Reliable Meshing Process vs. Push Button Mesher

As you already know, the holy grail of CFD meshing is to be able to push one button and automatically produce a perfect CFD solution, or what I like to call the push button approach. Of course, the tricky part of this is the grid generation. While this single click system may get you a grid that you can read into your solver, depending on the complexity of the simulation, it may produce a grid that requires a lot of time spent tuning solver settings to achieve convergence or it may not properly resolve important flow features resulting in a solution that is not correct. More often than not, adapting the solver to a particular mesh isn’t practical, or even possible when using commercial software. In the end you may be forced to go back to the preprocessor to adjust a few parameters and try pushing the grid generation button again. By the time a decent solution is found, the push button grid is already far from realized.

A major drawback of this approach is the lack of control that prevents users from being able to make desired adjustments. This closed process is termed automatic meshing because the software tries to figure out the best grid based on a very limited set of meshing parameters and constraints. But unless you’ve generated the grid before, how do you know what parameters to use? Is there an approach that provides the automation of a push button mesher but with the added benefit of control and a well defined meshing process? I call it the reliable meshing process, and it’s the way engineers typically generate grids using Pointwise.

The Reliable Meshing Process

From the beginning, Pointwise has offered tools that can automate portions of the meshing process while still giving users enough control to get a grid that meets their convergence and accuracy requirements. Each rung of the ladder from the construction of the grid curves through initialization of the volume grid is automated with default inputs that usually result in a good grid. Ultimately, Pointwise is working to provide an automated meshing experience without sacrificing the benefits of the bottom-up workflow as outlined in Rick Matus’s article, “Meshing Philosophy: Automated or Automatic?”

It’s really the best of both worlds. Automate the majority of the meshing process, but provide the user with the option to interrupt the process and make modifications if necessary. With the reliable meshing process user control is not sacrificed for automation, instead the two are married to provide a unique meshing technology to tackle problems of varying complexity.

Automation in the Real World

Putting the reliable meshing process to the test, I sat down to generate an unstructured grid for the Generic Conventional Model (GCM) semi truck and trailer. The geometry imported into our software fairly cleanly, and after a few minutes I had several watertight solid models that I could begin meshing. Had I been using an automatic tool, cleanup and defeaturing may have been faster than the approach I chose to take, but at the cost of control. I’m an engineer and a pilot and I like to be in control. That’s why I don’t enjoy rollercoasters or on-rails first person shooters, they are one-dimensional experiences in a three-dimensional world. But I digress. In the end I had the exact geometry I wanted to start meshing on.

Watertight solid models representing the GCM semi truck and trailer.

Watertight solid models representing the GCM semi truck and trailer.

With the geometry import out of the way, I began to focus my attention on surface meshing. Using the surface grid boundaries I defined on the geometry based on an arbitrary turning angle, I manually grouped them into more appropriate meshing regions to recover the original engineering intent. I didn’t have to do this, and with a push button mesher I would not have been able to, but it’s a valuable option when using our automated tools.

In this situation, Pointwise’s automated tools offer a clear advantage over fully automatic approaches by getting me ninety percent of the way there with little effort. However, I’m still left with the last ten percent to make any changes or adjustments I feel are necessary before passing things over to the surface mesher. The last ten percent really makes all the difference and ultimately reduces the time and effort I have to spend downstream in the meshing and CFD solution process.

The GCM meshing regions defined using quilts.

The GCM meshing regions defined using quilts.

With the hard work out of the way, the next step was to specify the default spacing constraints for curve and surface grids and then automatically put surface grids on CAD entities. Using the toolbar option, this was accomplished with a single click. While creating surface grids in Pointwise can be done automatically, I still had the ability to go back and make modifications where necessary. For example, in gap regions I locally refined the surface grid by adding internal edges and even refined the grids on the axles based on surface curvature.

A high quality, watertight, unstructured surface mesh.

A high quality, watertight, unstructured surface mesh.

Creation of the farfield surface grids was straight forward and is a perfect example of where scripting can be used to automate tedious parts of the meshing process. A Glyph script could be written to automate the creation of arbitrarily shaped farfield domains. In fact, if the process was streamlined enough and geometry changes within reason, a script could be written to automate the entire meshing process from start to finish. You could even use Glyph scripting to integrate Pointwise into a design optimization framework or conduct a grid dependency study.

Volume meshing was accomplished using T-Rex, Pointwise’s automated unstructured viscous boundary layer meshing tool. T-Rex starts from an existing surface grid then uses user defined boundary conditions and quality criteria to locally advance points and generate stacks of right angled tetrahedral elements. The final front of the extrusion is then passed to the isotropic tetrahedral mesher and the remainder of the volume is populated. On export, the stacks of right angled tetrahedra can be recombined to form triangular prisms using information stored during point advancement. The advantage T-Rex has over traditional boundary layer inflation techniques is the ability to stop locally if there are cell quality violations or interferences with adjacent cells or geometry while allowing surrounding points to continue advancing. The result is a high quality boundary layer resolved region created with little user interaction. This makes it ideal for geometries such as the GCM truck and trailer.

After applying the appropriate boundary conditions and adjusting a few parameters, the volume mesher was let loose. By the time I was back from the kitchen to get a cup of coffee, the extrusion was nearly complete. A couple more minutes and the volume mesh was finished.

The final unstructured viscous volume mesh.

The final unstructured viscous volume mesh.

In less than an hour I went from geometry import to a boundary layer resolved volume mesh. In fact, I even had time to back up and redefine a few inadequate meshing regions and adjust several surface meshing parameters I felt unhappy with. Pointwise’s bottom up meshing paradigm offers a clear advantage over the push button mesher, or top down meshing software in general, by allowing me to easily stop the process and make modifications and adjustments as necessary.

As you can see, Pointwise isn’t a push button mesher, and it probably never will be. We feel it’s much more important to you, the user, to control the level of automation and ultimately your overall exposure to the details of the meshing process. This can be accomplished either by using the automated toolbar shortcuts, the more advanced hands-on menu bar commands, or even by writing a Glyph script. Pointwise allows you to be the judge regarding the level of automation desired for your meshing applications.

About Travis Carrigan

A Pointwise engineer helping other engineers solve their meshing problems.
This entry was posted in Applications and tagged , , , , . Bookmark the permalink.

Leave a Reply