What is the matching relationship between the rupture model (srf) mesh and the computational domain mesh size? Sometimes null values appear in the calculation (SW4 (Seismic Waves, 4th order))
In seismic simulations like those conducted with SW4 (Seismic Waves, 4th order), the rupture model mesh (often defined in a .srf file) must align appropriately with the computational domain mesh to avoid inaccuracies, such as null values in the results. These null values typically arise when the spatial discretization of the rupture source does not match the computational grid resolution, leading to interpolation or aliasing issues in the waveform data.
Here's how to ensure mesh compatibility:
Mesh Resolution Matching:Ensure that the rupture model (srf file) mesh resolution aligns closely with the computational domain mesh. Typically, the rupture model’s spatial resolution should not be finer than the computational mesh, as this can lead to undersampling or ineffective interpolation. If the .srf mesh resolution is finer, use downsampling techniques to match the computational grid's resolution. Alternatively, increase the computational mesh size to match the .srf mesh, if computational resources allow.
Interpolation Techniques:In cases where exact matching is difficult, interpolation techniques can be applied to map rupture data onto the computational grid. SW4 can handle interpolation, but this may introduce errors if the mesh sizes differ significantly. It's important to ensure that the interpolation method in use minimizes distortion by accurately capturing high-frequency details of the rupture model within the computational domain.
Domain Size and Boundary Conditions:Adjust the computational domain size to ensure all rupture data points from the .srf file fit within the simulation boundaries. Null values often appear if portions of the rupture model extend outside the computational domain.
Testing and Validation:Run smaller simulations to validate the mesh compatibility, examining outputs for any unexpected null values. Modifying the simulation's resolution parameters or applying domain damping near boundaries can help mitigate inaccuracies.
Using a consistent mesh resolution and running smaller, validated test simulations can help you determine an optimal configuration to avoid these null values and improve simulation fidelity.
Thank you for your answer. Under difficult calculation conditions, I used a source model with a sub-fault size of 1km and a calculation grid size of 600m. The calculated PGA showed a large value, with an average value of more than 2g and even 8g. Have you encountered such a problem?