Wow those are some numbers.Terry Duncan wrote: ↑27 Oct 2019 19:30Having just run a very basic version of this into Springsharp I got the following returns for the basic hull with armour, no engines or guns at present;
Cost: $1,775,858 million.
It has a very high roll at 13.5 seconds, is an unsteady gunplatform at 14% (ave = 50%) and has the seaboat qualities of 0.08% where 1% is average.
Obviously these figures could be trimmed a fair amount, but on the positive side it will take 800k 6" shells or 1,800 torpedoes to sink!?
My intuition about a free program like this is it's based on correlation with historical designs and has basically no underlying economic analysis of factors like economies of scale.
This is why I prefer a fundamentals-based approach from the main parameters of drag, propeller thrust, bending stress, etc.
That said, let me interrogate your inputs a bit:
It says a displacement of ~2mn tons. Is that empty displacement? Does the program break it into armor and hull (assuming no guns and armor as you said).
Basic arithmetic can tell us the weight of the armor I propose; this is nearly an order of magnitude above what that basic arithmetic indicates (~1250x300ft citadel, 2ft thick belt and deck armor, 490 lbs/ft3 steel density = ~300k tons citadel).
Hull weight would, I posit, be around 10% of gross displacement. This may be where the program is most off if it doesn't account for the fact that moment of inertia of a beam increases with the fourth power of its linear dimensions. It is for that reason that the structural efficiency of a ship (empty/full ratio of weight) improves dramatically with size. Do we know the basic structural/static engineering parameters of this program?
This is the kind of thing I'd want transparency on to credit a model...Terry Duncan wrote:Probably to destroy all reserve bouyancy, the program is not clear on that sort of thing. I believe it is worked on the overall strength and size of the structure.
A program that doesn't account for non-linear physical and economic factors that vary with size can easily create exponential errors when extrapolated beyond its sample set, even if it's fairly accurate within that sample.