# Godzilla Design Speed Averaging Problems

Discussion in 'Software' started by TJ Cameron, Apr 10, 2012.

1. Joined: Mar 2012
Posts: 16
Likes: 0, Points: 1, Legacy Rep: 10
Location: USA

### TJ CameronJunior Member

I am having trouble getting the design speed averaging routine to work properly. I need to optimize my sailboat design over a wide range of speeds and so this is a critical function.

The Godzilla examples all run fine at one speed. When 2 speeds are used the average Rt seems to be way off. I have attached 2 versions of Corey’s canoe file as an example.

ORIGINAL CANOE mlt file with 2 speeds:
# Number of Design Speeds (min 1, max 50)
2
# Design Speeds and Weightings (one pair per design speed)
3,0.5
5,0.5

Rt= 0.036

MODIFIED CANOE mlt file with 1 speed:
# Number of Design Speeds (min 1, max 50)
1
# Design Speeds and Weightings (one pair per design speed)
3,0.5

Rt= 0.040

Summary:
The single lower speed drag is being calculated with a higher resistance than the average of the same lower speed with a higher speed which is obviously incorrect. Am I doing something wrong or is there a problem with the code in Godzilla version V9.02?

For multiple runs of this file:
At 3m/s => Rt = 0.040
At 5m/s => Rt = 0.100

The averaged drag Rt result should be approximately Rt = 0.070 rather than 0.036. It seems as thought the calculation engine might be dividing by 2 when it should not.

Any insight into this issue would be greatly appreciated.

Best regards, TJ

#### Attached Files:

File size:
6.3 KB
Views:
335
• ###### inCanoe-2Speed.mlt
File size:
6.3 KB
Views:
364
2. Joined: Jan 2002
Posts: 2,696
Likes: 154, Points: 63, Legacy Rep: 2229

### Leo LazauskasSenior Member

When does this occur? Before running the optimisation? Or after? If so, I would need to see the output file too.

There are several mistakes in your input files...

Don't use blank lines, and start each line in the 1st column.
Use decimal values for the design speeds, not integers.
You have used 1.0 for the offset series type. The prgram expects an integer, e.g. 1

Those "mistakes" might not always cause problems, but if they do it can
be a pain to find. Also, if other errors are made, it might be difficult to pin
down where they occurred.

In the file you attached, the last line is
3,1.0
not
3,0.5

Good luck,
Leo.

3. Joined: Mar 2012
Posts: 16
Likes: 0, Points: 1, Legacy Rep: 10
Location: USA

### TJ CameronJunior Member

Leo,

Thank you for the tips. I went through my mlt files and made the corrections suggested and most of them began working properly averaging the Rt values. So yes, another case of user error!

In one file I discovered that the heal speeds were required to exactly match the design speeds for the speed averaging to work.

On a related topic: I have no comprehension of the wave elevation parameters. Did I miss that explanation in one of the manuals?

Are the wave elevation parameters settings going to have a significant effect on my design so that I should be concerned about them?

Thanks for the coaching. TJ

4. Joined: Jan 2002
Posts: 2,696
Likes: 154, Points: 63, Legacy Rep: 2229

### Leo LazauskasSenior Member

No, they don't affect resistance calculations.
They are only important if you are "designing" for minimum wake height, or
for inverse problems where you are trying to find the arrangement of hulls
that makes a specified wave pattern.

5. Joined: Mar 2012
Posts: 16
Likes: 0, Points: 1, Legacy Rep: 10
Location: USA

### TJ CameronJunior Member

"No, they don't affect resistance calculations."

Thank you, That is a relief.

I noticed that some (most?) starting seeds result in evolutionary dead ends which have Rt's which are 1 to 2% above the minimum Rt found with other seeds. Is there any way to avoid testing every possible seed to find the lowest Rt for a specific set of hull constraints?

Does a seed which converges on a low Rt faster than others have a greater likelihood of converging on the ultimate minimum?

I have limited time to learn new programs. What will I find more effective for me to use to go from Godzilla to hull model and back to Michlet?

I am considering Delftship and Freeship. Will I find Delftship Pro to be worth 150 EUR ?.

6. Joined: Jan 2002
Posts: 2,696
Likes: 154, Points: 63, Legacy Rep: 2229

### Leo LazauskasSenior Member

No. Godzilla is a stochastic optimisation technique. It would be impractical
to test all possible combinations for problems with a large number of
parameters. In some cases that would take billions of years.
The only advice I can give is to use a "reasonable" number of different
seeds and to look at the hulls and the drag estimated for those hulls.
Only experience can be your guide here. As I say in the manual,
carbon-based naval architects are quite good at this sort of thing

No. Some seeds will give very good results with a small number of
iterations, but then they struggle to do better for a long time. Others will
seem to do poorly for a long time and then suddenly improve.

I can't help much there. All I know is that Delftship accepts some Michlet
files and can produce Michlet input files.

There is a program called "Loftsman" which can produce offsets in a format
suitable for input to Michlet, but I don't know anything about it apart from
that. See page 3.42 of