Inicio / ANSYS / Turbulence – What a Drag It Is When You Drive

Turbulence – What a Drag It Is When You Drive

Roopinder Tara posted on December 20, 2016

“We are about to encounter some turbulence.” 

Turbulence is a known troublemaker for air travelers. It can spill your drink, bounce you off the roof of the cabin if you haven’t fastened your seat belt—and even bring a plane down. Airline pilots are super aware of it. But turbulence, the swirly chaotic movement of fluid particles, while most feared and destructive in air travel, is all around us and occurs every time you drive.

Neat streamlines over race car mislead by ignoring the details of turbulence.(Image courtesy of ANSYS.)
Neat streamlines over race car mislead by ignoring the details of turbulence.(Image courtesy of ANSYS.)
In reality, the flow is highly turbulent, more the rule than the exception.(Image courtesy of ANSYS.)

“A car you drive has more turbulent flow than laminar flow,” said Dr. Sandeep Sovani of ANSYS, a self-declared “automotive person.”

Those pictures that computational fluid dynamics (CFD) vendors are fond of showing, with colored streamlines flowing nicely over race cars—they’re misleading. Or at least wishful thinking. In real life, the outside—and inside—of a car is a swirly mess of turbulence.

Analyze This, Einstein

Analyzing something you can’t see is no picnic. Welcome to the world of the aerodynamicist. It’s a tough job. You have to understand what is happening with air on the scale of molecules and apply it to cars, planes, entire buildings—all of it while blind to its forces. To compensate for a lack of vision,they must rely on physics. This actually works well when fluid flow is simple and well behaved with unmoving streamlines. It is an orderliness we crave. But what happens when fluid particles depart from order and fall into chaos, when flow is turbulent rather than laminar? That is an entirely different story.

How hard is it to solve for turbulence? CFD analysts joke that Einstein only resorted to solve time and space after failing to solve for turbulence.

 

Over, Under, Sideways, Down—and Through

Aerospace aerodynamics must be complex, right? After all, an airplane can fly around the speed of sound, where air must be treated as a compressible fluid. Also, with jet engines, fluid flow is complicated with rotation. Then there’s heat transfer, both inside engines and outside with skin friction.

But flow analysis on the ground is more difficult than in the air or space, said Sovani. With 10 years experience with Fluent, he has worked on both automotive and aerospace problems.

Flow separation occurs as air rushes over a golf club head, a bluff object.(Image courtesy of ANSYS.)
Flow separation occurs as air rushes over a golf club head, a bluff object.(Image courtesy of ANSYS.)

“Automobiles are bluff objects,” says Sovani, using the aerodynamicist’s term for blunt. They’re nothing like the sleek shape of a fighter jet, for example. As fluid tries to flow around a bluff object, it can separate from it, leading to regions of eddies and turbulence. Instead of smooth streamlines that represent particles behaving in an orderly manner, particles swirl, a situation that makes movement of particles difficult to predict.

Wishful thinking? Smoke trails such as these from a wind tunnel test enforce the idea of car slicing through the air. If you could see actual airflow, the airflow over most of the car is anything but smooth. (Image courtesy of JT Technical Solutions at SlideShare.net.)
Wishful thinking? Smoke trails such as these from a wind tunnel test enforce the idea of car slicing through the air. If you could see actual airflow, the airflow over most of the car is anything but smooth. (Image courtesy of JT Technical Solutions at SlideShare.net.)

“Very little of the flow over a car is smooth flow,” said Sovani, shattering the conception most would have of cars nowadays designed to look fast when standing still. With air dams and spoilers, rakish windshields, the smooth and seductive curves … how could they not be slipping effortlessly through the air? Indeed, simulations and testing confirm Sovani’s claim. Only small portions over the hood and windshield show streamline flow. The rest of the car is a swirly mess.

Rocket Scientists Have It Easy

While the automotive stylist concentrates on what a customer sees in the show room, the aerodynamicist is under the car—and he is not pleased. The control volume contained between the ground, wheels and the underbody may well be one the biggest challenges in CFD.

Aerospace engineers don’t have to contend with wheel rims and wheel wells, jagged mufflers and exposed hydraulics and fasteners. Hundreds of bits and pieces, hidden from public scrutiny, poke into the airstream, causing drag. A volume of air gets trapped under the car and scrapes it along the ground. Not a factor with flying objects, the ground is a hard fact of life for the automotive aerodynamicist.

Airflow that enters the engine space through the grille encounters a mess of wires, cables and components and an engine, all exacting another aerodynamic penalty.

It’s no wonder automotive aerodynamicists think CFD on an aircraft is a cakewalk.

Moore No Match for Little Eddy

Turbulent flow analysis can boil down to the study of eddies, the quick little swirls such as those visible in the wake of a powerboat. Eddies are considered big when on the scale of meters. Smaller eddies can interact with each other to form larger eddies. Accurately modeling a small eddy means cells of the order of 1/10 the size of the geometry causing them. When eddies can have radii of microns (10–6 m), that will make cells as small as 10–7 m. That’s trillions of cells in an ordinary analysis. Against this number of cells, the most powerful supercomputer is rendered powerless. Even with Moore’s law (computing power doubling annually), a solution would not happen in our lifetime.

To find an accurate solution for turbulence modeling without having to model with trillions of cells, aerodynamicists have had to resort to approximations. The trick is to have CFD models with much fewer cells but where each cell has the same behavior as the many smaller cells it replaces. Should the program make a wrong assumption, it should be able to correct itself dynamically and automatically and revert to smaller cells.

It is a quest that has kept aerodynamic problem solvers busy for more than a hundred years, said Sovani. To go to a aerodynamicist conference is to be subjected to hundreds of papers that run benchmarks of turbulence models that address the problem in different scenarios. You then compare it to experimental data and find the one that fits your application the best. Which method is best is a matter of hot debate and has been for over a hundred years, said Sovani.

Stoked on CFD


Figure 1. Don't worry, Einstein didn't get it either. Naiver–Stokes equations are the basis for 3D fluid flow. (Image courtesy of NASA.)
Figure 1. Don’t worry, Einstein didn’t get it either. Naiver–Stokes equations are the basis for 3D fluid flow. (Image courtesy of NASA.)
Fluid flow programs work by solving the Navier–Stokes equations (see Figure 1). For turbulence, however, most CFD codes traditionally use RANS, or Reynolds-averaged Navier–Stokes, in which instantaneous effects of flow are averaged over a portion of space and time in the form of additional viscosity terms.
The extreme density of cells for LES models makes it impractical for everyday analysis. (Image courtesy of ANSYS.)

The extreme density of cells for LES models makes it impractical for everyday analysis. (Image courtesy of ANSYS.)

However, a new breed of solutions called SRS, for scale-resolving simulation, has come into vogue for its ability to solve a wider range of problems, such as large-scale separations, rapid swirling and acoustics such as muffling needed in jet engines, airfoils at angles of stall and more.All SRS methods—and there are many—attempt to solve for turbulent fluid flow by using cells for most eddies except the very smallest.

LES Is More

Of the many SRS models, LES, or large-eddy simulation, is a favorite among modern turbulence researchers, giving some of the best correlations with lab results. However, the fineness of the mesh that SRS models require can overwhelm any supercomputer, so a model that uses only LES is less than practical for an everyday solution.
Dr.Florian Menter suggests a solution he has developed. Stress-blended eddy simulation, or SBES, which he presents as both accurate and efficient. He cites ANSYS benchmarks to prove it.

“A 1 million-cell RANS-based CFD model can quickly go up to a 100s of millions or even to a many billions of cell model with LES,” said Menter.

Florian Menter, senior research fellow at ANSYS, and creator of the SBES solution for turbulent modeling.
Dr. Florian Menter, senior research fellow at ANSYS, and creator of the SBES solution for turbulent modeling.
If there was a CFD hall of fame, Menter, who works from his Munich office for ANSYS, would already have a place in it. He has published over 50 research pieces and is the creator of the shear stress transport model.

As often in engineering, the optimal solution is a blend of existing methods – in current case of RANS and LES techniques. Boeing proposed such a model called Detached Eddy Simulation (DES) for treatment of aeronautical flows some years ago. However, DES has gotchas as well. Sometimes, CFD codes that use DES don’t know when to stop, an over-refinement that can lead to flow separation where it would really not occur, an error known as grid-induced separation. To protect against it, aerodynamicists have had to shield areas where there is no possibility of separation occurring from LES

“One cannot use LES in the boundary layer because the cell size needed for LES in that region will be impracticably small,” said Menter, by way of example.

SBES: A Mutually Agreed-Upon Separation

Menter’s new solution, which is available in ANSYS Fluent and CFX software, automatically decides which areas are right for LES and which ones can stay with the more compute-economical RANS method. SBES allows the CFD code to adapt the turbulence model to the flow conditions. The user gets automatically and optimal blend between existing RANS and LES models, without having to define different physics zones during pre-processing. .

“The whole trick is to be able to convert between RANS areas and LES intelligently—and on the fly,” said Menter. This is what SBES does.

SBES uses the RANS model in the thin layer (a few millimeters thin) bordering solid surfaces, where shear forces are having their maximum effect, and in outer flow, where the flow of air is unrestrained. SBES uses LES judiciously in between.

The importance of the thin layer near the surface cannot be overstated, said Sovani. It is where the velocity gradient is the largest. On a race car, the velocity of the air goes from 0 mph on the surface to 200 mph a few millimeters away in the airstream. It is in the high gradients that turbulence is generated.

The SBES method has proven benchmarks with DrivAer, the generic car model, somewhat of a standard among aerodynamicists. (Image courtesy of CFD Online.)
The SBES method has proven benchmarks with DrivAer, the generic car model, somewhat of a standard among aerodynamicists. (Image courtesy of CFD Online.)

“SBES is not so much a new method but a blend of two models, RANS and LES,” says Menter.

ANSYS, one of the leading vendors in CFD with its Fluent and CFX products, has been using many types of scale-resolving turbulence-modeling methods. A single analysis may vary its methods depending on the geometry and flow characteristics. But SBES closes the gap between lab results and theory across more cases, said Sovani of his colleague’s achievement.

Tests run by ANSYS with SBES turbulent modeling have been consistent with where separation occurs in real life for various geometries, orientations and Reynolds numbers.

Approximations Are the Best We Can Do

Any successful attempt to increase accuracy efficiently in simulation should be saluted, and ANSYS SBES is that sort of attempt in the field of aerodynamics. Turbulence, once thought to be utterly unsolvable, a mystery of swirling randomness and chaos, appears to be nearing accurate modeling and believable results thanks to pioneering efforts of researchers such as Menter. We’re still a long way from completely understanding the forces that act in the world of our scale, for lack of knowledge of the world on the scale of the smallest of its particles, but approximations like SBES may be considered a step in that direction.
For More Information

ANSYS has sponsored this article and provided access to their products and people. They have provided no other editorial input. All opinions are the authors, except what is quoted.

Acerca de Norma Alvarez

norma.alvarez@grupossc.com'
Marketing

Ver también

acredita-conacyt-acta-universitaria-revista-competencia-internacional-universidad-guanajuato-ug-ugto

Acredita Conacyt a Acta Universitaria como revista de competencia internacional

Guanajuato, Gto., a 12 de septiembre de 2017.- Luego de una evaluación por parte del Consejo ...

Deja un comentario

Tu dirección de correo electrónico no será publicada. Los campos necesarios están marcados *

*

Puedes usar las siguientes etiquetas y atributos HTML: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>