Likelihood of SRS overtest/undertest?

[jvian]

I have recently had some vibe and shock testing performed by a local test house and have ran into a question of test tolerance and the likelihood of over-test/under-test. Up to this point my experience deals with random vibration testing and classical sawtooth impulse shock testing; however, this round of testing utilized a shock response spectrum (SRS) which is new to me.

I feel comfortable with the concept of the SRS in general and have some (though limited via my own research) understanding of its derivation and also per MIL-STD-810G feel comfortable with the testing process. First input a reference SRS, this derives the reference transient time domain pulse, followed by the "actual applied" transient pulse via the shaker table, and finished with the "actual applied" resultant SRS (synthesized from the "actual applied" transient pulse).

When performing the test the technician was forced to widen the tolerance band of the reference SRS (from +/-3dB to +/- 6dB) to allow the shaker table to converge on a transient impulse. My requirement was to be within the +/- 3 dB band however if increasing to +/- 6 dB helped while not under-testing the unit, I would accept the results as a pass as well (with some robustness to over-test). However when the test was complete we never received any charts or data (nor can i get that data at this point) on the "actual applied" SRS and therefore have now way of knowing whether or not we over-tested or under-tested our unit by widening the tolerance band. In my experience (~3 years) with the dynamic world of shock and vibration every example or test has always resulted in either a "within limits" satisfied test, or an "acceptable over-test" condition due to dynamic amplification. I have never seen a system attenuate an input.

My hope is that someone here may be able to provide some insight as to the likelihood of my test being an over-test or an under-test. I have since passed a more severe crash safety shock and am confident the design is acceptable but would like to run this to ground for my own benefit and to widen my experience if possible. Had I not performed additional and more severe testing I feel an experienced "judgement" call would have been required.

Any help or insight would be appreciated and thanks in advance.

-J-

 

[btrueblood]

" However when the test was complete we never received any charts or data"

Well...how are you going to prove to anybody else that the test was conducted? Sounds like perhaps the shaker table operator knew there were problems, and buried the data?

 

[jvian]

Btrueblood,

I received some but not all of the data. I have the input reference SRS, the synthesized transient impulse and the "actual applied" transient impulse. What I did not get was the "actual applied" SRS (which would have been synthesized from the transient impulse). My hope was to have received the applied SRS and overlay the reference to determine if the test was high or low. Since by nature an SRS can synthesize an infinite number of time domain pulses I cannot use the time domain data that I have by itself (or at least I am not aware of how).

-J-

 

[btrueblood]

Um, I think you CAN reverse it.

i.e. i think you are right - you can synthesize an infinite number of time domain pulses that will give the same SRS, or put another way you can't definitely determine the time history given an SRS (this is directly from Cyril Harris, Shock and Vibration Handbook, 3rd ed., McGraw-Hill, 1987, p 23-9.

But any given input time history will only yield one SRS, for a given test object's known (or assumed) natural frequencies and damping...in fact the time history is how you calculate the SRS.

If you have the a(t) data values (FM tape?), you should be able to generate the SRS. Talk to the lab guys again, it should be do-able.

 

[GregLocock]

While you are at it remember to update your test spec so that all the info you might need is part of the deliverables.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[jimkinney]

If you don't have the data from the control accels mounted to the shaker head (or test fixture), or response accels mounted to the DUT, it doesn't count.

Inputs to the shaker rarely matches what the DUT actually sees.

Jim

Jim Kinney
Kennedy Space Center, FL

 

[jvian]

Thank you all for your responses but my main question is still unanswered. Is it more likely to be an under-test or an over-test? My gut feel is that it was more likely an over-test simply due to the fact that most of my experience tells me that dynamic systems are harder to control on the high side due to resonance. In the end I am looking for an "industry opinion" as to weather it is more likely that my test was harder to control within limits on the +3dB side or the -3dB side. I did speak with a field engineer for the software that was used for the test (spectral dynamics) and he seemed to have the same gut feel that it would be harder to limit an amplification than to boost up an attenuation. It has been a good learning experience especially with what is required from the test lab.

Thanks again for your responses.

-J-

 

[StrykerTECH]

Jvian,

StrykerTECH performs SRS testing for customers, SRS pulses are tricky little animals. They take for ever to range the system. So what generally happens is the system will range running lower level -12DB or -18DB pulses until it dials in the profile to your specific pulse pattern. The bad news is, table type along with fixture design play a large roll on the performance characteristics you can achieve. With out see your data it is hard to draw a conclusion but what I can tell you is where I find it hard to match up the pulse it at the peaks and the distance between the peaks. This is directly related to the velocity capabilities of your table. If I had to guess I would say it is an Under-test. What is probably happening is instead of getting sharp elongated opposing U's to match your profile you probably are getting more squatty triangular shaped opposing U's. I have attached a crude (got to love MS-paint) picture explaining what we usually run into with tough to range SRS pulses. This is most likely because your table can't perform the way it needs to with the load. Of course I am assuming your are doing this on an ED shaker table

StrykerTECH Engineering Staff
Milwaukee, WI

http://www.stryker-tech.com/

 

[jvian]

Stryker,

Thank you for your response. The "ranging" you mention was the same process that we performed (but called it convergence) with a -12dB input. I have attached a couple of images of our test data and was hoping you could provide some insight as to the difficulty or "ugliness" of my reference SRS. You mention the peaks and distance between the peaks as being difficult to match up and I assume that you are referring to how sharp and frequent the peaks are as being difficult. I was surprised by your images that the lower frequency pulse is worse than the higher frequency pulse.

If possible please look at my figures and let me know if you think you would have a hard time matching the SRS and/or what your thoughts are as to the overall data set.

Again I thank you and the others for your time.

-J-

 

[StrykerTECH]

Jvian,

Sorry I should clarify, the image I sent was suppose to be the same pulse, but the image on the bottom was the best attempt for a table to replicate the pulse image above. obviously not to scale. It is a problem we often run into when trying to replicate SRS pulses. The sharp transient response up and down is hard to replicate.
The Convergence is the correct terminology, We generally refer to it as "ranging" simply because it is trying to operate in the response range we desire. As you increase the acceptable DB range you allow that ramp up and down to be less and less steep.
Looking at the pulse the transient response on the first half of figure two is going to be almost impossible to replicate with any kind of load on the table. It has really jagged edges. Almost like it was recorded. Like a hammer hitting a bell. The second half looks o.k. but still would need a little smoothing. Hopefully we have been of assistance.

StrykerTECH Engineering Staff
Milwaukee, WI

http://www.stryker-tech.com/