Bridging Data Gaps in Structural Fragility Modeling through Transfer Learning: Methodology and Case Studies

2026-06-18 · Source: stat.ML updates on arXiv.org · Field: Science & Research — Artificial Intelligence & Machine Learning, Engineering & Applied Sciences · Depth: Expert, quick

Summary

A new methodology-centered transfer learning framework, submitted on June 17, 2026, addresses data gaps in structural fragility modeling by adapting models under domain shift, class imbalance, and scarce target labels. This framework maintains engineering interpretability and supports decision-making under uncertainty. It demonstrates four transfer learning strategies: instance-based, parameter-based, hierarchical Bayesian, and multi-source. Three case studies illustrate these strategies: coastal bridge fragility using Hurricane Katrina observations, residential building fragility with Hurricane Ian data, and seismic bridge fragility from the 2001 Nisqually earthquake. The research finds that direct transfer of existing models fails under these challenging conditions, while targeted adaptation significantly improves failure detection and predictive stability in low-data environments, highlighting a need for systematic guidance in diagnostics and strategy selection.

Key takeaway

For research scientists and engineers developing structural fragility models with limited or mismatched data, directly applying existing models is often insufficient. You should instead implement targeted transfer learning strategies, such as instance-based adaptation via importance weighting or multi-source fusion, to achieve robust failure detection and predictive stability. Prioritize systematic diagnostics, strategy selection, and uncertainty reporting to ensure reliable decision-making in low-data regimes.

Key insights

Transfer learning effectively bridges data gaps in structural fragility modeling under domain shift and scarcity.

Principles

• Direct model transfer often fails under domain shift.
• Targeted adaptation improves predictive stability.
• Systematic guidance is crucial for model adaptation.

Method

The framework applies instance-based, parameter-based, hierarchical Bayesian, and multi-source transfer learning strategies to adapt structural fragility models, preserving interpretability and quantifying uncertainty.

In practice

• Apply importance weighting for instance-based transfer.
• Use partial pooling with hierarchical Bayesian methods.
• Fuse multiple analytical models with learned source weights.

Topics

• Structural Fragility Modeling
• Transfer Learning
• Domain Shift
• Class Imbalance
• Uncertainty Quantification
• Coastal Bridge Fragility

Best for: AI Scientist, Research Scientist

Related on AIssential

• WaveGraphNet: Physics-Consistent Guided-Wave Damage Localization through Coupled Inverse-Forward Graph Learning — Coupling inverse graph localization with a physics-consistent forward model improves damage localization extrapolation in unseen regions.
• A green solvent screening tool for emerging materials via uncertainty aware, transformer enhanced transfer learning — Transfer learning with uncertainty quantification enables accurate green solvent screening despite limited data.
• Uncertainty Quantification of Engineering Structures by Polynomial Chaos Expansion and Multivariate Active Learning — Adaptive sequential sampling for multi-output PCE improves surrogate accuracy and stability by balancing variance exploitation and spatial exploration.
• MoSA: Motion-constrained Stress Adaptation for Mitigating Real-to-Sim Gap in Continuum Dynamics via Learning Residual Anisotropy — MoSA learns residual anisotropy and heterogeneity using physics priors to close the real-to-sim gap in continuum dynamics.
• Bridging Domain Expertise and Generalization for Performance Estimation — FRAP unifies foundation model generalization with base model expertise for robust performance estimation under distribution shift.
• How Useful is Causal Invariance for Domain Adaptation in Finite-Sample Settings? — Causal invariance improves supervised domain adaptation in finite-sample settings when target-risk margins between candidate models are sufficiently large.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by stat.ML updates on arXiv.org.