System reliability assessment of steel moment frames with different failure mechanisms using Bayesian Probability Network

Due to possibility of different seismic collapse scenarios as well as the complicated interaction of component ductility and redundancy in frame structures, system reliability has recently gained a great attention. This paper investigates system reliability of a one bay, two-story Special Steel Moment Resisting Frame (SSMRF) using combined Incremental Dynamic Analysis (IDA) and Bayesian Probability Network (BPN). Record to Record seismic response variability also known as inherent uncertainty is reflected in the chance node of Intensity Measure (IM) which is the spectral acceleration in first vibration mode. The parametric uncertainty represented by three chance nodes is related to the most significant parameters of nonlinear model. The randomness in IM extracted from seismic hazard curve and the uncertainty in model parameters are taken from experiments by others. A Failure Mode (FM) representing collapse is a sequential assembly of plastic hinges leading the whole structure to collapse. Different likely FM's were first recognized through a series of sensitivity pushover analyses. The conditional probability of each FM has been taken from intensive IDA conditioned on model chance nodes and an uninformative set of damage index in BPN. The collapse reliability index and mean annual rate of collapse as two conventional decision variables in performance based seismic design were determined inside the BPN. With the aid of BPN additional probabilistic processes were also conducted. Finally, updating as one of the significant advantages of a BPN was exemplified and influential parameters on system's response were recognized.