Odd acceleration data at high sampling rate 1

[ProjEngAnalyst]

Hi all,

I have some acceleration data that was taken from a very stiff location near the rear wheel mounting position on a 4 wheeled vehicle. The terrain traversed is mild to moderately aggressive (most aggressive would be a curb hit at approx 10 mph). The vehicle has no suspension, thus the drivetrain is hard mounted.

This data is sampled at 10000hz.

My concern is that it shows spikes (approx 40-50) of 60 g's whose duration are about 5 ten-thousanths of a second. This kind of acceleration load resolves to a force of something like 60,000 lbf. No way.

If the data is resampled at 50 hz it looks alot more appropriate, albeit because of the effects of aliasing. (MTS likes 50 hz data, which is why we resampled at this rate.)

I am looking to use this data in order to perform a FEA analysis by transposing the loading from the historical data into static cases and using the cumulative damage method to determine an approximate life.

If I apply a 60,000 lb static load to it, its just going to blow up. This cannot be the case.

Does anyone have experience taking accelerometer data and possibly know what the source of these odd spikes are?


Any input would be greatly appreciated.

Thanks,
Nick

 

[rb1957]

my 2c ... a accelerometer measures 50g doesn't equate to a static load of 50*Wt. what's the duration of the spike ? < 1/100th of a sec ?? I think sampling at 50 Hz is not fast enough, that these events last < 1/50th of a sec. IMO a 50 Hz sample will reasonably capture a signal with a period of about 1/10th of a sec.

you're reading experimental data, and know that the experiment didn't break the truck. therefore you know (well, to a reasonably high degree of certainty) that no impossibly high load was applied to the truck.

another day in paradise, or is paradise one day closer ?

 

[GregLocock]

"This kind of acceleration load resolves to a force of something like 60,000 lbf. No way. "

I agree, no way. High frequency (ie short pulse length) high g spikes are not a durability issue. You cannot turn them into equivalent static loads. Specifically they are decoupled from the main mass of the vehicle because you are seeing the local modal mass at that frequency and location, not the rigid body.

Using accelerometer data for durability loads for vehicles is actually pretty difficult, some people seem to claim to do it, but we currently measure forces and/or strains.

Since your FEA model is completely useless at 200 Hz, never mind 1000 Hz, it seems unwise to drive it with 10000 Hz signals.

Cheers

Greg Locock


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

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

 

[zeusfaber]

How secure is the attachment of accelerometer to structure?

A.

 

[IRstuff]

When you run the the spike through the transmissibility, you probably wind up with close to zero. If you want to consider the spike, you need to look at the equivalent energy, i.e., 60 g at 500 microseconds.

TTFN
faq731-376
Need help writing a question or understanding a reply? forum1529


Of course I can. I can do anything. I can do absolutely anything. I'm an expert!
There is a homework forum hosted by engineering.com:

http://www.engineering.com/AskForum/aff/32.aspx

 

[gruntguru]

Greg nailed it. You are measuring local accelerations of one small part of a flexible system, not the entire vehicle.

You can either filter the data or mechanically filter the accelerometer using a softer mounting and adding mass to the accelerometer. Either method is pretty hit and miss unless you have alternative measurements to validate the results.

je suis charlie

 

[MikeHalloran]

The vehicle has no suspension, thus the drivetrain is hard mounted.

As Greg alluded, not quite true.
The 'suspension' comprises flexure of the vehicle itself.
Past related discussions, if I recall correctly, suggest that you may be able to limit, or better control, the worst stresses, and maybe save weight, by adding a primitive suspension.

I went looking for an example, and couldn't find a trace, of an odd small tractor I saw maybe 20 years ago, comprising a simple sheet steel tub-like frame, each corner decorated with a post having knee-action articulated wheel at the bottom, with all wheels driven and all wheel steerable, and all driven by long v-belts, from a very modest engine. The suspension travel came from the articulated arms carrying the wheels, and from gross twist of the frame itself. It looked more like a kitchen table than a tractor, but it was very well suited for New England's rocky soils and hilly terrain. I think it was made in Vermont, or somewhere nearby, and the articulated wheels were the subject of a patent. I'm guessing that, with all the exposed v-belts, it had too many pinch points to sell today.




Mike Halloran
Pembroke Pines, FL, USA

 

[ProjEngAnalyst]

All,


Sorry for the delay.

"you're reading experimental data, and know that the experiment didn't break the truck. therefore you know (well, to a reasonably high degree of certainty) that no impossibly high load was applied to the truck."

This is absolutely true. Which is why I feel that something is up with the data.

"I agree, no way. High frequency (ie short pulse length) high g spikes are not a durability issue. You cannot turn them into equivalent static loads. Specifically they are decoupled from the main mass of the vehicle because you are seeing the local modal mass at that frequency and location, not the rigid body."

Well, not really TURNING them into static loads. Just using a general method to back out a rough estimate of life in order to move forward with design. The cumulative damage method has been used successfully in this manner many times before. We just didn't have ridiculous data.

"Since your FEA model is completely useless at 200 Hz, never mind 1000 Hz, it seems unwise to drive it with 10000 Hz signals. "

I don't understand this statement.

"How secure is the attachment of accelerometer to structure?"

It is fastly secured.

"When you run the the spike through the transmissibility, you probably wind up with close to zero. If you want to consider the spike, you need to look at the equivalent energy, i.e., 60 g at 500 microseconds."

I am interested in this transmissibility phenomenon. Any references? I have done a quick google and understand the principle. I don't see any explicit analytical solutions to transmissibility. I presume it is related to material/geometric stiffness and driving/resonant frequencies. Interesting for sure.

"As Greg alluded, not quite true. The 'suspension' comprises flexure of the vehicle itself. Past related discussions, if I recall correctly, suggest that you may be able to limit, or better control, the worst stresses, and maybe save weight, by adding a primitive suspension."

Well, there may be a small amount of deflection due to loading, but it sure isn't much. This thing is seriously stiff. A suspension would be out of the question, tight competition.



I appreciate all of the comments. If you have anything else to add, please to so.

I still would like to speak with someone who is familiar enough with accelerometers to difinitivley* tell me if there is some sort of measurement anomaly or not.

Thanks,

Nick

 

[rb1957]

I think we're saying (as far as we can) that your data is probably good. if you collected data 1000/sec then you probably have a good record of what happened. but when you say "60 g's ... This kind of acceleration load resolves to a force of something like 60,000 lbf" then you are making a static equivalent of a dynamic transient. To understand the internal stresses experienced (for a fatigue calc) then you probably need a dynamic FEA.

as alluded to above, the key issue is the energy imparted into the vehicle by the acceleration. maybe something worth doing is calculating displacements due to the input.

another day in paradise, or is paradise one day closer ?

 

[GregLocock]

How have you validated your FEA model at 200 Hz? 1000 Hz? 10000 Hz?

GIGO



Cheers

Greg Locock


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

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

 

[ProjEngAnalyst]

I haven't. I won't even be setting up the model. We have a more experienced analyst that will be doing it. I am somewhat of his protege. We have been discussing this data and neither of us feel comfortable with it.

BTW, we have been using the cumulative damage method in this manner for a long time with good results. We do look to move forward to FE Safe in the near future but we have a project that we can't risk a new type of analysis on. I will probably perform the FE safe analysis after we do what we are familiar with.

I'm not sure why it would need to be validated at those frequencies anyways, the 10000 hz sample is not the frequency of the events.

I forgot to mention before, the FFT of this data looks like a big bunch of mess. Nothing like you would expect with the events and corresponding frequencies.

 

[3DDave]

Perhaps the test setup needs to be moved to a more controlled system to validate the setup. Simply supported bar and a hammer or shaker table? You might have a broken lead or other noise source. It would also be good to have a second data source, such as strain gages to vote for the most likely loading. If they all agree, then it's probably good. If one is not with the program, then the others may reveal that.

For newbies just listening in - 10kHz is good for resolving 1kHz or less signals. While Nyquist suggests that a single signal will be aliased at any sample rate less than 2 times its frequency, that's the -frequency-, not the amplitude that is being discussed. Sampling at least 10 times the expected frequency is required to avoid seeing changes as shocks.

 

[gruntguru]

The results you are seeing are not unexpected. I see very similar results on an impact testing rig with an accelerometer quite rigidly coupled to the impact face of a 300 kg "hammer". Although the hammer is "rigid" for the purposes of the impact, it is somewhat short of being a "rigid" body. We see peak accelerations at the accelerometer at least one order of magnitude higher than the overall acceleration of the 300 kg body.

je suis charlie

 

[ProjEngAnalyst]

"The results you are seeing are not unexpected. I see very similar results on an impact testing rig with an accelerometer quite rigidly coupled to the impact face of a 300 kg "hammer". Although the hammer is "rigid" for the purposes of the impact, it is somewhat short of being a "rigid" body. We see peak accelerations at the accelerometer at least one order of magnitude higher than the overall acceleration of the 300 kg body. "

This is interesting, that you see extreme values within the data. I expected for the tire/ground interface to disallow this sort of behavior. But, maybe not. What sort of values are you seeing from this impact? Approximately what velocity is the hammer moving and what is the mass of the impacted body?

"Perhaps the test setup needs to be moved to a more controlled system to validate the setup. Simply supported bar and a hammer or shaker table? You might have a broken lead or other noise source. It would also be good to have a second data source, such as strain gages to vote for the most likely loading. If they all agree, then it's probably good. If one is not with the program, then the others may reveal that."

I agree, I think there should be some redundant instrumentation. Possibly even more accelerometers further up the chain to compare the data with as well. We could at least calculate transmissibility that way, Which would be useful to have visibility on.

 

[gruntguru]

300kg hammer with impact velocity 8 m/s, striking a concrete filled steel tubular beam, 165mm dia x 5.4mm wall thickness 3000mm long. Mass = 200 kg - supported at each end, impact at mid-span. Total deflection about 150 mm => average acceleration of 22g. Peak acceleration approx 200g. Peak acceleration at accelerometer location approx 2000g.

je suis charlie

 

[ProjEngAnalyst]

Very good, Thanks.

So how is it that you have resolved the peak accel of the member vs the peak accel of the instrument?

 

[gruntguru]

To clarify, the accelerometer is fitted to the hammer, with good coupling to the impact point. The actual acceleration of the hammer CG has not been accurately verified. The tests were modelled in LS-DYNA with strong validation of the model provided by several other measured parameters. Mechanical filtering of the accelerometer (by trial and error adjustment of compliant mounting material) produced an acceleration time history that correlated well with the model. Peak acceleration was approx 200g.

The 2000g peaks were very short duration and clearly due to local deformation of the hammer face rather than true acceleration of the hammer CG. We are investigating alternatives for instrumenting the hammer (accurate inference of impact force is the goal) - probably using laser displacement measurement.

je suis charlie