Effect of modeling uncertainty on multi-limit state performance of controlled rocking steel braced frames

Abstract

Quantifying and propagating uncertainties in structural response is critical to robust performance-based seismic assessments. This paper evaluates the influence of structural model parameter uncertainty on the seismic response and limit state assessment of controlled rocking steel braced frames (CRSBFs). The dead load on the rocking frame, fuse yield strength, initial post-tension (PT) force and yield strength of the braced frame elements are the random variables considered. Realizations of 3-, 6- and 9-story CRSBF nonlinear structural models are constructed using the random variable instances generated by Latin Hypercube Sampling (LHS). The structural model realizations are subjected to multiple stripe analyses using six sets of hazard-consistent ground motions. Three limit states are considered including immediate occupancy, repairability and collapse prevention. The probability distribution of maximum transient and residual story drift ratios, PT strain and fuse shear deformation are also established. These four engineering demand parameters also serve as the basis for the limit state assessments. The effect of modeling uncertainty is examined by comparing the dispersion of the responses and limit state performance from “mean models” (i.e., models constructed based on mean values of random variables) with those obtained from the LHS-generated models. The results showed that model uncertainty had the greatest impact on the response and performance of the 9-story CRSBF.