Dear Marc, and all.
Hello.
I'm using dinver for Rayleigh wave dispersion inversion.
My target sites seem to have very fast Vs half space layer >~3000 m/s at ~500m.
For the first time, I set the velocity range of the half space to have 50-4000 m/s, as the other layers for a broad search.
But no result model has a layer that exceeds ~3100m/s (approximately, it could be more or less).
So I tested the velocity range of the half space to be like 3100 - 5000 m/s.
Then it generates the following error message,
"ASSERT: "parameterSpace.isOkDebug()" in file src/Generator.cpp, line 132"
I guess it is because the parameter space of velocity is limited up to ~3100 m/s.
I'm using Geopsy version 3.4.2
Could you give me some comments on this issue????
Always thanks for your efforts.
Best regards,
Youngjun Jeon
Dinver parameter space (maximum velocity) issue
Dinver parameter space (maximum velocity) issue
- Attachments
-
- Sample.dinver
- Dispersion curve and Parameter set
- (2.77 KiB) Downloaded 564 times
Re: Dinver parameter space (maximum velocity) issue
Hi Youngjun,
I checked with 3.5.0-preview and I did not encounter this ASSERT. It happens with gradient layers due to an unsolved bug.
Why are you using so many layers? With 3 layers in the soft part and 2 below 15 m, I got a reasonable fit of the high frequency part. In any case, the low frequency part looks difficult to fit. A contrast around 25 m is inverted. Your dispersion curve has a big gap between 2.53 and 7.88 Hz. In general, it is safer to have a continuous sampling to avoid mode jumping during forward computation. It does not occur here, I test a resampled version of your dispersion curve and the original, I got approximately the same results. See figures.
I checked with 3.5.0-preview and I did not encounter this ASSERT. It happens with gradient layers due to an unsolved bug.
Why are you using so many layers? With 3 layers in the soft part and 2 below 15 m, I got a reasonable fit of the high frequency part. In any case, the low frequency part looks difficult to fit. A contrast around 25 m is inverted. Your dispersion curve has a big gap between 2.53 and 7.88 Hz. In general, it is safer to have a continuous sampling to avoid mode jumping during forward computation. It does not occur here, I test a resampled version of your dispersion curve and the original, I got approximately the same results. See figures.
- Attachments
-
- gap.png
- (143.82 KiB) Not downloaded yet
-
- continuous.png
- (146.19 KiB) Not downloaded yet
Re: Dinver parameter space (maximum velocity) issue
Dear Mark,
Thanks for your fast comment!
In terms of the number of layer, I'm using multiple layers larger like ~15 layers then it generally accepted for the first trial inversion process.
Then I reduce the layer number based on the visually constrained layers.
Best,
Youngjun Jeon
Thanks for your fast comment!
In terms of the number of layer, I'm using multiple layers larger like ~15 layers then it generally accepted for the first trial inversion process.
Then I reduce the layer number based on the visually constrained layers.
Best,
Youngjun Jeon
Re: Dinver parameter space (maximum velocity) issue
Usually, the approach is reversed: start with few and increase until getting an acceptable fit, and thirdly start an appraisal phase (depth limits, non-uniqueness,...)