An impressive CFD simulation on a Cray XE6 supercomputer

Discussion in 'Hydrodynamics and Aerodynamics' started by daiquiri, Nov 2, 2012.

  1. daiquiri
    Joined: May 2004
    Posts: 5,373
    Likes: 247, Points: 73, Legacy Rep: 3380
    Location: Italy (Garda Lake) and Croatia (Istria)

    daiquiri Engineering and Design

    Just saw this video of a CFD simulation of a flow around a transom-stern vessel: http://www.youtube.com/watch?NR=1&v=cenIBqda8zA&feature=endscreen

    The qualitative result is really impressive, and the flow features appear to agree very well with those of a model tested in a towing tank (also shown in the video). Too bad they don't say anything about the numerical results, forces and moments. How do they compare to the towing-test data?

    The simulation is done on a Cray XE6 supercomputer, here is the webpage of the beast: http://www.cray.com/Products/XE/CrayXE6System.aspx
    I am reading some data there: 50 kW of power absorbed and 5000 cu.m/h of airflow for cooling, for every cabinet. It means that a beast made of 8 cabinets shown in the brochure requires 400 kW and 40.000 cu.m/h of airflow? Wow, those folks have to pay some nice electric bills at the end of the month. ;) A 20 hrs computation should cost, how much, a $1500-$2000 just for the electric power consumed?

    This video has left me pretty much stuck. Evidently we now have the computing power to perform qualitatively excellent full-NS calculations around vessels travelling at high displacement speeds. But what about the quantitative side of the results? Do we have now a reliable tool for prediction of flows around transom-sterns, and the associated forces and moments?

    Anyone here who might have a first-hand insight into the state-of-the-art of this type of hydrodynamic research, and who could tell us something about the level of accuracy attained so far?

    Cheers
     
  2. J Feenstra
    Joined: Jan 2012
    Posts: 79
    Likes: 6, Points: 0, Legacy Rep: 100
    Location: The Netherlands

    J Feenstra Junior Member

    that electric bill really is significant, I always hear people say that CFD is cheaper and faster than model testing, but this really put's it in another perspective.

    Just for prediciting the restistance I believe that CFD is rather accurate, I'm currious to know if people use CFD for prediciting seakeeping capabiltiy
     
  3. daiquiri
    Joined: May 2004
    Posts: 5,373
    Likes: 247, Points: 73, Legacy Rep: 3380
    Location: Italy (Garda Lake) and Croatia (Istria)

    daiquiri Engineering and Design

    Flow around a transom stern has always been a problem so far, that's why I was asking. Cheers
     
  4. Leo Lazauskas
    Joined: Jan 2002
    Posts: 2,696
    Likes: 148, Points: 63, Legacy Rep: 2229
    Location: Adelaide, South Australia

    Leo Lazauskas Senior Member

    Another very beautiful simulation!
    It seems to be a recent refinement of one I pointed to several months ago...
    http://www.youtube.com/watch?v=A-0fVE_pPNg

    I agree with you that it would be nice to see some predictions of forces and
    moments to go with the nice pictures. My alter-ego "Ferocitus" commented
    on that in the original Youtube video but got no reply.

    From my point of view, it is important to emphasise that it is asking a lot
    from simple linear codes (e.g. Michlet, Hullspeed, Hydros etc) to even come
    close to that level of detail for the near wave field. However, given the
    uncertainties in measurement, those theories, combined with some empirical
    equations, can do quite well in predicting forces and moments.

    Furthermore, linear codes can do very well in predicting the wave field very
    far from the ship, say > 5 shiplengths back, which the Cray simulation might
    struggle with, unless it reverts to linear methods for that part of the field.

    Note, too, how there is no splash at the bow. I appreciate that they are
    concentrating on the stern flow, but splash and unsteadiness at the bow
    would make for a very messy flow near the stern at high speed. They seem
    to have side-stepped that complication. For good reason!

    So, yes, it is "state-of-the-art", but IMO it's still a long way from modelling
    "reality". I have read in some other forums that this group is also interested
    in simulations for video games and movies. For those applications,
    pathological accuracy and force calculations are irrelevant.
     
  5. Ad Hoc
    Joined: Oct 2008
    Posts: 6,610
    Likes: 617, Points: 113, Legacy Rep: 2488
    Location: Japan

    Ad Hoc Naval Architect

    I think Leo summed it up nicely.
    Impressive..but jury is still out.

    From Leo's thread on latest ITTC, here is their state of the art, as such on CFD:

    Pages from 03-2_Page_1.jpg Pages from 03-2_Page_2.jpg
    Pages from 03-2_Page_3.jpg Pages from 03-2_Page_4.jpg

    Trouble is, it is no different from the fancy plots of FEA of composite materials in their through thickness. They look great, but it is impossible to verify. Since to have a strain gauge in the layers, is not exactly easy (the gauge being generally thicker than the cloth!!), and the mere presence of it alters the strain anyway. :eek:
     
  6. daiquiri
    Joined: May 2004
    Posts: 5,373
    Likes: 247, Points: 73, Legacy Rep: 3380
    Location: Italy (Garda Lake) and Croatia (Istria)

    daiquiri Engineering and Design

    AH, it would be interesting to compare the costs of the two approaches (towing tank vs CFD) too.
    Let's put aside (just for a moment) the costs relative to the purchase of the equipment and the construction and maintenance of the test/computation facilities in two cases. How much, based on your experience, could a similar test program cost when performed in a towing tank?
     
  7. daiquiri
    Joined: May 2004
    Posts: 5,373
    Likes: 247, Points: 73, Legacy Rep: 3380
    Location: Italy (Garda Lake) and Croatia (Istria)

    daiquiri Engineering and Design

    Good observation, I take note of that. It looks as the hull in the simulation is a light displacement like a shallow sailboat hull, or a moderately high LDR hull, like the A-Yacht: http://www.youtube.com/watch?feature=player_detailpage&v=vbucJO3zDCE#t=7s so it makes no significant splash at the bow in the speed range considered. As you said, evidently they wanted to concentrate on transom-stern flow only.
     
  8. Leo Lazauskas
    Joined: Jan 2002
    Posts: 2,696
    Likes: 148, Points: 63, Legacy Rep: 2229
    Location: Adelaide, South Australia

    Leo Lazauskas Senior Member

    AdHoc might give you some idea of costs, but I think it depends on what you want.
    For example, do you just want video of the flow filed, or do you want
    laser-scans of the wave elevations on a fine grid so you have the elevations
    in a digital format? Many tanks won't have that capability.
     
  9. Leo Lazauskas
    Joined: Jan 2002
    Posts: 2,696
    Likes: 148, Points: 63, Legacy Rep: 2229
    Location: Adelaide, South Australia

    Leo Lazauskas Senior Member

    And the transom stern is not all that deep. I suspect that a larger transom
    might make for a less attractive video because there would be a lot more
    wave-breaking and viscous debris flying around.
    In any case, I'm glad I'm not paying the electricity bill!
     
  10. daiquiri
    Joined: May 2004
    Posts: 5,373
    Likes: 247, Points: 73, Legacy Rep: 3380
    Location: Italy (Garda Lake) and Croatia (Istria)

    daiquiri Engineering and Design

    A nice example of theory of relativity applied to fluid dynamics. :p

    I get your point Leo. I am not trying to find out which method is ultimately better - I'm sufficiently smart to understand that there is no such thing as "the best". What I am trying to understand is - how reliable and accurate is the CFD nowadays, with present computational power, and where does it becomes competitive with (or can even substitute) classic towing tank tests.

    Your reply has given a hint about one aspect of fluid dynamics research where CFD is indeed valuable - the visualization of flow details. The great videos posted by Mikko Brummer in the thread http://www.boatdesign.net/forums/hydrodynamics-aerodynamics/ultimate-validation-44809.html are another fine example of this visual delivery capability of CFD. Very hard to achieve such level of visual information with empirical methods.
     
  11. Leo Lazauskas
    Joined: Jan 2002
    Posts: 2,696
    Likes: 148, Points: 63, Legacy Rep: 2229
    Location: Adelaide, South Australia

    Leo Lazauskas Senior Member

    I can only point you to the latest ITTC reports because that is by far the
    best chance we have of getting some objectivity on the matter.

    Those reports have some minor errors, but they really are a terrific summary
    of current techniques and procedures. IMO they are the most significant documents
    available on hydrodynamics and experimental methods.

    I agree completely. It is just a case of choosing the right tool for the job.
    Using Cray's at the preliminary design stage would be an expensive over-kill.
     
  12. Ad Hoc
    Joined: Oct 2008
    Posts: 6,610
    Likes: 617, Points: 113, Legacy Rep: 2488
    Location: Japan

    Ad Hoc Naval Architect

    Exactly.

    We were forced by a client to use their preferred test tank. I wont mention their name.

    This was almost 10years ago, they charged €50k for simple tank testing. I was in attendance, and during the higher speed runs, there was a huge bow wave and all hell broke out....they (test tank) went back started doing endless CFD simulates about changing the entrance angle and then the customer rep was up in arms..blah blah blah....and on it went..was a nightmare, politically.

    Solution, simple, just alter the waterline, so she is "trimmed aft"...ie move the lcg/lcb aft....problem solved. Their CFD dint tell them that.....just plain old common sense :eek:

    BTW..our preferred test tank is some 70% cheaper!!
     
  13. daiquiri
    Joined: May 2004
    Posts: 5,373
    Likes: 247, Points: 73, Legacy Rep: 3380
    Location: Italy (Garda Lake) and Croatia (Istria)

    daiquiri Engineering and Design

    You mean, around $15k ? Does it include the costs relative to the creation of the model hulls?
     
  14. Leo Lazauskas
    Joined: Jan 2002
    Posts: 2,696
    Likes: 148, Points: 63, Legacy Rep: 2229
    Location: Adelaide, South Australia

    Leo Lazauskas Senior Member

    I can imagine that the experimental program that BMc and his colleagues ran
    for their "Sea Train" would have cost much more than that.

    Simple methods (e.g. Flotilla and other linear codes) can do fairly well in
    predicting the resistance and wave patterns for such thin hulls and relatively
    gentle air cushions, but I wouldn't trust computer simulations of the behaviour
    of that long, articulated vessel in waves.

    Sorry, Slavi, but I can't answer your question more precisely.
    Sometimes you can save money by relying on CFD or other codes, sometimes
    you have to resort to tank experiments. In both cases there are risks.
     

  15. daiquiri
    Joined: May 2004
    Posts: 5,373
    Likes: 247, Points: 73, Legacy Rep: 3380
    Location: Italy (Garda Lake) and Croatia (Istria)

    daiquiri Engineering and Design

    But you have already given some very good answers to my questions. What use can one make of them - it's up to reader's intelligence. ;)
     
Loading...
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.