Deep preconditioners and their application to seismic wavefield processing
Seismic data processing heavily relies on the solution of physics-driven inverse problems. In the presence of unfavourable data acquisition conditions (e.g., regular or irregular coarse sampling of sources and/or receivers), the underlying inverse problem becomes very ill-posed and prior information...
| Main Author: | |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Frontiers Media S.A.
2022-09-01
|
| Series: | Frontiers in Earth Science |
| Subjects: | |
| Online Access: | https://www.frontiersin.org/articles/10.3389/feart.2022.997788/full |
| _version_ | 1818017569611710464 |
|---|---|
| author | Matteo Ravasi |
| author_facet | Matteo Ravasi |
| author_sort | Matteo Ravasi |
| collection | DOAJ |
| description | Seismic data processing heavily relies on the solution of physics-driven inverse problems. In the presence of unfavourable data acquisition conditions (e.g., regular or irregular coarse sampling of sources and/or receivers), the underlying inverse problem becomes very ill-posed and prior information is required to obtain a satisfactory solution. Sparsity-promoting inversion, coupled with fixed-basis sparsifying transforms, represent the go-to approach for many processing tasks due to its simplicity of implementation and proven successful application in a variety of acquisition scenarios. Nevertheless, such transforms rely on the assumption that seismic data can be represented as a linear combination of a finite number of basis functions. Such an assumption may not always be fulfilled, thus producing sub-optimal solutions. Leveraging the ability of deep neural networks to find compact representations of complex, multi-dimensional vector spaces, we propose to train an AutoEncoder network to learn a nonlinear mapping between the input seismic data and a representative latent manifold. The trained decoder is subsequently used as a nonlinear preconditioner for the solution of the physics-driven inverse problem at hand. Through synthetic and field data examples, the proposed nonlinear, learned transformations are shown to outperform fixed-basis transforms and converge faster to the sought solution for a variety of seismic processing tasks, ranging from deghosting to wavefield separation with both regularly and irregularly subsampled data. |
| first_indexed | 2024-04-14T07:28:39Z |
| format | Article |
| id | doaj.art-d0480707d0a74d859d096fed7ffe2a9a |
| institution | Directory Open Access Journal |
| issn | 2296-6463 |
| language | English |
| last_indexed | 2024-04-14T07:28:39Z |
| publishDate | 2022-09-01 |
| publisher | Frontiers Media S.A. |
| record_format | Article |
| series | Frontiers in Earth Science |
| spelling | doaj.art-d0480707d0a74d859d096fed7ffe2a9a2022-12-22T02:05:57ZengFrontiers Media S.A.Frontiers in Earth Science2296-64632022-09-011010.3389/feart.2022.997788997788Deep preconditioners and their application to seismic wavefield processingMatteo RavasiSeismic data processing heavily relies on the solution of physics-driven inverse problems. In the presence of unfavourable data acquisition conditions (e.g., regular or irregular coarse sampling of sources and/or receivers), the underlying inverse problem becomes very ill-posed and prior information is required to obtain a satisfactory solution. Sparsity-promoting inversion, coupled with fixed-basis sparsifying transforms, represent the go-to approach for many processing tasks due to its simplicity of implementation and proven successful application in a variety of acquisition scenarios. Nevertheless, such transforms rely on the assumption that seismic data can be represented as a linear combination of a finite number of basis functions. Such an assumption may not always be fulfilled, thus producing sub-optimal solutions. Leveraging the ability of deep neural networks to find compact representations of complex, multi-dimensional vector spaces, we propose to train an AutoEncoder network to learn a nonlinear mapping between the input seismic data and a representative latent manifold. The trained decoder is subsequently used as a nonlinear preconditioner for the solution of the physics-driven inverse problem at hand. Through synthetic and field data examples, the proposed nonlinear, learned transformations are shown to outperform fixed-basis transforms and converge faster to the sought solution for a variety of seismic processing tasks, ranging from deghosting to wavefield separation with both regularly and irregularly subsampled data.https://www.frontiersin.org/articles/10.3389/feart.2022.997788/fullseismic processingseismic data analysisdimensionality reductionunsupervised learningdeep learninginverse problems |
| spellingShingle | Matteo Ravasi Deep preconditioners and their application to seismic wavefield processing Frontiers in Earth Science seismic processing seismic data analysis dimensionality reduction unsupervised learning deep learning inverse problems |
| title | Deep preconditioners and their application to seismic wavefield processing |
| title_full | Deep preconditioners and their application to seismic wavefield processing |
| title_fullStr | Deep preconditioners and their application to seismic wavefield processing |
| title_full_unstemmed | Deep preconditioners and their application to seismic wavefield processing |
| title_short | Deep preconditioners and their application to seismic wavefield processing |
| title_sort | deep preconditioners and their application to seismic wavefield processing |
| topic | seismic processing seismic data analysis dimensionality reduction unsupervised learning deep learning inverse problems |
| url | https://www.frontiersin.org/articles/10.3389/feart.2022.997788/full |
| work_keys_str_mv | AT matteoravasi deeppreconditionersandtheirapplicationtoseismicwavefieldprocessing |