Linearized stress results - design values?

[Masse69]

Does anyone have any experience from standards like ASME or similar, etc where linearized stress results are used for dimensioning?

I have no problem understanding the linearized stress query, but how is this result used for dimensioning, and how is it justified?

I recall that it is common in pressure vessel design and possibly other design situations. What is the design value for the linearized stress result, and how is this justified? Can reference be made to ASME (for example) for an arbitrary product that is not required to be approved according to a standard? In my case it is a sheet steel bracket that holds a component in a vehicle.

I'm working on a shell model, where I cannot do elastoplastic analysis (software limitation). I end up with some very small regions exceeding the yield stress. If the high stress area is sufficiently small, then the linearized stress will be considerably lower than the peak stress, and then it should (I guess?) be OK to use the linearized stress for dimensioning.

Can anyone confirm this?

What should the design value be, for the linearized stress?

How is this justified? What standard (ASME-xyz? or similar) should I refer to?

Are there other techniques to use results from a linear analysis with very small regions exceeding the yield stress? The challenge is to get the design approved without having to run the non-linear analysis, which I can't to because of limitations in my software (Creo Simulate, no material nonlinearity for shell models)

I would be most grateful for any feedback.

/Mats L/

 

[rb1957]

i suspect that you're bound up in red tape. i'm sure the processes you're bound to don't allow waving off some stress peaks as "not significant" ... is your peak greater than ftu ? as a element averaged stress ?? are stresses (von mises or max principal) less than ftu acceptable ? or does your approval process say stresses greater than fty need NL ?

can you test the detail ?

another day in paradise, or is paradise one day closer ?

 

[Masse69]

Thanks for swift response, rb1957.

In fact, there are no legal/formal requirements in this case. My client designs monitors for vehicles and public places. To my surprise there are no legal requirements other than that the design should be "safe and reliable". So me and my client have agreed on some realistic design loads.

In this case, I'm analyzing a sheet metal bracket that connects a monitor to the interior structure of a bus. But neither my client nor his client (bus operator) has any experience from FEA. So even though I feel confident about the design, I will have to explain the "red spots" somehow. Pointing to a procedure described by for example ASME would calm them I think.

Peaks are near, and sometimes, at singularities (sharp corners) above ftu.

I can work around or mitigate some of these "hot spots" by making some less conservative assumptions regarding constraints etc. I have assumed "worst case" for some of the constraints. The design contains sharp corners at the relief of sheet-metal bends where rounds can be added. I can even require the design engineer to create bends with obround reliefs instead of the sharp rip reliefs that he has used. But the issue is that I have a series of variants that need to be analyzed, so I would ideally like to have a stress evaluation technique that allows me to work through the list of designs without too much hassle for each design. Plus of course that I can't do non-linear analysis with shell elements. I have just had ANSYS installed, but I'm an Ansys novice, it would take me weeks to create a single model. I need training, but as always there is little time...

/Mats L

 

[rb1957]

if you have no Rigid procedure then you've got room to manoeuvre. If you're convinced that it's a trivial issue, hide it using element centroid averaged stress. if the peak is less than Ftu then you've got less of a "problem". if you want to be really sneaky ... tweek the legend levels ...

if your client makes a big deal over it ... how difficult to test it ?

another day in paradise, or is paradise one day closer ?

 

[corus]

In pressure vessel design these hot spot stresses would assessed against failure through fatigue damage, so they can't easily be dismissed if your load is cycling. If the high stresses are at a sharp corner then you could only dismiss them by saying it's an approximation of the model to assume a sharp corner rather than rounded. Of course if the stress is compressive, and away from a weld, then you wouldn't expect fatigue damage unless the stresses were above yield and you were getting residual tensile stresses once unloaded, and the load was cycling. Never try and 'fiddle' the results by changing the plot contours or using photoshop to blank them out. Once found out you'd never get work again.

 

[rb1957]

i wasn't suggesting photoshopping the results (though i know someone who would do this) ...

tailoring the legend levels is sneaky but only a presentation trick. your analysis should still mention the maximum/critical stresses found and justify their acceptability (rather than saying "fig 1 shows that the stresses are acceptable").

using element average stress i think is reasonably acceptable practice and will blend out nodal peaks.

another day in paradise, or is paradise one day closer ?

 

[realtime2]

In the vehicle industries its quite common to get hot-spot stresses
over the yield and even ultimate if one is doing an elastic analysis.
The trick is to perform a "plasticity correction", based on energy, and
then use the corrected local stresses and strains to check the fatigue life.
The most commonly used plasticity correction for automotive body or chassis bits for
example, is the Neuber Plasticity Correction. Details are available on the
internet. A lowly SAE1005 for example will have slight plasticity even at
10**6 cycles to failure, so its not that big of a concern by itself. Just don't
use the elastic "stresses" alone to predict capability.

 

[Masse69]

Thanks for valuable input.

I'm new to this forum, it has worked great for me so far...

B.R. Mats Lindqvist