AI2030 Responsible AI Tools Marketplace

Key Takeaways

SHAP provides interpretable explanations for "black box" ML models in fluid dynamics: The paper demonstrates that the Shapley Additive Explanations (SHAP) algorithm can successfully explain why neural networks and other ML models make specific predictions about fluid flow behavior, addressing a major challenge in applying powerful but opaque ML algorithms to scientific domains.
Feature importance rankings from SHAP align with established physical understanding: When tested on wall-bounded shear flows, SHAP identified only three factors as major predictors of relaminarisation events (laminar profile, streamwise vortex, and streak instability)—findings that matched known physics, validating that the algorithm captures genuine physical mechanisms rather than spurious correlations.
Explainable AI methods enable scientific discovery, not just prediction: By revealing which physical features drive ML predictions, SHAP allows researchers to gain new insights into fluid dynamics phenomena rather than simply accepting accurate predictions as "black boxes," demonstrating how interpretability serves the scientific process.
Explainability methods can scale from simple to complex datasets: The paper shows SHAP's effectiveness progresses from controlled low-dimensional models (where ground truth is known) to larger, real-world fluid dynamics datasets, suggesting interpretable AI approaches are practical for increasingly complex scientific applications.
Responsible AI in science requires transparency about model reasoning: This work illustrates why scientists should demand explainability from ML tools—understanding how models reach conclusions is essential for validating results and ensuring ML augments rather than replaces domain expertise.

Abstract

Machine Learning (ML) is becoming increasingly popular in fluid dynamics. Powerful ML algorithms such as neural networks or ensemble methods are notoriously difficult to interpret. Here, we introduce the novel Shapley Additive Explanations (SHAP) algorithm (Lundberg & Lee, 2017), a game-theoretic approach that explains the output of a given ML model, in the fluid dynamics context. We give a proof of concept concerning SHAP as an explainable AI method providing useful and human-interpretable insight for fluid dynamics. To show that the feature importance ranking provided by SHAP can be interpreted physically, we first consider data from an established low-dimensional model based on the self-sustaining process (SSP) in wall-bounded shear flows, where each data feature has a clear physical and dynamical interpretation in terms of known representative features of the near-wall dynamics, i.e. streamwise vortices, streaks and linear streak instabilities. SHAP determines consistently that only the laminar profile, the streamwise vortex, and a specific streak instability play a major role in the prediction. We demonstrate that the method can be applied to larger fluid dynamics datasets by a SHAP evaluation on plane Couette flow in a minimal flow unit focussing on the relevance of streaks and their instabilities for the prediction of relaminarisation events. Here, we find that the prediction is based on proxies for streak modulations corresponding to linear streak instabilities within the SSP. That is, the SHAP analysis suggests that the break-up of the self-sustaining cycle is connected with a suppression of streak instabilities.

29 pages, 10 figures

Key Takeaways

SHAP provides interpretable explanations for "black box" ML models in fluid dynamics: The paper demonstrates that the Shapley Additive Explanations (SHAP) algorithm can successfully explain why neural networks and other ML models make specific predictions about fluid flow behavior, addressing a major challenge in applying powerful but opaque ML algorithms to scientific domains.
Feature importance rankings from SHAP align with established physical understanding: When tested on wall-bounded shear flows, SHAP identified only three factors as major predictors of relaminarisation events (laminar profile, streamwise vortex, and streak instability)—findings that matched known physics, validating that the algorithm captures genuine physical mechanisms rather than spurious correlations.
Explainable AI methods enable scientific discovery, not just prediction: By revealing which physical features drive ML predictions, SHAP allows researchers to gain new insights into fluid dynamics phenomena rather than simply accepting accurate predictions as "black boxes," demonstrating how interpretability serves the scientific process.
Explainability methods can scale from simple to complex datasets: The paper shows SHAP's effectiveness progresses from controlled low-dimensional models (where ground truth is known) to larger, real-world fluid dynamics datasets, suggesting interpretable AI approaches are practical for increasingly complex scientific applications.
Responsible AI in science requires transparency about model reasoning: This work illustrates why scientists should demand explainability from ML tools—understanding how models reach conclusions is essential for validating results and ensuring ML augments rather than replaces domain expertise.

Abstract

29 pages, 10 figures

Interpreted machine learning in fluid dynamics: Explaining relaminarisation events in wall-bounded shear flows

Key Takeaways

Abstract

Paper Info

Responsible AI Topics

Links

Interpreted machine learning in fluid dynamics: Explaining relaminarisation events in wall-bounded shear flows

Key Takeaways

Abstract

Paper Info

Responsible AI Topics

Links