Reliability assessment of a multi-response bridge system considering climate change effects on wind and wave loads

This study proposes a novel reliability analysis framework to efficiently evaluate the long-term reliability of multi-response structures under climate change. In this framework, an error-controlled multi-input-multi-output surrogate model is developed based on a support vector regression method, which can propagate uncertainties of structural parameters to responses with satisfactory accuracy while minimising computational costs. By integrating the copula theory and reliability assessment method, the multi-response failure probability is derived, considering both correlations between responses and uncertainties of performance thresholds. In addition, future data on wind climate are collected to account for climate change, and a neural network model is trained using historical and simulated data to predict future wave heights. The framework is then applied to a long-span cable-stayed and suspension bridge system as a case study. Analysis results demonstrate that a comprehensive multi-responses reliability analysis, which considers correlations between responses, can reflect the overall fragility of structures and prevent the underestimation of structural risk. Though the trends in structural failure probabilities vary depending on the climate model, climate change significantly influences structural risk assessment. This study provides a novel perspective for the structural reliability assessment of complex structures.