An efficient entropy-based global sensitivity analysis of bridge seismic demand based on a novel three-point-estimate method

Global sensitivity analysis is important for quantifying uncertainties in the design of civil infrastructure, which is taken as a crucial step towards structural reliability or fragility. Most global sensitivity analysis uses Monte Carlo simulation, but that has high computation demands since many nonlinear time-history analyses are required to obtain the unconditional and conditional probability density functions. To address this limitation, a more efficient entropy-based global sensitivity analysis technique is proposed and tested for its ability to rank the importance of random parameters. A bridge under seismic loading serves as an example of its use. The unconditional and conditional probability of bridge seismic demands can be obtained by reconstructing the probability density function through a parameterized probability distribution based on the proposed new three-point-estimate method. Two examples show that the proposed method can effectively rank the importance of uncertain parameters. One uses a linear polynomial and the Ishigami function as mathematical test cases. The other is a finite element model of a continuous girder bridge. The entropy-based global importance of the uncertain material and structural parameters are obtained with only a few nonlinear time-history analyses. The results are compared with those from Monte Carlo simulation.