in "real life" there are no such things like "singularities" of stresses. Material will go plastic or degrade in other form because "unlimited stresses" does not exist.
You are simulating linear elastic material, so your stresses AT THE CRACK TIP will eventually be unlimited if your element size approaches zero .... Conclusion: Simply do not evaluate stresses at the crack tip. At the rest of the cracked area however you can evaluate stresses and you can even work with nodal stresses.
Thank you Martin,
As you probably noticed, I am modeling a planar crack. I do that by applying a symmetric constraint only on an area of the mid surface of the model. The rest of the surface is the crack. I think a better solution is to evaluate stresses close to the crack, not to the crack edge itself. Another approach would be to do anonlinear analysis?
I also think that using elemental stresses is a better solution, since I am avoiding the crack edge.
Am I right?
the pictures do not match to the others. Did you change something?
I am still thinking to the crack singularity.
I attached more images:
Elemental - assembly.jpg, that presents the elemental stress distribution for the assembly
Elemental averaged.jpg, that presents the stress distribution for the stell plate
Elemental nodal - assembly.jpg, hat presents the elemental nodal stress distribution for the assembly
Elemental nodal averaged.jpg, that presents the stress distribution for the stell plate
Plates-assembly.jpg, that presents a drawing of the assembly (stell plate and composite plate).
As you see, the maximal values appear at the point where the steel plate and composite plate touch in the symmetry plane. Can we speak here about a singularity??
As you can also see, there is a big difference between the elemental and nodal-elemental stresses, ans more than that, the area where the stresses are maximal is different for the elemental and nodal elemental values.
Many thatnks for your remarks
From my point of view I think that you started wrongly...
For first (in my opinion) you should clearly define what you want to do:
- check the area where the stress concetrations will appear or,
- to read the stresss from the notch.
If you would like to investigate cracks stress, you must not do this with linear solver (I assume you usesd SOL101). With use of linear elastic solver you can only investigate where you can expect stress concetrations in your structure (assuming that the model is correctly prepared). Everything esle, like stress checks or crack propagation you have to do by use of non-linear solver. SOL106 or SOL601/106, with includance of real sress-strain curve and material failure... and so on.
There is no way of obtaining 'real' results of stress in such notch areas by using linear solution.
What's more in such notches it always better to model 'full' model rather than applying symmetrical BC. I know that many will say that it is all the same. Maybe it is but for 'easy' cases, not when you want to read the stress from the edge where you apply BC. just see the approach for calculations of fatigue, where notch effects are critical. Any time you have to modell full structure and such simplifications are not acceptable.
What's more the area under your considerating is a region where more more than one BC could have an influence on the stress. Your 'simple support' of the block also influences on the stress in your notch (remember that your elemetns deforms in all directions, especiall under high loads).
So... that's my opinon for this. I hope it's helpful
Dear Tomek (I hope this is correct),
Thank you very much for your answer. It is very useful for me. Regarding my problem, I have to stress that I am not interested in the crack itself, but in the way the steel plate and the composite plate interact when the crack has different lengths. The finite element analysis tries to model a real experiment that has been done. In this experiment, the specimen (the two plates, bounded together) has been submitted to a variable load. We assume, analysing the way the crack looks, that it intiated for values of stress smaller than the yield stress (this is whay I am doing the analysis with 101 solution).
Regarding the use of symmetry, as far as I can imagine, it is the only solution to model a plane crack. If I would have used the whole model, I do not know how to model a plane crack, that is with zero thickness.
So, this is my problem and I am a little bit confused about the differences between the elemental and nodal-elemental stresses, and also about the great values of stresses in the area close to the crack, in the symmetry plane.