Podcast Beta
Questions and Answers
What are the main points of concern regarding data-driven turbulence models?
Generalisability and the consistency of the a posteriori results.
What computational approach is used for the progressive improvement of turbulence models in this study?
Simulation-driven Bayesian optimisation with Kriging surrogates.
What is the objective of augmenting secondary-flow prediction capability in the linear eddy-viscosity model?
To enhance prediction without violating the original performance on canonical cases.
Match the turbulence modeling approaches with their definitions:
Signup and view all the answers
What type of models are obtained through progressive data-augmented explicit algebraic Reynolds stress models?
Signup and view all the answers
What performance aspect is preserved when applying new models on channel flow cases?
Signup and view all the answers
The study reports that there is a significant improvement in the prediction of secondary flows and streamwise velocity with the new models.
Signup and view all the answers
What optimization method is presented in the study for turbulence model improvement?
Signup and view all the answers
What is the goal of augmenting the secondary-flow prediction capability?
Signup and view all the answers
The study focuses primarily on direct numerical simulation (DNS) methods for turbulence modeling.
Signup and view all the answers
What are progressively data-augmented explicit algebraic Reynolds stress models (PDA-EARSMs) used for?
Signup and view all the answers
What findings were observed with the new turbulence models on unseen test cases?
Signup and view all the answers
Study Notes
RANS
- Reynolds-averaged Navier–Stokes (RANS) equations are preferred over more computationally expensive methods like DNS and LES for industrial applications of CFD.
- RANS relies on Reynolds stress tensor (RST) models to predict turbulence physics.
Data-Driven Turbulence Models
- Generalisability and consistency of the a posteriori results are critical concerns in data-driven turbulence models.
- This study progressively improves turbulence models using a simulation-driven Bayesian optimisation with Kriging surrogates.
- The optimisation of the models is achieved through a multi-objective approach based on duct flow quantities.
PDA-EARSMs for 𝑘 − 𝜔 SST
- This research aims to augment the secondary-flow prediction capability of the linear eddy-viscosity model 𝑘 − 𝜔 SST without compromising its performance on canonical cases like channel flow.
- The study develops progressively data-augmented explicit algebraic Reynolds stress models (PDA-EARSMs) for the 𝑘 − 𝜔 SST model.
- These PDA-EARSMs can predict secondary flows that the standard 𝑘 − 𝜔 SST model fails to predict while maintaining the original model's performance on channel flow.
Results
- Numerical verification for unseen test cases demonstrates a significant improvement by PDA-EARSMs in predicting secondary flows and streamwise velocity.
- The progressive approach enhances the performance of data-driven turbulence models for fluid flow simulation while preserving robustness and stability of the solver.
Prior Data-Driven Approaches
- Existing research on the improvement of RANS turbulence models is focused on training models using high-fidelity RST data to improve predictions from empirical models.
- Other studies aim to correct RST predictions or to modify existing empirical models.
Introduction
- Reynolds-averaged Navier–Stokes (RANS) equations are preferred over DNS and LES for industrial CFD applications because of their robustness and computational speed.
- RANS predicts turbulence physics through a Reynolds stress tensor (RST) model, influencing simulation results.
- Despite the prevalence of RANS simulations, common empirical models have limitations.
Data-Driven Turbulence Models
- Generalisability and consistency of results are crucial considerations for data-driven turbulence models.
- This study enhances turbulence models using simulation-driven Bayesian optimisation with Kriging surrogates.
- The optimisation is multi-objective, targeting duct flow quantities.
Aim of the Study
- To augment the secondary-flow prediction capability of the 𝑘 − 𝜔 SST linear eddy-viscosity model without compromising its performance on canonical cases like channel flow.
Methodology
- Progressively data-augmented explicit algebraic Reynolds stress models (PDA-EARSMs) for 𝑘 − 𝜔 SST are developed, enabling prediction of secondary flows missed by the standard model.
- New models are tested on channel flow cases to ensure they retain the performance of the original 𝑘 − 𝜔 SST model.
Findings
- Numerical verification is carried out on various test cases.
- Unseen test cases demonstrate significant improvement in secondary flow and streamwise velocity predictions, highlighting the generalisability of the new models.
Conclusion
- The study promotes the potential of a progressive approach for enhancing data-driven turbulence models while preserving solver robustness and stability.
Studying That Suits You
Use AI to generate personalized quizzes and flashcards to suit your learning preferences.
Related Documents
Description
Explore the fundamentals of turbulence modeling in Computational Fluid Dynamics (CFD), focusing on RANS equations, data-driven approaches, and the PDA-EARSMs for the k-ω SST model. This quiz covers key concepts and advancements in turbulence prediction methods, particularly for industrial applications.
