Error detected for latest release, in the case of fluid-solid interface and model=binary

[pengguo-seismo]

Dear Developers,

I ran into error messages, for a simple two-layer model test, where the 1st layer is acoustic water, and second layer is elastic. The error message is Error detected, aborting MPI... proc 0 Error, program ended in exit_MPI

It only happens when I use model=binary. If I use model=default, the simulation runs without error messages.

Can you help me solve this?

Regards

[komatits]

Hi,

Let me cc Etienne Bachmann at Princeton, because I think model=binary is an option he added last year (but I am not 100% sure; if not, I will check the Git logs to see who developed it and could thus maybe help you).

Best regards, Dimitri.

On 12/18/18 5:37 AM, meatball880207 wrote:

Dear Developers,

I ran into error messages, for a simple two-layer model test, where the 1st layer is acoustic water, and second layer is elastic. The error message is Error detected, aborting MPI... proc 0 Error, program ended in exit_MPI

It only happens when I use model=binary. If I use model=default, the simulation runs without error messages.

Can you help me solve this?

Regards

— 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/1010, or mute the thread https://github.com/notifications/unsubscribe-auth/AFjDKYFxayWmAU1AI1MnlRFiorjS3Kvgks5u6HEXgaJpZM4ZXle3.

-- Dimitri Komatitsch, CNRS Research Director (DR CNRS) Laboratory of Mechanics and Acoustics, Marseille, France http://komatitsch.free.fr

[komatits]

Hi,

Do you have Stacey boundary conditions on for the simulation? We have a bug in the case of external models with Stacey + save forward, because we create files, acoustic or elastic, on the left right top bottom, before reading the external model.

Best regards, Etienne

Télécharger Outlook pour Androidhttps://aka.ms/ghei36

De : Dimitri Komatitsch Envoyé : mardi 18 décembre à 06:38 Objet : Re: [geodynamics/specfem2d] Error detected, in the case of fluid-solid interface and model=binary (#1010) À : geodynamics/specfem2d, meatball880207, geodynamics/specfem2d Cc : Subscribed, Etienne Bachmann, Alexis Bottero, Paul Cristini

Hi, Let me cc Etienne Bachmann at Princeton, because I think model=binary is an option he added last year (but I am not 100% sure; if not, I will check the Git logs to see who developed it and could thus maybe help you). Best regards, Dimitri. On 12/18/18 5:37 AM, meatball880207 wrote: > Dear Developers, > > I ran into error messages, for a simple two-layer model test, where the > 1st layer is acoustic water, and second layer is elastic. The error > message is > Error detected, aborting MPI... proc 0 > Error, program ended in exit_MPI > > It only happens when I use model=binary. If I use model=default, the > simulation runs without error messages. > > Can you help me solve this? > > Regards > > — > You are receiving this because you are subscribed to this thread. > Reply to this email directly, view it on GitHub > , or mute the > thread > . > -- Dimitri Komatitsch, CNRS Research Director (DR CNRS) Laboratory of Mechanics and Acoustics, Marseille, France http://komatitsch.free.fr

[pengguo-seismo]

Hi, Do you have Stacey boundary conditions on for the simulation? We have a bug in the case of external models with Stacey + save forward, because we create files, acoustic or elastic, on the left right top bottom, before reading the external model. Best regards, Etienne Télécharger Outlook pour Androidhttps://aka.ms/ghei36 De : Dimitri Komatitsch Envoyé : mardi 18 décembre à 06:38 Objet : Re: [geodynamics/specfem2d] Error detected, in the case of fluid-solid interface and model=binary (#1010) À : geodynamics/specfem2d, meatball880207, geodynamics/specfem2d Cc : Subscribed, Etienne Bachmann, Alexis Bottero, Paul Cristini Hi, Let me cc Etienne Bachmann at Princeton, because I think model=binary is an option he added last year (but I am not 100% sure; if not, I will check the Git logs to see who developed it and could thus maybe help you). Best regards, Dimitri. On 12/18/18 5:37 AM, meatball880207 wrote: > Dear Developers, > > I ran into error messages, for a simple two-layer model test, where the > 1st layer is acoustic water, and second layer is elastic. The error > message is > Error detected, aborting MPI... proc 0 > Error, program ended in exit_MPI > > It only happens when I use model=binary. If I use model=default, the > simulation runs without error messages. > > Can you help me solve this? > > Regards > > — > You are receiving this because you are subscribed to this thread. > Reply to this email directly, view it on GitHub > , or mute the > thread > . > -- Dimitri Komatitsch, CNRS Research Director (DR CNRS) Laboratory of Mechanics and Acoustics, Marseille, France http://komatitsch.free.fr

Hi Dimitri,

Thank you for the reply. I use the PML boundary condition.

I think I may figure out the reason: from the latest version of specfem2d, it didn't really give me useful error message rather than endless "Error, program ended in exit_MPI". So I tested it in a previous version, and I got the following message, "external velocity model cannot be both fluid and solid inside the same spectral element". So the error probably comes from the mesh I used at the fluid-solid interface.

I use the mesh generator embedded in specfem2d for meshing. Any comments how to address the fluid-solid interface using outer model?

Best Regards Peng

[qcliu0]

Hi Peng, Hi Dimitri,

I have some experiences when working with Daniel on this issue.

Peng, perhaps you can try to re-assign acoustic parameters to the GLL points at and above the fluid-solid interface.

Say, we have two cells within two media, respectively, as follows

0 0 0 0 0 0 0 0 0 <- fluid interface

and

1 1 1 <- solid interface 1 1 1 1 1 1

the two interfaces are coupled numerically but with different parameters.

Attached please find my figures. Hopefully my solution would work.

Best regards, Qiancheng

[image: image0007000.jpg] [image: image0015000.jpg]

On Wed, Dec 19, 2018 at 11:20 AM meatball880207 notifications@github.com wrote:

Hi, Do you have Stacey boundary conditions on for the simulation? We have a bug in the case of external models with Stacey + save forward, because we create files, acoustic or elastic, on the left right top bottom, before reading the external model. Best regards, Etienne Télécharger Outlook pour Androidhttps://aka.ms/ghei36 De : Dimitri Komatitsch Envoyé : mardi 18 décembre à 06:38 Objet : Re: [geodynamics/specfem2d] Error detected, in the case of fluid-solid interface and model=binary (#1010 https://github.com/geodynamics/specfem2d/issues/1010) À : geodynamics/specfem2d, meatball880207, geodynamics/specfem2d Cc : Subscribed, Etienne Bachmann, Alexis Bottero, Paul Cristini Hi, Let me cc Etienne Bachmann at Princeton, because I think model=binary is an option he added last year (but I am not 100% sure; if not, I will check the Git logs to see who developed it and could thus maybe help you). Best regards, Dimitri. On 12/18/18 5:37 AM, meatball880207 wrote: > Dear Developers, > > I ran into error messages, for a simple two-layer model test, where the > 1st layer is acoustic water, and second layer is elastic. The error > message is > Error detected, aborting MPI... proc 0 > Error, program ended in exit_MPI > > It only happens when I use model=binary. If I use model=default, the > simulation runs without error messages. > > Can you help me solve this? > > Regards > > — > You are receiving this because you are subscribed to this thread. > Reply to this email directly, view it on GitHub > , or mute the > thread > . > -- Dimitri Komatitsch, CNRS Research Director (DR CNRS) Laboratory of Mechanics and Acoustics, Marseille, France http://komatitsch.free.fr

Hi Dimitri,

Thank you for the reply. I use the PML boundary condition.

I think I may figure out the reason: from the latest version of specfem2d, it didn't really give me useful error message rather than endless "Error, program ended in exit_MPI". So I tested it in a previous version, and I got the following message, "external velocity model cannot be both fluid and solid inside the same spectral element". So the error probably comes from the mesh I used at the fluid-solid interface.

I use the mesh generator embedded in specfem2d for meshing. Any comments how to address the fluid-solid interface using outer model?

Best Regards Peng

— 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/1010#issuecomment-448508117, or mute the thread https://github.com/notifications/unsubscribe-auth/AU_WPlRrFQPZ5F3kLI-XMfjAMQ9CZR9mks5u6fbYgaJpZM4ZXle3 .

--

This message and its contents, including attachments are intended solely for the original recipient. If you are not the intended recipient or have received this message in error, please notify me immediately and delete this message from your computer system. Any unauthorized use or distribution is prohibited. Please consider the environment before printing this email.

[pengguo-seismo]

Hi Qiancheng and Dimitri,

I tried to make sure that the same spectral element (25 gll points) contains only either acoustic or elastic parameters. The error message disappears, however, there is another issue. I put the source at fluid, and the wavefield will only propagate in the acoustic media. There is no wavefield penetrating through the fluid/solid interface. For this test, I use model=ascii.

Regards Peng

[qcliu0]

Hi Peng,

Sorry for the late getting back because I was a little bit busy.

In specfem2d/EXAMPLES ( https://github.com/geodynamics/specfem2d/tree/master/EXAMPLES), please find "Industrial_Format_SEP https://github.com/geodynamics/specfem2d/tree/master/EXAMPLES/Industrial_Format_SEP" and "fluid_solid https://github.com/geodynamics/specfem2d/tree/master/EXAMPLES/fluid_solid ".

For the simple layered applications, the "fluid_solid" would be qualified.

For complicated investigations such as Marmousi2 please combine the two together.

Hope this helps.

Regards, Qiancheng

On Thu, Dec 20, 2018 at 10:04 AM meatball880207 notifications@github.com wrote:

Hi Qiancheng and Dimitri,

I tried to make sure that the same spectral element (25 gll points) contains only either acoustic or elastic parameters. The error message disappears, however, there is another issue. I put the source at fluid, and the wavefield will only propagate in the acoustic media. There is no wavefield penetrating through the fluid/solid interface. For this test, I use model=ascii.

Regards Peng

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/geodynamics/specfem2d/issues/1010#issuecomment-448895631, or mute the thread https://github.com/notifications/unsubscribe-auth/AU_WPhbDzrOodC9amj960swlgocbZuVsks5u6zZqgaJpZM4ZXle3 .

--

This message and its contents, including attachments are intended solely for the original recipient. If you are not the intended recipient or have received this message in error, please notify me immediately and delete this message from your computer system. Any unauthorized use or distribution is prohibited. Please consider the environment before printing this email.

[EtienneBachmann]

Hi,

Actually, there is indeed a few bugs when it comes to external models, combined with fluid solids. The fluid-solid interface array is not updated when loading the external model, which explains why nothing is transmitted between two media. The same thing happens to the acoustic free surface.

I will work on fixing that in the next days. Best regards,

Etienne

[pengguo-seismo]

Hi Etienne and Qiancheng,

A solution I found to let the waves pass through the fluid/solid interface is, to set the fluid/solid interface Par_file explicitly using nbregions, even if external model is used. But using this method didn't allow bathymetry to be included.

A more elegant way will be great :)

Regards Peng

[komatits]

Hi Etienne, Hi all,

Thanks a lot, and great if you know how to fix it.

Best regards, Dimitri.

On 12/20/18 11:54 PM, EtienneBachmann wrote:

Hi,

Actually, there is indeed a few bugs when it comes to external models, combined with fluid solids. The fluid-solid interface array is not updated when loading the external model, which explains why nothing is transmitted between two media. The same thing happens to the acoustic free surface.

I will work on fixing that in the next days. Best regards,

Etienne

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/geodynamics/specfem2d/issues/1010#issuecomment-449161997, or mute the thread https://github.com/notifications/unsubscribe-auth/AFjDKZJ4FxzkL60eFeA4Zp_SDh0Gle-Qks5u7BUsgaJpZM4ZXle3.

-- Dimitri Komatitsch, CNRS Research Director (DR CNRS) Laboratory of Mechanics and Acoustics, Marseille, France http://komatitsch.free.fr