Seismic fragility analysis under multidimensional limit state through Gaussian mixture model

Traditional seismic fragility analysis is generally proceeded based on lognormal distribution assumption, and most researchers adopt fixed threshold for limit state definition. However, these simplified strategies are often deviated from the reality. This paper presents a novel seismic fragility analysis method based on Gaussian mixture model under multidimensional performance limit state considering the threshold randomness without lognormal assumption. Multidimensional performance limit state function is used to measure the structural damage degree. Engineering demand parameters and thresholds are calculated based on the incremental dynamic analysis (IDA). The structural probability seismic demand model (PSDM) and probability seismic capacity model (PSCM) are fitted by Gaussian mixture model respectively instead of lognormal assumption. Monte Carlo simulation method is adopted to calculate failure probability, and the fragility curves are obtained. A reinforced concrete frame-shear wall structure is established as the research object. The results show that: thresholds have strong randomness, and it will increase with the increase of the damage degree. Ignoring the threshold randomness will lead to large deviation of the results. The fragility curves obtained by GMM are not coincide with those of the traditional lognormal assumption. With the increase of damage degree, the difference between GMM and lognormal assumption is getting larger. The results show that traditional method will lead to inaccurate fragility curves