Fluid simulations 101

Most engineers are familiar with finite-element analysis and its usefulness in structural problems, but they aren't as well versed in computational fluid dynamics (CFD). That is about to change as more developers simplify their CFD software for wider use. For example, FEA developer Cosmos, a division of SolidWorks, along with Nika Inc., a German firm, have tuned the latter's CFD algorithms so even bench-level engineers can apply Cosmos-FloWorks to frequently encountered engineering problems for better insight on products ranging from ball valves to computer enclosures.

The latest features in the newer CFD program, developers say, make it easy to use, let users monitor progress of solutions, and simulate rotating components so it models more complex products. But getting CFD programs up and running takes a bit more than just buying and loading the software.

To implement CFD software into an engineering department, Bill Dziedzic, a senior consulting engineer with Cosmos, suggests first taking a two-day class on the software. Other experts add that working through tutorial problems before going to class better prepares students because it familiarizes them with the software and prompts questions. And since most engineers have already been introduced to fluid mechanics, Dziedzic suggests reading the first few chapters in an introductory text on the subject to get reacquainted with frequently used terms and equations.

After examining the software, engineers will find the CFD is easy to use thanks to a close link with CAD software and a wizard that speeds model setup. “The CFD software runs inside Solid-Works so it's easy to set up a range of configurations for study,” says Suchit Jain, product manager with Cosmos. “The close link means a single button puts the modeling system aside and brings up the CFD program. The wizard helps apply a mesh and boundary conditions. And when it's necessary to adjust the geometry, a single button restarts the CAD functions. Returning to the CFD program allows reapplying the previous mesh and boundary conditions.” Bringing in IGES, Step files, and other types of geometry transfers are less efficient because mesh and boundary conditions must be manually applied each time making it more tedious to test, according to Jain.

The wizard simplifies preprocessing by turning it into a series of questions such as: Does the simulation have internal or external flow, or both? What is the ambient temperature and pressure? Is heat transfer an issue? What is the fluid?

Goal-based simulations, a relatively new idea, let users assign a sort of digital sensor to a component or surface of interest. “The goal can be a parameter such as velocity, pressure, or temperature,” says Dziedzic. “The parameter can be associated to a volume, surface, or a global item in the system. For example, if users want to know how hot a component gets in an enclosure, they could associate a goal to its temperature. Then while the problem is solving, users can see a plot of the value. Should the temperature exceed a particular value, the simulation can be stopped and the model adjusted, perhaps by adding a fan. Monitoring simulations is a good idea because it's not unusual for complex models to take 2 to 10 hours to finish, and finding a modeling or preprocessing error early in a long run can minimize the number of simulations,” he adds.

For now, users monitor problems by watching a meter in a window. It shows convergence history and a preview plot of the solution. In addition, values from a goal can be used in equations and those functions plotted. The feature could be used, for instance, to track the efficiency or thrust of a propeller. If its design is not meeting expectations, the solver could be stopped, design adjusted, and restarted.

Simulating rotating components is another recent feature. “Actually, the geometry doesn't move — only the frame of reference does. For example, a car in a wind tunnel does not move, but the air around it does. Similarly with a rotating frame of reference, fluids would spin around a blade.

As for solver speed, Dziedzic says the software's may not be the fastest, but it is more certain to converge than others. And that avoids headaches.

Interpreting results could be a topic of its own, but Jain and Dziedzic offer a few guidelines. For example, streamlines are paths that particles would follow if dropped into the flow. They are also color coded to indicate pressure or speed. Lines that seem to knot or tangle indicate turbulence a source of noise and drag. They may require engineering attention. To spot noisy products, users plot energy and dissipation from turbulence as functions or contour plots. “Users can also look for vortices in flow streams,” says Dziedzic.

Tools in the software allow looking at results several ways. “A few tools in the CFD program are the same as in the solid modeler,” says Dziedzic. “Say velocity through a certain cross section is of interest. Slicing the results model allows such an examination, and it is simple.”

Links to Microsoft Excel and Word are fairly standard in recent analysis software, but they speed writing and disseminating reports. A plot command in the CFD program moves CFD data into Excel spreadsheets for additional analysis.

—

MAKE CONTACT:
Cosmos, (310) 207-2800
www.cosmosm.com