CFD Online Discussion Forums

CFD Online Discussion Forums (
-   Main CFD Forum (
-   -   Programming own CFD, noone supports vortex rings? (

DanFrederiksen September 1, 2010 10:56

Programming own CFD, noone supports vortex rings?
I plan to implement a CFD for some car aerodynamics experiments if I can and I was wondering what the right physics model is and I thought about smoke rings. is it really true that the existing commercial CFDs can't simulate that? why not?
does it mean that something vital is missing in current models?
or is it even worse that the vortex ring mechanism is not even understood?

agd September 1, 2010 11:04

How are you defining support? Modern CFD techniques are perfectly capable of computing the vorticity in a flowfield, and higher-order methods do a nice job of conserving the vorticity as it is advected. Or are you talking about using vortex rings as computational elements? If so, what advantage do you see gaining by using those over other elements?

DanFrederiksen September 1, 2010 11:28

no I meant as a macroscopic phenomenon. can for instance Fluent produce a stabil 'smoke' ring if starting conditions are right? with no cheating or special case bullshit

agd September 1, 2010 14:22

I don't use fluent, so I don't know what it can or can't do. All of my work involves in-house software. However, in a general sense any modern CFD solver should compute a velocity field that gives rise to a vorticity distribution comprising a vortex ring given the correct BCs. The only real difficulty is that most current solvers are at most 2nd-order accurate, and vorticity tends to get smeared out as it is convected due to numerical diffusion, unless extremely fine grids are used. Higher-order solvers generally do a good job on convecting vorticity on coarser meshes.

DanFrederiksen September 1, 2010 14:28

ok. can your own software do it? have you ever seen a vortex ring in simulation that wasn't specially designed to simulate vortex rings?

do you understand how a vortex ring can travel stably?

agd September 1, 2010 15:29

While I have never performed a simulation of a vortex ring, I have both seen and performed numerous simulations where concentrated regions of vorticity are convected large distances with minimal diffusion. A vortex ring convects via self-induction, and can be modeled inviscidly with the Biot-Savart law. Once perturbations enter circumferentially the motion of the ring can become quite complex. This is a model for the real viscous world, where you have a ring of concentrated vorticity that arises due to a particular set of boundary conditions/initial conditions. Any flow solver that can convect vorticity with minimal numerical dissipation will produce a good solution.

DanFrederiksen September 1, 2010 19:01

perhaps you should try it?

agd September 1, 2010 21:24

Why do more? As I stated, I have done computations which involve regions of concentrated vorticity like vortex rings. A cylinder in crossflow is a simple example in 2D that contains much of the relevant physics and tests the numerics. I have done 3D simulations of flows over a range of bodies that shed complex vortical structures. I have done the Biot-Savart computations for inviscid vortex rings. A better question, given the ttitle of your thread, would be "why don't you try?"

DanFrederiksen September 2, 2010 10:39

how disappointing.
I'm just getting into CFD and you claim to be very established and have software of your own yet you conclude that it's my responsibility to test this archetypal phenomenon.
are you afraid it wont work? so dismissing it is the best course of action?

have you ever seen a vortex ring in any CFD sim anywhere? has it ever been done

agd September 2, 2010 10:57

What part of reading comprehension do you struggle with - as I noted, I have seen it, done it, got the free t-shirt. Your original question concerned the suitability of CFD to capture effects such as the motion of coherent regions of vorticity in a flowfield (of which the vortex ring you are continually harping is an inviscid approximation). That question has been answered, and now you seem determined to ignore the answer in favor of some idiotic and pointless challenge. If you can't be bothered to search the open literature and interpret the answers given in fora such as this, then you are simply wasting time - yours and mine.

DanFrederiksen September 2, 2010 11:13

well perhaps I am a bit pushy when rightly impatient but I did read and comprehend this sentence of yours "While I have never performed a simulation of a vortex ring".
isn't Biot-Savart an electromagnetism theory?

and would it really take that much effort to setup a 'smoke ring' sim? who knows, you might learn something new..

agd September 2, 2010 11:45

Let me try to clarify then - I have never done a numerical simulation of an inviscid vortex ring. A vortex ring is an INVISCID model of a viscous phenomenon, to wit a vorticity concentration in the shape of a ring, such as would develop from a pulse of fluid being ejected from an orifice. I have done simulations of such VISCOUS flows, with good results in terms of the vorticity convection and self-induced motion. As noted (in the original answer to your question concerning software capability) there are modern CFD tools which are quite capable of doing this, many of which you can probably find on the open market. It is simply a question of good numerics, which can be readily evaluated by looking at similar problems such as vortex shedding from a cylinder in cross-flow, or the tracking of a vorticity concentration undergoing advection. These are standard test cases that can indicate how good a solver does at hadling these regions of vorticity, and if you want to evaluate a particular algorithm then you should include them in your validation suite. I do that all the time when looking at algorithms. If you must restrict yourself to an inviscid vortex ring then do yourself a favor and look up how to use the Biot-Savart law to compute the self-induced motion of a vortex ring. It is a classic problem in fluid mechanics - I would recommend the texts by Streeter, or Batchelor, or Horace Lamb. But if you are genuinely interested in flow over automobiles (as stated in your original post) then you are going to have to include viscosity. So to answer your question one more time - there are modern solvers that can track vorticity (even vorticity in the shape of a ring). I have done it in a variety of viscous cases. I have not done it for an inviscid vortex ring, because I have no pressing reason to do so - I don't live in an inviscid world. I also have every confidence that the codes I use which can convect vorticity correctly in a viscous flow will work fine if I turn viscosity off.

At this point, either I have answered your original questions, or we are simply talking past one another and I have no real clue exactly what you were trying to ask.

DanFrederiksen September 2, 2010 12:11

you keep saying inviscid. are you saying a smoke ring somehow manages to remove all viscosity or that the phenomenon can only be replicated in simulation if viscosity is set to zero? you may not live in an inviscid world but you live in a world where smoke rings are real.

it also sounds like you're saying that vorticity requires special handling as opposed to a simple universal model akin to 'cheating' or hacking when you say "there are modern solvers that can track vorticity". am I understanding that correctly?

agd September 2, 2010 13:41

No - you are not understanding correctly. You keep referring to a vortex ring as if it is some unique phenomenon. A vortex ring is simply an inviscid model of a viscous fluid behavior. Any modern CFD code that can track a concentration of vorticity without numerically diffusing it (note that this is a numerical issue, not a question of physics) will do what you ask. There are no special buttons or knobs or tricks needed by these codes. We simply solve the equations of motion in a way that the numerics don't destroy the physics (which is basically just good CFD practice). If I want to use my viscous CFD tool to solve for a vortex ring (which to me refers to an inviscid approximation), then I'll do it by setting viscosity to zero in my flow solver. If I want to simulate a smoke ring, then the question to ask is "do I simulate this as a vortex ring (inviscid) or do I simulate this as a region of concentrated vorticity that can convect and diffuse (viscous) as well as induce its own motion?"

DanFrederiksen September 2, 2010 14:45

I suspected you meant something contextually specific rather than the more general meaning of vortex ring. these are vortex rings:
and I was just wondering if common CFDs could do that since I couldn't find any CFD demonstration of such. general CFD, not one where the outcome is hardwired.

DanFrederiksen September 3, 2010 10:36

would it kill you to try? you know you want to : )

also a couple of other interesting test cases. the golf ball. if the sim can't demonstrate the dramatic reduction in drag from the dimpels then the model seems critically limited.

another interesting test is a spinning shallow dome as a potential replacement for helicopter rotor.

adrin September 3, 2010 20:03

Traditional CFD methods will not be able to capture vorticity accurately. However, as mentioned by agd, you _can_ capture vorticity (nowadays) if you use high-order methods and/or AMR (adaptive mesh refinement) techniques. You will need _high_ order as in higher than 4th order discretization to maintain vorticity for long times using nominal grid points. Alternatively, you can use Lagrangian vortex methods to do the same using even lower order discretization. Biot-Savart is the name given for the integral evaluation of velocity due to a distribution of vorticity because the integrals are of the same form as that for electromagnetics. Check basic fluid mechanics books for that. As for vortex ring simulations, check papers that deal with Lagrangian vortex methods and invariably you will see results from 3D simulation of vortex rings (under various perturbation conditions, which, depending on the wavenumber of perturbation, will give you different forms of evolution).

As for simulation similar to your youtube reference, I simulated head-on collision of vortex rings more than 7 years ago using a technology that is far less sophisticated than what we're using now. Go to click on projects, then "Development of a grid-free Large Eddy Simulation package for vortex-dominated flow in complex geometries (NSF SBIR project)" to see two vortex ring simulations, including head on. Again, this was from a while back, and we can do much better. The latter use the so-called Lagrangian vortex methods.



adrin September 3, 2010 20:15

I have to disagree with you about a vortex ring being an inviscid flow mechanism or even a model. Vortex rings exist physically even at low Reynolds numbers. There is nothing inherently magical about vortex rings - it's just that they are shaped in the form of a ring. Take this ring to be an initial condition and submit this to the vorticity transport equation (contains convection, stretch _and_ diffusion) and so long as the numerics are correct the equation will faithfully let the ring evolve (and experience diffusion). We have run laminar vortex rings with Lagrangian vortex methods in-house without a problem.

As for your claim that you can "simply" turn viscosity off and your code will faithfully simulate the same problem in "inviscid" mode, nothing could be further from the truth. If you turn viscosity off and are able to run the "inviscid" case with "no" problem - read: without blow up - then your code/simulation contains numerical diffusion and strictly speaking it's not solving the inviscid flow problem. If your code does eventually blow up then there's a good chance that you're solving the inviscid flow case, but in that case you shouldn't be claiming that there is "no" problem!



Avr.Tomer September 4, 2010 02:20

Dan, Bio-Savart law has got nothing to do with any physical phenomenon its pure mathematics and proved by using an exitence of a divergence free vector whose curl is a divergence free vector and applying Green function for the laplacian and some vector identities. It doesnt metter if you deal with a vorticity or a magnet field. By this, the mathematical form describing the velocity of a thin vortex ring by means of the Bio-Savart law has nothing to do with viscinity or the lack of it because its not derived from a description of a specific physical dynamical system.
Since the circulation theorem conclusions about vorticity creation remain true for both viscous and invsicid scenarios one must crate it somehow.Then, once its created its evolving in time depends on the mechanism supplied by the NS equations (ie Physics). Therefore sustaining the structure in time is different whether you "add" a viscous term in the vorticity transport equation or "omitting" it (math again - "add" and "omitting" for clear reasons ).
Now modeled via CFD one must consider a sufficient numerical evalution for the mechanism that sustains the structure and generaly even using higher order methods might collapse if the solution is not regular enough. The use of an inhouse code (which its lack of robustness will be in its advantage for getting a valid justified description) that eventually describe the phenomenon and owns the correct physical justification for the model applied is not by any means "cheating" its the essence of CFD.

All times are GMT -4. The time now is 13:09.