SEISMIC CONFIDENCE LEVELS AND COLLAPSE CAPACITY ASSESSMENT OF STEEL MOMENT RESISTING FRAMES USING NEURAL NETWORKS

This paper employs neural network models to assess the seismic confidence levels at various performance levels, as well as the seismic collapse capacity of steel moment-resisting frame structures. Two types of shallow neural network models including back-propagation (BP) and radial basis (RB) models are utilized to evaluate the seismic responses. Both neural network models consist of a single hidden layer with a different number of neurons. The prediction accuracy of the trained neural network models is compared using two illustrative examples of ۶- and ۱۲-story steel moment-resisting frames. The obtained numerical results indicate that the BP model outperforms the RB model in predicting seismic responses.

Keywords:

seismic life cycle cost , performance-based design , nonlinear response history analysis , steel moment resisting frame