Reliability‐based assessment of a multilinear regression model for estimating the ultimate load of eccentrically loaded slender circular <scp>CFDST</scp> columns

Abstract This paper presents a reliability‐based assessment of the multilinear regression (MR) model to predict the ultimate load of slender circular concrete‐filled double‐skin tubular (CFDST) columns under eccentric loading. To assess the MR model's performance, reliability analyses of 10 experimental CFDST columns are conducted and then compared to design code formulae (EN1994, ACI 318‐11, and AISC 2010). Results show that the MR model insignificantly exceeds the target reliability index within a mean absolute error (MAE) margin of 0.37, whereas the opposite applies for the design codes with MAE values of 2.18, 2.74, and 2.13 for EN1994, ACI 318‐11, and AISC 2010, respectively. First‐order reliability method sensitivity coefficients computed show that the model uncertainty has a minimal impact on the MR model but significantly affects the design models. Capacity adjustment factors, based on model uncertainty and eccentricity, are proposed for the design models, leading to improved agreement with target reliability indices with MAE values of 0.17, 1.08, and 0.33. These findings emphasize that the MR model and the modified design models are applicable for use in accurately predicting the ultimate load of slender circular CFDST columns subjected to eccentric loading with acceptable levels of reliability and safety margins.