ANSYS Modeling and Meshing Guide, Chapter 6
"... some CAD models cannot be used for analysis and some models should not be used for analysis."
PLAN projects for the best possible coordination between CAD and FEA requirements. (Team work, concurrent engineering)
Realize that two different models may be needed (CAD vs. FEA)
CAD model (or design geometry) can be for manufacturing, prototyping, detailing, tolerance checking, and maybe FEA
Analysis model (for FEA) has only the elements necessary for simulation of the part in its application.
CAD models may be difficult to use for an analysis or may produce meshing failures due to sloppy modeling practices. Sometimes, just start over for FEA.
If possible, planning during the project concept period can lead to useful AND usable CAD models.
1. Surfaces having consistent size & shape ratios
2. Simplification is possible without compromising the structure
Remember, simplifying or suppressing features far from the area of interest is acceptable. Effects, local to the simplification will most likely not effect the global behavior of the model.
1. "SHORT" Edges (short, with respect to, model size or mesh size)
2. Sliver surfaces (faces with high aspect ratio)
3. Voids (holes inside)
4. Complex surface or edges (lofts or blends)
5. Corrupted geometry files (from translators)
6. Parent-Child relations (prohibit simplification)
outside corners (leave them sharp)
ignore small inside fillets far from region of interest
threads or splines are usually ignored
ignore small holes, not in the load carrying material
ignore small protrusions, not in the load carrying material
ignore decorative or cosmetic features
large sections not attached to the part of interest (large enough to be "rigid" relative to the section you are interested in).
1. Delay addition of fillets/chamfers as long as possible
2. Delay definition of draft as long as possible
3. Create parent-child relationship from permanent datums (whenever possible)
4. Never reference new features to fillets/chamfers/draft edges
5. Never leave unwanted or incorrect features in a model (even if they don't show)
>We all go through three phases. I've been in the FEA biz about 25 years, and that's pretty much how it goes. You can look at a given solution and tell which stage someone's in:
>1) fascinated
Believes any problem worth meshing is worth overmeshing. Rejects beam and plate elements as analytically impure. Prefers contact element algorithms over actual boundary conditions. Plots everything. Punches and keeps plots (even the ugly ones) Spends about 2-3 times more effort writing macros than the macros actually save Reports quite colorful; heavy on graphics and FEA-speak and light on insight.
>2) frustrated
Refines mesh selectively; shows resignation to dealing with ambiguity Relies less on clever elements; truly trusts only classical element types Abandons attempts to model welds with solid elements. No longer weeps at sight of tet elements. Time spent writing and debugging macros about equals time saved by macros. Reports contain caveats and warnings about applicability.
>3) healthy
Meshing aimed at specific problem areas; seldom models the entire airplane to find stress in the door latch. Element choice reflects engineering considerations; comfortable with approximation. Keeps obsession with computational efficiency under control, usually without medication. Makes frequent use of tabular results; understands use of numbers and text. Reports balanced between engineering issues and eye candy. Makes appropriate use of both.
Christopher Wright P.E.
chrisw@skypoint.com
http://www.skypoint.com/~chrisw