Feature Extraction from Simulations and Experiments: Preliminary Results Using a Fluid Mix Problem

Kamath, C ; Nguyen, T

Lawrence Livermore National Laboratory

关键词: 99 General And Miscellaneous//Mathematics, Computing, And Information Science;    99 General And Miscellaneous//Mathematics, Computing, And Information Science;    Computerized Simulation;    Approximations;    Simulation;   

DOI  :  10.2172/15014706
RP-ID  :  UCRL-TR-208853
RP-ID  :  W-7405-ENG-48
RP-ID  :  15014706

美国|英语

来源: UNT Digital Library

【 摘 要 】

Code validation, or comparing the output of computer simulations to experiments, is necessary to determine which simulation is a better approximation to an experiment. It can also be used to determine how the input parameters in a simulation can be modified to yield output that is closer to the experiment. In this report, we discuss our experiences in the use of image processing techniques for extracting features from 2-D simulations and experiments. These features can be used in comparing the output of simulations to experiments, or to other simulations. We first describe the problem domain and the data. We next explain the need for cleaning or denoising the experimental data and discuss the performance of different techniques. Finally, we discuss the features of interest and describe how they can be extracted from the data. The focus in this report is on extracting features from experimental and simulation data for the purpose of code validation; the actual interpretation of these features and their use in code validation is left to the domain experts.