Fragility assessment of transmission line supports under non-synoptic wind loads using multi-fidelity stochastic emulators

Fragility functions are central to the Performance-Based Wind Engineering (PBWE) framework for assessing the reliability of transmission lines (TLs) under tornado and downburst wind loads. Traditionally, these curves are constructed using Monte Carlo simulations; however, the extensive number of replications required by nonlinear dynamic time-history analyses renders such simulations computationally prohibitive for practical fragility estimation. In this study, a multi-fidelity stochastic surrogate modeling approach is applied to efficiently construct fragility functions for TL structures subjected to tornado-induced wind loads. The framework employs the recently developed multi-fidelity generalized lambda model (MF-GLaM) emulator combining low- and high-fidelity simulations to reproduce the full probabilistic relationship between intensity measures (IMs) and engineering demand parameters (EDPs). This approach avoids reliance on replication-based methods or restrictive assumptions regarding the probability distributions of the fragility functions. A case study is presented on a simplified linear-elastic transmission tower, where the MF-GLaM accurately predicts non-monotonic and heteroscedastic bi-dimensional fragility surfaces under stochastic tornadic wind fields. The results demonstrate that the MF-GLaM significantly outperforms single-fidelity emulators, achieving lower prediction errors, faster convergence, and enhanced accuracy while reducing computational costs compared to the single-fidelity approach. Overall, the proposed framework offers a general and scalable methodology for probabilistic fragility assessment of structures under complex, transient wind phenomena, facilitating efficient risk-based design within the PBWE paradigm.