Seismic Performance Uncertainty Estimation via IDA with Progressive Accelerogram-Wise Latin Hypercube Sampling

Abstract

An algorithm is proposed for the rapid estimation of the influence of model parameter uncertainties on the seismic performance of structures using incremental dynamic analysis (IDA) and Monte Carlo simulation with Latin hypercube sampling. It builds upon existing methods that quantify the uncertainty for structural models with nondeterministic parameters by performing IDA with multiple ground motion records on each model realization out of a predetermined sample. However, their practical application is restricted due to (1) the inability to determine a priori the required number of samples and (2) the disproportionate increase of the number of analyses in realistic multiparameter models. To address these issues, two fundamental changes are incorporated. First, Latin hypercube sampling is applied progressively by starting with a small sample that is doubled successively until the desired accuracy is achieved. Second, parameter sampling is performed on a record-by-record basis rather than maintaining the same model over an entire record suite, thus expanding the model sample size without increasing the number of nonlinear dynamic analyses. Using strong-column and weak-column models of a steel moment-resisting frame, the algorithm is shown to possess excellent scalability, extending the original methodology to be applicable to large-scale models with hundreds of random variables.