ansys

ansys / femap / gridgen / hypermesh / icem / mesher — View Comments
28
Dec 07

Standalone Meshers: Are they a dying art form?

I grew up in a world of CFD where the “code” consisted of a Pre-processor, Solver and Post-processor. Seemed pretty straight forward at the time. I now cringe when I hear those terms. Feels as though someone is dating themselves and using old slang. You know the feeling?!

However, there must be a large market for it as I still hear Harpoon, Gridgen and Hypermesh mentioned from time to time in various companies. Fluent still markets “GAMBIT” as a separate program/interfaces etc. ANSYS acquired ICEM CFD a few years back and morphed an FEA version, AI*Environment out of it as well. Plus, I always thought the guys at FEMAP had cool jobs. Their HQ was only a few miles from my home and I worked for a company that had close ties to them and visited their offices quite frequently.
Why did I think the FEMAP guys were cool? Well, they were able to fulfill a need out there in the market place that others could not. They were in an elite category but didn’t necessarily give the vibe of an “elitist“. They were hard working guys in this little office that were one of few in the engineering software world in the 90s. They have been since acquired by SDRC, then EDS, then UGS, then Siemens. You get the idea.
Meshing has come along way since then. The guys that were doing that stuff were way into it. It was not uncommon for meshing to take a couple weeks, literally. Seemed “streamlined” at the time, as I shake my head and type. Clearly there is still a market where large companies have “Meshing Engineers” on staff still. That market is getting smaller and smaller, but they will always be able to find jobs. They might have to look hard, but they are there.
The coolness of meshing is dying as well. We just take for granted that models can be meshed. We do not view it as an art form that it is. I still smile at the image to the left. Kudos ICEM! It is leap years away from being a push button, but it sure is pretty nonetheless. If you have numerous years of experience on a meshing program, I’d venture to say this would still take days to complete. This mesh was used for an aero-acoustics and aero-dynamics analysis, hence the need for a nice HEX mesh. This is hardcore analysis.
I do believe meshing is still an art form. However, the role of the artist has changed. I used to view the guy driving the mesher as the artist. But the real honor of artist should go to the developer. They are the ones that are up all hours of the night testing their masterpiece. They are being pushed constantly to develop a program that creates the perfect mesh in a click of a button. Anyone can take geometry and create a mesh of some sort. But to create an intelligent mesh that places the correct element type in the correct location with the correct quality everytime, he is the artist. You want to see automatic, check out the image below.

Model brought in directly from Pro/engineer and with one click, I created an optimal mesh. I tried refining the mesh, but found the original was more than adequate. Perhaps my taste in art is changing with age, but That! is a pretty mesh.

ansys / cad / cfd / cfdesign / fea / upfront cfd — View Comments
15
Nov 07

CFD is easier than other FEA Simulations!?!

Sounds as though I am going against the grain here a bit? There are many engineering management teams that feel CFD is beyond their reach. The perception is that it is complicated, difficult to use and require a PhD to drive it. However, Structrual FEA has become almost as common as 3D CAD in many organizations. Often done “upfront” in the design process. So are structural problems so much easier to setup and solve?

In fact, there are a fair amount of decisions to be made when setting up a structural simulation of a typical mechanical assembly. Often, constraining the model can be a challenge. Also, how the parts interact can pose some questions and requires some experience to choose the correct settings. Is it bonded, sliding, frictional contact etc? What exactly is a “penalty method”?Does the model undergo large deflection, causing the analysis to be non-linear? How about material properties – elastic or plastic? FEA vendors, such as ANSYS, MSC and COSMOS have improved their interfaces substantially over the years. Some allow you to read the CAD model directly into the FEA interface. Others leverage the CAD interface and are “CAD-embedded”.

So why is CFD so scary? Well, for one the equations are intrinsically non-linear. Meshing has posed some challenges, as CFD simulations tend to be more mesh sensitive than structural problems. But, the need is there. There are 100s, 1000s of companies that manufacture valves, pumps, manifolds, flow meters, fans, nozzles, medical devices, ducts, hoses etc that could substantially benefit from a flow simulation tool. Not to mention thermal effects. Think of how many mechanical designers are working in the electronics industry that would love to be able to ensure their products will not overheat once they hit the market.

If we think about the decisions that need to be made to set up a flow simulation, they are actually quite a bit less than their structural counterparts. Typically, a flow rate or pressure at the inlet and a pressure at the outlet. Material properties are just density and viscosity. Still cautious? You should be.

There are many CFD vendors that have been around for along time and have been very successful. The merger of CFX and Fluent under the ANSYS umbrella makes them the super power of CFD. Right? How about the guys at Star-CD? They speak the language of fluid guys, analysts. They are very good at what they do and serve their market very well. It just doesn’t seem like they are ready to enter the design community. Can’t say that I blame them. It is a complete cultural shift for them. Sure, they have “entry-level” tools but there is a huge difference between entry-level and an upfront design tool.

Entry level is exactly that. It implies a first step, the need for more at some point. Perhaps OK for some companies. But it can have a mixed message. It can be a risk for a company to make the entry level product part of their design process, but then hit the ceiling very early and then have to move on to “the real deal”. They are now required to learn yet another interface, a new process, make bigger more complicated decisions. This step typically requires a significant additional investment in implementation or worse, the need to hire a specialist. Regardless of which path is chosen, the end result is the same. Less productivity, more time wasted and a lack of ROI in the end.

Instead, companies should make the assessment early on as to what their needs are for say the next three years. What kind of problems will you need to solve? What resources do you have or could you get to solve those problems? But more importantly, partner with a vendor that is focused on solving your types of problems, your way. If you are an Inventor house and design and manufacture valves, then the folks you are working with should surely be versed in both of these. They should speak your language, not require you to speak theirs.
There are vendors out there that are focused on this market- the Upfront CFD market. Certainly, don’t want this to come across as a shameless plug. I am clearly a believer. But then again, I have seen all sides of the argument. Proof is in the pudding. Take a look for yourself. Jump on Google and see what all the hype is really about. I think you’ll be pleasantly surprised. More to come..