FEA Non-Linear Analysis

[NikonF6]

When "he" say: o man this is non-linear, he means result in his hand is last word, so no comment on accuracy should be made. This is in top level (or is suppose to be) engineering in Detroit automotive.
I can see more and more often that FEA word becomes excuse for non-ability to understand mechanical engineering at all. FEA C-A-N be made to be useful SOMETIMES.
Properly used, in calculation of stresses, among 100 lines of calculation, FEA result as input into calculation takes only 2 lines (Form Factor and Nominal Stress).

Now about that NON-LINEAR
When "FEA" calculates it does not calculates deformations and stress separately as two independent calculations. "FEA" calculates only ELONGATION, and stresses are post-calculation from elongation (it is why stresses are less accurate then deformation, one error more is made).

When elongation due to loads and constraints is calculated at the point, stress is: (total elongation) x (E modulus of elasticity)
example: elongation = 0.0002, E=210000, stress=0.0002*210000=42 N/mm^2 (try this on your computer)
This is calculation along straight line and it is why it is called linear.
When "yield" is past, real stress swings away from the straight line, and if we continue to use linear analysis error is made.
Now comes non-linear as need.
Is FEA useful at all in non-linear? If one is not close to expert in metal forming unknown error is on the way.
The best way “non-linear” is to do linear analysis and find elongation. Make FLOW curve for the material (by hand on a paper). Draw vetrical line with known elongation, up to curve, then draw to left ordinate showing stress. You got real stress.

It is that simple and accurate as your FLOW curve is. Flow curve is (Stress)-(Logarithmic elongation) curve. Engineers working in field as bending, drawing, forging, .. (all above yield, so "non-liner" guys) knows it.
No non-linear analysis, in so many cases, is not even and close to real stress. E.G. Hertzian stress, Press Fit, .. and yields very large errors done by computer.
FEA C-A-N be very useful only in hand of "good engineers" not computer cables.

So lets comment on non-linear FEA (actually FEM).
With High Respect

 

[inline6]

in theory it is simple to predict stress. simply obtain deformed shape, have enough gauss points in the correct area to capture stress distributions and use a suitable material model and hey presto.