• Citer

Deterministic and probabilistic finite element analysis of reinforced concrete structures subjected to extreme events using a multilayered approach: Application to column loss and fire scenarios

par Sosso, Batoma
Président du jury Bouillard, Philippe
Promoteur Berke, Peter
Co-Promoteur Massart, Thierry,Jacques
Publication Non publié, 2022-01-11

Thèse de doctorat

• Accès en ligne
• Détails
• Contenu
• Statistiques

Résumé :

This work focuses on the development and employment of advanced numerical finite element modelling in reinforced concrete (RC) structures subjected to two accidental loads: sudden column loss and fire. A corotational 2D multilayered Bernoulli beam approach is employed in a correlated reduced Latin Hypercube sampling-based framework where uncertain model parameters are incorporated for the material behavior and cross sectional geometry. These uncertainties are investigated and shown to influence significantly the structural response, being the main focus of this contribution. The structural response of a realistic RC structure subjected to column loss when incorporating material properties variability is investigated in a dynamic approach (geometrical non-linearities coupled to layer plasticity and fracture) where the influence of each random variable (RV) is quantified in a Latin Hypercube Sampling (LHS) approach spanning over 12.000 simulations and leading to recommendations to mitigate progressive collapse: correct design of stirrups and good choice of steel reinforcement quality. The set of uncertain model parameters is extended by geometrical parameters of the cross section of the beam/frame structural elements when a stochastic analysis of the structural response and fire resistance is targeted. The corotational layered beam is adapted for simulations at elevated temperature by the implementation of a closed from thermal model coupled to the calculation of thermally induced strains from different sources and ensuring that all of the material parameters and their temperature evolution are adjusted to experimental data. The resulting efficient computational model compares favourably to experimental data from the literature, validating the proposed model. Its application in an LHS framework (+20.000 simulations) is shown to allow identifying a reduced set of dominant uncertain material and geometrical parameters for the simulation of RC members subjected to fire: bottom concrete cover, steel and concrete yield strength.Considering the large computational effort of full probabilistic models an attempt is made to use beam-column macro-models in a structural simulation of column loss. A first step towards the computational identification of the joint model parameters form layered beam simulations is established.