Broadband Ground Motion Synthesis by Diffusion Model with Minimal Condition

2024-12-23

arXiv.org

AI Summary

Existing seismic waveform generation methods suffer from low fidelity and poor generalizability across diverse tectonic regions. Method: We propose the first end-to-end differentiable framework integrating a conditional latent diffusion model with a high-fidelity waveform reconstruction module, requiring only minimal inputs—e.g., moment magnitude and epicentral distance—to synthesize broadband 3D (E/N/Z) ground motions. The framework explicitly models P- and S-wave arrival times, envelope morphology, signal-to-noise ratio (SNR), and spectral characteristics. Contribution/Results: It achieves unified training and cross-regional generalization—across North America, East Asia, and Europe—on a single GPU. Joint optimization enforces consistency with ground motion prediction equations (GMPEs), minimizes spectral distortion, and maximizes spectrogram alignment. Experiments show P/S-phase timing errors < 0.3 s, envelope correlation > 0.92, and SNR improvement of 5.8 dB—substantially outperforming state-of-the-art methods—enabling real-time seismic risk simulation and engineering-based seismic design.

Technology Category

Computer Vision: Diffusion Models for VisionNatural Language Processing: GenerationSearch and Optimization: Sampling/Simulation-based Search

Application Category

Web Mining and Content Analysis: Web data generation and simulationGraph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphsUser Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems

Abstract

Shock waves caused by earthquakes can be devastating. Generating realistic earthquake-caused ground motion waveforms help reducing losses in lives and properties, yet generative models for the task tend to generate subpar waveforms. We present High-fidelity Earthquake Groundmotion Generation System (HEGGS) and demonstrate its superior performance using earthquakes from North American, East Asian, and European regions. HEGGS exploits the intrinsic characteristics of earthquake dataset and learns the waveforms using an end-to-end differentiable generator containing conditional latent diffusion model and hi-fidelity waveform construction model. We show the learning efficiency of HEGGS by training it on a single GPU machine and validate its performance using earthquake databases from North America, East Asia, and Europe, using diverse criteria from waveform generation tasks and seismology. Once trained, HEGGS can generate three dimensional E-N-Z seismic waveforms with accurate P/S phase arrivals, envelope correlation, signal-to-noise ratio, GMPE analysis, frequency content analysis, and section plot analysis.

Problem

• Generating realistic earthquake ground motion waveforms
• Improving waveform quality using diffusion models
• Validating performance across diverse global earthquake datasets

Innovation

• Uses conditional latent diffusion model
• End-to-end differentiable generator design
• Generates 3D seismic waveforms accurately