Influence of Slenderness on the Evaluation of Epistemic Uncertainty Related to Non-Linear Numerical Analysis of RC Columns

This investigation is devoted to quantify the epistemic uncertainty related to the nonlinear analysis of reinforced concrete columns characterized by high slenderness using numerical codes. The adoption of refined numerical tools, which are able to consider both mechanical and geometric non linearities, implies to perform assumptions and approximations with respect to reality. Whit reference to reliability analysis, these simplifications lead, inevitably, to additional uncertainties which are of epistemic nature. In fact, these uncertainties may be reduced by the engineers/analysts by increasing the level of refinement of the numerical model and/or increasing knowledge about parameters associated to material models. However, also numerical model established by expert engineers/analysts are affected by this kind of epistemic uncertainty. Accepting that the level of uncertainty associated to the experimental tests set are minimized, the epistemic uncertainty associated to non-linear numerical simulations can be quantified characterizing the model uncertainty random variable comparing the outcomes of numerical results to the associated experimental ones. The present investigation proposes the quantification of the model uncertainty related to non-linear numerical simulations of slender RC columns. A total number of 40 experimental results known from literature are herein selected in coherence with current Eurocodes specifications. The experiments are reproduced adopting non-linear numerical analysis differentiating between several modelling hypotheses (i.e., numerical code; materials models). The comparison between experimental and numerical results is adopted to characterize the most suitable probabilistic model for the model uncertainty random variable associated to non-linear numerical simulations of RC columns subjected to significant slenderness. The outcomes of the research are useful to provide background to the characterization of partial safety factor for model uncertainty in non-linear numerical analysis using the approach of the global resistance format for safety verifications.