Papers
Machine learning for physics, as it is published.
New work from arXiv on CFD surrogates, neural operators and physics-informed learning, refreshed every hour. Search it, filter by topic, or start with the papers we keep coming back to.
Papers
New work from arXiv on CFD surrogates, neural operators and physics-informed learning, refreshed every hour. Search it, filter by topic, or start with the papers we keep coming back to.
564 papers in Turbulence · newest first
physics.flu-dyn
Efficient tools for predicting the drag of rough walls in turbulent flows would have a tremendous impact. However, methods for drag prediction rely on experiments or numerical simulations which are costly and time-consuming. Data-driven regression methods have the potential to provide a prediction that is accurate and fast. We assess the performance and limitations of linear regression, kernel methods and neural networks for drag prediction using a database of 1000 homogeneous rough surfaces. Model performance is evaluated using the roughness function obtained at friction-scaled Reynolds number 500. With two trainable parameters, the kernel method can fully account for nonlinear relations between and surface statistics (roughness height, effective slope, skewness, etc). In contrast, linear regression cannot account for nonlinear correlations and display large errors and high uncertainty. Multilayer perceptron and convolutional neural networks demonstrate performance on par with the kernel method but have orders of magnitude more trainable parameters. For the current database size, the networks' capacity cannot be fully exploited, resulting in reduced generalizability and reliability. Our study provides insight into the appropriateness of different regression models for drag prediction. We also discuss the remaining steps before data-driven methods emerge as useful tools in applications.
physics.flu-dyn
Machine learning techniques have been applied to enhance turbulence modeling in recent years. However, the "black box" nature of most machine learning techniques poses significant interpretability challenges in improving turbulence models. This paper introduces a novel unit-constrained turbulence modeling framework using symbolic regression to overcome these challenges. The framework amends the constitutive equation of linear eddy viscosity models (LEVMs) by establishing explicit equations between the Reynolds stress deviation and mean flow quantities, thereby improving the LEVM model's predictive capability for large separated turbulence. Unit consistency constraints are applied to the symbolic expressions to ensure physical realizability. The effectiveness of the framework and the generalization capability of the learned model are demonstrated through its application to the separated flow over 2D periodic hills and a backward-facing step. Compared to the standard k-ε model, the learned model shows significantly improved predictive accuracy for anisotropic Reynolds stresses, velocity and skin friction, while exhibiting promising generalization capabilities across various scenarios.
physics.flu-dyn
Simulations of complex turbulent flow are part and parcel of the engineering design process. Eddy viscosity based turbulence models represent the workhorse for these simulations. The underlying simplifications in eddy viscosity models make them computationally inexpensive but also introduce structural uncertainties in their predictions. Currently the Eigenspace Perturbation Method is the only approach to predict these uncertainties. Due to its purely physics based nature this method often leads to unrealistically large uncertainty bounds that lead to exceedingly conservative designs. We use a Deep Learning based approach to address this issue. We control the perturbations using trained deep learning models that predict how much to perturb the modeled Reynolds stresses. This is executed using a Convolutional Neural Network that learns the difference between eddy viscosity based model predictions and high fidelity data as a mapping of flow features. We show that this approach leads to improvements over the Eigenspace Perturbation Method.
cs.LG
Several related works have introduced Koopman-based Machine Learning architectures as a surrogate model for dynamical systems. These architectures aim to learn non-linear measurements (also known as observables) of the system's state that evolve by a linear operator and are, therefore, amenable to model-based linear control techniques. So far, mainly simple systems have been targeted, and Koopman architectures as reduced-order models for more complex dynamics have not been fully explored. Hence, we use a Koopman-inspired architecture called the Linear Recurrent Autoencoder Network (LRAN) for learning reduced-order dynamics in convection flows of a Rayleigh Bénard Convection (RBC) system at different amounts of turbulence. The data is obtained from direct numerical simulations of the RBC system. A traditional fluid dynamics method, the Kernel Dynamic Mode Decomposition (KDMD), is used to compare the LRAN. For both methods, we performed hyperparameter sweeps to identify optimal settings. We used a Normalized Sum of Square Error measure for the quantitative evaluation of the models, and we also studied the model predictions qualitatively. We obtained more accurate predictions with the LRAN than with KDMD in the most turbulent setting. We conjecture that this is due to the LRAN's flexibility in learning complicated observables from data, thereby serving as a viable surrogate model for the main structure of fluid dynamics in turbulent convection settings. In contrast, KDMD was more effective in lower turbulence settings due to the repetitiveness of the convection flow. The feasibility of Koopman-based surrogate models for turbulent fluid flows opens possibilities for efficient model-based control techniques useful in a variety of industrial settings.
physics.flu-dyn
We develop time-series machine learning (ML) methods for closure modeling of the Unsteady Reynolds Averaged Navier Stokes (URANS) equations applied to stably stratified turbulence (SST). SST is strongly affected by fine balances between forces and becomes more anisotropic in time for decaying cases. Moreover, there is a limited understanding of the physical phenomena described by some of the terms in the URANS equations. Rather than attempting to model each term separately, it is attractive to explore the capability of machine learning to model groups of terms, i.e., to directly model the force balances. We consider decaying SST which are homogeneous and stably stratified by a uniform density gradient, enabling dimensionality reduction. We consider two time-series ML models: Long Short-Term Memory (LSTM) and Neural Ordinary Differential Equation (NODE). Both models perform accurately and are numerically stable in a posteriori tests. Furthermore, we explore the data requirements of the ML models by extracting physically relevant timescales of the complex system. We find that the ratio of the timescales of the minimum information required by the ML models to accurately capture the dynamics of the SST corresponds to the Reynolds number of the flow. The current framework provides the backbone to explore the capability of such models to capture the dynamics of higher-dimensional complex SST flows.
physics.flu-dyn
Neural-network models have been employed to predict the instantaneous flow close to the wall in a viscoelastic turbulent channel flow. Numerical simulation data at the wall is utilized to predict the instantaneous velocity-fluctuations and polymeric-stress-fluctuations at three different wall-normal positions in the buffer region. The ability of non-intrusive predictions has not been previously investigated in non-Newtonian turbulence. Our analysis shows that velocity-fluctuations are predicted well from wall measurements in viscoelastic turbulence. The models exhibit enhanced accuracy in predicting quantities of interest during the hibernation intervals, facilitating a deeper understanding of the underlying physics during low-drag events. The neural-network models also demonstrate a reasonably good accuracy in predicting polymeric-shear stress and the trace of the polymer stress at a given wall-normal location. This method could be used in flow control or when only wall information is available from experiments (for example, in opaque fluids). More importantly, only velocity and pressure information can be measured experimentally, while polymeric elongation and orientation cannot be directly measured despite their importance for turbulent dynamics. We therefore study the possibility to reconstruct the polymeric-stress fields from velocity or pressure measurements in viscoelastic turbulent flows. The results are promising but also underline that a lack of small scales in the input velocity fields can alter the rate of energy transfer from flow to polymers, affecting the prediction of the polymer-stress fluctuations. The present approach not only aids in extracting polymeric-stress information but also gives information about the link between polymeric-stress and velocity fields in viscoelastic turbulence.
physics.flu-dyn
This chapter provides an introduction to data-driven techniques for the development and calibration of closure models for the Reynolds-Averaged Navier–Stokes (RANS) equations. RANS models are the workhorse for engineering applications of computational fluid dynamics (CFD) and are expected to play an important role for decades to come. However, RANS model inadequacies for complex, non-equilibrium flows and uncertainties in modeling assumptions and calibration data are still a major obstacle to the predictive capability of RANS simulations. In the following, we briefly recall the origin and limitations of RANS models, and then review their shortcomings and uncertainties. Then, we provide an introduction to data-driven approaches to RANS turbulence modeling. The latter can range from simple model parameter inference to sophisticated machine learning techniques. We conclude with some perspectives on current and future research trends.
physics.flu-dyn
The transformative impact of machine learning, particularly Deep Learning (DL), on scientific and engineering domains is evident. In the context of computational fluid dynamics (CFD), Physics-Informed Neural Networks (PINNs) represent a significant innovation, enabling data-driven fluid simulations while incorporating physics-based laws described by partial differential equations (PDEs). While PINNs have demonstrated efficacy in various fluid flow scenarios, a noticeable gap exists in their application to simulate jet flows - an essential category in engineering. Jets, crucial for downburst outflow, ventilation, and heat transfer, lack comprehensive exploration through PINNs in existing literature. This study addresses this gap by focusing on the application of PINNs to simulate steady jet flows, specifically 2D planar turbulent jet flow scenarios. The novelty lies not only in adapting PINNs for simulating jet flows but also in overcoming challenges such as poor convergence during training, attributed to imbalances in loss terms. We propose a novel PINN architecture for Reynolds-Averaged Navier-Stokes (RANS) simulation of steady turbulent jet flows, without the need for turbulence models and simulation data. Additionally, an extended dynamic weighting strategy is introduced to enhance the balance of loss term contributions in the PINN, resulting in improved convergence and more accurate predictions.
physics.flu-dyn
A wavelet-based machine learning method is proposed for predicting the time evolution of homogeneous isotropic turbulence where vortex tubes are preserved. Three-dimensional convolutional neural networks and long short-term memory are trained with a time series of direct numerical simulation (DNS) data of homogeneous isotropic turbulence at the Taylor microscale Reynolds number 92. The predicted results are assessed by using flow visualization of vorticity and statistics, e.g., probability density functions of vorticity and enstrophy spectra. It is found that the predicted results are in good agreement with DNS results. The small-scale flow topology considering the second and third invariant of the velocity gradient tensor likewise shows an approximate match. Furthermore, we apply the pre-trained neural networks to coarse-grained vorticity data using superresolution. It is shown that the superresolved flow field well agrees with the reference DNS field and thus small-scale information and vortex tubes are well regenerated.
physics.flu-dyn
Obtaining system parameters and reconstructing the full flow state from limited velocity observations using conventional fluid dynamics solvers can be prohibitively expensive. Here we employ machine learning algorithms to overcome the challenge. As an example, we consider a moderately turbulent fluid flow, excited by a stationary force and described by a two-dimensional Navier-Stokes equation with linear bottom friction. Using dense in time, spatially sparse and probably noisy velocity data, we reconstruct the spatially dense velocity field, infer the pressure and driving force up to a harmonic function and its gradient, respectively, and determine the unknown fluid viscosity and friction coefficient. Both the root-mean-square errors of the reconstructions and their energy spectra are addressed. We study the dependence of these metrics on the degree of sparsity and noise in the velocity measurements. Our approach involves training a physics-informed neural network by minimizing the loss function, which penalizes deviations from the provided data and violations of the governing equations. The suggested technique extracts additional information from velocity measurements, potentially enhancing the capabilities of particle image/tracking velocimetry.
physics.flu-dyn
In this work, we consider the problem of inferring the initial conditions of a Rayleigh-Taylor mixing zone by measuring the 0D turbulent quantities at an unspecified time. To this aim, we have generated a comprehensive dataset through direct numerical simulations (DNS), focusing on miscible fluids with slight density contrasts. The initial interface deformations in these simulations are characterized by an annular spectrum which is parametrized by four non dimensional numbers. %In order to study the sensitivity of 0D turbulent quantities to the initial interface perturbation distributions, we build a surrogate model for the simulations using a physics-informed neural network (PINN). This allows us to compute the Sobol indices for the turbulent quantities, disentangling the effects of the initial parameters on the growth of the mixing layer. Within a Bayesian framework, we use a Markov chain Monte-Carlo method to determine the posterior distributions of initial conditions given various state variables. %This sheds light on the inertial or diffusive trajectories along with how the initial conditions are progressively forgotten during transition to turbulence. Moreover, it identifies which turbulent quantities are better predictors for the dynamics of Rayleigh-Taylor mixing zones by more effectively retaining the memory of the flow. By inferring the initial conditions and forward propagating its maximum a posteriori (MAP) estimate, we propose a strategy to model the Rayleigh-Taylor transition to turbulence.
cs.LG
The paper reflects on the future role of AI in scientific research, with a special focus on turbulence studies, and examines the evolution of AI, particularly through Diffusion Models rooted in non-equilibrium statistical mechanics. It underscores the significant impact of AI on advancing reduced, Lagrangian models of turbulence through innovative use of deep neural networks. Additionally, the paper reviews various other AI applications in turbulence research and outlines potential challenges and opportunities in the concurrent advancement of AI and statistical hydrodynamics. This discussion sets the stage for a future where AI and turbulence research are intricately intertwined, leading to more profound insights and advancements in both fields.
physics.flu-dyn
Predicting the large-scale dynamics of three-dimensional (3D) turbulence is challenging for machine learning approaches. This paper introduces a transformer-based neural operator (TNO) to achieve precise and efficient predictions in the large-eddy simulation (LES) of 3D turbulence. The performance of the proposed TNO model is systematically tested and compared with LES using classical sub-grid scale (SGS) models, including the dynamic Smagorinsky model (DSM) and the dynamic mixed model (DMM), as well as the original Fourier neural operator (FNO) model, in homogeneous isotropic turbulence (HIT) and free-shear turbulent mixing layer. The numerical simulations comprehensively evaluate the performance of these models on a variety of flow statistics, including the velocity spectrum, the probability density functions (PDFs) of vorticity, the PDFs of velocity increments, the evolution of turbulent kinetic energy, and the iso-surface of the Q-criterion. The results indicate that the accuracy of the TNO model is comparable to the LES with DSM model, and outperforms the FNO model and LES using DMM in HIT. In the free-shear turbulence, the TNO model exhibits superior accuracy compared to other models. Moreover, the TNO model has fewer parameters than the FNO model and enables long-term stable predictions, which the FNO model cannot achieve. The well-trained TNO model is significantly faster than traditional LES with DSM and DMM models, and can be generalized to higher Taylor-Reynolds number cases, indicating its strong potential for 3D nonlinear engineering applications.
physics.flu-dyn
This study introduces the Conditional Neural Field Latent Diffusion (CoNFiLD) model, a novel generative learning framework designed for rapid simulation of intricate spatiotemporal dynamics in chaotic and turbulent systems within three-dimensional irregular domains. Traditional eddy-resolved numerical simulations, despite offering detailed flow predictions, encounter significant limitations due to their extensive computational demands, restricting their applications in broader engineering contexts. In contrast, deep learning-based surrogate models promise efficient, data-driven solutions. However, their effectiveness is often compromised by a reliance on deterministic frameworks, which fall short in accurately capturing the chaotic and stochastic nature of turbulence. The CoNFiLD model addresses these challenges by synergistically integrating conditional neural field encoding with latent diffusion processes, enabling the memory-efficient and robust probabilistic generation of spatiotemporal turbulence under varied conditions. Leveraging Bayesian conditional sampling, the model can seamlessly adapt to a diverse range of turbulence generation scenarios without the necessity for retraining, covering applications from zero-shot full-field flow reconstruction using sparse sensor measurements to super-resolution generation and spatiotemporal flow data restoration. Comprehensive numerical experiments across a variety of inhomogeneous, anisotropic turbulent flows with irregular geometries have been conducted to evaluate the model's versatility and efficacy, showcasing its transformative potential in the domain of turbulence generation and the broader modeling of spatiotemporal dynamics.
physics.flu-dyn
A modelling framework based on the resolvent analysis and machine learning is proposed to predict the turbulent energy in incompressible channel flows. In the framework, the optimal resolvent response modes are selected as the basis functions modelling the low-rank behaviour of high-dimensional nonlinear turbulent flow-fields, and the corresponding weight functions are determined by data-driven neural networks. Turbulent-energy distribution in space and scales, at the friction Reynolds number 1000, is predicted and compared to the data of direct numerical simulation. Close agreement is observed, suggesting the feasibility and reliability of the proposed framework for turbulence prediction.
physics.flu-dyn
Small-scale turbulence can be comprehensively described in terms of velocity gradients, which makes them an appealing starting point for low-dimensional modeling. Typical models consist of stochastic equations based on closures for non-local pressure and viscous contributions. The fidelity of the resulting models depends on the accuracy of the underlying modeling assumptions. Here, we discuss an alternative data-driven approach leveraging machine learning to derive a velocity gradient model which captures its statistics by construction. We use a normalizing flow to learn the velocity gradient probability density function (PDF) from direct numerical simulation (DNS) of incompressible turbulence. Then, by using the equation for the single-time PDF of the velocity gradient, we construct a deterministic, yet chaotic, dynamical system featuring the learned steady-state PDF by design. Finally, utilizing gauge terms for the velocity gradient single-time statistics, we optimize the time correlations as obtained from our model against the DNS data. As a result, the model time realizations statistically closely resemble the time series from DNS.
physics.flu-dyn
Within the scope of reacting flow simulations, the real-time direct integration (DI) of stiff ordinary differential equations (ODE) for the computation of chemical kinetics stands as the primary demand on computational resources. Meanwhile, as the number of transport equations that need to be solved increases, the computational cost grows more substantially, particularly for those combustion models involving direct coupling of chemistry and flow such as the transported probability density function model. In the current study, an integrated Graphics Processing Unit-Artificial Neural Network (GPU-ANN) framework is introduced to comply with heavy computational costs while maintaining high fidelity. Within this framework, a GPU-based solver is employed to solve partial differential equations and compute thermal and transport properties, and an ANN is utilized to replace the calculation of reaction rates. Large eddy simulations of two swirling flames provide a robust validation, affirming and extending the GPU-ANN approach's applicability to challenging scenarios. The simulation results demonstrate a strong correlation in the macro flame structure and statistical characteristics between the GPU-ANN approach and the traditional Central Processing Unit (CPU)-based solver with DI. This comparison indicates that the GPU-ANN approach is capable of attaining the same degree of precision as the conventional CPU-DI solver, even in more complex scenarios. In addition, the overall speed-up factor for the GPU-ANN approach is over two orders of magnitude. This study establishes the potential groundwork for widespread application of the proposed GPU-ANN approach in combustion simulations, addressing various and complex scenarios based on detailed chemistry, while significantly reducing computational costs.
physics.flu-dyn
While many physics-based closure model forms have been posited for the sub-filter scale (SFS) in large eddy simulation (LES), vast amounts of data available from direct numerical simulation (DNS) create opportunities to leverage data-driven modeling techniques. Albeit flexible, data-driven models still depend on the dataset and the functional form of the model chosen. Increased adoption of such models requires reliable uncertainty estimates both in the data-informed and out-of-distribution regimes. In this work, we employ Bayesian neural networks (BNNs) to capture both epistemic and aleatoric uncertainties in a reacting flow model. In particular, we model the filtered progress variable scalar dissipation rate which plays a key role in the dynamics of turbulent premixed flames. We demonstrate that BNN models can provide unique insights about the structure of uncertainty of the data-driven closure models. We also propose a method for the incorporation of out-of-distribution information in a BNN. The efficacy of the model is demonstrated by a priori evaluation on a dataset consisting of a variety of flame conditions and fuels.
physics.flu-dyn
Nonlinear machine learning for turbulent flows can exhibit robust performance even outside the range of training data. This is achieved when machine-learning models can accommodate scale-invariant characteristics of turbulent flow structures. This study presents a data-driven approach to reveal scale-invariant vortical structures across Reynolds numbers that provide insights for supporting nonlinear machine-learning-based studies of turbulent flows. To uncover conditions for which nonlinear models are likely to perform well, we use a Buckingham Pi-based sparse nonlinear scaling to find the influence of the Pi groups on the turbulent flow data. We consider nonlinear scalings of the invariants of the velocity gradient tensor for an example of three-dimensional decaying isotropic turbulence. The present scaling not only enables the identification of vortical structures that are interpolatory and extrapolatory for the given flow field data but also captures non-equilibrium effects of the energy cascade. As a demonstration, the present findings are applied to machine-learning-based super-resolution analysis of three-dimensional isotropic turbulence. We show that machine-learning models reconstruct vortical structures well in the interpolatory space with reduced performance in the extrapolatory space revealed by the nonlinearly scaled invariants. The present approach enables us to depart from labeling turbulent flow data with a single parameter of Reynolds number and comprehensively examine the flow field to support training and testing of nonlinear machine-learning techniques.
physics.flu-dyn
This paper addresses the issue of predicting separated flows with Reynolds-averaged Navier-Stokes (RANS) turbulence models, which are essential for many engineering tasks. Traditional RANS models usually struggle with this task, so recent efforts have focused on data-driven methods such as field inversion and machine learning (FIML) to correct this issue by adjusting the baseline equations. However, these FIML methods often reduce accuracy in attached boundary layers. To address this issue, we developed a "conditioned field inversion" technique. This method adjusts the corrective factor \b{eta} (used by FIML) in the shear-stress transport (SST) model. It multiplies \b{eta} with a shield function f_d that is off in the boundary layer and on elsewhere. This maintains the accuracy of the baseline model for the attached flows. We applied both conditioned and classic field inversion to the NASA hump and a curved backward-facing step (CBFS), creating two datasets. These datasets were used to train two models: SR-CND (from our new method) and SR-CLS (from the traditional method). The SR-CND model matches the SR-CLS model in predicting separated flows in various scenarios, such as periodic hills, the NLR7301 airfoil, the 3D SAE car model, and the 3D Ahmed body, and outperforms the baseline SST model in the cases presented in the paper. Importantly, the SR-CND model maintains accuracy in the attached boundary layers, whereas the SR-CLS model does not. Therefore, the proposed method improves separated flow predictions while maintaining the accuracy of the original model for attached flows, offering a better way to create data-driven turbulence models.