peterdsharpe/AeroSandbox

What it solves

AeroSandbox simplifies the design and optimization of aircraft and other engineered systems. It addresses the difficulty of optimizing large-scale design problems—which often involve tens of thousands of variables—by providing a framework where physics models are end-to-end differentiable, allowing for rapid convergence and high performance on standard hardware.

How it works

The package combines a NumPy-like syntax for ease of use with modern automatic differentiation. This allows the software to calculate gradients automatically, eliminating the need for tedious manual derivative math. It includes a library of differentiable aerospace physics models covering aerodynamics, structures, propulsion, and mission trajectories, but also allows users to integrate their own custom physics models or use it as a standalone solver for nonlinear systems of equations, ODEs, and PDEs.

Who it’s for

It is designed for aerospace engineers, researchers, and hobbyists who need to perform conceptual aircraft design, shape optimization, or complex system trade-off analysis.

Highlights

• End-to-End Differentiability: Enables simultaneous optimization of multiple disciplines (aerodynamics, propulsion, etc.) with high speed.
• Broad Aerospace Toolkit: Includes tools for Vortex Lattice Method (VLM) simulations, airfoil optimization, and detailed weight estimation.
• Flexible Integration: Interfaces with external tools like XFoil, AVL, and CAD software via STEP export.
• General Purpose Solver: Can be used independently of aircraft design as a general optimization solver or for solving differential equations.