A practical global sensitivity analysis methodology for multi-physics applications. This paper describes a global sensitivity analysis methodology for general multi-physics applications that are characterized by strong nonlinearities and interactions in their input-output relationships, expensive simulation runs, and large number of input parameters. We present a four-step approach consisting of (1) prescription of credible input ranges, (2) parameter screening, (3) construction of response surfaces, and (4) quantitative sensitivity analysis on the reduced set of parameters. Details of each step will be given using simple examples. Numerical results on real applications are available in another paper. Motivated by our experience with some large-scale multi-physics applications, we also propose enhancements to the various steps in the methodology for improving its robustness. The essential computational techniques targeted for this methodology have been implemented in a software package called PSUADE.
Keywords for this software
References in zbMATH (referenced in 3 articles , 1 standard article )
Showing results 1 to 3 of 3.
- Hadjidoukas, P.E.; Angelikopoulos, P.; Papadimitriou, C.; Koumoutsakos, P.: $\Pi$4U: a high performance computing framework for Bayesian uncertainty quantification of complex models (2015)
- Chen, Xiao; Ng, Brenda; Sun, Yunwei; Tong, Charles: A flexible uncertainty quantification method for linearly coupled multi-physics systems (2013)
- Tong, C.; Graziani, F.: A practical global sensitivity analysis methodology for multi-physics applications (2008)