Is this a stress singularity? And why can't I avoid it with a fillet?

[Anonymous]

Hi again,

I have an assambly with a few parts. In a FEA analysis I get a high stress value between two faces of two parts that are perpendicular. See figure below.

I did read a lot about the Stress Singularity problems. Could this be such a case?

When using finer meshes, the stress does not increase. So maybe it is not and the indicated stress is the real stress at that point.

Because I need a big safety factor, I want to get the stress down. So I placed a fillet.

(I did not want to convert it to a welding assembly so I used a extrude for the fillet shape).

Now the maximum stress is moved to some other similar point as indicated in the figure below.

The stress is a little lower, but it mainly moved to the edge of the filler. I did not expect that.

How can I get rid of this point with (relative) high stress?

Why is there such a stress peak at the edge of the fillet. Is that expected? I hoped the stress peak with dissapear.

Any suggestions?

I also added the design (pack and go), but without the large FEA files (you can skip them when opening).

I would be very glad if someone could make some comments and give me hints to minimize this stress peak.

[Dan_Margulius]

Hey,

Stress singularty occurs when A=area its 0 in the equation so the stress is infinte. This is not your case. Color red doesnt always mean bad results. From calculation I see FS is 3.21 so from the FS view you are on the safe side.Also the Displacement is 0.3 mm...

What is the max displacement allowed? 

What is the FS required in your case? 

A design change can also be a different material  with better material properties and that will take your stress down.

Regards,

Dan 

[Anonymous]

---

@Dan_Margulius wrote:

Hey,

Stress singularty occurs when A=area its 0 in the equation so the stress is infinte. This is not your case.

...

What is the max displacement allowed? 

What is the FS required in your case? 

A design change can also be a different material  with better material properties and that will take your stress down.

Regards,

Dan 

---

Thanks Dan,

Good to know this is not a stress singularity.

You say: this is not your case. Do you conclude this because the stress doesn not get higher with finer mesh?

Maximumdisplacement is no problem.

I will need a better safety factor because of regulation and certification.

I assume a better material gives the same stress, but larger safety factor.

What I still do not understand is why the edge of the fillet gives a stress peak. Is there a problem with the mesh?

Can I avoid this somehow

[Anonymous]

Hi

I have removed the h-refinement to 0 and unselected part based measure for assembly mesh.. This gives a more uniform mesh rather than having a highly densed mesh around the smaller parts. The resuly obtained is shown below

I assume all the parts are made from the same material. If this is the case I suggest create shrinkwrap subsitute to convert the assembly into a single part this help to create a more uniform mesh and more likely remove the red hotspots you are getting.

I have a similar problem like this in my book and shows suggested workflows on ow to remove/avoid hotspots like you are getting. You can get more info about my book from Amazon.com

Hope this helps

[Anonymous · Accepted solution]

Younis:

I compared the meshes:

A: meshes with h-refinement and ''part based measure for assembly'mesh'' both on (my approach).

 

B: meshes with h-refinement and "part based measure for assembly mesh'' both off (your solution):

I think your approach looks more realistic.

What I do not understand is that the resulting meshes seems identical. But the stress is rather different.

How is this possible? Why is there a stress peak at the edge of the fillet in my approach and not in yours?

About your book: I bought yout e-book ''Up and Running with Autodesk Inventor Simulation 2011''.

Is this topic about hotspots new in the 2012 book?

Will this 2012 book give me much more/new information compared to the 2011 e-book?

[Anonymous]

With your types of models I would suggest not to use h-refinement. and part based mesh. Just change the average element size to different values, maybe local mesh control, to see if the results have converged.

Secondly I also recommend that you shrinkwrap the assembly to a single part this will helps towards removing the problems you are having.

This what I always recommend to trainee's and customers.....

Finally I happy to hear you have purchased my book. and fyi there is no difference between 2011 and 2012 books. Hoping to extend material for 2013 edition.

[kmeldfreyssinet · Accepted solution]

Hi,

I did analisys for you case.

There is nothing rong with geometry. You are getting stress peek because your mesh is not well defined.

My highest value is 46,5 MPa, and with local mesch controll about 32 MPa.

Problem in this case seem to be way of conection between Rails:n and  HokerProfileVoor:n. Rails "go in side" the profilem and filet is done in Hoker... With settings for mesh you have and local mesh control you use you are getting some really nusty elements in area of conection.

Try to put:

Avrage element size =0,2

minimum element size =0,05

gradi factor =1,05

max turn angle = 90 deg

and some local mesch control. I tried fiew things, for some It was not possible to compute results.

Genarally elements of "bad" shape are cosing problems. try to make your mesch as uniform as risonably possible. Inventor uses triangles. Try to force them to have decent edges ratios (all edges schould be of similar lenght) at leest in areas you would like to have good results.

Hope this helps.

Cris.

[Anonymous]

Thanks both for the comments and suggestions.

So it is all about better (cleaner) meshes!

I tried your suggestions and the mesh can indeed be made much cleaner.

But (in my opinion) it is not easy. There is so little mesh control available.

I would like to be able to specify a mesh size per part.

Also I would like to be able to specify a mesh size for some area of the assembly. Like local mesh control but then for multiple parts at once. So that you can increase mesh seze at some area of interrest.

When is a mesh good enough? Is there some literature available about that?

What are do's and do'nts for meshing?

Are there any  rule of thumbs?

What can I do to check if a mesh is good enough? Decresing mesh sizes and if stress doesn't change much then it is ok?

Thanks again.

[Anonymous]

Hi what I suggest is to run a analysis once then run the analysis again with a different mesh and see whether the stress in the area of interest has not changed mych (within 10% is the norm)

[kmeldfreyssinet]

Hi,

I realy can not clearly and in scientific way answer your questions 😉

In my case knowlagde of basics of FEM is very usefull. I had this on studies. It was not realy very extencive but in the end very usefull. I am now aware of how stifness matrix is calculated and what has influence on it, and how are after stresses calculated.

I realy do not have any spetial experience, nor did special courses.

My only adwise to you could be:

If you are getting something that schould not be there because there is no fisical explanation\justification to that = something is wrong.

and also:

always try to force mesch to look "good". generally quad elements are better than triangle but since you realy have no influence on element type in this softwer you schould enshure getting goog mesch by "hand".

You can ofcourese send your wishes to Autodesk via feedback page but schould not count this will result soon. Full control over mesch is practicaly only avialable in advanced FEM softwear, or sometiones in open source, but in such case interface is very poor, althought calculation methods are the same.

Opensource for exumple is ELMER.

Cris.

[Anonymous]

Youniis,

I have been reading your book as well. In terms of adding the fillets to avoid singularities, I have attempted to do so. Unfortunately, it is not allowing my to select the second set of faces for the 'face fillet' tool. I used frame generator to create the structure. Is using frame generator the reason I cannot utilize the face fillet feature between components created by the frame generator?

Thank you.

Jay