PREDICTION OF NONLINEAR INTER-STORY DRIFTS IN STEEL MOMENT FRAMES UNDER SEISMIC LOADING USING CASCADE FORWARD NEURAL NETWORKS

This paper aims to predict the maximum inter-story drift ratios of steel moment-resisting frame (MRF) structures under seismic loading, corresponding to different performance levels, using cascade-forward back-propagation (CFBP) neural network models. To this end, CFBP networks with varying numbers of hidden layer neurons are trained on nonlinear time-history analysis results of ۶- and ۱۲-story planar steel MRFs subjected to a suite of earthquake ground motions. The predictive performance of the trained models is systematically compared. Numerical results demonstrate that CFBP networks with ۱۵ neurons in the hidden layer consistently outperform other network architectures, yielding more accurate predictions of the maximum inter-story drift ratios at each seismic performance level for both frame heights. These findings highlight the potential of moderately sized CFBP networks as efficient surrogates for nonlinear dynamic analysis in performance-based seismic assessment.

کلیدواژه ها:

nonlinear time history analysis ، steel moment resisting frame ، performance level ، cascade-forward back-propagation ، neural network