Machine Learning for RANS Turbulence Modelling of Variable Property Flows

Total Page:16

File Type:pdf, Size:1020Kb

Machine Learning for RANS Turbulence Modelling of Variable Property Flows Machine Learning for RANS Turbulence Modelling of Variable Property Flows Rafael Diez Technische Universiteit Delft Machine Learning for RANS Turbulence Modelling of Variable Property Flows by Rafael Diez in partial fulfillment of the requirements for the degree of Master of Science in Mechanical Engineering at the Delft University of Technology, to be defended publicly on Friday September 14, 2018 at 10:00 AM. Supervisor: Prof. dr. ir. R. Pecnik Thesis committee: Prof. dr. ir. B. J. Boersma, TU Delft Prof. dr. S. Hickel, TU Delft Prof. dr. ir. M. J. B. M. Pourquie, TU Delft An electronic version of this thesis is available at http://repository.tudelft.nl/. Abstract Turbulence modelling corresponds to one of the greatest unsolved problems in physics and mathemat- ics. This phenomenon is marked by the emergence of chaotic vortex structures in the solution of the Navier-Stokes equations, and it corresponds to the leading-order effect in the majority of the flows observed in nature. Due to the importance of turbulence modelling, researchers have designed RANS (Reynolds-Averaged Navier Stokes) turbulence models to understand their mean flow behavior. How- ever, one important limitation present in traditional RANS turbulence models is given by their focus on isothermal incompressible fluids, which present constant molecular properties. In order to overcome these limitations, the research previously done at the Process & Energy Department of the TU Delft has established new scaling theories for variable-property flows, which bring the greatest degree of univer- sal collapse observed to date. The application of these scaling theories to traditional RANS turbulence models has greatly improved their performance, although multiple-equation models still require further tuning. The present thesis work thus consisted in applying the principles of Machine Learning (ML) to build improved data-driven RANS turbulence models; using the DNS database assembled by the research group. Special emphasis was placed in an emerging technique called FIML (Field Inversion Machine Learning). This technique splits the process of building ML corrections into an initial data mining stage known as Field Inversion, and a subsequent Machine Learning stage where a predictive system is trained using the corrections found. During the present study, several advances were proposed for the previous methodology. The first contribution is given by the synthesis of a direct Hessian-free Field Inversion optimizer, which con- tinuously seeks the best learning hyper-parameters to use while remaining unconditionally stable. As a result, the Field Inversion optimization process can be fully automatized for any RANS turbulence model. This marked a large contrast with respect to the Field Inversion optimizers employed in leading publications, where probabilistic Bayesian inversion algorithms are used. These algorithms can result unnecessarily complex, since they perform a side-study of the probabilities associated to obtaining certain parameters, whereas the FIML approach only requires the final (or average) parameter distri- butions chosen. Beyond the previous discussion, performing a rigorous probabilistic analysis of a Field Inversion problem requires an expensive technique called MCMC (Markov-Chain Monte Carlo) sampling, which results intractable in large-scale systems. As a result, the probabilistic analysis performed by standard Field Inversion optimizers is based on strong Gaussian assumptions, whose effect cannot be immediately quantified. The overall Field Inversion methodology developed during the present work was further enhanced by creating a Sympy script to derive the algebraic expressions required to perform Field Inversion, which can be imported into any programming language. Thanks to this methodology, the running times registered in Python were improved by a factor of x1000 compared to the use of Automatic Dif- ferentiation libraries. Additionally, a strategy was created to build explicit source term corrections (휹) for any RANS turbulence model without compromising the numerical stability of the optimizers. This in- novation allowed Field Inversion optimizers to perform an unbiased analysis of any governing equation. Regarding the Neural Network systems built, the theoretical review performed revealed that loga- rithmic neurons could be added to the first layer of a Neural Network, such that the optimizer was able to identify the best parameter groups available. Important parameters can be found in this category, such as the Reynolds number, the Prandtl number, or any intermediate combinations (e.g., 푅푒푃푟). As a result, it was possible to build Neural Networks which could automatically discover adequate input features for the subsequent hidden layers, and the overall modelling process was greatly simplified. Furthermore, the present thesis work includes the derivation of a new intelligent relaxation factor methodology, which was created to eliminate any spurious corrections predicted by Neural Networks iii iv Abstract that may cause divergence. The methodology created scales well to large-scale systems, and it was able to perform an intelligent selection of the best corrections to keep in less than a millisecond during the tests performed. The final Field Inversion study centered in the Spalart-Allmaras (SA) and the MK turbulence models, since these models presented the lowest and the highest initial modelling errors respectively. The Field Inversion corrections obtained for the SA turbulence model presented a high degree of over-fitting and inconsistent patterns, which did not form a good basis to train Neural Networks. These issues were at- tributed to the high initial accuracy exhibited by the SA model, which implied that improvements could only be achieved through direct over-fitting. Similar trends were obtained during a secondary study performed for the turbulent Prandtl number (푃푟), which also presented inconsistent trends marginally deviating from 푃푟 ≈ 1. However, the Field Inversion corrections built for the MK turbulence model presented consistent trends, which were suitable for a Machine Learning study. After an exhaustive analysis, it was detected that the predictions made by Neural Networks for the source terms required by the k-equation of the MK turbulence model resulted more stable, and that the use of the intelligent relaxation factor methodology corresponded to an important asset. The FIML study for the MK turbu- lence model thus concluded with satisfactory results. From a global perspective, it was detected that the Field Inversion corrections built for RANS tur- bulence models presented a high degree of variability, which seems to be caused by their numerical properties instead of the actual flow physics. This insight was based on the derivation of a new tur- bulence parameter for rectangular channel flows, which was able to collapse the turbulence statistics of all the DNS flows studied into similar trajectory curves. As a result, it was finally concluded that the FIML methodology developed can be employed to perform successful Machine Learning studies, but also that new turbulence parameters should be developed in order to create more reliable predictive systems in the future. Contents List of Figures vii List of Tables xi Acknowledgements xiii 1 Introduction 1 1.1 Overview . 1 1.2 Thesis Objectives and Contributions . 2 1.3 Thesis Outline . 3 2 Theoretical Background 5 2.1 Fluid Mechanics . 5 2.1.1 Navier-Stokes Equations . 5 2.1.2 Energy Equation . 6 2.1.3 Dimensionless Form of the Governing Equations . 8 2.1.4 Reynolds-Averaged Navier-Stokes Equations . 11 2.1.5 Alternatives to RANS Turbulence Modelling . 13 2.1.6 Boundary Layer Structures. 15 2.1.7 RANS Turbulence Models . 17 2.2 Parameter Optimization . 21 2.2.1 General Paradigm . 21 2.2.2 Gradient-based Optimization. 22 2.2.3 Computation of the Jacobian Matrix . 23 2.2.4 Hessian Optimization . 25 2.2.5 Jacobian Optimization Algorithms. 27 2.3 Machine Learning Applied to CFD . 30 2.3.1 Neural Networks . 31 2.3.2 Convolutional Neural Networks . 33 2.3.3 Long Short Term Memory Neural Networks . 34 2.3.4 Logarithmic Neural Networks. 35 2.3.5 Model Selection . 36 2.3.6 Other Machine Learning Techniques . 37 2.3.7 Field Inversion Machine Learning . 39 3 Preliminary Examples 41 3.1 Field Inversion for a 1-D Radiative Heat Transfer Problem . 41 3.2 Recovery of a Laminar Viscosity Profile . 44 3.3 Field Inversion Optimization for the 푘-휔 Turbulence Model . 46 3.4 Deep Dreaming the Existence of the Reynolds Number . 49 4 Data-Driven RANS Turbulence Modelling for Variable Property Flows 53 4.1 RANS Equations Applied to Channel Flows . 53 4.2 DNS Database Assembled. 54 4.2.1 Performance of Traditional RANS Turbulence Models in the Dataset . 56 4.3 Field Inversion Optimization . 58 4.3.1 Field Inversion Formulation for the MK Turbulence Model . 58 4.3.2 Field Inversion Optimization for the Spalart-Allmaras Turbulence Model . 60 4.3.3 Field Inversion Study for the Turbulent Prandtl Number . 61 v vi Contents 4.4 Field Inversion Optimization Results . 62 4.4.1 MK Turbulence Model . 62 4.4.2 Field Inversion Results for the SA turbulence model . 70 4.4.3 Field Inversion Results for the Turbulent Prandtl Number . 73 4.5 Machine Learning Systems . 75 4.5.1 Derivation of an Intelligent Relaxation Factor . 75 4.5.2 Neural Network Architectures . 78 4.6 Final Remarks . 85 5 Alternative Machine Learning Formulation for Turbulent Channel Flows 87 5.1 Mathematical Derivation of a New Turbulence Parameter (Φ). 87 5.2 Machine Learning Predictive System . 90 5.3 Final Remarks . 94 6 Conclusions and Recommendations 95 6.1 Summary and Conclusions . 95 6.2 Recommendations . 97 A Source Code for the Examples Presented in Chapter 3 99 A.1 Field Inversion Optimization of a 1-D Radiative Heat Transfer Profile . 99 A.1.1 Version A: Build 휷 multipliers for 푃 (Inefficient) . 99 A.1.2 Version B: Build Explicit 휹 Source Term Corrections (Efficient) . .104 A.2 Recovery of a Dynamic Viscosity Profile . .109 A.3 Field Inversion Optimization for the 푘 − 휔 Turbulence Model .
Recommended publications
  • Multiscale Computational Fluid Dynamics
    energies Review Multiscale Computational Fluid Dynamics Dimitris Drikakis 1,*, Michael Frank 2 and Gavin Tabor 3 1 Defence and Security Research Institute, University of Nicosia, Nicosia CY-2417, Cyprus 2 Department of Mechanical and Aerospace Engineering, University of Strathclyde, Glasgow G1 1UX, UK 3 CEMPS, University of Exeter, Harrison Building, North Park Road, Exeter EX4 4QF, UK * Correspondence: [email protected] Received: 3 July 2019; Accepted: 16 August 2019; Published: 25 August 2019 Abstract: Computational Fluid Dynamics (CFD) has numerous applications in the field of energy research, in modelling the basic physics of combustion, multiphase flow and heat transfer; and in the simulation of mechanical devices such as turbines, wind wave and tidal devices, and other devices for energy generation. With the constant increase in available computing power, the fidelity and accuracy of CFD simulations have constantly improved, and the technique is now an integral part of research and development. In the past few years, the development of multiscale methods has emerged as a topic of intensive research. The variable scales may be associated with scales of turbulence, or other physical processes which operate across a range of different scales, and often lead to spatial and temporal scales crossing the boundaries of continuum and molecular mechanics. In this paper, we present a short review of multiscale CFD frameworks with potential applications to energy problems. Keywords: multiscale; CFD; energy; turbulence; continuum fluids; molecular fluids; heat transfer 1. Introduction Almost all engineered objects are immersed in either air or water (or both), or make use of some working fluid in their operation.
    [Show full text]
  • Deep Learning Models for Turbulent Shear Flow
    DEGREE PROJECT IN MATHEMATICS, SECOND CYCLE, 30 CREDITS STOCKHOLM, SWEDEN 2018 Deep Learning models for turbulent shear flow PREM ANAND ALATHUR SRINIVASAN KTH ROYAL INSTITUTE OF TECHNOLOGY SCHOOL OF ENGINEERING SCIENCES Deep Learning models for turbulent shear flow PREM ANAND ALATHUR SRINIVASAN Degree Projects in Scientific Computing (30 ECTS credits) Degree Programme in Computer Simulations for Science and Engineering (120 credits) KTH Royal Institute of Technology year 2018 Supervisors at KTH: Ricardo Vinuesa, Philipp Schlatter, Hossein Azizpour Examiner at KTH: Michael Hanke TRITA-SCI-GRU 2018:236 MAT-E 2018:44 Royal Institute of Technology School of Engineering Sciences KTH SCI SE-100 44 Stockholm, Sweden URL: www.kth.se/sci iii Abstract Deep neural networks trained with spatio-temporal evolution of a dy- namical system may be regarded as an empirical alternative to conven- tional models using differential equations. In this thesis, such deep learning models are constructed for the problem of turbulent shear flow. However, as a first step, this modeling is restricted to a simpli- fied low-dimensional representation of turbulence physics. The train- ing datasets for the neural networks are obtained from a 9-dimensional model using Fourier modes proposed by Moehlis, Faisst, and Eckhardt [29] for sinusoidal shear flow. These modes were appropriately chosen to capture the turbulent structures in the near-wall region. The time series of the amplitudes of these modes fully describe the evolution of flow. Trained deep learning models are employed to predict these time series based on a short input seed. Two fundamentally different neural network architectures, namely multilayer perceptrons (MLP) and long short-term memory (LSTM) networks are quantitatively compared in this work.
    [Show full text]
  • Development of a Wake Model for Wind Farms Based on an Open Source CFD Solver
    Development of a wake model for wind farms based on an open source CFD solver. Strategies on parabolization and turbulence modeling PhD Thesis Daniel Cabezón Martínez Departamento de Ingeniería Energética y Fluidomecánica, Escuela Técnica Superior de Ingenieros Industriales, Universidad Politécnica de Madrid (UPM), Madrid, Spain Abstract Wake effect represents one of the most important aspects to be analyzed at the engineering phase of every wind farm since it supposes an important power deficit and an increase of turbulence levels with the consequent decrease of the lifetime. It depends on the wind farm design, wind turbine type and the atmospheric conditions prevailing at the site. Traditionally industry has used analytical models, quick and robust, which allow carry out at the preliminary stages wind farm engineering in a flexible way. However, new models based on Computational Fluid Dynamics (CFD) are needed. These models must increase the accuracy of the output variables avoiding at the same time an increase in the computational time. Among them, the elliptic models based on the actuator disk technique have reached an extended use during the last years. These models present three important problems in case of being used by default for the solution of large wind farms: the estimation of the reference wind speed upstream of each rotor disk, turbulence modeling and computational time. In order to minimize the consequence of these problems, this PhD Thesis proposes solutions implemented under the open source CFD solver OpenFOAM and adapted for each type of site: a correction on the reference wind speed for the general elliptic models, the semi-parabollic model for large offshore wind farms and the hybrid model for wind farms in complex terrain.
    [Show full text]
  • Simulation of Turbulent Flows
    Simulation of Turbulent Flows • From the Navier-Stokes to the RANS equations • Turbulence modeling • k-ε model(s) • Near-wall turbulence modeling • Examples and guidelines ME469B/3/GI 1 Navier-Stokes equations The Navier-Stokes equations (for an incompressible fluid) in an adimensional form contain one parameter: the Reynolds number: Re = ρ Vref Lref / µ it measures the relative importance of convection and diffusion mechanisms What happens when we increase the Reynolds number? ME469B/3/GI 2 Reynolds Number Effect 350K < Re Turbulent Separation Chaotic 200 < Re < 350K Laminar Separation/Turbulent Wake Periodic 40 < Re < 200 Laminar Separated Periodic 5 < Re < 40 Laminar Separated Steady Re < 5 Laminar Attached Steady Re Experimental ME469B/3/GI Observations 3 Laminar vs. Turbulent Flow Laminar Flow Turbulent Flow The flow is dominated by the The flow is dominated by the object shape and dimension object shape and dimension (large scale) (large scale) and by the motion and evolution of small eddies (small scales) Easy to compute Challenging to compute ME469B/3/GI 4 Why turbulent flows are challenging? Unsteady aperiodic motion Fluid properties exhibit random spatial variations (3D) Strong dependence from initial conditions Contain a wide range of scales (eddies) The implication is that the turbulent simulation MUST be always three-dimensional, time accurate with extremely fine grids ME469B/3/GI 5 Direct Numerical Simulation The objective is to solve the time-dependent NS equations resolving ALL the scale (eddies) for a sufficient time
    [Show full text]
  • An Introduction to Turbulence Modeling for CFD
    An Introduction to Turbulence Modeling for CFD Gerald Recktenwald∗ February 19, 2020 ∗Mechanical and Materials Engineering Department Portland State University, Portland, Oregon, [email protected] Turbulence is a Hard Problem • Unsteady • Many length scales • Energy transfer between scales: ! Large eddies break up into small eddies • Steep gradients near the wall ME 4/548: Turbulence Modeling page 1 Engineering Model: Flow is \Steady-in-the-Mean" (1) 2 1.8 1.6 1.4 1.2 u 1 0.8 0.6 0.4 0.2 0 0 0.5 1 1.5 2 t ME 4/548: Turbulence Modeling page 2 Engineering Model: Flow is \Steady-in-the-Mean" (2) Reality: Turbulent flows are unsteady: fluctuations at a point are caused by convection of eddies of many sizes. As eddies move through the flow the velocity field changes in complex ways at a fixed point in space. Turbulent flows have structures { blobs of fluid that move and then break up. ME 4/548: Turbulence Modeling page 3 Engineering Model: Flow is \Steady-in-the-Mean" (3) Engineering Model: When measured with a \slow" sensor (e.g. Pitot tube) the velocity at a point is apparently steady. For basic engineering analysis we treat flow variables (velocity components, pressure, temperature) as time averages (or ensemble averages). These averages are steady (ignorning ensemble averaging of periodic flows). ME 4/548: Turbulence Modeling page 4 Engineering Model: Enhanced Transport (1) Turbulent eddies enhance mixing. • Transport in turbulent flow is much greater than in laminar flow: e.g. pollutants spread more rapidly in a turbulent flow than a laminar flow • As a result of enhanced local transport, mean profiles tend to be more uniform except near walls.
    [Show full text]
  • Modeling and Simulation of High-Speed Wake Flows
    Modeling and Simulation of High-Speed Wake Flows A DISSERTATION SUBMITTED TO THE FACULTY OF THE GRADUATE SCHOOL OF THE UNIVERSITY OF MINNESOTA BY Michael Daniel Barnhardt IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY Graham V. Candler, Adviser August 2009 c Michael Daniel Barnhardt August 2009 Acknowledgments The very existence of this dissertation is a testament to the support and en- couragement I’ve received from so many individuals. Foremost among these is my adviser, Professor Graham Candler, for motivating this work, and for his consider- able patience, enthusiasm, and guidance throughout the process. Similarly, I must thank Professor Ellen Longmire for allowing me to tinker around late nights in her lab as an undergraduate; without those experiences, I would’ve never been stimulated to attend graduate school in the first place. I owe a particularly large debt to Drs. Pramod Subbareddy, Michael Wright, and Ioannis Nompelis for never cutting me any slack when I did something stupid and always pushing me to think bigger and better. Our countless hours of discussions provided a great deal of insight into all aspects of CFD and buoyed me through the many difficult times when codes wouldn’t compile, simulations crashed, and everything generally seemed to be broken. I also have to thank Travis Drayna and all of my other friends and colleagues at the University of Minnesota and NASA Ames. Dr. Matt MacLean of CUBRC and Dr. Anita Sengupta at JPL provided much of the experimental results and were very helpful throughout the evolution of the project.
    [Show full text]
  • Turbulence: V&V and UQ Analysis of a Multi-Scale Complex System Presented by Parviz Moin, Curtis W
    Turbulence: V&V and UQ Analysis of a Multi-scale Complex System Presented by Parviz Moin, Curtis W. Hamman and Gianluca Iaccarino 1 Introduction Turbulent motions of liquid and gases are ubiquitous and impact almost every aspect of our life, from the formation of hurricanes to the mixing of a cappuccino [1, 2, 3, 4, 5]. Nearly all human endeavors must contend with turbulent transport: turbulence is the rule, not the exception, in fluid dynamics. Energy, transportation, and the environment function on length and time scales where turbulence rapidly develops: even a simple, slow stroll leaves behind a turbulent wake with a wide range of scales (see table 2 for examples). With the advent of faster computers, numerical simulation of turbulent flows is becoming more prac- tical and more common [6, 7]. In this short note, we describe the fundamental physics and numerics needed for accurate turbulent flow simulations. Several illustrative examples of such simulations are presented. In addition, we show how simulations of turbulent flows contribute to breakthroughs in clean energy technologies, reduce dependence on volatile oil reserves and develop carbon-free sources of energy [8]. We also show how increased industrial competitiveness in high-tech sectors such as transportation and aerospace is made possible by high-performance computing, especially where physical experiments are impossible, dan- gerous or inordinately costly to perform [8]. 1.1 Aircraft and jet engines Aerospace industry was one of the early pioneers to apply computational fluid dynamics (CFD) to the design process. The primary motivation has always been to reduce expensive physical testing (e.g., wind tunnel tests) and the number of prototypes built during the design cycle.
    [Show full text]
  • Anisotropic RANS Turbulence Modeling for Wakes in an Active Ocean Environment †
    fluids Article Anisotropic RANS Turbulence Modeling for Wakes in an Active Ocean Environment † Dylan Wall ∗,‡,§ and Eric Paterson ‡,§ Department of Aerospace and Ocean Engineering, Virginia Polytechnic Institute and State University, Blacksburg, VA 24061, USA; [email protected] * Correspondence: [email protected]; Tel.: +1-865-816-0632 † This paper is an extended version of our paper published in 15th OpenFOAM Workshop. ‡ Current address: Randolph Hall, RM 215, 460 Old Turner Street, Blacksburg, VA 24061, USA. § These authors contributed equally to this work. Received: 14 November 2020; Accepted: 14 December 2020; Published: 18 December 2020 Abstract: The problem of simulating wakes in a stratified oceanic environment with active background turbulence is considered. Anisotropic RANS turbulence models are tested against laboratory and eddy-resolving models of the problem. An important aspect of our work is to acknowledge that the environment is not quiescent; therefore, additional sources are included in the models to provide a non-zero background turbulence. The RANS models are found to reproduce some key features from the eddy-resolving and laboratory descriptions of the problem. Tests using the freestream sources show the intuitive result that background turbulence causes more rapid wake growth and decay. Keywords: stratified wakes; turbulence; RANS; stress-transport 1. Introduction Oceanographic flows include a broad variety of turbulence-generating phenomena, and the associated unsteady motions are in general inhomogeneous, non-stationary, and anisotropic. The thermohaline stratification of the ocean introduces a conservative body force which must be considered when examining flows in such an environment. Numerous effects including buoyancy, shear, near-free-surface damping, bubbles, and Langmuir circulations complicate any attempt to describe turbulent motions.
    [Show full text]
  • An Overview of Turbulence Modeling
    An overview of turbulence modeling P.A. Durbin1 & T.I-P. Shih2 1Department of Mechanical Engineering, Stanford University, Stanford, California, USA 2Department of Aerospace Engineering, Iowa State University Ames, Iowa, USA Abstract Despite its great importance and the tremendous efforts that have been made to understand it, turbulence remains a thorny problem. Though great strides have been made and a plethora of predictive turbulence models have been developed, there is only a modicum of guidance available on what is practical and what can be relied upon for design and analysis. This chapter seeks to provide an introduction to turbulence modeling and to make some sense of it. First, the physics of turbulence are described briefly. Then, different mathematical approaches used to predict turbulent flows are outlined, with focus on their essence, their ability to predict correctly, and their turn-around time. Next, widely used, single-point closure models for Reynolds-averaged Navier–Stokes equations are summarized, with some reference to the principles used to develop them. Finally, an appraisal is given of the current state-of-art models, in order to provide guidelines on their usefulness. The chapter concludes with comments on transition prediction. Nomenclature bij anisotropy tensor Cp constant pressure specific heat d shortest distance to the wall k turbulent kinetic energy or thermal conductivity P rate of turbulent kinetic energy production WIT Transactions on State of the Art in Science and Engineering, Vol 15, © 2005 WIT Press
    [Show full text]
  • A Deep Learning Framework for Turbulence Modeling Using Data Assimilation and Feature Extraction
    A deep learning framework for turbulence modeling using data assimilation and feature extraction Atieh Alizadeh Moghaddam1 & Amir Sadaghiyani2 Abstract Turbulent problems in industrial applications are predominantly solved using Reynolds Averaged Navier Stokes (RANS) turbulence models. The accuracy of the RANS models is limited due to closure assumptions that induce uncertainty into the RANS modeling. We propose the use of deep learning algorithms via convolution neural networks along with data from direct numerical simulations to extract the optimal set of features that explain the evolution of turbulent flow statistics. Statistical tests are used to determine the correlation of these features with the variation in the quantities of interest that are to be predicted. These features are then used to develop improved partial differential equations that can replace classical Reynolds Averaged Navier Stokes models and show improvement in the accuracy of the predictions. I. Introduction, motivation and objectives Although today there are many precise methods for modeling problems in turbulent flow by computational fluid dynamic tools such as Large Eddy Simulation, Direct Numerical Simulation, etc., considering their high computational cost and difficulties, they cannot be used widely for industrial purposes. Currently the predominant tool for engineering problems with turbulent flows are Reynolds Averaged Navier Stokes (RANS) models. The accuracy of the RANS models is limited due to closure hypothesis that induce uncertainty into the RANS models. RANS model predictions are known to be unreliable for flows with strong pressure gradient, streamline curvature or flow separation. This includes both structural uncertainty and parametric uncertainty. These closure assumptions are used for predicting the Reynolds stresses by introducing functional dependency between mean flow properties and turbulent quantities that could be different in various turbulent problems.
    [Show full text]
  • Eddy Viscosity Turbulence Models Employed by Computational Fluid Dynamic
    Eddy Viscosity Turbulence Models employed by Computational Fluid Dynamic Ï WOELKE General Electric Company Polska ABSTRACT This paper is discussing the advantages and disadvantages of the two-equation eddy- viscosity turbulence models employed to carry out computational fluid dynamic analyses. Simulating turbulence by means of numerical methods is one of the most critical problems in modeling fluid flow. The most popular, two-equation eddy-viscosity turbulence models are reviewed in this paper. These models rely on the eddy-viscosity, which according to Boussinesq approach relates Reynolds stress with mean velocity gradients. The consequences of such approach for two- ǯ abilities in predicting fluid motionH are analyzed. Moreover, the current work discusses: (i) Approaches taken to obviate theZ shortcomings of the Boussinesq assumption; (ii) Performance of the k- model and some of its modifications in predicting fluid Hstructure;Z (iii) Advantages offered by the k- model in simulating some class of the flows; (iv) The Shear Stress Transport (SST) model benefiting from the features of the k- and k- models. This paper discusses the vital problems related to simulating turbulence flows by 1. INTRODUCTION means of CFD codes. International industry develops very fast, imposing high demands on the technical ǯ Ǥ the computer technology enables numerical modeling of the most complicated phenomenon described by differential equations. Simulating fluid motions described by Navier-Stokes equation is one of the most challenging problems in aerodynamics and has therefore received a lot of attention from the researchers and engineers. Industry demands more accurate computer codes, which would enable numerical testing of new applications prior to the experimental investigation in order to reduce time required for developing new products and improving their performance.
    [Show full text]
  • Turbulence Closure Models for Computational Fluid Dynamics Paul A
    Aerospace Engineering Publications Aerospace Engineering 2-23-2018 Turbulence Closure Models for Computational Fluid Dynamics Paul A. Durbin Iowa State University, [email protected] Follow this and additional works at: https://lib.dr.iastate.edu/aere_pubs Part of the Aerodynamics and Fluid Mechanics Commons, and the Computational Engineering Commons The ompc lete bibliographic information for this item can be found at https://lib.dr.iastate.edu/ aere_pubs/136. For information on how to cite this item, please visit http://lib.dr.iastate.edu/ howtocite.html. This Book Chapter is brought to you for free and open access by the Aerospace Engineering at Iowa State University Digital Repository. It has been accepted for inclusion in Aerospace Engineering Publications by an authorized administrator of Iowa State University Digital Repository. For more information, please contact [email protected]. Turbulence Closure Models for Computational Fluid Dynamics Abstract The formulation of analytical turbulence closure models for use in computational fluid mechanics is described. The ubjs ect of this chapter is the types of models that are used to predict the statistically averaged flow field – commonly known as Reynolds‐averaged Navier–Stokes equations (RANS) modeling. A variety of models are reviewed: these include two‐equation, eddy viscosity transport, and second moment closures. How they are formulated is not the main theme; it enters the discussion but the models are presented largely at an operational level. Several issues related to numerical implementation are discussed. Relative merits of the various formulations are commented on. Keywords turbulence modeling, RANS, k ‐epsilon model, Reynolds stress transport, second moment closure, turbulence prediction methods Disciplines Aerodynamics and Fluid Mechanics | Aerospace Engineering | Computational Engineering Comments This is a chapter from Durbin, Paul A.
    [Show full text]