Seismosoft | Earthquake Engineering Solutions

Dear Seismo,

I am modeling a single column with several springs using static time-history analysis, when I set the curve multiplier =0.09 or more in the Applied Loads, then the analysis will terminate with the "fbd_Ite" message. Would you please tell me why and how to solve it? I have sent you the model just in case you may want to have a look. Thanks

Hi Tracy,
This message appears when the maximum number of iterations (default=300 iterations) is reached in the internal element loop of the force-based elements (infrmFB and infrmFBPH), before internal equilibrium having been achieved. It is a typical divegence message.

As a first step increase either the number of iterations or the convergence tolerance from the Elements tab of the Project Settings. Alternatively, the ¡Do not allow element unbalanced forces in case of fbd_ite¢ option may be unchecked.

Further, measures on the global level may be taken, for instance the analysis load step can be decreased (by increasing the load factor increments or the time-steps), and the global convergence criteria can be increased.

For a better understanding, refer to the specific documentation of the force-based elements [e.g. Spacone et al. 1996; Neuenhofer and Filippou 1997].

Thank you for your answer. I have increased the number of iterations as the model I sent you, but it still doesn't work.

If we decrease the analysis step,the same problem comes out, it terminates at small strain and can not reach the ductility I want. Thank you.

May I know how to solve this problem?

Have you tried all the possible solutions that we suggested (5-6 in total)? For instance have you unchecked the ¡Do not allow element unbalanced forces in case of fbd_ite¢ checkbox?

Note also that we did not receive anything at support@seismosoft.com, but in any case support through the forum is always faster and preferable, since other users will be able to read and learn from the answers provided.

SeismoSoft Support

Yes, I have tried all the methods you mentioned. change the maximum iteration number to 1000;uncheck the 'Do not allow element unbalanced forces in case of fbd_ite' option;decrease the load steps. But it doesn't work.
Now, I tried to set up a single column (18'), fix in the bottom, use 'infrmFB' element, static time-history analysis. Model A is set up by one element (the element is 18'); Model B is set up by 72 elements (each element is 3 inches, 3-inch*72= 18'). The result is Model A runs well, but the Model B terminates with the 'fbd_ite' message. May I know how it comes?

Dear Tracy
Change your element type to "infrmDB" in B model.I am sure it will run well.
I have tried many times to verify experimenatal column like PEER column data with "infrmDB" and in that case didn't have converge problem. make no doubt in that case analysis "infrmDB" is much better than "infrmFB".

Tracy,

I carried out a search in this Forum on "convergence difficulties" and found a post (http://www.seismosoft.com/forum/topic.asp?TOPIC_ID=287) where the following is stated:
quote:use a smaller number of integration points (2-3) for short force-based elements (you can actually use displacement-based elements in such cases)

Then I searched this Forum for the expression "element length" and found a post (http://www.seismosoft.com/forum/topic.asp?TOPIC_ID=196) where a similar statement is made:
quote:I experienced numerical instability when using a very fine element mesh, while with a coarser mesh numerical instability improved and delivered satisfactory pushover results. I investigated the results of the coarser mesh and found that the results were accurate.

I then also read the 'Element Connectivity' page of the program's Help System/User Manual and saw the following statement there:
quote:Whilst a too coarse finite element mesh may lead to the impossibility of accurately reproducing certain response shapes/mechanisms, an exaggeratedly mesh refinement (e.g. resulting in frame elements with length much smaller than their cross-section height) may lead to unnecessary long analyses and, in some instances, to less stable solutions. Hence, users are advised to make well balanced and judged decisions on the level of mesh refinement that they decide to introduce, ideally carrying out sensitivity studies in order to define the point of optimum balance between accuracy, numerical stability and analysis' run times.

Subsequently, I read the paper by Calabrese etal. (2010) suggested in the 'Material Inelasticy' page of the Help System/User Manual, and it became evident that the employment of very small force-based elements will inevitably lead to erroneous or unstable analyses, due to numerical integration difficulties and/or localisation issues (if you search this Forum for "localisation" or "discretisation" you will again find very interesting and informative past discussions on this topic).

In conclusion, it seems thus to me that the fact that your model with very small force-based elements fails to converge is not a surprise, but rather the expected behaviour.

Rui

Rui,

Thank you so much for you reply and advice. Now I am reading the paper by Calabrese etal. (2010) and it is useful. The small elements cause the fail to converge, but if I want to use the springs to model the soil (3 in/spring) so the pile has to be separated by several nodes, do you know how to solve such problem? Thank you.

Tracy

Hi tracy
your work is really needed to be verify by other software like ABAQUS.
I recommend you:
suppose your pile remain elastic behavior because of confining of soil so,Try to have one example in elastic Pile element with nonlinear link for behavior of soil and nonlinear element for column.
Then see your result.
Your work need trail and error solution to get which element can get the best result. maybe with elastic pile you can get well result!