Analysis is getting easier and results are coming faster, but so are mistakes. How to avoid them.
Modern simulation software does a lot with a little. You give it some simplified geometry and a few boundary conditions, and you get back brilliant results. But before you run into your boss’s office saying, “It works” or “It broke,” take a moment to check your results. They may be dead wrong. And you could be quite embarrassed.
The stresses or fluid flows that unfurl on your screen seconds after you hit “solve” sure may look right. You trust the program. Computer-aided engineering (CAE) programs can be counted on to work without oversight. Your simulation program has dutifully taken your geometry and run with it. A lot of work occurs that is unseen—meshing, converting inputs into a boatload of partial differential equations, assembling an immense matrix, solving it—with 100 percent accuracy. Your program has trusted in you and assumed that you knew what you are doing, that you have used sound engineering judgment—and have not made a single typo.
Like a dog happily and repeatedly fetching a stick, your simulation program will be eager to please and do what you request of it, no questions asked. It will be up to you, its master, to put this trust to good use.
Here, we will list several mistakes that are frighteningly easy to make—and describe how you can try to avoid them.
- Mass Is Critical
- Count Your Zeroes
- Units: Nothing If Not Consistent
- Incorrect Material Properties
- The Wrong Stress
- Use Checks
- Intuition—This is often based on experience with similar problems.
- Physical testing—Build a scale version of your design for validation of the finite element model. With the finite element model verified, you will have confidence to proceed with the FEA model for variations in loads, materials and so on.
- “Back-of-the-envelope” checks—These are useful in determining whether the results lie in the right order of magnitude. Handbook formulas may apply only to a gross approximation of your part, but with the absence of physical tests, this could still determine if FEA results are in the right ballpark.
- Bracketing the result with checks—This is also helpful. For example, suppose you have a structure whose support behaves somewhere between a simple support and a fixed support. From handbooks, get the result for both cases and check the FEA result to make sure it falls inbetween.
Figure 3. In a dynamic analysis, mass should be used, not weight. (Image courtesy of MSC Software.)
Think back to your first science class when you found out that weight—how you measured almost everything—mattered only to you on the Earth’s surface. For the rest of the universe, and in most CAE programs, you learned that you had better switch to mass. If you have since neglected this early warning and used weight instead of mass, your results will be off by the gravitational constant. That will affect every place where the mass of a structure is important—even on Earth—as well as dynamic analyses where mass is involved.
The inputs and outputs of FEA programs are usually entered without units; therefore, it’s up to users to make sure that the units of measure are used consistently. For example, if a part is measured in inches, take care to use inches consistently. A unit that is based on length will be affected. Density, for example, is mass over length cubed. If acceleration is being entered for a dynamic analysis, acceleration due to gravity should be in inches per second squared, not the commonly expressed feet per second squared. Otherwise, the result will be off by a factor of 12.
Tables of material properties are easy enough to find, but there is more to material properties than meets the eye.
Material data presented as a single number implies a certain, but fictitious, precision. Yield stress, which may be reported as a definite number, may be an average of data points from lab experiments. Designing to an average yield stress without any safety factor would mean that half of your parts would fail. Look for statistical qualifiers in the data. Some references give material properties that satisfy 99 percent of samples tested. Also, some industry codes may suggest that you use 50 percent of the stated value for the sake of safety.
Most FEA programs will default to apply similar material properties in all directions, a condition analyst call isotropic. But many materials (e.g., wood, carbon fiber composites) do not behave isotopically. In fact, materials that are very strong in one direction can be extremely weak in the other two.
You will need to consider if parts will be in the same temperature range as the samples from which the material data was obtained. A property table may have been generated at “room temperature,” but the actual parts may be subjected to extreme temperature conditions where materials respond quite differently. Metals get very brittle at extremely low temperatures and become soft when heated. A change of phase when you are counting on a material to stay solid could prove to be a disaster.
The more sophisticated CAE programs do allow you to input temperature-related material properties.
FEA solutions are often based on the assumption that the material will behave elastically, which is fine as long as deflections stay small. But if there is too much loading, all bets are off. For one thing, this will distort the finite elements, and the more distorted it gets, the less trustworthy the results will be. If you are using a material that experiences a lot of stretch before failure, look into using elements that can accommodate distortion from large displacement.
FEA programs can display any number of different stresses. For example, an analysis will give a choice of several directional stresses, principal stresses, shear stresses and the von Mises stress. Looking at the wrong stress could potentially cause you to miss the stress that causes failure.
Failure for ductile materials, such as steel and aluminum alloys, can be well predicted using the von Mises, or an equivalent stress. If the maximum von Mises stress is less than the yield stress, the part will not fail.
Maybe the biggest mistake you can make in simulation is not checking for mistakes. Most gross errors like the ones listed above, which can skew results by an order of magnitude or more, can be eliminated with the following simple checks:
Are there any other simulation blunders we have missed? Please let us know in suggestion box below.