Prediction of peak ground acceleration for Himalayan region using artificial neural network and genetic algorithm

Abstract

Level of ground shaking in terms of peak ground acceleration (PGA) is used to carry out seismic hazard analysis of a region and has a great importance for designing earthquake resistant structures. Thus, prediction of PGA for an earthquake is very important. This paper presents a study to predict PGA of strong ground motion using the earthquake data recorded in rock sites of Himalayan region, India by using artificial neural network (ANN) and genetic algorithm (GA). In the present study, magnitude and hypocentral distance of an earthquake event has been used as input parameters and PGA recorded at different hypocentral distances have been used as output parameter for construction of ANN and GA model. Ninety percent data have been used for training and 10% testing purposes for both the ANN and GA model. Chi-square test results also reveals the successful application of ANN and GA for predicting PGA data site even with the lesser available data of a study region. The comparison of predicted PGA values from current study holds quite well with the other available developed attenuation relationships for the study region. The present study depicts the successful application of ANN and GA for predicting PGA data site even with the lesser available data of a study region.