AUTOMATIC RANKING OF DESIGN PARAMETER SIGNIFICANCE FOR FAST AND ACCURATE CAE-BASED DESIGN SPACE EXPLORATION USING PARAMETER SENSITIVITY FEEDBACK

摘要

A computer-implemented method for ranking design parameter significance includes a computer receiving an input dataset representative of a physical object. This input dataset includes a baseline parameters and associated probabilities. The computer also receives performance requirements. For each respective baseline parameter, the computer performs an analysis process. During this analysis process, a range of parameter values are selected for the respective baseline parameter based on its corresponding probability distribution. The range of parameter values are segmented into parameter subsets and multiple instances of a simulation are executed using the performance requirements to yield snapshots. A Proper Orthogonal Decomposition (POD) basis is derived using the snapshots. A sensitivity analysis is performed based on the POD basis to yield a sensitivity measurement representative of an effect of variation of the respective parameter on the performance requirements. The computer may then generate a ranking of the baseline parameters according to their corresponding sensitivity measurements.