USING BAYESIAN STATISTICS TO EVALUATE THE ACCURACEY OF PROBABILISTIC SEISMIC DEMAND MODELS

Probabilistic Seismic Demand Model (PSDM) is a mathematical expression relating the structural demand at the component level to the demand at system level by using an Intensity Measure (IM) parameter. Naturally, the seismic demand is estimated accurately when the most important component, i.e. PSDM, is selected carefully. This selection must be based on several inherent properties such practicality, sufficiency, effectiveness and efficiency. The purpose of this research is to evaluate the accuracy of different PSDM for steel moment-resisting frames by using Bayesian statistics, which is a strong tool to simultaneous modelling of uncertainties and randomness. Generally, the results show that there is no efficient and sufficient PSDM with single IM parameter to cover all structural height and the accuracy of these models is a functionof the number of stories. Since it is not possible to use one combined IM which includes twospectral values because of the difficulties in producing a multiple-parameters seismic hazard curve, it is recommended to use a model with multiple IMs which is believed to be the best model and covers all structural heights, i.e. a model which is the linear combination of spectral accelerations of the first and second periods of vibration