FEA questions

[ahad602]

I have some questions regarding FEA and will highly apppreciate your responses:

1) What is the main reason of fatigue: Stress or strain?

2) Could one use linear static simulation to predict the potential fatigue area(s)?

3) Could one use linear static simulation to predict the deformational failure for a pressurized vessel made of steel?

4) When the FEA predicts areas of high stress, how can one make sure that is valid and not a result of poor mesh?

5) When doing modal analysis, how accurate could the predicted frequencies and amplitudes be?

 

[rstupplebeen]

1. Stress = Strain * Modulus so they are both equally responsible. I tend to gravitate towards strain because 2% strain is the same regardless of material, conversely you need to know what the material is to know if 1MPa is high or low. However, this is just a preference of mine.
2. Yes, linear static simulation can be used provided those assumptions are accurate. As fatigue progresses the loading pattern will change. Because of this additional factor of safety should be used. There are commercial codes that extend FEA to more accurately predict fatigue.
3. Near failure loading a linear simulation is rarely accurate. This is typically overcome by the increasing the factor of safety as in #2.
4. Mesh sensitivity studies can be performed. If it is at a stress concentration such as an internal sharp corner more detailed geometry may be needed such as including a fillet even if it's just the cutting tool radius.

I hope this helps.

Rob Stupplebeen

http://www.optimaldevice.com

 

[GregLocock]

1)It was always explained to me that stress alone does not cause fatigue, that is in order to get the grain boundaries to move there has to be motion, ie strain.

5)It can be very good. It usually isn't. You don't know the damping. You don't model the boundary conditions properly. You don't model joints properly.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[corus]

Linear analysis is generally used when assessing stresses against design code limits as these have various safety factors built in. In these design codes fatigue is based upon linear elastic calculated stresses in the data provided for the SN curves. If you want to justify a design beyond design code limits then you'd need to use non-linear methods and lots of validation.

 

[ukbridge]

My contributions

1) What is the main reason of fatigue: Stress or strain?

For elastic analysis these are obviously very closely related, but strictly it is the change in stress an element will undergo that contributes to fatigue, rather than actual stress magnitudes. The type of detail is also very important as some are much more susceptible than others.

4) When the FEA predicts areas of high stress, how can one make sure that is valid and not a result of poor mesh?

There are many means of doing this. You could do a numerical check using tables from classical textbooks such as Roark et al, assuming that you're using the right table! You could also concentrate your mesh in areas of interest by using non-uniform mesh divisions i.e. apply a mesh with an element size of 0.5m, 0.25m, 0.125m etc starting from one end. You can also do a quick parametric study with different mesh fineness. A model sanity check never hurts either.

5) When doing modal analysis, how accurate could the predicted frequencies and amplitudes be?
The question is far too broad, but bear in mind some analyses are only SDOF whereas in practice there are infinite DOF's. It's also difficult to assess the damping constant. In short our understanding is good enough for all practical purposes over many years of research and the advancement of technology. Download an accelerometer app and take readings on a source of vibration and you'll probably answer your own question!