Number of records

02-Analytical capabilities
KGK
Posts: 4
Joined: 28 Apr 2016, 16:12

Number of records

Post by KGK »

Hello, congratulations for your software!

I have a question regarding the number of artificial recordings; which would be the appropriate procedure in order to extract more than 8 artificial records using SeismoArtif and for example the Gasparini & Vanmarcke approach (for a given target spectrum and envelope function)?

Thank you in advance.
Regards,
KGK
huffte
Posts: 978
Joined: 22 Jul 2011, 10:19
Location: Cookeville, Tennessee, USA
Contact:

Re: Number of records

Post by huffte »

There should be a number of options I would think. One might be to run twice, slightly tweaking parameters such as times on the envelope spectrum, magnitude and distance on the generation, etc.
Tim Huff
KGK
Posts: 4
Joined: 28 Apr 2016, 16:12

Re: Number of records

Post by KGK »

Thanks for the prompt response huffte; I am a bit concerned, however, that by slightly changing parameters as the rise and fall times or the duration of the envelope function (or magnitude and distance in case a different approach is adopted), then slightly different records will be extracted too (with not identical, but quite similar spectra) introducing bias in structural analysis results.

Regards,
KGK
huffte
Posts: 978
Joined: 22 Jul 2011, 10:19
Location: Cookeville, Tennessee, USA
Contact:

Re: Number of records

Post by huffte »

Hi KGK. I am not certain what you mean by " slightly different records will be extracted too (with not identical, but quite similar spectra)". It seems that the point is to obtain different records. And, in reality, even the 8 initial records with all parameters identical will indeed have different spectra. Two real records recorded at similar sites the same distance from the same fault source will be different. So I'm not certain of the introduction of bias into the results. Perhaps you could elaborate a bit? Best of luck with your work KGK.
Tim Huff
KGK
Posts: 4
Joined: 28 Apr 2016, 16:12

Re: Number of records

Post by KGK »

Thanks huffte,
Just to clarify things; the point is of course to retrieve different and statistically independent artificial time histories.
I believe that this will not be achieved if you slightly change the envelope function parameters or the magnitude-distance input. For example, if you initially extract 8 histories with M=7 and D=20km and and then 4 additional with M=7 and D=21km, the last 4 will be quite similar with 4 records from your initial bin, hence your structural analysis results (based let's say on the mean response from 12 histories) will be biased. Recall here that my initial question was about the appropriate procedure to extract more than 8 records for a given target spectrum and a given envelope function.

I suspect that the answer to my question is related to the ability to generate a string of different phase angles (Gasparini & Vanmarcke approach) in each run. Is this controlled somehow by the parameter "Random Process Starter" in settings? I wasn't able to find many details about this parameter in the manual.
Regards,
KGK
huffte
Posts: 978
Joined: 22 Jul 2011, 10:19
Location: Cookeville, Tennessee, USA
Contact:

Re: Number of records

Post by huffte »

That does indeed clarify things, KGK. I see your point. If you are bound to use the frequency domain spectral matching, then perhaps one of the other options (e.g., adjusting real accelerograms to a target spectrum) might be a means of providing the supplemental synthetic records? Or complementing with some time-domain matched records using SeismoMatch? Or using 2 or 3 completely different envelope functions? Or, certainly, you could ask the good folks at SeismoSoft to simply increase the max number permitted from 8 to something more? Might be worth a shot.

My original line of thinking was that if you had a characteristic magnitude and distance for your site, then given the uncertainty inherent in each, you could quickly obtain many artificial records using the specific barrier model in SeismoArtif. For example, say your characteristic, or modal, (M,R) pair is (7.0, 20 km). A common M-range is +/- 0.2. So you could generate 8 records using (6.8, 20 km), 8 records using (7.0, 20 km), and 8 records using (7.2, 20 km). You could further specify a range of source-to-site distances and quickly come up with an enormous number of records.

Best of luck, KGK.
Tim Huff
User avatar
seismosoft
Posts: 1184
Joined: 06 Jul 2007, 04:55

Re: Number of records

Post by seismosoft »

Hi KGK,

In the project settings and in the Accelerogram Generation tab, there is a setting called Random Process Starter that can take any integer value.
If you change this setting, the new set of generated accelerograms will be different from the initial one.

Regards,
Seismosoft Support
KGK
Posts: 4
Joined: 28 Apr 2016, 16:12

Re: Number of records

Post by KGK »

Thank you for your straightforward answer and solution.
May I also kindly suggest to provide this info (about the "Random Process Starter") in your manual too? Users may find useful this capability.
Thanks again!
Regards,
KGK
User avatar
seismosoft
Posts: 1184
Joined: 06 Jul 2007, 04:55

Re: Number of records

Post by seismosoft »

Thank you for your suggestion KGK. In fact we are thinking of removing the Starter completely in future releases, making the process entirely random.

Best Regards,
Seismosoft Support
rojand
Posts: 1
Joined: 02 Nov 2017, 09:42
Contact:

Re: Number of records

Post by rojand »

Dear huffte,
Thank you for these useful information( both the appropriate forum and modelling imperfections)
Post Reply

Return to “02-Analytical capabilities”