/uav_drag_dse

This repository contains scripts for generating and analyzing drag data for UAV design and analysis

Primary LanguagePythonApache License 2.0Apache-2.0

ATHENS Drag Study Tools

This repository contains a cli for generating drag data for UAV and UAM vehicles from the SwRI Athens Corpus.

At a high level, this repo runs the new drag model with altered parameters for the specified vehicle. It is currently possible to randomly adjust cylinder/tube length, propeller diameter, and wing shape (chord1, chord2, span) and store the resulting drag values. Note that the drag model requires the existence of the designData.json and designParameters.json which are generated in buildcad.py and are transformations of the 8 json files produced by the autograph.py. Thus, we assume that the vehicle has been built in Creo at least once. This is because the drag model needs the transforms (i.e. orientation, rotation, and translation) for each component in the vehicle in order to create representative meshes.

The benefit of this tool is that, instead of updating parameters and rebuilding the vehicle using Creo (slow), we can regenerate the mesh with modified parameter values and record the resulting drag data. Then, for parameters that produce "optimal" drag, we can run the vehicle fewer times through the slower Creo pipeline. Additionally, by operating directly on the dictionaries representing the designData.json and designParameters.json files, we save a little file I/O by reading from these files once and then writing to the dictionaries on each iteration of the drag model. Currently, to update parameters of interest and run the drag model using this tool takes ~1s / iteration.

Installation & Usage

  1. To install this package, run pip install -e . from the root of this repository (assumes Python >= 3.8 or so)

  2. To store the baseline info for a vehicle including x, y, z plots, the stl mesh, and default drags and centers, run athens-dragstudy --vehicle <vehicle_name> --baseline. This will exit if the data already exists for the vehicle's default parameters and assumes that designData.json and designParameters.json files are present in data/uav|uam/<vehicle_name>/.

  3. To store results from n iterations of the drag model with specified study parameters, run athens-dragstudy --vehicle <vehicle_name> --rand-length --rand-prop --rand-wing --runs n

    1. The rand-length flag will randomly alter the LENGTH of each cylinder/tube in the vehicle, the rand-prop flag will alter the DIAMETER of each propeller in the vehicle, and the rand-wing flag will alter the CHORD_1, CHORD_2, and SPAN values of each wing in the vehicle.

  4. Results are written to data/uav|uam/<vehicle_name>/results/*.csv where each row represents a single iteration through the drag model and contains the resulting 6 output values from the drag model iteration and the value of each parameter in that iteration.

  5. To add a new vehicle, create a directory named after the vehicle in the appropriate uam or uav directory and copy the designData.json and designParameters.json files (this will be a command line option, eventually). Note uam and uav are still separated because of CAD part and parameter naming differences

  6. It is also possible to run a drag study on a zip file which is output from Jenkins. To do this, move the zip file of interest to the design_zips folder. Then, from the command line, use the --from-zip argument along with the --vehicle vehicle_name argument where vehicle_name is the name of the zip folder of interest, without the ".zip" extension.

See the drag_model notes file for FAQs about the code itself

Usage Examples

Generate data for a single component as the drag subject (called a "primitive"). Results are stored under data/primitive/<drag_subject>/results/*.csv

  • Capsule (UAV fuselage): athens-dragstudy -subject capsule --runs 1000
  • refactoring/ cleanup needed for other commands at this time, WIP

TODO

  • Add curve fitting
  • Resolve positional changes in motors when changing lengths
  • Create data format for running experiments using files from Jenkins data.zip
  • Clean up data format for studying multiple params (e.g. length vs length + prop)
  • Specify vehicle, study parameters, number of runs
  • Create command option to generate baseline results (original vehicle) with plots and mesh