It is well known that traditional compressible flow solvers have difficulties simulating small Mach number flows, which is the reason why low Mach preconditioning has been developed. Although this technique allows for a fast convergence towards an accurate solution, it does suffer from robustness issues. In other words, the use of inaccurate initial conditions that are not a solution of the governing equations or even strong transient changes in flow conditions quite often lead to divergence. Several different criteria are employed in the present study to try and understand the origin of such a lack of robustness. It is traced to a strong sensitivity of the preconditioning and eigenvector matrixes to numerical perturbations. When low Mach preconditioning is removed however, this strong sensitivity disappears. This led to the development of a delayed preconditioning strategy that is able to greatly improve robustness without significantly delaying convergence.
展开▼
