Global sensitivity analysis in high dimensions with partial least squares-driven PCEs

We develop an efficient method for the computation of variance-based sensitivity indices using a recently introduced latent-variable-based polynomial chaos expansion, which is particularly suitable for high dimensional problems. By back-transforming the surrogate from its latent variable space-basis to the original input variable space-basis, we derive analytical expressions for these sensitivities that only depend on the model coefficients. Thus, once the surrogate model is built, the variance-based sensitivities can be computed at negligible computational cost as no additional sampling is required. The accuracy of the method is demonstrated with a numerical experiment of an elastic truss.