EMA: Measuring Material Aging with Experimental Modal Analysis

[skrat]

We are looking into Experimental Modal Analysis in one of my classes at university. The goal is to measure the Frequency Response Function and from it determine the mass and stiffness modal matrices.

But my question is the following: My professor is leading lectures with his measurements that are roughly 3-4 years old. And the first eigenfrequency of that beam at that time was about 320 Hz. Today the first eigenfrequency is 311 Hz.

So my question here is plain and simple (I did not ask my professor that because at that time I didn't think of that): Is it possible to estimate the material aging/condition by doing the EMA on a structure? Any papers on that topic?

Cheers

 

[anorlunda]

It sounds promising to me as a diagnostic tool. But no, I never heard of it being done or being studied.

 

[Work Hard Play Hard]

I can't suggest any actual studies but it's hard to believe it isn't out there. What I have seen is documentaries comparing the old to new. Particularly I'm thinking of a documentary I saw on joinery used in Japanese construction through time and how the study of techniques from the past are being used in skyscrapers to meet earthquake building codes. I doubt a documentary about data collection of the different forces on buildings during earthquakes would hold a viewers interest like the stories and empirical evidence does. From the documentary I remember the single center pole design of Pagodas being studied and a skyscraper being designed and built on the same principles because they were once the tallest buildings in Japan and they don't come down in earthquakes. What you are looking for might come from specific queries rather than generic queries.

If you don't find what you are looking for the automotive industry is full of studies on the effects of aging and time. In the last week I have seen studies on the aging and life cycle of glass batteries compared to lead acid batteries and the sun's effect on the degradation of paint and interiors over time. Obviously what you are looking for isn't in the automotive industry but you might find insights into the methodology used you could reverse engineer for your applications and you also might find some references to more relevant information to suit you. It's not often you come across the wheel being reinvented.

 

[dawin]

When I was in college there was a modal research group which was looking to correlate FRFs to structural damage, which you could argue could include effects of aging. In general, you're assuming that damage affects local stiffnesses which in turn affect your final modal response. I think what you've proposed is in theory doable, and I believe some people do use modal information to monitor structural health now. However, I didn't follow my University's progress, but the methods in general require some model correlation to an ideal structure and historical data, and some assessment of the measurement uncertainty. My experience with modal testing is somewhat limited, but the results you get can be heavily influenced by the quality of your test setup and execution, and I think in any case you need to know the sensitivity of your method (i.e., how gross does the damage need to be to push the frequency response to something that is statistically significantly different?).

311 Hz is about 97% of the original measurement. That's not extreme. I'd suggest while you learn this topic you look into what you expected vs. what you observe and go through explanations why this might have happened. For example, what's the beam material (this is critical, especially if it's a composite beam), and how is it mounted? What mode do you think you're measuring and what do you expect its natural frequency to be? How would a change in material properties or geometry affect your modal resposne? How does measurement location affect your measurement? How does beam mounting (i.e. boundary conditions) affect the final modal response? If you used an accelerometer, what is its sensitivity, accuracy and linearity; what would be the extreme differences in results that are still considered "good" using this sensor and data acquisition setup?

If you wanted to have some fun, you could even make a finite element model of this beam and see if the 320 Hz matches your predictions, and then add some localized damage (e.g. set some areas to zero stiffness) and see how it if affects your response.

Have fun!