You are here
Fatigue life calculation - reporting stress
Thu, 2008-09-18 02:31 -onkar bhise
For fatigue life estimation (HCF & LCF), I am not sure which stress shall I consider i.e. max. principal stress or von Mises stress & what should be the criteria for selection of the same? Shall we be driven by material model (i.e. ductile or brittle) or mode of failure or some thing else?
»
- onkar bhise's blog
- Log inorregisterto post comments
- 16922 reads
Comments
Stress as the control variable for fatigue calculation
Hi Onkar,
Well, it is quite an important question. It should be also related to the question on how complex is the loading of the component you examine. If you have e.g. several non-correlated load channels acting simultaneously or even correlated but with some phase shift, the effects of strain hardening caused by this non-proportional loading can be observed. As far as I know, no simple modification of von Mises or principal stresses works then. Unfortunately, it is also hard to say, if use of these simple variables can give you conservative or non-conservative results. You should work with a direct load effect on some specific planes - but such a solution is a bit demanding as regards the time to get to the results. Take a look on my webpagehttp://www.pragtic.com, where you can download freeware PragTic that can be of use to you in such a case.
On the other hand, if the loading is proportional, e.g. von Mises equivalent stress can work quite well. Still it depends on what is your base S-N curve - if it is the one for fully reversed loading (load ratio R=-1), you should do also some modification to introduce the sign into the otherwise positive von Mises stress. If you do not do that, the fully reversed load cycle will be transformed to a repeated one - i.e. with the amplitude being half of the expected and with non-zero mean stress of the same level. The sign can be (e.g.) attributed in dependency of the sign of the stress tensor first invariant, or of the sign of the extreme principal stress - it is hard to say, what is better then, I do not know about any real comparison of these methods. (If anybody does, please let me know, I am highly interested in it.)
At last, if you are talking also about LCF domain, be aware that the S-N solution is not perfect there and the use of Manson-Coffin curves is more appropriate. Since this is a bit more complicated and necessitates iterative search for final number of cycles, I recommend you the use of PragTic once more.
Cheers,
Jan