Why are additional nodes resulting in failure + convergence

[asmith]

Hi,

I am modelling a two-storey RC frame to undertake some dynamic analysis. My standard frame runs perfectly for my dynamic loading. It is a simple one bay structure.

To develop my analysis I am attempting to strengthen the structure using confinement of the concrete. This occurs at 1/4 length of the columns at top / bottom of columns. To achieve this, I am adding additional nodes.

If the original structure was (for one vertical of the structure, where 0 = node, ---- = column.
0-------------------------0-------------------------0

this becomes
0----0-------------0-----0----0--------------0-----0

So in the smaller areas i can adjust the material properties to reflect the confinement.
When I do this step, i start getting convergence errors and some errors stating the materials are failing in compression before i get the convergence error.

I thought I would obtain continuity of the element as everything is the same except for the strength / strain / modulus. Is there a step i need to do to prevent this? I believe the non-linear analysis is the cause but i'm unsure how to fix it.

Any help would be great.

Thanks

[ismailqeshta]

Are you using force beam-column elements?

I have observed significant differences in the post-peak behaviour using different nodes.

[asmith]

Thanks for your reply.

Yes, I am using the nonlinearBeamColumn element. From some testing, I noticed that using a column with one node between beams results in no issues. When i add an additional node (IE the column is now split into 3 elements), i get errors with the Plain and RCM numberer no longer working. To try and work around this I have used a force beam column element with user defined integration points and changed defined the section properties at each integration point as opposed to breaking the column into three. It seems to have worked OK but I still feel this approach is sub-optimal.

Do you know why adding the additional node may result in RCM/Plain numberers no longer giving any results from dynamic analysis?

Thanks for your help

[ismailqeshta]

I am using Plain numberer and got differences in results (post-peak) when I used more than one element.

In my case, the differences are actually in static analysis. I did not check the dynamic analysis.

I am not actually sure about the source of this problem, but if you really need to model your structure using more than one element, you may try to use displacement beam-column element, but you will need to provide many elements to get close results to the one element model using force beam-column.

[EricsonEncinaZ]

You may be having localisation issues: Coleman&Spacone 2001 "Localization Issues in Force-Based Frames Elements".

Despite you use FBE you have to ensure you have enough IPs to actually capture an objective response of the element non-linearity. Remember that this is a distributed plasticity element, so the number of IPs have an impact on the response. You can see this by doing an analysis with 2IP and then run the same analysis but with 6 or 9 IPs, for example.

You must ensure that the number of IPs are enough to capture the NL behaviour of the element, and if you need more IPs (max is 10 according to the source code) then you need to use more than 1 element to add more IPs.

I have tried DBEs as well and, as it is known, you need way more elements to get an objective response, but they do have less convergence problems... at least as I can say based on the models I have run. Be aware that you can also have localisation issues in DBEs, but instead of localise the response in 1 IP the localisation will occur over an entire element.

[asmith]

Thank you all for your replies.

I am almost at a loss with how to progress with this. I have tried everything I can think of. I have managed to get certain numberers to work, but when I get to higher strength earthquake the dynamic analysis simply doesn't continue. It doesn't give any convergence errors or any indication it is initiating my iterative solver.

Has anyone done something similar with splitting an element into three and knows how I may solve this?

Many thanks

[ismailqeshta]

Yes, I am currently facing a similar issue with convergence for higher intensities. I just posted a question on it recently:

http://opensees.berkeley.edu/community/ ... =2&t=65708

[asmith]

Just an update that may help people who can solve this: I have run the dynamic analysis for a single column, then two columns (no beam between them) and its runs fine. As soon as I add a beam between the two I get issues with the analysis not wanting to run.

[ismailqeshta]

You may need to check the rigidity of the beam between the two columns. It may be hard to obtain convergence if it has large rigidity. I have faced a similar issue in one of my models.