Hydrogen combustion, high-fidelity CFD and the future of aircraft propulsion are the focus of this conversation with Dr. Daniel Mira, Head of the Propulsion Technologies Group at the Barcelona Supercomputing Center. Neil and Dani discuss why reacting flows are so difficult to simulate, how hydrogen changes combustion and aircraft design, the limits of RANS, LES and DNS, GPU-native solvers, coding agents and AI surrogate models.

Full episode, corrected transcript and resources:

https://neilashton.co.uk/podcasts/s4-e6-daniel-mira-on-hydrogen-combustion-modelling-and-future-propulsion/

Topics

Why reacting flows are so computationally difficult

Hydrogen versus hydrocarbon combustion

When hydrogen could reach commercial aviation

How engines and aircraft must be redesigned

Industrial trust in high-fidelity combustion CFD

RANS, LES and DNS for reacting flows

Chemistry, load balancing and computational cost

Wall modelling in combustion LES

GPU acceleration and solver redesign

Coding agents for scientific software

AI surrogate models and digital engineering workflows

Selected resources

Daniel Mira and the Propulsion Technologies Group

https://ptg.bsc.es/?p=44

Propulsion Technologies Group — research lines

https://ptg.bsc.es/research-lines/

BSC — Combustion research

https://www.bsc.es/research-development/research-areas/engineering-simulations/combustion

Center of Excellence in Combustion (CoEC)

https://coec-project.eu/

High-fidelity simulations of the mixing and combustion of a technically premixed hydrogen flame

https://upcommons.upc.edu/entities/publication/08a27c10-cb13-4357-a3ab-8e9ec1d706cc

Chapters

00:00 Podcast intro

00:39 Introducing Daniel Mira

03:00 Conversation begins

04:55 Why combustion CFD is so hard

10:23 Daniel’s path into hydrogen and jet-engine combustion

12:48 Hydrogen versus hydrocarbon combustion

17:58 Industrial adoption of hydrogen

20:54 Gas turbines, aviation and fuel infrastructure

25:35 How jet engines must change

30:43 Redesigning the whole aircraft

34:46 What will trigger commercial adoption?

37:27 Why aerospace projects take a decade

42:14 RANS, LES and DNS for reacting flows

44:31 Replacing expensive tests with high-fidelity CFD

46:01 The biggest accuracy gaps in combustion LES

49:26 Where the computational cost goes

52:06 Chemistry, species and source-term bottlenecks

55:35 Wall modelling in combustion LES

59:49 GPUs, algorithms and solver redesign

01:08:52 Can coding agents accelerate combustion CFD?

01:12:27 AI surrogate models for combustion

01:24:20 Closing thoughts

Differentiable physics, neural emulators and foundation models for PDEs are the focus of this conversation with Professor Nils Thuerey, head of the Physics-based Simulation group at TUM. Neil and Nils discuss PhiFlow, PICT, Tadpole, scalable 3D transformers, online synthetic data, open datasets, world models and agents that call physics simulators.

Full episode, corrected transcript and resources:

https://neilashton.co.uk/podcasts/s4-e5-prof-nils-thuerey-on-differentiable-physics-and-foundation-models/

Topics

Differentiable physics and physics-based deep learning

PhiFlow and differentiable simulation across ML frameworks

When neural emulators can outperform their training data

Foundation models for PDEs and synthetic online training

Scalable 3D transformers and high-resolution simulations

LES, temporal data and correlated CFD datasets

Open-source tools, startups and physics-aware world models

AI agents that call physics simulators

Papers

Neural Emulator Superiority: When Machine Learning for PDEs Surpasses its Training Data

https://arxiv.org/abs/2510.23111

Tadpole: Autoencoders as Foundation Models for 3D PDEs with Online Learning

https://arxiv.org/abs/2605.15284

P3D: Scalable Neural Surrogates for High-Resolution 3D Physics Simulations with Global Context

https://arxiv.org/abs/2509.10186

PICT — A Differentiable, GPU-Accelerated Multi-Block PISO Solver for Simulation-Coupled Learning Tasks in Fluid Dynamics

https://arxiv.org/abs/2505.16992

PhiFlow: Differentiable Simulations for PyTorch, TensorFlow and JAX

https://proceedings.mlr.press/v235/holl24a.html

Physics-based Deep Learning

https://arxiv.org/abs/2109.05237

Learning to Control PDEs with Differentiable Physics

https://arxiv.org/abs/2001.07457

Solver-in-the-Loop: Learning from Differentiable Physics to Interact with Iterative PDE-Solvers

https://arxiv.org/abs/2007.00016

tempoGAN: A Temporally Coherent, Volumetric GAN for Super-resolution Fluid Flow

https://arxiv.org/abs/1801.09710

Deep Learning Methods for Reynolds-Averaged Navier-Stokes Simulations of Airfoil Flows

https://arxiv.org/abs/1810.08217

WeatherBench: A Benchmark Dataset for Data-Driven Weather Forecasting

https://arxiv.org/abs/2002.00469

SuperWing: A Comprehensive Transonic Wing Dataset for Data-Driven Aerodynamic Design

https://arxiv.org/abs/2512.14397

Links

Nils Thuerey and the Physics-based Simulation group

https://ge.in.tum.de/about/n-thuerey/

Chapters

00:00 Podcast intro

00:39 Introducing Prof. Nils Thuerey

04:13 Conversation begins

05:13 From Computational Numerics to Graphics and Visual Effects

07:17 Physics-Based Deep Learning Before ChatGPT

10:01 CNNs, Graphics and the Move into Engineering Applications

12:37 PhiFlow and Differentiable Physics

14:13 Can Neural Emulators Surpass Their Training Data?

18:00 The Promise and Limits of Foundation Models for PDEs

20:43 Tadpole and Synthetic Online Pre-Training

24:07 From Canonical PDEs to Navier-Stokes and Industrial CFD

26:35 What Do Foundation Models Actually Learn?

28:36 PDE Pre-Training vs. Millions of CFD Simulations

33:08 Scaling 3D Transformers and Training Infrastructure

35:58 Generating and Training on Data in Real Time

38:00 LES, Temporal Data and Turbulence

42:15 Overfitting and Correlated Simulation Data

44:27 Bringing Differentiable Solvers Back into the Loop

45:31 WeatherBench, APEBench and the Value of Benchmarks

47:09 SuperWing, Open Datasets and Commercial Data

51:31 Open Source, Commercial Models and a Technical Oscar

56:17 Academia, Startups and Industry

01:00:55 What Will Change Over the Next Five Years?

01:02:07 World Models and the Need for Physics

01:08:19 Agents, Tool Use and Calling Physics Simulators

01:11:22 Career Advice for AI and Simulation

01:13:54 Closing Thoughts

RANS uncertainty, data-driven turbulence modeling and AI for Science are the focus of this conversation with Professor Paola Cinnella, Professor of Fluid Mechanics at Sorbonne University and Director of SCAI. Neil and Paola discuss high-order methods, dense gases, Bayesian uncertainty, AirfRANS, surrogate modeling, scientific publishing and education in the AI era.

Full episode, corrected transcript and resources:

https://neilashton.co.uk/podcasts/s4-e4-prof-paola-cinnella-on-ai-for-science-and-fluid-mechanics/

Topics

Fluid mechanics, CFD and high-order schemes

Dense gases, real-gas effects and expansion shockwaves

Uncertainty quantification and Bayesian methods

RANS turbulence-model uncertainty

AirfRANS and CFD datasets for machine learning

Turbulence modeling vs. surrogate modeling

Scientific publishing and ML-for-CFD standards

SCAI and AI for Science

Education, ChatGPT and centaur scientists

Papers

Quantification of model uncertainty in RANS simulations: A review — Heng Xiao, Paola Cinnella

https://doi.org/10.1016/j.paerosci.2018.10.001

Discovery of Algebraic Reynolds-Stress Models Using Sparse Symbolic Regression — Martin Schmelzer, Richard P. Dwight, Paola Cinnella

https://doi.org/10.1007/s10494-019-00089-x

Bayesian estimates of parameter variability in the k-epsilon turbulence model — W.N. Edeling, P. Cinnella, R.P. Dwight, H. Bijl

https://doi.org/10.1016/j.jcp.2013.10.027

AirfRANS: High Fidelity Computational Fluid Dynamics Dataset for Approximating Reynolds-Averaged Navier-Stokes Solutions

https://arxiv.org/abs/2212.07564

Data-driven turbulence modeling — Paola Cinnella

https://arxiv.org/abs/2404.09074

Direct numerical simulations of supersonic turbulent channel flows of dense gases — Luca Sciacovelli, Paola Cinnella, Xavier Gloerfelt

https://doi.org/10.1017/jfm.2017.237

Links

Paola Cinnella named Director of SCAI

https://scai.sorbonne-universite.fr/news/paola-cinnella-new-director

SCAI

https://scai.sorbonne-universite.fr/

Paola Cinnella — HAL publications

https://cv.hal.science/paola-cinnella

Paola Cinnella — Google Scholar

https://scholar.google.com/citations?hl=fr&user=wBRA0JAAAAAJ

ERCOFTAC SIG 54 — Machine Learning for Fluid Dynamics

https://www.ercoftac.org/special_interest_groups/54-machine-learning-for-fluid-dynamics/master-of-science-internships/

Chapters

00:00 Podcast intro

00:39 Introducing Prof. Paola Cinnella

03:28 Conversation begins

03:56 How Paola Found Fluid Mechanics

07:09 Moving from Italy to France

08:37 High-Order Schemes and Compressible Flows

09:30 Building an Academic Career

12:06 Dense Gases and Uncertainty Quantification

15:16 Expansion Shockwaves and Real-Gas Effects

19:17 Returning to Paris and Academic Mobility

24:52 Academia, Passion and Persistence

27:51 Bayesian Methods and Turbulence Uncertainty

30:47 Learning Statistics Across Disciplines

33:07 LearnFluidS, AirfRANS and CFD Datasets

36:33 Skepticism and Physics in ML Turbulence Modeling

40:41 Could ML Lead to a Universal Turbulence Model?

42:59 Turbulence Models, Surrogate Models and RANS

45:03 Why LES Alone Cannot Solve Optimization

47:15 Multi-Fidelity Modeling

49:08 What Computers & Fluids Looks for in ML-for-CFD Papers

54:05 CFD Metrics vs. Machine-Learning Metrics

57:13 Overselling, Publication Pressure and Quality

01:02:22 SCAI and AI for Science

01:06:07 Cross-Disciplinary AI for Science

01:09:26 Education in the AI Era

01:12:44 Critical Thinking and AI Outputs

01:18:15 AI as a Companion, Not a Replacement

01:21:42 AlphaFold and the Future of Discovery

01:23:43 Training Centaur Scientists

01:25:11 Closing Thoughts