A semi-empirical method for shear response modelling of masonry infilled frame structures

Abstract

This paper presents a novel experimental and theoretical methodology for the fragility assessment of masonry infilled frame structures subjected to seismic loads. The method uses a Hamiltonian Monte Carlo Bayesian Neural Network trained with laboratory tests, to obtain the constitutive parameters of a non-linear spring model that represents the masonry shear behaviour. The resulting model accounts for several types of masonry units, structural steel and reinforced concrete frames along with the effects of windows and/or doors openings. The results show that the use of deterministic models lead to poor estimations about the in-plane behaviour of the system, whereas the application of the proposed semi-empirical method results in more robust predictions according to the measured data. Also, the proposed approach is tested against two extra data-sets to evaluate its extrapolation capabilities, with satisfactory results. Moreover, the proposed method has been applied to an engineering case study which demonstrates that it can be efficiently applied to robustly assess the safety against collapse of MIF buildings. Finally, a discussion between the proposed method and the current structural standards is provided within the context of the case study.