Use of the model parameter sensitivity analysis for the probabilistic-based seismic assessment of existing buildings

Seismic assessment of existing buildings is usually treated by international codes and guidelines through a semi-probabilistic approach based on the use of the so-called confidence factor (CF). Many authors revealed the inadequacy of such an approach, proposing alternative procedures based on: the updated calibration of the CF values together with its application to a parameter better representative of the structural response than the material strength, as usually adopted by codes; or the fully probabilistic approach by explicitly considering the propagation of uncertainties. Although the latter constitutes the most rigorous approach, it is still computationally demanding and difficult to be integrated as standard tool in the engineering practice. In this paper, the model parameter sensitivity analysis is proposed to support the seismic assessment in various aspects such as: pointing out, in an explicit way, the influence each uncertain parameter has on the structural response; supporting the set of an effective investigation plan; computing the essential parameters for a probabilistic-based verification on basis of a limited number of analyses. To the latter aim, the results from the model parameter sensitivity analysis executed according to the star design with central point approach are used to determine the median intensity measure (IM LS ) and, with the help of the surface response technique, its dispersion (β LS ), that are the two parameters of the fragility curve representing the capacity in the assessment. The proposed methodology is applied on two case studies, representative of existing URM buildings. Firstly, the IM LS and β LS values are calculated and thus compared, for the aim of validation, with the reference ones obtained from nonlinear static analyses performed on a large number of models generated using Monte Carlo simulations. Results obtained show a good estimate of the fragility curve parameters, compared to the rigorous probabilistic approach, highlighting the potential of the procedure proposed.