Output error estimation using super-resolution neural networks
T.P. Hunter (TU Delft - Aerospace Engineering)
SJ Hulshoff – Mentor (TU Delft - Aerodynamics)
More Info
expand_more
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.
Abstract
Numerical simulations and optimisation methods, such as mesh adaptation, rely on the accurate and inexpensive use of error estimation methods. Output-error estimation is the most accurate method; however, it relies on the use of approximations in order to be implemented in practice. The proposed method in this thesis relies on the use of super-resolution neural networks to reconstruct the fine adjoint solution from a computed coarse adjoint solution. The proposed method is compared to reference error estimators on an unsteady Burgers’ equation using the method of manufactured solutions, as well as a lid-driven cavity flow. For both of these test cases, it was shown that super-resolution neural networks were able to reconstruct the fine adjoint solution and provide robust and inexpensive output-error estimates at the cost of lower accuracy.
Nonetheless, the accurate estimation of the error indicators gives great confidence in the proposed method’s ability to perform similarly to the adjoint-weighted residual output-error estimate with a mesh adaptation procedure. A cost metric for the computational overhead of the output-error estimate is proposed. This highlights the superior performance of the lower up-scale ratio super-resolution neural networks due to their higher accuracy and lower computational cost than those with higher up-scaling factors.