Plz help, dual problem (shell failed & unable to mirror fillet) in SW

1. Shell failed

I am able to shell this part
http://home.comcast.net/~wangphk/SolidWorks/Parts/Shell-Failed.jpg
after rounding off shown edges (Top left: Fillet 35, 33 and 39).
However, if I add the edges shown (Middle and Bottom left), I received
this message:

"**Shell: The shell operation failed to complete. One of the faces may
offset into an adjacent face, a small face may need to be eliminated,
or one of the faces may have a radius of curvature which is smaller
than the shell thickness. Please use Tools Check to find the minimum
radius of curvature on appropriate faces. If possible, eliminate any
unintended small faces or edges."

I try different combination e.g either increase or decrease the shell
thickness (1mm ~ 10mm) and round feature (1mm ~ 15mm) and still unable
to fix this. Could someone please take a look and let me know how can
I get around this problem.

2. Unable to mirror fillet feature

If I try to mirror the Extrusion 3 along with its fillet features, I
receive this message:

" Mirror: Unable to create instances for the pattern. Try again by
decreasing the number of instances or the distance between instances."

Removing all the fillet feature will cure the problem. Does anyone
know why SW doesn't let me mirror the fillets?

Thank you all for your time and help.

 

It's hard to say why this fails without having the part in hand. (Often
times it is hard to say with the part in hand!) I suspect that the corner
where the two added fillets meet is too complex for SW to figure it out in
the offset for the shell. The other end of the last fillet may be too tricky
as well.

My first pass at a work around would be to shell the part without the two
fillets, then add the fillets and the matching internal fillets.

I don't know why, but I know at least part of when. SW has had trouble
mirroring some fillet since I started using it with SW98+ The obvious, but
sometimes painful, workaround is to add the fillets after the mirror.


Jerry Steiger
Tripod Data Systems
"take the garbage out, dear"

 

Jerry:

Thanks for your comments and suggestion.

I would hope that there are other painless and time consuming way.
But I think, I have to cope with this for now. I will ask these
questions in my next SW advanced training session and see if someone
has a better idea.

 

John,

One of the ways to find out if the part will shell/offset or not is to
run Tools/Check and check the "minimum radius of curvature" option,
select the "results list" and then subtract your shell/offset and that
should help you understand the areas of concern and if it will
shell/offset or not?

...

 

btw, personally, I wish the shell/offset functionality had direct
graphical feedback (or an option for this feedback) of the failed or
problem areas of the surface(s)/solid.

Again, this makes total sense but having this feedback would help the
users understand the model as well as give the user options or
interaction which may allow problem areas to be over ridden or allow
shell/offset algorithms for overlapping or consuming the areas which are
failing or allow for holes which could be repaired later..,
Most users, I could be totally wrong here, want information about
what/where/why.. the shell/offset is failing and ways to solve the
problem? Basically, they want to get their job done asap!?

Oh well, just a silly idea... ;^)

 

It does (added 2004). You can get SWx to show you which faces can offset,
and which it thinks are the problems. You can even get SWX to make an offset
surface comprised fo the faces that are not problems so you can manually
patch the manifold and make your own shell (using the tried and true
workaround)
This being SWx, though, it only is useful around 50% of the time. It also
can misidentify the problem area. But in all fairness, it is there. And
perhaps it is working better - I only had one occasion to use it in SP4,m
and the diagnosis tools worked great.
BTW- I don't think I would ever, ever let SWx offset the faces for me
because I have no direct control over what happens after rebuilds (much
better to copy the body, delete the bad faces without patching, then make
the offset myself)

Getting close to what you suggest, i have used the diagnosis to find the
trouble area, close out of the shell, delete and fill the bad faces
(sometimes changing the boundary, but usually just 'delete and tangent fill
'right from the 'delete face' PM), then the shell works.

 

As Jerry said, it is hard to diagnose without the part, and though the
pictures were good I couldn't tell which faces you wish to remove with the
shell (if any). That is a pretty important deal.

When diagnosing a shell problem, you have to think what is happening behind
the scenes. Every face needs to offset the shell distance and the lot needs
to come together into a single 'manifold'.

If a face can't be offset the shell distance, you get an error.... that is
what Paul alluded to with 'check minimum radius of curvature'. You
shouldn't expect to shell a 1/8" round to a 1/4" thickness, because the 1/8"
radius face would wink out of existance as it was shelled. Of course, that
check is only relevant in a small minority of shell problems - maybe one a
year for me. Minimum radius of curvature only matters on concave surfaces -
convex surfaces are just fine (the 1/8" fillet would become a 3/8" round on
the inside of the part). Also, SWx is smarter than it ever has been before,
and almost always recognises that the 1/8" concave surface can no longer
exist and just brings the adjacent faces together in a sharp corner on the
shell interior (see next apragrpah for why this can cause problems). You
can run into sneaky minimum radius issues on lofted faces, but they usually
(75%) arise from wrinkles or defects and its your own damn fault for not
checking the results of your loft until you got around to shelling the
model. Even these can be quickly fixed with surface fills that wipe out the
bad bits.

The much more prevelant source of problems in my recent experience is that
of bringing all the offset faces together into a single manifold. I suspect
that is where your problem comes from. I do not know of an instance where
the shell command created a face - it needs to extend faces that are already
in the model and try to keep everything in contact, which is a problem that
pops up on fillets too. I would read the filleting section under tutoirals
at www.dimontegroup.com and an article on filleting in the archives of SWx
express newsletter (avaialble through the SWx website) to get some
introduction to how the faces that need to extend and keep in contact with
each can cause a filleting operation to choke. This mindset can be used to
dissect the shell problem - I imagine that you are shelling the flat bottom,
and see problems with the fillets along the side making some disjoint stuff
when they are offset any shell distance, and some edges that move yet don't
seem to have faces nearby that can be extended to fill the resulting gaps.

One helpful trick to see if you have the right idea is to cut away the model
portion that you find suspicious and see if the model shells. Then modify
that cut until you really isolate where the problem is - it can be where
multiple faces come together, complex corners, or edges.

 

Ed,

Ok, yeah, there is a show option for "shell" (offset shows as default
but does not have diagnostic options) now in 04.
Looking at it in sp4.1, it's does not really tell you anything or give
enough insight to what is failing or degenerating and it's not very
consistent graphically.

Now, as for diagnostics.. in SW2004,... well, in my examples, it does
not work, I do not get the error diagnostics option
Even testing this with a simple solid box with holes it it, it does not
give me a diagnostics option or show anything useful. So, what you are
seeing, I'm not seeing, .. although, the help does say it's there? Or
is it broken in sp4.1?

But in SW2005, using the same examples, it is giving me the option and
this is getting better and I admit I did not really test out the other
options because it was not working for the examples I tried in SW2005
beta (reason for my writing).
For the diagnostics, the failing surfaces, display mesh and display
curvature do not show what is failing in the examples I'm looking in
PR1. So, I don't see it working "yet" (why I'm writing, like many new
SW features, they don't work or they take time to actually function).
So, if you talking about SW2004, you're right, it is giving something
but not the point I'm making in getting feedback to the areas of concern
or options in omitting bad surfaces (allowing and not allowing SW to
figure it out for you or using different algorithms and of course what
you don't allow you get a surface offset (which again, offset has no
tools/diagnostics) and manually patch and cut to shell). (btw, yes,
there is a option to shift out of shell, if it fails (which is not
consistent) and allow your to surface offset (which again does not have
diagnostic tools and btw, it funny since it is basically offsetting
surface anyhow what's more funny is when surface offset works but shell
does NOT???)
Granted, it's new and it's a start and it will have to go through the
paces (at the cost of the users testing it).. but it's basic and the
reference I'm making is/was something I had access to using in Pro/e 6
plus years ago.

Is it getting better? Compared to past SW, yes.

Ditto. Manual is a safer way on complex parts and even simple ones.

 

Ok, in 04 sp4.1 and 05 pr1, I got a model to show the failed shell but
again, imho, it's not clear in what the problems are.
It reminds me of working with trimmed surfaces ('04 and '05), the
graphical feedback is vague or useless or fails to represent the
boundaries in the selection list or trim list or washes out graphically
as you make selection changes.

What would be more useful are converging boundaries or intersections
which show the overlaps as well as in mesh and color thereby
highlighting the suspected check body/faces.
As it is now, it's a collage of vague color and meshed graphical
representations which are vague, imho.

We have a ways to go with this interface.

...