Machine Learning Techniques for Analysing the Seismic Response in Multistorey Steel Structures

Abstract

This paper provides a detailed analysis that uses an Artificial Neural Network (ANN) in forecasting the seismic response in multistorey steel structures. A comprehensive framework has been developed to conduct a parametric study, with the intention of leveraging the outcomes from dynamic analysis, as inputs in the ANN. The framework explores the process of selecting earthquakes by utilizing seismic response spectrums and a nonlinear finite element (FE) model that introduces varied geometric properties, member sizes, and peak ground accelerations to derive eigenfrequencies, horizontal drift and base shear. Results indicate a satisfactory accuracy of the trained Artificial Neural Networks to predict the dynamic response.