Michlet-Info

Leo
Improving the resolution gives a more familiar shape.

If I was doing this for my own purposes I would also look at what a form factor might achieve given the blunt entry.

For what it is worth I also note that after about 8000 iterations the two hulls are not identical shape. I was not prepared to take the time to see if they actually arrive at the same shape.

Rick W

 

1. I usually run optimisations at least 10 times with different seed values. Sometimes they all converge to the same hull (+/- a small bit), sometimes a very different hull shape comes out of the blue, especially with the 32-parameter hull series.

2. I have built in some empirical form factor methods into newer version of Godzilla, but I haven't had time to release these with the free versions yet.

Leo.

 

If I am doing a hull that I intend to build I play with GODZILLA for a long while.

I typically have a very tight power constraint so I try to get as much benefit as I can before I start building.

I am happy with what can be done with 7 parameters. I expect that applying constraints usually have a greater impact than what could be done with more shape parameters. I always start with only power and displacement just to see what lowest drag looks like. I then close in on various constraints to determine what they cost.

As an example, if the cat in question here was allowed to go to the lowest drag length of 60m and beam down to 2.2m then the power drops to 1829kW. So the cost of the length constraint, in terms of power, is around 17%. I would sneak up on the length constraint gradually coming down from 60m. Likewise with the beam constraint.

Imagine trying to do this with tank testing or parameter analysis!

Rick W

 

Thank you for the work.
How much control do you have over the hullform?
Can you fix key sections to control the hullform better?
The hullform it spit out is in no way practical for the intended use. It looks like I will be going with the maxsurf/rhino/NAVCAD. Optimisation such as this is not a prime or realistic goal of our project, but the program is interesting.

Nick

 

Nick
I have attached the in.mlt file that has the higher calculation resolution that Leo suggested for your case. It produces a smooth resistance curve so a better result no doubt but is slower to do iterations.

In this one I have set the maximum length to 320m; effectively unconstrained. The minimum beam is also set so that this is not constraining the hull. You will see that there is provision to constrain a whole range of hull parameters for each hull between max and min values.

There are also 7 hull shape functions that can have their range set. If you change the shape function limits you need to make sure the starting seed is set within the range. I have noted with comments in the file a couple of the ones I limit. There is information on the others in the manual. Leo has versions with different numbers of hull shape functions. I have played with the 18 version and he likes his 32 version, which I have not tried.

Rick W

 

Nick
I do not use the exact result from GODZILLA for the final shape. I always make slight adjustments with fairing for example. I may also change the shape a little to suit construction from flat panels if that is the method being used.

After I have made the modifications I use Michlet to determine the drag of the end result to ensure I am not far from what GODZILLA produced. By that stage I usually have a good idea of what minor changes will cost.

Rick W

 

1. Yes, you can hold some sections fixed. E.g. to control the size of a transom stern, or a midship section.

2. Even though I wrote Michlet, I actually disagree with Rick - I don't think that you should be considering computer "optimisation" at this stage. It is far more important for you to grind your way around the design spiral by hand. Just my opinion.

All the best,
Leo.

 

I use the 32 parameter series to constrain, for example, the amount of rocker and/or the flare at the waterplane. I also use it to get a mathematical representation of existing hulls. That's not possible with some of the smaller parameter series.

For your applications 7 is probably enough to get a rough idea of the prinicipal dimensions and section shapes before fine tuning by hand to get the final design.

Leo.

 

Leo
When my tenth hull was a retrograde step I realised I needed something better as a basis of design than just using current hulls as a reference.

I recognised getting a handle on wave drag was a key requirement. If I had a year or two to spare I could sit down and rediscover my uni maths, study Michell's work and learn how to translate that for numerical solution. After that I could then build an optimising routine to sit on it.

However given my age and needs I prefer to fiddle with Michlet/GODZILLA. I agree Nick may get value out of getting more to the basics. The one thing I would suggest though is to always look for sound analytical methods of analysis in preference to empirical methods. If you can get an accurate model in terms of the basic physics it is far more powerful than extrapolating from the currently available data sets.

Rick W

 

And it shouldn't be forgotten that Michlet is NOT a ship design program - it's more of a hydrodynamics workbench.

I originally wrote it because it wasn't possible to do the things I wanted with existing ship design packages e.g. identify ships from their wave patterns, search for multihull arrangements that cancel waves in weird ways etc.

I agree that analytical methods are often better, but it is also a good thing to have a database of existing designs to start from.

Get back to work, slacker!
Leo.