modeling symmetry vs whole model

[tr6]

I have a 24" dia tubesheet plate 1" thick with 2" dia holes on a 60^ pitch on 2-5/16" centers welded on the OD. The tubesheet has a surface pressure on one side and zero on the backside. Also, there are surface loads around the hold diameters. I first modeled the plate as a quarter plate using Z & Y symmetry boundary conditions. This resulted in an extremely high stress point (twice the ultimate)at the intersection of the two symmetry planes; the center of the plate. I then modeled the entire plate. Same loads. The stress results were about half.

My question is: why do the two modeling methods produce such different results?

 

[johnhors]

They shouldn't !

Check that your boundary conditions in both models are correctly applied.

 

[tr6]

Could the "common edge", where the two symmetry planes meet and has the boundary conditions for both symmetry plains applied, be causing a problem?

Other than that, the boundary conditions check out.

 

[johnhors]

"Could the "common edge", where the two symmetry planes meet and has the boundary conditions for both symmetry plains applied, be causing a problem? "

No, it shouldn't.

Perhaps you can attach an image of your model and the BC's applied

 

[tr6]

johnhors,
I hate to ask a dumb question, but how do I attach an image to a thread. I have run the Algor report with the constraints listed and copied an image of the model. But I don't know where to go from there.

I just registered with Engineering.com, but can't figure out how to upload, as the message in the "Attachment" says.

 

[johnhors]

In the reply window there is step 3 below for an attachment or upload.

 

[tr6]

johnhors,
I have uploaded the file below.

thanks for your time and assistance.

 

[LOUDOG]

Was you mesh size larger when you did the whole part? At first glance it looks as though your results are accurate.

 

[johnhors]

I can't really comment without knowing how the surface and part ID's relate to the image.

One thing that strikes me as odd, is that you have rotational supports specified with solid brick elements. I have never used Algor, does it really have rotational degrees of freedom for nodes in its brick elements?

 

[LOUDOG]

No. But when you select the surface and "fix it". All the possible DOF's are restrained. For brick elements the rotational DOF's dont do anything.

 

[GBor]

Something doesn't look right. It seems like the stress pattern would be more uniform unless the loads on the surfaces of the holes are not consistent.

You are correct that the rotational degrees of freedom are meaningless in the context of the analysis.

Can you remove the stress results, turn on loads and boundary conditions and upload that picture? Color by surface, please.

Thanks,
Garland

 

[tr6]

Attached is the loads and boundary conditions. With the surface load and loads around the holes, the surfaces did not show very well, so I've added a second page with just the surfaces shown.

Thanks for all of your help.

For what it is worth, I also ran a full plate and then did a quick check using Roarke for a circular plate, no holes, with a uniform surface load. The Roarke formula fell in the area of the "full Plate" analysis.

 

[GBor]

It is difficult to see the applied loads, but this is what it looks like:

Pressure load on the entire face
Load around the rim of the tube penetrations
Load inside the tube penetrations
Partially constrained boundaries

The boundary conditions should be as follows:

for the face perpendicular to the y-axis, the BC's should be TyRxz

For the face perpendicular to the z axis, you should have TzRxy

I'm assuming you are pinned (Txyz) around the outer edge

The center "corner" of the model should be TyzRxyz

I just looked at the report again and noticed you have multiple parts? What are the three parts? This looks like it should be a single part. Are there different materials for the parts?

When you get a chance, is it possible for you to upload an archive without results? It's a lot easier to troubleshoot when you have the actual model.

Garland E. Borowski, PE
Engineering Manager
Star Aviation

 

[tr6]

Garland,
I can confirm the following to you:
Applied loads are:
1. Pressure on the entire face
2. Load around the rim of the tube penetrations

Boundary conditions:
1. For face perpendicular to the y-axis: Ty, Rx, Rz
2. For face perpendicular to the z-axis: Tz, Rx, Ry
3. For the "center" corner: Ty, Tz, Rx, Ry, Rz
4. The outer edges are fixed. Although I did a "pinned" version that did not change the results.

There are (3) parts: the plate and then (2) fillet welds at the perimeter on each side of the plate. The materials are essentially the same: 316 SST for the plate and 309L for the welds.

The purpose of this analysis was to evaluate the welds, and try to reduce the size.

You have requested an upload of "an archive without results". Could you be a bit more specific? I don't understand what you are asking for.

 

[GBor]

If you go to "File: Archive: Create", it will ask if you want to create an archive with or without results. If you choose "with results", you will get a very large file with all the stress and displacement results zipped up into a file with the extension ".ach". Without results just zips up the nodes, elements, materials, etc.

If you are able to upload the .ach file, I can download it, extract it, and run it just to see what's going on, but it sounds like you have everything set up properly. As mentioned previously for brick elements, "pinned" and "fixed" are identical in the Algor context. Sounds like there is no reason why your results should be different, which is why I am curious about the model...

 

[tr6]

Garland,
I am having problems uploading the file to Engineering.com. The zipped file is 28 M. I don't know if this is too large, but the site "times out". I am trying to find an FTP site, or I can email the file.

 

[GBor]

That is MUCH larger than I expected...doesn't seem like your file should be that large without the results. Perhaps just upload the report file...I can figure out the rest from there. If that doesn't work, try

http://www.yousendit.com

 

[tr6]

Garland,
I ran a larger mesh, which reduced the file size. Same basic problem though.

 

[GBor]

It looks like your loads are applied twice...or, actually, like two different loads of the same type are each applied twice. You have two pressure loads over the entire surface (150 psi and 218 psi) and you have two different linear loads applied as a surface force around each rim (5XX and 828 pounds). Was this intentional? Or could this be the problem?

Also, your boundary conditions were a bit of a mess. You can select multiple surfaces against which to apply the boundary conditions by holding down the control key after selecting your first surface.

The nodes, elements, etc. look fine. Nothing that should cause problems.

When I deleted one set of each type of load (I maintained the higher loads in each case: 218 psi and 828 pounds), I ended up with deflections of 0.116 inch and stress of 114 ksi. Not sure how this compares to your full model results, but when I deleted the symmetry boundary conditions and copied the quarter around, cleaned it up and reran it, it compared well.

 

[tr6]

I had the load sets (218 & 150) set up to be "enabled", or not, for different loading conditions, 218 & 150 were not applied at the same time.

The results do not come close to the full plate model. In the full plate model, stress at the center is 62,000 and displacement is .084".

So, I am not sure where to go from here.

Thanks for your input and assistance.