Prediction of fatigue load from FEA

[meher634]

Is there a way to predict fatigue load from FEA performed?
I always perform FEA on the model and compare the stresses to yeild strength of the material. For most of the materials I think fatigue strength of the material is about 50-60% of the ultimate strength of the material. In that case my assumption is that the fatigue load should be the load that gives stress which is about 50-60% of the ultimate strength of the material. Is my assumption correct? Any suggestions?

 

[cbrn]

Hi,

the way I prefer is the following:
1- perform FEA to determine high-stresses and peaks; simulate of course the two "extremes" of the possible load cycle
2- log in a file the stress results for every high-stressed points
3- import it into Excel, in order to calculate - for each of those points - stress amplitude and stress average
4- obtain from literature or experiments the Woeller-curve of the material
5- correct the curve using Haigh method, after reducing it in order to account for the desired failure probability (for work reasons, I always use 99.97% probability of NO failure - Woeller curve is refered to 50% probability, generally)
6- use Miner theory to calculate damage due to the loadcycle, or predict possible lifespan allowed by the loadcycle.

Of course, if the loadcycle is complex and involves several loadcases instead of a "min-max", I'll need one FEA for each of these loadcases...

Your assumption is a simplified one which works, if I remember well, if your loadcycle is such that R=-1, i.e. that sigma-mean=0 and sigma-min=-sigma_max.

Regards

 

[rb1957]

your assumption may be valid for the materials you usually work with, but it isn't typical. Acceptable fatgiue design means combining cross-section details (stress peaks), service spectrum, and material.

now you may be able to simplify your spectrum to one cycle per unit of time (for me it'd be per flight, but it could be per hour, minute, second); this takes some experience with the application. i'd use the FE data to tell me the ratio between the stress peak and the applied stress, this is equivalent to a classical stress concentration. then there are plenty of handbaooks around (is use MIL HDBK 5 or AR-MMPDS-01 mostly) to allow you to calculate the fatigue life for that applied load and cross-section. These curves are a "nominal" fatigue life, and a typically heavily factored, between 4 and 10, to determine a safe life.

good luck

 

[cbrn]

Hi,

good point, rb1957; especially when you refer to "design curves" for fatigue strength that are normated in literature and, generally, application-specific.
It's more or less what I suggest with what I call the "full method" (Woeller + probability correction + Haigh + Miner Rule or oligocyclic fatigue + corrections for aggressive environment + Rainflow or others, +...), but with much less effort.
For example, when I must verify the resistance of an hydraulic conduit, I apply ASME VIII,2,5.
But when the object can not fall within the application range of any norm, I prefer the "full method" (OK, I recognize that I'm a bit lucky because in my organization we can obtain data for the materials very easily) rather than a direct comparison with a design curve or, worse, with a single pre-determined ratio.

Regards

 

[rb1957]

diferent industries, different approaches ... same skinned cat !

are the ASME books on-line (i'm sure they are), but free ?
yes, cheap i know, but i'm willing to spend a little time looking into other approaches (but not much money !??)

 

[cbrn]

Hi,
no, ASME norms are not free. None of the industry norms are, as far as I know... :-( If you're lucky, the nearest University with a serious Mech Dept will have a bibliotheque where you can hire these books, I think. ASME are very common, so it's likely that you will find them...

Regards