Global sensitivity analysis for multivariate output model and dynamic models

Global sensitivity analysis has mainly been analyzed for scalar output and static models, though many mathematical and computational models used in engineering produce multivariate output that show some degree of correlation, and most physical systems are dynamic models. This paper focuses on global sensitivity analysis for multivariate output and dynamic models and a novel procedure is proposed to research the influence of inputs and model modes on the synthetic uncertainty of output. Introducing an additional variable to represent the variation of model modes which is viewed as model framework uncertainty, the variance decompositions of multivariate output and dynamic models are obtained and the significance of variance contributions is presented in detail. Two numerical examples and two practical models are employed to illustrate the validity and usefulness of the novel global sensitivity analysis approach.