iiinova — where your digital self keeps evolving

Bernat Font

@fontb.bsky.social

360 followers 293 following 63 posts

Assistant Professor in Data-Informed CFD at TU Delft. Interested in turbulent flow simulations 🌀 data-driven models 🤖 numerical methods ∫→∑ and high-performance computing ⚡️. I generally talk about open source, music, and science.

Posts Replies Media

Bernat Font @fontb.bsky.social · 28/09/2025

000

Bernat Font @fontb.bsky.social · 28/09/2025

the best things are also thanks to fluid dynamics!

140

Bernat Font @fontb.bsky.social · 22/07/2025

The WaterLily.jl article is now available on Computer Physics Comms! @gabrielweymouth.bsky.social and I review our CFD solver's design, performance, accuracy, and also show some very cool examples (like the one below). Check it out! 🔗 doi.org/10.1016/j.cpc.2025.109748 #CFD #GPU #HPC #JuliaLang

0144

Bernat Font @fontb.bsky.social · 25/05/2025

New WaterLily version is out 🌀! github.com/WaterLily-jl... #JuliaLang #CFD #GPU #LES 1/2

2111

Bernat Font @fontb.bsky.social · 23/05/2025

Our latest preprint on "Mesh-free sparse identification of nonlinear dynamics" is out! In collaboration with Mars Gao and Nathan Kutz, we explore a novel SINDy framework to uncover dynamical systems from scarce, noisy, and non-uniformly sampled data. 🔗 arxiv.org/abs/2505.16058 #SINDy #PDE #SciML #ML

Mesh-free SINDy framework and 1D Burgers' equation as example.

140

Bernat Font @fontb.bsky.social · 05/05/2025

WaterLily and @makie.org make a great team! #CFD #JuliaLang #GPU #DataViz

1153

Bernat Font @fontb.bsky.social · 27/03/2025

It does! With N=64^3 the whole inertial subrange is mostly resolved. Here are the results with N=32^3. I had to fine-tune the Smagorinsky constant from 0.18 (used in the N=64^3) to 0.20 to nicely match the reference experimental data.

010

Bernat Font @fontb.bsky.social · 26/03/2025

using an explicit LES model and a central difference scheme for the convective flux, we match nicely experimental results from the 70s! (doi.org/10.1017/S002...) Results with N=64^3 cells and Smagorinsky-Lilly turbulence model

Energy cascade of decaying homogeneous isotropic turbulence with N=64^3 cells

170

Bernat Font @fontb.bsky.social · 20/03/2025

We would like to invite you to join the next Large-scale CFD online seminar by Dr. Francesc Verdugo (UV Amsterdam) on "Large-scale sparse linear algebra with PartitionedArrays" github.com/PartitionedA... Date: 27/03/2025, 5pm CET Seminar link: tinyurl.com/4hy47ezv #HPC #MPI #JuliaLang

Abstract: We present PartitionedArrays, a large-scale sparse linear algebra library fully implemented in the Julia programming language. The key differentiating feature of PartitionedArrays with respect to related libraries is that it introduces an alternative programming model to implement its distributed algorithms. This model allows one to decouple the logic behind data partition and communication from the number of processing units used to run the code. A logically distributed algorithm can be run both on a single and multiple process. This delivers an improved balance between code productivity and performance in the development of large-scale scientific computing applications. The parallel logic can be debugged on a single process, avoiding deadlocks, race conditions, and other practical challenges of debugging parallel code, and permits using standard sequential debuggers. At the same time, the resulting code can be deployed on a supercomputer with performance similar of native MPI-based applications. PartitionedArrays provides distributed linear algebra data structures (vectors and sparse matrices), and key computational kernels (distributed sparse matrix-vector product and sparse matrix-matrix products), as well as distributed algebraic multigrid preconditioners. These tools have been successful in the implementation of finite element solvers scaling to tens of thousands of CPU cores. In this presentation, we show the software design of PartitionedArrays and a hands-on overview to its main features. The source code of the library is freely available at https://github.com/PartitionedArrays/PartitionedArrays.jl.

040

Bernat Font @fontb.bsky.social · 12/03/2025

Playing a bit with our WaterLily.jl #CFD #JuliaLang solver and homogeneous isotropic decaying turbulence 🌀 128^3 grid running on my local #GPU ⚡ visualizing in real-time with @makie.org github.com/WaterLily-jl...

0266

Bernat Font @fontb.bsky.social · 06/03/2025

Another great output from our DRL for active flow control team! Lead by Pol Suarez, in this paper we show how DRL helps to achieve significant drag reduction for flow around a cylinder at Re=3900 with reduced mass cost efficiency! #CFD #RL #DRL #turbulence 📎 doi.org/10.1007/s10494-025-00642-x

Flow past a circular cylinder at Re=3900 with/without active flow control

062

Bernat Font @fontb.bsky.social · 13/02/2025

I only realised during the second season of #Severance !

Actor John Turturro in The Big Lebowski (movie) and Severance (TV show)

030

Bernat Font @fontb.bsky.social · 13/02/2025

Tuesday next week (18th February, 5pm CET), Dr. Soledad Le Clainche from Universidad Politécnica de Madrid will present at the Large-scale CFD online seminar series: "From modal decompositions to machine learning: new tools and applications in fluid dynamics".

Details of the seminar in image format. Written format available in https://www.linkedin.com/events/7295741410593443841

120

Bernat Font @fontb.bsky.social · 10/02/2025

We have succeeded in controlling a turbulent separation bubble at a high Reynolds number using deep reinforcement learning! The paper, published in @naturecomms.bsky.social (@natureportfolio.nature.com), can be accessed here: www.nature.com/articles/s41... #ML #DRL #CFD #LES

A turbulent separation bubble visualized with Q-criterion vortical structures.

2143

Bernat Font @fontb.bsky.social · 21/11/2024

Following the Scholar Goggler trend, hello BlueSky! Try yours here: scholargoggler.com

111