21st EANN 2020, 5 -7 June 2020, Greece

Accomplished Reliability Level for Seismic Structural Damage Prediction using Artificial Neural Networks

Magdalini Tyrtaiou, Antonios Papaleonidas, Anaxagoras Elenas, Lazaros Iliadis

Abstract:

  This research aims to determine the optimal Multi-Layer Feed-Forward Artificial Neural Network (MLFF) capable of accurately estimating the level of seismic damage on buildings, by considering a set of Seismic Intensity Parameters (SIP). Twenty SIP (well established and highly correlated to the structural damage) were utilized. Their corresponding values were calculated for a set of seismic signals. Various combinations of at least five seismic features were performed for the development of the input dataset. A vast number of Artificial Neural Networks (ANNs) were developed and tested. Their output was the level of earthquake Damage on a Reinforced Concrete Frame construction (DRCF) as it is expressed by the Park and Ang overall damage index. The potential contribution of nine distinct Machine Learning functions towards the development of the most robust ANN was also investigated. The results confirm that MLFF networks can provide an accurate estimation of the structural damage caused by an earthquake excitation. Hence, they can be considered as a reliable Computational Intelligence approach for the determination of structures’ seismic vulnerability.  

*** Title, author list and abstract as seen in the Camera-Ready version of the paper that was provided to Conference Committee. Small changes that may have occurred during processing by Springer may not appear in this window.

*** Title, author list and abstract as seen in the Camera-Ready version of the paper that was provided to Conference Committee. Small changes that may have occurred during processing by Springer may not appear in this window.