An intelligent neural system for predicting structural response subject to earthquakes

An efficient method is introduced to predict the time history responses of structures subject to earthquakes employing neural network techniques. In order to achieve this purpose, a new intelligent neural system (INS) is designed by combining competitive and radial basis function (RBF) neural networks. In the INS the input space is classified by a competitive neural network (CNN) based on natural frequencies of the structures. Afterward an RBF network is assigned to each class and is trained by using the data located in the class. Results of illustrative examples demonstrate high performance and computational advantages of INS comparing with the single RBF network.