A few Michlet/Godzilla questions

Discussion in 'Software' started by NoEyeDeer, Nov 13, 2010.

  1. NoEyeDeer
    Joined: Jun 2010
    Posts: 983
    Likes: 32, Points: 28, Legacy Rep: 300
    Location: Australia

    NoEyeDeer Senior Member

    I've done a search but can't find the answers in the existing threads. Questions are:

    1/ Leo stated elsewhere that Michlet should preferably not be used for L/B < 6 or B/T > 6. In qualitative terms, what sort of inaccuracies occur outside this range?

    2/ Are there any quantitative guides to how fast the inaccuracies escalate?

    I'm trying to avoid GIGO here. :)

    Also, I was trying to use Godzilla to work through some shapes that have the LCB aft of midships. It doesn't seem to like this and keeps saying all hulls have the LCB at 0 and all hulls violate the imposed constraints, meaning it wont generate any results. I'm using this under "First hull constraints":

    # Minimum LCB
    -0.15
    # Maximum LCB
    -0.14

    So question 3/ What am I doing wrong here?
     
  2. Leo Lazauskas
    Joined: Jan 2002
    Posts: 2,696
    Likes: 155, Points: 63, Legacy Rep: 2229
    Location: Adelaide, South Australia

    Leo Lazauskas Senior Member

    1. and 2. I don't know of any systematic studies that can answer these questions. At best, all you can hope for is that Michlet will correctly order the hulls with respect to their resistance. Thus, if Michlet predicts that hull 1 has lower total resistance than hull 2, that is most likely to be found experimentally, even if the actual predicted resistances are in error. This is also generally true for CFD and several other methods.

    3. It could be for several reasons.
    You have a special hull constraint set so that only symmetirc hulls are generated.
    Or it might be that you are using a hull series that always has LCB=0.

    I'd need to see the input file because it could be something else.

    Leo.
     
  3. NoEyeDeer
    Joined: Jun 2010
    Posts: 983
    Likes: 32, Points: 28, Legacy Rep: 300
    Location: Australia

    NoEyeDeer Senior Member

    Wrong hull series was it. Thanks.
     
  4. NoEyeDeer
    Joined: Jun 2010
    Posts: 983
    Likes: 32, Points: 28, Legacy Rep: 300
    Location: Australia

    NoEyeDeer Senior Member

    Got another one for you. I've been running some hulls which have the LCB slightly aft of midships (around 52-53%). This of course also means that the maximum cross section is aft of midships (around 58-59%).

    Michlet knows the areas of all the sections but it doesn't give an accurate prismatic. It apparently defaults to using a section at 50% DWL. This is fine for a normal large ship but not so good for some other hulls. It means, in the case of the hulls I'm checking, that for a hull with a prismatic of 0.593 Michlet will give the prismatic as being 0.617.

    This will obviously have an effect on wave drag calculations, and it means that comparing custom hulls with Series 7 hulls for optimisation becomes a bit hit and miss.

    Is there any way of making Michlet use the actual maximum section for calculating the prismatic?

    Com to think of it, does it matter anyway? Does Michlet use the prismatic in calculations at all or does it just go with the actual hull shape?


    ETA: For those that aren't aware, Delftship does the same thing. You have to set the midship location manually to get an accurate prismatic in Delftship. Easy once you know how (project settings).
     
  5. Leo Lazauskas
    Joined: Jan 2002
    Posts: 2,696
    Likes: 155, Points: 63, Legacy Rep: 2229
    Location: Adelaide, South Australia

    Leo Lazauskas Senior Member

    1. NO. The prismatic coefficient is NOT used in wave drag calculations.

    2. Sorry, but you need to manually calculate the prismatic coefficient based on maximum section area. Section areas are available in the .out file and also as a separate .csv file.

    Leo.
     
  6. NoEyeDeer
    Joined: Jun 2010
    Posts: 983
    Likes: 32, Points: 28, Legacy Rep: 300
    Location: Australia

    NoEyeDeer Senior Member

    Ok, no problem then. If it's not used in calculations for Michlet I can just get the prismatic from Delftship. Thanks.

    By the way, I found it interesting that the Series 7 hulls can be made as low in resistance as the Series 1 hulls, but the Series 7 hulls do it with a lower prismatic. This is even allowing for the displayed prismatic being slightly out in Michlet.

    A Series 1 will commonly optimise to around 0.66 but for the same Froude number (0.4) a Series 7 will optimise with a prismatic of around 0.58.
     
  7. Leo Lazauskas
    Joined: Jan 2002
    Posts: 2,696
    Likes: 155, Points: 63, Legacy Rep: 2229
    Location: Adelaide, South Australia

    Leo Lazauskas Senior Member

    It is crucial to remember that the "optimised" hulls could lie in a very shallow region of the search space. By this I mean that there could be a wide range of potential candidate hulls for your problem, and that their resistances might all lie within (say) 1% of each other. Other considerations could then be far more important and making it worthwhile to sacrifice 1% of the total drag.

    That is why it is important to put in as many constraints as you can. In many cases, these constraints (which are often a small subset of a statement of requirements) will cause the "optimal" hull to be quite different to those found with no constraints.

    Once you have a reasonable first (mathematical) hull that satisfies all your constraints, and has acceptable drag, you can proceed to the next step of the design spiral. There are many more notches and turns still left!

    Have fun!
    Leo.
     
  8. NoEyeDeer
    Joined: Jun 2010
    Posts: 983
    Likes: 32, Points: 28, Legacy Rep: 300
    Location: Australia

    NoEyeDeer Senior Member

    Yes I understand about the range of suitable hulls. As per your recommendations elsewhere I was running Godzilla with a range of seed values, and I made sure to give each option plenty of iterations to stabilise (100,000 seemed about right).

    In these cases (the infamous rowboat stuff) there was a lot of latitude for length, depth and prismatic. The only things I really had to constrain were displacement and (for a recreational boat) beam. Anything else could be left to do its own thing in the quest for lower resistance.

    What I found was exactly what you say: there isn't one standout hull but instead a range of fairly similar hulls whose resistance could be regarded as identical for all practical purposes. I did still notice some consistent trends though.

    1/ Series 7 hulls always seem to optimise with a lower prismatic than Series 1 hulls.

    2/ The best Series 7 hulls seem to be every bit as good in terms of resistance, despite the lower prismatic.

    3/ From this is follows that the Series 7 hulls would be a better choice in some ways. For instance, a lower prismatic for the same resistance opens up the possibility of increasing stability (larger midship section) without increasing resistance. This is assuming that the calculated results for Series 1 and Series 7 translate accurately to real life.

    4/ Moving the LCB aft of midships seems to consistently increase resistance very slightly. Mind you, it's only around 0.5% for LCB at 53% aft so it's not the sort of thing that is likely to make a difference in practice. For practical purposes the LCB can probably be put where desired without adverse effects.

    5/ Any significant rocker forward seems to increase resistance. This applied even when the rockered and straight hulls had the same length and prismatic. The best hulls in this speed range had almost straight keel lines.
     
  9. Leo Lazauskas
    Joined: Jan 2002
    Posts: 2,696
    Likes: 155, Points: 63, Legacy Rep: 2229
    Location: Adelaide, South Australia

    Leo Lazauskas Senior Member

    Michell's thin-ship theory is used to estimate the wave resistance.
    In this theory, it can be shown that the wave resistance is a minimum for fore-aft symmetric hulls. A small amount of asymmetry (because you have constrained the LCB) is unlikely to have much of an effect on resistance, except perhaps at very low Froude numbers.

    The wave resistance is also identical whether the boat is travelling forwards or backwards. This also holds true for multihulls. This is quite remarkable at first glance, because the wave patterns for a staggered trimaran are quite different when it is going forwards or backwards. Some Japanese researchers calculated the wave resistance using wave patterns measured in a towing tank and verified the theory.
    Of course, all this assumes that the effect of the boundary layer is small.

    I'd be surprised if the waterlines of the hulls you found are not parabolic (or two parabolas joined at the position of maximum beam).

    Leo.
     
  10. NoEyeDeer
    Joined: Jun 2010
    Posts: 983
    Likes: 32, Points: 28, Legacy Rep: 300
    Location: Australia

    NoEyeDeer Senior Member

    Then you're surprised. They have a slight hollow in the waterlines, like a truncated sine wave. In fact they look very much like the classic fast wherry: keel mostly straight, with short rounding in the forefoot and none aft. Waterlines basically parabolic-ish but with a slight reverse curve over the last one or two stations.

    That's just the way they came out of Godzilla. When you think about it they would have to be like this, since parabolic waterlines with a straight keel line will always give a prismatic of 0.667 for a longitudinally symmetrical hull. Any hull with a pretty-much-straight keel line and a prismatic less than 0.6 has to have waterlines that are not parabolic.
     
  11. Leo Lazauskas
    Joined: Jan 2002
    Posts: 2,696
    Likes: 155, Points: 63, Legacy Rep: 2229
    Location: Adelaide, South Australia

    Leo Lazauskas Senior Member

    The small indentation in the waterlines could be just a graphical inaccuracy. It might be less noticeable if you use more stations and waterlines.

    For Hull Series 7, the waterline shape is parabolic if the 1st and 4th shape parameters are equal to 1.0 (or close to it).

    Leo.
     
  12. NoEyeDeer
    Joined: Jun 2010
    Posts: 983
    Likes: 32, Points: 28, Legacy Rep: 300
    Location: Australia

    NoEyeDeer Senior Member

    I was using 21 waterlines and 81 stations. The hollow doesn't look like a graphical inaccuracy. Those are usually little jags in odd places, rather than being a definite reverse curvature that appears to be fair. Also, as mentioned, the prismatic indicates that the waterlines could not be parabolic. Even if they were a power curve that had no reverse curvature they would still have to have an exponent of 1.5 or less to give a prismatic under 0.6.

    ETA: If it's possible to import an out.mlt file into Delftship I could draft up one of the "optimised" hulls pretty quickly, but Delftship doesn't seem to like Michlet's offsets. I can export to Michlet but I can't seem to import. I'm happy to send you some files with the seed values I was using if you'd like to check them out.
     
  13. Leo Lazauskas
    Joined: Jan 2002
    Posts: 2,696
    Likes: 155, Points: 63, Legacy Rep: 2229
    Location: Adelaide, South Australia

    Leo Lazauskas Senior Member

    I have most of the implementation ready for Michlet to export to Delftship but it's not a priority for me. Sorry, but that's the downside of free stuff :)

    There's no need to send the files. It sounds like you are pretty comfortable with the program, and you are sensible enough to explore the search space instead of accepting a single design.

    Leo.
     
  14. NoEyeDeer
    Joined: Jun 2010
    Posts: 983
    Likes: 32, Points: 28, Legacy Rep: 300
    Location: Australia

    NoEyeDeer Senior Member

    Fair enough re the free stuff. I'm not complaining. :)

    I looked in the out.mlt for the best Series 7 hull I've found (so far) and the shape parameters are as follows:

    SHAPE PARAMETERS
    #, Hull 1
    0, 1.666645
    1, 0.556308
    2, 0.000141
    3, 1.356283
    4, 0.017607
    5, 0.537803
    6, 0.481221

    So as you can see the waterlines are a fair way off being parabolic. Don't ask me why. I just pushed the buttons. :)
     

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

    Leo Lazauskas Senior Member

    If you don't want hollows in the waterline, i.e. that they are always convex, you can constrain shape parameters 0 and 3 so that they are never greater than 1.0.

    Leo.
     
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.