My former Ph.D. advisor sent me an email recently asking for some of the geometries I used during my graduate studies. A student had read some papers we wrote and was interested in performing some similar CFD calculations. “No problem!” I said and proceeded to rummage through the bowels of my old computer looking for the appropriate files.
I managed to find the two original geometry files – both of which were in Gridgen’s native geometry format. In fact, this is how and when I was first introduced to Gridgen, but that’s a story for another day. I packaged them up and sent them off to the student.
While doing so, I began to reminisce about the “old” days at the beginning of my Ph.D. and the efforts I invested in meshing. I wondered, What would it take for me to mesh these geometries now? I now had a ton of industrial-scale meshing experience under my belt and a new software (Pointwise) at my disposal. After 45 minutes, I was done. Within about five minutes of finishing, the reality of what I had just done began sinking in. I scrambled to find some Prozac.
Seriously though, in the time it takes me to shower, brush my teeth and dress myself in the mornings, I had gone from dirty CAD to a high-quality, fully resolved boundary layer mesh. During my studies, I guesstimated I spent the better part of the year cleaning up crappy geometries, struggling to get enough prism layers and constantly tweaking the solver to converge with marginal quality cells before I was happy with my meshes. And I’m pretty confident this new mesh was even better (in terms of cell quality and total number of prism layers) than what I had created during my studies. So what gives?
Upon letting this marinate in my brain a little longer and talking with my colleagues about it, I now feel somewhat better about what had happened. Two major things had changed between then and now: 1) I have almost five years of industrial-scale meshing experience 2) Our meshing software improved significantly since that time. Let me expand on both of these items.
As I blogged about last month, as a technical sales engineer my job involves helping people solve their meshing problems. I help all kinds of engineers, in academia and industry, from every discipline, create meshes for a huge range of applications. Every day. If you think about that for second, most CFD engineers are trying to do the opposite – spend as little time meshing as possible. It certainly varies by person, company and application, but if I were to take a wild guess, a CFD engineer might spend on average of one day per week on pre-processing. That means I have as much meshing experience as a typical CFD engineer would in 25 years. Whatever the actual numbers are, I now have a lot of meshing experience.
I first used Gridgen in 2003. Pointwise was released in 2007. Since grad school, not only is there a newer, more powerful and easier to use pre-processor but it’s had five years to mature. Two big ticket capabilities that I used for my new mesh that weren’t at my disposal nine years ago are Solid Meshing and T-Rex. Solid Meshing gives users the ability to heal dirty geometry, recover engineering intent and more efficiently get to a watertight mesh. There’s a nice explanation of it in a past issue of our Connector newsletter. T-Rex which is short for anisotropic teTRahedral EXtrusion, is a robust and highly automated method for creating unstructured boundary layer meshes on complex geometries. T-Rex has it own page on our website: http://www.pointwise.com/T-Rex/.
So what is my point? Don’t be easily discouraged by CFD and in particular, meshing. No one said it would be easy. But that doesn’t mean it can’t get easier. Here’s are some suggestions to make your CFD meshing experience better:
- Spend time learning it: Like anything else, you need to invest time educating yourself. Meshing in itself is a full fledged science. Read websites, books and journal papers. There are people (both commercial and academic) who happily spend their careers researching and improving meshing technology.
- Attend training: If you are a current customer of ours, training is include with your license for the courses we hold in our Fort Worth, TX, office. If you want to get really good at it, work for a meshing company (we’re currently looking for developers and support engineers so check us out).
- Investigate meshing technologies: Aside from ours, there are dozens of other meshing tools, both commercial and free, out there. We don’t profess to be the “be all do all” mesher. We know our strengths and weaknesses. For over 25 years, our bread and butter has been structured meshing but as applications have become more complex and bigger we have developed new techniques to be able generate high-quality, complex meshes quicker (e.g. Solid Meshing and T-Rex).
- Talk with experts: Read forums and ask questions. Join a mailing list and solicit advice. Attend CFD and meshing conferences and speak with users. Call me and discuss your applications (queue the shameless plug). I have always found the CFD community small, tight-knit and as a whole very friendly. Everyone is approachable. We’re all nerds wanting to learn new things and talk shop about each others applications.
Meshing doesn’t need to be the bain of a CFD engineer. Educate yourself. Seek expert advice. Check us out. Give me a call.
Awesome! I like it.
A suggestion. Take it to the next level. Run it in a RANS flow solver. Most likely a turb model such as SA will hit steady state. SST might not. Refine the grid and do a grid convergence study on skin friction. Integrated values and contours. That’s what I’m looking for. If I can’t get reasonable behavior on solution and grid convergence then I’m reaching for the Prozac! And this is just a suggestion. I know such work always takes time.
Martin, I did all of this and more for my thesis. One of the chapters was specifically about grid convergence, near wall resolution, turbulence model influence, etc, i.e. verification and validation, for these types of applications. If you would like a copy, I’d be more than happy to send it to you.
Oh, the idea was just to show a couple plots on the blog so that your readers see a simplification of the analysis process. As you know, it takes many grids to gain confidence in the solution and your grid. I assume you spent a lot of time on verification and validation. I’m not sure if everyone gets what it takes…
great post, thank you. When I have time, I practice, practice and practice. I re-read and re-read the tutorials and the user manual, load old geometries, practice geometry cleaning and re-build mesh.
chris, could you post a pdf of your thesis?? (you recommend to learn from experts)
Pingback: How Meshing Stole an Hour of My Life – Applied CCM