Hello again,
This time I used curve network to create a surface, and later on tried insert knot and remove knot, move uvn to adjust the surface. However when i turned on curvature graph, it showed the isoparm is not in good shape, there is some “pointed region” indicated that it does not go smooth , but I smoothed the U direction points and looks even already. May I know why and how to fix? thanks
interesting
not sure how much more you could smooth it, but doesn’t look done imo:
rebuild srf:
I just pick the neaby region surface points and smooth by U with move uvn command
yeah, that works to an extent. that’s one of the first ways I learned to do smoothing of srf’s
unfortunately Rhino’s GUI wont let the user smooth U and V simultaneously with those sliders.
technically ‘rebuilding the srf’ does do some smoothing due to the tolerance factor etc.
I usually prefer to rebuild srf’s so that their UV’s are more squarely/evenly spaced.
So I usually tend to ‘balance’ the UV densities.
I rarely use rebuild at this stage as I think it is changing the shape out of my control and no more at the shape of guide curves, but I always stuck here and maybe it is the only way
part of the reason I rebuild and make the density balanced is to keep the shape more stable during these other transformations.
thereby I think it helps control any deviations.
so compare original density:
with rebuilt density:
imo, the rebuilt version shown here, will keep the shape more stable during smoothing.
there are weird density near the boundaries which I do delete temporarily in some workflows, and then add them back later with another rebuild stage.
for example:
‘removecontrolpoint’
I highly recommend this approach for high precision situations that might require more meticulous procedure:
honestly though imo, the curvature graphs don’t make very much sense because it seems like different smoothing techniques can make the graph look “worse”.
but here’s kinda what it looks like after messing around with the points a little bit:
At the end of the little bit of time I put in the transformation, I did try to throw in a ‘fitsrf’ stage but rhino didn’t really let it work unless I used 0.01 tolerance so I just didn’t use it.
Surface curvature problem_emod.3dm (2.3 MB)
There’s still lots of improvements that could be done, but depending on design intent, the best thing might to be to break the srf back into a crv netwrk and renetwrk the srf at some point.
it really depends on the objectives 
Yes that looks alot better, I may need to in depth study the rebuild command, of the input value i set.
So after drag and move the surface points, we should always rebuild the surface, is it?
I highly recommend the ‘rebuild’ stage in many workflows, that I know of.
I think alot of users are used to dealing with all kinds of deviations anyways – imo.
so, the rebuilding technique imo is just simply another way to handle what’s going on.
plus you can always go back and forth from different scenarios.
some users like to do things like ‘fitsrf’ but I’m not a fan of that really, although it’s good to know sometimes if you really really want to simplify a srf as much as possible.
while I don’t like it cause the srf becomes more unpredictable afterwards when moving control points.
Thank you for reply in detail. 
np anytime. Rhino is fun! 
Maybe, in such cases, rather than performing time consuming CP gymnastics, the original shape should be revolved from a simple degree 6 single span curve, and then transformed to make it elliptical?
It depends on the tolerance. I want to simplify it as well if the simplified shape meet the tolerance…just like command “ fitsrf”.. It won’t simplify anything if shape changes are not accepted.
