**Francisco Pinto**
@
on

Problem with deconvolution

Hi,

I have been trying to deconvolve my seismic data, but it isn’t working properly. I have this soil profile:

Thickness 2.00 m // 6.00 m // 15.00 m // Half-Space Bedrock Vs 280 m/s // 390 m/s // 325 m/s // 1500 m/s UnitWeight 17 kPa // 20 kPa // 19 kPa // 22 kPa

I have tried different damping and G/G_max models (Darendeli, Vucetic, Ishibashi), but it doesn’t solve the problem. Input Motion: PAG=0,12 and PGV=41,35 (specify depth = 0)

When I compute to get the acceleration time series on the bedrock it returns an acceleration time series with values around 1E+19 g, which is very unrealistic. However, when I change the input motion to one with lower PAG and PGV, it works just fine.

Anyone knows how to solve this problem? Thanks.

0 Like 0 Dislike

Ellen M. Rathje@ onIn deconvolution, the surface motion Fourier Amplitude Spectrum (FAS) is DIVIDED by the transfer function (TF). This is the opposite approach used when propagating the motion to the surface, where the rock FAS is multiplied by the TF.

At large intensities, the TF at high frequencies will be close to zero because of soil nonlinearity (i.e., increased damping and reduced G at large strains). When you divide the zurface motion by this TF, which is close to zero, you can get massive motions at the base. This problem is not apparent at smaller intensities because the TF does not approach zero for lower intensities/smaller strains.

There really is no way to solve this problem…it clearly shows there is a problem with 1D site response analysis (EQL in this case, but I also believe it is a problem with nonlinear analysis) at large strains.

Reply Report abuse

Please login to answer the question.