Global sensitivity analysis using a Gaussian Radial Basis Function metamodel

• RBF based sensitivity analysis method is proposed.
• Sobol' decomposition of Gaussian RBF metamodel is obtained.
• Sobol' indices of Gaussian RBF metamodel are derived based on the decomposition.
• The efficiency of proposed method is validated by some numerical examples.

Sensitivity analysis plays an important role in exploring the actual impact of adjustable parameters on response variables. Amongst the wide range of documented studies on sensitivity measures and analysis, Sobol' indices have received greater portion of attention due to the fact that they can provide accurate information for most models. In this paper, a novel analytical expression to compute the Sobol' indices is derived by introducing a method which uses the Gaussian Radial Basis Function to build metamodels of computationally expensive computer codes. Performance of the proposed method is validated against various analytical functions and also a structural simulation scenario. Results demonstrate that the proposed method is an efficient approach, requiring a computational cost of one to two orders of magnitude less when compared to the traditional Quasi Monte Carlo-based evaluation of Sobol' indices.