ESTIMATION OF PEAK GROUND ACCELERATION, VELOCITY AND DISPLACEMENT USING ARTIFICIAL NEURAL NETWORK

Abstract

Dynamic analyses require many seismic data such as design acceleration and spectra of the site and earthquake records. This paper aims to predict peak ground acceleration, speed, and displacement using Artificial Neural Network (ANN). The proposed method of estimating PGA, PGV and PGD from earthquake parameters via a neural network has been applied to the database of the Next Generation Attenuation (NGA) project which includes 1950 recorded earthquake accelerograms classified into three sets according to the faulting mechanism for the training of the neural networks. The number of earthquake accelerograms for each mechanism (the strike-slip, reverse and reverse-oblique) is 650. Earthquake records have moment magnitudes between 4.5 and 7.9 Richter, distances from the recording site to epicentre ranging from 2.3 to 195 kilometres, hypocenter depths between 0.5 to 29 kilometres, and average shear-wave velocities in the top 30m ranging from 116 to 2016 m/sec. In the selected learning algorithm, the average speed of the shear wave in the top 30 metres, focal depth, magnitude and distance to source are the input variables, and Peak Ground Acceleration, Velocity and Displacement (PGA, PGV and PGD) values are used as output. Close match between the predicted values of the deployed method with the observed values and its ability to reduce or even eliminate the uncertainties in the attenuation relationships show that this method can be used as a reliable method for predicting the main parameters of strong ground motions. The results indicate successful performance for the artificial neural network algorithm in predicting the expected results.

Keywords

• earthquake parameters
• neural network
• attenuation relationship
• peak ground acceleration
• average shear-wave velocities