JAX-SPH: A Differentiable Smoothed Particle Hydrodynamics Framework

TL;DR

JAX-SPH introduces a differentiable, Python-based Smoothed Particle Hydrodynamics framework compatible with deep learning, enabling advanced physics simulations and inverse problem solving.

Contribution

It presents the first Lagrangian fluid simulator in JAX with verified gradients, extending SPH algorithms and demonstrating applications in inverse problems.

Findings

Built a differentiable SPH framework in JAX

Verified gradient correctness through the solver

Enabled inverse problem solving and Solver-in-the-Loop applications

Abstract

Particle-based fluid simulations have emerged as a powerful tool for solving the Navier-Stokes equations, especially in cases that include intricate physics and free surfaces. The recent addition of machine learning methods to the toolbox for solving such problems is pushing the boundary of the quality vs. speed tradeoff of such numerical simulations. In this work, we lead the way to Lagrangian fluid simulators compatible with deep learning frameworks, and propose JAX-SPH - a Smoothed Particle Hydrodynamics (SPH) framework implemented in JAX. JAX-SPH builds on the code for dataset generation from the LagrangeBench project (Toshev et al., 2023) and extends this code in multiple ways: (a) integration of further key SPH algorithms, (b) restructuring the code toward a Python package, (c) verification of the gradients through the solver, and (d) demonstration of the utility of the gradients for solving inverse problems as well as a Solver-in-the-Loop application. Our code is available at https://github.com/tumaer/jax-sph.