Evaluation of effect of soil characteristics on the seismic amplification factor using the neural network and reliability concept

Abstract

Peak ground acceleration (PGA) is one of the most important parameters in structural seismic design. Since many variables such as soil characteristics and attenuation relations are probabilistic parameters, the PGA cannot be determined using deterministic approaches. In this paper, by using the concept of reliability, effects of soil characteristics on PGA have been investigated for Kermanshah (Iran) as the case study. Several deterministic analyses were performed using the PLAXIS software, and the amplification factor of soil was obtained. From the obtained results an artificial neural network (ANN) was trained according to the input–output pairs and these emulators were used as an alternative for PLAXIS software. As the next step for effective parameters on PGA, a great number of input data according to the probability density function, average, and standard deviation were generated. According to these data, PGA was determined on the ground by using ANN, and reliability was calculated according to Monte Carlo simulation. In fact, reliability was calculated by comparing the PGA in deterministic and probabilistic approach. Finally, sensitivity analyses for the effect of soil characteristics on PGA were performed.