Using SIF from FEA without k factor for branch

Posted by: Van Ha

Using SIF from FEA without k factor for branch - 10/08/19 07:55 AM

In case D/T>100:

+ Update SIF and k from FEA for bends
+ Update SIF from FEA, leave k=1 for branches

Is it OK with reason that leaving k=1 for branches will give conservative output?
Posted by: Michael_Fletcher

Re: Using SIF from FEA without k factor for branch - 10/08/19 01:30 PM

Yes and no.

You should be accounting for the flexibility of the connection, but not through the k factor.

You can accomplish this through CNODE connections or "nozzle flex" feature, as applicable.
Posted by: Van Ha

Re: Using SIF from FEA without k factor for branch - 10/08/19 08:25 PM

k=1 imply branch flexibilities are not applied while k for bends are still be applying to the system.

Regarding to you reply. Why Yes and No?
Posted by: Michael_Fletcher

Re: Using SIF from FEA without k factor for branch - 10/09/19 08:32 AM

Yes, you should leave k as 1.

No, you should not ignore the flexibility of the thin wall fittings as calculated within FEA. You just shouldn't input it through the k factor, but through CNODEs or through the nozzle flex feature.

I don't know what your piping system in particular looks like, but if elbow stiffness contributes to failure in all components (pipe, equipment, and steel, etc), it makes sense, sure.

But I can envision scenarios where this isn't necessarily true.

So, if you're certain that what you're doing is more conservative than reality, then yes.

After all, buying a dozen eggs when you only need six still means you can make omelets.
Posted by: Bob Zimmerman

Re: Using SIF from FEA without k factor for branch - 10/09/19 08:51 AM

The loads in beam element programs such as CII will be distributed based on the stiffness of the system. Typically a more rigid system will yield conservative overall results but the results may not match the actual installed system out in the field.


All the components have some kind of stiffness and loads will be distributed throughout the system accordingly. From basic Statics the loads will distribute more to the stiff items. If you want more realistic results you should include Branch connection, Nozzle and Support stiffness if these can be obtained without excessive manhours being expended.


If you have the SIfs from an FEA, then you may already have the flexibitities also if you used PRG software products. Just go ahead and input this data. All you need is a little patience and the work will be done.

Granted for the "average" moderate size and mild operating condition systems there may not be much benefit or change in results for this extra work. BUT for more critical systems the NEW TREND should be to obtain the correct answers (within reason) and then evaluate how much real margin we have in the design. Also a more accurate analysis tends to help troubleshoot issues that arise from the Job Site.