Running the MAXVCG command, I was getting LIMIT margins of more than 1000% at the high displacements. The run file uses 3 perfectly valid area limits with relatively small required areas.
So I began thinking what method of failure would trip up the process of raising the VCG
until the righting arm margin returns a zero.
I had already looked at an individual case to see if I could best the MAXVCG algorithm,
sort of a rematch of Kasparov vs Deep Blue. I tried one of the high-margin cases with a
VCG slightly above the reported maximum. To my surprise there was a FAIL in LIMIT AREA from EQU to FLD or RA0. So why were the margins reported by MAXVCG so high???
Upon close inspection of the table it was clear that a water-tight opening near the aft end of the
vessel was immersed at equilibrium causing the sudden failure. Due to increasing VCG, the trim of the vessel increased to the point where the tight point was immersed even at upright. (A large range of LCG values were used as part of a series of MAXVCG evaluations that were later combined into a single composite curve.)
Figuring that the very large margins reported by MAXVCG represented the case with a VCG just low
enough to prevent the tight point immersion, it seemed it would be impossible to find a case which had precisely the required amount of area under the righting arm curve. Just then in a blinding flash it came to me: what I need is a limit that is sensitive to the vessel trim, that way as the tight point approaches the waterline, the limit margin will decrease. I found the FLOODHT limit type which could directly track the height of the tight point. That would make it clear in the MAXVCG results
what was governing. By requiring a positive flood height, it would allow a
small safety margin rather than teetering on the edge of failure.