Publications

Conference Paper

Mixed continuation methods for robust CFD Newton solvers

authors

J. L. Anibal, S. Seraj, A. Yildirim, and J. R. R. A. Martins

journal

AIAA SciTech Forum, 2023

doi

10.2514/6.2023-2296

PDF 
@inproceedings{Anibal2023a,
  author = {Anibal, Joshua L. and Seraj, Sabet and Yildirim, Anil and Martins, Joaquim R. R. A.},
  booktitle = {AIAA SciTech Forum},
  doi = {10.2514/6.2023-2296},
  keywords = {CFD, solver, continuation},
  month = jan,
  title = {Mixed continuation methods for robust {CFD} {Newton} solvers},
  year = {2023}
}
Google Scholar

Computational fluid dynamics (CFD) has become a vital tool for aerospace engineers. The ability to rapidly assess the performance of a design without a wind tunnel experiment or flight test has greatly increased the rate at which engineers can iterate on a design. Despite CFD’s widespread adoption, its automated use remains challenging. One challenge is the development of robust and high-performance solver algorithms for a broad range of problems. Globalized Newton’s methods offer a promising solution to this issue. However, selecting a suitable globalization strategy for this task requires careful analysis of possible candidates. In this work, we investigate the use of combinations of existing globalization strategies over a range of test cases. Specifically, we will compare the relative performance of combinations of pseudo-transient, boundary condition, and dissipation-based continuation. We find well tuned methods that use a mix of pseudo-transient and dissipation-based continuation to be the most robust for our test problems.