Damping in buildings: its neural network model and AR model

The results of full scale measurements of damping as well as other research es on damping show that damping in buildings exhibits randomness and amplit ude dependent behaviour in the case of tall buildings subjected to dynamic loading. In this paper, based on full scale measurements of damping in a ta ll building, a time series analysis method (TSA) is employed to obtain the relationship between damping and vibration amplitude. Then, two models of d amping in a tall building, the artificial neural network (ANN) model and th e auto-regressive (AR) model, are established by employing ANN and AR metho ds, and used to predict the damping values at high amplitude level, which a re difficult to obtain from field measurements. In order to Set high accura cy, a genetic algorithm strategy is employed to aid in training the ANN. Co mparison analysis of the neural network model and the AR model of damping i s made, and the results are presented and discussed. (C) 2000 Elsevier Scie nce Ltd. All rights reserved.