Exploring Challenges in Deep Learning of Single-Station Ground Motion Records

Ümit Mert Çağlar,Baris Yilmaz,Melek Türkmen,Erdem Akagündüz,Salih Tileylioglu

2024-03-12

Abstract:Contemporary deep learning models have demonstrated promising results across various applications within seismology and earthquake engineering. These models rely primarily on utilizing ground motion records for tasks such as earthquake event classification, localization, earthquake early warning systems, and structural health monitoring. However, the extent to which these models effectively learn from these complex time-series signals has not been thoroughly analyzed. In this study, our objective is to evaluate the degree to which auxiliary information, such as seismic phase arrival times or seismic station distribution within a network, dominates the process of deep learning from ground motion records, potentially hindering its effectiveness. We perform a hyperparameter search on two deep learning models to assess their effectiveness in deep learning from ground motion records while also examining the impact of auxiliary information on model performance. Experimental results reveal a strong reliance on the highly correlated P and S phase arrival information. Our observations highlight a potential gap in the field, indicating an absence of robust methodologies for deep learning of single-station ground motion recordings independent of any auxiliary information.

Signal Processing,Computer Vision and Pattern Recognition,Machine Learning

What problem does this paper attempt to address?

This paper mainly discusses the limitations of deep learning in the analysis of single-station seismic records. Although deep learning models have shown potential in the field of seismology and earthquake engineering, the extent to which they effectively learn from complex time series signals has not been fully explored. The researchers evaluated the performance of two deep learning models (ResNet and Temporal Convolutional Networks, TCN) in learning seismic records through hyperparameter search, and analyzed the impact of auxiliary information (such as arrival times of seismic waves) on the model performance. The experimental results show that the models highly depend on the highly correlated arrival time information of P-waves and S-waves, indicating that the current methods may not fully utilize deep learning from a single seismic record. The paper points out that the lack of robust deep learning methods that do not rely on auxiliary information to process single-station seismic records is a potential research gap. In addition, the study conducted experiments using the Stanford Earthquake Dataset (STEAD), which contains a large amount of seismic waveforms. Through experiments on different hyperparameters and models, the authors found that the P/S wave arrival time information significantly affects the prediction of epicentral distance. The paper also provides a discussion on how to avoid excessive reliance on this auxiliary information to improve the model's generalization ability. In summary, the paper aims to address the issue of whether deep learning models overly rely on auxiliary information when predicting epicentral distance using single-station seismic records, revealing the limitations of current methods, and proposing directions for future research.

Exploring Challenges in Deep Learning of Single-Station Ground Motion Records

Deep Learning-based Epicenter Localization using Single-Station Strong Motion Records

Deep Learning-based Average Shear Wave Velocity Prediction using Accelerometer Records

Toward Fully Autonomous Seismic Networks: Backprojecting Deep Learning-Based Phase Time Functions for Earthquake Monitoring on Continuous Recordings

An End-to-End Earthquake Detection Method for Joint Phase Picking and Association using Deep Learning

Winning with Simple Learning Models: Detecting Earthquakes in Groningen, the Netherlands

Deep Learning in Earthquake Engineering: A Comprehensive Review

Deep vs. Shallow Learning: A Benchmark Study in Low Magnitude Earthquake Detection

Deep Learning-based Small Magnitude Earthquake Detection and Seismic Phase Classification

Seismic event and phase detection using deep learning for the 2016 Gyeongju earthquake sequence

Deep learning for low-magnitude earthquake detection on a multi-level sensor network

A Practical Approach to Automatic Earthquake Catalog Compilation in Local OBS Networks Using Deep-Learning and Network-Based Algorithms

Real-time Earthquake Monitoring using Deep Learning: a case study on Turkey Earthquake Aftershock Sequence

Using a Deep Neural Network and Transfer Learning to Bridge Scales for Seismic Phase Picking

Multi-task multi-station earthquake monitoring: An all-in-one seismic Phase picking, Location, and Association Network (PLAN)

Testing the Potential of Deep Learning in Earthquake Forecasting

SCALODEEP: A Highly Generalized Deep Learning Framework for Real‐Time Earthquake Detection

An Unsupervised Machine Learning Approach for Ground-Motion Spectra Clustering and Selection

Deep learning for seismic structural monitoring by accounting for mechanics-based model uncertainty

Earthquake transformer—an attentive deep-learning model for simultaneous earthquake detection and phase picking

Generalization of Deep-Learning Models for Classification of Local Distance Earthquakes and Explosions across Various Geologic Settings