Neural networks for the rapid seismic assessment of existing moment-frame RC buildings

The study presented in this paper analyses and investigates the possibility of introducing a general and rapid methodology based on an artificial neural network (ANN) to assess the seismic response of existing reinforced concrete (RC) buildings. Starting from investigations carried out on buildings located in the outskirts of Bologna, 928 finite element models have been developed on the basis of the most recurrent data. The input parameters representing the characteristics of the structures were systematically varied and, through modal dynamic and non-linear static analyses, the outputs representing the seismic response were recorded. The resulting dataset was used to create a function, based on ANN, that can reliably predict the seismic behaviour of a RC structure. Finally, by means of k-fold cross-validation, the instruction of the function was optimised and simultaneously verified, obtaining a coefficient of determination for the totality of the samples and the previously unseen cases of 0,94 and 0,88, respectively. The result obtained not only aims at enriching the existing framework on the subject, increasing the awareness of the seismic issues affecting this building typology, but also constitutes a prioritization system that could highlight the need for structural renovation.

Automated summary

This study investigates the use of an artificial neural network (ANN) to assess the seismic response of existing reinforced concrete (RC) buildings. Through extensive modeling and data analysis, a reliable function is established to predict seismic behavior, achieving high coefficients of determination through validation. This research enriches the existing framework, increases awareness of seismic issues, and provides a prioritization system for structural renovation.