Machine Learning-based Seismic Signal Denoising with a Case Study of the 2006 Yogyakarta Earthquake Aftershock Sequence
Year: 2025
Type: conference-paper
Abstract: Seismic signal processing is essential for earthquake data analysis, but it is often hindered by noise, especially in data with low Signal-to-Noise Ratio (SNR). This noise, originating from human activity or environmental factors, can obscure key components like P-waves and S-waves. This study aims to evaluate the effectiveness of a machine learning-based denoising technique, specifically the deep denoising method in enhancing seismic signal quality. Focusing on the 2006 Bantul, Yogyakarta aftershock sequence, characterized by high noise variability and low SNR, this research investigates how deep neural networks can separate signals from noise, even when their amplitudes are nearly identical. The results demonstrate that the Deep Denoiser significantly improves signal clarity, outperforming conventional methods in identifying P-waves and S-waves. This study confirms the potential of machine learning approaches in advancing seismic data interpretation, contributing to more accurate earthquake analysis and hazard assessment in future research.
Authors: Revaldi Kukuh Putra, Iktri Madrinovella, Mohamad Ramadhan
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W4413077985