Improved seismic fragility analysis method using BOX-COX regression and Monte Carlo Sampling technique

Abstract

The cloud method is one of the main tools for calculating analytical fragility curves. However, assumptions of this method, viz., linearity, normality and homoscedasticity, which are aimed at simplifying calculation process, are usually not consistent with actual results. To solve the problem, a new fragility analysis method is proposed by combining the BOX-COX regression and the Monte Carlo Sampling technique. The proposed method does not increase the times of the nonlinear time history analysis, and also need not restrict to the three basic assumptions of the cloud method. A three-span concrete continuous girder bridge is taken as an example to verify the validity of the method developed. Subsequently, the results obtained from the proposed method are compared with those of cloud method employing determination coefficient (R 2), Kernel Density Estimation curve, and rank-correlation coefficient. Results indicate that BOX-COX regression can improve the linearity, normality and homoskedasticity of probabilistic seismic demand model (PSDM), ensuring the accuracy of fragility analysis. The fragility curves derived by cloud method will lead to errors when the basic assumptions are not satisfied, and the errors increase with damage state increasing.