Curvature Question

I'm perhaps overreaching my abilities here, but I'm attempting to develop a definition that will take a surface (or, ideally, multiple surfaces at once) and tell me, roughly, what percentage of the surface area is 1) Flat, 2) Curved in One Direction, or 3) Curved in Two Directions.

I've attached what I've got so far - a first step where I try to sort the sample points into 'developable' (singly curved) and 'non developable' (doubly curved).

What's confusing me now is the Gaussian curvature - when I check for equality (G=0 would mean a flat or single-curved surface at the sample point), I get apparently wrong results - in this case every point is shown as 'non developable' ---- However, when I use an approximate check (G≈0) it gives the result I expected... but I'm worried about the accuracy of using (G≈0) as the equation (what's the tolerance of ≈?)

I hope someone can help as I try to refine this definition. I've seen this great definition by LAN - http://www.livearchitecture.net/archives/3381 - but its purpose seems to be vizualization, and I want graphical AND numerical output.

Does anyone have any ideas?