Earthquake engineering problems in parallel neuro environment

Abstract

The aim of the paper is to explore the application of Parallel Neuro Simulator for the generation of artificial earthquake. Parallel Neuro Simulator is a neural network code developed on PARAM 10000 using 'C' language and MPI library subroutines. In this study, two artificial neural network (ANN) models have been proposed to replace the auto-regressive moving average (ARMA) model. First ANN model substitutes the polynomial model that represents the relation of initial site information and coefficients of polynomial and the second ANN based model substitutes the estimated parameters of the ARMA model. Several Indian earthquake records have been used for present study on PARAM 10000. The variation in computational time with increasing number of processors has also been studied. © Springer-Verlag 2004.