Full Version: Root Fillet Example Model

Show messages:  1-20  21-27

Hi Michael,

Here is the model shown in the screencast from a different thread. I was able to get it to fillet with a .08mm radius. Anything bigger would fail to calculate volume. This is a Rhino 5 model. Let me know if it is helpful or if you have questions. It's very indicative of the kinds of problems I see often. The red solid can have a 1mm root fillet. The green one only .08mm. The end user issue is that it requires a lot of trial and error. I also spent a lot of time trying all kinds of ways of creating the solid prior to narrowing down the issue.

Some things that cause problems here: I used ellipses for the le and te. This seems to be part of the problem. The fillet has problems where the ellipse meets the next curve. The airfoils are projected onto cylinders in the green model. When Rhino does this, it screws up the curves. This seems to make the filleting a lot worse. I think if I cut all the curves and rebuilt them it would help. But this is a lot of extra work. For the non-projected airfoils, there is no issue. They keep their original structure and fillet fine. Had I used circular arcs for the le and te, I think it would help the filleting out a lot, but not sure. Basically, doing the root fillet has been a colossal pain in the arse with Rhino 5. There is no sure fire way of getting it to do what you want, that I have found. I follow the same basic build process but it randomly does not work. I think it might just be a matter of you can't fit a rolling ball fillet into certain geometry. So if the software told you that, instead of create crap geometry, that would be incredibly helpful. Even more helpful would be if it calculated and prompted you, saying this is the max fillet that can fit, do you want to use that value and proceed. That would fix the whole problem entirely. I've been living with this hell for 10 years. I know filleting isn't good with the nurbs kernel. It's been brought up a lot, so I didn't mention it. I was just trying to help the poster in the other thread. So that is why it's come up again.

I should note, with Rhino 5 at least, you can't have a seam in the area you want to fillet. So I have to rotate the seam out of the way. If you know why this is, I'd love to know. I'm not sure if MoI has that same issue. I have been using the calculate volume function at each step along the way to know if the solid is still valid. I have had a few occasions where what Rhino creates and exports as a step file is still corrupt, even if it did calculate a volume. I can't have corrupt geometry. I have to be able to mesh it for FEA. So this whole thing of filleting the root area of the blade, like the other poster wants to do, is still a big problem. It can be done, but not easily. I've lost many many hours to this over the years.

Anthony

PS; these are the links to the two screencasts mentioned in the other thread:

One; https://youtu.be/R5DF822U89w

Two (follow-up); https://youtu.be/jNkNNawmfJ0

---

Hi Anthony, over here filleting your model in MoI I seem to be able to get up to a 1.5mm fillet on your object. Any more than that and there wouldn't be room to fit it up near the top:





Cases like in the other thread where there's a sharp trailing edge are a lot more difficult though, I was wondering if your case was similar to that but yours doesn't have that.

- Michael

Image Attachments:
anthony_fillet1.jpg  anthony_fillet2.jpg 

---

usually i use the naca 65a009 and it has a sharp trailing edge made of a circular arc when i need to make a root fillet. the le is an ellipse. the model attached here is for a computer case fan. the airfoils are all stalled so it is made of a flat plate. usually you would use a circular arc for the le and te. however, i wanted to experiment with ellipses for the le and te. i wonder if your solid is valid. the line in your second picture, in the fillet. that is where rhino fails almost all of the time, until i shrunk the fillet way down to .08mm. it does create fillets at bigger values, however, they are all invalid. that's the problem i have with rhino. it either fails to make a fillet or creates a garbage fillet. if you battle with it forever you can usually get a good fillet out of it. but it's not at all easy. i'm not sure how to test if your model is valid. i saw you can calculate volume with the v4 beta. the other way may be to export a step file. i can see if it will mesh with netgen. in rhino for the larger value fillets if you zoom in where that line is you can see the fillet is actually mangled. but as the radius shrinks it becomes impossible to actually see what's going on.

---

Just saw this, would a simple sweep give you the profile you want?

I swept the "oval" along the 2 rails, made cylinder boolean combine with the profile, then filleted the cylinder profile edge with 1.0.




I went all the way to 10 mm, still worked...

Attachments:
Projecting Airfoils Onto Cylinders Example-1.3dm

Image Attachments:
15mm fillet.jpg  Projecting Airfoils Onto Cylinders Example-1.png 

---

hi death,

originally, about 10 years ago, the idea was to use a 2 rail sweep. several years later, the developer of mecway discovered that rhino was tilting the airfoils. i did some testing and confirmed this. i also found it had a huge impact on the natural frequencies. the mass distribution was way off throughout the model. basically, rhino can't be trusted to do anything right. over time i found just a normal loft worked best. but what you are finding is part of the problem. rhino screws up various things along the way. one of them is the quality of the airfoils. in the model i attached, when you go from cartesian to cylindrical it messes up the curves. this ends up messing up the fillet operation. but one thing you show, seems to be what michael has shown too. that being, moi seems to do a much better job at filleting than rhino. so that's great news. i look forward even more to switching. thanks for your efforts wrt topic. it's helpful

---

Hi Anthony yes I see that Rhino's filleter (both Rhino v5 and v6) does not seem to like your blade surface very much and has a gap in it at the blade's closing seam edge.

Your blade surface is not 100% exactly smooth at its closing point, there is a tangent deviation of around 0.01 degrees. It's so small that I wouldn't think it should make a difference but it seems that Rhino's filleter happens to be particularly sensitive to that. It looks like it is treating it as a sharp juncture and trying to extend and intersect the fillet pieces and that is what is causing the problem.

I think you can actually repair it in Rhino - if you extract out the fillets and untrim and retrim the fillets with an isocurve at the blade's closing seam point then it looks like the pieces join up ok. They just shouldn't have been extended and intersected. You'll also have to untrim the cylinder piece and retrim that as well.

The one made by MoI's filleter looks to be all ok to me, I have attached a STEP file export of it.

The new volume calculator for MoI v4 does not seem to have any problem with it, it says it has a volume of 8122.33 mm³.

- Michael

Attachments:
moi_airfoil_fillet_step.zip

---

that's amazing you figured out the problem. i'm more impressed that MoI doesn't have these problems. i don't know cad well enough to do the things you mentioned. but you provided me with enough information to know switching to MoI will be well worth it. i'll check out your step file now. thanks for looking into it. at least the original poster in the other thread will have an easier time than i have been having with rhino.

---

i tried to mesh your step file in netgen and it won't work. so something is still messed up. perhaps a 1mm fillet. not sure. visually, it looks fine. but it won't run.

---

i attached some pics of the netgen mesh for the model i sent. the one with the .08mm fillet. the fillet is so small the mesh is absolutely huge. over 1 million elements. there's no way i could solve that on my laptop. the other thing is, i don't even need to solve it to know the stress would be way too high. you need decent sized fillets to keep the blade from breaking off and killing people. .08mm isn't going to cut it. but the model does mesh, that's the thing that can be a problem. as the model you created shows. that's one of the problems i run into as well. for this particular model a 1mm root fillet that is valid is the goal. i gave up on it a long time ago. i had to go with the version not projected onto cylinders. i think the only thing that might help is recreating the curves that rhino screws up. i'll give that a try next. i'll have to cut them all and rebuild them all. it somehow takes four curves and turns it into two, when you do the projection. the problem is it screws up the le and te.

---

Hi Anthony, what happens if you try to mesh your original object without any fillet on it at all, does that work ok in netgen or not? Your blade does have a pretty complex structure with some of its knots quite close together. It could be that netgen doesn't like that surface.

I've attached a 1mm filleted version here too though.

- Michael

Attachments:
moi_airfoil_fillet_1mm_step.zip

---

see my previous post for the answer to your question. netgen will mesh the rhino .08mm model. but it's way to big to solve and the stress in a fillet that small would be way too high.

i did fix the model just now. what happens is when i do the projection in rhino it screws up the airfoils. it takes four curves and turns them into 2. it also alters the structure of the curves in a bad way. so i have to go back in and manually split the curves into four then rebuild. the rebuilding losses some accuracy. for both these reasons, it's not an ideal solution. but it will mesh in netgen with a small model of excellent quality. one weird thing i encounter with all this fillet headache is most of the time you get one curve to fillet. if you get more than one it usually won't fillet correctly. in this case, i get multiple curves i have to select to do the fillet but it does work out correctly. it's just a mess. there are so many problems and oddities with rhino i couldn't begin to explain them all. i attached an updated model with the fix in a before and after fillet state. also, some pics of the netgen mesh.

edit; oh sorry i didn't read your question properly. i thought you were asking about the .08mm fillet. i tried the no fillet model after you asked about it again. the post is further down.

---

unfortunately, your 1mm model hangs the same way as the previous step file you provided. the model looks fine but it won't mesh. it has an issue with a surface, but i don't know which one. when i get moi i can experiment. your filleting seems better than rhino. perhaps if i did the projections in moi those would be better too. it seems the rhino projections are the base of the problem. the filleting can get messed up by all sorts of things so it's hard to debug. thanks for your insights. last question; do you know if you have to rotate seams out of the way of filleting in moi, like you do in rhino.

---

this is the end goal of creating the cad model. stress evaluation and modal analysis. i attached some pics. fea model in mecway. load was 4600 rpm. the aero loads are very very small so they don't need to be modeled. i have checked that out previously. this is a delta computer case fan. the real part is much thinner and made with plastic. my part is aluminum and needed to be much thicker, because of that. this is just an example. i don't have the material properties for the real product.

---

Hi Anthony, yes seam edges can be a problem in filleting with MoI as well. A fillet in MoI will not be able to cross over some other edge in the model that is not being filleted but that the fillet is going to run into. So that part is similar probably, sorry.

- Michael

---

Hi Anthony, what about this version. This is with no fillet but exported out from MoI instead of from Rhino. Maybe it doesn't like something about MoI's STEP export. If this one does mesh ok then I guess it's something about the fillet that it doesn't like.

- Michael

Attachments:
moi_airfoil_no_fillet_stp.zip

---

hi michael,

the last model you sent with no fllet will mesh with netgen but it hangs a little bit on a surface. not sure which one. i imported the rhino model with no fillet and it meshes faster without hanging. but both meshed without a fillet. this is for the solid with the projected curves from rhino. so no fix was done to those profiles. if i fix the profiles i can put a 1mm fillet on it. not really sure what to make of this. just another nurbs filleting oddity. i attached the step file from rhino.

---

given the results of the last test, i went back to your 1mm model. i got it to mesh with netgen this time. not sure what's going on there. it seems to hang on an optimize surface step. last nite i left it there for a few minutes before killing it. today it hung but not for too long and finished meshing. i attached pics. i'll run it through fea just to be sure. but it looks like moi can create a quality step model, even using the messed up rhino projected airfoils.

---

hi michael,

your model ran through fea fine. i attached the results. i'm not sure what was going on the first time i tried it. maybe the mesh settings i tried were too high. i'm going to try the first model you sent with the even bigger fillet radius. attached are the fea results for your 1mm fillet model.

---

the first step model you sent has problems with netgen. on import a number of errors show up. it will mesh but in the area of the fillet seam the elements are all mangled. i tried two different mesh sizes and got mangled elements both times in that area. but your 1mm model worked and that was the goal. not really sure what to make of all this. it's kind of like what i was saying. you can follow the same process over and over but get different results. it's not clear what is causing the problems most of the time. so you end up just trying all sorts of things. so a matter of will power to get a good model out. but it seems like it's easier with moi. i look forward to trying it out.

netgen is free and available here, in case you want to play with it; https://ngsolve.org/news/releases

there isn't really decent help for it, but if you have questions, i can try to assist

i attached some pics of the error messages and of the mangled mesh fillet area. this was for the same exact mesh settings i used for your 1mm model.

---

i had some ideas to test. i brought in all the models to netgen and checked to see if those warnings came up. they came up for all of the models you sent me. but two of your models meshed anyways. there were no warnings for the rhino no fillet model and the .08mm fillet model. Also, no warning for my correct 1mm fillet model. you may be getting affected by those bad projected curves that rhino created. i think i posted an updated file that has the fixed profiles in there. and the solid with the fixed profiles before and after filleting. if you try working with that model, the netgen errors should go away. but it seems like those errors don't indicate whether the model will mesh or not. however, netgen seems to work better if there are no errors present, the errors could explain the hanging i'm getting with the models you sent.

when i get moi v4, i will try making the projections in moi. if it does it better than rhino, there may be no issues with filleting and step export. rhino takes my two ellipses and two curves, that have been joined, and turns it into two curves. they break the ellipses up and merge parts of them with the curves. they also re-create the curves in a bad manner. so i have to go back split their two curves to get the ellipses back then rebuild all four curves and join them. the problem here is you loose accuracy doing the rebuild. plus it's yet another time suck. ideally, if the program could project the original curves and keep their structure. rhino can't do that apparently. not sure about moi.

---

Show messages:  1-20  21-27