Pure unstructured hexahedra in Pointwise

wernerhe · February 3, 2020, 12:16

Good morning CFDers,

I'm a software developer working on some high-order in house-code for moving and deforming grids. Our base testing case will be an infinite span NACA-0012 airfoil at high Reynolds number (Re = 12,000) and we will be pitching the airfoil up to a static AOA and then holding it there for a specified number of chord times. We will compare with results from our water tunnel before testing more complex geometry variants and eventually be moving to marine propellers.

My code requires me to have a purely hexahedral grid, and the complexity of future geometry will require me to use a combination of structured hexahedra for my airfoil surface mesh and boundary layer, and unstructured hexahedra for the volume mesh (block) to reduce cell count and avoid high aspect ratio cells in the far-field.

My problem is this. I am able to generate a pure unstructured hexa mesh for low Re simulations using ANSYS Fluent and multi-zone method, however at the target Re our flow becomes turbulent and I require a structured boundary layer mesh with first layer height 3e-03 normalized by chord length to achieve a y+ of 1.0. ANSYS Fluent refuses to generate an inflation layer on my complex geometry cases. I know that Pointwise is capable of handling the inflation, however much of my research has revealed that Pointwise is not capable of generating pure unstructured hexahedra or quads automatically. I'm getting conflicting opinions from my research group and I think some of what I'm trying to do is being lost in the language barrier (I'm the only native English speaker here).

Does anybody have any experience with manipulating the settings in PW to achieve pure unstructured hexahedra around an airfoil, or would you be able to point me to some reference materials I can read and teach myself? For my complex geometry, I might be able to generate pure unstructured quads along a base NACA 0012 profile and extrude along a path or apply equation based deformation to achieve my geometry, but I have no experience with this method.

RcktMan77 · February 27, 2020, 14:12

In Pointwise you can go about this several ways. Simply create a structured mesh using the structured meshing tools in Pointwise and export the grid into an unstructured grid format. The grid is retained, you just lose the structured i-, j-, k- connectivity information, which gets replaced with cell types, associated nodes, and neighbors (typical unstructured connectivity information).

There is also an algebraic extrusion capability in Pointwise that will allow you to directly extrude blocks of unstructured hexahedra from a quad surface. Additionally, you could extrude a boundary layer of hexahedra using Pointwise's T-Rex advancing layer technique around your geometry, and then stitch that together with a farfield consisting of voxels. It just depends on which approach is most suitable for your problem, but the capability certainly exists within Pointwise. You could even elevate the elements of your grid in Pointwise directly.

wernerhe · May 26, 2020, 15:28

Hi,

I ended up tabling this problem to work on a few other issues that were more time sensitive, and now I am revisiting the issue of unstructured hexahedra in Pointwise.

The issue isn't a mesh file format and solver compatibility problem (though it is a good suggestion). I've already made several structured grids for both simple and complex geometry cases and the cell counts are simply unacceptable when I add in the necessary grid refinement zones to capture vortex formation and convection in the wake. I unfortunately cannot share the complex wing geometry I am working on with you, but what I'm after is extremely high resolution for about 4 chords above the suction surface of the airfoil and 4 chords down stream of the airfoil, with a structured boundary layer profile. The mesh should have very low resolution outside of the boundary layer on the pressure side of the airfoil and in the far field mesh.

What I'd like to do to accomplish this is to create a normal extrusion of the structured airfoil surface mesh to get a nice structured boundary layer, apply source geometry to define an unstructured refinement zone in the wake and on the suction side of the near-airfoil mesh, and transition everything to a purely unstructured hex volume grid. The entire mesh cannot have anything but hexahedra and quads, our solver code will not do triangle/prism/pyramid/etc...

I'm extremely inexperienced in Pointwise unstructured meshing and am coming from a Gambit/ANSYS Fluent background, where much of the desired mesh behavior can be accomplished by clicking a few buttons. I really need to learn this, but the software is so powerful I really feel at a loss as to where to start. Could you recommend some materials that I could work through to be more proficient with this software?

February 3, 2020, 12:16	Pure unstructured hexahedra in Pointwise	#1
wernerhe New Member Harry Join Date: Feb 2020 Location: Madrid, New York, USA, Earth Posts: 3 Rep Power: 6	Good morning CFDers, I'm a software developer working on some high-order in house-code for moving and deforming grids. Our base testing case will be an infinite span NACA-0012 airfoil at high Reynolds number (Re = 12,000) and we will be pitching the airfoil up to a static AOA and then holding it there for a specified number of chord times. We will compare with results from our water tunnel before testing more complex geometry variants and eventually be moving to marine propellers. My code requires me to have a purely hexahedral grid, and the complexity of future geometry will require me to use a combination of structured hexahedra for my airfoil surface mesh and boundary layer, and unstructured hexahedra for the volume mesh (block) to reduce cell count and avoid high aspect ratio cells in the far-field. My problem is this. I am able to generate a pure unstructured hexa mesh for low Re simulations using ANSYS Fluent and multi-zone method, however at the target Re our flow becomes turbulent and I require a structured boundary layer mesh with first layer height 3e-03 normalized by chord length to achieve a y+ of 1.0. ANSYS Fluent refuses to generate an inflation layer on my complex geometry cases. I know that Pointwise is capable of handling the inflation, however much of my research has revealed that Pointwise is not capable of generating pure unstructured hexahedra or quads automatically. I'm getting conflicting opinions from my research group and I think some of what I'm trying to do is being lost in the language barrier (I'm the only native English speaker here). Does anybody have any experience with manipulating the settings in PW to achieve pure unstructured hexahedra around an airfoil, or would you be able to point me to some reference materials I can read and teach myself? For my complex geometry, I might be able to generate pure unstructured quads along a base NACA 0012 profile and extrude along a path or apply equation based deformation to achieve my geometry, but I have no experience with this method.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Pointwise Unstructured Meshing Between Objects Problem	ona	Pointwise & Gridgen	3	July 15, 2018 17:22
Pointwise 2D unstructured mesh problem	raye_xie	Pointwise & Gridgen	13	March 15, 2018 10:54
Pointwise Unstructured Meshes: PISOFOAM	Ranjan	Pointwise & Gridgen	1	June 17, 2016 16:53
High skewness for a 2D unstructured domain in Pointwise	Stefanodm01	Pointwise & Gridgen	1	August 2, 2015 17:57
unstructured grid, pure convection, code	harbinyg	Main CFD Forum	1	May 26, 2010 07:39

February 27, 2020, 14:12		#2
RcktMan77 Senior Member Zach Davis Join Date: Jan 2010 Location: Los Angeles, CA Posts: 101 Rep Power: 16	In Pointwise you can go about this several ways. Simply create a structured mesh using the structured meshing tools in Pointwise and export the grid into an unstructured grid format. The grid is retained, you just lose the structured i-, j-, k- connectivity information, which gets replaced with cell types, associated nodes, and neighbors (typical unstructured connectivity information). There is also an algebraic extrusion capability in Pointwise that will allow you to directly extrude blocks of unstructured hexahedra from a quad surface. Additionally, you could extrude a boundary layer of hexahedra using Pointwise's T-Rex advancing layer technique around your geometry, and then stitch that together with a farfield consisting of voxels. It just depends on which approach is most suitable for your problem, but the capability certainly exists within Pointwise. You could even elevate the elements of your grid in Pointwise directly.

May 26, 2020, 15:28		#3
wernerhe New Member Harry Join Date: Feb 2020 Location: Madrid, New York, USA, Earth Posts: 3 Rep Power: 6	Hi, I ended up tabling this problem to work on a few other issues that were more time sensitive, and now I am revisiting the issue of unstructured hexahedra in Pointwise. The issue isn't a mesh file format and solver compatibility problem (though it is a good suggestion). I've already made several structured grids for both simple and complex geometry cases and the cell counts are simply unacceptable when I add in the necessary grid refinement zones to capture vortex formation and convection in the wake. I unfortunately cannot share the complex wing geometry I am working on with you, but what I'm after is extremely high resolution for about 4 chords above the suction surface of the airfoil and 4 chords down stream of the airfoil, with a structured boundary layer profile. The mesh should have very low resolution outside of the boundary layer on the pressure side of the airfoil and in the far field mesh. What I'd like to do to accomplish this is to create a normal extrusion of the structured airfoil surface mesh to get a nice structured boundary layer, apply source geometry to define an unstructured refinement zone in the wake and on the suction side of the near-airfoil mesh, and transition everything to a purely unstructured hex volume grid. The entire mesh cannot have anything but hexahedra and quads, our solver code will not do triangle/prism/pyramid/etc... I'm extremely inexperienced in Pointwise unstructured meshing and am coming from a Gambit/ANSYS Fluent background, where much of the desired mesh behavior can be accomplished by clicking a few buttons. I really need to learn this, but the software is so powerful I really feel at a loss as to where to start. Could you recommend some materials that I could work through to be more proficient with this software?