Safety Factor...how to calculate with FEA results

[drillrig]

Hello all,

I've searched the forums and cannot find what I'm looking for so I'll try here.

I do alot of FEA's here and often am unsure about how to determine the safety factor. For example, assume I have a model that is "clean", i.e., critical features are included, the mesh size is appropriate, the solution is converged, boundary conditions are sensible, etc. The maximum VonMises stress (at the fine meshed areas) is about 40,000psi. The material has a 46,000psi yield. Do I calculate the safety factor to be 46/40 = 1.15? Or do I use some nominal value for the stress. I've read in some places that using the max stress to calculate safety factors can be too conservative.

Thanks in advance!

- Chris


Chris C. Mechanical Engineer
SW07x64 SP4.0
Dell 490, Xeon Dual Core, 8GB ram
Nvidia Quadro FX3500

 

[rb1957]

there are so many things to take into account.

if you've a reasonably solid section, thick and chunky, then you could use Ftu or Fty. this is pretty conservative, if the stress peak is only over a quite small area ... local plasticity which shouldn't affect things significantly could be acceptable ... you're assuming that the materail will yield and the load will redistribute locally.

but if you section is thin gauge and/or you've got lots of compression and shear loading it, then you need to consider other failure mechanisms (like column buckling, crippling, shear buckling, etc). in these cases the allowable is less than Fty.

You also need to account for repeated loading (fatigue) ... if the applied loading is repeated, then the allowable is probably less than Fty.

then there might be required margins, required by authorities, or the customer, or your own office.

no simple answer !

 

[GBor]

Chris,

It can be too conservative, but it sounds like you "have your head on straight". I generally look at the "mean" stress in the area of greatest interest, but I do also look to make certain that my mesh has converged as well as the solution. A mesh convergence should show a fairly smooth stress pattern with no single element containing greater than about 10% of the stress range. If you are following a proper FEA method, I think you are probably OK to calculate safety factor the way you are doing so now.

My 2 cents...

 

[cbrn]

Hi,
assuming... all that you have assumed (mesh convergence, BCs, etc...), and assuming in addition that the main failure mechanism is strictly structural (linear elastic) and not, for example, buckling, then...
Then it depends on your sector's "how to's"... Because in most cases, even local yielding is not allowed (but it's not mandatory from a <real> resistance point of view); in other cases, only the loss of functionality is prohibited, but frank yielding can occur as a design condition...
As a general rule, and even is ASME allows local yielding, prohibiting stresses to overcome yield is the rule. Note that the allowable general stress can undergo much stricter limits (for example, 0.5 of the yielding).

Regards

 

[drillrig]

All,

My component (shaft loaded in tension only) has to fulfill the requirements of American Petroleum Institute (API) specification 8C. It calls for a safety factor of 3.0 based on yield. For the sake of argument, if my material has a 100,000psi yield and the FEA tells me the maximum stress at a fillet is 50,000psi, is the appropriate safety factor 2.0 and the shaft therefore does not fulfill the requirement? Is this the case even though the yield stress is never reached and the nominal stress (simple P/A calculation) is, say 20,000psi?

- Chris




Chris C. Mechanical Engineer
SW07x64 SP4.0
Dell 490, Xeon Dual Core, 8GB ram
Nvidia Quadro FX3500

 

[EdR]

DRILLRIG:

I think you got it right on your last post....A safety factor can be defined with respect to anything so you need to read your spec very closely to see exactly what the definition is...If it is wrt nominal stress then your last post is not correct, however if it is wrt maximum stress then your last post is correct...I would also do some simple checks using standard tables for the stress concentration factor based on your fillet radius to be sure the FEA results are in the ball park.....i.e. max stress = K * nominal stress where K comes from a table for stress concentration factors...

Ed.R.

 

[cbrn]

Hi,
I am not familiar with API standards and regulations, but it would be mechanically / structurally absurd to set a safety factor of 3 on peak or even primary+secondary stresses. So I believe that this criterion refers to the primary stress (nominal stress).

Regards

 

[corus]

rb1957 had it right, or to summarise, you need to follow a design code as there are different allwoable stresses for different loads, different locations, different applications. Probably the most comrprehensive design guidelines are found in pressure vessel codes, which in principle can be applied to any structure. Basically though most design codes are based on the assumption that you are doing hand calculations, are making wild assumptions, and missing out secondary effects, don't know what you are doing, and hence apply huge safety margins. Just to pick the maximum yield stress from the FEA model and compare it to yield isn't correct though and the structure may well fail by other means.

corus