MET 415 Lecture Notes
Chapter 9
Text: Building Better Products with FEA,
by V. Adams & A. Askenazi, (Read pp. 303-312)
Do NOT take your solver for granted !
Multiple Load and Constraint Cases: (p. 304-5)
Is it more effective to solve in a single run, or in
separate runs ?
Can your postprocessor handle (and combine) separate load
cases ?
In ANSYS: see LCASE, LCOPER, etc.
FINAL MODEL CHECKS (especially for big models): (p. 305-9)
- Free Nodes: Un-used nodes
- Model Continuity: Un-merged boundaries (check with a /EDGE
plot)
- Sanity Checks:
- list material props., element types, real constants
- plot and visually check boundary conditions,
list loads
- check mass (in ANSYS, during the solution)
- RAM and disk space estimates can be
requested
(p. 310) "... most solvers still provide cryptic error
messages which provide little help" like the "small pivot" error
Read the message carefully,
try to interpret in light of what you know about FEA and your
specific model.
If you must reduce model size:
- use symmetry, if possible
- clean-up questionable features
- use local refinement, not small elements
everywhere
- manual mesh (all bricks or quads)
First MODE Solution (dynamic analysis) = 1st
natural frequency (p. 311)
Can show improper constraint, un-merged boundaries.
ANSYS Solvers (see
HELP,EQSLV):
- Frontal = frontal direct equation solver
(small problems: < 50k DOF's)
(smallest RAM, highest disk
space demand)
- Sparse (often the default) = sparse matrix direct solver (more
RAM req'd.)
All the other choices are "iterative"
solvers, typically, requiring large RAM, moderate disk space:
- JCG = Jacobi Conjugate Gradient (for 3D
"harmonic" analyses, acoustics, and multifield problems)
- ICCG = Incomplete Cholesky Conjugate Gradient
(good for "ill-conditioned" cases and multiphysics)
- PCG = Preconditioned Conjugate Gradient (RAM =
2*JCG, but faster. Very good for large models w/solid elements)
- ITER = Iterative solver (Automatically chooses
an iterative solver that is appropriate for the physics of the problem)
ANSYS Workbench: Design Simulation > Solution
> Solver Type:
- Direct (works better with thin, flexible
models)
- Iterative (works better with bulky models)
- Program Controlled (default)