Non Numerical Input Error

03-Unexpected behaviour/errors
Post Reply
Kirkaguel
Posts: 40
Joined: 31 Dec 2018, 10:04

Non Numerical Input Error

Post by Kirkaguel »

Hi Technical Support,

I was doing a trial spectral matching where I have my target MCE design spectrum as a CSV file I was just confused why it reads that line 3 is not a numerical input and it gives an error. Line 3 is just a period-acceleration pair of 0, 0.64005 respectively?

Here is the CSV file https://www.dropbox.com/s/70gckihavys12 ... 0.csv?dl=0
huffte
Posts: 918
Joined: 22 Jul 2011, 10:19
Location: Cookeville, Tennessee, USA
Contact:

Re: Non Numerical Input Error

Post by huffte »

The problem is actually the commas, I believe. I open the file in Excel, then copy and paste all the data (including headers) to a text file in notepad. The commas are gone from the resulting text file and the file loads fine into SeismoMatch. Of course, you could manually replace all commas with a space, but the procedure outlined above is very quick for a csv file.
Tim Huff
Kirkaguel
Posts: 40
Joined: 31 Dec 2018, 10:04

Re: Non Numerical Input Error

Post by Kirkaguel »

Thanks Hufftie I managed to make it work as per your suggestion I think seismosoft does not support CSV file. It would be good if they can also add this feature in future release.

After running the spectral matching I get a reasonably nice fit for the acceleration and velocity graphs except for the displacement that clearly shows drift. I am using the Abrahamson algorithm which i thought as per the technical reference does not need any baseline correction. Any suggestion which parameter to tweak to avoid the drift on the displacement graph?

https://www.dropbox.com/s/471zzhgny7uhv ... t.jpg?dl=0
huffte
Posts: 918
Joined: 22 Jul 2011, 10:19
Location: Cookeville, Tennessee, USA
Contact:

Re: Non Numerical Input Error

Post by huffte »

I have had similar results from time to time. Sometimes, you can save the matched accelerogram, load it into SeismoSignal, and baseline correct (without filtering) to get rid of the displacement drift. If that doesn't work, the you could add filtering to the baseline correction in SeismoSignal, making certain you preserve all frequencies of interest (probably by using a bandpass filter with frequency limits as wide as possible, but narrow enough to successfully eliminate the drift). In such a case, you probably should compare spectra from the originally matched record with the corrected/filtered version to ensure you still have a good match to the target.
Tim Huff
Kirkaguel
Posts: 40
Joined: 31 Dec 2018, 10:04

Re: Non Numerical Input Error

Post by Kirkaguel »

Thanks Tim will give it a try
Post Reply

Return to “03-Unexpected behaviour/errors”