A method for predicting great earthquake magnitude by HR-GNSS combining physical perception and attention mechanism
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHANDONG UNIV OF SCI & TECH
- Filing Date
- 2026-03-20
- Publication Date
- 2026-05-29
AI Technical Summary
Existing magnitude estimation methods based on HR-GNSS and deep learning struggle to adaptively differentiate between high signal-to-noise ratio and low-quality stations under multi-station input conditions. Furthermore, they fail to adequately incorporate the geometric relationship between the seismic source and the station, resulting in insufficient stability and generalization ability of the model in different regions and complex observation environments.
A large earthquake magnitude prediction method based on HR-GNSS that integrates physical perception and attention mechanisms is adopted. The method extracts waveform features from multiple stations through a convolutional neural network, introduces a station-dimensional dot product self-attention mechanism for feature interaction, and combines physical auxiliary features of epicentral distance for feature fusion to output the magnitude prediction value.
It enables rapid and accurate estimation of large moment magnitudes under complex observation conditions, improves the robustness and stability of the model, can adaptively identify the characteristics of high-quality stations, suppress interference from low signal-to-noise ratio stations, and enhances the model's adaptability in different regions and observation environments.
Smart Images

Figure CN121899887B_ABST