Seismosoft | Earthquake Engineering Solutions

Hello everybody!

I am running a time history analysis with a set of accelerograms to assign as input to the nodes at the base of the structure. I need to assign 9000 steps, but each load curve permit me to assign only 5000. I have tried to assign two different load curves, one of them having a delay. But the software tells me that a load curve already exists for the nodes. So basically it means that I am not able to study the behaviour of the structure during all the seismic event time, if it consists of more than 5000 steps.
Of course I thought of different ways to solve the situation. Two main ones are:
1 - take final displacements of the analysis carried out with the first 5000 steps and assign them together with the accelerograms for a second analysis. Of course I lose the history of many important information and it implies many efforts.

2 - Create two nodes in the vertical direction that are very close (1mm) and link them rigidly. So I can assign the two load histories (the second is delayed) in order to have a single analysis. Of course I don't have to care of the output influence of such rigid link.

Now I would like to ask if there exists another method I am not aware of, and why a limit of 5000 load steps has been assigned.
( Being a time step of about 0.001 s, very common in many research applications, an accelerogram of only 5 seconds can be applied).

Thank you for your attention

Leandro,

If you carry out a search in this Forum for "maximum number data points" (or simply "maximum points"), you will find some guidance or comments on this matter.

We would like to note, however, that even if you will manage to run analyses with records featuring a large number of data points, we would actually argue that a time-step of 0.001 seconds does not really seem very common to us, but rather excessively small. Are you sure you need such a small time-step, which will render your analyses very long? If you use the much more common time-step of 0.01 seconds, which will speed up significantly your analyses, will your structural response really change that much? We would expect not to.

Seismosoft Support

I'm not sure what's happening with your model Leandro. I have run many time history analyses in SeismoStruct with up to about 32000 steps and a step size as small as 0.004, but never as small as 0.001.

A rough rule of thumb is to use a time step size of 0.1 times the smallest natural period of interest. The rule of thumb would thus indicate a step size of 0.01 seconds to accurately capture the response at a period as small as 0.1 seocnds.

Of course, the rule of thumb should take into consideration the step size of the input accelerogram too. Don't use a step size for analysis larger than the step size of the input accelerogram.

Oe way to study the response of a structure to different loads in teh same model is to have a duplicate, unconnected structure in the same model.

I do really thank you all for your replies.
Unfortunately the step time is assigned, and I cannot modify it. Anyway I have run many analyses for research purposes and I do not find strange such a step time. Basically, I have a simple model and not a big structure. No matter if it takes all night long to run an analysis.

Anyway, it seems I need to assign two coincident nodes with rigid link so that I can use up to 10000 steps.
If next release will permit to increase maximum number of data points, that would be really welcomed.

Thank you again!

My apologies for not quite understanding your situation Leandro, but I'm still having difficulty.

As I stated earlier, 9000 steps is no problem for SeismoStruct. I have used up to 32000 steps.

Pardon me if I am wrong, but it almost sounds as if you simply need to change the default "last line" from 5001 to 9001 when you "Load" the curve from file. Or make certain that there is not a blank or non-numeric line after line 5000 in your curve file. Or you can also "Create" instead of "Load" and simply paste in the time versus acceleration data from Excel.

Does your load curve have time and acceleration on each line, one pair per line?

I hope you can find a satisfactory solution Leandro, because, as I say, 9000 steps in a loading curve is no problem.

You should also check your loading stages specification and make certain that - for the example you give - the number of steps is set to 9000 and the end of the stage is 9 seconds. If you specify something beyond the end of the loading curve you will get problems.

You are totally right! I do not believe I could not see that I just needed to set up the number of input lines!
Sorry for making you waste your time and thank you so much!