Skip to Main content Skip to Navigation
New interface
Theses

Contrôle qualité et étude du traitement sismique par apprentissage statistique : développement d'un outil industriel autonome

Abstract : Seismic data processing has the goal of converting raw field records into images of geological structures below the ground. It may also allow us to recover some physical properties of the sub-surface. Modern seismic surveys can take months to acquire, resulting in several terabytes of seismic data. The huge size of those datasets makes their analysis and processing increasingly long and expensive. Considering this amount of data, it is impossible for geophysicists to systematically verify the coherency of processing results for the whole datasets.To do the processing and quality control of seismic data, a geophysicist needs a lot of expertise in seismic. The different works of this thesis deal with the use of machine learning methods as a tool to aid in this processing and proposes techniques for quality control of the data. An overview of the current state of the art of processing methods and quality control is outlined. A number of approaches are highlighted: firstly, those which are based on our observations of how geophysicists usually build their quality control workflow combined with machine learning approaches. Secondly, new approaches are proposed for the processing and quality control which are based on supervised deep learning in the context of a small amount of labeled data. An alternative method is given, based on semi-supervised learning for the model part and on active learning which takes advantage of labels assigned by the geophysicis
Document type :
Theses
Complete list of metadata

https://pastel.archives-ouvertes.fr/tel-03868270
Contributor : ABES STAR :  Contact
Submitted on : Wednesday, November 23, 2022 - 4:55:15 PM
Last modification on : Thursday, November 24, 2022 - 3:54:00 AM

File

2021UPSLM060_CHAMBEFORT_archiv...
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-03868270, version 1

Citation

Mathieu Chambefort. Contrôle qualité et étude du traitement sismique par apprentissage statistique : développement d'un outil industriel autonome. Sciences de la Terre. Université Paris sciences et lettres, 2021. Français. ⟨NNT : 2021UPSLM060⟩. ⟨tel-03868270⟩

Share

Metrics

Record views

0

Files downloads

0