.png){kind=icon}
.png){kind=icon}
Return
RecFusionED: A recursive deep learning framework for structural response prediction with extrapolation capability
.png){kind=icon}
PRE
AI
S
J
F
J
S
DOI：10.1016/j.oceaneng.2026.127486.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• A novel hierarchical deep learning framework, RecFusionED, is proposed for structural response prediction under multi-hazard coupling. • FusionED unit achieves deep fusion of structural parameters with multi-source excitations via a feature-wise linear modulation mechanism. • A recursive extrapolation architecture enables stable long-sequence prediction beyond training horizons with dynamic teacher forcing. • Ablation studies confirm the critical roles of FiLM fusion and teacher forcing in maintaining prediction stability and generalization.
Keywords:
Structural response prediction
Jacket offshore platform
Multi-source data fusion
Recursive prediction strategy
Long-sequence extrapolation
Journal
IF:
5.5
Papers:
5.5K
Citations:
7.6W
Organization
Cited Papers