Global sensitivity analysis of composite columns in steel and concrete

This study investigates the flexural buckling of composite columns in steel and concrete subjected to centric compressive force. The load bearing capacity of the composite columns under compressive force is influenced by many parameters (random variables) such as: compressive strength of concrete, steel and reinforcement yielding stress, geometry of the cross-section, geometrical and material imperfections etc. Being so, it is in our interest to know the impact of each parameter on global resistance, and one way to do it, is by performing sensitivity analysis. In this study, the Sobol indices is employed to perform sensitivity analysis. In finite element analysis (FEA), the input parameters e.g., geometry of the structure, materials, loads, and other structural properties are deterministic, while the stochastic finite element method (SFEM) is an extension of FEM, which allows variation of these parameters as random variables using normal, lognormal distribution or other type of distribution. To generate random samples arising from variation (fluctuation) of the geometry and material parameters of the composite columns, the Latin Hypercube Sampling (LHS) is used. Regarding the number of simulations, per each column length varying from 2 m up to 10 m, a series of 2000 Latin Hypercube Sampling (LHS) were executed. The results of Sobol indices sensitivity analyses show correlation between each random variable and the ultimate resistance of the composite column without considering the interaction with other random variables called Main Sobol index and the Total Sobol index accounting also for the interaction between random variables.