This package provides some convenience functions for ADflow. It allows to:
- Easily calculate a Polar sweep with ADflow.
- Automatically create the output folder for ADflow.
- Automatically restart ADflow if there is a solution available.
- Plot realtime ADflow state variables in the terminal.
- Plot ADflow state variables from a logfile
The following script sweeps through a list of Angle of Attacks (the full example can be found at example/polar_sweep.py).
from adflow_util import ADFLOW_UTIL
options = {
'name': 'n0012_sweep',
'resetAP': True,
}
aeroOptions = {
'alpha': [1, 2, 3, 4],
'reynolds': 3e6,
# ...
'evalFuncs': ['cl','cd', 'cmz']
}
solverOptions = {
# Common Parameters
'gridFile': 'n0012.cgns',
'outputDirectory':'output',
# Physics Parameters
'equationType':'RANS',
# ...
'L2Convergence':1e-12,
}
au = ADFLOW_UTIL(aeroOptions, solverOptions, options)
au.run()The dict aeroOptions holds all the variables that normally baseclasses.AeroProblem would. If one variable is a list, this is considered the sweep variable. All variables except coefPol, cosCoefFourier, sinCoefFourier, momentAxis, solverOptions, evalFuncs can be sweeped.
The dict options holds some ADFLOW_UTIL specific options. More about them can be found here.
This script will generate a file called n0012_sweep.out with this content:
n0012_sweep
Aero Options
-------------- -----------
alpha 1, 2, 3, 4
...
evalFuncs cl, cd, cmz
-------------- -----------
RESULTS
alpha cd cl cmz totalRes iterTot
------- ---------- ---------- ---------- ---------- ---------
1 0.01011288 0.11602992 0.00066774 0.00085824 633
...
4 0.01080607 0.46238766 0.00248672 0.00089012 541
It is also possible to have multiple sweep variables. But all must have the same length. There will be no cross calculation. For example, if you set alpha = [0, 1] and reynolds = [5e6, 3e6] there will be two simulations in total. The first with alpha = 0, reynolds = 5e6 and the second with alpha = 1, reynolds = 3e6.
If this package was installed using pip, the command adflow_plot should be available in your terminal. To use it, simply type adflow_plot -i yourADflowScript.py. As this utility reads the stdout stream, it should work with all scripts as long as the ADflow option printIterations is True.
If you want to parallelize your ADflow calculation, simply add -np number_of_cores oder -H list_of_nodes. As a default, mpirun is used to start mpi. If you have a different installation of mpi, you can change it with -mpi some_different_mpi_command. Type adflow_plot -h to get a list of all available start options.
The output looks something like this:
If you want to see the raw ADflow output, you can list it at the top by entering hlog number_of_lines.
To close it, type q or quit.
Type h or help to get a list of all commands. type h a_command oder help a_command to get additional information about this specific command.
If the inputfile does not end with .py it is assumed to be a logfile. The file is read continously with the linux-command tail -f. This makes it possible to plot the variables allmost in realtime while the script is executed on a cluster for example.
Simply execute this command (if pip and git is installed)
pip install git+https://github.com/DavidAnderegg/adflow_util.git
As for now, this whole suite was only tested with steady simulations. Additionally, some terminals handle input keys differently and it might be possible, that for example ENTER ist not properly recognised. Please file a bug report so I can fix it as soon as possible.
If you find any bugs and/or you have ideas that I could add, file an issue on github or write me at andv@zhaw.ch.
This package uses a modified version of plotext. The raw package is available at https://github.com/piccolomo/plotext.
The Naca0012 Grid in the example folder is from https://github.com/mdolab/MACH-Aero.