Welcome to the forums. Please post in English or French.

You are not logged in.

#1 2011-10-03 11:30:02

TEOL11
Member
Registered: 2010-09-15
Posts: 45

generation of artificial accelerograms

Hello,

I wonder if there exists in code aster a command to synthesize an accelerogram (time history) from a pseudo-acceleration spectrum.

thank you for any info

Offline

#2 2011-10-04 05:46:45

boyere
Member
Registered: 2009-03-02
Posts: 115

Re: generation of artificial accelerograms

Hello,

no, there isn't any "operator" to produce accelerograms from an response spectrum, as this task is quite specific. It's not an easy job (non linear signal processing with many traps) which deserves its own tools.

But, as I understand your use of Code_Aster is for your own personal research, I can give you my own "quick & dirty" way to synthesize accelerograms, based on the following publication :
P. Cacciola, A stochastic approach for generating spectrum compatible fully nonstationary earthquakes. Computers & Structures 88, 889-901, 2010

The part of the method to go from the Pseudo-Acceleration Response Spectrum to the DSP  (inspired by the publication) is rigorous and sound although you must check carefully that the hypothesizes are true. You can find other methods which are more refined and precise but also far more complex too. The implementation was kindly given to me by a colleague and I've translated it into python, for which I'm not particularly gifted. The second part, to compute the accelerogram from the DSP, is maybe more questionable. So, if you can find any improvement, don't hesitate to tell me.

best greetings

Emmanuel

NB : I hope that the .comm file is self-explanatory enough.
NB2 : Repeating the GENE_FONC_ALEA command, you can obtain as much independent accelerograms as you which.


Attachments:
sro2dsp.comm, Size: 3.56 KiB, Downloads: 610

Offline

#3 2011-10-04 08:38:13

humbert
Member
From: Seoul
Registered: 2007-11-26
Posts: 81
Website

Re: generation of artificial accelerograms

TEOL11 wrote:

Hello,

I wonder if there exists in code aster a command to synthesize an accelerogram (time history) from a pseudo-acceleration spectrum.

thank you for any info

No, and the reason is there is an infinity of accelerograms corresponding to a single pseudo-acceleration spectrum.
Some methods exist though (one was proposed above), but none can ensure you that the generated accelerogram will cover "all" loading cases.
They can (at least most of them) only ensure you that the generated accelerogram's spectrum will match the target input spectrum.
Indeed as you (may) know the "shape" of the accelerogram has an effect on structures, and similar accelerograms may be more or less severe to a same structure.
If you need to cover "all" possible cases of loadings (I mean a significant part) then you need either a parametric study with a recommanded value of n >= 30 accelerograms (can't remember the exact reference here sorry), or use a probabilistic modelling of the input loading. In either case this is a long and complex process though.
To conclude, from the engineering point of view, these methods are not applicable (for evident reasons of time and complexity), and one often simply generate one or a few synthetic accelerograms based on a summation of hundreds of sinusoidal signals with pseudo-random frequency and amplitude. The accelerogram is then modified to match the target spectrum. This is not a precise definition but you can get the idea. Unfortunately I do not know the implementation details of such method.

PS: I'm sorry if this response is not quite accurate in a scientific/mathematical point of view (e.g. "all", "shape", ...) but this is more to get the general concept rather than to give a proof.

Offline

#4 2011-10-04 11:55:48

boyere
Member
Registered: 2009-03-02
Posts: 115

Re: generation of artificial accelerograms

Thanks, Jérôme, for the explanation concerning the non-unicity of the accelerograms.

I had this in mind as I said that the process to go from the Response Spectrum to the Seismic DSP is non-linear and also as I mentioned that you can use GENE_FONC_ALEA many times to get many instances of accelerograms corresponding to the same DSP.
In my humble opinion, it is due to the stochastic nature of earthquakes. And engineers don't get satisfied with just ONE accelerogram, even it is real and not synthetic.
You have to be sure that you've run your simulation through a collection of accelerograms corresponding to the scope of the earthquake you can expect (or fear).
And there are rules to be sure of that, even if they are not so simple as a mere number.
The advantage of a Response Spectrum is that the statistical nature of the earthquake is already taken into account in its definition.

All in all, with the .comm file I've written, you should have enough to get some preliminary results to evaluate the efficiency and precision of a method.
With a loop you can even get statistical (and more significant) results.
But, if you want authority-proof results, you'll have to get a little further and work a little more.

Emmanuel

Offline

#5 2011-10-04 15:03:11

paster
Member
Registered: 2009-08-14
Posts: 73

Re: generation of artificial accelerograms

Hello

Very interesting topic popular in the early 70's you can find information in or at
http://phm7833.perso.neuf.fr/ thanks to my friend Philippe
CEA which has developped long time ago a specific code called TIROIR
The litterature of preumont (specially in 80's)
The code pythagore of setec

Have a nice day

Offline

#6 2011-10-05 09:49:07

TEOL11
Member
Registered: 2010-09-15
Posts: 45

Re: generation of artificial accelerograms

Thank you all, for help and code, your posts have been enlighting

Offline

#7 2011-10-05 11:48:39

humbert
Member
From: Seoul
Registered: 2007-11-26
Posts: 81
Website

Re: generation of artificial accelerograms

boyere wrote:

The advantage of a Response Spectrum is that the statistical nature of the earthquake is already taken into account in its definition.

No, it is not. The spectrum itself, be it from a natural event, synthetic, and/or from design codes, does not account for future events (which are unpredictable -- otherwise Japan wouldn't be in such a distress), and may be (e.g. in design codes) an envelope.
I would say that in either case the response spectrum as well as the accelerogram are both random by nature, and numerous studies exist on the challenging topic of modeling earthquake signals, none of them being AFAIK acknowledge to be the "best" one.
Using either of those depends mainly of what type of simulation you want to achieve and what kind of result you are looking for about your structure.

An interesting reference, the PhD thesis of Guillaume Pousse (mainly in French, with a bit of English):
* Pousse G., Analyse de données accélérométriques de K-net et Kik-net: implications pour la prédiction du mouvement sismique- et la prise en compte des effets de site non-linéaire, Thèse, Université J. Fourier, Grenoble, 257 p., (2005).

Last edited by humbert (2011-10-05 11:51:54)

Offline

#8 2011-10-06 00:43:28

boyere
Member
Registered: 2009-03-02
Posts: 115

Re: generation of artificial accelerograms

Please, Jérôme, don't misquote me.
What we are talking about is, of course, a Design response spectrum.
And it's got a statistical nature. Read the first pages of its definition in Eurocode 8 and you will find such expressions as "Return Period" or "Probability of being exceeded" applied to it.
Now, if you find the definitions of the code too crude or not cautious enough, you can lobby the legislator to enforce a more mathematical engineering practice.
As I'm no expert in this field, I wont' say which logic is better and I won't argue further.

Concerning the accelerogram, it can be viewed as deterministic or as random, depending of its use.
For example, when I test an algorithm or a behavior law for a seismic application, I compare the numerical results with experimental measures on a scale model submitted to an accelerogram on a shaking table. The accelerogram is then deterministic and perfectly known. Incertitudes come from other sources and, to explain discrepancies between my simulation and the experiment I won't suspect in the first place the variability of the seismic input .
On the other hand, if I want to design a building to withstand to earthquakes and check it with time-history simulations, the accelerograms will be random variates of possible earthquakes. And I agree with you that one run won't be sufficient to cover all the field of the possible earthquakes. I let experts tell me how many I have to use and how to pick them.

As for the reasons why Japan suffered recently such a tremendous earthquake and why the consequences were so devastating, I think we've already largely trespassed the scope of this Code_Aster forum. But if you've got any clues, don't hesitate to mail me. I would be interested.

cordialement

Emmanuel

Offline

#9 2022-03-11 10:02:47

guido.tognan
Member
Registered: 2020-03-27
Posts: 3

Re: generation of artificial accelerograms

Hello,
there is a useful software tool called REXEL which you can download from the Reluis (Consorzio della Rete dei Laboratori Universitari di Ingegneria Sismica e Strutturale) site - just goggle it -, it allows to define the design spectra according to the Eurocode 8 or to a completely user-defined one and based on these spectra, it produces sets of Spectrum Compatible Accelerogram.

Guido

Offline

#10 2022-03-16 14:24:45

Viktor.Vlaski
Member
Registered: 2015-10-16
Posts: 20

Re: generation of artificial accelerograms

For generation of artificial acceleration time histories out of response spectra, the command GENE_ACCE_SEISME, which is described in document U4.36.04, is available.

The document is for example available at code-aster.org/doc/default/en/man_u/u4/u4.36.04.pdf

Offline