Abstract:
This study explores classical machine learning methods for fault detection in two seismic volumes: F3 (Netherlands) and Thebe (Australia). For each volume, we created a training dataset by extracting features through seismic attribute computation and a sliding window method. We compared multiple classifiers through a 10-fold cross-validation, with the Support Vector Machine (SVM) achieving the highest F1 score. The selected model was tested on both intra-volume (same volume, different region) and inter-volume scenarios. Results showed that models performed best when trained and tested on data from the same volume.
Referência:
CUSTÓDIO, Gustavo Torres; SAAR, Hugo Ferreira; HELENO, Aline Fernandes; GAMBA, Carlos Tadeu de Caralho; SILVA, Celso Luciano Alves da, SILVA, Cristina Maria Ferreira da; VIRISSIMO, Denis Bruno; SALES, Elisa Morandé; SILLES, Felipe Silva; GUIRELI NETTO, Leonides; GUERRA, Ney Ferreira de Souza; GANDOLFO, Otavio Coaracy Brasil; RUBO, Rafael Andrello; AOYAGI, Thiago Yuji. Fault detection in 3D seismic data with machine learning methods and multiple atributes. In: SBGf CONFERENCE SUSTAINABLE GEOPHYSICS AT THE SERVICE OF SOCIETY, 2025, Rio de Janeiro. Lecture… 27 slides.
Acesso aos slides da apresentação da Palestra em PDF: