Re: [fluka-discuss]: geometry import from cad file

From: JR Cary <cary_at_colorado.edu>
Date: Thu, 4 Feb 2016 05:56:38 -0700

Thanks Andrew and Chris,

As Andrew knows, we (at Tech-X) and the DAGMC team have been in
contact, and given our experience in integrating CAD with
electromagnetics and plasma modeling, we hope to help out with
DAGMC at some point. (As usual, all depends on funding.)
Their work on fast MC with tessellated surfaces is what first
caught my eye.

So far I am not hearing that the problems of using CAD with MC
are any more difficult than the problems of using CAD with any
other CAE tool. In fact, the opposite may be true, in that
meshing may require more cleaning/healing than MC.

Did I capture this correctly?

Where we have seen the need for CAD is not generation of the
basic geometries, but instead in fitting into the overall
design flow, where after radiation modeling, the part goes
out for mechanical design, is modified, then comes back as
a CAD file for further radiation modeling. One of our partners
is in this quandary in moving the part between SolidWorks
and another radiation modeling tool.

This brings me to Andrew's point,

   However, for the models of the complexity that would warrant
   the use of CAD, we find that we can turn around analysis 2 or 3
   times faster than the translation route with something like
   McCAD or MCAM.

for which I am not sure I got the intended meaning.

Andrew, are you saying that when CAD is necessary, you are finding
that the DAGMC route, possibly involving SpaceClaim healing, is
faster than the route of trying to re-create native CSG from the CAD?

Thanks......John

On 2/4/16 2:11 AM, Chris Theis wrote:
>
> Hi Andy and John,
>
> Andy already expressed my point perfectly. I would just like to add
> one point regarding CFD/FE as this often comes up in the discussion
> and is given as an example where the use of CAD models works easily.
> Actually it doesn’t – at least not without human intervention when it
> comes to halfway complex 3D models. Often FE calculations are done in
> 2D where the meshing is numerically much more forgiving and in 3D
> people spend a lot of time cleaning up meshes. So they suffer from the
> same problems but the code packages usually come with cleanup tools.
>
> In addition there are also mesh free FE approaches based on compact
> smoothing kernels. They avoid the issue already from the start but the
> mathematical foundation is completely different and much more complex.
>
> So at the end of the day, there is no such thing as a free lunch. But
> I’m certainly looking forward to seeing DAGMC in combination with
> FLUKA, which I think is very good initiative.
>
> Cheers
>
> Chris
>
> *From:*owner-fluka-discuss_at_mi.infn.it
> [mailto:owner-fluka-discuss_at_mi.infn.it] *On Behalf Of *Andrew Davis
> *Sent:* 04 February 2016 07:48
> *To:* cary_at_colorado.edu
> *Cc:* fluka-discuss_at_fluka.org
> *Subject:* Re: [fluka-discuss]: geometry import from cad file
>
> Hi John
>
> Just to expand some of what Chris is saying, there are a number of
> things that limit using CAD immediately.
>
> 1) overlaps - there can be very slight overlaps in geometry, imagine a
> cylinder sliced through the circular cross section. The two surfaces
> should be topological equivalent and usually are at the point of
> creation. Sometimes going through a save and import these can drift
> from equivalence.
>
> 2) surface representation - CAD can operate with non analytic
> surfaces, which require some form of simplification in order to be
> translated to use the CSG geometry of the MC code
>
> Both 1 and 2 hinder an automatic transfer from CAD to MC, in our
> experience most time is spent fixing that CAD before it can be
> translated to MC or used in DAGMC for example. However, for the models
> of the complexity that would warrant the use of CAD, we find that we
> can turn around analysis 2 or 3 times faster than the translation
> route with something like McCAD or MCAM.
>
> Furthermore, there are tools like Ansys-Spaceclaim which make fixing
> the CAD very intuitive and straightforward.
>
> The FE approaches (including Attila) require dramatic simplifications
> and cleaning in order to be able to mesh (which is not the case for
> Dagmc), which are often the show stoppers that Chris is speaking of.
>
> Chris, appologies for putting words in your mouth, feel free to
> correct if I've misspoken on your behalf.
>
> Thanks
>
> Andy
>
> ################
>
> Hi Chris,
>
> I am very interested to learn more about your thoughts on this.
>
> There is a big CAE industry for fluid dynamics, electromagnetics,
> and structural mechanics, all of which do seem to integrate CAD,
> at least well enough so that companies buy their stuff. Do you
> have an idea of what distinguishes MC?
>
> There are also commercial radiation modeling products, like
> FastRad (Geant4 based I
>
> On 2/3/16 1:19 PM, Chris Theis wrote:
>
> >
> > >
> > > Dear Marlon,
> > >
> > > The complex geometries which you seem to be referring to have
> been built "manually" without CAD import. CAD import is a widely
> discussed topic but the technical difficulties discussed in the
> thread that you're referring to still stand. Additionally, in the
> more than 11 years that I've been working computer graphics
> industry I have not come across one single CAD file in an exchange
> format which was numerically 100% robust and correct so that it
> would not require substantial manual corrections. Often the
> problems are visually not apparent but as soon as you do some
> mathematical modeling with the geometry (Monte Carlo, FEM, etc.)
> presumably small errors can become show-stoppers.
> > >
> > > It is not impossible to use CAD with MC but, depending on the
> application, it can be a rather laborious and non-trivial process.
> > >
>
> >>
> >> >> Currently, Is using SimpleGeo and creating bodies from a 2D background
> image still the best way to generate complex geometries for Fluka?
>
>
> >
> >
> > >
> > >
> > > This is one possible way and there is also the option in SG to
> use 3D CAD files for verification of the model built by the user.
> But there are several tools out there (e.g. GeoViewer, manual
> input via Boolean algebra), each with its advantages and
> disadvantages. In the end it depends on the user which tool suits
> him best.
> > >
> > > Cheers Chris
> > >
> > > On 03 Feb 2016, at 20:42, Marlon Saveri Silva
> <marlon.saveri_at_lnls.br <mailto:marlon.saveri_at_lnls.br>> wrote:
> > >
>
> >>
> >> >> Dear experts,
> >> >>
> >> >>
> >> >>
> >> >> I would like an update regarding this
> 2010
> discussion:http://www.fluka.org/web_archive/earchive/new-fluka-discuss/2726.html
> >> >>
> >> >>
> >> >>
> >> >> I’ve seen some of yours studies with entire sites
> simulated at Fluka, I wonder if they have been drawn without
> importing from a CAD file.
> >> >>
> >> >>
> >> >>
> >> >> Currently, Is using SimpleGeo and creating bodies from a
> 2D background image still the best way to generate complex
> geometries for Fluka?
> >> >>
> >> >>
> >> >>
> >> >>
> >> >>
> >> >> Att.
> >> >>
> >> >>
> >> >>
> >> >> Marlon Saveri Silva
> >> >>
> >> >> Mechanical Engineer
> >> >> Beamlines Instrumentation and Support Group – SIL
> >> >> Brazilian Synchrotron Light Laboratory– LNLS – CNPEM
> >> >> + 55 (19) 3512-2490 <tel:%2B%2055%20%2819%29%203512-2490>
> >> >>
> >> >>
>



__________________________________________________________________________
You can manage unsubscription from this mailing list at https://www.fluka.org/fluka.php?id=acc_info
Received on Thu Feb 04 2016 - 15:36:49 CET

This archive was generated by hypermail 2.3.0 : Thu Feb 04 2016 - 15:36:55 CET