Abstract
The method of Elementary Effects (EE) is a parameter screening type sensitivity analysis technique that combines advantages of inexpensive one-at-a time methods and expensive variance decomposition based global Sensitivity Analysis (SA) techniques. Most of the sampling strategies for EE either use random sampling or maximize sample spread through oversampling. The Sampling for Uniformity (SU) is the only available strategy that combines the principle of sample spread with the principle of uniformity. In this work, we proposed modifications to SU (eSU) to further improve sample uniformity. Performance of eSU was compared to that of SU based on uniformity, sample spread, sample generation time, and screening efficiency. Importance of the concept of uniformity was strengthened as eSU outperformed SU across all evaluation criteria. Further, it was found that eSU does not need oversampling and can result in better screening with relatively few trajectories indicating significantly reduced requirement on computational resources.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call