PyFR

An open-source platform for fast, robust and accurate Scale-Resolving Simulations of industrial flow problems.

Get PyFR v2.1 Documentation

Open source

PyFR is released under a 3-Clause BSD License to enable sharing and collaboration.

High-order accurate

PyFR utilises the Flux Reconstruction approach to achieve arbitrary-order spatial accuracy.

Unstructured meshes

PyFR can operate with mixed elements and unstructured meshes to enable simulations of complex geometries.

Cross-platform

PyFR can target a range of hardware including CPUs, Nvidia GPUs, AMD GPUs, Intel GPUs, and Apple GPUs.

Challenge and opportunity

Reynolds Averaged Navier-Stokes (RANS) simulations are cheap and robust, but accuracy is limited for complex unsteady industrial flows. Scale-Resolving Simulations (SRS) overcome the accuracy limitations of RANS, but can lack robustness, and have historically been too expensive for industrial adoption.

The use of CFD in the aerospace design process is severely limited by the inability to accurately and reliably predict turbulent flows with significant regions of separation.

Mission and vision

Our mission is to develop an SRS capability that is fast and robust, with a vision to enable industrial adoption of SRS, and hence a paradigm-shift in industrial design capability.

An international community

"PyFR has been a game changing tool for investigating complex turbulent flow phenomenon in the wake of reentry capsules and predicting their dynamic stability characteristics. It’s capability to run on modern GPU supercomputers has been an enabler for running high-fidelity scale resolving simulations in a matter of hours, which would otherwise have taken many weeks to months to perform using legacy CFD tools."

Rathakrishnan Bhaskaran

Branch Technical Lead | NASA Ames Research Center

"PyFR is a great example of what's next in classical computing. We've been using it as as a performance benchmark for years, but have now progressed to advancing its use in industrial applications to deliver a step change in modelling fidelity of turbulent flows. It's usable out-of-the-box, extensible and has a vibrant ecosystem and support community."