformer-Based Person Re-Identification

This is a simple plugin for Vim that allows you to use OpenAI Codex.

The lightweight PyTorch wrapper for high-performance AI research. Scale your models, not the boilerplate.

Official PyTorch implementation of "Proxy Synthesis: Learning with Synthetic Classes for Deep Metric Learning" (AAAI 2021)