Skip to content

Aircraft design optimization made fast through computational graph transformations (e.g., automatic differentiation). Composable analysis tools for aerodynamics, propulsion, structures, trajectory design, and much more.

License

Notifications You must be signed in to change notification settings

peterdsharpe/AeroSandbox

Repository files navigation

by Peter Sharpe (<pds [at] mit [dot] edu>)

Downloads Monthly Downloads Build Status PyPI Documentation Status License: MIT

AeroSandbox is a Python package that helps you design and optimize aircraft and other engineered systems.

At its heart, AeroSandbox is an optimization suite that combines the ease-of-use of familiar NumPy syntax with the power of modern automatic differentiation.

This automatic differentiation dramatically improves optimization performance on large problems: design problems with tens of thousands of decision variables solve in seconds on a laptop. AeroSandbox also comes with dozens of end-to-end-differentiable aerospace physics models, allowing you to simultaneously optimize an aircraft's aerodynamics, structures, propulsion, mission trajectory, stability, and more.

Keeping AeroSandbox easy to learn and use is a top priority. Complexity is optional - you can use AeroSandbox's built-in physics models where helpful, or you can drop in arbitrary custom physics models of your own.

pip install aerosandbox[full]

What can I do with AeroSandbox?

Use AeroSandbox to design and optimize entire aircraft:

Feather (an ultra-lightweight 1-meter-class RC motor glider)

Feather first page

SEAWAY-Mini (a solar-electric, 13' wingspan seaplane)

Seaway-Mini first page

Use AeroSandbox to support real-world aircraft development programs, all the way from your very first sketch to your first-flight and beyond:

Initial concept sketches + sizing of Dawn (a solar-electric airplane for climate science research) in AeroSandbox, Spring 2020

Dawn initial design

Dawn (later renamed SACOS) in first flight, Fall 2022

(A massive build effort with excellent engineering and coordination by Electra.aero!)

SACOS first flight

Use AeroSandbox to explore counterintuitive, complicated design tradeoffs, all at the earliest stages of conceptual design where these insights make the most difference:

Exploring how big a solar airplane needs to be to fly, as a function of seasonality and latitude

Dawn seasonality latitude tradespace

Exploring how the mission range of Firefly, a Mach 0.8 rocket drone, changes if we add an altitude limit, simultaneously optimizing aircraft design and trajectories

Firefly range ceiling trade

Exploring how many LH2 aircraft classes an airline fleet needs to cover the market, considering off-design performance

LH2 Market Coverage

Use AeroSandbox as a pure aerodynamics toolkit:

VLM simulation of a glider, aileron deflections of +-30°

VLM simulation

Aerodynamic shape optimization of a wing planform, using an arbitrary objective and constraints

Wing optimization

Optimize airfoil shapes with aerodynamic, structural, and manufacturing considerations

Airfoil optimization

Among many other discplines:

Structural optimization of a composite tube spar

Beam optimization

Electric motor analysis for propeller matching

Motor performance

Tools to analyze unconventional propulsion (e.g., LH2)

LH2 airplane three-view

Detailed weights estimation for aircraft ranging from micro-UAVs to airliners

Mass Budget

Easily interface AeroSandbox with all your favorite tools:

Other conceptual design tools (AVL, XFLR5, XFoil, ASWING, MSES, etc.)

XFoil

CAD tools via STEP export (SolidWorks, Fusion 360, etc.)

(STL, OBJ, etc. supported too)

CAD

User-provided models + code (for custom aerodynamics, structures, propulsion, or anything else - e.g., for optimizing flight through a probabilistic wind field, shown below)

Wind speed

Or, throw all the airplane-design-specific code out entirely, and use AeroSandbox purely as an optimization solver or as a solver for nonlinear systems of equations (or ODEs, or PDEs):

Optimize the 2D Rosenbrock function

Optimization

Specify the Falkner Skan ODE (nonlinear, 3rd-order BVP) and let AeroSandbox automatically take care of the discretization, solution, and even inverse solving.

FS ODE

And much, much more. Best of all, combine these tools arbitrarily without any loss in optimization speed and without any tedious derivative math, all thanks to AeroSandbox's end-to-end automatic-differentiability.

Getting Started

Installation

In short:

  • pip install aerosandbox[full] for a complete install.

  • pip install aerosandbox for a lightweight (headless) installation with minimal dependencies. All optimization, numerics, and physics models are included, but optional visualization dependencies are skipped.

For more installation details (e.g., if you're new to Python), see here.

Tutorials, Examples, and Documentation

To get started, check out the tutorials folder here! All tutorials are viewable in-browser, or you can open them as Jupyter notebooks by cloning this repository.

For a more detailed and theory-heavy introduction to AeroSandbox, please see the author's PhD thesis and master's thesis.

For a developer-oriented description of AeroSandbox internal modules, please see the developer README.

For fully-detailed API documentation, see the documentation website.

You can print documentation and examples for any AeroSandbox object by using the built-in help() function (e.g., help(asb.Airplane)). AeroSandbox code is also documented extensively in the source and contains hundreds of unit test examples, so examining the source code can also be useful.

Usage Details

Units

One final point to note: all inputs and outputs to AeroSandbox are expressed in base SI units, or derived units thereof (e.g., m, kg, sec, N, m/s, J, Pa). Since this unit system is coherent, an enormous number of quantities can be converted without any scaling factors. This improves readability and reduces the likelihood of errors.

There are only two exceptions to this SI-everywhere rule:

  1. If alternate units are noted in a variable name's suffix. For example:

    • battery_capacity → Joules
    • battery_capacity_watt_hours → Watt-hours
    • aircraft_endurance → Seconds
    • aircraft_endurance_hours → Hours
  2. Angle of attack (alpha, α) and sideslip angle (beta, β) are given in degrees due to long-standing aerospace convention. All other angles and angular rates use radians.

Also, in case of any confusion on the units of a function's inputs and outputs, units are listed on all function docstrings.

If you wish to use other units, consider using aerosandbox.tools.units to convert easily.

Project Details

Contributing

Please feel free to join the development of AeroSandbox - contributions are always so welcome! If you have a change you'd like to make, the easiest way to do that is by submitting a pull request.

The text file CONTRIBUTING.md has more details for developers and power users.

If you've already made several additions and would like to be involved in a more long-term capacity, please message me! Contact information can be found next to my name near the top of this README.

Donating

If you like this software, please consider donating to support development via PayPal or GitHub Sponsors! Proceeds will go towards more coffee for the grad students.

Bugs

Please, please report all bugs by creating a new issue!

Versioning

AeroSandbox loosely uses semantic versioning, which should give you an idea of whether or not you can probably expect backward-compatibility and/or new features from any given update.

For more details, see the changelog.

Citation & Commercial Use

If you find AeroSandbox useful in a research publication, please cite one of the following publications:

The author's PhD thesis:

@phdthesis{aerosandbox_phd_thesis,
   title = {Accelerating Practical Engineering Design Optimization with Computational Graph Transformations},
   author = {Sharpe, Peter D.},
   school = {Massachusetts Institute of Technology}, 
   year = {2024},
}

The author's Master's thesis:

@mastersthesis{aerosandbox,
    title = {AeroSandbox: A Differentiable Framework for Aircraft Design Optimization},
    author = {Sharpe, Peter D.},
    school = {Massachusetts Institute of Technology},
    year = {2021}
}

Commercial users: I'm more than happy to discuss consulting work for active AeroSandbox support if this package proves helpful - use the email address in the header of this README to get in touch.

License

MIT License applies, full terms here. In short: use AeroSandbox for anything you want (commercial or non-commercial). AeroSandbox is released in hope that it will be useful but without any warranty of merchantability (either express or implied).

If you use AeroSandbox, attribution is appreciated.

Stargazers over time

Stargazers over time

Supported by

JetBrains logo.