Font Size: 

Global sensitivity indices based on variance can effectively apportion the output uncertainty to the inputs. How to efficiently and accurately perform the global sensitivity analysis is of great concern for researchers. In this work, the employment of sparse grid integration to the estimate of global sensitivity indices is discussed. The new method can be used for sensitivity analysis of the structural models involving independent variables or correlated variables, and can further decompose the variance contribution in the correlation cases. Advantage of the sparse grid integration in estimating integrals is well inherited by the new method, to ensure the accuracy while keeping the computational burden controllable. Numerical and engineering examples have been studied to test the applicability of the proposed method.

numerical methods; function approximation