Скачать или смотреть IDWSDS 2024 - S30: Using Functional Data Analysis for Surrogate Model Development

JHU/APL has developed a physics-based modeling and simulation software suite that simulates the performance of a system. That physics-based model has dozens of input and output variables and can take several minutes to run for each test case. To support risk assessment, concept studies, and system trades, development of a surrogate model would simplify model usage, reduce run time, and allow opportunity to share with other organizations. An effort was conducted to develop a surrogate model of that physics-based model by 1) exploring a robust down selection of critical input parameters to cover sampling of the analysis space and 2) leveraging functional data analysis to model two transient output curves to those critical input parameters. The goal of this brief will be to showcase the analysis process taken to develop such a surrogate model.