Digital Engineering the Test and Modeling Process

From Moore Analytics and Dynamics Laboratory

Overview

Funded by AFOSR Young Investigator Award

The proposed research focuses on data-driven and deep learning approaches for autonomizing the validation and updating of digital models using Test and Evaluation (T&E) data. The first part of this research will create novel overlapping neural networks that leverage the principle of time reversibility to autonomously repair T&E data with missing data segments. The second portion will produce advanced mathematical techniques for infusing physics into autoencoder neural networks for extracting corresponding universal representations from both test and model results, facilitating the comparison of similar but disparate datasets. The third part will introduce and deploy new generator-discriminator-translator networks by leveraging the power of generative adversarial networks to autonomously update digital model parameters using T&E data. The new deep learning frameworks will be employed on data taken from computer-generated signals, numerical simulations, and experimental measurements.

Objective 1

Objective 2

Objective 3