Generación de señales volcánicas artificiales del Volcán Cotopaxi usando modelos GAN (Generative Adversarial Networks) : Generación de señales volcánicas artificiales del tipo VT (Volcano-Tectonic) y LP (Long Period) usando un modelo cGAN (Conditional Generative Adversarial Network).

Abstract

Ecuador is a country that has many active volcanoes, including Cotopaxi, whose eruption could cause irreparable damage such as loss of human life, as well as environmental, economic and social problems, among others; For this reason, the study of volcanic seismic signals for prevention is of the utmost importance, but the amount of original data available is a limitation. The present work uses a cGAN model to generate new volcanic seismic signals of the type VT and LP. This model was developed in the Pytorch library for Python and was trained with the IGEPN public database. The generated signals were compared with the original signals using the FID metric, reaching an acceptable value for this type of signals and being similar to the original ones both in the time domain and in the frequency domain.