Physics-Preserving Latent Compression for Zero-Shot Resolution Transfer in 3D Turbulence

2026-06-19

AI Summary

This work addresses the challenge of high-resolution turbulence modeling, which is hindered by the prohibitive cost of direct numerical simulation and the scarcity of full-resolution data, while existing compression methods struggle to preserve physical fidelity and generalize across resolutions. To overcome this, the authors propose PPLC, a local patch-based latent-space compression framework that, for the first time, enables zero-shot cross-resolution transfer without retraining. Leveraging scale similarity in the inertial subrange, PPLC integrates a shared variational autoencoder, exact mean preservation, zero-mean fluctuation encoding, an invertible Haar wavelet front-end, translation-consistency regularization, and overlap-aware reconstruction to rigorously enforce turbulence physics in the latent representation. Evaluated on 1024³ turbulence fields, PPLC substantially outperforms both classical and learning-based baselines, accurately preserving key diagnostic quantities—such as dissipation rate, vorticity, energy spectrum, and incompressibility—in zero-shot reconstructions.

Abstract

High-resolution turbulence modeling is essential for scientific computing, but remains constrained by the cost of direct numerical simulation and the scarcity of full-resolution data. Existing scientific compressors reduce storage but typically operate on per-frame representations, whereas learned compressors yield compact latents that are often resolution-dependent and weakly aligned with the physics of turbulence. This raises the need for a compression framework that reduces data size, preserves physical diagnostics, and transfers from low-resolution training fields to high-resolution test fields without retraining. In this paper, we propose Physics-Preserving Latent Compression (PPLC), a patch-local latent compressor for three-dimensional turbulence. Motivated by inertial-range scale similarity, PPLC treats fixed-size patches as transferable units and applies a shared variational autoencoder independently of the global grid size. It combines exact mean preservation, zero-mean fluctuation encoding, an invertible Haar wavelet front-end, shift-consistency regularization, and overlap-aware reconstruction. Instantiated on forced isotropic turbulence, PPLC is trained only on stride-downsampled 256^3 fields and transfers zero-shot to 1024^3 fields. Experiments show that PPLC improves the balance between reconstruction accuracy and physical fidelity over classical and learned baselines, keeping diagnostics such as dissipation, enstrophy, energy spectra, and incompressibility closer to the ground truth. Beyond turbulence compression, PPLC offers a general strategy for physics-preserving latent representations that support data-efficient scientific surrogate modeling.

Problem

• turbulence compression
• zero-shot resolution transfer
• physics-preserving representation
• latent compression
• scientific computing

Innovation

• physics-preserving compression
• zero-shot resolution transfer
• latent representation
• 3D turbulence
• variational autoencoder