nonlinear time history analysis does not converge with viscous damper link

[crimaval]

I am trying to model a steel RC frame of 5 floors and 3 bays with a viscous damper (link element) using seismosoft 2021. However, I had problems with the convergence of the model under nonlinear time history analysis. If someone could help me I can share the file to verify what is the problem.

I model the viscous damper with the link element in "element classes". then I connect the damper link to steel bars that connect to the upper and lower floors. However, after 3 sec of simulation, the software stop and convergence is not achieved.

I have to say that if I delete the viscous damper link and the steel bars, then convergence is achieved. I think some numerical difficulty arises in the way I model the viscous damper.

Please Help, I can share the file with the problem if you can help me.

[seismosoft]

There are different reasons why the analysis diverges. Have you carried a search on this forum on 'convergence problems'? There are numerous posts with useful information. Besides these you could try the following:
1) There is a special section on the User's Manual on how to fix convergence problems
2) Check the Convergence Details page of the post processor to see the area or elements that cause the problems
3) Make sure that the damping parameters (mass and/or stiffness parameters) are correctly calculated
4) Run eigenvalue analysis and check the modes to see if the model's members are well connected to each other
Seismosoft Support

[crimaval]

I get the solution to this problem, changing the constitutive model of the concrete from chang-mander to mander-model. It seems that the chang-mander model is complex to be handle by the soft.

Now I got another problem. In the 2018 version of seismosoft structure, I model a Steel RC frame with viscous damper using the lump damper model of the software in that version. Now I am trying to reproduce the same model but using a link damper, however, I got a totally different result with link damper. I am working with the same damper constant 637 ton/sec. Any idea about the problem?.

[seismosoft]

Make sure that the hysteretic behaviour of the two dampers is similar. Since you have not change anything else in the model, focus of the local response of the damper.
Seismosoft Support

[crimaval]

I solved this problem too, Now my problem is that I get the peak top displacement from the RC frame (5 floors, 3 bays) at the 5 second, then the simulation continues, however at time 13 seconds the analysis stop due to "convergence problems". The problem is that I run the simulation for 20 seconds with input earthquake time of 50 seconds.

I had to increase the step time from 0,01 sec to 0,04 sec in order to get convergence at longer times, but still, I did not reach the 15 seconds analysis that I require. I also try reducing the step time to 0,005 sec but it is worse. The usual message is "Elm_Inv". I already increase the tolerance at the "convergence criteria", also increase tolerance at "element iterative strategy". I deactivate "do not allow element unbalanced force in the case of Elm_Inv". I also use only three "integration section" per element and every element is model using the InfrmFB (except the damper that is model as a link, however even in the model without damper I have problems).

I did not get a problem when I worked with steel as reinforcement with the "mander model", but the convergence problem appears when I use the GFRP bar as reinforcement with the "trilinear model ("frp_tl")". I work with a GFRP Elastic modulus between 45-60 GPa and tensile strength between 800-1200 MPa. I checked the model with eigenvalue analysis and the model is ok. help, please.

[crimaval]

I solved this problem too, Now my problem is that I get the peak top displacement from the RC frame (5 floors, 3 bays) at the 5 second, then the simulation continues, however at time 13 seconds the analysis stop due to "convergence problems". The problem is that I run the simulation for 20 seconds with input earthquake time of 50 seconds.

I had to increase the step time from 0,01 sec to 0,04 sec in order to get convergence at longer times, but still, I did not reach the 15 seconds analysis that I require. I also try reducing the step time to 0,005 sec but it is worse. The usual message is "Elm_Inv". I already increase the tolerance at the "convergence criteria", also increase tolerance at "element iterative strategy". I deactivate "do not allow element unbalanced force in the case of Elm_Inv". I also use only three "integration section" per element and every element is model using the InfrmFB (except the damper that is model as a link, however even in the model without damper I have problems).

I did not get a problem when I worked with steel as reinforcement with the "mander model", but the convergence problem appears when I use the GFRP bar as reinforcement with the "trilinear model ("frp_tl")". I work with a GFRP Elastic modulus between 45-60 GPa and tensile strength between 800-1200 MPa. I checked the model with eigenvalue analysis and the model is ok. help, please.

[seismosoft]

The fact that you get the max. repsonse at 5sec and you have large drifts and convergence problems at t=13sec might mean that you need to input viscous damping in your model.
Apart from this, there are numerous posts in this forum on how to solve convergence problems. Note also the corresponding section on how to solve convergence problems in the Users Manual.
Seismosoft Support