Improving Simulation with test
I got a chance to talk with some folks from National Instruments about how physical test can improve (or at least, inform) simulation.
National Instruments (www.ni.com) is a very well known test and measurement company ($571M sales for 2005) with their LabVIEW software as the leading test platform software. They do not make sensors (they partner with sensor vendors) but provide the data acquisition hardware and LabVIEW software which controls the test and collects the test data. (In a previous version of this entry, I screwed up and called National Instrument "National Semiconductor." I can only plead temporary stupidity.)
What National has done that's interesting is write an application that captures the test data, and feeds it back into Cosmos (the FEA software from SolidWorks.) By doing this, a simulation engineer can correlate the results of their FEA analysis against the physical test. (And, personally, I hope airframe manufacturers don't stop doing physical test any time soon.)
One intersting result that National found when doing a mock-up of a wing (with sensors) was that what should have been a linear system was actually non-linear. That is, you press down on the end of the wing, and the deflection is not linearly proportional to the force. Now, this is not something I'd like to find out when sitting in an airplane during a windshear.
It turns out that the cause of this result lies deep within the theory of finite element analysis: systems are only linear within certain limited ranges of deflection. In the case of wings, it's about 5% of the length of the wing. If you've looked out the window of a plane, and watched its wings bounce up and down, you can probably surmise that they deflect more than 5%.
While it might be fair to say that a simulation engineer ought to know enough to do a non-linear analysis when they expect relatively large deflections, there's a difference between "ought to know" and actually knowing.
All the information necessary to determine if a deflection is too large for linear analysis is available to the CAE software -- so, why doesn't the CAE software give the engineer that information?
In fact, as a result of the conversations between National and SolidWorks, the Cosmos FEA software now does tell users when there is a chance that the simulation has enough deflection to possibly be non-linear.
Another example of software making engineers smarter: Instead of just reporting a "blue is good, red is bad" result, Cosmos software can now tell the user "you might not want to trust these results."
Evan Yares
3 Comments
Reader Comments (3)
Evan Replies:
I think you're right -- this will be more interesting to people who don't make their living with FEA. My comment about "Blue is good, red is bad" was a reference to a statment made by Paul Teutul Sr, of Orange County Choppers, about CosmosWorks -- not a man who's noted for his engineering acumen. Thanks for your comment.
There are multiple ways for a problem to become nonlinear. Geometry, as the software considers, material nonlinearity as Brian adds, and also the loads can be nonlinear (aerodynamic being a prime example!). COSMOSWorks already provides facilities for including material nonlinearities. It is basically impossible for the software to warn you that your applied load is really nonlinear...
Not knowing much about the wing example cited, I won't speculate too much except maybe it was the sensor engineer rather than the wing designer that thought that a long, riveted and/or glued structure that sags visibly under its own weight (and may have been filled with liquid fuel) "should have been a linear system"...
Evan Replies:
The wing example was actually a simple foot-long aluminum plate (shaped sortof like a wing), with a bunch of sensors, and lightly loaded within the elastic range by a lead-screw with a force sensor at its free end. It was a little demo system, to show off National Instruments' software. Just looking at it, it looked like it would be linear -- but it wasn't.
By the way -- I admit to being (shudder) an electrical engineer. So feel free to make fun of me if and when I stay stupid things that are obvious to mechanical engineers.