XRePIT: AI-CFD Hybrid Solver Tutorial

XRePIT Tutorial: Automated Hybrid CFD-ML Simulations with OpenFOAM A quick walkthrough of XRePIT (eXtensible RePIT), an open-source framework that couples machine learning acceleration with solver-based correction for stable, long-horizon CFD rollouts. In this video I cover installation, simple setup, running 2D, 3D, and different neural architecture case. XRePIT builds on the Residual-based Physics-Informed Transfer learning (RePIT) strategy, turning what was a manual 2D proof-of-concept into a fully automated, physics-aware pipeline. It achieves the first stable, accelerated rollouts past 10,000 timesteps, scaling from 2D benchmarks to 3D applications. What's covered: 00:00 Introduction 01:41 Installation 04:53 Configuration explanation 13:27 2D case initialization and training 19:16 Hybrid simulation from checkpoint 20:55 3D case in XRePIT (parallelized OpenFOAM) 26:20 Changing the neural accelerator 28:57 Closing remarks Read the paper: "XRePIT: A deep learning-computational fluid dynamics hybrid framework implemented in OpenFOAM for fast, robust, and scalable unsteady simulations" arXiv: https://arxiv.org/abs/2510.21804 PDF: https://arxiv.org/pdf/2510.21804 Links: Code / repo: https://github.com/POSTECH-NINE/repitframe... #CFD #MachineLearning #OpenFOAM #Hybrid #AI-CFD #ScientificComputing #FluidDynamics #DeepLearning #XRePIT