Continuous footing FE modeling: different results

[mpiciu]

Hi everyone!

I'm trying to model a continuous footing foundation: Winkler springs bed under a continuous slab with one or more loaded nodes (corresponding to the columnus).

The problem is that I modeled it with two different softwares (Strand7 and Ansys) and I obtained two different results.

As long as I consider just one column (i.e. one nodal force in the middle of the beam) I have the same results with Strand and Ansys, while considering a longer slab with more loadings they grow different:
-Strand 7: approximately same values for max momentum and displacement between one node and multiple nodes models;
-Ansys: max momentum and displacement get lower when the number of loaded nodes increases.
With 50 loaded nodes Ansys max momentum is 20-25% lower than Strand 7 one.
Of course the distance between loaded nodes is always the same.

My intuition would make me think that Ansys model is (more) correct, but on the other hand I'm way more confident with Strand7, therefore I cannot understand where I was wrong.

Any tips?

Thank you very much!

 

[IDS]

Can post your models (both Strand7 and Ansys). At the moment it's not clear to me exactly what you are doing.

Have you compared the results with a simple "hand calculation"?

Doug Jenkins
Interactive Design Services

http://newtonexcelbach.wordpress.com/

 

[mpiciu]

Ok then here I have a few screenshots.

Problem data:
Beam section: 200*200 mm2 (square)
Elastic modulus: 10000 MPa (Poisson ratio=0)
Length between vertical forces: 4000 mm
Nodal forces: -12000 N
Spring axial stiffness: 1000 N/mm

ONE nodal force - Strand 7 results:

ONE nodal force - ANSYS results:

10 nodal forces - Strand 7 results:

10 nodal forces - ANSYS results:

 

[Denial]

As a non-user of ANSYS (but a one-time regular Strand7 user) I am having trouble working out what those Ansys diagrams are trying to show me. Can you somehow arrange for Strand7 / Ansys diagrams that enable us to compare apples with apples rather than apples with meteorites?

Having admitted total ignorance I will now venture an opinion that, qualitatively at least, the Strand7 diagrams look generally sensible. I would expect a region of hogging moment midway between the load-application points, which that wavy Ansys line in your bottom diagram does not seem to show. That same wavy line also does not seem to show the expected angular discontinuity in the bending moment diagram under the points of load application, but that might simply be an artifact of the low resolution at which the diagram is displayed.

If I was tackling this problem (the problem of reconciling differing results from different software, not the problem of designing the foundation beam) I would simplify the model considerably by assuming the beam was infinitely long, then modelling just a typical 4000mm length. This model would run from the midpoint of one bay to the midpoint of a neighbouring bay, and would model the infiniteness of the actual length by the use of appropriate symmetry boundary conditions at its ends. (These boundary conditions would restrain in-plane rotation while allowing vertical movement.)

 

[IDS]

It looks like a units problem; the Ansys results look like the spring stiffness is much less.

What vertical deflections are you getting?

With Strand7 I get 739 mm for the single load and 713 mm for 10 loads (under the central load).

If you uploaded the models I wouldn't have to ask.

Doug Jenkins
Interactive Design Services

http://newtonexcelbach.wordpress.com/

 

[mpiciu]

Thank you all, I'm sorry I'm new to the forum now I get what you meant by "upload the models".

@IDS: I get the same values, now I should have posted the models so you can check everything.

@Denial: I started to solve a 4000m beam because I have the "hand calculated" solution for that, which is the same as given by Strand7 and which Ansys get very close to.
I'm sorry but I don't know why the beam doesn't display the contour as it should. I'm still trying to solve this thing, I'll post a better image as soon as I can.

 

[mpiciu]

Looks like I coudn't upload more than one file.
The one in the previous post was the STRAND 7 model.
This one is the ANSYS input. Note: "NUMERO PILASTRI" means "COLUMNS NUMBER".

 

[IDS]

I have downloaded your Strand7 model and run it. It gives almost identical results to mine (difference in 5th or 6th significant figure). I used the Node-Stiffness attribute, rather than spring-damper elements, which probably explains the small difference. A third option is to use the Beam-Support attribute (applied to the horizontal beams), which has the option of zero tension if you run a non-linear analysis.

A couple of points on the output display:
To reduce the height of the bending moment diagram, use the slider control in the Results Settings dialog (under "Relative size")
To control the number of significant figures displayed click on the "Listings" icon (right hand end of the top tool bar), then on the icon showing 100 1E2. Select the number format you want, close the results listing, then press F5 to redraw the display.

I had hoped to import the Ansys file into Strand7 (I don't have Ansys), but I couldn't get it to find any beams. The deflection diagram looks like the spring stiffness is less, so I tried reducing the stiffness to 100 n/mm in Strand7. That increases the bending moment, rather than reducing it, but the deflections look more like the Ansys output. I have attached the data file.

That still leaves the exact reason for the difference in the Ansys results unresolved, but it seems likely to be a difference in the spring stiffness and/or beam bending stiffness.

Doug Jenkins
Interactive Design Services

http://newtonexcelbach.wordpress.com/

 

[mpiciu]

Thank you for your help and your suggestion for Strand7 output display.

By the way I also modeled the problem with Strand7 using node stiffness and beam support, results are very close. I was now focusing on the spring model in order to compare it with Ansys analysis.

Actually I guess the value for spring stiffness is correct, because is the same I used for Strand7 model and I also modeled just a spring-beam connection and results were the same.

I'm sorry I cannot open your Strand7 file, I have an older release (2.3.3). What kind of Ansys file do you guess you might open with Strand7? Because I uploaded the input file I wrote for Ansys, but maybe I can save it in a different format/extension.

 

[Denial]

I do not have access to the full Strand7 software, so I had to resort to the (freely downloadable) Strand7 "Viewer" to look at the model file you provided. As far as I can tell within the limitations of the Viewer it is NEARLY fully consistent with the parameters you give in your 28Feb13@09:16 post above.

HOWEVER. I was puzzled by why the model has 400 beam elements and 410 spring elements. I would have expected 401 spring elements. With the limited toolset of the Viewer it was a bit tedious, but I have demonstrated to myself that you have duplicated some of your spring elements. See, for example, the connections between node 2 and node 82. (Strand7, but not its Viewer, can automatically remove such duplications via its "Clean" option.)

 

[IDS]

I have attached the Strand7 file in txt format. You should be able to read it with File-import.

I haven't tried importing ANSYS files before, but Strand7 looks for .dat or .cbd. Looking at the .txt file you sent, it looks like it might be a language issue.

Doug Jenkins
Interactive Design Services

http://newtonexcelbach.wordpress.com/

 

[IDS]

Denial - I cleaned the mesh in my downloaded copy, and now it gives exactly the same results as I was getting (to at least 6 S.F.)

Doug Jenkins
Interactive Design Services

http://newtonexcelbach.wordpress.com/

 

[mpiciu]

@Denial: thank you for your hint, I didn't notice it, by the way results hold the same.

@IDS: your model give me the same results as mine actually.
Are you confident with Ansys? If so, how can I use LINK180 for nonlinear analysis as a compression only element?
Because I define the real costant set as "R,1,1,0,-1" but when I solve it with several load options I see no changes. Moreover if try to set that costant value through the GUI I see the only options "Area" and "Added mass".

 

[IDS]

Are you confident with Ansys?

No, I haven't used it.

Doug Jenkins
Interactive Design Services

http://newtonexcelbach.wordpress.com/