Physics-Guided Machine Learning for Uncertainty Quantification in Turbulence Models

TL;DR

This paper introduces a hybrid physics-guided machine learning approach that enhances the Eigenspace Perturbation Method for turbulence models, resulting in more accurate and better-calibrated uncertainty quantification in turbulent flow predictions.

Contribution

It proposes a CNN-based modulation of EPM perturbations to improve uncertainty calibration while maintaining physical consistency in turbulence modeling.

Findings

Hybrid ML-EPM provides tighter uncertainty bounds.

Improved calibration over baseline EPM.

Effective across canonical turbulence cases.

Abstract

Predicting the evolution of turbulent flows is central across science and engineering. Most studies rely on simulations with turbulence models, whose empirical simplifications introduce epistemic uncertainty. The Eigenspace Perturbation Method (EPM) is a widely used physics-based approach to quantify model-form uncertainty, but being purely physics-based it can overpredict uncertainty bounds. We propose a convolutional neural network (CNN)-based modulation of EPM perturbation magnitudes to improve calibration while preserving physical consistency. Across canonical cases, the hybrid ML-EPM framework yields substantially tighter, better-calibrated uncertainty estimates than baseline EPM alone.