Michlet

Discussion in 'Software' started by pavel915, Dec 8, 2009.

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

    Leo Lazauskas Senior Member

    Thanks for the thought-provoking post.

    The issue of roughness is a vexed one. There is no consensus on what friction line to use for perfectly smooth plates, so getting agreement on how to handle roughness is still many years away. Candries thesis has some interesting results. I can get the exact reference if you want.

    Just as worrying is the tangled web of turbulence stimulation. Different tanks use different methods, some follow ITTC procedures, some are ignorant of the fact that there are recommended standards. So for me it still comes down to which towing tank results do you believe, and which do you just accept out of hand because you assume they adhered to guidelines. It all seems a touch too random for mine at the moment. The world-wide facility experments might show just how random it is. :)

    But as you said, what can you believe if you don't accept experimental evidence?

    I appreciate what you have said about form factors, but you must also admit that they are far from universally accepted. For example, out of 8 respondents to a recent ITTC survey of towing tank operators, one used an unspecified method, one used the standard Prohaska method, and none of the others used a form factor at all, even for full ships.
    See: Table 5-3, Final Report of the Specialist Committee on Powering
    Performance Prediction, 24th ITTC, 2005.

    As I mentioned, F. Stern and the ITTC have done a lot of work to establish procedures for validating CFD and experiments. These include accounting for facility biases in physical quantities such as water density etc, carriage speeds and the whole shebang. These too are not without their critics. There are several ITTC documents discussing proposed procedures and their strengths and failings.

    Guilloton's method seemed like a huge advance at the time. Guevel et al, used it to predict the wave resistance of high-speed ships at the Workshop on Ship-Wave Computations, Bethesda, 1979. They admitted in the discussion of their paper that:
    1. It couldn't handle bulbous bows,
    2. It didn't handle flat bottoms and low draft hulls well,
    3. It was not valid for hulls with too wide a bow angle. For those
    hulls their results did not converge and the wave resistance was negative.

    An improved method was presented by the authors at the Workshop on Ship-Wave Computations, Japan, 1980. They including sinkage and trim and got reasonable agreement with some Wigley hull experiments.
    I don't know where the method went after that. Obscurity, I suspect, as did Keller's ray tracing method and several others.

    All the best,
    Leo.
     
  2. Guest625101138

    Guest625101138 Previous Member

    Pat
    If you get a chance to experience the thrill of a fast pedal boat you will ditch the bikes.

    Rick W
     

    Attached Files:

  3. Guest625101138

    Guest625101138 Previous Member

    See inserted notes.

    From my own developments I am more than satisfied with the value that Michlet/GODZILLA offers. It would be nice to see hull shapes morph in response to squat/lift/trim and ambient wave pattern but I guess that is a while off. Watching a hull morph as it goes through the GODZILLA optimisation is a learning experience in itself. It has given me considerable insight that would take an eternity to acquire building and testing hulls.

    As noted a couple of years ago it would be nice to have similar optimising routines for propeller designs as there are some gains still to be made here. I will live with what I have for now but with the knowledge it could be done better.

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

    Leo Lazauskas Senior Member

    Just a few more glitches in your "Min of the Desert" paper.
    Hope you don't mind the pedantry.

    Nomenclature:
    The symbol for rho has been rendered as a ? in the pdf.

    Consistency. Some quantities have units, e.g. [m/s], others such as GMt do not.

    Page 4: 2nd column, 2nd line up from bottom.
    Semicolon should be a comma

    Page 5: 3rd line of section 3.1(d)
    Is there a space in
    mo ment
    or is it just bad formatting because you used MS-Word or some other non-LaTeX muck?

    Page 9:
    1st column, 3rd para, 3rd last line
    due to: increased...
    Either use a colon and then semi-colons to separate clauses, or don't use it.

    1st column, last two paras and 2nd column, 2nd line: rho has been rendered as ? in the pdf.

    Thanks for the Min offsets!
    Leo.
     
  5. Leo Lazauskas
    Joined: Jan 2002
    Posts: 2,696
    Likes: 155, Points: 63, Legacy Rep: 2229
    Location: Adelaide, South Australia

    Leo Lazauskas Senior Member

    It's even weirder and more wondrous the more shape parameters you use. I have a 32-parameter series I use regularly and I did have a 200-parameter version of Godzilla on the Unix workstations at Uni a while ago. They are a bit of a nightmare to set up in the input file. Every parameter needs to be constrained, initialised etc. And sometimes the best hull under the given constraints and objectives ends up as quite a simple shape anyway.

    I think there are two main reasons you get good results.
    1. Your hulls are thin and smooth.
    2. You use a full-size hull in your tests so you don't have to frig around with scaling and all its attendant uncertainties.

    Funny you should mention that. Ernie Tuck said he'd supervise a thesis in either ship hydro or on optimising propellers. I chose the former and now have no time for the latter. Anyway, one has to leave something for the younger boffins. :)

    Leo.
     
  6. Guest625101138

    Guest625101138 Previous Member

    For the sake of getting to the bottom of the Formsys presented graph I have normalised my graph and subtracted viscous drag.

    In doing the normalisation I realised how small measurement errors can lead to misleading comparisons particularly at low speed so included an error band.

    The attached shows the measured data as an error band of +/-1% assuming a force measurement range of 20N. So an error in force measurement of 0.2N.

    Both the measured data and Michlet have the viscous drag excluded. This means the hydrostatic transom drag is included as wave drag in the Michlet data.

    This makes it clear how data can get distorted when it is manipulated in various ways. To publish experimental data without error bands is common but readers need to be astute enough to appreciate that experimental error exists and make due allowance for it rather than presuming it to be the holy grail.

    Rick W
     

    Attached Files:

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

    Leo Lazauskas Senior Member

    1% error bands? In your and my dreams!

    Take a look at Figure 2 of
    http://www225.pair.com/magic/ftp/Maxsurf/TechnicalPapers/SlenderBody_Couser.pdf (The figure is mislabelled - it is definitely not CT).

    The experimental error around 0.23 could be more than 300%

    Maybe now you can see why some mathematicians/hydroboffins don't trust experimental data, are suspicious of form factors, and why that data should not be used unthinkingly to discredit numerical methods.

    Cheers,
    Leo.
     
  8. Guest625101138

    Guest625101138 Previous Member

    Leo
    I was giving the experimental data favourable treatment. I would aim to get force measurement within 1% of measured range if I was working in a lab environment so used my acceptable standard for it.

    The interesting thing is that the experimental curve would be very suspicious if an error band showing negative wave resistance was presented. It would immediately make people question the experimental data or how it was manipulated.

    Subtracting a large component of the force then reducing the force to a coefficient based on inverse of velocity squared and plotting against a factor based on sqrt of velocity amplifies any error in force measurement at low speed. To make comparisons of the coefficients in percentage terms after doing all this manipulation (normalisation) leads to ridiculous conclusions in my view.

    As far as the experimental results are concerned a gnat's fart would be significant in terms of percentage variation of wave resistance at the low speed end. There may be some merit in normalisation for comparison purposes but there is also a need to stay grounded enough to know the order of magnitude of the measured data. You also have to wonder who is really interested in wave drag with Froude Numbers as low as 0.2 - what bearing will it have on the boat design.

    When I first looked at the Formsys curve I was more concerned with the variation at the high speed end but the omission of the transom drag explains this.

    I can understand your skepticism when it comes to experimental data. I trust my heart in assessing hull resistance. I even have a fair understanding of when that data become unreliable but from week-to-week and boat-to-boat the comparison is incredibly sensitive to small variation. Flotilla might give a bit more relevant data but Michlet is mighty close to what I measure.

    Rick W
     
  9. daiquiri
    Joined: May 2004
    Posts: 5,371
    Likes: 258, Points: 93, Legacy Rep: 3380
    Location: Italy (Garda Lake) and Croatia (Istria)

    daiquiri Engineering and Design

    How do you perform the calibration of the measuring instrument?
     
  10. Guest625101138

    Guest625101138 Previous Member

    I don't get too hung up about the absolute value because I want to detect minor improvements.

    For absolute calibration I have a few methods. Preferred method is to spend an hour or so on a recumbent cycle machine monitoring heart rate and power output. You need to compare a few machines to make sure they are consistent. I have found some to be poorly calibrated.

    Another is to walk up a long steep hill near my home using the GPS for speed and elevation. I eliminate the effort to move at the same speed on flat ground so only look at the increase related to climbing for both bpm and power. This gives my W/bpm over my baseline which is higher than my resting heart rate when I am on the boat. This correlates reasonably well with the recumbent machines.

    The other is to climb the stairs in a tall building at a rate where the heart rate has stabilised. I assume all power is going into the lifting.

    But like I say I am mainly interested in comparative performance and my most reliable testing is done in calm conditions for periods of 2 to 3 hours similar to the chart posted earlier.

    I have been collecting this data for a number of years so I can go back through it and sometimes get old boats out to confirm where I have come from.

    There are certainly variation as I gain fitness so there is a need to recalibrate as I train up. There is also variation after long periods at sustained effort when blood conditions change.

    On a particular day I can detect variation as small as 1 to 2% by heart rate. For example I know if there is a leaf stuck on the shaft strut. I know what wind drag costs. I know what ambient waves cost.

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

    Leo Lazauskas Senior Member

    From Rick's earlier post which I couldn't extract properly...
    Leo: I'm not sure which file you used as input to Michlet.
    Were experimental values of squat specified in the input file?
    Rick: Yes the in.mlt file did have squat included. I think they are identical to what was included in the Version 8.07 I have.


    But squat for a catamaran, in particular for the narrowly-spaced w/L=0.2 case, is different than for a monohull.

    Leo.
     
    Last edited: Jan 14, 2010
  12. Guest625101138

    Guest625101138 Previous Member

    Leo
    I used the in_npl.mlt and sploff1.csv (as txt) files supplied by Pat. I have attached them.

    Rick W
     

    Attached Files:

  13. patc
    Joined: Jan 2010
    Posts: 20
    Likes: 2, Points: 0, Legacy Rep: 36
    Location: Area65

    patc Junior Member

    lest keep on subject?

    personally I cannot think of anything worse. Riding a bike on the flat is bad enough. Cycling on a lake must be like riding a HPV down a motorway
     
    1 person likes this.
  14. patc
    Joined: Jan 2010
    Posts: 20
    Likes: 2, Points: 0, Legacy Rep: 36
    Location: Area65

    patc Junior Member

    yeah sorry vertical axis is incorrectly labelled Ct -- should just be resistance coefficient. Its essentially Cwp -- ie wave pattern resistance measure in the far field -- either by experiment or one of the numerical methods.

    Full dots are Ct - (1+k) Cf
    Which I called Cw -- of the part of the resistance that scales with Fn and not Rn
    k is determined from the measurements of wave pattern resistance at Fn > 0.5 or 0.6
    Remember that this work was on _High_Speed_Catamarans_ over the hump speed -- we were trying to improve the method for going from model to full scale. What was tending to happen was that not enough of the model CT was being scaled by Rn and hence the resistance at fullscale was being over predicted -- look at the work of Tony Armstrong for more info.

    So I would agree that the experimental uncertainty at 0.23 Fn, is quite high (but the experimental point are all the measurements -- with no filtering -- so you can see how much scaller there is in the are that was of interest to us at the time (0.5 < Fn < 1.0). If we were interested in the low speed end we would have used bigger models.

    cheers,
    pat
     

  15. patc
    Joined: Jan 2010
    Posts: 20
    Likes: 2, Points: 0, Legacy Rep: 36
    Location: Area65

    patc Junior Member

    It looks like you are comparing with the the Cw from experiment (the filled dots in Figure 2) which is
    Cw = Ct - (1+k) Cf
    with Cf from the ITTC57 model-ship correlation line

    I feel you should be comparing to the measured wave energy in the far field wave system (which is what the slender/thin ship method is supposed to be calculating) -- these are the empty dots

    Below about Fn = 0.4 the transom is running wet so probably you shouldnt be using a hydrostatic correction?

    In any case I cannot reproduce the Michlet data you present from the spreadsheet you emailed me. If I take Ct-Cv I get the blue line in the attached image. Note that I have included the full spreadsheet as an attachment in case anyone is interested.

    Now I really do have to get on with some work :)
    cheers,
    pat

    [​IMG]
     

    Attached Files:

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.