Departments to which FEA is more recent may not have yet established a set of best practices. To jump-start that effort, Bob Williams, product manager at Algor Inc., Pittsburgh, makes these suggestions:

Take formal training when getting started. "While many engineers might be able to teach themselves to use different aspects of an FEA program, most engineers also benefit from traditional classroom training and Web-based distance learning, especially new users," says Williams. "There are techniques you'll learn in class you might not pick up on your own," he adds. "Selecting the right material model and analysis types are examples. You may have extensive material data, but what material model should you choose? If you have piezoelectric materials, how do you analyze their behavior?" Proper training provides the needed guidance.

One reason traditional training pays off is that easy-to-use FEA software lets users tackle increasingly complex problems. "A training best practice is to bring a representative problem or model to class. Seeing how people at other companies use the software is one benefit of classroom training that self-study does not provide. Students leave class with a greater appreciation for applying FEA to the design process," he says. For more training best practices, see "Planning for FEA training" (Machine Design, May 8, 2003, pp. 48-50).

Calculate approximate results before building an FEA model. This best practice is a reality check. "FEA is not a black box into which users place incorrect data and receive useful information. Therefore, it's best to have some idea of how a model will behave beforehand," says Williams.

Another aspect of this best practice is to assemble a list of frequently used calculations or methods that provide a double check of FEA results. "Expected results can also come from prior experience and experiments," says Williams. "Remember, FEA is just one engineering tool of many and does not replace experience and experiments," he adds. When used properly, FEA is a powerful complementary tool.

Reduce CAD models to the minimum required detail. "CAD models can be enormous because they often carry complete manufacturing information. Think first about what really needs to be simulated. What is essential to the analysis? Most likely it is not the entire product. For example, the CAD model of a circuit board may include a heat sink, screws with threads, wires, and other details. These parts could be meshed with thousands of elements, but are they necessary to the analysis? They may only complicate and slow down the required heat transfer simulation," says Williams.

To handle large CAD models, FEA developers have added defeaturing tools to remove the unnecessary details. However, Williams says the tools might never be used because it's so easy to import entire models. "Just a little more time up front removing unnecessary details saves time later on," he adds.

Think about all the environmental conditions that could affect the design. "Structural and mechanical stress are not the only causes of product failure," Williams points out. "Consider whether motion, heat transfer, fluid flow, and electrostatics will affect your design as well." FEA software has made it easier to directly couple different analysis types so that engineers can simulate a product's behavior when multiple physical phenomena interact.

Build and simulate complex conditions in steps. Before sitting down at the computer, break the system into a series of models with increasing complexity. For instance, an engineer designing a computer tape drive first modeled the tape in static tension, then modeled it around one roller, then a second, and finally with motion. Each step let him improve the model and build confidence in the final simulation's accuracy.

"Prove to yourself that you understand how to set up each model and that the smaller simulations work," says Williams. "Starting with large models and complex analyses with lots of loads and constraints usually leads to wondering why the results don't seem correct."

Double check material properties. "Loads and constraints are important, but materials are crucial to a model's accuracy," says Williams. An online database such as MatWeb.com is an invaluable resource that provides free access to extensive material data. Be sure to check material properties at the temperatures the actual product will experience, not just the standard handbook values.

Check the selected material model as well. For instance, Mooney-Rivlin and Ogden work well for rubber, and Drucker-Prager models concrete. There are many others.

Consider a range of loads and constraints instead of just one. "For instance, you may have a 500-lb load acting on an object at a 50° angle. After building a model that simulates this condition, go the extra step and add in two more loading scenarios, one at 400 lb and one at 600 lb as well as the 30 to 60° range."

Software has already begun automating steps that apply load ranges with the additional benefit of sensitivity analyses, those that tell which components are most affected or sensitive to loads. In the past, it was common practice to build safety factors into products to cover design uncertainties. Today, knowing exactly how a product will perform throughout the operating range becomes the safety factor.

Finally, conduct a peer review. Let a resident expert look over the model and results before passing them on for engineering use. "Collaboration is an increasingly important part of the design process because better products result from bringing more than one point of view to the table." Web-ready report generation and real-time collaboration tools help FEA users communicate with others throughout the design process.

Double checking material properties is a best practice for FEA users. One way to do so is through MatWeb.com. It provides free access to material properties. The results, for example, are from a search for plastics with coefficient of friction between 0.1 and 0.3. Picking on a material name in blue pulls up the values needed for an analysis.

Web-ready report generation and real-time collaboration tools help FEA users communicate with others for peer reviews and throughout the design process. A Web-ready report generated in Algor FEA software (lower right), was prepared to let a simulation expert review and comment on the model and results before releasing results to engineering. Real-time collaboration made it easy for design team members to discuss the design's progress (lower left).