Research map / Probabilistic and scientific modeling / Differential equations and PDEs
TurtLES: A Large-Scale Benchmark for Turbulent 3D Neural PDE Surrogates
Jarno Platenburg, Armand Kassaï Koupaï, Paola Cinnella, Patrick Gallinari
Read the paperCheck your idea against 137,217 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Introduces TURTLES, a 13 TB benchmark of implicit large-eddy simulations of 3D turbulent cylinder wakes for evaluating neural PDE surrogates.
Problem
Progress in machine-learning surrogates for CFD is limited by the lack of benchmarks capturing realistic, time-dependent turbulent flows.
Approach
A large-scale dataset of high-fidelity implicit LES trajectories of 3D turbulent cylinder wakes in the shear-layer transition regime, intended as a benchmark for turbulent 3D neural PDE surrogates.
Key results
- Dataset comprises 400 trajectories of 400 time steps each, totaling 13 TB of high-fidelity implicit LES data of 3D turbulent cylinder wakes.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use CALM-PDE continuous-convolution operator instead of UPT | neural operator surrogate · TurtLES long rollout | VRMSE steps 301-399: 5.12 -> 7.89e-1 | Improved |
| Use CALM-PDE instead of Transolver++ | neural operator surrogate · TurtLES next frame | VRMSE: 1.06 -> 7.26e-2 | Improved |
| Use Transolver++ for rollout | Transolver++ · TurtLES rollout | VRMSE steps 301-399: 2.33 -> 31.24 | Hurt |
| Use GINO baseline | GINO · TurtLES training | convergence | Hurt |
| Compare Transolver++ spectral error on high band vs other models | Transolver++ · TurtLES next frame | high-frequency spectral error: 4.41e-1 vs about 4.4e-2 | Hurt |
| Use GINO on low-frequency band | GINO · TurtLES next frame | low-frequency spectral error: 1.32e-1 vs 4.69e-2 (UPT) | Hurt |
| Replace PyFR solver with neural surrogates | surrogates · TurtLES inference cost | GPU-seconds per trajectory: 13160 -> 835 (UPT) | Improved |
| Use persistence baseline (repeat initial state) for spectral comparison | persistence baseline · TurtLES spectra | time-averaged energy spectrum | No clear effect |
Experimental setup
Benchmarks: TURTLES
Closest papers by meaning
- EddyFormer: Accelerated Neural Simulations of Three-Dimensional Turbulence at Scale (NeurIPS 2025 · similarity 0.88)
- From Zero to Turbulence: Generative Modeling for 3D Flow Simulation (ICLR 2024 · similarity 0.87)
- Deep Learning-based Algebraic Reynolds Stress Closures for RANS simulations of Turbulent Flows (NeurIPS 2026 · similarity 0.87)
- Geometric and Physical Constraints Synergistically Enhance Neural PDE Surrogates (ICML 2025 · similarity 0.86)
- PEINR: A Physics-enhanced Implicit Neural Representation for High-Fidelity Flow Field Reconstruction (ICML 2025 · similarity 0.86)
- PIORF: Physics-Informed Ollivier-Ricci Flow for Long–Range Interactions in Mesh Graph Neural Networks (ICLR 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
