Pattern and Mirror not Correctly Modeling Fillet (Bug?)

[neljoshua]

I am working on creating a model of a plate heat exchanger (not our product) for testing purposes. Fusion does not like something that I am doing, but I am unsure what.

In this image you can see that I have one fillet command, yet not all of the divots have the same fillet applied. This error showed up after I tried to change the fillet radius.

The Mirror operation is similar--it did not correctly mirror the fillets.

Does anyone know what is going on here?

https://a360.co/2EB8sxP

[TrippyLighting]

@paul.clauss @James.Youmatz you may want to take a look at this one.

The "funniest" think is that the mirrored instances are correct, only these two are not. Very odd!

[Phil.E]

Hi,

The fillet is the same, the divot is not.

(edit: I take it back, it actually depends on the pick order of the top vs bottom edge of the fillet. I'm not sure this is a defect, it may just be how the solver works on overlapping fillets - one has to take precedence. Did you do the selection individually or by window select all the edges at once?)

Also, the mirror works fine if you pick Optimized option from the fillet dialog.

Thanks and let me know if you have more questions.

Regards,

[paul.clauss · Accepted solution]

Hi @TrippyLighting and @neljoshua

Thanks for posting! This is a strange problem -  the fillet command appeared to have problems with two of the holes (the same two as shown in Joshs initial post) and the Mirror command previewed the fillets correctly, but they are not shown correctly after the feature is finalized.

This appears to be some sort of a bug (everything in the timeline looks like it should be working correctly) and I've passed your file along to our development team to have a deeper look (internal reference FUS-39353). I went back to pretty early in the timeline and redid the model with the same results and have been unable to find out what is triggering this issue at this time.

One thing I did notice is that everything is modeled as one body. Splitting the component up into multiple bodies may allow you to mirror bodies instead of faces or features, which often gives more stable results. With that said, I still got the two incorrectly sized fillets regardless of how the model was split up - the mirror command worked, though. EDIT: As @Phil.E mentioned, the mirror command will also work if set to "Optimized."

I was able to "brute force" a fix by deleting the problem fillets and recreating them in with a new mirror feature and then mirroring faces instead of features. The problem fillets were created in the incorrect size when I tried to fix them with a new fillet feature. I've attached the file to this post. 

We appreciate both of you bringing this to our attention! Please let me know if you have any questions.

[neljoshua]

@TrippyLighting, @Phil.E, @paul.clauss,

Thanks for your help and input.

The dimples/divots started out as the same size, so I am not really sure what is going on.

I did notice some strangeness with the Mirror command. I have two questions; perhaps you can help:

1) When working on creating this design I used a number of Mirror commands. I had to go back to add some features that I had missed (notably the 45˚ chamfer on the back of the inlet/outlet pipes. When I did this, I tried to edit the Mirror that I had used to create the second pipe (the first Mirror in the timeline). While so doing, the command would show the correct preview, but the end result would not be updated. The only way I was able to do it was to delete and then re-create the Mirror. The same thing happened with mirroring the divots. What is going on there?

2) I know there are different options for the Mirror command and was able to get the divots to mirror correctly after changing the option to "Optimize" (thanks for that, by the way). When is it appropriate to use each of the different options for Mirror?

Thanks.

[neljoshua]

@Phil.E,

I selected the edges one at a time.  As such, I cannot be sure about the order in which I did so.

[Phil.E]

Okay thanks. In this case, because the fillets overlap, the pick order matters. You can see the difference if you edit this file, or one like it, and apply the fillet to one hole at the bottom edge, and one at the top. The preview shows what will happen when you pick the next edge, and how the compute would be different.

Here are a couple of holes just like the ones you made. One has the top edge picked, one has the bottom edge picked.

Here is the result, prior to picking the second edge. You can see where you'd get a different result picking the second edge for both of these holes.

Please let me know if you have more questions.

Regards,

[neljoshua]

Thanks, @Phil.E. That really helps.

Any thoughts on the Mirror questions above?

[Phil.E]

Sure, sorry about that.

"1) When working on creating this design I used a number of Mirror commands. I had to go back to add some features that I had missed (notably the 45˚ chamfer on the back of the inlet/outlet pipes. When I did this, I tried to edit the Mirror that I had used to create the second pipe (the first Mirror in the timeline). While so doing, the command would show the correct preview, but the end result would not be updated. The only way I was able to do it was to delete and then re-create the Mirror. The same thing happened with mirroring the divots. What is going on there?"

(Phil.E: Not sure. I'd need to see an example I could test. Perhaps related to the need for Optimize option we found with the divots? If you see it again, try to catch the file in that state and please send to me.)

"2) I know there are different options for the Mirror command and was able to get the divots to mirror correctly after changing the option to "Optimize" (thanks for that, by the way). When is it appropriate to use each of the different options for Mirror?"

(Phil.E: I know @jeff_strater could explain it better than I... Personally, I just use whatever works. If the default doesn't work, I start looking at the mirror options. If none of those work, I start changing my modeling strategy.

Just a note about mirroring strategy: I have found that mirroring the body is the #1 way to capture all the features/faces. If you place a mirror body command at the end of the timeline, before you have any features to mirror, and then roll back before it to add features, you can see the complete body mirror update when you roll to the end of the model timeline. Sometimes this is not possible, but it tends to work well with mirrors that contain many features.

Steps:

1. Make a body 

2. Mirror it

3. Roll prior to the mirror, only the body from step 1 is in focus.

4. Add features to the body.

5. Roll timeline to end to see result of mirror.

6. Repeat as necessary.)

[neljoshua]

@Phil.E,

1) I actually saw this earlier today, but of course I did not record it. I will see if I can promote an older version of the design and make it happen again.

2) Thanks for the explanation & your time.

[jeff_strater]

I'll try to explain the various pattern/mirror options and how they work under the hood.  Just FYI, my approach to pattern/mirror is pretty close to what Phil suggests:  Start with Optimized.  If it doesn't work, try Identical, then, if all else fails, use Adjust.

OK, here is the lowdown on what these options do.  First, I'll limit this to Face patterns and Feature patterns.  Body and Component patterns are different beasts, and are, IMO, much easier to understand.  Hopefully the explanations below make some sense.

1. Face pattern.  Face pattern does pretty much what it sounds like it does.  You can select any collection of faces and try to pattern them.  The modeler will copy those selected faces, transform them by the pattern transform, then try to stitch them back into the model.  If the stitch will fail (the transform sends them into space, or they don't line up exactly with body geometry), then this pattern will fail.  This is the most efficient type of pattern, and why I always try this one first.
2. Feature "Optimized" pattern.  For all intents, this is identical to a Face pattern.  There are a few subtle differences, but I've forgotten what they are.  The main difference is the selection filter is set to Features.  You cannot select just any random set of faces with a Feature pattern.  You must select one or more Features in the timeline.  But, Fusion turns this Feature selection into the set of faces that were created by this Feature.  From then on, it uses the Face pattern algorithm to achieve the pattern.  Because of that, it is also very efficient.
3. Feature "Identical" pattern.  This is a bit slower than Optimized, but still pretty efficient.  This pattern type works by extracting the "tool body" from the selected features.  For instance, in this example, the box + fillet + hole defines a clear, closed set of faces that can be replicated:
   
   However, if you just selected the fillet by itself, that does not define a closed body which can be replicated.
   You can use Identical to pattern part of a set of features.  In the above case, if I do an Identical pattern, and choose just the Box, I can get this, which is not possible with Optimized:
   
   
   
4. Finally, we have "Adjust".  This is the most expensive pattern option available.  It is unfortunate that this is the default, but it is.  In this method, the feature recipe for the selected features is replicated.  That is, if you pattern 3 features in a 3x3 grid, there will then be 9 copies of those 3 features (for a total of 27) that get executed (though they are not shown in the timeline, trust me, they are there).  This type exists to support patterns like this:
   
   the cylinder here is a to-face Extrude.  If I pattern this with Identical, we get this:
   
   but, if I change it to Adjust, you get this:
   

So, what I usually do is go through that exact progression:  Face pattern, Feature Optimized, Feature Identical, then Feature Adjust until one works.  This progression is from most efficient to least efficient, and, in theory, should be least robust to most robust.  However, I have found cases where Optimized works, and Adjust fails.  I'm not 100% certain why that is, to be honest.

[neljoshua]

@jeff_strater,

Thanks a ton for that description. It is very helpful!