Moment Independent Sensitivity Analysis: H-PCFE-Based Approach

This work presents a novel approach for moment independent sensitivity analysis. The proposed approach integrates hybrid polynomial correlated function expansion (H-PCFE) into the framework of moment independent sensitivity analysis. H-PCFE is a novel surrogate model formulated by coupling polynomial correlated function expansion (PCFE) with kriging. H-PCFE performs a bilevel approximation: firstly on the global scale via PCFE and secondly on the local scale via kriging. It is argued that by incorporating H-PCFE into the framework of moment independent sensitivity analysis, the computational cost can be significantly alleviated. Performance of the proposed approach has been illustrated with three numerical examples. For the first two examples, results obtained have been compared with other popular techniques for moment independent sensitivity analysis. It is observed that the proposed approach is accurate as well as efficient. In the third example, the proposed approach has been used for sensitivity analysis of a transmission tower. The sensitivity indexes of the variables, with respect to the maximum lateral displacement, have been computed. It is observed that elastic modulus and loads in the direction of deflection are the most sensitive parameters. (C) 2016 American Society of Civil Engineers.