Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
162 tokens/sec
GPT-4o
7 tokens/sec
Gemini 2.5 Pro Pro
45 tokens/sec
o3 Pro
4 tokens/sec
GPT-4.1 Pro
38 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Classically studied coherent structures only paint a partial picture of wall-bounded turbulence (2410.23189v2)

Published 30 Oct 2024 in physics.flu-dyn

Abstract: For the last 140 years, the mechanisms of transport and dissipation of energy in a turbulent flow have not been completely understood due to the complexity of this phenomenon. The dissipation of energy due to turbulence is significative, and understanding turbulence physics is crucial for fighting the present climate emergency. Previous research has focused on analyzing the so-called coherent structures of the flow (Q events, streaks, and vortices), which are regions of high turbulence transport, high/low streamwise fluctuation, and rotation, respectively. However, the connection between these classically studied structures and the flow development is still uncertain. To fill this gap, here we show a data-driven methodology for objectively identifying high-importance regions in a turbulent flow. A deep-learning model is trained to predict a future state of a turbulent channel flow and the gradient-SHAP explainability algorithm is used to calculate the importance of each grid point for such a prediction. Finally, high-importance regions are computed using the SHAP data, analyzing and comparing their characteristics with those of the other coherent structures. The SHAP analysis provides an objective way to identify the regions of highest importance in the turbulent flow, which exhibit different levels of agreement with the classically studied structures.

Definition Search Book Streamline Icon: https://streamlinehq.com
References (1)

Summary

  • The paper presents a deep learning and explainable AI framework to identify high-importance regions in wall-bounded turbulence, offering a new perspective beyond classical methods.
  • Using SHAP values, the study identifies structures that align with classical coherent structures like sweeps and ejections but also reveal critical regions previously undefined.
  • This objective identification of key turbulent regions can lead to improved theoretical models and targeted flow control strategies for enhanced energy efficiency.

Overview of the Paper on Coherent Structures in Wall-bounded Turbulence

The paper presents an innovative approach to understanding wall-bounded turbulence through the application of deep learning and explainable artificial intelligence (XAI). By leveraging these advanced methodologies, the research provides a novel framework for identifying and analyzing high-importance regions within turbulent flows. This approach departs from classical methods that often focus on specific coherent structures such as Q events, streaks, and vortices.

Methodology and Findings

The primary objective of the paper is to bridge the gap in understanding the intricate dynamics of turbulence, specifically the transport and dissipation of energy across different regions of a flow. The authors introduce a data-driven methodology employing a deep-learning model to predict future states of turbulent channel flows. A U-net architecture trained on a comprehensive dataset of 10,000 instantaneous flow fields serves as the foundation for this predictive model, achieving a notable relative error of around 1% across all velocity components.

Upon establishing a reliable predictive model, the explainability algorithm known as gradient-SHAP is used to determine the significance of each grid point in influencing the model's predictions. This produces a quantifiable measure of importance for various spatial regions within the flow, termed SHAP values. These values are then utilized to form SHAP-based structures, objectively identifying regions of utmost significance in the flow's evolution.

A key result of the paper is the high degree of agreement observed between SHAP-based structures and previously studied coherent structures, particularly sweeps and ejections. Notably, approximately 60% of SHAP structures coincide with intense Reynolds-stress regions, although some critical areas remain unexplained by traditional definitions, indicating the finely tuned ability of SHAP values to capture complex flow dynamics beyond the existing paradigms.

Theoretical and Practical Implications

This research suggests several implications for both the theoretical paper and control of turbulent flows. The work provides a more comprehensive framework for defining and analyzing coherent structures, potentially influencing future turbulence modeling and containment strategies in industrial applications. The identification of high-importance regions could enable targeted interventions for energy consumption reduction, addressing critical environmental and engineering challenges.

Moreover, the method holds promise for extending beyond the present paper, such as exploring longer prediction horizons or varying predictive targets like wall-shear stresses. These future directions could yield further insights into the control and optimization of fluid dynamics systems.

Conclusion

The integration of deep learning and XAI in the examination of turbulence presents a robust and objective approach to understanding the complex behavior of wall-bounded turbulent flows. This paper not only enhances the theoretical framework surrounding coherent structures but also lays the groundwork for practical applications in turbulence management and energy efficiency, representing a significant stride towards advancing turbulence research in both computational and experimental domains.

Youtube Logo Streamline Icon: https://streamlinehq.com