Monolithic adjoint-based shape sensitivities from topology-preserving mesh perturbations скачать в хорошем качестве

Monolithic adjoint-based shape sensitivities from topology-preserving mesh perturbations

A SlideCast of the presentation given by Grégoire Varillon at SoTiC 2024 at NTNU in Trondheim, NO. DOI 10.13140/RG.2.2.25618.52160 Here's the abstract: This work introduces adjoint-based shape sensitivities from the Linearized Reactive Flow (LRF), a monolithic approach to the linear stability of reactive flows. This enables the optimization of all geometrical parameters of a burner, including those affecting the flame dynamics and flame shape. Shape optimization based on LRF builds on standard shape optimization frameworks based on Helmholtz solvers, which are limited to optimizing geometrical parameters that do not affect the flame (e.g., plenum shape). This owes to the use of an external flame model in which the flame dynamics is frozen. This limitation is overcome in the present work by the use of LRF equations. In addition, this work demonstrates a topology-preserving perturbation method to compute adjoint-based shape derivatives without the need to derive adjoint boundary conditions and use the Hadamard formalism for shape derivatives. The method is incrementally verified against established shape derivatives stemming from continuous adjoints, with test cases ranging from an acoustic duct to a slit flame that exhibits thermoacoustic instability. The proposed method is in quantitative agreement with the reference results. We could verify the application of the perturbation method to a reactive flow by computing shape sensitivities with respect to geometry parameters that change the flame dynamics. This work paves the way to a monolithic and robust shape-optimization framework based on the adjoint.