First, "laminar" does not necessarily mean "no change in one of the dimensions". Second, any numerical model is an approximation of reality. You need to read much more and study this subject in order to better understand the details. Simply running a computer model of something doesn't mean you understand the physical processes involved.

Understanding what "laminar regime" stands for is a topic of basic fluid dynamics.

The flow over an obstacle can be laminar even in case it is unsteady, as happpens in the classic case of laminar vortex shedding.

Turbulence is charactertized by a wide range of vortical structures.

Howdy Yi-Shian Ciou,

The earlier replies have good advice. I shall only add a bit of information/opinion that I think also fits your questions, but study on your own, too.

Fluids are composed of atoms, molecules and/or small clusters of each in some cases. These fluid components are in continual random motion called molecular unrest. A fluid will respond (strain) to any force (stress) however small. If the response is like the reaction of a deck of playing cards when the top card is pushed parallel to the table and the stack shears to an angle the flow of the fluid is called laminar. This can occur along the surface of a channel when a liquid flows along that surface, say due to a slope.

The critical issue of "laminar" is smooth variation of the motion normal to the flow with "no" mixing between layers. This happens with the deck of cards because they are solids, but fluid "layers" may mix easily. With very low stress there is molecular diffusion due to the molecular unrest that shears the layers of fluid from no-slip at the surface so they move like the card deck - very low stress! The flow may curve as it does off the top of your cylinder, but no bulk mixing is permitted between layers in the laminar case.

The character of "turbulence" is that groups of fluid elements many times larger than the molecular unrest scale form as structures (eddies, etc.) and cause bulk mixing between the layers. The way this happens has many explanations, so pay attention to the conditions for any specific theory.

A convenient engineering approach to when turbulence will start in a laminar flow is given by the Reynolds Number which provides a rough indication. It is defined as a ratio of bulk diffusion to molecular diffusion by use of flow velocity, a dimension of the flow location and a measure of molecular diffusion. Text books are full of examples, but I have chosen to write about it in it's original purpose to encourage understanding "how" and "why" issues. The critical Reynolds number varies over a wide range that depends on geometry of the flow location, like a pipe versus a flat sheet, BUT is is a very good way to compare cases with different flow speed, geometry and viscosity IN A SPECIFIC GEOMETRY. The Reynolds Number is used in many other cases where the transition to turbulence is not the issue, because it is convenient, so beware and be sure you understand what is being claimed.

The scale of turbulence structures that do bulk mixing across a flow has a wide range depending on the intensity of the turbulence. At the small end it is barely discernible and may be ignored in practical cases, at the large end the eddies are as large as the space available permits.

Look up flow visualization papers and books that include good captions, say An album of fluid motion, by Milton van Dyke, Parabolic Press, Stanford, Ca. (1982) 174 pages, or more simply, spend some time online with the images of "flow visualization" you find there and read the captions. If you are visually oriented and add some study, you'll understand soon and well.

Happy Trails, Len

P. S. A fellow camera man it seems from your picture. Okay. lfh

The main parameter affecting transition from laminar to turbulent flow is Reynolds number and major factor causes separation of boundary layer from the body surface is positive pressure gradient. These two phenomena may occur simultaneously on flow passing a circular cylinder. So one should not confuse while analysing fluid flow in such cases. It is better before any numerical simulation, to carry out certain experiments, so that one get better feeling on actual flow situation.

I am not going to talk about the theory as there were already many good answers explaining it, and I would not be able to add much more.

But let me ask you a couple of questions. Are you performing CFD simulations. Which approach are you using? LES, URANS, RANS etc?

If you are performing RANS, which turbulence model are you using?

I am asking it because you need to use turbulence models that predict (try to predict) the laminar to turbulent transition, such as the the k-omega SST with intermittency, or the Re-Theta model.

Then it is very important that you have a fine enough mesh to solve your case.

These are just basic suggestions, then the challenge would be to validate your problem. Not easy. Good luck!

Fabio Malizia hi,

I'll use ANSYS two-way coupling to capture the motion of rigid body forced by Karman Vortex Street. And I'm struggling with Fluent which calculates the fluid domain 'cause residuals didn't show a downward trend, i.e. it was unstable. Well to reduce complexity, so I set the model laminar (Re≤200)[1]; as for turbulence, just leave it for when I successfully enrol postgraduate school.

Last night, I read a paper discussing about the "end effect" of finite cylinder [2], it seems normal that inherent downwash near the top interconnects wake. So should I still use turbulence model? To use any turbulence model, building a correct boundary layer mesh is a challenge 'cause total thickness should match δ99, but I didn't find any paper about empirical formula of δ99 for cylinder. what do you think?

[1]Book Hydrodynamics Around Cylindrical Structures

[2]Article The finite-length square cylinder near wake

Milne Ando

I am not sure of what you want to do. "Laminar" is not a model, it is just the setting of the solver for NSE without any other assumptions.

The numerical solution you get is laminar if the regime is laminar but can become transitional and turbulent with the same NSE solver. The real issue would be the unresolved components when the flow become turbulent on a certaain grid. That would implicitly produce a LES-like solution without SGS model. You can turn to a full LES by adding the dynamic Smagorinksy model or, if your computational power is sufficient, you can afford a DNS with the proper grid resolution.

There are RANS formulations that try to consider transitional conditions but I would not suggest this kind of approach. Generally, RANS is a statistical approach where it is supposed to work with fully developed turbulence. A lot of variants are present in literature but you should be aware of the fundaments of the statistics.

Filippo Maria Denaro hi,

considering the oscillating motion may turn wake into turbulence (even upstream Re≤200), it's conservative to use turbulence solver, is my understanding correct?

Someone suggested me to use Spalart-allmaras, well but I meet two problem:

what do you think?

There is no formula for d99 on the cylinder …

it is just a rude way define a limit for the BL that is, actually, extending in theory up ti Infinity. and the theory is for a flat plate.

Again, if you solve the NSE you are assuming nothing about the regime of the flow.

Your first task is to evaluate if your flow problem can be solved this way, without any assumption.

Generally, Navier-Stokes eqns are a good model for laminar as well as turbulent flows as long as we talk about relatively low Reynolds numbers, even for non-Newtonian fluids and dispersions. Regarding the transition to turbulence, much depends on geometry. The Reynolds number gets only first-guess of the character of the flow. For fully developed turbulence I recommend to look at the extreme case of Re = \infty and flow along plane boundaries where statistics helps very much, see e.g. the attachment. However, even at Re = \infty "outer" geometry of flow and details like curvature of pipes possibly matter.

As I said it all depends on flow Reynolds number. Other parameter that could enhance turbulence on any surface includes; approaching free stream turbulence, flow pressure gradient, flow curvature, flow temperature, free stream divergence or convergence.

No It depends on the flow speed and Reynolds number.

Dear Milne,

based on the data reported in the literature, you get the vortex shedding regime already at values of the Reynolds number as low as 50 (see works from Williamson). The fact that vortices are produced in the wake (therefore the flow exhibits large-scale unsteadiness) is not a sufficient condition to trigger the break-up of these vortices into smaller scales, typically regarded as "turbulence". Provided you increase the Reynolds number (typically to the order of 10^3), eventually you will observe such a cascade of scales.

The fact that the flow has a vertical component atop your truncated cylinder cannot be associated to the laminar, steady or unsteady, or turbulent condition. It is just a result of the geometry, where the flow passes this obstacle generating a specific pattern of streamlines.

About laminar and turbulent, I agree with the previous indications from Leonard and Filippo. I use often the book from Milton van Dyke to show students how a given flow regime "looks like".

Hi Milne and all.

Since you used the word “always” in your question, the answer is unequivocal: no.

In short, the occurence of turbulence depends on the prevailing Reynolds number, which in turn requires the specification of a finite dimension of the body under study. For the classical problem regarding flow over a wide flat finite plate, no lateral border effects are present and Reynolds is defined using the plate length. Within the boundary layer, the flow is 2-D and velocity depends on the longitudinal distance as well as on the height above the plate. In your case, depending on the streamline considered over the circular cylinder cross-section, the Reynolds number will vary. I guess, at least in principle, the existence of turbulence would depend on the chord considered over that cross-section. Thus, for a given U∞ (undisturbed approach velocity), turbulence could occur for a large chord but not for a small one. A profile of Reynolds numbers would then exist. In this case, microeddies could diffuse laterally.

Milne Ando

Focusing only on the flow pattern you get (pathlines), I would say you have a steady solution but that is not physical in your case for Re=200.

Check your setting, something can be wrong in your simulation.

Do not use first order scheme, set a fully second order unsteady solver. Put a probe behind the body and register the velocity components in time.

Let the solution to fully develop and forget the arbitrary initial conditions.

You should see the onset of the vortex shedding.

it depends on flow Reynolds number

Hello

The flow past circular cylinder is not always turbulent.

It depends on Reynolds number.

For smooth cylinder the flow is laminar for Reynolds numbers less or equal than 150 to 180 it depends on many other parameters.

Even though, the vortex shedding start at Reynolds around 50, it remains laminar until up to Reynolds 150.

In your case Re = 200 you should have vortex shedding already.

The vortex shedding starts to be turbulent but very close to the cylinder (on the boundary layer and the separated shear layer) you should have laminar flow.

The down-wash you've seen in the edge of the cylinder is natural since this zone behind the cylinder the pressure is low.

The flow in this region is 3D.

Not always. the Reynolds number affect much

Dear Milne,

Laminar doesn't mean that the flow is always unidirectional. In fact it possible to obtain laminar vorteces. In simple words, laminar flow is a flow in which fluid layers have a "smooth" motion and do not mix with other layers. Also the Raynolds number depends on the geometry, hence the Reynolds number is a relative concept (e.g. Re = 200 can be low in some configuration but high in other).

Matteo Baldelli hi,

thanks for your explanation. But Is there any method to judge whether layers mix or not via post-processing? 'cause I don't know what parameter I should monitor. For 2D case, end-effect is not considered, while for 3D case end-effect (downwash or upwash) interacts with wake. Thus, I guess wake near the end of cylinder transits to turbulence first when Re increases over 200.

What do you think?

Milne Ando

be careful, turbulence is never 2D.

The tool to analize the flow is the statistical analysis.

But the plot you put has nothing to do with a turbulent pattern.

There are basically 5 flow regimes around a smooth circular cylinder placed with axis normal to a uniform incoming flow. Firstly there is a 2D laminar flow without flow separation for Re

Howdy Folks,

Avraham Hirschberg has provided an excellent review of flow around a cylinder from an engineering/experiment viewpoint. I wish to add only a couple observations from a river watcher/scientist viewpoint, especially since the illustration and Milne Ando's question involve a cylinder of limited height in the flow.

I believe one will discover that the 2D laminar flow regime exists only in a full channel with the cylinder occupying the full depth, else "downwash" in the wake will occur even with creeping flow. That it occurs in the simulation is real 3D without turbulence at that flow rate and that Milne has found a "downwash" article that addresses that departure from Oseen flow is good. Even a young mountain stream knows how to get over a pebble, but the math is complicated and often "streamlined" in practice.

Turbulence with a focus on vortex stretching and L. F. Richardson's ditty "Big whirls have little whirls . . ." is a good approach in the pursuit of results. "The cascade of transfer of energy from large eddies down to scales at which energy is dissipated by viscosity is the essence of turbulence." No, it's just an important portion of the process. When one looks more deeply into the processes of turbulence generation one finds that asymmetric diffusion on the molecular scale and larger scales also is important because it creates lateral accelerations and vortex cores. This topic normally is not considered so I offer it for the next generation. To look at the vortex in the wake of a bluff body and to explain how it develops would be a good start for anyone interested.

Also the flow toward the cylinder will develop a shear layer with higher velocity away from the surface, The stagnation pressure of this higher velocity will cause flow down toward the surface and eventually a full horseshoe vortex wrapped around the cylinder and stretched downstream that increases the complexity of the flow regime.

"Around Re>2000 the full wave has become turbulent, but the flow in the viscous boundary layers at the wall of the cylinder remains laminar. The so called resistance crisis (sudden decrease in drag) somewhere in the range 5+5

All these answers are very good but the real issue is in the pattern of the flow showed in the picture. That numerical solution is wrong. Why the result is wrong is still unclear…

Howdy Filippo Maria Denaro,

The laminar/turbulence issue has been well addressed here as you note. Several replies have demonstrated knowledge of the program and programming involved, and I must rely on them to evaluate the model constraints and initial/boundary conditions for that part of the answer to your observation.

Since you noted that the pattern of the flow in the picture is wrong, I looked again and agree. What you see as wrong will be welcome when provided, but for me the parallel "pathlines" across the top of the cylinder are hard to accept for a real flow. After all, the pressure distribution at the top edge of the cylindrical surface would vary from stagnation through a minimum to stagnation and no indication of that is present in the parallel pathlines on the circular surface. If they were an initial condition the calculation may be "right" but very "unreal," which has been the case on occasion. Interesting issue.

Happy Trails, Len

Leonard Hall

dear Len, the pattern is wrong since it reproduces a specular steady solution at Re=200.

The real laminar flow would produce an unsteady pattern of separated vortical structures in 2D and the finite lenght in this 3D case would produce also a further effect due to separation on the top plate.

The author has not yet addressed the set-up of his simulation. I could immagine that a strong numerical viscosity is introduced by the scheme he used but without details I cannot say why he got a solution resembling a very lower Re number.

Regards.

The flow pattern and the drag on a cylinder are functions of the Reynolds number ReD = U∞D/n, based on the cylinder diameter D and the undisturbed free-stream velocity U∞. Recall that the Reynolds number represents the ratio of inertial to viscous forces in the flow.

Interface boundary is used, when the surface is shared by both Fluid and solid regions. If you can explain the problem in detail, may be i can describe it in more detail. Interfaces, in the sense of Fluent, are matching surfaces with different mesh topology.