Parallel Wavelet BP Neural Networks for Approximate Structural Dynamic Analysis

analysis of structures for earthquake induced loads is very expensive in terms of the computational burden. In this study, to reduce the computational effort a new neural system which is called parallel wavelet back propagation (PWBP) neural networks has been introduced. Training of PWBP is implemented in two phases. In the first phase, the input space is classified by using competitive neural networks. In the second phase, one distinct WBP neural network is trained for each class. Comparison the numerical results obtained by PWBP with the corresponding ones obtained by the single WBP neural network reveals the better performance generality of PWBP.