Machine-Learning-Assisted Blending of Data-Driven Turbulence Models

Locations

  • Flow Turbulence and Combustion
  • arXiv (Cornell University)

Ask a Question About This Paper

Summary

This work introduces a significant advancement in turbulence modeling for computational fluid dynamics (CFD) by presenting a machine learning-based framework for adaptively blending specialized data-driven turbulence models. The primary motivation stems from the long-standing challenge of achieving generalizability and accuracy in Reynolds-Averaged Navier-Stokes (RANS) simulations across diverse, complex flow regimes. Traditional RANS models, while computationally efficient, often struggle with non-equilibrium turbulence, strong gradients, and separated flows, and data-driven models, while promising, typically specialize to narrow flow classes and may fail to generalize or even degrade performance in conditions outside their training.

The core innovation lies in its intrusive internal blending strategy. Unlike prior “external” blending approaches that combine the predictions of multiple RANS models, which often violate conservation laws and require computationally expensive multiple simulations, this method directly blends the corrective terms applied to a baseline RANS turbulence model (specifically, the k-ω SST model). By working at the level of model corrections rather than full solutions, the resulting blended model inherently satisfies the governing RANS equations, preserving physical conservation principles and requiring only a single RANS simulation for prediction.

A second key innovation is the adaptive, localized weighting mechanism. The framework employs a Random Forest Regressor (RFR) to dynamically determine the blending weights for each specialized model at every spatial location within the flow domain. This RFR is trained to map local physical flow features—such as turbulent kinetic energy, pressure gradients, and strain rates—to the optimal weights, which are derived from a Gaussian kernel that quantifies how well each specialized model performs compared to high-fidelity data in specific flow regions. This allows the blended model to intelligently select the most appropriate “expert” in real-time, based on the local flow characteristics, effectively leveraging the strengths of individual models where they are most accurate.

The foundational prior ingredients enabling this work include:
* Reynolds-Averaged Navier-Stokes (RANS) Equations and Baseline Turbulence Models: The standard k-ω SST model forms the bedrock upon which the data-driven corrections are applied and blended.
* Data-Driven Turbulence Modeling: The general concept of using machine learning to improve RANS closures, particularly the approach of augmenting baseline models with learned correction terms for the Reynolds stresses and transport equations.
* Sparse Bayesian Learning (SBL) and Symbolic Regression: These techniques, particularly the SBL-SpaRTA algorithm, are crucial for training the specialized “expert” data-driven models. These experts are pre-trained for distinct flow classes like turbulent channel flows, separated flows, and axisymmetric jets, providing the individual components that the blending framework then combines.
* Mixture-of-Experts (MoE) Concepts: While deviating significantly in its intrusive nature, the paper builds on the conceptual foundation of MoE architectures and space-dependent model aggregation techniques, which aim to combine the predictive power of multiple individual models.
* Physical Flow Features for Machine Learning: The selection of specific, physically interpretable flow features (e.g., non-dimensional strain and rotation rates, pressure gradients) is essential as inputs for the RFR to learn the local blending weights. These features allow the ML model to understand the underlying physics of different flow regions.
* Random Forest Regression: This ensemble machine learning algorithm is chosen for its robustness and ability to capture complex, non-linear relationships between the input flow features and the desired blending weights.
* High-Fidelity Data (DNS/LES/Experiments): Extensive high-fidelity data from Direct Numerical Simulations (DNS), Large Eddy Simulations (LES), and experiments are indispensable for both training the specialized expert models via symbolic regression and for training the RFR to determine the optimal blending weights.

The robust performance demonstrated across various test cases, including untrained scenarios like a NACA0012 airfoil at various angles of attack, underscores the potential of this framework to provide more generalizable, accurate, and physically consistent turbulence models for a wide range of engineering applications.

A machine learning-based methodology for blending data-driven turbulent closures for the Reynolds-Averaged Navier-Stokes (RANS) equations is proposed to improve the generalizability across different flow scenarios. Data-driven models based on sparse … A machine learning-based methodology for blending data-driven turbulent closures for the Reynolds-Averaged Navier-Stokes (RANS) equations is proposed to improve the generalizability across different flow scenarios. Data-driven models based on sparse Bayesian learning and symbolic regression are pre-trained for specific flow classes, including turbulent channel flows, separated flows, and axisymmetric jets. The specialized models (called "\textit{experts}") are then blended using a set of weighting functions that reflect the local likelihood of a model to capture the reference data. The weighting functions are expressed as functions of a set of local flow features, and Random Forest regressors (RFR) are trained alongside the expert models to learn the mapping between the features and the weights. The training dataset includes additional data for flows ranging from shear flows at equilibrium to separated flows. The resulting blended model is then used to predict unseen flows, including a turbulent zero-pressure-gradient flat plate and a wall-mounted hump involving attached and separated boundary layer regions. The results show that the blending strategy adapts to the local flow conditions and effectively leverages the strengths of individual models, promoting the best-performing models locally, in a manner consistent with the training scenarios. The blended model provides a robust approximation for unseen flows that exhibit similar underlying physical characteristics to the training cases, while also successfully extrapolating to flows with different geometries and physical configurations.
As part of a larger effort on data-driven turbulence modeling, this paper investigates machine learning models in their capability to reconstruct the functional forms of spatially distributed quantities extracted from … As part of a larger effort on data-driven turbulence modeling, this paper investigates machine learning models in their capability to reconstruct the functional forms of spatially distributed quantities extracted from high fidelity simulation and experimental data.Such datasets typically involve very high dimensional feature spaces with sparsely populated and noisy data.A new multiscale Gaussian process regression technique is described and is compared to 'conventional' Gaussian process regression and artificial neural networks.All these techniques are applied to the reconstruction of functions arising from Bayesian inference applied to turbulent channel flow and bypass transition.The efficiency, accuracy and effectiveness of each learning algorithm as well as factors that influence their output is assessed.The results highlight the potential of machine learning as an enabling tool in data-driven turbulence modeling.
Reynolds-averaged Navier-Stokes (RANS) equations are widely used in engineering turbulent flow simulations. However, the mean flow fields predicted by RANS solvers could come with large discrepancies due to the uncertainties … Reynolds-averaged Navier-Stokes (RANS) equations are widely used in engineering turbulent flow simulations. However, the mean flow fields predicted by RANS solvers could come with large discrepancies due to the uncertainties in modeled Reynolds stresses. Recently, Wang et al. demonstrated that machine learning can be used to improve the RANS modeled Reynolds stresses by leveraging data from high fidelity simulations (Physics informed machine learning approach for reconstructing Reynolds stress modeling discrepancies based on DNS data. Physical Review Fluids. 2, 034603, 2017). However, solving for mean flows from the machine-learning predicted Reynolds stresses still poses significant challenges. The present work is a critical extension of (Wang et al. 2017), and it enables the machine learning model to yield improved predictions of not only Reynolds stresses but also mean velocities therefrom. Such a development is of profound practical importance, because often the velocities and the derived quantities (e.g., drag, lift, surface friction), and not the Reynolds stresses per se, are the ultimate quantities of interest in RANS simulations. The present work has two innovations. First, we demonstrate a systematic procedure to generate mean flow features based on the integrity basis for a set of mean flow tensors, which is in contrast to the ad hoc choices features in (Wang et al. 2017). Second, we propose using machine learning to predict linear and nonlinear parts of the Reynolds stress tensor separately. Inspired by the finite polynomial representation of tensors in classical turbulence modeling, such a decomposition is instrumental in overcoming the instability of RANS equations. Several test cases are used to demonstrate the merits of the proposed approach.
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 … 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.
We present a comprehensive framework for augmenting turbulence models with physics-informed machine learning, illustrating a complete workflow from identification of input/output to prediction of mean velocities. The learned model has … We present a comprehensive framework for augmenting turbulence models with physics-informed machine learning, illustrating a complete workflow from identification of input/output to prediction of mean velocities. The learned model has Galilean invariance and coordinate rotational invariance.
Data-driven turbulence modeling is a newly emerged research area in thermal hydraulics simulation of nuclear power plant (NPP). The most common CFD method used in NPP thermal hydraulics simulation is … Data-driven turbulence modeling is a newly emerged research area in thermal hydraulics simulation of nuclear power plant (NPP). The most common CFD method used in NPP thermal hydraulics simulation is Reynolds-averaged Navier-Stokes (RANS) method, which still has acknowledged deficiencies not only in the calculation speed but also in the complexity of choosing turbulence model and parameters for different flow patterns. Data-driven turbulence modeling aims to develop a RANS-based method which not only computationally efficient but also applicable to different flow patterns. To achieve this goal, the first step is to develop an approach to properly perform RANS for selected flow patterns. In this work, a machine learning approach is selected to achieve this goal. The main purpose of this study is to perform a data-driven approach to model turbulence Reynolds stress leveraging the potential of massive direct numerical simulation (DNS) data. The approach is validated by a turbulence flow validation case: a parallel plane quasi-steady state turbulence flow case. The work contains three parts. The first part is database preparation. In this step, turbulence properties (Reynolds stress) are extracted from DNS results, which are considered as "physically correct data". Meanwhile, flow features are extracted from RANS results, which are considered as "data to be corrected". The second part is surrogate model establishment. In this step, a data-driven regression function is trained between flow features and turbulence properties obtained from the previous step. The last part is model validation, which is applying trained data-driven regression function to a test case to validate this approach.
The objective is to provide clear and well-motivated guidance to Machine Learning (ML) teams, founded on our experience in empirical turbulence modeling. Guidance is also needed for modeling outside ML. … The objective is to provide clear and well-motivated guidance to Machine Learning (ML) teams, founded on our experience in empirical turbulence modeling. Guidance is also needed for modeling outside ML. ML is not yet successful in turbulence modeling, and many papers have produced unusable proposals either due to errors in math or physics, or to severe overfitting. We believe that "Turbulence Culture" (TC) takes years to learn and is difficult to convey especially considering the modern lack of time for careful study; important facts which are self-evident after a career in turbulence research and modeling and extensive reading are easy to miss. In addition, many of them are not absolute facts, a consequence of the gaps in our understanding of turbulence and the weak connection of models to first principles. Some of the mathematical facts are rigorous, but the physical aspects often are not. Turbulence models are surprisingly arbitrary. Disagreement between experts confuses the new entrants. In addition, several key properties of the models are ascertained through non-trivial analytical properties of the differential equations, which puts them out of reach of purely data-driven ML-type approaches. The best example is the crucial behavior of the model at the edge of the turbulent region (ETR). The knowledge we wish to put out here may be divided into "Mission" and "Requirements," each combining physics and mathematics. Clear lists of "Hard" and "Soft" constraints are presented. A concrete example of how DNS data could be used, possibly allied with ML, is first carried through and illustrates the large number of decisions needed. Our focus is on creating effective products which will empower CFD, rather than on publications.
A data-driven approach to the modeling of turbulent and transitional flows is proposed in this work, with the goal of developing more robust and accurate closure models.The key idea is … A data-driven approach to the modeling of turbulent and transitional flows is proposed in this work, with the goal of developing more robust and accurate closure models.The key idea is to (i) infer the functional form of deficiencies in known closure models by applying inverse problems to computational and experimental data, (ii) use machine learning to reconstruct the improved functional forms, and (iii) to inject the improved functional forms in simulations to obtain more accurate predictions.The inverse modeling step, on its own, can yield valuable insight to the modeler, essentially converting data to information.The machine learning step is a tool to convert information into modeling knowledge.Representative examples are used to describe the methodology and to demonstrate its viability.The first example investigates the modeling of a non-equilibrium turbulent boundary layer, and the second involves the modeling of bypass transition to turbulence.Evidence from these problems emphasizes the utility of the proposed approach in offering new routes to closure modeling in general computational physics disciplines.
Although an increased availability of computational resources has enabled high-fidelity simulations of turbulent flows, the RANS models are still the dominant tools for industrial applications. However, the predictive capabilities of … Although an increased availability of computational resources has enabled high-fidelity simulations of turbulent flows, the RANS models are still the dominant tools for industrial applications. However, the predictive capabilities of RANS models are limited by potential inaccuracy driven by hypotheses in the Reynolds stress closure. Recently, a Physics-Informed Machine Learning (PIML) approach has been proposed to learn the functional form of Reynolds stress discrepancy in RANS simulations based on available data. It has been demonstrated that the learned discrepancy function can be used to improve Reynolds stresses in different flows where data are not available. However, owing to a number of challenges, the improvements have been demonstrated only in the Reynolds stress prediction but not in the corresponding propagated quantities of interest. In this work, we introduce the procedures toward a complete PIML framework for predictive turbulence modeling, including learning Reynolds stress discrepancy function, predicting Reynolds stresses in different flows, and propagating to mean flow fields. The process of Reynolds stress propagation and predictive accuracy of the propagated velocity field are investigated. To improve the learning-prediction performance, the input features are enriched based on an integrity basis of invariants. The fully developed turbulent flow in a square duct is used as the test case. The discrepancy model is trained on flow fields obtained from several Reynolds numbers and evaluated on a duct flow at a Reynolds number higher than any of the training cases. The predicted Reynolds stresses are propagated to velocity field through RANS equations. Numerical results show excellent predictive performances in both Reynolds stresses and their propagated velocities, demonstrating the merits of the PIML approach in predictive turbulence modeling.
The recent surge in machine learning augmented turbulence modelling is a promising approach for addressing the limitations of Reynolds-averaged Navier-Stokes (RANS) models. This work presents the development of the first … The recent surge in machine learning augmented turbulence modelling is a promising approach for addressing the limitations of Reynolds-averaged Navier-Stokes (RANS) models. This work presents the development of the first open-source dataset, curated and structured for immediate use in machine learning augmented turbulence closure modelling. The dataset features a variety of RANS simulations with matching direct numerical simulation (DNS) and large-eddy simulation (LES) data. Four turbulence models are selected to form the initial dataset: $k$-$\varepsilon$, $k$-$\varepsilon$-$\phi_t$-$f$, $k$-$\omega$, and $k$-$\omega$ SST. The dataset consists of 29 cases per turbulence model, for several parametrically sweeping reference DNS/LES cases: periodic hills, square duct, parametric bumps, converging-diverging channel, and a curved backward-facing step. At each of the 895,640 points, various RANS features with DNS/LES labels are available. The feature set includes quantities used in current state-of-the-art models, and additional fields which enable the generation of new feature sets. The dataset reduces effort required to train, test, and benchmark new models. The dataset is available at https://doi.org/10.34740/kaggle/dsv/2044393 .
The recent surge in machine learning augmented turbulence modelling is a promising approach for addressing the limitations of Reynolds-averaged Navier-Stokes (RANS) models. This work presents the development of the first … The recent surge in machine learning augmented turbulence modelling is a promising approach for addressing the limitations of Reynolds-averaged Navier-Stokes (RANS) models. This work presents the development of the first open-source dataset, curated and structured for immediate use in machine learning augmented turbulence closure modelling. The dataset features a variety of RANS simulations with matching direct numerical simulation (DNS) and large-eddy simulation (LES) data. Four turbulence models are selected to form the initial dataset: $k$-$\varepsilon$, $k$-$\varepsilon$-$\phi_t$-$f$, $k$-$\omega$, and $k$-$\omega$ SST. The dataset consists of 29 cases per turbulence model, for several parametrically sweeping reference DNS/LES cases: periodic hills, square duct, parametric bumps, converging-diverging channel, and a curved backward-facing step. At each of the 895,640 points, various RANS features with DNS/LES labels are available. The feature set includes quantities used in current state-of-the-art models, and additional fields which enable the generation of new feature sets. The dataset reduces effort required to train, test, and benchmark new models. The dataset is available at this https URL .
Turbulence modeling in a Reynolds Averaged Navier-Stokes (RANS) setting has traditionally evolved through a combination of theory, mathematics, and empiricism.The problem of closure, resulting from the averaging process, requires an … Turbulence modeling in a Reynolds Averaged Navier-Stokes (RANS) setting has traditionally evolved through a combination of theory, mathematics, and empiricism.The problem of closure, resulting from the averaging process, requires an infusion of information into the various models that is often managed in an ad-hoc way or that is focused on particular classes of problems, thus diminishing the predictive capabilities of a model in other flow contexts.In this work, a proof-of-concept of a new data-driven approach of turbulence model development is presented.The key idea in the proposed framework is to use supervised learning algorithms to build a representation of turbulence modeling closure terms.The learned terms are then inserted into a Computational Fluid Dynamics (CFD) numerical simulation with the aim of offering a better representation of turbulence physics.But while the basic idea is attractive, modeling unknown terms by increasingly large amounts of data from higher-fidelity simulations (LES, DNS, etc) or even experiment, the details of how to make the approach viable are not at all obvious.In this work, we investigate the feasibility of such an approach by attempting to reproduce, through a machine learning methodology, the results obtained with the well-established Spalart-Allmaras model.In other words, the key question that we seek to answer is the following: Given a number of observations of CFD solutions using the Spalart-Allmaras model (our truth model), can we reproduce those solutions using machine-learning techniques without knowledge of the structure, functional form, and coefficients of the actual model?We discuss the challenges of applying machine learning techniques in a fluid dynamic setting and possible successful approaches.We also explore the potential for machine learning as an enhancement to or replacement for traditional turbulence models.Our results highlight the potential and viability of machine learning approaches to aid turbulence model development.
In this article, we propose a data-driven methodology for combining the solutions of a set of competing turbulence models. The individual model predictions are linearly combined for providing an ensemble … In this article, we propose a data-driven methodology for combining the solutions of a set of competing turbulence models. The individual model predictions are linearly combined for providing an ensemble solution accompanied by estimates of predictive uncertainty due to the turbulence model choice. First, for a set of training flow configurations we assign to component models high weights in the regions where they best perform, and vice versa, by introducing a measure of distance between high-fidelity data and individual model predictions. The model weights are then mapped into a space of features, representative of local flow physics, and regressed by a Random Forests (RF) algorithm. The RF regressor is finally employed to infer spatial distributions of the model weights for unseen configurations. Predictions of new cases are constructed as a convex linear combination of the underlying models solutions, while the between model variance provides information about regions of high model uncertainty. The method is demonstrated for a class of flows through the compressor cascade NACA65 V103 at Re~3e5. The results show that the aggregated solution outperforms the accuracy of individual models for the quantity used to inform the RF regressor, and performs well for other quantities well-correlated to the preceding one. The estimated uncertainty intervals are generally consistent with the target high-fidelity data. The present approach then represents a viable methodology for a more objective selection and combination of alternative turbulence models in configurations of interest for engineering practice
A probabilistic machine learning model is introduced to augment the k-ω SST turbulence model in order to improve the modelling of separated flows and the generalisability of learnt corrections. Increasingly, … A probabilistic machine learning model is introduced to augment the k-ω SST turbulence model in order to improve the modelling of separated flows and the generalisability of learnt corrections. Increasingly, machine learning methods have been used to leverage experimental and high-fidelity data, improving the accuracy of the Reynolds Averaged Navier Stokes (RANS) turbulence models widely used in industry. A significant challenge for such methods is their ability to generalise to unseen geometries and flow conditions. Furthermore, heterogeneous datasets containing a mix of experimental and simulation data must be efficiently handled. In this work, field inversion and an ensemble of Gaussian Process Emulators (GPEs) is employed to address both of these challenges. The ensemble model is applied to a range of benchmark test cases, demonstrating improved turbulence modelling for cases with separated flows with adverse pressure gradients, where RANS simulations are understood to be unreliable. Perhaps more significantly, the simulation reverted to the uncorrected model in regions of the flow exhibiting physics outside of the training data.
A probabilistic machine learning model is introduced to augment the $k-\omega\ SST$ turbulence model in order to improve the modelling of separated flows and the generalisability of learnt corrections. Increasingly, … A probabilistic machine learning model is introduced to augment the $k-\omega\ SST$ turbulence model in order to improve the modelling of separated flows and the generalisability of learnt corrections. Increasingly, machine learning methods have been used to leverage experimental and high-fidelity data, improving the accuracy of the Reynolds Averaged Navier Stokes (RANS) turbulence models widely used in industry. A significant challenge for such methods is their ability to generalise to unseen geometries and flow conditions. Furthermore, heterogeneous datasets containing a mix of experimental and simulation data must be efficiently handled. In this work, field inversion and an ensemble of Gaussian Process Emulators (GPEs) is employed to address both of these challenges. The ensemble model is applied to a range of benchmark test cases, demonstrating improved turbulence modelling for cases with separated flows with adverse pressure gradients, where RANS simulations are understood to be unreliable. Perhaps more significantly, the simulation reverted to the uncorrected model in regions of the flow exhibiting physics outside of the training data.
This work presents a review and perspectives on recent developments in the use of machine learning (ML) to augment Reynolds-averaged Navier--Stokes (RANS) and Large Eddy Simulation (LES) models of turbulent … This work presents a review and perspectives on recent developments in the use of machine learning (ML) to augment Reynolds-averaged Navier--Stokes (RANS) and Large Eddy Simulation (LES) models of turbulent flows. Different approaches of applying supervised learning to represent unclosed terms, model discrepancies and sub-filter scales are discussed in the context of RANS and LES modeling. Particular emphasis is placed on the impact of the training procedure on the consistency of ML augmentations with the underlying physical model. Techniques to promote model-consistent training, and to avoid the requirement of full fields of direct numerical simulation data are detailed. This is followed by a discussion of physics-informed and mathematical considerations on the choice of the feature space, and imposition of constraints on the ML model. With a view towards developing generalizable ML-augmented RANS and LES models, outstanding challenges are discussed, and perspectives are provided. While the promise of ML-augmented turbulence modeling is clear, and successes have been demonstrated in isolated scenarios, a general consensus of this paper is that truly generalizable models require model-consistent training with careful characterization of underlying assumptions and imposition of physically and mathematically informed priors and constraints to account for the inevitable shortage of data relevant to predictions of interest. Thus, machine learning should be viewed as one tool in the turbulence modeler's toolkit. This modeling endeavor requires multi-disciplinary advances, and thus the target audience for this paper is the fluid mechanics community, as well as the computational science and machine learning communities.
This work aims to incorporate basic calibrations of Reynolds-averaged Navier-Stokes (RANS) models as part of machine learning (ML) frameworks. The ML frameworks considered are tensor-basis neural network (TBNN), physics-informed machine … This work aims to incorporate basic calibrations of Reynolds-averaged Navier-Stokes (RANS) models as part of machine learning (ML) frameworks. The ML frameworks considered are tensor-basis neural network (TBNN), physics-informed machine learning (PIML), and field inversion & machine learning (FIML) in J. Fluid Mech., 2016, 807, 155-166, Phys. Rev. Fluids, 2017, 2(3), 034603 and J. Comp. Phys., 2016, 305, 758-774, and the baseline RANS models are the one-equation Spalart-Allmaras model, the two-equation $k$-$\omega$ model, and the seven-equation Reynolds stress transport models. ML frameworks are trained against plane channel flow and shear-layer flow data. We compare the ML frameworks and study whether the machine-learned augmentations are detrimental outside the training set. The findings are summarized as follows. The augmentations due to TBNN are detrimental. PIML leads to augmentations that are beneficial inside the training dataset but detrimental outside it. These results are not affected by the baseline RANS model. FIML's augmentations to the two eddy viscosity models, where an inner-layer treatment already exists, are largely neutral. Its augmentation to the seven-equation model, where an inner-layer treatment does not exist, improves the mean flow prediction in a channel. Furthermore, these FIML augmentations are mostly non-detrimental outside the training dataset. In addition to reporting these results, the paper offers physical explanations of the results. Last, we note that the conclusions drawn here are confined to the ML frameworks and the flows considered in this study. More detailed comparative studies and validation & verification studies are needed to account for developments in recent years.
Theory of wing sections, including a summary of airfoil data , Theory of wing sections, including a summary of airfoil data , مرکز فناوری اطلاعات و اطلاع رسانی کشاورزی Theory of wing sections, including a summary of airfoil data , Theory of wing sections, including a summary of airfoil data , مرکز فناوری اطلاعات و اطلاع رسانی کشاورزی
Reynolds Averaged Navier Stokes (RANS) models are widely used in industry to predict fluid flows, despite their acknowledged deficiencies. Not only do RANS models often produce inaccurate flow predictions, but … Reynolds Averaged Navier Stokes (RANS) models are widely used in industry to predict fluid flows, despite their acknowledged deficiencies. Not only do RANS models often produce inaccurate flow predictions, but there are very limited diagnostics available to assess RANS accuracy for a given flow configuration. If experimental or higher fidelity simulation results are not available for RANS validation, there is no reliable method to evaluate RANS accuracy. This paper explores the potential of utilizing machine learning algorithms to identify regions of high RANS uncertainty. Three different machine learning algorithms were evaluated: support vector machines, Adaboost decision trees, and random forests. The algorithms were trained on a database of canonical flow configurations for which validated direct numerical simulation or large eddy simulation results were available, and were used to classify RANS results on a point-by-point basis as having either high or low uncertainty, based on the breakdown of specific RANS modeling assumptions. Classifiers were developed for three different basic RANS eddy viscosity model assumptions: the isotropy of the eddy viscosity, the linearity of the Boussinesq hypothesis, and the non-negativity of the eddy viscosity. It is shown that these classifiers are able to generalize to flows substantially different from those on which they were trained. Feature selection techniques, model evaluation, and extrapolation detection are discussed in the context of turbulence modeling applications.
Keywords: ecoulement : compressible Note: + disquette Reference Record created on 2005-11-18, modified on 2016-08-08 Keywords: ecoulement : compressible Note: + disquette Reference Record created on 2005-11-18, modified on 2016-08-08
Standard statistical practice ignores model uncertainty. Data analysts typically select a model from some class of models and then proceed as if the selected model had generated the data. This … Standard statistical practice ignores model uncertainty. Data analysts typically select a model from some class of models and then proceed as if the selected model had generated the data. This approach ignores the uncertainty in model selection, leading to over-confident inferences and decisions that are more risky than one thinks they are. Bayesian model averaging (BMA)provides a coherent mechanism for accounting for this model uncertainty. Several methods for implementing BMA have recently emerged. We discuss these methods and present a number of examples.In these examples, BMA provides improved out-of-sample predictive performance. We also provide a catalogue of currently available BMA software.
After an extensive survey of mean-velocity profile measurements in various two-dimensional incompressible turbulent boundary-layer flows, it is proposed to represent the profile by a linear combination of two universal functions. … After an extensive survey of mean-velocity profile measurements in various two-dimensional incompressible turbulent boundary-layer flows, it is proposed to represent the profile by a linear combination of two universal functions. One is the well-known law of the wall. The other, called the law of the wake, is characterized by the profile at a point of separation or reattachment. These functions are considered to be established empirically, by a study of the mean-velocity profile, without reference to any hypothetical mechanism of turbulence. Using the resulting complete analytic representation for the mean-velocity field, the shearing-stress field for several flows is computed from the boundary-layer equations and compared with experimental data. The development of a turbulent boundary layer is ultimately interpreted in terms of an equivalent wake profile, which supposedly represents the large-eddy structure and is a consequence of the constraint provided by inertia. This equivalent wake profile is modified by the presence of a wall, at which a further constraint is provided by viscosity. The wall constraint, although it penetrates the entire boundary layer, is manifested chiefly in the sublayer flow and in the logarithmic profile near the wall. Finally, it is suggested that yawed or three-dimensional flows may be usefully represented by the same two universal functions, considered as vector rather than scalar quantities. If the wall component is defined to be in the direction of the surface shearing stress, then the wake component, at least in the few cases studied, is found to be very nearly parallel to the gradient of the pressure.
We present a tree-structured architecture for supervised learning. The statistical model underlying the architecture is a hierarchical mixture model in which both the mixture coefficients and the mixture components are … We present a tree-structured architecture for supervised learning. The statistical model underlying the architecture is a hierarchical mixture model in which both the mixture coefficients and the mixture components are generalized linear models (GLIM's). Learning is treated as a maximum likelihood problem; in particular, we present an Expectation-Maximization (EM) algorithm for adjusting the parameters of the architecture. We also develop an on-line learning algorithm in which the parameters are updated incrementally. Comparative simulation results are presented in the robot dynamics domain.
Many tasks in fluids engineering require knowledge of the turbulence in jets. There is a strong, although fragmented, literature base for low order statistics, such as jet spread and other … Many tasks in fluids engineering require knowledge of the turbulence in jets. There is a strong, although fragmented, literature base for low order statistics, such as jet spread and other meanvelocity field characteristics. Some sources, particularly for low speed cold jets, also provide turbulence intensities that are required for validating Reynolds-averaged Navier-Stokes (RANS) Computational Fluid Dynamics (CFD) codes. There are far fewer sources for jet spectra and for space-time correlations of turbulent velocity required for aeroacoustics applications, although there have been many singular publications with various unique statistics, such as Proper Orthogonal Decomposition, designed to uncover an underlying low-order dynamical description of turbulent jet flow. As the complexity of the statistic increases, the number of flows for which the data has been categorized and assembled decreases, making it difficult to systematically validate prediction codes that require high-level statistics over a broad range of jet flow conditions. For several years, researchers at NASA have worked on developing and validating jet noise prediction codes. One such class of codes, loosely called CFD-based or statistical methods, uses RANS CFD to predict jet mean and turbulent intensities in velocity and temperature. These flow quantities serve as the input to the acoustic source models and flow-sound interaction calculations that yield predictions of far-field jet noise. To develop this capability, a catalog of turbulent jet flows has been created with statistics ranging from mean velocity to space-time correlations of Reynolds stresses. The present document aims to document this catalog and to assess the accuracies of the data, e.g. establish uncertainties for the data. This paper covers the following five tasks: Document acquisition and processing procedures used to create the particle image velocimetry (PIV) datasets. Compare PIV data with hotwire and laser Doppler velocimetry (LDV) data published in the open literature. Compare different datasets acquired at roughly the same flow conditions to establish uncertainties. Create a consensus dataset for a range of hot jet flows, including uncertainty bands. Analyze this consensus dataset for self-consistency and compare jet characteristics to those of the open literature. One final objective fulfilled by this work was the demonstration of a universal scaling for the jet flow fields, at least within the region of interest to aeroacoustics. The potential core length and the spread rate of the half-velocity radius were used to collapse of the mean and turbulent velocity fields over the first 20 jet diameters in a highly satisfying manner.
Some new developments of explicit algebraic Reynolds stress turbulence models (EARSM) are presented. The new developments include a new near-wall treatment ensuring realizability for the individual stress components, a formulation … Some new developments of explicit algebraic Reynolds stress turbulence models (EARSM) are presented. The new developments include a new near-wall treatment ensuring realizability for the individual stress components, a formulation for compressible flows, and a suggestion for a possible approximation of diffusion terms in the anisotropy transport equation. Recent developments in this area are assessed and collected into a model for both incompressible and compressible three-dimensional wall-bounded turbulent flows. This model represents a solution of the implicit ARSM equations, where the production to dissipation ratio is obtained as a solution to a nonlinear algebraic relation. Three-dimensionality is fully accounted for in the mean flow description of the stress anisotropy. The resulting EARSM has been found to be well suited to integration to the wall and all individual Reynolds stresses can be well predicted by introducing wall damping functions derived from the van Driest damping function. The platform for the model consists of the transport equations for the kinetic energy and an auxiliary quantity. The proposed model can be used with any such platform, and examples are shown for two different choices of the auxiliary quantity.
The commonly used linear K-l and K -ε models of turbulence are shown to be incapable of accurately predicting turbulent flows where the normal Reynolds stresses play an important role. … The commonly used linear K-l and K -ε models of turbulence are shown to be incapable of accurately predicting turbulent flows where the normal Reynolds stresses play an important role. By means of an asymptotic expansion, nonlinear K-l and K -ε models are obtained which, unlike all such previous nonlinear models, satisfy both realizability and the necessary invariance requirements. Calculations are presented which demonstrate that this nonlinear model is able to predict the normal Reynolds stresses in turbulent channel flow much more accurately than the linear model. Furthermore, the nonlinear model is shown to be capable of predicting turbulent secondary flows in non-circular ducts - a phenomenon which the linear models are fundamentally unable to describe. An additional application of this model to the improved prediction of separated flows is discussed briefly along with other possible avenues of future research.
Low-speed flow separation over a wall-mounted hump, and its control using steady suction, were studied experimentally in order to generate a data set for the development and evaluation of computational … Low-speed flow separation over a wall-mounted hump, and its control using steady suction, were studied experimentally in order to generate a data set for the development and evaluation of computational methods. The baseline and controlled data sets comprised time-mean and unsteady surface pressure measurements, flowfield measurements using particle image velocimetry, and wall shear stress obtained via oil-film interferometry. In addition to the specific test cases studied, surface pressures for a wide variety of conditions were acquired for different Reynolds numbers and suction rates. Stereoscopic particle image velocimetry and oil-film flow visualization indicated that the baseline time-averaged separated flowfield was two-dimensional. With the application of control, mild three-dimensionality was evident in the spanwise variation of pressure recovery, reattachment location, and spanwise pressure fluctuations.
Highly resolved large-eddy simulation (LES) is used to investigate the characteristics of a canonical boundary layer separating from a curved step in a channel of height 8.5 times that of … Highly resolved large-eddy simulation (LES) is used to investigate the characteristics of a canonical boundary layer separating from a curved step in a channel of height 8.5 times that of the step. The flow is treated as statistically spanwise homogeneous, in line with the conditions of a related experimental study in a large aspect ratio channel, undertaken within a companion research programme. Primary attention focuses on the details of the separation process and the properties of the separated region, including reattachment. Results are reported and analysed, from a flow physical perspective, for a wide variety of properties, including wall pressure and skin friction, mean velocity, Reynolds stresses and related anisotropy maps, two-point-correlation functions, unsteadiness indicators, budgets of the Reynolds stresses and length scales characterising the turbulence, and mean strain fields. The study highlights a range of distinctive features of separation from gently curved surfaces: the separation process is highly unsteady in time and space; turbulence is highly non-local in character; the mean reverse-flow region is thin and highly elongated; no part of the flow is reversed at all times; the level of production is extremely high following separation, resulting in massive departures from turbulence energy equilibrium, very high anisotropy and a trend towards one-component turbulence in the separated shear layer. The result, apart from offering insight into the physics of separation, constitutes a valuable data set for benchmarking model solutions and investigating statistical turbulence-closure proposals.
Reynolds-averaged Navier-Stokes (RANS) simulations are a practical approach for solving complex multi-physics turbulent flows, but the underlying assumptions of the turbulence models introduce errors and uncertainties in the simulation outcome. … Reynolds-averaged Navier-Stokes (RANS) simulations are a practical approach for solving complex multi-physics turbulent flows, but the underlying assumptions of the turbulence models introduce errors and uncertainties in the simulation outcome. The flow in scramjet combustors is an example of such a complex flow and the accurate characterization of safety and operability limits of these engines using RANS simulations requires an assessment of the model uncertainty. The objective of this paper is to present a framework for the epistemic uncertainty quantification of turbulence and mixing models in RANS simulations. The capabilities of the methodology will be demonstrated by performing simulations of the mixing of an underexpanded jet in a supersonic cross flow, which involves many flow features observed in scramjet engines. The fundamental sources of uncertainty in the RANS simulations are the models used for the Reynolds stresses in the momentum equations and the turbulent scalar fluxes in the scalar transport equations. The methodology consists in directly perturbing the modeled quantities in the equations, thereby establishing a method that is completely independent of the initial model form to overcome the limitations of traditional sensitivity studies. The perturbations are defined in terms of the decomposed Reynolds stress tensor, i.e., the tensor magnitude and the eigenvalues and eigenvectors of the normalized anisotropy tensor. The turbulent scalar fluxes are perturbed by using the perturbed Reynolds stresses in a generalized gradient diffusion model formulation and by changing the model constant. The perturbations were parameterized based on a comparison between the Reynolds stresses obtained from a baseline RANS simulation and those obtained from a large-eddy simulation database. Subsequently an optimization problem was solved, varying the parameters in the perturbation functions to maximize a quantity of interest that quantifies the downstream mixing. The result encompasses the value for the quantity of interest obtained from the LES database. It is shown that a traditional sensitivity study, in which the turbulent Schmidt number is varied, cannot capture this uncertainty, which further demonstrates the effectiveness of the proposed approach.
In this paper, we provide a comprehensive survey of the mixture of experts (ME). We discuss the fundamental models for regression and classification and also their training with the expectation-maximization … In this paper, we provide a comprehensive survey of the mixture of experts (ME). We discuss the fundamental models for regression and classification and also their training with the expectation-maximization algorithm. We follow the discussion with improvements to the ME model and focus particularly on the mixtures of Gaussian process experts. We provide a review of the literature for other training methods, such as the alternative localized ME training, and cover the variational learning of ME in detail. In addition, we describe the model selection literature which encompasses finding the optimum number of experts, as well as the depth of the tree. We present the advances in ME in the classification area and present some issues concerning the classification model. We list the statistical properties of ME, discuss how the model has been modified over the years, compare ME to some popular algorithms, and list several applications. We conclude our survey with future directions and provide a list of publicly available datasets and a list of publicly available software that implement ME. Finally, we provide examples for regression and classification. We believe that the study described in this paper will provide quick access to the relevant literature for researchers and practitioners who would like to improve or use ME, and that it will stimulate further studies in ME.
The skin-friction distribution on a wall-mounted hump model has been obtained using oil-film interferometry. This effort is part of a larger study to provide validation cases for simulations of unsteady … The skin-friction distribution on a wall-mounted hump model has been obtained using oil-film interferometry. This effort is part of a larger study to provide validation cases for simulations of unsteady flows. The challenges of using oil-film interferometry on this model, including model curvature and close camera proximity, are discussed. Skin-friction measurements are obtained over most of the hump model, including especially high-quality measurements in the separated and reattachment regions. These results highlight the method’s ability to capture a wide range of skin friction including measurements in reverse-flow and high-gradient regions. The wall skin-friction data are shown to complement other experimental data, and the use of independent skin-friction measurements for scaling in wall-bounded flows is emphasized. A comparison with results from several computational simulations of the same flow is presented. The comparison indicates that, for the most part, the computations accurately predict the skin-friction ahead of separation, but fail to predict the reattachment point correctly, and thus the comparison in the separated and recovery regions of the flow is poor. The ability of the skin-friction measurements to pinpoint regions where the computation performs poorly in the near-wall region is also presented. From these results, it is evident that independent skin-friction measurements should be a part of all validation experiments conducted in wall-bounded flows.
The control of a separated flow over a wall-mounted hump, by means of two-dimensional zero mass-flux perturbations, was studied experimentally to generate a data set for the development and evaluation … The control of a separated flow over a wall-mounted hump, by means of two-dimensional zero mass-flux perturbations, was studied experimentally to generate a data set for the development and evaluation of computational methods. The companion paper (Part 1) considered details of the baseline (uncontrolled) case and a steady-suction control case. The data set for a specific zero mass-flux control case comprised static surface pressures together with phase-averaged unsteady surface pressures and particle image velocimetry flowfield measurements. Additional surface pressures were acquired for a variety of control frequencies, control amplitudes and Reynolds numbers. Due consideration was given to characterizing the flow in the vicinity of the control slot, with and without external flow, and to perturbation two-dimensionality. Triple-decomposition of the fluctuating velocity and pressure fields was employed for presenting and analyzing the experimental data. This facilitated an assessment of the mechanism of separation control and the quantification of the coherent and turbulent surface pressures, Reynolds stresses, and energy fluxes. Spanwise surface pressures and phase-averaged stereoscopic particle image velocimetry data revealed an effectively two-dimensional flowfield despite highly three-dimensional instantaneous flow structures.
High-dimensional data can be converted to low-dimensional codes by training a multilayer neural network with a small central layer to reconstruct high-dimensional input vectors. Gradient descent can be used for … High-dimensional data can be converted to low-dimensional codes by training a multilayer neural network with a small central layer to reconstruct high-dimensional input vectors. Gradient descent can be used for fine-tuning the weights in such “autoencoder” networks, but this works well only if the initial weights are close to a good solution. We describe an effective way of initializing the weights that allows deep autoencoder networks to learn low-dimensional codes that work much better than principal components analysis as a tool to reduce the dimensionality of data.
Explicit algebraic stress models that are valid for three-dimensional turbulent flows in non-inertial frames are systematically derived from a hierarchy of second-order closure models. This represents a generalization of the … Explicit algebraic stress models that are valid for three-dimensional turbulent flows in non-inertial frames are systematically derived from a hierarchy of second-order closure models. This represents a generalization of the model derived by Pope (1975) who based his analysis on the Launder, Reece & Rodi model restricted to two-dimensional turbulent flows in an inertial frame. The relationship between the new models and traditional algebraic stress models – as well as anisotropic eddy viscosity models – is theoretically established. A need for regularization is demonstrated in an effort to explain why traditional algebraic stress models have failed in complex flows. It is also shown that these explicit algebraic stress models can shed new light on what second-order closure models predict for the equilibrium states of homogeneous turbulent flows and can serve as a useful alternative in practical computations.
High-resolution large-eddy simulation is used to investigate the mean and turbulence properties of a separated flow in a channel constricted by periodically distributed hill-shaped protrusions on one wall that obstruct … High-resolution large-eddy simulation is used to investigate the mean and turbulence properties of a separated flow in a channel constricted by periodically distributed hill-shaped protrusions on one wall that obstruct the channel by 33% of its height and are arranged 9 hill heights apart. The geometry is a modification of an experimental configuration, the adaptation providing an extended region of post-reattachment recovery and allowing high-quality simulations to be performed at acceptable computing costs. The Reynolds number, based on the hill height and the bulk velocity above the crest is 10595. The simulated domain is streamwise as well as spanwise periodic, extending from one hill crest to the next in the streamwise direction and over 4.5 hill heights in the spanwise direction. This arrangement minimizes uncertainties associated with boundary conditions and makes the flow an especially attractive generic test case for validating turbulence closures for statistically two-dimensional separation. The emphasis of the study is on elucidating the turbulence mechanisms associated with separation, recirculation reattachment, acceleration and wall proximity. Hence, careful attention has been paid to resolution, and a body-fitted, low-aspect-ratio, nearly orthogonal numerical grid of close to 5 million nodes has been used. Unusually, the results of two entirely independent simulations with different codes for identical flow and numerical conditions are compared and shown to agree closely. Results are included for mean velocity, Reynolds stresses, anisotropy measures, spectra and budgets for the Reynolds stresses. Moreover, an analysis of structural characteristics is undertaken on the basis of instantaneous realizations, and links to features observed in the statistical results are identified and interpreted. Among a number of interesting features, a distinct ‘splatting’ of eddies on the windward hill side following reattachment is observed, which generates strong spanwise fluctuations that are reflected, statistically, by the spanwise normal stress near the wall exceeding that of the streamwise stress by a substantial margin, despite the absence of spanwise strain.
A discussion of the applicability of an effective-viscosity approach to turbulent flow suggests that there are flow situations where the approach is valid and yet present hypotheses fail. The general … A discussion of the applicability of an effective-viscosity approach to turbulent flow suggests that there are flow situations where the approach is valid and yet present hypotheses fail. The general form of an effective-viscosity formulation is shown to be a finite tensor polynomial. For two-dimensional flows, the coefficients of this polynomial are evaluated from the modelled Reynolds-stress equations of Launder, Reece & Rodi (1975). The advantage of the proposed effective-viscosity formulation, equation (4.3), over isotropie-viscosity hypotheses is that the whole velocity-gradient tensor affects the predicted Reynolds stresses. Two notable consequences of this are that (i) the complete Reynolds-stress tensor is realistically modelled and (ii) the influence of streamline curvature on the Reynolds stresses is incorporated.
Abstract A computational large eddy simulation (LES) study is presented of the interaction between a turbulent boundary layer separating from a rounded ramp in a duct and a pair of … Abstract A computational large eddy simulation (LES) study is presented of the interaction between a turbulent boundary layer separating from a rounded ramp in a duct and a pair of spanwise-periodic, round synthetic jets, actuated upstream of the nominal separation line. Several scenarios are considered, for different injection angles and velocity ratios. In all cases, the actuation frequency corresponds to the shedding-instability mode of the separated shear layer. Experimental data, available for both the baseline flow and one actuated configuration, are used to verify the validity of the computational solutions. The analysis includes a separation of coherent and stochastic contributions to the time-averaged statistics of the auto- and cross-correlations of the fluctuations. The control authority is examined by reference to the effects of the actuation on the size of the separated zone, the momentum thickness of the boundary layer, the velocity field, various turbulence quantities and phase-averaged properties. The study demonstrates that the principal aspect of the interaction, at mean-flow level, is an increase in mixing provoked by the formation of strong streamwise vortices away from the wall, the induction of much weaker streamwise vortices close to the wall, and the extra production of stochastic turbulence caused by unsteady straining. The coherent stresses arising from the periodic perturbations are high – typically 5 times the levels of the unperturbed flow – but only within about 5–7 diameters of the jet orifice, and 2 orifice diameters on each side of the jet, and these are dominant primarily in the outer parts of the boundary layer. Stochastic turbulence is also elevated, but more modestly. The global effect of the actuation is a reduction of 10–20 % in the length of the separated region and 20–40 % in the thickness of the reverse-flow layer, depending on the actuation scheme, counter-flow actuation being the most effective. This reduction is mainly associated with a delay in separation. These results highlight the need for synthetic jets to be placed close to the separation zone and for the inter-jet distance to be of order 5 or lower to achieve a high level of separation-control authority.
Two new versions of the k - w two-equation turbulence model will be presented. The new Baseline (BSL) model is designed to give results similar to those of the original … Two new versions of the k - w two-equation turbulence model will be presented. The new Baseline (BSL) model is designed to give results similar to those of the original k - w model of Wilcox. but without its strong dependency on arbitrary freestream values. The BSL model is identical to the Wilcox model in the inner SOC7£; of the boundary-layer but changes gradually to the standard k - f. model (in a k - w fonnulation) towards the boundary-layer edge. The new model is also virtually identical to the k - f. model for free shear layers. The second version of the model is called Shear-Stress Transport (SSn model. It is a variation of the BSL model with the additional ability to account for the transport of the principal turbulent shear stress in adverse pressure gradient boundary-layers. The model is based on Bradshaw's assumption that the principal shear-stress is pro­ portional to the turbulent kinetic energy, which is introduced into the definition of the eddy-viscosity. Both models are tested for a large number of different fiowfields. The results of the BSL model are similar to those of the original k - w model, but without the undesirable free stream dependency. The predictions of the SST model are also independent of the freestrearn values but show better agreement with exper­ imental data for adverse pressure gradient boundary-layer flows.
The performances of four turbulence models are evaluated against eight selected experimental flow fields. The four models are the two-equation k-e model of Launder and Sharma, the two-equation k-a> model … The performances of four turbulence models are evaluated against eight selected experimental flow fields. The four models are the two-equation k-e model of Launder and Sharma, the two-equation k-a> model of Wilcox, the twoequation k-03 SST model of Menter, and the one-equation eddy-viscosity model of Spalart and Allmaras. The eight turbulent flows of the validation are four fully-developed freeshear flows, an incompressibl e boundary layer, and three complex flows with flow separation. The free-shear layer flows comprise a mixing layer, a round jet, a plane jet, and a plane wake flow. The three complex flows are comprised of an adverse-pressure-gradient boundary layer, an axisymmetric shock-wave/boundary-layer interaction, and a transonic RAE 2822 airfoil flow. The experimental data for these flows is well established and has been extensively used in model developments. The numerical predictions include mean velocity profiles, spreading rates, surface pressure coefficients, skin friction, and shear-stress profiles. Most significantly, this research includes a sensitivity study on the accuracy of the solutions with respect to the effects of freestream turbulence, grid resolution, grid spacing near the wall, initial conditions, numerical methods and codes, and free stream Mach number effects on incompressible flows.
Turbulence closure models are central to a good deal of applied computational fluid dynamical analysis. Closure modeling endures as a productive area of research. This review covers recent developments in … Turbulence closure models are central to a good deal of applied computational fluid dynamical analysis. Closure modeling endures as a productive area of research. This review covers recent developments in elliptic relaxation and elliptic blending models, unified rotation and curvature corrections, transition prediction, hybrid simulation, and data-driven methods. The focus is on closure models in which transport equations are solved for scalar variables, such as the turbulent kinetic energy, a timescale, or a measure of anisotropy. Algebraic constitutive representations are reviewed for their role in relating scalar closures to the Reynolds stress tensor. Seamless and nonzonal methods, which invoke a single closure model, are reviewed, especially detached eddy simulation (DES) and adaptive DES. Other topics surveyed include data-driven modeling and intermittency and laminar fluctuation models for transition prediction. The review concludes with an outlook.
Computational fluid dynamics analyses of high-Reynolds-number flows mostly rely on the Reynolds-averaged Navier–Stokes equations. The associated closure models are based on multiple simplifying assumptions and involve numerous empirical closure coefficients, … Computational fluid dynamics analyses of high-Reynolds-number flows mostly rely on the Reynolds-averaged Navier–Stokes equations. The associated closure models are based on multiple simplifying assumptions and involve numerous empirical closure coefficients, which are calibrated on a set of simple reference flows. Predicting new flows using a single closure model with nominal values for the closure coefficients may lead to biased predictions. Bayesian model-scenario averaging is a statistical technique providing an optimal way to combine the predictions of several competing models calibrated on various sets of data (scenarios). The method allows a stochastic estimate of a quantity of interest in an unmeasured prediction scenario to be obtain by 1) propagating posterior probability distributions of the parameters obtained for multiple calibration scenarios, and 2) computing a weighted posterior predictive distribution. Although step 2 has a negligible computational cost, step 1 requires a large number of samples of the solver. To enable the application of the proposed approach to computationally expensive flow configurations, a modified formulation is used where a maximum posterior probability approximation is used to drastically reduce the computational burden. The predictive capability of the proposed simplified approach is assessed for two-dimensional separated and three-dimensional compressible flows.
A direct numerical simulation of incompressible channel flow at a friction Reynolds number ( $\mathit{Re}_{{\it\tau}}$ ) of 5186 has been performed, and the flow exhibits a number of the characteristics … A direct numerical simulation of incompressible channel flow at a friction Reynolds number ( $\mathit{Re}_{{\it\tau}}$ ) of 5186 has been performed, and the flow exhibits a number of the characteristics of high-Reynolds-number wall-bounded turbulent flows. For example, a region where the mean velocity has a logarithmic variation is observed, with von Kármán constant ${\it\kappa}=0.384\pm 0.004$ . There is also a logarithmic dependence of the variance of the spanwise velocity component, though not the streamwise component. A distinct separation of scales exists between the large outer-layer structures and small inner-layer structures. At intermediate distances from the wall, the one-dimensional spectrum of the streamwise velocity fluctuation in both the streamwise and spanwise directions exhibits $k^{-1}$ dependence over a short range in wavenumber $(k)$ . Further, consistent with previous experimental observations, when these spectra are multiplied by $k$ (premultiplied spectra), they have a bimodal structure with local peaks located at wavenumbers on either side of the $k^{-1}$ range.