Mesh adaptation for the numerical observation of bi-stability in turbulent flows
Extreme and bi-stable events are phenomena that occur in many weather and climate systems. Their predictability is a major concern, as they are difficult to observe despite their potentially significant impact. Similarly, extreme rare events as well as bi-stable events have also been observed in laboratory experiments and in numerical simulations on simplified configurations. However, capturing such a phenomenon numerically remains challenging, since very long simulation time are required. To address this issue, HPC methods that are particularly efficient must be employed.
In the first part of my presentation, I will present one of these methods : mesh adaptation techniques. Mesh adaptation allows the mesh to be refined or coarsened locally where the flow requires it. However, most existing approaches rely on dimensional criteria, which require a new target value not known a priori for each case studied, limiting the potential computational gain. I will introduce a new non-dimensional meshing criterion, based on a cell-size Reynolds number, which enables the construction of near-optimal meshes for complex geometries without the need for parameter tuning. This criterion is evaluated on several academic and industrial test cases.
In the second part, I will apply this new criterion, together with recent developments done in the YALES2 code, to perform numerical simulation of configurations known to be subject to a bi-stable behaviour.
Recent numerical results showing bi-stability and extreme events for a NACA0012 airfoil will be presented. Bi-stable events at lower Reynolds numbers will also be shown for the Ahmed body.
