Artificial Neural Network Application for Predicting Seismic Damage Index of Buildings in Malaysia

Abstract

An effective, convenient and reliable intelligent seismic evaluation system for buildings in Malaysia has been developed in this study by using Back-Propagation Artificial Neural Network (ANN) algorithm. A total of forty one buildings with 164 sets of input data spreading throughout Peninsular and East Malaysia were chosen and analyzed using IDARC-2D finite element software under seismic loading at peak ground accelerations of 0.05g, 0.10g, 0.15g and 0.20g respectively. Non-linear dynamic analysis was performed in order to obtain the damage index of each building. The ANN algorithm comprising 15 hidden neurons with 1 hidden layer outperformed other combinations in predicting the damage index of buildings with accuracy statistical value of 93% in testing phase as well as 75% in validation stage. From the results, the ANN system is suitable to be used for predicting the seismic behaviour of their buildings at any given time.