CTR Seminars Archive 2021
Simulating flow field around propeller blades, and other examples
Date and Time: Friday, December 3, 2021 - 16:15
Location: Building 300, Room 300
Event Sponsor: Parviz Moin, Director of Center for Turbulence Research
Speaker(s): Prof. Jelena Svorcan, Associate Professor at the Department of Aeronautics of the University of Belgrade, Faculty of Mechanical Engineering
Rotors form an integral part of many flow machines, e.g. propellers, helicopters, wind turbines, turbo-machinery, etc. Although many computational models (differing in complexity and starting assumptions) for the estimation of their aerodynamic properties exist, accurately simulating flows around rotors still presents a challenge (due to unsteadiness, turbulence, flow separation, and other flow phenomena). This presentation will focus on some of the commonly most employed computational approaches and present some of the most characteristic quantitative and qualitative results obtained for a small-scale rotor designed at the University of Belgrade, Faculty of Mechanical Engineering including a background story about the blade design. Most important conclusions and recommendations shall be provided. Also, a brief summary of the latest work performed at the Department of Aeronautics shall be given.
Speaker Bio:
Dr. Jelena Svorcan is an Associate Professor at the Department of Aeronautics of the University of Belgrade, Faculty of Mechanical Engineering (UB-FME) since 2020. Currently, she is associated to Stanford University, Center for Turbulence Research through a Fulbright grant (till August 2022). She received her MSc and PhD in 2010 and 2014, respectively, at UB-FME where she has been employed since 2011. Her main research interests are focused to contemporary engineering problems in aeronautics and include topics from: computational aerodynamics, turbulence, rotors, renewable energy, aircraft design and optimization.
--------------------------------
Energy stable numerical schemes for simulating two-phase flows by solving Cahn-Hilliard Navier Stokes equations on adaptive octree meshes
Date and Time: Friday, November 5, 2021 - 16:15
Location: Building 300, Room 300
Event Sponsor: Parviz Moin, Director of Center for Turbulence Research
Speaker(s): Dr. Makrand Khanwale, Postdoctoral Fellow at the Center for Turbulence Research at Stanford University
Developing accurate, stable, and thermodynamically consistent numerical methods to simulate two-phase flows is critical for many applications. We develop numerical methods to solve thermodynamically consistent Cahn-Hilliard Navier-Stokes equations to simulate two-phase flows with deforming interfaces at various density contrasts. We develop three essentially unconditionally energy-stable time integration schemes. The first two time-integration schemes are fully implicit based on the pressure-stabilization technique. The third approach utilizes the projection method to decouple the pressure to improve the efficiency of the fully implicit scheme. We rigorously prove the energy stability of the time-discrete numerical schemes for the approaches with pressure-stabilization approaches. We also prove the existence of solutions of the advective-diffusive Cahn-Hilliard operator. We use a conforming continuous Galerkin (CG) finite element method in space equipped with a residual-based variational multiscale (RBVMS) procedure to stabilize the pressure in the first two approaches. In the third approach, we present a projection based framework extending the fully implicit method to a block iterative, hybrid semi-implicit-fully-implicit in time method. We use a semi-implicit time discretization for Navier-Stokes and a fully-implicit time discretization for Cahn-Hilliard equations. Pressure is decoupled using a projection step resulting in two linear positive semi-definite systems for velocity and pressure instead of the saddle point system of a pressure-stabilized method. All the resulting linear systems are solved using the efficient and scalable algebraic multigrid (AMG) method. We deploy all three approaches on a massively parallel numerical implementation using fast octree-based adaptive meshes in a computational framework called "Proteus". We perform a detailed scaling analysis of all three solvers. A comprehensive set of numerical experiments showing detailed comparisons with results from the literature for canonical cases are used to validate the methods for an extensive range of density ratios. This presentation is an overview of the main developments of Khanwale's PhD research. Khanwale will also discuss his plans to push the boundaries of phase-field methods and develop robust multiphysics solvers which couple scalar transport, electroconvection with multiphase flows.
Speaker Bio:
Dr. Makrand Ajay Khanwale is currently a Postdoctoral Fellow at the Center for Turbulence Research at Stanford University. He received his PhD in the Summer of 2021 from Iowa State University (ISU), co-majoring in Mechanical Engineering and Applied Mathematics. At ISU, he was co-advised by Dr. Baskar Ganapathysubramanian and Dr. James Rossmanith. For his dissertation, he worked on developing and analysing numerical schemes for high fidelity simulations of multiphase flows. Specifically, he developed energy stable numerical methods to simulate two-phase flows using Cahn-Hilliard Navier-Stokes equations. Before joining Iowa State for his graduate work, he had a brief stint as a research associate in Dr. Krishnaswamy Nandakumar‘s group at Louisiana State University (LSU). At LSU, he worked on developing closure models for energy cascades in multiphase flows. He received his Bachelors in Chemical Technology from the Institute of Chemical Technology, Mumbai.
----------------------------
Near-wall filtering and splitting the dependent variables in Large Eddy Simulations
Date and Time: Friday, October 22, 2021 - 16:15
Location: Building 300, Room 300
Event Sponsor: Parviz Moin, Director of Center for Turbulence Research
Speaker(s): Dr. Nagi N. Mansour
A new formulation for near-wall filtering the Navier-Stokes equations without commutation errors will be detailed in the presentation. In addition, a new split of the dependent variables into large scales and sub-filter scales is formulated where the filtered large scales are not affected by further filtering (i.e. the filtered “filtered scales” are the “filtered scales”), thus removing the Leonard terms from the filtered equations. Finally, closure models for the near-wall effects on the large scales are proposed.
Speaker Bio:
Dr. Nagi N. Mansour earned his Ph.D. in Mechanical Engineering from Stanford University in 1978 where he pioneered one of the earliest Large-Eddy Simulations of a turbulent mixing layer. Dr. Mansour pioneered the use of direct numerical simulation for turbulence model development, as well as the use of boundary integral methods for studying drop formation in surface-tension driven flows. He is also a pioneer in the development of high-fidelity models for modeling material response to high-enthalpy flows. Dr. Mansour has served in technical management positions including lead of the Physics Simulation and Modeling Office, chief of the Reacting Flow Environments Branch, Chief of the Computational Physics Branch, and as deputy director of the Stanford Center for Turbulence Research, operated jointly with NASA. He is a fellow of the American Physical Society (APS, and associate fellow of the American Institute of Aeronautics and Astronautics (AIAA). Dr. Mansour’s current research interests include the development of high-fidelity models for Thermal Protection Systems used for spacecraft atmospheric entry, and development of realistic models of plasma with coupled radiative transfer and magneto-hydrodynamics effects on exascale computing systems; large eddy simulations of the solar convection zone, modeling sound propagation in plasma; and the development of emerging magnetic flux models.
-----------------------------
Coupling CFD with detailed chemistry for heterogeneous catalysis: application to methane valorization
Date and Time: Friday, September 24, 2021 - 16:15
Location: Bldg 530, Rm 127
Event Sponsor: Parviz Moin, Director of Center for Turbulence Research
Speaker(s): Dr. Laurien Vandewalle, Postdoctoral Fellow in the Mechanical Engineering Department (NanoEnergy Lab) at Stanford University
Many innovative catalytic technologies have been developed in the past decade as a response to the world’s rapidly growing demand for a more efficient and sustainable exploitation of energy and material resources. An example is oxidative coupling of methane (OCM), which is considered one of the most promising processes to valorize methane directly into ethylene. The performance of a heterogeneous catalytic process is the result of a complex interaction of phenomena at very different time and length scales. Even in the simplest reactor configurations (i.e., a packed bed reactor), conversion and selectivities are affected by transport phenomena at the pellet scale as well as concentration, temperature, pressure and velocity gradients at the reactor scale. Fundamental multiscale modeling of catalytic processes is the key to obtain a better understanding of catalytic systems, improve existing technologies and develop novel reactor concepts. Computational fluid dynamics (CFD) is hereby needed to predict flow fields and transport phenomena, while the use of microkinetic models for both gas and surface chemistry allows an accurate description of each elementary step at the microscale.
Nowadays, reactive CFD studies with detailed chemistry are still mostly limited to combustion applications. The coupling of CFD with detailed kinetic models for heterogeneous catalysis is only a recently emerging research field, but one with an enormous potential, both scientifically and economically. During Dr. Vandewalle’s PhD, she developed catchyFOAM (CATalytic CHemistrY FOAM), an OpenFOAM-based CFD code, targeted at Euler-Euler simulations of catalytic fluidized bed reactors using detailed microkinetic mechanisms for both the gas phase and catalytic surface chemistry. The catchyFOAM framework was then used to study OCM in a gas-solid vortex reactor. In this talk, she will give an overview of the main outcomes of her PhD research, as well as discuss her future plans for particle-resolved modeling of both packed and fluidized beds.
Speaker Bio:
Dr. Laurien Vandewalle is currently a Postdoctoral Fellow in the Mechanical Engineering Department (NanoEnergy Lab) at Stanford University. She obtained her BSc (2012) and MSc (2014) degree in Chemical Engineering from Ghent University in Belgium. After her studies, she worked as a Process Engineer in industry for one year and then started her PhD at the Laboratory for Chemical Technology (LCT) at Ghent University, under the supervision of Prof. Guy Marin and Prof. Kevin Van Geem. During Vandewalle’s PhD she focused on coupling CFD with detailed chemistry for heterogeneous catalysis, specifically for process intensification of oxidative coupling of methane in a gas-solid vortex reactor. She defended her PhD in May 2020 and worked as a Postdoctoral Assistant at the LCT. In July 2021, she started a Postdoctoral Fellowship at Stanford University, funded by the Belgian American Educational Foundation.
------------------------------------
Input-output analysis of receptivity and instability of hypersonic boundary layers
Date and Time: Friday, May 21, 2021 - 16:30
Location: Zoom
Event Sponsor: Parviz Moin, Director of Center for Turbulence Research
Speaker(s): Prof. Joseph Nichols, University of Minnesota
Spatial amplification owing to modal instability plays a significant role in determining when and how a hypersonic boundary layer transitions from a laminar state to turbulence. Traditional stability analysis methods rely on the strong assumption of a slowly-varying baseflow, which limits their predictive power. Such methods can calculate downstream amplification only relative to upstream points where the flow has already passed through shock waves (such as a vehicle’s bow shock) and that are far away from complex geometry (such as a blunt nose tip). The estimation of perturbation amplitudes at these upstream points from freestream disturbance levels then requires empirical formulae, calibrated for different geometries on a case-by-case basis. As a further complication, hypersonic boundary layers also support non-modal amplification mechanisms not captured by traditional stability analysis.
We instead investigate this problem using Input/Output (I/O) analysis which directly relates freestream perturbations to the total system response through the resolvent operator, eliminating the need for a slowly-varying base flow. As a global method, I/O analysis simultaneously incorporates receptivity, modal amplification, and non-modal amplification mechanisms. Furthermore, as a decomposition method, I/O analysis provides a technique to distinguish between different mechanisms competing in a single flow. Crucial to the accurate representation of receptivity, our method employs a linear model for the transmission and reflection of small perturbations through a shock wave. We validate our method through comparison to schlieren measurements of hypersonic instabilities over ogive-cylinders taken in the AFRL Mach-6 Ludwieg Tube. In addition to the expected Mack 2nd mode instability, I/O analysis predicts a new type of modal instability, as well as non-modal entropy-layer instability, in good agreement with experimental observations. The interaction between these different mechanisms may explain the phenomenon of “transition reversal” with increasing nose-tip bluntness.
Speaker Bio:
Professor Joseph Nichols’ current interests are in the areas of stability and sensitivity analysis of hypersonic flows and the aeroacoustics of high-speed jets. He has performed some of the largest computational fluid dynamics simulations in the world, involving massively parallel high-fidelity Large Eddy Simulation of tactical aircraft engine exhausts, running on more than a million computer processors simultaneously. He is currently developing novel and scalable stability and transition analysis tools for hypersonic flow over complex geometry. Before joining the faculty at the University of Minnesota, he held postdoctoral research positions at the Laboratory d’Hydrodynamique (LadHyX) at the École Polytechnique in France, and at the Center for Turbulence Research (CTR) at Stanford University. At the University of Minnesota, Nichols teaches a number of courses in Fluid Mechanics, Computational Fluid Mechanics, and Hydrodynamic Stability Analysis.
------------------
Machine Learning for Turbulence
Date and Time: Friday, May 7, 2021 - 16:30
Location: Zoom
Event Sponsor: Parviz Moin, Director of Center for Turbulence Research
Speaker(s): Dr. Daniel Livescu
This talk summarizes part of the work performed during our 3-year Laboratory Directed Research and Development - Directed Research (LDRD-DR) project, titled MachinE Learning for Turbulence (MELT). Started in October 2018, the project partially covered ~10 staff members, 7 postdocs, several more summer students, and addressed a diverse set of topics related to turbulence and applications in climate and astrophysics. Today, in order to highlight the exploratory aspect of such projects, I will survey some of our results on a) neural network (NN) models of scalar turbulence, b) learning Lagrangian dynamics, c) Mori-Swanzig structure of turbulence and, finally, d) unsupervised identification of dynamical regimes.
Link to video file:
Machine Learning for Turbulence
Speaker Bio:
Dr. Daniel Livescu has been a scientist at Los Alamos National Laboratory since he received his Ph.D. in 2001 and, currently, is leading the fluid dynamics team within the CCS Division and is the PI for OE/NNSA Office of Experimental Sciences program on DNS. His research interests are in the general areas of theoretical and computational fluid mechanics, with emphasis on turbulence and turbulent mixing simulation, theory, and modeling. Dr. Livescu is a Fellow ASME, Associate Fellow AIAA, and the recipient of the 2017 (inaugural) Frank Harlow Distinguished Mentor Award. He also serves as Associate Editor for AIAA Journal, ASME Journal of Fluids Engineering, and has recently joined the Editorial Committee of the Annual Review of Fluid Mechanics.
----------------------------
Modeling Turbulent Liquid/Gas Phase Interfaces using a dual-scale Large Eddy Simulation approach
Date and Time: Friday, April 23, 2021 - 16:30
Location: Zoom
Event Sponsor: Parviz Moin, Director of Center for Turbulence Research
Speaker(s): Prof. Marcus Herrmann, Professor in the School for Engineering of Matter, Transport and Energy at Arizona State University.
While significant progress has been made in the past decade to predict atomization using detailed numerical simulations, these simulations come at significant computational cost since the range of scales that must be resolved typically exceeds those of a single phase turbulent flow significantly. A switch to a Large Eddy Simulation (LES) approach would be desirable, however, the underlying assumption of LES methods that the dynamics of the unresolved sub-filter scale can be inferred from the dynamics of the resolved scales is questionable when atomization occurs. Similar to viscosity in single-phase flows, surface tension forces scale with the inverse of a length-scale, but unlike viscosity, surface tension can act to either dissipate surface corrugations preventing breakup, giving rise to the Hinze scale/Kolmogorov's critical radius, or enhance surface corrugations due to the Rayleigh-Plateau instability, resulting in breakup. Which process is dominant on the sub-filter scale seems to depend entirely on the sub-filter interfacial geometry, i.e., if the interface is in the shape of ligaments, the surface tension can lead to breakup, whereas in other cases, surface-tension forces can inhibit breakup. Unfortunately, the sub-filter geometry cannot be inferred from the filtered interfacial geometry alone. LES approaches going beyond the traditional single-phase cascade hypothesis may be required for two-phase flows with atomization.
In this presentation, a dual scale LES modeling approach will be discussed that can handle the dual nature of surface tension on the sub-filter scale. The model maintains a fully resolved realization of the phase interface, shifting the modeling task to reconstructing the fully resolved interfacial advection velocity from the filter-scale velocity, taking the effects of sub-filter surface tension, turbulent eddies, and shear into account. Results showing the viability of the dual-scale LES approach in canonical test cases will be discussed.
Link to video file:
Modeling Turbulent Liquid/Gas Phase Interfaces using a dual-scale Large Eddy Simulation approach
Speaker Bio:
Marcus Herrmann is a Professor in the School for Engineering of Matter, Transport and Energy at Arizona State University. He received his PhD in Mechanical Engineering from the Technical University in Aachen, Germany in 2001 and was a post-doc and research associate at CTR and CITS at Stanford University from 2002 to 2007. His research is in the area of computational fluid dynamics for turbulent liquid/gas interfacial flows in both incompressible and supersonic flow environments. His specific area of interest is in understanding and predicting the primary atomization processes of injected liquids with applications ranging from fuel injection systems to medical sprays. He currently serves as the Editor-in-Chief for the Americas of the journal Atomization and Sprays.
------------------------
Towards the Best Practice in Wall-modeled Large-eddy Simulation of Heat Transfer Problems
Date and Time: Friday, April 9, 2021 - 16:30
Location: Zoom
Event Sponsor: Parviz Moin, Director of Center for Turbulence Research
Speaker(s): Prof. Xiang Yang, Assistant Professor in the Mechanical Engineering Department at the Pennsylvania State University
The ambitious performance goals set by the aerodynamic as well as the turbomachinery industries call for more accurate, scale-resolving simulation tools. A viable path towards industrial-level scale-resolving simulations of flows at high Reynolds numbers is through wall-modeled large-eddy simulation (WMLES). This talk will discuss the best practice in WMLES of problems in which heat transfer plays a role, covering topics including the gird resolution, the LES/wall-model matching location, the low Mach number limit, and the turbulent Prandtl number. Special attention is given to problems at high Mach numbers. Our results will answer the outstanding question: why our wall model works well in the context of WMLES yet the temperature transformations that give rise to our wall model fail at collapsing data.
Speaker Bio:
Dr. Xiang Yang is an Assistant Professor in the Mechanical Engineering Department at the Pennsylvania State University since 2018. He received his Ph.D. in Mechanical Engineering from Johns Hopkins University in 2016. Yang joined the Center for Turbulence Research (CTR) 2016-2017 as a CTR Postdoctoral Fellow. At Penn State, Yang’s group conducts high-fidelity numerical simulation, builds physics- and data-based models, and finds efficient solutions for real-world engineering problems. His group uses many tools, including direct numerical simulation, large-eddy simulation, Reynolds-averaged Navier Stokes, and more recently, machine learning models.
-----------------------------
Subgrid-scale modeling of turbulent bubble breakup in oceanic breaking waves
Date and Time: Friday, March 5, 2021 - 16:30
Location: Zoom
Event Sponsor: Parviz Moin, Director of Center for Turbulence Research
Speaker(s): Dr. Ronald Chan
Turbulent breaking waves entrain air cavities that break up and coalesce to form polydisperse clouds of bubbles. We recently provided theoretical and numerical justification that the dominant mechanism for super-Hinze-scale bubble generation is a fragmentation cascade from large to small sizes sustained by turbulent velocity fluctuations. This behavior should be universal across various turbulent bubbly flows because of the size locality inherent in a cascade. Universality simplifies the development of subgrid-scale (SGS) breakup models in two-phase large eddy simulations (LES). We formulate an SGS breakup model based on a breakup cascade in accordance with the LES paradigm, where large bubbles are resolved through a two-phase Eulerian description, while small bubbles are separately modeled and tracked as Lagrangian point particles. This model requires the generation and breakup of Lagrangian particles from underresolved Eulerian bubbles with suitable distributions for breakup rates and child bubble sizes. A priori testing and a posteriori implementation of the model in a coarse breaking-wave simulation are discussed, and the need for an accurate modeled breakup mass flux between sizes is underscored. Relations to the dynamic model in single-phase turbulence simulations are also explored, with direct analogy to the central role of the turbulent kinetic energy interscale flux in the single-phase model.
-------------------------
Development of the Dynamic Localization Model – a personal story
Date and Time: Friday, February 19, 2021 - 16:30
Location: Zoom
Event Sponsor: Parviz Moin, Director of Center for Turbulence Research
Speaker(s): Prof. Sandip Ghosal
When I first arrived at CTR as a postdoctoral fellow in 1992, I could feel deep rumblings beneath my feet. I knew that California was famous for its seismic activity but this was of a different kind. It originated from something called the “Dynamic Model”: the greatest idea in turbulence modeling in a long time that “worked” but did not make sense. I would like to talk
about our efforts to make sense of it all and the intellectual atmosphere at CTR in the 1990’s.
Link to video file:
Development of the Dynamic Localization Model – a personal story
Speaker Bio:
Department of Mechanical Engineering and by courtesy Engineering Sciences and Applied Mathematics, Northwestern University
-------------------------
The dynamic model for LES and variations on the theme
Date and Time: Friday, February 5, 2021 - 16:30
Location: Zoom
Event Sponsor: Parviz Moin, Director of Center for Turbulence Research
Speaker(s): Prof. Charles Meneveau, Louis M. Sardella Professor of Mechanical Engineering, and Associate Director, Institute for Data Intensive Engineering and Science, Johns Hopkins University
At 30 years old, the Germano identity-based dynamic model for turbulence simulations can be said to have become a classic result for the field of turbulence. In this presentation, I will recall how several of its variants (e.g. Lagrangian, scale-dependent, multiplicative) came about and summarize the main ideas underlying these variations on an elegant theme.
Link to video file:
The dynamic model for LES and variations on the theme
---------------------------------
At last, Turbulence closure modeling minus tuning parameters!
Date and Time: Friday, January 22, 2021 - 16:30
Location: Zoom
Event Sponsor: Parviz Moin, Director of Center for Turbulence Research
Speaker(s): Prof. Parviz Moin, Franklin P. and Caroline M. Johnson Professor in the School of Engineering and Director, Center for Turbulence Research, Stanford University
In the immediate aftermath of the 1990 Summer Program, there was a flurry of activities at CTR seeking to exploit the dynamic modeling concept in canonical turbulent flows which heretofore had challenged the universality of turbulence models. I will discuss extensions to compressible flow, scalar transport, reacting flows and other potential formulations
of the dynamic modeling concept that were considered.
Link to video file:
At last, Turbulence closure modeling minus tuning parameters!
-----------------------------
A model is born
Date and Time:
Friday, January 15, 2021 - 16:30
Location:
Zoom
Event Sponsor:
Parviz Moin, Director of Center for Turbulence Research
Speaker(s):
Prof. Ugo Piomelli, Canada Research Chair in Turbulence Simulation and Modeling from the Department of Mechanical and Materials Engineering at Queen's University.
It was a sunny morning in July 1990, the first week of the CTR Summer Program, Massimo Germano, Parviz Moin, Bill Cabot and Ugo Piomelli met to discuss the possible uses of an identity that Massimo Germano had recently derived. This talk will chronicle how the events of that month produced one of the most influential subgrid-scale models for large-eddy simulations.
Link to video file:
https://stanford.box.com/s/8x6oamqqc9v68ot9d860wq411sgiyzk4
--------------------