Earthquake ground-motion duration estimation using general regression neural network

Document Type: Article


Department of Civil Engineering, University of Tabriz, Tabriz, Iran


Accurate prediction of earthquake duration could control seismic design of structures. In this paper, a new simple method was developed to estimate such important parameter by employing artificial neural networks (ANN) capability. A generalized regression neural network (GRNN) as a special class of RBF networks was implemented in this study to reduce the computation steps required for the searching process on sparse data sets. This network with quick-design capability does not need to impose a prescribed form for mapping of the observed data. The independent variables used in the predictive model of this study were earthquake magnitude, distance measure and site conditions. The designed models were trained using the 950 accelerograms recorded at Iranian plateau. The performance of proposed approach was compared with predicted results of feed forward back propagation networks. Analyses show that the designed GRNN performs well in estimating earthquake record duration and could be applied for prediction of common measures of earthquake ground-motion duration.


Main Subjects