Banca de QUALIFICAÇÃO: ADÈLE MEGUEDONG YOTA

Uma banca de QUALIFICAÇÃO de MESTRADO foi cadastrada pelo programa.
STUDENT : ADÈLE MEGUEDONG YOTA
DATE: 23/02/2026
TIME: 09:30
LOCAL: DIMAp (presencial) e Google Meet (remoto)
TITLE:

Identification and Classification of Seismic Events

KEY WORDS:

Keywords: Seismic Catalogs, PhaseNet, Deep Neural Networks, GaMMA, Computational Geophysics, Linear Predictive Coding

PAGES: 61
BIG AREA: Ciências Exatas e da Terra
AREA: Ciência da Computação
SUBÁREA: Metodologia e Técnicas da Computação
SPECIALTY: Processamento Gráfico (Graphics)
SUMMARY:

The detection and association of seismic activities is a time-consuming procedure that involves the analysis of geophysical data from seismographic
stations by specialized technicians. This analysis has applications ranging from the geological characterization of the Earth’s crust to supporting the prediction of seismic incidents and hydrocarbon exploration. Due to the development and innovation of Artificial Intelligence tools, the objective of this work is to optimize computational time in the automatic elaboration of the seismic catalog with precision for the municipality of Jacobina in Bahia, located in the Northeast region of Brazil. To achieve this, we identify patterns in seismic events using the transfer learning technique and the deep neural network PhaseNet algorithm for selecting the arrival times of primary (P) and secondary (S) waves from seismic events, and the GaMMA algorithm for associating these phases, followed by the hypocentral location of each event. An intermediate step is the representation of the spectral envelope of each seismic event using the Linear Predictive Coding (LPC) method to measure similarity between events, supported by expert analysis from Geophysics professionals to generate the gold standard for the dataset used. A seismic catalog consists of the epicentral coordinates, origin time, depth, and local magnitude of the events. For this study, a seismic database from Bahia, pertaining to March 2022 and managed by 6 stations from the Seismological Laboratory of UFRN (LabSis), was used. The performance indicators, such as detection precision, aftershock filtering precision, association and location precision, were used to evaluate the algorithms. The initial results for the selection of P and S seismic phases achieved a detection precision greater than or equal to 60%, with a total of 1135 picks across the 6 stations. Applying these results to GaMMA (Gaussian Mixture Model Association), the association resulted in 51 seismic events with a minimum of 5 picks per event, and a precision greater than or equal to 5. The processing of this data (1135 picks) was executed in 34 seconds.

 

 

COMMITTEE MEMBERS:
Presidente - 1221251 - MARTIN ALEJANDRO MUSICANTE
Interno - 2177445 - BRUNO MOTTA DE CARVALHO
Externo ao Programa - 2497950 - SELAN RODRIGUES DOS SANTOS - UFRN