In view of the current calibration problems in a lack of unified calibration method and difficulty to sample expansion, the distribution parameter calibration method for probability damage model and sample expansion method based on Bayesian statistical method were proposed. The typical damage model was considered, and damage indexes of each limit state were calibrated by hysteretic test results of conventional reinforced concrete beam-column components. The impact of the damage model on the performance evaluation was examined from single component damage to overall structural component loss in four RC frame structures with different layers in 8 intensity region. The results show that the sample expansion method proposed can balance the prior information and sampling information, and the calibration process proposed can facilitate the expansion of the following sample. The prominent damage state can be recognized by the calibrated damage models; the displacement damage model is suggested under small and moderate earthquakes as its convenience for computation and the earthquake average losses are basically the same, and the Park-Ang damage model is suggested under rare earthquakes for prudence.