Seismic Analysis Method selection 3

[amini06]

I am doing a model of a dam(concrete dam) subject to seismic loads, i.e. El Centro earthquake accelerogram record & want to select a method in ANSYS.
Which method must be selected?
1- Transient Dynamic Analysis-full method or
2- Spectrum Analysis - SPRS?
How can I input time steps (say 0.02 sec) Vs earthquake displacement?
* Results must include the history of stress during earthquake.

 

[Drej]

You haven't stated the type of seismic data you have. This will determine which method you use. However, it sounds as though you have time-history data (displacement vs time?), which would mean you should use the transient approach. This method is VERY expensive in terms of time. You can convert this data into a response spectrum (see the RESP command in /post26) which will allow you to carry out a response spectrum analysis, which is frequency domain based and is consequently waaay less expensive. It also sounds as though you're new to this field, so I would recommend you read some texts on the different approaches before you do any analysis work, as this is a very complex area and takes some time to understand all of the fundamental principles involved. The ANSYS help file is a good place to start.


------------
See faq569-1083 for details on how to make best use of Eng-Tips.com

 

[amini06]

Dear Drej
Thanks for your attention.
Considering your comments, let we have a response spectrum & do a response spectrum analysis, but here is a problem. Response spectrum as you know deals with only the max values, not history of data. For instance, for a seismic loads i.e. El Centro earthquake accelerogram & known damping ratio(say 5%),we have a deformation response spectrum vs natural frequencies. In this curve I didn’t see the history of ground shaking!
So If we apply it to the structure, how we can have a history of results (say stress or displacements)?

Regards

 

[Drej]

Correct, in response spectrum analyses the solution is carried out in the frequency domain, which means there is no (time) history of the seismic event. What you get is a single result, which is simply the combination of all of your modal responses due to the response spectrum. The assumption is that all of your structural modes combine in a specific way. The load on your structure is calculated based on the mass participation of each mode and its direction (the response spectrum is used as a look up table: the modal frequency is found and the corresponding acceleration is applied to the structure. This happens for all modes and these are then combined to give you the final "single" result.). To get the time history you need to carry out the analysis in the time domain i.e. by carrying out a transient analysis.


------------
See faq569-1083 for details on how to make best use of Eng-Tips.com

 

[amini06]

Dear Drej
Thanks again.
About Transient Analysis:
Once you replied to a question about Tran. Anal. with the title of "EARTHQUAKE TIME HISTORY" from "civil33 (Structural)" in this forum.
His main question was about the accuracy of his method to modeling earthquake force.

You did emphasis that in "Tran. Anal." ANSYS, didn’t accept accelerogram record (ground acceleration Vs time) directly & it must be convert to displacement Vs time. But if you remember,"civil33" use accelerogram directly by this method:
TIME,T1
ACEL,g1
TIME,T2
ACEL, g2,etc.Considering to the above comments:
Do you confirm this method?
Lets we have a simple structure, say a column. Gravity force, as you know is a "body force" and acting to the whole of the structure. But in a seismic shaking, we have these steps:
1-Ground starts to shake.
2-The structure foundation starts to shake.(The base of the structure suffers some displacements vs time)
3-At last the structure, considering its damping properties, starts to shake (displacements) and it causes a new stress distribution vs time.
In these steps, the ground shaking transfer from the foundation nodes to the structure base nodes & at the next steps transfer from the structure base nodes to the upper structure nodes.


When we apply "g" as a variable vs time (ACEL,g1 TIME,T1 ,...) are we sure to consider the above steps?

How we can model the seismic load "transferring steps"?

Regards

 

[cbrn]

Hi,
I think Drej will also confirm this thought we seem to share: I believe that applying time-variant ACEL is not correct.
This would not capture correctly the structure's response because in reality a time-variant displacement (which, twice-derivated, gives the time-variant acceleration) is applied to the foundation nodes, not uniformly over the structure.
For ex., if you deal with an extremely "soft" structure, the nodes "far" from the foundations would not be affected at all by what happens at the foundations, whether if you use the ACEL method you have them be excited anyway.
Transfering an acceleration spectrum to a displacement time-history is not easy: it involves Inverse Fourier Transform, but the problem is how to handle the information about phase (the earthquake's accel spectra that I know don't have info about phase)...
I also had this problem some time ago, Drej gave some very valuable advices, sure you can find the post by searching in this thread...

Regards

 

[amini06]

Thanks cbrn.
I believe that applying time-variant ACEL is not correct, too.
Also, I read your post & comments titled as, “earthquake simulation with non-linearities and displacement spectrum?”
We have some commands in Spectrum analysis such as (sed,svtype,...) to apply Base nodes DISPLACEMENT vs Frequencies. If I want to use TRANSIENT analysis is there any way to apply DISPLACEMENTS vs TIME (modeling a earthquake shakes on the structure base joints)?
You confirm me if I need a time history results, say stresses changes vs time due to a earthquake, it is unavoidable to apply base nodes displacements vs time by caring out a transient analysis, don’t you?

Regards

 

[Drej]

cbrn - yes, I agree with you, that method is incorrect.
amini06 - search through the archives here for information on time-history seismic analysis using displacements. There is lots and lots of information on this already.


------------
See faq569-1083 for details on how to make best use of Eng-Tips.com