Introduction

Welcome to the documentation for Fireshape.

Overview

Fireshape is a shape optimization toolbox for the finite element library Firedrake.

To set up a shape optimization problem, all you need to provide is the mesh on an initial guess, the shape functional, and the weak-form of its PDE-contraint.

Fireshape computes adjoints and assembles first and second derivatives for you using pyadjoint, and it solves the optimization problem using the rapid optimization library (ROL).

Features

Fireshape neatly distinguishes between the discretization of control and state variables. To discretize the control, you can choose among finite elements, B-splines, and the free-form approach. To discretize the state, you have access to all finite element spaces offered by Firedrake.

Fireshape relies on the mesh-deformation approach to update the geometry of the domain. By specifying the metric of the control space, you can decide whether meshes should be updated using Laplace or elasticity equations. In 2D, you can also use the elasticity equation corrected with Cauchy-Riemann terms, which generally leads to very high-quality meshes.

Where to go from here

If you are interested in using Fireshape, do not hesitate to get in contact. You can do so by sending an email to a.paganini@leicester.ac.uk or to wechsung@nyu.edu.

You can find information on how to install Fireshape on the page Installation.

On the page Example 1: Level Set, we show how to solve a toy shape optimization problem.

On the page Example 2: L2-tracking, we show how to solve a shape optimization problem constrained to a linear boundary value problem.

On the page Example 3: Kinetic energy dissipation in a pipe, we show how to solve a shape optimization problem constrained to a nonlinear boundary value problem and a volume constraint.

On the page Using ROL in Fireshape, we give a very brief introduction to ROL.

Finally, you can find a manuscript about Fireshape here.