Banca de QUALIFICAÇÃO: KATERINE DE JESUS RINCON PEREZ
Uma banca de QUALIFICAÇÃO de DOUTORADO foi cadastrada pelo programa.STUDENT : KATERINE DE JESUS RINCON PEREZ
DATE: 11/12/2024
TIME: 16:00
LOCAL: http://meet.google.com/dci-ixyc-fma
TITLE:
Deep learning applied to time-lapse seismic
KEY WORDS:
Time-lapse seismic, OBN, GANs, non-repeatability effects, deep learning
PAGES: 57
BIG AREA: Ciências Exatas e da Terra
AREA: Ciência da Computação
SUMMARY:
Time-lapse or 4D seismic imaging is a method for monitoring reservoir changes, providing reservoir knowledge that aids in strategic management. However, a significant limitation lies in the non-repeatability of data that occurs in time-lapse (TL) seismic due to different environmental factors and variations in equipment positioning between surveys, which can mask changes in the reservoir. This work addresses this limitation by employing Generative Adversarial Networks (GANs) to mitigate the effects of non-repeatability in 4D seismic data acquired with Ocean Bottom Node (OBN) seismic. By preprocessing these data with a cGAN-like network to adjust for non-repeatability effects and implementing them later in a Full Waveform Inversion (FWI) workflow, we aim to improve data quality before FWI is applied. Building on recent advances in deep learning (DL) applied to 4D seismic data; the research integrates a matching technique between TL acquisitions, which uses conditional Generative Adversarial Networks (cGANS). Initial results demonstrate consistent performance on the central traces of each seismogram, validated through metrics such as mean squared error (MSE), root mean squared error (RMS), and structural similarity index (SSIM). Subsequently, a reverse time migration (RTM) was performed as a preliminary evaluation step before FWI.
COMMITTEE MEMBERS:
Presidente - 1673543 - SAMUEL XAVIER DE SOUZA
Externo ao Programa - 1379465 - GILBERTO CORSO - UFRNExterno ao Programa - 2492756 - JOAO MEDEIROS DE ARAUJO - UFRNExterno ao Programa - 3216921 - TIAGO TAVARES LEITE BARROS - UFRN