FE Update: Translation Tools Make CAD Models FEA-Ready
9/8/2008
by
Abdul Shammaa
The sharing of data between CAD, CAE, and CAM has
grown to include machining, molding, and PLM.
So true interoperability is a necessity to ensure product-data
quality (PDQ) throughout the enterprise.
Interoperability issues even can arise early in the game. For
instance, FE analysts need data forms, tolerances, and structures that
are different from their counterparts in product development who begin
with a CAD model. When there’s a misfit between design data and FEA
parameters, meshes generated from CAD contain a host of errors. FE
analysts must then spend appreciable time fixing bad data, instead of
running actual simulations. Such wasted time reduces development team
productivity, particularly as the cycle between design and analysis
shrinks and becomes more interactive.
The main reason for a CAD and FEA mismatch is that the 3D model
contains too much detail for FEA. Automated algorithms that generate FEA
meshes have difficulty accounting for slivers or geometric faces too
small to generate well-shaped elements. Run times rise. When the mesh is
fed into a solver, the skewed or collapsed elements crash the code. For
a solver to work, every single mesh element must meet some parameter of
minimum quality.
A CADdoctor screenshot of a CAD model shows narrow faces (highlighted
in red). These kinds of details interfere with accurate FEA meshing.
Make it simple
Simplifying CAD models makes the FE analyst’s job easier and also speeds
meshing and brings more accurate physics. Many question whether removal
of CAD details will degrade analysis accuracy. The answer: Not
functionally. That’s because FEA is an approximate method that provides
answers to within 5 or 10% of a theoretical solution. The accuracy
provided by a cleaned-up CAD model is more than sufficient for analyzing
degree of contact, bending, impact, and other stress information FEA
provides.
Furthermore, removed details are generally insignificant for an
analysis. For instance, they might be in an area containing no stress
concentration, or where details don’t affect the structural integrity of
the model. An example of an extraneous detail is a company logo embossed
on a part. The logo doesn’t impact the part’s structural integrity and
would be difficult to mesh.
Simplified geometry is necessary for CAD-generated FEA models about
80% of the time. When analysts need more detail, such as a contact
surface around a bolt hole, the original geometry should be retained.
Hidden geometries
Visible CAD-to-FEA issues are fairly easy but time consuming to fix
manually. However, the naked eye cannot see everything that needs
simplification. For example, curvature values, surface discontinuities,
and tangent continuity can be visually difficult to identify. So can
complex geometric errors such as twists and turns that create “surface
normal” problems. These also affect FEA (and, ultimately, cutting tools
— a surface error can flip a cutting path from inside to outside).
Again, gaps or overlapping faces (resulting in two surfaces where there
should be only one), small fillets, holes, and bosses, as well as free
edges, sharp-edge angles, and intersecting and self-intersecting loops
also cause problems.
It’s important to note that “small” in FEA is usually orders of
magnitude larger than CAD modeling tolerances. Small edges and faces
unintentionally created when a CAD model is built are usually above the
modeling tolerance for the CAD designer, who therefore considers the
model acceptable. For example, when the tolerance is 0.01 mm and there
is no edge, face, or surface in the model less than that dimension, the
CAD designer will see the model as “clean.”
But tolerance issues work the opposite way for FEA. For the FE
analyst, the 0.01-mm tolerance is much too small. It essentially means,
“Place this mesh on this node to within 0.01 mm.” To meet such a
miniscule tolerance, meshing algorithms are forced to generate tiny
elements which cannot be used without lots of manual correction.
Simplifying CAD data
Dedicated data-translation software provides a useful solution to
problems that arise during handoff from CAD to FEA. Translation software
essentially deconstructs CAD files mathematically and then reconfigures
the data to meet the computational expectations of the FEA program.
Part and parcel with data transformation, translation tools automate
the simplification of CAD models where necessary to ensure elements
generated by FEA are of usable quality. Translation software also
detects tolerance issues and other “bad geometry” such as small holes.
When designers use a data translator, the tool highlights areas that are
too small as well as untenable geometries. Designers can then select
easy-to-use, semi-automated click, see, accept, or reject options for
simplification, removal, or surface healing. The result is a clean,
viable model that will let FE analysts create proper meshes.
It can be seen that the advantages of using a dedicated data
translator are considerable. For example, simplified geometries from our
CADdoctor tool are neutral to any CAD system. In other words, original
data can come from any CAD software, be cleaned up, and then be sent to
any other CAD system. What’s more, the translator tolerates all
intermediate “limbo” states, during data dissection and cleaning, when
the actual model becomes invalid to the CAD software.
Alternatives are limited
One alternative to dedicated data translation is the evolving “virtual
topology” technology being developed within certain design software
programs to facilitate the transition from CAD to FEA. The approach is
sound but has limitations, among which are less flexibility and the
inability to support intermediate simplification states.
That said, virtual topology works well within its own suite of
software tools, but is of no use to an external solver. This could
certainly be an issue in a supply chain with lots of different CAD
programs, or even within a single, large organization.
Other alternatives to dedicated data translation are complex
preprocessors that manually clean up, delete, and merge dirty mesh
elements that result from CAD-FEA interface issues. However, such
preprocessors are used after the FE model has been created,
necessitating the mastery of new commands and more work for analysts.
PDQ for analysts
What FE analysts really need is the same high level of PDQ that has
become the Six Sigma holy grail in the manufacturing phase. Yes,
companies can go from CAD to meshing and then ask their analysts to
hand-edit thousands — or even millions — of erroneous elements. Or firms
can save their design teams countless headaches by preparing CAD
geometry with a dedicated data translator before sending models to FEA.
The cost of fixing CAD-FEA interoperability problems early on in this
fashion is small.
About the Author
Abdul Shammaa
Elysium Inc.
Article edited by Leslie Gordon,
Sr Editor, Machine Design
 |
Article reprinted by permission of Penton Media,
publisher of Machine Design |
| 
