Actually, a score of 100% "percentage of good geom" is not very common and 80% or better is considered a good result. Here’s the explanation:
TransMagic’s geometry engine is extremely precise; in fact the value that TransMagic considers zero (0) is .000001 mm. That’s a very small figure in the CAD world and in fact a higher precision working value than most engineering systems. So percentage of good geom refers to the percentage of surface-to-surface intersections that are precise to “TransMagic Zero” or 10e-6mm. Every single edge in the model could be precise to 10e-5mm and you would get a 0% percentage of good geom result after Full Repair – even though that’s still beyond most engineerings systems precision and still a very precise water-tight solid. We’ve never seen that happen but it’s theoretically possible.
TransMagic’s Full Repair function actually attempts to extend and re-intersect surfaces with the goal always being TransMagic Zero but that is not always possible. For example, near-tangent conditions can often not be resolved without actually moving surfaces in 3D space – a big no-no for a translation application – so we get those as precise as we can without moving them and then leave them alone. Another example is a spline surface whose trim boundary is exactly the same as its underlying surface domain. In that case we have no math upon which to extend the surfaces. So we get those as tight as we can but maybe not perfect TransMagic Zero.
So given that, we consider anything over 80% to be a high-quality, water-tight solid. Less than 80% and you could potentially have some downstream issues even if the file is still a solid. However, if you we’re going to CATIA V4 – which is a lower precision modeler – then even 70% would probably be fine. So that figure also depends on the receiving system but 80% is a good general figure to stick with….90% is great, 95% is excellent, 100% is perfect…but all would be considered high-quality water-tight solids – or in the case of sheet-bodies – high quality water-tight intersections.