An Iterative Machine-Learning Framework for RANS Turbulence Modeling

TL;DR

This paper introduces an iterative machine-learning framework integrated with RANS turbulence modeling, demonstrating improved accuracy and generalization in predicting turbulent flows across different scenarios.

Contribution

It proposes a novel iterative ML-RANS framework that ensures reproducibility and effectively predicts turbulent flows, including separated flows, with limited training data.

Findings

Accurately predicts mean flow and turbulence variables in channel flows.

Reliable interpolation between different Reynolds numbers.

Outperforms traditional models in flow over periodic hills.

Abstract

Machine-learning (ML) techniques provide a new and encouraging perspective for constructing turbulence models for Reynolds-averaged Navier--Stokes (RANS) simulations. In this study, an iterative ML-RANS computational framework is proposed that combines an ML algorithm with transport equations of a conventional turbulence model. This framework maintains a consistent procedure for obtaining the input features of an ML model in both the training and predicting stages, ensuring a built-in reproducibility. The effective form of the closure term is discussed to determine suitable target variables for the ML algorithm, and the multi-valued problem of existing constitutive theory is studied to establish a proper regression system for ML algorithms. The developed ML model is trained under a cross-case strategy with data from turbulent channel flows at three Reynolds numbers and \textit{a posteriori} simulations of channel flows show that the framework is able to predict both the mean flow field and turbulent variables accurately. Interpolation tests for the channel flow show the proposed framework can reliably predict flow features that lie between the minimum and maximum Reynolds numbers associated with the training data. A further test related to the flow over periodic hills also demonstrates a better result than a traditional turbulence model, indicating a promising predictive capability of the developed ML model for separated flow even though the model is only trained with planar channel flow data.