%0 Journal Article
%T Estimating Structural Collapse Responses Considering Modeling Uncertainties using Artificial Neural Networks and Response Surface Method
%J Amirkabir Journal of Civil Engineering
%I Amirkabir University of Technology
%Z 2588-297X
%A Bayari, Mohammad Amin
%A Izadi Zaman Abadi, Esmaeel
%A Shabakhty, Naser
%D 2021
%\ 08/23/2021
%V 53
%N 6
%P 2249-2276
%! Estimating Structural Collapse Responses Considering Modeling Uncertainties using Artificial Neural Networks and Response Surface Method
%K Uncertainty analysis
%K Incremental dynamic analysis
%K Structural Collapse Responses
%K Response Surface Method
%K artificial neural network
%R 10.22060/ceej.2020.17312.6539
%X This research investigates the collapse responses of a concrete moment frame considering modeling uncertainties. These modeling uncertainties are considered for evaluating a collapse response related to the modified Ibarra-Krawinkler moment-rotation parameters for beam and column elements of a given structure. To analyze these uncertainties, the correlations between the model parameters in one component and between two structural components were considered. Latin Hypercube Sampling (LHS) method was employed to produce independent random variables. Moreover, Cholesky decomposition was adopted to produce correlated random variables. Performing 281 simulations for the uncertainties involved considering their inter-correlations, incremental dynamic analysis (IDA) was done using 44 far-field accelerograms to determine structural collapse responses. Collapse responses of each simulation, including mean collapse capacity, mean collapse drift and mean annual frequency, were obtained. Then, the collapse responses were predicted using the response surface method and artificial neural network. The results show that the Correlation coefficients (R) between the target data resulted from incremental dynamic analysis (IDA), output data resulted from response surface method (RSM), and artificial neural network (ANN) were obtained for the collapse responses above 0.98. The maximum prediction errors for mean collapse capacity and mean collapse drift are less than 5% and for mean annual frequency less than 10% under the response surface method (RSM) and artificial neural network (ANN).
%U https://ceej.aut.ac.ir/article_3795_f87c87b33a2dcd4feff390128fe2b6fc.pdf