March 01, 2010
CFD: Of Meshes and Math
Please note that contributed articles, blog entries, and comments posted on MCADcafe.com are the views and opinion of the author and do not necessarily represent the views and opinions of the management and staff of Internet Business Systems and its subsidiary web-sites.
| by Jeff Rowe - Contributing Editor
Each MCAD Weekly Review delivers to its readers news concerning the latest developments in the MCAD industry, MCAD product and company news, featured downloads, customer wins, and coming events, along with a selection of other articles that we feel you might find interesting. Brought to you by MCADCafe.com. If we miss a story or subject that you feel deserves to be included, or you just want to suggest a future topic, please contact us! Questions? Feedback? Click here. Thank you!
Among the mechanical designer's many responsibilities, there is one task that routinely causes a pang of hesitation: fluid flow analysis. Why? Because most engineers remember studying the subject in college and being grateful that their career plans were taking them in a different direction.
The study of fluid flow has long been regarded as a science by some, an art by others. The underpinning is a mathematical regime so complex and mysterious that it has its own name: the Navier-Stokes equations. Developed by a pair of 19th-century geniuses exploring the realm of mechanics, the equations have been a rite of passage for generations of engineering students. Some aspects of the Navier-Stokes equations are actually said to be insoluble by human hands and instead must be solved with the use of numerical approximations known as Computational Fluid Dynamics (CFD).
Fluid flow is an increasingly common part of product design. Increasingly, devices that were once purely mechanical incorporate control electronics and sensors and are expected to operate with essentially no human intervention. Consider a fire suppression system's valves and conduits. They must deliver their payload flawlessly when it is needed. And only by thoroughly evaluating the “what-ifs” of the suppressant flow can this level of performance be assured. Any mechanical engineer developing a device that conducts air, gases or liquids-or even sand or oatmeal-needs a way to predict the fluid flow response.
CFD is the most widely accepted solution, but it can present a challenge for mechanical engineers who lack special training. Imagine designing, for example, a manifold that conducts atomized fuel. Prior to a CFD analysis, voluminous details about the product's dimensions and topology must be fed into the process. If the fluid flows through hollow areas in the design, these must be separately defined and modeled as solids. Something called a “mesh” must be devised and boundary conditions defined. It's no wonder that mechanical engineers usually turn to CFD specialists to handle the translations and computations.
Unfortunately, many of today's design project deadlines do not allow for the time it takes for conventional CFD analysis. Iterative cycles of refinement are often needed but CFD complexities can limit the number of repetitions. Equally important, conventional CFD usually cannot commence until the design is at a relatively advanced stage.
Look at the prevailing method of mechanical design today. The engineer works at a PC or workstation running one of the big MCAD applications. The MCAD toolset automates many labor-intensive calculations and record-keeping operations. It imposes design rules and checks the user's work. It draws and dimensions the emerging device on-screen while recording details about solids, surfaces, apertures, outlets, and chambers.
Finding and Fixing a Disconnect
The astute observer will notice that this is exactly the information the CFD process needs. But there is a long-standing disconnect between MCAD and CFD, and this must change. There is no expedient way for the MCAD application to deliver its knowledge to the flow analysis tool. The MCAD and CFD domains have always been disparate worlds, speaking different languages. Complex transfers between the environments add time and increase the likelihood of errors. Where the engineer sees smooth curves and angles, the CFD tool needs to see a “mesh” derived by discretizing continuous features into small localized domains that can be analyzed individually. Figure 1 depicts a
cross-sectioned valve whose flow areas have been meshed for CFD analysis. This is a very complex operation requiring decisions about the size, shape, and placement of the cells that make up the mesh.
Figure 1: A valve and its cross-section that has been meshed for CFD analysis.
Clearly, there is a need for some form of integrated toolset that will enable the designer to run CFD analysis procedures from the comfort of his or her MCAD workstation Absent such a solution, CFD analysis will continue to be a pacing item in the mechanical design timeline.
Fortunately the simulation and modeling industry has listened and alternatives are emerging. Some tools have evolved to allow design engineers to conduct analysis “up-front” by offering special interfaces CFD from to CAD. These interfaces make it easier to conduct analysis but require users to deal with two different GUIs when transferring geometry between the two environments. And meshing is still left to the designer.
Concurrent CFD is another, newer approach. Concurrent CFD is CAD-embedded, therefore it enables mechanical design engineers to analyze and optimize their designs directly within leading mainstream MCAD environments. This can save up to 75% in analysis time when compared to conventional CFD. Initiating a CFD analysis is no more difficult than running commands under the customary MCAD menus.
True concurrent CFD tools such as Mentor Graphics' FloEFD spare the designer from time-consuming operations associated with traditional and up-front CFD tools and provide the following advantages:
A concurrent CFD application shares MCAD geometry data without the need for export or translation.
These same tools generate optimized meshes automatically. The importance of this feature is hard to overstate, since it applies “expertise” that formerly came only with specialized CFD training and experience.
The tools also perform cavity modeling automatically, rendering “hollow” spaces as solids so they can be meshed to predict their flow response.
Concurrent CFD speeds what-if analysis. Models can be set up, cloned, and subjected to multiple analysis scenarios without the need to reassign boundary conditions or material properties.
A fairly routine CFD analysis can produce an impressive result. Figure 2 illustrates a lamp design showing the air flow around the exterior of the device, with color-graded details about temperature changes as the air travels over the heat source-a power LED.
Figure 2: Concurrent CFD analysis of air flow over an LED luminaire
The Transformation Ahead
I believe concurrent CFD can have a transformative effect on the mechanical design discipline. Integrating once-daunting CFD procedures into the MCAD realm will free engineers to try out their most adventurous ideas and see the results almost immediately. It is a situation that fosters experimentation, learning, and innovation.
Of course there are other factors. First, designers who have efficient, easy-t0-use CFD capabilities at their fingertips will check fluid flow behaviors early and often. This will prevent errors and misconceptions from being perpetuated through the design process.
Second (and directly related to the first issue), embedded concurrent CFD tools will enable more thorough testing without unreasonably extending project timelines. Of course, this will have a favorable impact on project costs as well. In the past it has been common practice to bring in external CFD resources late in the process, only to learn that the MCAD prototype's behavior is not as expected. Back to the drawing board!
Third, designers will gain a more systemic view of their device's behavior. As they become better acquainted with the intricacies of fluid flow (thanks to their hands-on work with CFD) they will learn to anticipate the fluid flow characteristics and needs of their designs-even if these attributes are incidental to the device's main functions.
We are on the threshold of a paradigm shift in mechanical design. Fluid flow analysis and CFD have always played a necessary role in product development. But the shift of CFD from discrete external discipline to integrated MCAD function will enable a faster, more agile, more productive design process.
By Erich Buergel, General Manager, Mechanical Analysis Division, Mentor Graphics Corp.
Commentary By Jeffrey Rowe, Editor
Our thanks go out to Erich Buergel of Mentor Graphics for providing this article on the increasing necessity of integrating CFD with MCAD. Once the domain of specialists, I see CFD going much the same direction that FEA took a few years ago - wider acceptance and use by non-specialists. A definite trend for the future.
I do remember, though, in engineering school that fluid mechanics (of which CFD is a branch) was one of the “weed-out” classes that separated those of us who would go on to become engineers, and those who were forced to switch to other curricula, such as liberal arts (not that there's anything wrong with that).
As Mr. Bruegel points out, the fundamental basis of most CFD problems are the Navier-Stokes equations that define fluid flow through approximation. These equations can be simplified by removing terms describing viscosity to yield Euler equations. Further simplification, by removing terms describing vortices produces full potential equations, and these equations can be linearized to yield linearized potential equations.
Historically, the first CFD methods were developed to solve linearized potential equations. Two-dimensional methods, using conformal transformations of the flow about a cylinder or the flow around an airfoil were developed in the 1930s. As computer power became more available three-dimensional CFD methods were developed.
The fundamental consideration in CFD is treating a continuous fluid using a discrete method. One method is to divide a space (spatial domain) into smaller volume spaces to form a volume mesh or grid, and then apply a suitable algorithm to solve the equations of motion (Euler equations for non-viscous, and Navier-Stokes equations for viscous flow).
Although usually complex, it is possible to directly solve Navier-Stokes equations for laminar flows and turbulent flows when they can be resolved by a grid. Generally, however, the range appropriate to the problem is larger than even today's massively parallel computers can handle. In these cases, turbulent flow simulations require turbulence models, such as large eddy simulations (LES) and the Reynolds-averaged Navier-Stokes (RANS) equations.
For really complex problems, other equations are solved simultaneously with the Navier-Stokes equations. These other equations can include those describing concentration (mass transfer), chemical reactions, heat transfer, etc. More advanced algorithms can simulate more complex fluid problems involving multi-phase flows (such as liquid/gas, solid/gas, liquid/solid), non-Newtonian fluids (such as blood), or chemically reactive flows (such as combustion).
Regardless of the approach, the same basic procedure is followed for solving CFD problems:
A preprocessor defines the geometry (physical bounds) of the problem, the volume occupied by the fluid is divided into a uniform or non uniform. mesh, the boundary conditions and physical modeling are defined
The simulation is started and the equations are solved.
A postprocessor analyzes and visualizes the result.
Obviously, what goes on behind the scenes is a lot more complicated than that, but those are the basic steps that are followed.
So, will CFD be the next “big thing” in the simulation/analysis arena that makes the grade in MCAD products? Tomorrow? Maybe not. Going forward? I'd say the odds are pretty good that that's a pretty safe bet as all types of simulation and analysis move forward in the design process.
The Week's Top 5
At MCADCafé we track many things, including the stories that have attracted the most interest from our subscribers. Below are the five news items that were the most viewed during last week.
Dassault Systèmes (DS) launched Release 20 of its collaborative V5 PLM portfolio, including CATIA, ENOVIA, SIMULIA, and DELMIA. The launch of V5R20 includes enhancements to DS' ENOVIA SmarTeam multi-CAD collaboration software, integration of SIMULIA's nonlinear and thermal realistic simulation capabilities into the V5 platform, as well as composites design and simulation capabilities in the CATIA and SIMULIA portfolios. V5R20 also features enhancements across the entire product portfolio, including digital manufacturing. V5's openness is advanced in Release 20 with updated multi-CAD integrations and a new 3D viewer available directly within ENOVIA SmarTeam. In addition, V5R20
introduces a new product, CATIA Extended STEP Interface. It enhances large assembly archiving with nested assembly support and in V5R20's Q2 2010 update will feature unique comprehensive support for composites design attributes and functional tolerancing and annotation data. Q2 2010's update to the entire V5 portfolio will also include support for Windows 7.
Luxion, makers of advanced rendering and lighting technology, announced the first release of KeyShot, an interactive raytracing and global illumination program that brakes down the complexity of creating photographic images from 3D models. KeyShot supports many native file formats, including SolidWorks, Rhino, SketchUp, Obj, IGES, and STEP. KeyShot is the replacement for the HyperShot software, the first interactive raytracing and global illumination program, also developed by Luxion. KeyShot is the realtime ray tracing program formerly know as HyperShot. Luxion will provide a free upgrade to all existing users of HyperShot. Pricing for KeyShot starts at $995, while KeyShot Pro sells
for $1,995. KeyShot is available for immediate purchase for both PC and Mac from
A Utah design team used its metalworking experience and SolidWorks to create a skeleton sled for an Olympic racer in a fraction of the time and for tens of thousands of dollars less typically required to produce one of the highly specialized sleds. A skeleton sled is a compact racing vehicle that riders steer down high-speed ice tracks by flexing and shifting their bodies. Suddenly without a working sled late last year when hers was damaged, U.S. Olympic team member, 2005 and 2007 World Cup champion Noelle Pikus-Pace, turned to her husband, Janson Pace, an industrial designer, and his employer, NuQuest, a design, engineering, and manufacturing company. NuQuest produced the sled in
four months, which is much less time than it usually takes to perfect a new design. NuQuest co-owner Troy Beckstead estimated that SolidWorks Simulation saved the company “tens of thousands of dollars” in development costs. Pikus-Pace will use the skeleton sled at the Vancouver Olympics.
ANSYS, Inc. announced that Volkswagen AG has signed a master agreement with ANSYS and intends to widen its use of ANSYS' comprehensive engineering simulation software. The strategic decision to use ANSYS software was due both to the bandwidth of applications that can be addressed as well as to the innovative ANSYS Workbench platform that allows for a substantial process compression. In its research and development, Volkswagen uses structural mechanics, fluid dynamics and explicit analysis tools from ANSYS to perform, among other applications, studies on climate control, headlights, and engine internal flow.
A growing number of refrigeration and cooling companies are designing their machinery with Autodesk Inventor. Companies such as Mammoth-WEBCO Inc. (Mammoth), a leading custom HVAC operation, and KYSOR Panel Systems (KYSOR), a leading developer of walk-in coolers and freezers based on innovative insulated panels, use Inventor software to go beyond 3D design to Digital Prototyping. By using Inventor software, Mammoth, which manufactures large-scale commercial custom rooftop equipment that weighs up to 60,000 pounds per section, no longer needs to construct physical prototypes and experiences far fewer errors when manufacturing its massive equipment. Since implementing Inventor software,
Mammoth has helped reduce manufacturing time by nearly 15 percent and now saves anywhere from $500 to $9,000 per project. KYSOR has streamlined much of the engineer-to-order process for its insulated panels by using Autodesk Inventor and Autodesk Vault software. KYSOR panels are used for food storage in supermarkets and convenience stores, as well as scientific applications where environmental control is critical and cold storage facilities where temperature, humidity, and light must be precisely controlled. With Inventor software, KYSOR says it has increased the overall efficiency of its design automation by about 45 percent.
Jeffrey Rowe is the editor of MCADCafé and MCAD Weekly Review. He can be reached at email@example.com or 408.850.9230.
Product and Company News
Related MCAD News
You can find the full MCADCafe event calendar here
To read more news, click here
-- Jeff Rowe, MCADCafe.com Contributing Editor.
For more discussions, follow this link
- CFD: Don't be Scared March 06, 2010
Reviewed by 'John Chawner'
Thanks for giving mesh generation top billing (or at least equal billing) with CFD in the recent article "CFD: Of Meshes and Math". Products like Mentor Graphics' FloEFD and Blue Ridge Numerics' CFDesign (among others) are helping make CFD more accessible to design engineers (aka non-expert analysts).
But I have to take minor issue with the tone. This whole "fluid dynamics is hard" business reminds me of Barbie's "math is hard" but without the corresponding outrage. Yes, engineering *is* hard. That's why we go to school for 4 (or many more) years to become adept at it. Yes, CFD is hard but so is FEA (which also requires a mesh) but no one thinks twice about pounding out a bunch of FEA simulations. Yes, some engineers don't fully get CFD and then others of us (like me) get the "deer in the headlights" look when presented with a structural fatigue problem.
Stepping down from the soapbox, I agree with your assessment that CFD will continue to grow as the overall use of computational simulation increases. After all, simulation during the early stages of design leads to all sorts of good things including better products and quicker to market. And, mesh generation will be a driving factor in making use of CFD more efficient and effective. After all, your simulation is only as good as your grid.
Thanks again for some superb insights.
Full disclosure: At Pointwise we deliver mesh generation software and services to CFD practitioners worldwide.
One person found this review helpful.
- NICE FOLLOW UP! March 03, 2010
Reviewed by 'Russ Henke'
Thanks for your MCAD Weekly piece on "CFD: Of Meshes and Math". You caught the mood perfectly of many of us mechanical engineering students in our undergraduate years in fluid mechanics classes.
You're article was an excellent follow-up to the November 09, 2009 EDA Weekly posting of "MAD Progress":
Keep up the good work!
March 03, 2010
One person found this review helpful.