MECHANICA: How do I correlate results to test data?

[Flask]

How should I modify my model to match test data results?

I've completed a modal analysis and dynamic frequency analysis of an assy. I have graphs plotting frequency vs acceleration at specific points.

I also have vibe test data from a sine sweep test.

My Mechanica results are close, but a little different than the test data. How should I calibrate my model so that is matches the test data?

(1) If the acceleration response is too high, should I increase the damping of the assembly?

(2) If the frequency is too high, can I add mass to try and bring the Fn down? Or should I somehow subtract features to decrease the stiffness?

As far as I understand, I can modify
A) stiffness (modify geometry)
B) mass (change densities or add/subtract geometry)
C) dampening -- in the dyn freq analysis definition)

I'm just not sure what the best practices are for changing the behavior of the model.

Thanks!

PS - "Part I" - initial thread where I got help setting up my models:

http://www.eng-tips.com/viewthread.cfm?qid=296658&page=1

 

[Twoballcane]

Hi Flask. I don’t think there is a best practice method other than experience and intuition. Thus from my experience:

1. First you will have to verify that the UUT is correct in fit and form and that the data from the accelerometers truly make sense. An accelerometer’s data that sits in the middle of the cover will have different data than if it was placed near a screw. Also, I would go as far as using torque values to fasten the hardware than willy nilly tighten everything down. You get more and more tired from the first to last screw thus causing unknown effects. Also, you may want to do a survey of the fixture to make sure it is not causing any forced frequencies.

2. Once you have a good UUT and data that makes sense, you can compare this to your results. If your FEM results are off by 10%, than I would not worry too much about it. If your FEM results where off by 30% or more, than there should be something obviously wrong between the UUT and FEM. Now for that pesky 20% or less, you can use a few tricks to push your FEM in the right direction. The one place you should start to check is how the model is constrained and how many degrees of freedom you are modeling. When you constrain, you want the model to be loose as possible. Even though you think a cover with all of the screws used, that the cover is fixed on all sides, this is not the case. When the UUT is excited, everything starts to act like wet noodles and start to act as if everything is loose. So on the model, you can constrain in directions that make sense and have free rotation in all axis. Now for degrees of freedom. If you are modeling the whole UUT in one model, you will have more degrees of freedom which will cause the math to skew. Also, ProM “fuses” all of its parts together making it artificially stiff. It is better to model each part of the UUT independently and study its effects.

3. For what you’ve mentioned, manipulating the stiffness (Youngs modules) and weight (density) will be the last resort. However, make sure that you are using the correct Youngs modules and density. I’ve even create my own material instead of getting them from the library. I’ve been bitten once or twice in the past that the material from the library was wrong (not sure why).


Tobalcane
"If you avoid failure, you also avoid success."
“Luck is where preparation meets opportunity”