sail aerodynamics

Discussion in 'Hydrodynamics and Aerodynamics' started by Guest, Mar 21, 2002.

  1. tspeer
    Joined: Feb 2002
    Posts: 2,319
    Likes: 303, Points: 83, Legacy Rep: 1673
    Location: Port Gamble, Washington, USA

    tspeer Senior Member

    It sounds like you're talking about the two principal approaches, the Lagrangian approach using particles (like DualSPHysics), or the Eulerian approach that divides the flow into boxes (like OpenFOAM). There are open source and commercial codes that use both approaches. A lot of development has gone into these, so there's no need to start from scratch if your objective is to get engineering answers. Of course, if your objective is to have fun programming the code, the by all means have at it.

    The problem is the number of simultaneous equations that have to be solved in order to get the flowfield resolved adequately to get an accurate answer, especially in the boundary layers around solid surfaces. Coarse meshes add numerical dampening and don't capture enough of the detailed interaction within the flow.

    Your grid will need to be a lot finer than that in the direction away from the surface.

    Look up "wall functions."

    The resolution needed for the direct simulation of turbulence needs to be very fine indeed. You won't be able to do it on a practical scale, like a whole boat. even just doing it for a segment of a sail or foil requires a LOT of computation.

    Turbulence is usually modeled by adding an additional equation or two to the boxes that account for the general level of turbulence in the box without actually simulating the turbulent motion. The turbulence effectively changes the viscosity (damping) of the flow and is convected downstream with the flow. The equations account for the convection, as well as the production and destruction of the turbulence as a result of local flow conditions. It's as if turbulence is something added to the flow, like marker dye, that needs to be tracked and can dissipate or become concentrated as it goes.

    You might want to think of the three conservation laws - conservation of mass, conservation of momentum, conservation of energy - and how they affect the flow. When the flow is curved, there must be more pressure on the outside face of the box compared to the inside face of the box in order to generate the centripetal acceleration of the turning flow. If you then stack the boxes up from the inside of the turn to the outside, you have a gradually increasing pressure. Eventually, you arrive at the outside boundary of your flow, where the pressure is equal to the ambient pressure. That anchors the value of the pressure. Now, as you retrace the boxes inward, you get a pressure that progressively decreases from the ambient pressure, and when you arrive at the surface of the sail, you'll find low pressure there.

    If you do the same exercise starting from the windward side of the sail, you arrive at the opposite boundary, where the flow is also at ambient pressure. But now, when you trace the boxes back toward the sail, the pressure is increasing and you end up with high pressure at the windward side. So you have low pressure on the leeward side and high pressure on the windward side, all from the fact that the sail is curving the flow from the apparent wind direction.

    As to why the low pressure occurs near the luff, you have to look at the whole flow picture and see where the flow is being curved most strongly, which is at the luff. You also need to trace the boxes in the streamwise direction, and you'll see that the flow is accelerating into the low pressure point and decelerating away from there. At both the windward and leeward boundaries, the flow is again at ambient pressure.

    The value of the pressures in all the boxes has to be consistent with the local turning of the flow, the acceleration of the flow flowing from one box to the next, and the boundary conditions at the edges of the flow domain.

    Getting all that to balance out is exactly what leads to the high computational cost. It requires millions of cells to adequately represent a 3D flow, and the computational cost is a power function of the number of cells.

    I should add that just creating the 3D grid, with appropriate resolution in the right places (like near the surfaces) is not an easy thing in its own right, and is where the majority of the labor is involved. Probably the most useful thing you can do is to work on the grid generation and use an existing code to solve the flow. Then use an open visualization tool (like Paraview) to examine the results.
     
  2. Paul Scott
    Joined: Sep 2004
    Posts: 584
    Likes: 106, Points: 43, Legacy Rep: 84
    Location: San Juan Island, Washington

    Paul Scott Senior Member

    Well, the farfield is where dragons be, especially given the continual adjustment of a rule of thumb for infinity. But I'm still grumbling at the insistance by lazy mentats that if you multiply the square root of 2 by the square of two = 2. If you go out enough decimal places, it gets closer....

    I found a website with the square root of 2 to over a million decimal places, but multiplying that by itself gets into the same problems you are describing.

    Maybe a sojourn through chapter 10 of Doug McLean's Understanding Aerodynamics might help?

    Edit- Tom :p beat me 2 :p it......
     
  3. David Cooper
    Joined: Jan 2015
    Posts: 167
    Likes: 14, Points: 18, Legacy Rep: 42
    Location: Scotland

    David Cooper Senior Member

    Thanks Tom - you've provided me with a high density of quality information there which may save me a considerable amount of time, and you've pointed me in the right direction for a number of things I would have taken a long time to find otherwise. One of the things I want to explore with my simulation is whether pattern recognition can reduce the amount of processing required - there may be a lot of shortcuts that can be taken that won't reduce the quality of the results, but time will tell.

    Thanks too to Paul - I've put the book in my wishlist at Amazon, but I'll try to get it through the library first.
     
  4. Erwan
    Joined: Oct 2005
    Posts: 460
    Likes: 28, Points: 28, Legacy Rep: 97
    Location: France

    Erwan Senior Member

    Thanks for this interesting discussion.

    Especially this point:

    Turbulence is usually modeled by adding an additional equation or two to the boxes that account for the general level of turbulence in the box without actually simulating the turbulent motion. The turbulence effectively changes the viscosity (damping) of the flow and is convected downstream with the flow. The equations account for the convection, as well as the production and destruction of the turbulence as a result of local flow conditions. It's as if turbulence is something added to the flow, like marker dye, that needs to be tracked and can dissipate or become concentrated as it goes.

    It triggers a dummy question, considering a modern tear-drop mast like an A-Cat
    Trying to guess the contribution of the windward turbulent separation bubble to the profile drag (2D).

    Is there any additional equation which could be used for this purpose ?
    While lee side laminar separation bubble and reattachement are very well documented
    for windward side turbulent bubbbles I could not find much doc.

    Best regards


    EK
     
  5. brian eiland
    Joined: Jun 2002
    Posts: 5,067
    Likes: 216, Points: 73, Legacy Rep: 1903
    Location: St Augustine Fl, Thailand

    brian eiland Senior Member

    Attached Files:

  6. tspeer
    Joined: Feb 2002
    Posts: 2,319
    Likes: 303, Points: 83, Legacy Rep: 1673
    Location: Port Gamble, Washington, USA

    tspeer Senior Member

    2D programs like XFOIL and MSES can calculate the drag of the windward separation bubble. It's really no different than the leeward bubble, just different in size and located in the slower flow speed of the windward side.
     
  7. David Cooper
    Joined: Jan 2015
    Posts: 167
    Likes: 14, Points: 18, Legacy Rep: 42
    Location: Scotland

    David Cooper Senior Member

    Can the resolution not be generated automatically? Where neighbouring boxes have near identical content, there's no need to subdivide them, but any that are sufficiently different could be turned into eight smaller boxes, and if those become sufficiently similar again later on they could return to being a single box. That would also make it easy to adjust the set of sails and modify their shape (design) with the grid resolution evolving continually to keep up with the changes so that you're never doing any more processing than necessary. (The part downwind of the boat needn't be treated the same way though as it's of less interest.)
     
  8. DCockey
    Joined: Oct 2009
    Posts: 5,229
    Likes: 634, Points: 113, Legacy Rep: 1485
    Location: Midcoast Maine

    DCockey Senior Member

    Sometimes called an "adaptive grid" or "adaptive mesh". At least 40 years old, probably considerably older. Sometimes worthwhile but sometimes the overhead in moving data and extra calculations is more than the savings.
     
  9. DCockey
    Joined: Oct 2009
    Posts: 5,229
    Likes: 634, Points: 113, Legacy Rep: 1485
    Location: Midcoast Maine

    DCockey Senior Member

    David Cooper, perhaps you are not aware of the huge amount of work and innovation that has occurred in the last 70 years or so in modeling fluid flows using computers. If not you might want to spend some time reviewing it. And that work is based on decades, actually several centuries, of work with equations (algebraic, differential and integral) to describe and model fluid flow.

    Numerical methods for modeling flows need to model the physics with sufficient accuracy, be stable, and converge to an answer. The tricky part in developing such methods is an improvement in one area may cause problems in another.
     
  10. David Cooper
    Joined: Jan 2015
    Posts: 167
    Likes: 14, Points: 18, Legacy Rep: 42
    Location: Scotland

    David Cooper Senior Member

    By finding my own solutions to problems, I think I'll have a better chance of finding something new: when people spend most of their time following what other people have done before, they can easily all miss the same things. I have a number of off-the-wall ideas which I want to try out which may cut the amount of calculations that need to be done, and I have a track record of finding new things which other people have missed for decades. If it turns out that all bases have already been covered though, it won't matter: my main priority is just to learn how the physics of it all actually works at the lowest level, and working through it from scratch is a good way to learn. I'm ready to start writing code now and I'll just see how it goes.
     
  11. philSweet
    Joined: May 2008
    Posts: 2,682
    Likes: 451, Points: 83, Legacy Rep: 1082
    Location: Beaufort, SC and H'ville, NC

    philSweet Senior Member

    A historical perspective helps though. You could spend your entire life trying to reach the state of development that existed in 1947. The modern computer was invented to do fluid flow problems. The first program written for a Von Neuman machine was a Navier Stokes solver. Prior to the machine's construction, the solver was modeled using a loop of punch card machines with custom mechanicals. It also helps to know that off-the-shelf computers are intentionally crippled when it comes to these sort of problems. This has been true since the early 1950s. Americans didn't want anyone with an early IBM office mainframe to repurpose it for weapons development. In addition to monopolizing nuclear materials, key technical resources were controlled, including CPU design and instruction sets, in such a way to prevent accurate fluid modeling. Early machines did most of the calculations with hardware and instruction sets were massive. They can still outsolve modern crippled RISC machines operating millions of times faster clock speeds.
     
  12. Doug Halsey
    Joined: Feb 2007
    Posts: 640
    Likes: 212, Points: 53, Legacy Rep: 160
    Location: California, USA

    Doug Halsey Senior Member

    Unless you're dealing with hypersonic vehicles, you're not going to be finding any areas of "empty space".
     
  13. David Cooper
    Joined: Jan 2015
    Posts: 167
    Likes: 14, Points: 18, Legacy Rep: 42
    Location: Scotland

    David Cooper Senior Member

    That's only because the air moves in to fill the space so fast. If you want to understand the chain of cause and effect events, you have to recognise that the space that's being opened up (and filled as quickly as it opens) is the main driver of the action on the leeward side of the sail. If you take a sail shaped piece of wood and drag it through a tray of sand in the direction of the leading edge of the "sail", the biggest hole in the sand appears behind the back part of the sail. The effect of that is hidden quickly when a sail moves through air because air molecules are able to move into that space at enormous speed, and when you see where the lowest pressures end up you can easily imagine that the back part of the sail isn't doing anything useful and that you could do away with it, but, as some of you have already been saying in this thread, it is where the real power comes from. That doesn't necessarily mean an aft-mast rig with a very short "mainsail" is automatically going to be less effective though: if it has the same leech return it should still be generating a sizable "hole", and if you have a couple of sails ahead of it, the middle sail will be doing some of the work in generating that "hole" that a longer main would do.
     
  14. markdrela
    Joined: Jun 2004
    Posts: 307
    Likes: 30, Points: 28, Legacy Rep: 324
    Location: MIT Aero & Astro

    markdrela Senior Member

    Yes, it's called "Adaptive Mesh Refinement", or "Adaptive Gridding", or other similar terms. People have worked on this for at least 2 decades.

    Adaptive refinement does slow down the computation quite a bit, because when you change the grid you in effect have to generate a new solution, and this needs to be repeated a number of times. You do have the previous-grid solution as a good starting point for the new grid, but that does not help as much as one would think. For example, improving the grid near the trailing edge, or anywhere on the airfoil upper surface for that matter, will tend to change the overall lift by some amount which has a global effect on the entire flowfield.
     

  15. markdrela
    Joined: Jun 2004
    Posts: 307
    Likes: 30, Points: 28, Legacy Rep: 324
    Location: MIT Aero & Astro

    markdrela Senior Member

    It's important to realize that in a low-speed flowfield, everything affects everything else. There is no well-defined "chain of cause and effect". The lift on the main affects the lift on the jib, which affects the lift on the main, which affects the lift on the jib, etc. Likewise, the a separating boundary layer affects the airfoil surface pressures, which affect the boundary layer, which affect the surface pressures, etc.

    All well-posed CFD methods must address these global mutual influences in one way or another. Explicit Navier-Stokes methods like Fluent or Open-Foam do this by simultaneously evolving the entire flowfield in time, or pseudo-time, or "iteration count". Viscous/inviscid methods like XFOIL or MSES do this by repeatedly linearizing the entire numerical problem, and solving the resulting large linear system for all the viscous and inviscid variable changes simultaneously. These methods do not assume any sequential cause/effect chain between regions of the flow.
     
Loading...
Similar Threads
  1. Revboat
    Replies:
    28
    Views:
    5,770
  2. Forecaddie
    Replies:
    5
    Views:
    763
  3. Forecaddie
    Replies:
    9
    Views:
    1,631
  4. Forecaddie
    Replies:
    14
    Views:
    1,952
  5. Remmlinger
    Replies:
    45
    Views:
    5,869
  6. dustman
    Replies:
    25
    Views:
    3,832
  7. JerryWo
    Replies:
    4
    Views:
    1,330
  8. BMP
    Replies:
    12
    Views:
    1,503
  9. Howlandwoodworks
    Replies:
    2
    Views:
    1,601
  10. jmf11
    Replies:
    11
    Views:
    2,933
Forum posts represent the experience, opinion, and view of individual users. Boat Design Net does not necessarily endorse nor share the view of each individual post.
When making potentially dangerous or financial decisions, always employ and consult appropriate professionals. Your circumstances or experience may be different.