Deterioration parameters that are commonly used to simulate nonlinear behavior of steel components were mainly calibrated based on the results from experiments on steel beams. Recently, the state of knowledge for deterioration behavior of steel columns has been improved by experimental and analytical studies on the wide-flange steel columns. These deteriorating characteristics are introduced as regression relationships in which the associated uncertainties are represented by the coefficient of variation (COV). Accounting for these uncertainties in estimating collapse fragility curves through the incremental dynamic analysis (IDA) and simulation-based reliability methods is impractical due to the large amount of required computational effort. In this study, two main goals are pursued. The first goal is comprehensive evaluation of the main, interaction, and quadratic effects of the modeling random variables on the collapse capacity of steel structures. The second goal is to propose an efficient approach to create response surface (RS), which in combination with the Monte Carlo (MC) sampling method will be used to incorporate modeling uncertainties into the collapse fragility. This efficiency will be achieved by employing screening design techniques to reduce the amount of analysis required to create a quadratic RS and also endurance time (ET) analysis as an efficient alternative nonlinear dynamic analysis method with less computational time to estimate the structural responses. In order to develop a reliable probabilistic model, the Bayesian model inference approach is applied to account for the uncertainties in the created model. The proposed procedure is performed with both IDA and ET methods on a prototype 5-story steel frame. Results indicate that collapse capacity is highly influenced by the strength modeling variables of beam as well as the ultimate rotation capacity of column components. In addition, ET method by a considerable reduction in the computational costs provides comparable responses with IDA in a probabilistic framework.
Read full abstract