Probabilistic Sensitivity Analysis of Structures: Methods

Sensitivity gradient with respect to various design parameters characterizes the structural response. The conventional deterministic sensitivity analysis is insufficient to provide complete information regarding structural response. Here the applicability of the NE-MCS technique within the frame work of SFEM has been assessed in terms of numerical accuracy and time efficiency for the computation of response sensitivity when design parameters are random. The covariance matrix has been decomposed using Cholesky decomposition for digital simulation of random design parameters that modeled as homogeneous Gaussian field. Finally the stochastic stiffness matrix is expanded deploying the NE-MCS technique to derive at stochastic response sensitivity. For computation of sensitivity of dynamic responses, complex dynamic stiffness matrix has been expanded through the NE-MCS series and the analysis is performed in frequency domain. The direct simulation and first-order perturbation based sensitivity analysis are also presented for comparative study.