Probabilistic nonlinear computer simulations for realistic prediction of structural response

Abstract

The effect of inherent uncertainties in material properties on the global response of substandard reinforced concrete (RC) structural members was investigated by the stochastic study. An experimentally validated finite element model (FEM) was, therefore, combined with a suitable stochastic sampling technique (Latin Hypercube Sampling (LHS)). Then, the effect of inherent uncertainties on the material mechanical properties was studied by uncertainty analysis, while the uneven distribution of concrete mechanical properties over the specimen was accurately characterized by random fields theory. The partial correlation coefficient between material parameters and response variables was also evaluated to outline the parameters which mainly contribute to the global response (i.e., sensitivity analysis). Such an advanced modelling strategy was implemented on three different testing programs comprising RC members designed with structural details and material properties non-conforming to current codes and guidelines. The first testing program deals with experimental performance of an over-reinforced and shear critical beams together with stochastic assessments of beam members via computational stochastic mechanics. The effect of uncertainties on the response of shear critical and carbon fiber reinforced polymer (CFRP) retrofitted beam-column joints, which were selected from available testing programs in the literature, was also discussed. The stochastic-based numerical prediction of beam-type RILEM bond specimens characterized variability in the identical tests satisfactorily. Owing to the more realistic assessment capability of the stochastic-based nonlinear finite element (FE) analysis, the global response of the substandard RC members (over-reinforced and shear critical beams; shear critical and CFRP retrofitted beam-column joints; beam-type RILEM bond specimens) was accurately reproduced.