Sparse Bayesian Learning of Explicit Algebraic Reynolds-Stress models for turbulent separated flows

Soufiane, Cherroud,; Merle, Xavier; Cinnella, Paola; Gloerfelt, Xavier

doi:10.1016/j.ijheatfluidflow.2022.109047

Cited by 13 publications

(5 citation statements)

References 56 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…3. One specific turbulence closure methodology that has received much attention is the algebraic Reynolds stress model (ARSM), where data-driven techniques are used to improve coefficient determination in the tensor representation of constitutive relations [50][51][52][53][54][55][56]. 4.…”

Section: Machine-learning and Ransmentioning

confidence: 99%

Turbulence closure modeling with machine learning: a foundational physics perspective

Girimaji

2024

New J. Phys.

View full text Add to dashboard Cite

Turbulence closure modeling using machine learning is at an early crossroads. The extraordinary success of machine learning (ML) in a variety of challenging fields had given rise to an expectation of similar transformative advances in the area of turbulence closure modeling. However, by most accounts, the current rate of progress toward accurate and predictive ML-RANS (Reynolds Averaged Navier-Stokes) closure models has been very slow. Upon retrospection, the absence of rapid transformative progress can be attributed to two factors: theunderestimation of the intricacies of turbulence modeling and the overestimation of ML’s ability to capture all features without employing targeted strategies. To pave the way for moremeaningful ML closures tailored to address the nuances of turbulence, this article seeks toreview the foundational flow physics to assess the challenges in the context of data-drivenapproaches. Revisiting analogies with statistical mechanics and stochastic systems, the keyphysical complexities and mathematical limitations are explicated. It is noted that the currentML approaches do not systematically address the inherent limitations of a statistical approachor the inadequacies of the mathematical forms of closure expressions. The study underscoresthe drawbacks of supervised learning-based closures and stresses the importance of a morediscerning ML modeling framework. As ML methods evolve (which is happening at a rapidpace) and our understanding of the turbulence phenomenon improves, the inferences expressedhere should be suitably modified.

show abstract

Section: Machine-learning and Ransmentioning

confidence: 99%

Turbulence closure modeling with machine learning: a foundational physics perspective

Girimaji

2024

New J. Phys.

View full text Add to dashboard Cite

show abstract

“…These models, formulated as tensor polynomials, were inferred from high-fidelity data and allowed for a cost-effective correction of the k-ω SST model. Additionally, the works of Huijing et al [42], Zhang et al [43], Cherroud et al [44] have been instrumental in the further promotion and enhancement of SpaRTA.…”

Section: Introductionmentioning

confidence: 99%

Data-driven RANS closures for improving mean field calculation of separated flows

Chen,

Deng

2024

Front. Phys.

View full text Add to dashboard Cite

Reynolds-averaged Navier-Stokes (RANS) simulations have found widespread use in engineering applications, yet their accuracy is compromised, especially in complex flows, due to imprecise closure term estimations. Machine learning advancements have opened new avenues for turbulence modeling by extracting features from high-fidelity data to correct RANS closure terms. This method entails establishing a mapping relationship between the mean flow field and the closure term through a designated algorithm. In this study, the k-ω SST model serves as the correction template. Leveraging a neural network algorithm, we enhance the predictive precision in separated flows by forecasting the desired learning target. We formulate linear terms by approximating the high-fidelity closure (from Direct Numerical Simulation) based on the Boussinesq assumption, while residual errors (referred to as nonlinear terms) are introduced into the momentum equation via an appropriate scaling factor. Utilizing data from periodic hills flows encompassing diverse geometries, we train two neural networks, each possessing comparable structures, to predict the linear and nonlinear terms. These networks incorporate features from the minimal integrity basis and mean flow. Through generalization performance tests, the proposed data-driven model demonstrates effective closure term predictions, mitigating significant overfitting concerns. Furthermore, the propagation of the predicted closure term to the mean velocity field exhibits remarkable alignment with the high-fidelity data, thus affirming the validity of the current framework. In contrast to prior studies, we notably trim down the total count of input features to 12, thereby simplifying the task for neural networks and broadening its applications to more intricate scenarios involving separated flows.

show abstract

“…These results suggest that the use of wall models cannot predict the impending boundary layer detachment when using the k − ω SST model. Simultaneously, k − ω SST is known to overestimate boundary layer separation and re-attachment by the literature [86,87]. Future work suggests performing high-fidelity numerical simulations and/or experimentally investigating these regions to validate the results and which methodology is more accurate.…”

Section: V3 Resultsmentioning

confidence: 95%

“…There are, however, slight deviations in the mean value of the numerical prediction seen at x = 1D. These deviations suggest a slight underprediction of the axial velocity at the centre and top sections of the pipe and could indicate that the CFD methodology underestimates the turbulence diffusion in some regions of the flow as well as boundary-layer detachment and reattachment [29,86,87]. Nevertheless, all numerical results are confined inside experimental uncertainty bounds of 2σ in the analysed profiles, where the slight disagreement of high Re cases is expected due to their added physical complexity and the prediction of turbulent fields.…”

Section: Iii32 Experimental Validation Of Axial Velocitymentioning

confidence: 99%

See 1 more Smart Citation

Robust Optimisation of Ultrasonic Flow Meters by Computational Fluid Dynamics and Enhanced Turbulence Modelling

Rincón

View full text Add to dashboard Cite

The accurate measurement of fluid flow is essential in a wide range of industrial applications. In this regard, ultrasonic flow meters present an accurate and cost-effective solution; however, these systems introduce challenges due to their complex geometry and interaction with fluid flow. Hence, understanding their flow dynamics is paramount in the improvement of their design and operation. Simultaneously, data-driven turbulence modelling techniques have emerged as promising tools to improve the precision of fluid flow simulations. This doctoral thesis focuses on bridging these two domains by predicting complex flows in ultrasonic flow meters, developing a robust methodology for geometrical optimisation, and enhancing existing Reynolds-averaged Navier-Stokes (RANS) turbulence models with progressive and generalisable data-driven methods. A methodology based on computational fluid dynamics (CFD) is developed to predict the flow dynamics throughout ultrasonic flow meters by the use of RANS k −ω SST. The method is experimentally validated with laser Doppler velocimetry and pressure drop experiments for the whole dynamic range of operation of these systems. Subsequently, a series of flow simulations are driven by surrogate and Bayesian multi-objective optimisation methods to provide improved geometrical designs. These improved geometries are further optimised by an adjoint-based shape optimisation method that enables the 3-dimensional morphing of the geometry, allowing higher improvement gains which cannot be achieved by traditional engineering ingenuity. In the realm of data-driven turbulence modelling, two critical aspects take centre stage: generalisability and the consistency of a posteriori results. This thesis addresses these challenges through a systematic approach that combines the CFD-driven optimisation techniques developed for flow meter improvement with a progressive augmentation strategy in the k −ω SST turbulence model. The goal is to improve the prediction of turbulence anisotropybased secondary flows, and flow separation—two common stumbling blocks in fluid flow simulations — without altering the successful law of the wall prediction by the original model. Two correction terms are carefully defined and introduced in the momentum and specific dissipation rate (ω) transport equations. After extensive numerical verification, the enhanced models exhibit significant improvements in the prediction of secondary flows, boundary-layer detachment and reattachment, and friction coefficients in both training and testing cases, reaffirming their a posteriori reliability and generalisability. In conclusion, this doctoral thesis presents a robust and cost-efficient methodology to systematically improve the performance of complex engineering systems based on CFD simulations. Additionally, two generalisable methods are developed to enhance turbulence modelling based on data-driven progressive approaches. By addressing challenges related to flow separation and secondary flow predictions in complex geometries together with design and shape optimisation techniques, this research contributes to more accurate fluid flow predictions and simulation-based optimisation; benefiting industries dependent on precise flow measurements and simulations. The synergy between data-driven modelling and complex system optimisation demonstrates the potential for innovative solutions in both turbulence modelling and flow measurement technologies, with broad implications across academic and industrial engineering sectors.

show abstract

Sparse Bayesian Learning of Explicit Algebraic Reynolds-Stress models for turbulent separated flows

Cited by 13 publications

References 56 publications

Turbulence closure modeling with machine learning: a foundational physics perspective

Turbulence closure modeling with machine learning: a foundational physics perspective

Data-driven RANS closures for improving mean field calculation of separated flows

Robust Optimisation of Ultrasonic Flow Meters by Computational Fluid Dynamics and Enhanced Turbulence Modelling

Contact Info

Product

Resources

About