Navier-Stokes

Experience 21st century turbulence

Differentiable fluid solver written in Julia

Getting started

GPU acceleration

Fast and efficient 2D/3D kernels for CPU and GPU with CUDA.jl

Differentiable physics

Backpropagate through the solver using Zygote.jl to optimize closure models

Neural network integration

Integrate neural network closure models with Lux.jl

Problems

Solve for decaying turbulence, channel flows, actuator disks, Rayleigh-Bénard convection, and more