A novel data-driven force–displacement macro-model for nonlinear analysis of infilled frames: development, validation and reliability comparison

Equivalent strut macro-models are largely employed in research and practice, to reproduce the effect of infill walls in frame structures. These models are simple and at the same time computationally effective, but because of their phenomenological nature, have the big commitment to summarize a complex mechanical interaction with a relatively simple analytical relationship. Because of this, the overall inelastic response of an infilled frame is quite sensitive to that adopted for the equivalent strut. Most of literature models make use of mechanical approaches to define the equivalent strut force–displacement response but, given the large uncertainty in predicting collapse modes actually occurring, results from these models are many times conflicting and raise doubts on their reliability. Based on the limitations of mechanics-based models, this paper proposes a new quadri-linear empirical force–displacement relationship to be used for the equivalent struts. The model definition follows a data-driven approach rather than a mechanical one, therefore parameters defining the force–displacement curve are analytically evaluable by means of empirical correlations. The analytical correlations are derived from an experimental data-set augmented with data obtained data from additional numerical simulations by a refined finite element model. Blind validation tests of the model are carried against six experimental tests not previously included in the data-set definition. A reliability comparison between mechanics-based models and empirical models is finally presented and discussed.