stress convergence

[madfem]

Hello,

I am doing a non linear analysis, displacement controlled.
i am able to get a good residual force convergence & displacement converegnce.

When i have a visual look on the elements, the displacements are well distributed among all elements
However when i check the stresses,stresses are concentrated on few points( like a stress singularity)
although this is not be the real situation. Its a large symmetric structure and stress need to distributed.

Do anybody have a solution to arrive at stress convergence ?

 

[johnhors]

Your model description is very lacking in detail.

"non linear analysis" - is too vague, what kind of NL , material, large displacement , contact ....?


You are applying displacements, how? To a surface or individual nodes? How is the model restained? Again at a surface or individual nodes?


The normal means to get stress convergence is to increase mesh density, but with stress spikes in your results this is unlikely to help. You need to remove the cause of the spikes first.

 

[corus]

Why seek convergence on what is not real? (No replies from Buddhists please). Unless you've made a mistake in your model, the singularities are probably due to point restraints or loads, and should be located away from the area you're interested in.

corus

 

[madfem]

Let me give some more information.

its an incompressible FEM simulation using rigid plastic formulation.So its a very unstable simulation.

it is to simulate pressing of glue (glue is modelled as non newtonian fluid & rigid plastic formulation).
The glue is pressed against a large struture with a heavy load. Target is to compute the displacements and stresses on the structure which is still elastic
displacements and loads are applied on the nodes and the model is restrained at nodes only.

My problem is i find stress concentration on this large structure which is not realistic.
Can anyone give me ideas to overcome this problem.

 

[johnhors]

Do you really need to model the glue, if it is only the large elastic structure that you are interested in?

As corus said point loads and restraints create singularities, therefore consider applying the "heavy load" as a distributed pressure on the large structure and forget about the glue. Ideally you should be able to apply pressure loads where the supports are to balance this load out and then use minimal supports, this will eliminate all your singularities.

 

[kellnerp]

Ok, so you are looking at stresses that occur while bonding two parts. You are modeling the glue line (much less rigid than the structure) and looking at how stresses transfer across the glue line. The bond line is very thin and you are using solids in a single layer for that and connecting to either side with contact elements. The aspect ratio of the elements in the bond line is also very high. You are seeing the high stresses at certain nodal locations along the bond line and you are using higher order elements.

Is my Crystal Ball working here?

TOP
CSWP
BSSE

http://www.engtran.com

http://www.niswug.org

"Node news is good news."

 

[gwolf2]

How about a picture?

 

[corus]

I'd consider the worst case and just look at the stresses and dispacements of the structure without the glue, if that's the most important factor. You can assume that the heavy load is rigid possibly and just apply the load directly either as a pressure load, or as contact between the two surfaces.

corus

 

[kellnerp]

stresses are concentrated on few points

Out of curiosity, is the displacement anywhere in the model a significant portion of the thickness of the element? Particularly where these peaks occur?

If you plot stresses along a continuous edge where these spike occur is it a smooth plot or jagged?

TOP
CSWP
BSSE

http://www.engtran.com

http://www.niswug.org

"Node news is good news."

 

[madfem]

kellnerp,

your crystallball is almost there. thats what i am trying to do.i trying to evaluate the stresses on a structure where a heavy load is bonded onto it. i am using a few layers of solid to model the glue and will have a pressed glue at end of simulation but i am not using higher order element but normal(linear) solid elements.my stresses are not at the bondling line. quite far away from the bonding line. but anyway i will follow yur advice and look at the displacment plot and let you know.

gwolf2, i may not be able to give you a picture becoz of confidential work.but i am free to discuss the topic.

corus,
i cannot discard the glue. apart from stress and displacement, i need to evaluate the force required to press the glue.since the bonding is a robot operation, the forces are also of interest.
but yes if i need to evaluate only stress, then i can do it as worst case scenario and neglect the glue.

 

[kellnerp]

What is the aspect ratio of the glue line elements?

TOP
CSWP
BSSE

http://www.engtran.com

http://www.niswug.org

"Node news is good news."