TimesNet-Gen: Deep Learning-based Site Specific Strong Motion Generation

2025/12/04 by Barış Yılmaz, Yilmaz, Baris, Bevan Deniz Çılğın +5
Computer Science · Engineering · #Artificial Intelligence (cs.AI) #FOS: Computer and information sciences #Machine Learning (cs.LG) #Seismic Performance and Analysis #Seismology and Earthquake Studies #Structural Health Monitoring Techniques

paper · pdf · doi:10.48550/arxiv.2512.04694

openalex publication_date 2025/12/04 · openalex created_date 2025/12/06 · openalex updated_date 2026/07/28

Abstract

Effective earthquake risk reduction relies on accurate site-specific evaluations, which require models capable of representing the influence of local site conditions on ground motion characteristics. We address strong ground motion generation from time-domain accelerometer records and introduce the TimesNet-Gen, a deep generative framework. In this framework, site-specific generation is directly achieved through a station-restricted, Dirichlet-based latent space resampling strategy, without relying on explicit conditioning inputs or dimensionality reduction. Pre-trained on the AFAD dataset via self-supervised learning, the frozen model demonstrates robust cross-regional generalization by successfully generating station-specific NGA-West2 records without any fine-tuning. Model performance is evaluated by comparing the distributions of generated and real records in the log-HVSR space, alongside the joint analysis of peak ground acceleration and fundamental site frequency. As a baseline, we construct a spectrogram-based conditional variational autoencoder (CVAE) explicitly formulated for station-specific latent space modeling. The results show strong station-wise alignment, consistent cross-regional ground motion synthesis, and a favorable comparison with a spectrogram-based conditional variational autoencoder baseline, demonstrating that the model empirically maintains the essential physical coupling between frequency content and peak amplitude. Our codes are available via https://github.com/brsylmz23/TimesNet-Gen.

Citations

• Broadband Ground Motion Synthesis by Diffusion Model with Minimal Condition
• High Resolution Seismic Waveform Generation using Denoising Diffusion
• Deep Learning-based Average Shear Wave Velocity Prediction using Accelerometer Records
• Learning Physics for Unveiling Hidden Earthquake Ground Motions via Conditional Generative Modeling
• Deep Learning-based Epicenter Localization using Single-Station Strong Motion Records
• TimeMixer: Decomposable Multiscale Mixing for Time Series Forecasting
• Site-Specific Ground Motion Generative Model for Crustal Earthquakes in Japan Based on Generative Adversarial Networks
• Exploring Challenges in Deep Learning of Single-Station Ground Motion Records
• Broadband Ground Motion Synthesis via Generative Adversarial Neural Operators: Development and Validation
• A Time Series is Worth 64 Words: Long-term Forecasting with Transformers
• TimesNet: Temporal 2D-Variation Modeling for General Time Series Analysis
• Data-driven Accelerogram Synthesis using Deep Generative Models
• Broadband Ground-Motion Simulation Using a Hybrid Approach

Related

• Learning Physics for Unveiling Hidden Earthquake Ground Motions via Conditional Generative Modeling
• Exploring Challenges in Deep Learning of Single-Station Ground Motion Records
• QuakeFormer: A Uniform Approach to Earthquake Ground Motion Prediction Using Masked Transformers
• Data-driven Accelerogram Synthesis using Deep Generative Models
• PGVnet: a machine learning framework for the generation of rapid, physics-consistent PGV maps