Note: This is a continuing series on CFD validation. Please subscribe to the blog or check back on a regular basis for previous and subsequent installments.
A properly refined mesh is just one of the key factors in obtaining reliable design insight from FEA and CFD software tools. Meshing is a specialized skillset that combines idealization of CAD models and meshing techniques to obtain a mesh that is not only accurate, but economical as well.
There are a number of alternatives in developing a mesh, including “classic” manual refinement approaches and semi-automated routines. Many simulation software vendors also offer adaptive meshing, where the software evaluates results on the fly and then decides where it thinks the mesh should be improved. This technology will be put to the test by comparing it to the results obtained in Part 2 of this blog series.
The adaptive mesh produced by the software during the simulation run, will be compared to a default autosized mesh along with a mesh that was refined down to 25% of the default global size. These meshes are compared in Figure 1.
Figure 1: Default autosize, 25% autosize and adaptive mesh
When comparing the results, the adaptive meshing option clearly has the same benefit as the 25% mesh of improving the results from the relatively coarse default mesh. However, note that the adaptive mesh while accurate at the ceiling, increased the error down near the floor for the return of the recirculating air.
Figure 2: Vertical velocity profile @ X = 3 meters
The initial adaptive mesh was generated using the standard settings. However, there are advanced parameters available to further “tune” the meshing algorithm. For example, when some parameters are adjusted to improve the wall boundary layer (which impacts the y+ value, a topic that will be covered later), the results for the adaptive mesh are much more improved, as depicted in Figure 3.
Figure 3: Vertical velocity profile @ X = 3 meters
Adaptive meshing is very powerful technology that can help to automate the task of generating a suitable mesh. It is also very surgical in placing elements exactly where they are needed. However, that convenience does come at an expense in that it will take longer to complete a simulation. Also it is not necessarily 100% automated; as the above results show, there are situations where it could also benefit from manual tuning at the hands of an experienced user. As simulation consultants, we do leverage adaptive meshing for certain jobs, but also still rely on classic meshing techniques for a lot of our work.