The “RefitTrim” tool is very inaccurate and almost always produces surfaces whose deviation is way too bad, including border that’s moved away from its original position. I often try to use it and the majority of times I simply give up and use “MoveUVN” instead. Or I simply trim the surface and accept the fact that the trimmed edge will lead to multi-span adjacent surfaces. If possible, I use “Match surface” with the “OnSurface” option to minimize the amount of spans and control points (compared to matching to a trimmed surface edge), though a robust “RefitTrim” like the one in Alias and ICEM would be the best way to do surfacing.
I think this is normal whenever you refit/“Icem Trim” at curve boundaries. Also in Icem you always need to double check the other continuities once you do that. You can minimize the error by simplifying the boundary.
In any case the difference is, that in Icem you can quickly rematch with minimal deviation in case this is the case. But you should also double check if you use MoveUVN.
In general, a manual corner fillet is always more work. No matter if you do that fast or not. So there will always be people wanting the One-Click solutions, and those who want/need a proper solution. For me doing corner fillets remains a post-process. At least with that in mind you can switch to other tools or minimize the effort if you work organized enough. Too many people want to see perfect models on early design stages, and I think this is also part of the problem. Therefore I would rather like to see a robust automation producing garbage corner blends. Honestly as long as its water-tight, it would be ok. Only for the final stage I would do the manual process once.
Oh, I have partial answer to the original question in this thread as to why “Solidworks usually fillets easily”. I’ve been back and forth with Siemens support a fair bit over the export behavior in NX, and as I understand it, they roughly do this:
Fillets in Parasolid (the geometric kernel in NX/Solidworks) aren’t really Nurbs fillets until you export as STEP. What you see in the CAD software viewport while editing is their own secret sauce (note, not talking about the mesh here… they have their own “recipe” for the math as well). Basically, Parasolid remembers what you wanted to do for the fillet for each edge and shows you a representation of it, but the underlying traditional math never gets implemented (ie converted to B-spline surfaces) until the moment you export. And if you export a Parasolid native file, you get all surfaces except the fillets, for which only the parameters are included. Kinda neat!
6 Likes
Hi eobet, that’s really cool I never new about that. Does that mean that another software that reads that Parasolid can Edit those fillets? Does this happen between which softwares?
Thx,
G
The simple answer is that the fillets that SWX makes so easily are true rolling ball constant radius fillets.
All this talk about about G2 blends is pure nonsense. Anybody reading this thread who has ambitions to make CAD models for an automobile manufacturer would be wise to become competent at making true fillets. In the hundreds of parts I have worked on for major automobile manufacturers not once has their ever been any G2 blends. I have made tens of thousands of rolling ball fillets on auto parts because that is what the manufacturers want on their parts. And the same thing holds true for parts made for other industrial sectors like aerospace, transportation, machinery, mining, energy, agriculture, etc.
True fillets are far more accurate than goofy blends. If a part drawing requires rounding or filleting corners with a specified radius the manufacturer has the expectation that will look the same regardless of which CAD program was used. True fillets are easy to edit and easy to measure to confirm that they are accurate.
Wake up folks, Boeing is not going to hire you to model the outer shell of airplane if you are unable to model a small strut on the landing gear correctly.
Here is the true fillet solution for when the radii on the top surface line up.
True_Fillets.3dm (99.6 KB)
3 Likes
I would assume so. What I can say for certain is that what I wrote answers how NX is able to resize or delete fillets imported from Rhino. By running the “optimize/simplify” command, NX identifies everything that looks like a fillet, removes them from the model and replaces it with Parasolids own procedural (their word, not mine) representation which is easy to edit.
And yes, this applies only to circular G1 fillets, so it doesn’t answer why NX/SW is so much better with G2, chord or even variable fillets. I just thought it was an interesting implementation detail.
3 Likes
that make sense, feature recognition is a wonderful tool for many MCAD tools, but yeah it’s always only on basic geometry.
Last month we released finished tooling geometry to a new client and they said: “Beautiful model, but it’s so weird, Solidworks did not recognize a single fillet, arc or extrusion in the model, I cannot change anything” (…there was not a single fillet, arc or extrusion in the model, that’s why it looked so beautiful 
)
G
6 Likes
The requirement to work with fillets applies mostly due to the fact that exact radius fillets offer the same deformation of sheet metal along the entire length of the bent area. This alone reduces the stress and fatigue of the material. This also applies to the majority of composite panels for airplanes where the thick carbon-fibre weave could be damaged if the bending is way too sharp.
Another reason why fillets are so popular in many industries is the time/cost ratio. Fillets are quick and efficient to CNC-machine with round tip tools, whereas G2 blends require multiple runs every 0,1 or less to look good. We already had a similar talk in another thread:
However, the exterior composite panels for cars do not use thick carbon-fibre weave, so they could take advantage of G2 blends to enhance the looks of the car (mainly sports cars)
3 Likes
@jim Am not sure this area of the model is transitioning at the highest level possible. This may or may not what you were aiming for.
Maybe I’m misreading where you were going, are you saying that CAD software can’t do higher than G2?
1 Like
What I’m telling you is what I know from personal experience.
What I’m saying is that on the vast majority of manufactured parts that go into the products made by large mfg. companies like Ford, GM, Honda, Caterpillar, Volvo, Boeing, GE Cummins, Alstom, Siemens, etc, the people paying the bills do not want to see any G2 blends.
As far as I know in The US there is no large manufacturing company that actually makes their own parts. Parts are made by suppliers (usually a chain of suppliers) and all parts have to conform to specifications for fit, form and function. Sure you can make G2 blends, but if you do, you may have your work rejected because that is not what they asked for. What they want is something they can measure and passes quality control.
First of all, its impossible to manufacture a part with light reflections as shown in your image. Even on an auto body part its impossible to form sheet metal and paint it and end up with light reflections that transition like that.
1 Like
@jim Yes there’s a difference between production, highly engineered models that need to be LCD. That’s not what the crux of this discussion is about. There are times when understanding what the highest order “can be” that helps with modeling skillset overall.
There are some car companies, and this is speaking from personal experience, that do what to start with the highest level and back down in areas to be more practical.
Yes the zebra stripes were exaggerated a bit it was to show that the area circled is not as optimized, it’s practical, which is again, not what most of the thread is about.
If you want to talk about the automotive industry, when car companies start full scale in clay they aren’t worried about the practical in terms of things having to be exact. They’re doing a form study. When Formula 1 cars are looking for things to be as aero dynamic as possible they’re looking for the highest degree possible. So it’s not just about practical, LCD only, this is about how to achieve a desired result that is as close to the designers intent as possible. Then the engineers get their hands on it and then it get butchered… 
K, but all your radii are 5mm except the one that’s 14.2539053 mm 
It’s funny too, cause my $42k truck has body panels that don’t even line up and aren’t even symmetrical.
They deviate probably +/-0.200" or more in some areas.
So I don’t even know why ClassA is a thing when I doubt ppl actually achieve it in the end.
Most products aren’t even deburred properly – if at all.
It’s sad when hardly any manufacturing is going on these days, and the production that’s happening is low quality and designed to fail in less than 5-10 yrs.
But we’re worried about the polynomial iso curvature of surface fabric geometry of fillets – instead of “how can we help the manufacturers add fillets easily and quickly”.
You think manufacturers that decide to do away with deburring completely, care about the polynomial math behind the darn fillets? or G1,2,3 when they’ve already given up on that and just use G0 and leave burrs all over everything?
Show me a product made out of plastic or sheet metal that doesn’t have sharp burrs or edges…
Gee I wonder why manufacturers are making short cuts when it comes to fillets lol.
Not to mention the signs of mold-sprew-port-fill zones, jagged mold seams etc., where they just ‘chop off’ the connection and leave signs of it or use a low quality grinder/sander process to barely mitigate the burr in half a second.
I think the main problem is most designers don’t seem to know how stuff is actually made.
They think the G1,2,3 continuities are actual manifestation in reality and tolerances can be +/-0.00000000000" lol.
Most true craftsman just want a dam round edge ffs. And if they can’t get it, they bring out the sandpaper.
And in the case where Ford and the autobody shop can’t make my body panels line up symmetrically, I might have to bring out some wrenches and hammers or hack saw cause G,123 is a fantasy.
2 Likes
The discussion centers on the question - how come Solidworks and other CAD programs can make standard fillets easily while Rhino can’t?
The answer is Rhino can make the same fillets that SWX and others make. Its just that Rhino users have to make those same fillets one fillet at a time and the Rhino users have to tell Rhino where each one goes and so that makes it far less easy to do the same thing in Rhino.
Whenever, this question is asked, there are always a few trolls that show up telling everybody that nobody needs true fillets. Tell that to the many millions of CAD users who are making a living creating models with true arc fillets.
1 Like
Yeah, that’s right.
If that construction were shown on a blueprint, its dimension would be designated as “R”. The R means whatever radius is required to make it line up with the 5mm radius
like this
1 Like
@jim On this we agree, it’s about why can’t the automation of fillets be done in Rhino like it is in Solidworks. Not about production or manufacturing… that is a different topic was my point.
I don’t think anyone in here has said that no one needs true fillets, so where is the trolling happening?
@lander Much of this depends on the industry that you’re in, production amounts… etc. I’d be willing to bet that 99% of truck owners could care less about Class A surfacing or what it nets as a result, their main thing is functionality. You can dam sure bet that in F1 Racing it’s a big deal.
Hence, that’s not the same thing.
What’s a “true” “fillet”?
Sounds like an overstatement.
What manufacturing process requires your “true arc”? +/-(how many units)?
Show me a GD&T/Blueprint-Reading reference that backs up this nonsense.
The vertical distance between those centers appears to be free. Therefore “R” is free.
Assuming the vertical alignment of centers isn’t free.
Guess I’d have to assume other constraints from that image.
I’m curious if you have a blueprint reading book that proves that annotation strategy isn’t nonsense.
I’m hoping u prove me wrong.
I probably should go dust my books off too 
I have consulted for the automotive and transportation industry from Japan, France and Germany, and tier-1 consumer product manufacturers in Europe and South Korea, and they all take formal aesthetics seriously. And they don’t back down from anything, because a design’s formal aesthetics are part of the company’s DNA, and so their suppliers have to stick to the rules. Not once was G2 or G3 surface continuity rejected. I can’t speak for the U.S., maybe it’s different there.
2 Likes
lol, it all boils down to reality. Reality requires tolerances.
“true fillet” is literally a fallacy of logic, a ‘strawman fallacy’. Define “true fillet” and how it differs from a ‘fake fillet’?
You mean like ‘tolerance magnitude’ written on the blueprint?
Right. Except when they’re fixing something that breaks due to a hard inside corner that lead to a fracture because someone failed to add an inner-round fillet to there manufacturing process.
Or also when they get cut by a sharp edge from something they had to deal with due to a hard corner that someone failed to round off with an outer-round fillet manufacturing flaw / short-cut.
K, how much of a big deal? Typical F1 Racing ‘true-fillet’ ‘class A’ R value GD&T deviation blue-print callout? +/- how many units deviation from G1,2,3?
I bet they use some dude skilled with sandpaper, and they dgaf about the fillets in your CAD file.
Prove me wrong.
Oh, n’ btw, how many F1 racing engines have perfect ‘true fillets’ on every hard outside and inside corner and to +/- how many units of deviation?
And how many engine blocks in any racing industry or any automotive industry have failed due to a crack forming on an inside or outside corner due to design/manufacturing shortcomings?
Ultimately, point being, a bad fillet is better than no fillet.
Even a bad fillet can prevent utter failure in the long run.
Whereas no fillet, is utter failure. Even a chamfer can be interpreted as a fillet with a few kinks.
I wonder if this same error would occur via “chamfer”:
I’ll check it later. For now I’ll be running some deburring toolpaths on my HAAS, which btw don’t require fillets on the CAD model.