Hot Spots 1

[jike]

When using finite element analysis, how much attention should I be paying to the hot spots of stress that might just be over a very small area like one or two inches? The high stress levels die out very quickly in the adjacent areas. Don't the stress levels have a tendency to "average" out? If some fibers yield in steel, the adjacent fibers will pick up some of the leftover stress.


I am doing mostly building analysis, things like reinforced concrete, unreinforced and reinforced masonry and structural steel and don't want to be overly conservative.

 

[GBor]

jike,

Good questions which are virtually unanswerable. "one or two inches" could mean that they are less than 1% of your model or that it is the bulk of a local model. "one or two inches" could mean less than an element width or twenty elements wide. The responses should vary depending on your situation. Can you give us a little more detail: "building analysis" meaning the entire building or local joint models, "one or two inches" in context of a mesh density or model size, software (whether it produces "smoothed" or "unsmoothed" stress results), etc.

Thanks

 

[jike]

One particular example is a masonry wall 30 ft wide x 16 ft high. 120 plate elements. each plate element about 2 feet square.

Another example might be a foundation mat with similar sized elements.

 

[TGS4]

In evaluating stresses from a FEA, differentindustries use different approaches. One that I am familiar with (from the pressure vessel and pipign industry) is that the engineer linearizes the stress across a section, with specific limits on the membrane (average) and membrane-plus-bending (linearized) stresses. The "hot spots" that you refer to are likely caused by inappropriate boundary conditions and are ficticious. A linearization technique might elminate that.

However, there are many the eschew the linearization techniques and go straight to the failure mechanism and place appropriate design margins on things such as plastic collapse. For a masonry wall, I am sure that you have a failure mechanism that is appropriate (maximum directional tensile forces which would crack grout for example). Find the "load" that corresponds to the "failure", and back the load off by your design margin.

Unfortunately, too many engineers focus their energies on the "stress" and the "allowable" without focusing on the "failure mode". From my perspective, that is misguided at best, and potentially disasterous at worst.

BTW, don't be afraid to throw a few more elements into your models. 2ft square elements seem to me to be huge. A quick mesh refinement study will quickly point you towards where your "hot spots" are a problem and where they are artifacts from bad BCs or numerical issues.

 

[GBor]

I'm assuming you are seeing spikes in the corners near boundary conditions? Probably not a problem, but one way to be more confident is to increase your mesh in the corners and rerun the analysis. This is called a "mesh sensitivity" analysis and is often overlooked even by analysts. If your results change by greater than about 10%, continue to refine the mesh until the results stabilize. You can refine the whole model, but this may become too much for your system.

You may also want to consider a "local model" where you take the displacements from your global model and apply them as constraints to a model of the area of concern. If you need more information on this, several of us on this forum could "talk" you through it.

Garland E. Borowski, PE
Borowski Engineering & Analytical Services, Inc.

http://www.borowskiengineering.com

 

[jike]

Thanks for the advice!