search

SPECFEM / specfem2d

SPECFEM2D simulates forward and adjoint seismic wave propagation in two-dimensional acoustic, (an)elastic, poroelastic or coupled acoustic-(an)elastic-poroelastic media, with Convolution PML absorbing conditions.
GNU General Public License v3.0
194 stars 146 forks source link

Acoustic inversion not converging #1035

Open pengguo-seismo opened 5 years ago

pengguo-seismo commented 5 years ago

Dear Developers,

I am trying to use specfem2d for acoustic inversion, p_sv=.true., and Vs=0, for updating Vp. I have tried both the pressure wavefield and potential for adjoint source, however, neither gave converging results.

I use the same model for elastic inversion, and SH mode inversion, both of them works. So it is not because of the starting model I used. Any suggestions?

bottero commented 5 years ago

Hi "Meatball",

Can you send me you example folder with you Par_file, script and everything so that I have a look?

Thank you,

Have a good day,

Alexis Bottero Marine Physical Laboratory Scripps Institution of Oceanography (+1) 619 677 9887 (+33) 695 170 097

Le mar. 26 mars 2019 à 07:32, meatball880207 notifications@github.com a écrit :

Dear Developers,

I am trying to use specfem2d for acoustic inversion, p_sv=.true., and Vs=0, for updating Vp. I have tried both the pressure wavefield and potential for adjoint source, however, neither gave converging results.

I use the same model for elastic inversion, and SH mode inversion, both of them works. So it is not because of the starting model I used. Any suggestions?

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/geodynamics/specfem2d/issues/1035, or mute the thread https://github.com/notifications/unsubscribe-auth/AGuO4mHY-eBguHSr9H-lO6b1qFKYfY0aks5vai-PgaJpZM4cLmSK .

rdno commented 4 years ago

Hi,

I have also encountered similar problem. I have noticed that adjoint sources for Z component does not contribute to the kernel. If you only give non-zero adjoint source to Z component, acoustic kernel comes as zero. I have setup a repo which replicates this for a simple case. https://github.com/rdno/specfem2d_kernel_test

Periodic boundaries are also change the sign of the kernel. If you want, I can provide an example.

Also, Hessian2 kernels come up as zero for acoustic simulation.

These are some of the problems I could identify.