structured vol. mesh on complex geometry
I am currently creating a point-point matching multiblock structured mesh on a fairly complex aircraft. The AC includes v-tail with 3 engines. I know that hyperbolic grid generator is fairly unstable, especially for geometries with lots concave regions. However, I do need a viscous mesh and hyp. grid generator produces high quality mesh. Is it worth trying to create the vol. mesh using the hyperbolic grid generator or would be easier to draw the boundaries of the volume grids and use elliptical smoothing?
sorry about the vagueness of the question.
Without seeing your geometry it's hard to make definitive statements but I generally agree the hyperbolic extrusion is tough to use for complex geometries. I tend to take the second approach you suggest.
I also find there's a second reason for choosing not to use the hyperbolic extrusion. If I need to change the mesh dimensions or distribution for some reason (e.g. creating a family of grids for a refinement study), it's difficult to modify the extruded domains or blocks because the point spacing and distribution is a result of the extrusion. Specifically, the connectors will have a general distribution type with unconstrained spacing constraints and domains will have a point distribution that may or may not be attainable with the elliptic smoother.
This means if one really wants to use the hyperbolic extrusion, it's best to re-run the extrusion, using the re-extrude command, with the new parameters rather than explicitly modify connectors or domains that are the result of an extrusion. The only difficulty with this approach is that any topology you've subsequently created around the hyperbolic will be adjusted to adapt to the new location of extrusion front, the result of which may be undesirable. If the new location of the extrusion front is close to the original it might be OK but if it's drastically different the outer topology will require major changes.
My general rule of the thumb is I will use the hyperbolic extrusion if I can get it to create the entire mesh (near field and far field). Otherwise I create the all topology manually and assemble the domains and blocks subsequently. Now this may sound tedious my suggestion is to create your topology with the analytic curve tools (either db curves or connectors) and use simple curve types (straight 2-pt curves or Bezier curves with as few control points as needed). Once you get an eye for it I find it doesn't take very long. For more complex topology, I look around for other structured meshes on similar geometries to give me hints. Here's a link to an article I wrote about this approach (toward the bottom) a while back when I was employed at Pointwise:
If you need specific help or suggestions with your topology, feel free to ask - I have done a ton of structured meshing in Pointwise.
Thank you for your reply. As I try out various things, I am finding out similar thing. Do you typically layout your topology using the database curves first? Is there advantage of laying out the topology using the database curve instead using the connectors?
The only advantage to using db curves, as I mention in the article, is once you start assembling the mesh should you modify your connectors so they are no longer analytical (e.g. through smoothing) but subsequently want to revert back to part or all of the original topology simply select the db curve(s) and click Con on DB.
Otherwise curves drawing for db curves or connectors is identical.
|All times are GMT -4. The time now is 16:25.|