.png){kind=icon}
Return
Seismic reliability analysis framework based on robust graph data representation and efficient selective state space model
.png){kind=icon}
PRE
AI
DOI：10.1016/j.istruc.2026.112111.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
This study aims to develop a versatile framework, named GraphS6R, for accurately and effectively conducting reliability analysis of multiple 3D building structures subjected to seismic loads. The core idea of the proposed approach lies in its applicability to diverse structural configurations without requiring alterations to its settings or retraining. This feature distinguishes GraphS6R from most existing learning-based reliability analysis models, which, once trained, apply only to specific structural layouts and require reconfiguration and retraining for other structures. To achieve this robustness, we first propose a modular workflow that seamlessly integrates advanced techniques to address various aspects of seismic reliability, including: converting structural geometry and properties into graph representations, aggregating features to combine diverse data types into a 3D feature tensor, constructing a multi-output, multi-channel forecasting metamodel based on a selective state-space deep learning architecture, and incorporating a reliability analysis component that delivers reliability results for various safety criteria. The applicability and performance of GraphS6R are consistently demonstrated through five structural problems with varying configurations and complexities. The framework achieves reasonably accurate prediction results with significantly faster computation times compared to Monte Carlo simulation and other deep learning-based counterparts.
Keywords:
GraphS6R
seismic reliability analysis
graph data representation
selective state space model
3D building structures
Journal
IF:
4.3
Papers:
1.3W
Citations:
2.7W
Organization
No organization information available
Cited Papers
No cited papers available