Quantification of the resistance modeling uncertainty of 19 alternative 2D nonlinear finite element approaches benchmarked against 101 experiments on reinforced concrete beams

Abstract

Nineteen 2D nonlinear finite element analysis (NLFEA) solution strategies were benchmarked against a wide variety of 101 experiments on reinforced concrete beams failing in bending, flexural shear, or shear compression. The relatively high number of solution strategies was motivated by the conviction that choices for the constitutive models, the finite element kinematics and equilibrium settings will interact, and must therefore be tested in conjunction. Modeling uncertainty distribution parameters are presented for the 19 solution strategies, using all beams, and using beams with and without stirrups separately. The resulting statistics are discussed against the correctness of the simulated failure modes and failure loads, revealing that rotating crack models perform well for the relatively ductile failures in beams with stirrups, while fixed crack models perform better for the more brittle failures in beams without stirrups. The tailored solution strategies that predict failure modes correctly, imply a log-normal distribution of the modeling uncertainty with relatively low coefficients of variation. The outlook is that these estimates of the statistical properties of the modeling uncertainties could serve as a basis within safety formats.