An Intelligent Framework based on Parametric Reduced-Order Model Database for Uncertainty Quantification in Aeroelasticity

The goal of the proposed effort is to develop an intelligent framework based on parametric reduced-order models (ROM) for uncertainty quantification (UQ) in aeroelasticity (AE). In Phase 1, the framework for automated ROM construction, adaptive sampling, and UQ was built. Feasibility of utilizing eigenvalues obtained non-intrusively from AE ROM as a stability measure was successfully validated. Stochastic modeling techniques based on neural network and Kriging were also developed to establish the relationship between uncertain AE inputs and outputs. For the first time, adaptive sampling-guided AE ROM generation and UQ with NASArsquo;s FUN3D was developed in a broad flight envelope and demonstrated to accurately capture flutter boundary in a computationally conscious manner. In Phase 2, the framework will be improved in terms of inclusion of structural parameters, ROM adaptation, state consistent ROM, and multifidelity data for accelerated ROM generation. The stochastic model engine will be extended to polynomial chaos expansion. FEM models will be also incorporated within the framework to accommodate varying structural parameters. The capabilities will be provided as modular software environment for integration into NASA workflow for technology transition. The software will be extensively validated and demonstrated for automated AE ROM and UQ development using vehicles of NASA interest.