Reduced-Order and Multi-Fidelity Approaches with Statistical and Modeling Uncertainties for Naval Applications
Abstract
Naval applications have increasingly relied on reduced-order models (ROMs) operating at different levels of fidelity. For example, ship motions have been modeled by ODE-like systems at the lowest level of fidelity, to potential-flow solvers at higher, engineering-level fidelity, to CFD solvers approaching highest levels of fidelity. Other examples include modeling of loads, maneuvering, usingdata from free roll decay tests, and so on.Advantages of using ROMs are several, including computational speed, physical understanding, and others. But using ROMs also brings numerous questions and challenges. How does one quantify modeling uncertainty for ROMs, that is, the error of using potentially useful but strictly speaking inaccurate ROMs? How does one integrate this with and quantify statistical uncertainty associated with finiteness of data? How does one work with the data and calibrate the models across the different levels of fidelity? Developing physics-informed ROMs in the first place and supplying them with data-fitting procedures are also at the center of these approaches.The main goal of the project is to advance reduced-order and multi-fidelity approaches with statistical and modeling uncertainty quantifications (UQs) for several naval applications, concerning motions in waves, including maneuvering and wave loads. This is envisioned along three interconnected directions: Development of statistical ROMs for loads, including submerged deck effect, integration of wave and impact loads, peaks formulation, local impact loads, and data-fitting procedures with statistical UQ; Modeling UQ in fitting statistical ROMs and regression-based ROMs, and long memory effects on statistical UQ for ship processes at non-zero speed; Development and validation of multi-fidelity approaches for extremes, including multi-dimensional settings and the use of deep learning methods.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jan 12, 2023
- Source ID
- N000142312176
Entities
People
- Vladas Pipiras
Organizations
- Office of Naval Research
- United States Navy
- University of North Carolina at Chapel Hill