Improved Rapid Assessment of Earthquake Hazard Safety of Structures via Artificial Neural Networks

Abstract

The vulnerability of structures mainly depends on the structural resistance system of buildings to earthquake. It is unlikely that all existing buildings can be inspected in detail. Therefore, rapid methods for evaluating buildings have been developed over the last decades. This paper investigates the earthquake susceptibility through the combination of buildings’ geometrical attributes that affect the vulnerability of building and can be used to obtain an optimal prediction of the damage state of reinforced concrete (RC) buildings using artificial neural networks (ANNs). In this regard, a multi-layer perceptron (MLP) network has been trained and optimized using a database of 145 damaged buildings from the Haiti earthquake. The results demonstrate the practicability and effectiveness of the selected ANNs approach to classify actual structural damage that can be used as a preliminary assessment procedure to recognize vulnerable buildings.