A sensitivity analysis technique for uncertainty, risk and reliability analysis based on iterative quasi-Monte Carlo sampling

This paper provides an introduction to the use of variance decomposition methods when applied to structural reliability and uncertainty analyses based on a new iterative quasi-Monte Carlo (qMC) method. Over the past year, this new qMC sampling method has proven to provide a very accurate characterization of the uncertainty in the response of complex systems with a substantial reduction in the number of simulations that are required. From a theoretical point of view, the new iterative qMC methods will always converge faster than the conventional Monte Carlo and generally provides a more accurate uncertainty description (for the same number of simulations) than Latin hypercube sampling (LHS). The sensitivity analysis method developed to support this new uncertainty analysis methodology is based on the use of variance decomposition techniques very similar to that developed by M. McKay and I. Sobol'. Particular attention will be made to comparing sensitivity methods employed in the uncertainty analysis area with the sensitivity methods traditionally used in the structural reliability area.