Abstract:In this paper, an evolutionary algorithm is implemented for the purpose of performing symbolic regression in an attempt to improve Reynolds-Averaged-Navier-Stokes models predictions. In contrast to most machine learning algorithms, Gene Expression Programming generates a mathematical expression that can be directly interpreted and implemented into the Computational Fluid Dynamics solver. In this paper, the latter feature is exploited based on high-fidelity data to ascertain novel correlations for the pressure-… Show more
“…The machine-learning wall-shear-stress model developed in this paper relies on a database of direct numerical simulations or high-fidelity wall-resolved LESs which have been produced by various research groups (Figure 1). Namely, the database includes the simulations of: a fully developed channel flow at friction Reynolds number Re τ ¼ 180 (CF1), performed at Imperial College London (Agostini and Vincent, 2020) using the high-order flux reconstruction method of Huynh (2007) in the CFD code PyFR (Witherden et al, 2014); a fully developed channel flow at friction Reynolds number Re τ ¼ 950 (CF2), performed at the Polytechnic University of Madrid (Del Álamo and Jiménez, 2003;Lozano-Durán and Jiménez, 2014; using a hybrid Fourier-Chebyshev spectral method; a three-dimensional diffuser (3DD) corresponding to the geometry "Diffuser 1" of Cherry et al (2008), performed at Barcelona Supercomputing Center (Ercoftac, 2022) using the low-dissipation finite element scheme of Lehmkuhl et al (2019); a BFS, performed at CERFACS (Pouech et al, 2019;Pouech et al, 2021) using a cell-vertex finite-element method (Schönfeld and Rudgyard, 1999 with second-order accurate convection and diffusion schemes (Lax and Wendroff, 1960); a curved BFS based on the geometry of Disotell and Rumsey (Disotell and Rumsey, 2017;Alaya et al, 2020), hereafter referred to as adverse-pressuregradient (APG) simulation and performed at the University of Bergamo (Ercoftac, 2022) using discontinuous Galerkin method and a fifth order linearly implicit Rosenbrock scheme (Bassi et al, 2015;; and a NACA 65-009 blade cascade (N65) such as studied experimentally by Ma et al (2011) and Zambonini et al (2017), performed using a cell-vertex finite-element method (Schönfeld andRudgyard, 1999 anda two-step Taylor-Galerkin scheme (Colin andRudgyard, 2000). The N65 case is subdivided into two sub-configurations that differ only by the inlet boundary condition: a simulation with an incidence angle of 4°(N65a) and a simulation with an incidence angle of 7°(N65b).…”
As the Reynolds number increases, the large-eddy simulation (LES) of complex flows becomes increasingly intractable because near-wall turbulent structures become increasingly small. Wall modeling reduces the computational requirements of LES by enabling the use of coarser cells at the walls. This paper presents a machine-learning methodology to develop data-driven wall-shear-stress models that can directly operate, a posteriori, on the unstructured grid of the simulation. The model architecture is based on graph neural networks. The model is trained on a database which includes fully developed boundary layers, adverse pressure gradients, separated boundary layers, and laminar–turbulent transition. The relevance of the trained model is verified a posteriori for the simulation of a channel flow, a backward-facing step and a linear blade cascade.
“…The machine-learning wall-shear-stress model developed in this paper relies on a database of direct numerical simulations or high-fidelity wall-resolved LESs which have been produced by various research groups (Figure 1). Namely, the database includes the simulations of: a fully developed channel flow at friction Reynolds number Re τ ¼ 180 (CF1), performed at Imperial College London (Agostini and Vincent, 2020) using the high-order flux reconstruction method of Huynh (2007) in the CFD code PyFR (Witherden et al, 2014); a fully developed channel flow at friction Reynolds number Re τ ¼ 950 (CF2), performed at the Polytechnic University of Madrid (Del Álamo and Jiménez, 2003;Lozano-Durán and Jiménez, 2014; using a hybrid Fourier-Chebyshev spectral method; a three-dimensional diffuser (3DD) corresponding to the geometry "Diffuser 1" of Cherry et al (2008), performed at Barcelona Supercomputing Center (Ercoftac, 2022) using the low-dissipation finite element scheme of Lehmkuhl et al (2019); a BFS, performed at CERFACS (Pouech et al, 2019;Pouech et al, 2021) using a cell-vertex finite-element method (Schönfeld and Rudgyard, 1999 with second-order accurate convection and diffusion schemes (Lax and Wendroff, 1960); a curved BFS based on the geometry of Disotell and Rumsey (Disotell and Rumsey, 2017;Alaya et al, 2020), hereafter referred to as adverse-pressuregradient (APG) simulation and performed at the University of Bergamo (Ercoftac, 2022) using discontinuous Galerkin method and a fifth order linearly implicit Rosenbrock scheme (Bassi et al, 2015;; and a NACA 65-009 blade cascade (N65) such as studied experimentally by Ma et al (2011) and Zambonini et al (2017), performed using a cell-vertex finite-element method (Schönfeld andRudgyard, 1999 anda two-step Taylor-Galerkin scheme (Colin andRudgyard, 2000). The N65 case is subdivided into two sub-configurations that differ only by the inlet boundary condition: a simulation with an incidence angle of 4°(N65a) and a simulation with an incidence angle of 7°(N65b).…”
As the Reynolds number increases, the large-eddy simulation (LES) of complex flows becomes increasingly intractable because near-wall turbulent structures become increasingly small. Wall modeling reduces the computational requirements of LES by enabling the use of coarser cells at the walls. This paper presents a machine-learning methodology to develop data-driven wall-shear-stress models that can directly operate, a posteriori, on the unstructured grid of the simulation. The model architecture is based on graph neural networks. The model is trained on a database which includes fully developed boundary layers, adverse pressure gradients, separated boundary layers, and laminar–turbulent transition. The relevance of the trained model is verified a posteriori for the simulation of a channel flow, a backward-facing step and a linear blade cascade.
The accuracy of machine-learned turbulence models often diminishes when applied to flow cases outside the training data set. In an effort to improve the predictive accuracy of data-driven models for an expanded set of cases, an extension of a computational fluid dynamics (CFD)-driven training framework consisting of three key steps is proposed. Firstly, a list of candidate flow-related parameters is selected to supplement Pope’s general tensor basis hypothesis. Secondly, modeling an additional production term may benefit the overall predictions in certain situations. Finally, the Reynolds-averaged Navier–Stokes (RANS) evaluations of candidate models are performed on several different flows simultaneously during the model training iterations. Five free-shear and five wall-bounded flow cases are chosen to train or test data-driven turbulence models. It is shown that the machine-learned models from the present multicase CFD-driven framework can significantly improve the predictive accuracy for the test cases where the baseline RANS results showed significant error from the ground truth. Meanwhile, for cases in which the baseline produced good results, the new models do not perform worse. Further analysis shows that the new models can adapt to opposite trends of turbulent diffusion required for the different cases with a common correction. Moreover, the trained models can be simplified and still achieve similar improvement as the whole expressions.
This paper discusses the generalizability of a data-augmented turbulence model with a focus on the field inversion and machine learning approach. It is highlighted that the augmented model based on two-dimensional (2D) separated airfoil flows gives poor predictive capability for a different class of separated flows (NASA wall-mounted hump) compared to the baseline model due to extrapolation. We demonstrate a sensor-based approach to localize the data-driven model correction to tackle this generalizability issue. Furthermore, the applicability of the augmented model to a more complex aeronautical three-dimensional case, the NASA Common Research Model configuration, is studied. Observations on the pressure coefficient predictions and the model correction field suggest that the present 2D-based augmentation is to some extent applicable to a three-dimensional aircraft flow.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.