FE analysis of steel bellows 4

[elogesh]

Hi,

I am analyzing steel bellows made of SS-304 stainless steel material subjected to axial displacement +/-12mm has to withstand 0.5 million cycles. Endurance strength of the material for infinite life is approx. 160 MPa. Stress from the linear static analysis for axial load is around 400 MPa.
My generic doubt is whether linear static model is adequate enough to represent the load-deflection characteristics of the bellows?

This is because the designer believes,stresses are over predicted from FE analysis.

If additional information is required in this regard, kindly let me know.

Logesh

 

[johnhors]

With a requirement for half a million cycles, anything that is not in the material's linear elastic range is surely not going to last? If you mean non-linear geometry, then this also could be a concern. Can you post a picture of your mesh?

http://www.Roshaz.com

 

[Andries]

Hi elogesh,

Bellows are supposed to absorb large axial and lateral displacements while still remaining within the elastic range of stress. You may have to consider a large displacement, small strain analysis that will require a non-linear code.

Andries

 

[gwolf2]

Large displacement (nonlinear geometry, elastic material) will dramatically reduce elastic stresses on a bellows as the thin walls transition from bending to membrane stresses.

Half a million cycles for a metal bellows does not sound right though. At a guess the nonlinear analysis will drop the bending stresses by about 30% still leaving you with 280 MPa to deal with.

I question the overall concept here, metal bellows are usually used for low number of cycle thermal applications.

gwolf.

 

[btrueblood]

I've modelled several bellows recently, using a cheap, canned FEA code that came with my CAD software. Modelled one half of a single convolution, using symmetry. Had results that matched the bellow's manufacturer's numbers within 1 or 2%, and matched our test data within a few %. Is this a welded or formed bellows?

Post your model, or at least a jpeg of the mesh, and show/describe your boundary conditions.

 

[elogesh]

Hi,

Thanks for your response.

The mesh model is included as an attachment.

I will be re-running the analysis with large deformation effects included and also will share the results.

The bellow needs to be analyzed for axial and bending loads.
For axial loads it needs to withstand half a million cycles, for bending loads it needs to withstand 50000 cycles.



Logesh.E

 

[btrueblood]

Well, again, for axial loads there is little to be gained by modelling the entire bellows. The mesh you post looks overly coarse, especially at the convolution bends. Try modelling one convolution, and see if you can't improve the mesh density by a factor of 10 or better through the bend. Just this change alone may change the apparent stiffness of the bellows significantly.

Modelling the bending condition may be a little bit trickier, but a suitable moment condition at the boundaries of a single convolution ought to be do-able.

 

[corus]

I'd agree with btrueblood. In general the advice is to use a minimum of 1 element per 15 degrees, if you wish to capture the stiffness of an arc. For your axial loads you should be able to use a thin shell axisymmetric model.

corus

 

[gwolf2]

If you have ABAQUS you can put non axisymmetric loads onto axisymmetric models

Then all you need to do is find a post processor to look at the results! I think CAE might do it.

 

[kellnerp]

What type of elements are you using?

TOP
CSWP, BSSE

http://www.engtran.com

http://www.niswug.org

"Node news is good news."