A few Michlet/Godzilla questions

Yes I know that. I did read some parts of the manual (and I promise I'll read the rest).

The thing is that in this case I don't want to put such a restriction in place because I'm getting better results without it. I just checked again with a Series 1 compared to a Series 7 for a speed of 6 knots.

The Series 1 goes to a rectangular profile with parabolic waterlines and a prismatic of 0.67, on a length of 6.10 metres with 2.64 square metres of wetted surface.

The Series 7 goes to an almost rectangular profile but with the reverse curves in the ends of the waterlines and a prismatic of 0.59, on a length of 6.51 metres with 2.59 square metres of wetted surface.

At 6 knots the Series 7 has 3.5% less total resistance, which is a significant amount. Furthermore, because it has 2% less wetted surface (despite being quite a bit longer) and a lower prismatic it should still be better even at lower cruising speeds.

tl:dr; version is that the hollow waterlines seem to be a good thing for this application.

 

My advice now would be to optimise for a range of speeds, e.g 5.5kts, 6.0kts and 6.5kts, (i.e. 2.8 , 3.1, and 3.3 m/sec).

Leo.

 

I already did a bit of checking for 5.6 knots. Best hull I've found so far for that speed is only slightly different from the best for 6 knots, which sort of makes sense for long and skinny boats. The differences were a slightly lower prismatic and a slightly deeper hull. Length seemed to work out the same. I'd guess that for 6.5 knots a fractionally higher prismatic would do the trick. I'll take a look at it. TBH though, given that I'm not one of these terribly serious Finnish supermen I'm not sure I'll ever be going much over 6 knots for long.

 

Don't be such a bloody goose!
You should be posting here that you are optimising for 8 knots, while secretly aiming for 6 knots at best.

 

[Doug] I'm gonna build this sucker out of kryptonite, fill it full of helium, and optimise it for 57 knots. [/Doug]

Back to Godzilla for a serious question. I've been checking things out and it's pretty obvious that when using a parabolic midship section the origin is at the DWL/Bmax point. This means a vee'd section when for minimum wetted surface on the same beam and displacement it would better to use a classic "right side up" parabola with the origin on the centreline at maximum depth (y=kx^2 and all that).

It's also pretty obvious that for tending to rectangular you're just messing with exponents on some basic function (since anything to the power of zero = 1 and my old man's a dustman and all that). For the centreline profiles I think there would be some advantage in flipping the origin too (so that as the ellipse tends towards rectangular the flattest part of the curve is along the keel rather than up the stem).

So, realising that you have saner and better paid things to do at the moment, how hard would it be for some mug to get into the source code and flip these bits around? The basic mathematics of it is not really a worry as I'm familiar with doing rude things to ellipses and power curves. Getting it into a runnable form may be a worry though. My coding experience is limited to php/html/css.

ETA: Just did some checking and it looks like you're not offering source code anyway, so you probably don't want people trying to mess around inside the code. I suppose what I am really suggesting is another hull series with "flipped" origins for the section and profile curves. Mathematically, this should be easy enough to arrange and it would give a much better basis for estimating recreational craft which cannot use a semicircular midship section. It would enable Godzilla to output lines that were extremely close to what was wanted for a real boat (ie: immersed sections pretty much parabolic-right-side-up with flare above the waterline).

 

There any many changes I could include, but Michlet is primarily a hydrodynamics workbench rather than a hull design program.

I take your point about having cross-sections that are parabolas rounded at the keel-line. The hull series I developed are convenient for hydrodynamic investigations. As you noticed, using a basis function and simple exponents allows me to generate ellipsoids, Wigley hulls, parabolic and elliptical struts, and many others. Very useful for benchmarking, but not much else.

I have also put in many new hull shape functions since version 8.07. (I can't even remember what was in 8.07 it was so long ago!)

For example, it is now possible to use over 40 shape parameters including constraints on the flare at the waterline. New objective functions also allow the user to search for a hull with the same principal dimensions (Displacement ,Length, Beam and Draft) as an existing hull, as well as demanding it to have the same section area curve and the same waterplane area curve. That makes it quite good at representing a real hull with a reasonable number of mathematical functions.

Godzilla can go from the weird initial hull (in the first two pics) to the smoother realistic hull in the last two frames.

All I need now is time and money to get around to updating the manual!

Cheers,
Leo.

 

Sounds great. I'll wait for that one then.

 

Hi Leo,

I've been playing with version 8.07 and can get Series 8 hulls to run fine in both Michlet and Godzilla, but cannot get either Series 9 or 20 to run, or even open with Godzilla. Are these types supported in this version or do I need to look to a more recent version?

Thanks

Mark

 

There was an extended version kicking around on this board a couple of years ago that had series 9 hulls but I can't find it here any more. It also allowed a wider range of viscosity so cold water boffins could play around with it. Sorry, but I don't support old versions so I can't send you a copy.

Hull series 20, 32, and 42 are in version 9.2 which I haven't released into the public domain yet.

Have fun!
Leo.

 

Bump. Got another question. Might as well go in the same thread.

I started applying a viscous drag form factor, since some graphs you showed in another thread had good correlations with experimental results by using Scragg and Nelson's empirical form factor in combination with the IITC line.

Now if I'm applying a viscous drag form factor I would expect the viscous drag to increase. The factor works out at 1.13 for the hull I'm interested in, so I'd expect a roughly 13% increase in viscous drag. Instead, the out.mlt file shows viscous drag increasing by 3-2%. Ok, I can live with that.

I wouldn't expect the wave drag to more than double for exactly the same hull. It does though. Just by introducing a viscous drag form factor of 1.13 the wave drag jumps from 3.71 N to 7.99 N at the same speed.

This seems a bit funny to me, to put it mildly. Yes, I have got the right factor. Wave drag factor is still set to 1.0.

Got any ideas about this?

ETA: This is with Michlet 9.20. I have double and triple checked that the viscous form factor is the only thing being changed. I have noticed something else too: when I view the calculated viscous resistance graph while Michlet is running, I can clearly see the two components graphed (resistance and form factor) and they add up to what I'd expect (about 13% more than without the form factor), but for some reason this is not being reflected in the out.mlt file.

The huge increase in wave drag is shown in the graph inside Michlet.

 

If the wave drag was only a small percentage of the viscous drag, this is to be expected. The new hull will be shorter, beamier, and deeper than an unconstrained hull with lower form factor. If the viscous drag was 10X wave drag to begin with, when the new form factor is applied, a hull with 1% less viscous drag would be "better" if it had less than a 10% increase in wave drag. If you went from 1.13 down to 1.03 in viscous drag, that would allow for roughly a doubling of wave drag.

In case Leo shows up, I have a little request of my own. When trying to analyse a hull with a beam limit and high stability requirement, the program becomes quite inefficient because beam is a calculated residual quantity. It would be nice to be able to toggle between beam-as-residual and draft-as-residual in order to better work with beam constraints.

NoEyeDeer, I think this might have some bearing on the b/t < 6 isssue you mentioned earlier. Very small changes in t result in too large a change in b and the method loses robustness. I have found it much easier to get stuck in local, nonoptimal solutions with high b/t.

 

You're missing the point. I realise that increasing the viscous form factor will result in hulls optimising to a shorter length, and that the resulting hull would naturally have a higher wave drag as a trade off to save some wetted surface. The basic concept there is a no brainer IMO.

What I am saying is Michlet itself does odd things even when I am not running an optimisation series through Godzilla. I realise they both rely on the same basic equations to perform their calculations.

The problem is this: say I start by running an optimisation series of hulls through Godzilla to get my benchmark. This gives me a "best" target to aim for when designing the lines for an actual boat. So I get those lines sorted and run them through Michlet to calculate the predicted drag. An actual example is this:

U (m/sec) 3.076102,
Rr 0.003711
Rt 0.036246

Then I decide I want to see what happens when I apply a viscous drag form factor of 1.13, without making any other changes at all. I do that and run the exact same boat through Michlet. The result is this:

U (m/sec) 3.076102,
Rr 0.007990
Rt 0.040747

Splitting this down to extract the viscous drag, we'd get 0.032535 in the first case (no form factor) and 32757 in the second case (viscous drag form factor 1.13). In other words, although the viscous drag form factor has increased by 13%, the calculated viscous drag as outputted to the out.mlt file has only increased by 0.68%. This seems unreasonably low.

Despite no alteration at all being made to the hull lines though, the calculated residuary resistance has increased to 215% of the residuary resistance wthout the viscous drag form factor. Note that in both cases the wave drag form factor is left at the default 1.0.

 

There might be some mistake with the residuary resistance, but it might just be sloppy "accounting" on my part.

Have you looked at the value of the form drag in the table, i.e. the column headed Rform? Do the values look correct?

Incidentally, I wasn't going to include Scragg and Nelson's form drag at all because it can be very sensitive to the value of the entrance and exit angles.

Cheers,
Leo.

 

Well TBH I don't know what the values of Rform are supposed to be. When I use the default viscous form factor of 1.0, the Rform is shown as 0 in all cases.

When I use a form factor of 1.13, the Rform values seem to vary roughly as the square of the speed (judging by a quick scan and some mental arithmetic).

ETA: Ok, I broke out the old Hewlett Packard HP28S (bless its little heart). Rform is actually varying as about U^1.83, which would make sense for viscous drag given that it's what I'd expect for U² modified by the Reynolds Number effect.

I've attached a zip with the in.mlt and out.mlt files for both form factors (1.00 and 1.13).

 

Incidentally, I don't really need help with this as such. It's not hard to start without the form factor (in which case Michlet seems to be fine) and then do a bit of manual recalculation with a few geosims to take the increase in viscous drag into account. I've already done that anyway for the hulls I'm currently interested in. I just included the files in the last post so Leo had an example. Debugging is always easier if you can see exactly what is going on.