peak ground acceleration prediction for critical aftershockes

This paper proposes a methodology using learning abilities of artificial neural networks in order to predict the peak ground acceleration of critical aftershocks based on the features of successive earthquakes. at first, a set of recorded consecutive earthquakes which has been contained critical main shocks and aftershocks is selected based on effective peak acceleration (epa) from peer and usgs. in the following, the idealized multilayer artificial neural networks were designed and trained to estimate the peak ground acceleration of critical aftershocks. in this regard, two-layer feed-forward (mlff) neural networks are used. the results indicate that the networks have learned to generalize the unseen information very well and reflect good precision in the simulation of the peak ground acceleration of critical aftershocks.