The use of computers in boat design

Discussion in 'General Computing' started by Stumble, Aug 12, 2011.

  1. MikeJohns
    Joined: Aug 2004
    Posts: 3,192
    Likes: 208, Points: 63, Legacy Rep: 2054
    Location: Australia

    MikeJohns Senior Member

    It's good to understand the problems associated with numerical methods but the computer has made my design work much easier over the years.

    I'm an over 50's greybeard and did my schooling and first degree with slide rules and log tables.
    Then the HP calculators arrived and suddenly slide rules were passé! We have moved on very rapidly from there.

    When the first PC's with decent spreadsheets and data storage we moved all our analysis and even hull design and analysis to spreadsheet. Weights and moments became much easier. The spreadsheet really shortened the initial design spiral, and still does for most of us.

    FEA in those days was mainframe computer time and very expensive and quite limited.

    When 2D cad became available it made drawing office staff redundant and designers could easily archive and alter drawings that saved a lot of time in later revisions and alterations.

    For structural design work I use a lot of safe load lookup tables from years before that I know I can trust and they are instant.

    For hull design I'm happy now using the computer for the whole process from scratch but I’ll often work from a sketch sitting on my desk while I define the lines on screen.

    I have no faith in CFD for hull drag analysis and can report several abysmal failures at drag prediction. Model testing is still in my view the only sensible tool to check a hullform and it’s not just the numbers that the computer spits out but just as importantly a look at the wave patterns peeling off that model hull at different Froude numbers.

    Verification seems to be often overlooked by computer jockies using numerical methods. I've seen huge and very expensive errors in shipbuilding projects because a team of young NA's didn't have the experience to know when the data coming out of the computer was obviously wrong, and the modern propensity to hire young people with limited experience.
     
  2. Leo Lazauskas
    Joined: Jan 2002
    Posts: 2,696
    Likes: 155, Points: 63, Legacy Rep: 2229
    Location: Adelaide, South Australia

    Leo Lazauskas Senior Member

    And I have little faith in either CFD or tank tests.

    If I was at the tank at the time of the test, I might be a little more confident. I have seen all manner of simple mistakes made in reports of tests. The latest ITTC study from 40ish tanks around the world doesn't inspire much confidence.
    Several tanks sent back reports with exactly the values the ITTC supplied as examples in spreadsheet templates.


    There has been a huge effort to emphasise V&V for ship CFD over the last 10 years or so, notably by Fred Stern at IIHR and others at INSEAN. Unfortunately, many people are just unwilling to put in the effort.

    I wish there more blind tests with CFD, e.g. supply the ship geometry and ask for results before releasing the tank data. That would sort out a lot of the CFD shills. I've only seen a couple of exercises like that, and some of the results were awful. Of course, once the tank data is made available, there is a lot of scurrying and tweaking to show that CFD could have got the results very accurately. :)

    Leo.
     
  3. MikeJohns
    Joined: Aug 2004
    Posts: 3,192
    Likes: 208, Points: 63, Legacy Rep: 2054
    Location: Australia

    MikeJohns Senior Member

    Leo that's the problem being an applied mathematician ;) If you can't predict it mathematically within 2% you don't trust it !

    I find tank tests much closer in the end design to any of the prediction algorithms which need significant tweaking to match the tank test data.

    CFD is not a tool for generating a resistance curve and I doubt it ever will be I've seen large models run on powerful workstations that chewed up 10 hours of processing time and failed to converge ! Then if the results do finally come out half sensibly, what do you tie them to?
     
  4. Leo Lazauskas
    Joined: Jan 2002
    Posts: 2,696
    Likes: 155, Points: 63, Legacy Rep: 2229
    Location: Adelaide, South Australia

    Leo Lazauskas Senior Member

    Personally, I love the methods and the mathematics more than the answers and whether anyone finds them useful.
     
  5. MikeJohns
    Joined: Aug 2004
    Posts: 3,192
    Likes: 208, Points: 63, Legacy Rep: 2054
    Location: Australia

    MikeJohns Senior Member


    I love watching the waves peel of the model and ripple across the surface. I don't trust grad div and curl in chaotic vector fields. :p
     
  6. Leo Lazauskas
    Joined: Jan 2002
    Posts: 2,696
    Likes: 155, Points: 63, Legacy Rep: 2229
    Location: Adelaide, South Australia

    Leo Lazauskas Senior Member

    Yes, it's best to trust tank results, like these from two different tanks. :rolleyes: (The ITTC left out the results that were completely flat!)
     

    Attached Files:

    • z.gif
      z.gif
      File size:
      51.9 KB
      Views:
      609
  7. Leo Lazauskas
    Joined: Jan 2002
    Posts: 2,696
    Likes: 155, Points: 63, Legacy Rep: 2229
    Location: Adelaide, South Australia

    Leo Lazauskas Senior Member

    Everybody believes tank tests, except the person conducting the experiments.
    Nobody believes CFD, except the person running the code :p
     
  8. Alik
    Joined: Jul 2003
    Posts: 3,075
    Likes: 357, Points: 93, Legacy Rep: 1306
    Location: Thailand

    Alik Senior Member

    Leo, I believe there is no conflict between tank testing and CFD in this topic: computers are used to conduct and process tank test data. I would only trust tank tests and CFD if they match with empirical methods and results of sea trials. At the end, we do not need to predict absolutely exact resistance, we need to get resistance with predictable/known accuracy margin.
     
  9. Leo Lazauskas
    Joined: Jan 2002
    Posts: 2,696
    Likes: 155, Points: 63, Legacy Rep: 2229
    Location: Adelaide, South Australia

    Leo Lazauskas Senior Member

    Of course, Alik. I was only poking Mike back!

    I also appreciate that for "standard" boats NAs have many ways of finding those margins.

    However, for innovative or unusual boats (e.g. UAV, SES, cats with unusual demihulls or tris, quads etc), that might not be so easy. Using sea trials with a completed full-size vessel is what you are trying to avoid by tank testing, and from what I have read of the tests used in the 26th ITTC, choosing the right tank can sometimes be a bit of a gamble. For example, the uncertainty for trim and sinkage can be around 40% or higher. Maybe even more for models under 3m in length.

    Using one tank test to show that a CFD calculation is reasonable, or way off the mark doesn't seem right. Nor is it right to use CFD to show that a single, random tank's procedures are sloppy.

    As Mike (and many others have pointed out), it's experience that is the key, and you can't get that in a 5 year Uni course.
     
  10. BATAAN
    Joined: Apr 2010
    Posts: 1,614
    Likes: 101, Points: 0, Legacy Rep: 1151
    Location: USA

    BATAAN Senior Member

    I wish that more young computer-dependent graduates would read Weston Farmer's book FROM MY OLD BOAT SHOP, even if it does not concern mouse use, but common sense in design.
    This doesn't mean to do things his way (60 years of designing mostly motorboats) but to acquire his habit of self-doubt and thoughtful analysis and never being quite satisfied, but always trying for simpler and better.
     
  11. Tim B
    Joined: Jan 2003
    Posts: 1,438
    Likes: 59, Points: 0, Legacy Rep: 841
    Location: Southern England

    Tim B Senior Member

    Thankyou, jehardiman and Leo. I am well aware of the numerical limitations of digital computers, but for 99% of typical computer usage they do not manifest themselves as a problem. This includes most of the analysis we do, until you get into high-end computational methods (eg. CFD with turbulence modelling). Even then I'd question whether the machine accuracy is actually the problem.

    The prevailing opinion seems to be that usage without experience is not good. Obvious really, the machine only does what you tell it. However, it does allow the user to explore more of the "problem space", and that has to be a good thing. Experience will only tell you what worked in the past (in the little bit of "problem space" you know about), it will not give you a radical (possibly better) solution to your problem (talking to people from other disciplines also achieves this quite well). That's where advances are made, but there are few people who stick with the radical solution for long enough to see the benefit.

    The computer, however, can be a tool to convince other people of the benefits of your idea by doing more simulation work than anyone 50 years ago could dream of, at much lower cost than producing physical models.

    Cheers,

    Tim B.
     
  12. Leo Lazauskas
    Joined: Jan 2002
    Posts: 2,696
    Likes: 155, Points: 63, Legacy Rep: 2229
    Location: Adelaide, South Australia

    Leo Lazauskas Senior Member

    We have converged on something sensible, then, Tim!

    Michlet/Godzilla is a good case in point. It can evaluate thousands of possible multihull arrangements in a few minutes but this would not be possible without some very clever numerical maths (devised by E.O. Tuck) because brute force (e.g. Simpson's rule) fail or take far too long. It would be very foolish to use Godzilla for anything except thin or slender demihulls and with sensible constraints, or the results are likely to be silly at best.

    The same can be said for CFD - it is a very useful tool for some numerical investigations, but it can also sometimes be inappropriate or risky for others. I'll go out on a limb and say the same applies to experiments with small models in finite width tanks.
     
  13. MikeJohns
    Joined: Aug 2004
    Posts: 3,192
    Likes: 208, Points: 63, Legacy Rep: 2054
    Location: Australia

    MikeJohns Senior Member

    The complexity of reality is the problem. For example with CFD we try and model an analogue complex system with a necessarily discrete model. Whether the result bears any resemblance to reality needs a very necessary validation for each new investigation.

    When CFD started to become available I had great hopes for it to be as suseful a tool as FEA/FEM is. I don't think it ever will be for something novel.

    But it does appear useful for predicting the effects of limited changes to already validated models.

    I saw one long and useless exercise recently which simply tried to reproduce tag test (flow field) indications from a flow tank around the control surfaces and stern of a powered vessel with a different design of kort nozzle. Despite a lot of expertise and frantic efforts from the software support desks there was no correlation. The same CFD package produces some really impressive graphics!

    So if at times you can't get CFD to match reality with any amount of tweaking how can you you possibly use it in the first instance with any reliability ? Certianly not for something completely novel, but rather for small alterations to existing forms.

    There have been some quite notable stuff-ups relying on cheaper CFD based testing.
     
  14. MikeJohns
    Joined: Aug 2004
    Posts: 3,192
    Likes: 208, Points: 63, Legacy Rep: 2054
    Location: Australia

    MikeJohns Senior Member

    Yes Leo I saw that in their report.
     

  15. Leo Lazauskas
    Joined: Jan 2002
    Posts: 2,696
    Likes: 155, Points: 63, Legacy Rep: 2229
    Location: Adelaide, South Australia

    Leo Lazauskas Senior Member

    The funniest one I saw (at your AMC) was some work on a Series 64. The student struggled to get sensible results and was told by the software suppliers that he needed to run the code in "linear mode" for hulls with large transoms. So, for that application that $100K code was no more sophisticated than the free version of Michlet.

    But, as you noted, the colours were very impressive!
     
Loading...
Similar Threads
  1. LP
    Replies:
    4
    Views:
    3,750
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.