15

For air flowing over a wing, both flow separation and turbulent flow involve disturbed flow next to the surface and smooth flow further away. At what point does one say "oh, this flow has changed from turbulence to separation" (or vice versa) and why?

I am wondering if separation involves only a disturbed boundary layer, while turbulence can involve a wider disturbance such as in a stall?

For example, is it correct to say that in the stall, an already-turbulent flow (sometimes experienced as burbling) becomes detached?

Or that vortex generators, designed to re-energize a stagnating boundary layer, do so by creating turbulence in order to prevent separation?

Guy Inchbald
  • 7,042
  • 18
  • 36

3 Answers3

17

At what point does one say "oh, this flow has changed from turbulence to separation"

At the point where the flow reverses direction.

enter image description here

Flow separation. The bold curve is the surface/wing.

Yes, that can happen.

Both turbulent and laminar flow can separate. Turbulent flow is in fact less likely to separate than laminar flow. This is why aircraft wings often have devices that deliberately create turbulence on the wing.

(Yes, separated flow produces negative skin friction, but at the price of huge pressure drag)

Here's a (badly hand-drawn) diagram showing the difference between laminar, turbulent and separated flows.

enter image description here

Just a clarification about stall. The stall is when the lift reduction caused by flow separation overwhelms the lift increase caused by flying at an increased angle of attack. Flow separation can happen without stall, and it will reduce the benefit gained from a higher angle of attack in proportion to the extent to which the flow is separated, but stall cannot happen without flow separation.

Indeed, many wings have separated flow at the trailing edge sometime before "stall" is reached. As one draws closer to "stall", the region of separated flow expands forward. The turbulence created by the wake of this separated flow hits the tail, causing "buffeting", which gives the pilot a warning that he is approaching stall. Airfoils that lack this feature, such as supercritical airfoils or sharp supersonic ones, tend to be dangerous to fly at slow speed with it's inherent high angles of attack.

And as you can see from the diagram, the flow separation at a given angle of attack is much worse for laminar flow than for turbulent flow. So the laminar separated case is more likely to be a stall than the turbulent separated case.

enter image description here

Lift vs angle of attack for thin, sharp wings vs thick ones. Airfoils designed for laminar flow fall in the thin category. And as above, just having or not having laminar flow on a wing can make a similar difference.

And yes, vortex generators prevent separation by creating turbulence, which causes high speed freestream air to get mixed with low-speed boundary layer, speeding up the boundary layer. It's a tradeoff between the drag of a turbulent boundary layer and the even greater drag and lift loss from flow separation.

Abdullah is not an Amalekite
  • 3,571
  • 1
  • 15
  • 58
  • But what is the technical distinction between separation and turbulence? Which is your diagram depicting and what would a diagram depicting the other one look like? – Guy Inchbald Jun 17 '20 at 15:59
  • @GuyInchbald Sorry, It depicts separation. The bold line is the wing. The normal lines with rows of arrows show the velocities of the boundary layer. – Abdullah is not an Amalekite Jun 17 '20 at 16:11
  • Thank you. It makes sense now. – Guy Inchbald Jun 18 '20 at 19:28
  • Would that last, detached turbulent flow be what is called the stall condition? – Guy Inchbald Jun 20 '20 at 06:11
  • @GuyInchbald see above – Abdullah is not an Amalekite Jun 20 '20 at 08:07
  • Thanks. I can see that the laminar separated case is more likely to lead to a stall than the turbulent separated case, but if there is laminar separation prior to the stall, then surely the snap transition to stalled condition must be the sudden onset of turbulence? Would separation not grow smoothly? (BTW the DH.88 Comet racer is neither supersonic nor supercritical but has a notoriously sharp stall with no warning) I am also uncertain how the Kutta condition and reverse flow near the trailing edge relate, but that is going a bit off-topic. – Guy Inchbald Jun 20 '20 at 08:52
  • @GuyInchbald apparently the DH.88 had thin wings. Also, see above again. – Abdullah is not an Amalekite Jun 20 '20 at 09:13
  • @GuyInchbald Separated flow may expand gradually or suddenly, depending on the type of wing. A wing that, as above, has sudden, sharp stalls will also have sudden, sharp increases in flow separation just before the stall point, since after all, flow separation is what leads to stall. – Abdullah is not an Amalekite Jun 20 '20 at 09:20
  • @GuyInchbald Think of a stall as a condition where pitching up decreases lift rather than increase it as would normally happen. – Abdullah is not an Amalekite Jun 20 '20 at 09:21
  • But wings can stall while still perfectly stable in pitch. So pitch instability cannot explain the sudden onset of the stall. – Guy Inchbald Jun 20 '20 at 09:44
  • @GuyInchbald added a new diagram. – Abdullah is not an Amalekite Jun 20 '20 at 09:49
  • @GuyInchbald Buffeting doesn't necessarily mean you're unstable, it just means that you're being bumped around because your tail is being pounded by turbulence. – Abdullah is not an Amalekite Jun 20 '20 at 09:52
  • But what causes the sudden violent breakdown which characterises the stall of a thin wing? Is it a sudden change in separation or in turbulence or in something else? So far we have established only that it is not pitch instability. – Guy Inchbald Jun 20 '20 at 10:07
  • 1
    @GuyInchbald: The wing is mildly unstable in pitch (more so with more camber) and only the wing-tail combination with an installed tail will be perfectly stable in pitch when the wing stalls. The sudden, sharp stall is caused by sudden flow separation right past the nose of the airfoil (causing separated flow over much of the airfoil) while the benign stall is caused by slowly increasing separation originating from the trailing edge and creeping forward with increasing angle of attack. – Peter Kämpf Jun 20 '20 at 10:07
  • @PeterKämpf Thank you. The pitch instability bit is wholly irrelevant though, I wish you guys would stop bringing it up. – Guy Inchbald Jun 20 '20 at 10:17
  • @GuyInchbald just FYI I was talking about the empennage being hit by the wing wake. And yes that's not that relevant. The whole point is that the thick airfoil has a gradually increasing level of flow separation as stall angle-of-attack is approached, which warns the pilot – Abdullah is not an Amalekite Jun 20 '20 at 11:52
7

Think of the boundary layer as a multi-lane highway with rubber cars which can bump into each other. This highway has a sticky curb on one side and the cars are a bit sticky themselves, so cars near that curb get the slower the nearer they are.

In one case the cars stay in their lanes and the rightmost lane, right next to the curb, (sorry, you Australians, Japanese or Indians: For you that would be the leftmost lane) is occupied by the slowest vehicles. Speed increases with each lane more distant from this slowest lane since cars rub along nicely. This is like laminar flow.

Now traffic changes and the drivers switch lanes frequently. The result is that cars in the slowest lanes have to speed up. New lanes join the fastest lane from time to time so the speed in the fastest lane will not slow down. Speed is now much more equal across lanes but the whole highway grows wider to accommodate all those new lanes with fast vehicles. This is like turbulent flow.

While in laminar flow the parcels of air all flow in the predominant flow direction, in turbulent flow there is a lot of crossflow, so those parcels get bumped along if friction with the wall (the sticky curb of the highway, to stay in the picture) slows them down too much. This needs a constant addition of new, high-energy parcels so the whole boundary layer is thicker and has a fuller speed profile.

However, if the speed gradient along the predominant flow direction is negative (say, in the recompression area in the rear upper half of an airfoil), the cars in the joining lanes become slower and the slower lanes slow down, too. It's as if they obey a sequence of speed limits that tell everyone to reduce their speed by some MPH. And then some more. If the speed near the curb (in the slowest lane) drops to zero and then reverses, flow separation has occurred. Now the slowest lane fills up with vehicles from both directions which pushes the cars in the adjoining lanes further out. The highway width explodes.

This can both happen with no or much lane changing; the result is the same. When it happens with no lane changing and drivers change their mind about that detail further downstream, the new cars joining will now bump all others along and get traffic moving again. This describes a laminar separation bubble with reattachment downstream.

I am wondering if separation involves only a disturbed boundary layer, while turbulence can involve a wider disturbance such as in a stall?

Every flow separates at the trailing edge. With too much angle of attack, this separation creeps forward on the upper side on thick airfoils or a new separation starts past the suction peak near the nose on thin airfoils. This separation, when extensive enough, causes loss of lift and defines the stall. Both laminar and boundary layers can experience this.

A special case is a laminar separation bubble which occurs past the suction peak but the subsequent transition to turbulent flow causes reattachment. This can still be followed by a separation of the turbulent boundary layer later on.

For example, is it correct to say that in the stall, an already-turbulent flow (sometimes experienced as burbling) becomes detached?

Yes, but also a laminar boundary layer can separate and cause stall (mostly at model airplane scales and smaller). The "burbling" you mention is not caused by this but by larger eddies hitting the tail. This indicates a major separation near the trailing edge on the inner wing but with no or little loss of lift. This kind of turbulence is different from that in a boundary layer and of a much larger scale.

Or that vortex generators, designed to re-energize a stagnating boundary layer, do so by creating turbulence in order to prevent separation?

Yes. Vortex generators add more high-speed lanes to the traffic in the boundary layer. They also help to fix the location of shocks in transsonic flight.

Peter Kämpf
  • 231,832
  • 17
  • 588
  • 929
  • Great. Next time I get on a plane, I'm going to look at the wing and see it full of little rubber bumper cars bounding all over the place. :) – FreeMan Jun 17 '20 at 17:45
  • 2
    Now what has the UK done to you that it deserves no mention? :-) – TooTea Jun 18 '20 at 10:50
  • 1
    I would really really avoid mentioning molecules at all. Turbulent and laminar flows are all about the continuum. The molecules are completely chaotic in either. Individual molecules start to matter at completely different scales, the mean free path in air is around 70 nm. There is a good reason why fluid parcels or particles were invented https://en.wikipedia.org/wiki/Fluid_parcel – Vladimir F Героям слава Jun 18 '20 at 13:06
  • @VladimirF: Yes, that makes sense. I replaced them with "parcels of air". – Peter Kämpf Jun 18 '20 at 18:18
  • 1
    @TooTea: They infected too many countries with the disease of driving on the wrong side of the road. But maybe I should give the south of Africa a honorary mention. – Peter Kämpf Jun 18 '20 at 18:20
  • As one who understands the good nature of Yankee jibes in ineffectual defence of driving on the wrong side of the road, I would note that we Brits chose the left in the interests of road safety - to face the oncoming horses and pedestrians. Even today our Highway Code recommends that where there is no pavement (which some understand only as "sidewalks") pedestrians should face the oncoming traffic. More paranoid countries followed the horses in avoiding the oncoming rider's sword arm. But I only ever met rubber bumper cars at the fairground and none had a reverse gear. – Guy Inchbald Jun 19 '20 at 09:42
  • @PeterKämpf I wish you would have a look at the comments section of Phil's answer – Abdullah is not an Amalekite Jun 20 '20 at 15:07
  • @Abdullah. Done. – Peter Kämpf Jun 20 '20 at 17:16
3

Flow separation and turbulent transition are completely different phenomena.

Flow separation is driven by an adverse pressure gradient in the flow direction. On the top surface of a lifting surface, the flow has to decelerate and return to farfield pressure as it approaches the trailing edge of the surface. So there is an adverse pressure gradient near the back of foil topsides. The problem is that this pressure gradient penetrates the boundary layer right down to the skin of the foil, and the boundary layer has been slowed due to skin friction. The result is that getting the air outside the boundary layer slowed down to freestream velocity can result in the boundary layer flowing the wrong way, forwards over the wing. The flow has to go somewhere, so a bubble forms and the streamlines lift away from the skin. Laminar flow boundary layers are prone to this happening due to the velocity profile of laminar boundary layers.

An area with a strong adverse pressure gradient can also develop just behind the leading edge suction peak. This may form a bubble and flow often reattaches behind it. One common occurrence is for a laminar separation bubble to form and for turbulent flow to reattach behind it. These can be stubborn and tend to produce hysteresis in the lift vs AoA curve.

Separation is less likely to happen in turbulent flow, as it needs a greater adverse pressure gradient to happen.

Turbulence is strongly a function of freestream velocity, and only weakly a function of pressure gradients. Indeed, many turbulence models just use flat plate turbulence data (zero pressure gradient) and ignore the pressure gradients completely.

So the difference is that they are caused by different conditions. Separation needs an adverse pressure gradient strong enough to back up the boundary layer, and turbulence doesn't much care about the pressure gradient.

Guy Inchbald
  • 7,042
  • 18
  • 36
Phil Sweet
  • 316
  • 1
  • 5
  • Creating turbulent flow on a smooth wing needs an adverse pressure gradient. Don't forget to mention that – Abdullah is not an Amalekite Jun 19 '20 at 10:02
  • @Abdullah I'm guessing you are referring to this bit - " All boundary layers start off as laminar. Many influences can act to destabilize a laminar boundary layer, causing it to transition to turbulent. Adverse pressure gradients, surface roughness, heat and acoustic energy all examples of destabilizing influences. Once the boundary layer transitions, the skin friction goes up. This is the primary result of a turbulent boundary layer. The old lift loss myth is just that—a myth." Adverse pressure gradient does have an weak effect on turbulent transition, but it is not required. – Phil Sweet Jun 19 '20 at 19:48
  • @Abdullah here is an example where the standard wall function of turbulent boundary layer, which does not consider pressure gradients, is upgraded to one that does. - https://www.afs.enea.it/project/neptunius/docs/fluent/html/th/node100.htm – Phil Sweet Jun 19 '20 at 19:51
  • @Abdullah You most certainly do not require an adverse pressure gradient to transition into turbulent flow. – JZYL Jun 19 '20 at 21:24
  • That explains flow separation nicely, thank you. But it does not characterize turbulence - for example, might adverse flow not be seen as a simple form of turbulence? Why not? – Guy Inchbald Jun 20 '20 at 06:13
  • @JZYL if we have a rough surface, and a very high Reynolds number, then yes, we can't get much laminar flow even with a favorable pressure gradient. I guess I didn't stress that enough. But laminar flow airfoils are designed to maximise the extent of the favorable pressure gradient, assuming of course that everything else is fine. Also, notice that boundary layer suction is a method to keep laminar flow from transitioning. – Abdullah is not an Amalekite Jun 20 '20 at 07:46
  • @PhilSweet I'm referring to A favorable pressure gradient is required to maintain laminar flow. Laminar flow airfoils are designed to have long favorable pressure gradients. It's right below your quote. – Abdullah is not an Amalekite Jun 20 '20 at 07:50
  • @PhilSweet I agree that turbulent boundary layers care much less about adverse pressure gradients. That's the reason why they delay stall, as you mentioned. – Abdullah is not an Amalekite Jun 20 '20 at 07:53
  • @Abdullah Not really. You don't need a "very rough surface" to trigger TS wave naturally. TS wave will be triggered on a smooth surface on a flat plate with zero incidence against the flow. – JZYL Jun 20 '20 at 11:29
  • @JZYL But that still doesn't invalidate the fact that favorable pressure gradients are stabilizing for laminar flows, does it? – Abdullah is not an Amalekite Jun 20 '20 at 11:38
  • @Abdullah If you're talking about turbulence transition, then no, the shape parameter would usually continue to increase through the favorable pressure gradient. In any case, the local Reynolds number would continue to increase, so it would not be "stabilizing". If you're talking about how separation, that's a completely different topic than what I've taken issue with. – JZYL Jun 20 '20 at 12:29
  • @JZYL so that quote is wrong? – Abdullah is not an Amalekite Jun 20 '20 at 15:06
  • @JZYL also, I most certainly am not talking about flow separation. – Abdullah is not an Amalekite Jun 20 '20 at 15:09
  • @JZYL Are you saying that by "maintain laminar flow" they mean "maintain attached laminar flow"? – Abdullah is not an Amalekite Jun 20 '20 at 15:10
  • @Adullah I'm not going to speculate on some random source. You can consult reputable books by Anderson or Drela on boundary layer, and they would indicate similar results on turbulence transition. An adverse pressure gradient will definitely promote laminar separation bubble at low moderate Reynolds number (as a source of natural transition), but it's not a necessary mechanism. – JZYL Jun 20 '20 at 15:53
  • 1
    Guys – why don't you ask a new question? It's actually quite simple: An adverse pressure gradient slows speed in the main flow direction only, leaving crossflow speeds untouched. So these get higher relative to main flow speed which helps transition along. And regarding the "myth" of no lift loss from early turbulent transition: Just ask any owner of a glider with the Wortmann 67-170 airfoil and they can tell you it is anything but a myth. Explaining all this with sufficient depth won't fit here, though, therefore a new question would help. – Peter Kämpf Jun 20 '20 at 16:52
  • 1
    On swept wings there is no need for the adverse gradient. The changing flow direction in the boundary layer is fully sufficient to trip transition. You should probably add that your answer is only valid for straight wings. – Peter Kämpf Jun 20 '20 at 17:27
  • @PeterKampf No one mentioned anything about lift loss, or swept wing, here. This discussion is limited to whether "creating turbulent flow on a smooth wing needs an adverse pressure gradient." is a correct statement. No, one does not need an adverse pressure gradient to transition. That's all. – JZYL Jun 20 '20 at 19:54
  • @JZYL Phil wrote in a comment "The old lift loss myth is just that—a myth." I addressed my comment to him and meant that false statement. No need to be upset. – Peter Kämpf Jun 20 '20 at 20:38