Abstract
In the seismic fragility assessment for geotechnical structures, the selection of input motion set for nonlinear dynamic finite element (FE) analyses has solely been based on the methods used in an incremental dynamic analysis (IDA), despite methods adopted in structural engineering for cloud analysis (CA), and multiple stripe analysis (MSA). This study investigates uncertainties in the seismic fragility curve of slopes arising from input motion sets used in nonlinear dynamic FE analysis to develop a probabilistic seismic demand model (PSDM). We consider a FE slope model and four sets of input motions based on CA (Set 1: 632 unscaled ground motion records), IDA (Set 2: random sampling with four sample sizes of 7, 14, 28, and 50 from Set 1, scaled to five PGA values, iterated 20 times), and MSA [Sets 3 and 4: different suites of ground motion records at five PGA levels spaced equally on linear and logarithmic scale, respectively, from Set 1, considering as many records as possible (maximum sample size) and 20 iterations of random sampling for three sample sizes of 3, 6, and 9 out of the maximum sample size]. Comparisons of the seismic fragility curves from Sets 2, 3, and 4 relative to the curves from Set 1 reveal that Set 4 (with a sample size of 9, involving 45 analyses) is more computationally efficient than Set 2 (with a sample size of 50, totaling 250 analyses) and would yield the curve stochastically closer to the one from Set 1. It is worthwhile to consider the input motion set based on the MSA with PGA levels evenly distributed on a logarithmic scale than the IDA-based set when compiling large numbers of ground motion records is limited.