• From: Finland
  • Registered: 2019-03-28
  • Posts: 138

Topic: Different results after recalculation

Hello!

I have every time new results after I recalculate the model. Below you can see table with number of iterations of 3 different analyses. "*" means non-converged step.

Is it normal that same model gives different results after after each solution? Maybe there is a problem with solution parameters...

        N1               N2                N3
1         2               2              2
2         5               5              5
3         26             31            31
4         29             39            44
5         136           200*        200*
6         200*                        200*
7         191                          200*
8         171                          200*
9          77    
10        182    
11        67    
13        117       
14        180       
15        200*       

GiD+Atena model

Quote Post 1

  • dpryl
  • CC software developer
  • Offline
  • Registered: 2007-07-13
  • Posts: 1,107

Re: Different results after recalculation

Dear Pavlo,
1. I would strongly recommend to run the ATENA Update Check utility, as the easiest way download the current installer (ATENA 5.6.1i or at least 5.3.5i). If your maintenance has already expired, please contact us for a renewal offer.

2. I see elements with problematic shapes (element edge aspect ratio about 5:1, while the recommended maximum is about 2-3:1 for volume elements).

3. We generally recommend to set the Conditional Break Criteria such that the analysis stops when the convergence errors are large at the end of a step, see ATENA Troubleshooting, 2.4.2 I get the message "The execution is killed due to violation of stop iteration criteria", what does it mean?

4. I suggest to consider Shell elements for the ceiling plate (ATENA-GiD User's, 5.3.2 Shell Material) and possibly beam elements for the columns (5.3.3 Beam Material).

Regards.

dpryl's Website

Quote Post 2

  • From: Finland
  • Registered: 2019-03-28
  • Posts: 138

Re: Different results after recalculation

Dear Dobromil,

Thank you for prompt replies.

1. I have restarted the model in Atena 5.6.1. Now I have all steps non converged which didn't happen before in older version.

2. Can you advice how to check mesh in GiD "Mesh quality"? Which criteria should be used and which values are OK for practical use?

3. Now I have all steps non converged but I don't see any troubles with structure behavior. How can I see that convergence cannot be achived in the specific elements? I checked residual forces but they seem distributed uniformly. Is this the correct way to estimate problematic elements in terms of convergence? By the way, mostly problems are located in Master Slave connections.
Screenshot: https://drive.google.com/open?id=1z5fdb … m4fqlx55i5

4. In fact, I have solved the same problem in another software with multilayer shells and frames. I'd like to verify the results obtained from simplified model with ATENA 3D modeling results. Also, I have modeled all rebars in 3D and will will take too much time to recreate the model with reinforcement zones. Is there a User guide how to model RC slab with shell elements and real reinforcement? Can't discrete rebars be used together with shell concrete?

5. I found that Atena Studio works very slow in some simple procedures like turning off cracks. E.g., I have some scalar results activated and cracks. Then I turn off cracks and I have to wait about 1 minute to see same view without cracks. One step analysis relults is 2 GB.

Quote Post 3

  • dpryl
  • CC software developer
  • Offline
  • Registered: 2007-07-13
  • Posts: 1,107

Re: Different results after recalculation

Dear Pavlo,
ad 1.: Please check + correct following Troubleshooting, 2.1.19 Problems reaching convergence and understanding ATENA convergence parameters and 2.4.6 Floating point problems - Division by zero, Loss of >7 digits on the matrix diagonal, CCFEModelExc: CCPardiso solver error: zero pivot,  Warning: Sign of diagonal changed when reducing equation, Floating point exception: Multiple floating points traps, Zero or negative jacobian,  InvertA: zero determinant, CCMaterials Extended AExc: CC5ParamYield F: project_stress_on_f_division by zero, and similar.
If you still have problems, you can follow Troubleshooting, 2.1.1 to sen us your model etc.

Ad 2.: You can see some information in GiD under Mesh - Mesh Quality. However, the basic (=most critical) for most volume and surface elements in ATENA is the edge length aspect ratio (to be kept under 2-3:1). The next are the angles (shown in the Mesh Quality dialog in GiD) - narrow angles are problematic, from about (20-)30 degrees OK.

Ad 3.: Most of the convergence errors are global measures (for more information, see Troubleshooting, 2.1.19). Moreover, problematic elements or extreme stiffness ratios typically make the system conditioning ill, which results in degraded precision / numerical problems globally (not just localized at the problematic regions).
If you see unexpected localizations at the M-S connections, I suggest to make sure the orientation of the connection is as recommended (side with coarser mesh as Master; for comparable element sizes, larger surface as Master).

Ad 4.: No special guide is recommended :-) 1-D bar elements can be used in Shells the same way as in Volume elements (with the exception of the older Ahmad Shell elements and bars in thickness direction, which would make no sense due to the undeformable thickness of these elements; we generally recommend the newer IsoShells). Moreover, you can freely combine reinforcement layers and discrete bars.

Ad 5. We are continuously working on performance improvements of ATENA Studio (and also of the solution kernel), however, with such an extreme model, I am afraid you can not expect instant response...
I suggest to see ATENA Troubleshooting, 2.2.17 The analysis runs too long. How can I speed it up? and 2.4.1 Out of memory error during ATENA analysis for basic related tips.

dpryl's Website

Quote Post 4