Peak Stress&fatigue analysis

[josedipofi]

I have question for you gents:
I’m checking a stress calculation from our vendor (vertical reactor) performed with FEM.
The loads are: weight, pressure and thermal,no cycle loads.
At a certain point he calculate PL+PB+Q+F=562MPa ( connection between the head and nozzle) larger then allowable that is 444MPa (3*Sps)
The next step in his cal. is not clear to me, follow what he wrote:

"For the secondary stress assessment, the ASME code section VIII div 2 only considers (linearized ) secondary stresses without the contribution of the peak stress (F).
Therefore it is allowed to perform a linearization to obtain the secondary stress range PL+PB+Q”

The value of PL+PB+Q he calculate afterword is still higher then allowable stress so the next step he performed is to apply a simplified elastic-plastic analysis according to 4-136-7 ( ASME VIII div 2 2004), shortly he drop from PL+PB+Q the thermal bending stress and compare the result with Sps (the stress now are lower than Sps), after that he calculate Ke to enter in fatigue curve with Ke*Sa to get the max. numbers of cycles allowable.
My question is: The procedure he is following is correct?
The combination PL+PB+Q+F must be check only if there is cycling service, if not when?
Thank you

 

[corus]

He's correct to look at all the stress categories and separate off the peak stress using stress linearisation. The resulting total primary plus secondary stresses are then limited to twice yield to prevent gross deformation. Thermal stresses are usually classed as secondary (strain dependent) so to remove those stresses will (conservatively) leave primary membrane and bending stresses with their stress limits of the allowable design stress, and yield stress respectively.
Checking the limit for the number of cycles seems a waste of time from what you say, but with thermal loads I'd suspect that there'd be some cycling of loads at least.

Tata

 

[adamuk]

Finite Element Anaylsis can produce any localised stresses you want dependent on the geometry (which may or may not be actually how it is in service) and the mesh (have a chat with the analyst and maybe you can tweak the model to see this in action).

This is recognised in the industry hence why linearisation is carried out.

Adam Potter MEng CEng MIMechE

http://www.ax-ea.co.uk