On the generalizability of machine-learning-assisted anisotropy mappings for predictive turbulence modelling

TL;DR

This study evaluates the generalizability of machine learning models for turbulence closure, finding they perform well within trained flow parameters but struggle to generalize across different flow types, highlighting their use for specialized applications.

Contribution

The paper systematically compares different ML models for turbulence closure and demonstrates the limitations in their ability to generalize across flow types, emphasizing their suitability for specialized modeling.

Findings

XGBoost outperforms other models in low-tuning cost and performance

ML models generalize well within trained flow parameters

Limited generalizability to new flow types

Abstract

Several machine learning frameworks for augmenting turbulence closure models have been recently proposed. However, the generalizability of an augmented turbulence model remains an open question. We investigate this question by systematically varying the training and test sets of several models. An optimal three-term tensor basis expansion is used to develop a model-agnostic data-driven turbulence closure approximation. Then, hyperparameter optimization was performed for a random forest, a neural network, and an eXtreme Gradient Boosting (XGBoost) model. We recommend XGBoost for data-driven turbulence closure modelling owing to its low-tuning cost and good performance. We also find that machine learning models generalize well to new parametric variations of flows seen in the training dataset, but lack generalizability to new flow types. This generalizability gap suggests that machine learning methods are most suited for developing specialized models for a given flow type, a problem often encountered in industrial applications.