Wheelbase filtering effect on the frequency response of a vehicle

[NP32]

Hello all,

I am carrying out some measurements on a vehicle using accelerometers placed on each of the wheel hubs, each of the suspension towers (body side) and on each seat rail.

My investigation involves driving down a defined stretch of road at a fixed speed with a known frequency input characteristic and determining the frequency response of the vehicle by producing FFTs and transmissibility plots using the measurement data collected.

Attached is the typical frequency response measured at a wheel hub, suspension tower and at the seat rail.

I understand the first two plots for hub and body response but not for the seat response. I am lead to believe that the multiple large peaks are due to the harmonics associated with the pitch natural frequency of the vehicle. But why is a similar response not measured at the suspension tower? Why are the harmonics associated with the heave and wheel hop modes not as evident at the suspension towers? Is it due to a lack of damping in the pitch mode?

I apologies for the long explaination!

 

[GregLocock]

The tower response looks ok, the other two look plain wrong, not enough below 5 Hz. Are these in the vertical direction?

Your explanation for the seat rail plot is dubious, but since the data looks wrong then I'm not going to invest much thought in it.

Wheelbase filtering, as mentioned in your thread title, is easy enough to work out, what is your wheelbase and vehicle speed?



Cheers

Greg Locock


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

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

 

[NP32]

Yes this is the response in the vertical axis for each accelerometer.

I'm surprised with your comments on the hub response, seeing as this is measured on the unsprung mass, I thought it would be normal to expect a greater concentration of energy around the wheel hop frequency? (16/17 Hz from what I can see)

The wheelbase is 2.6m and the vehicle speed is 30mph.

 

[GregLocock]

OK I've got spindle vibration for all sorts of roads, not too sure about strut top. I've got seat track for a few roads. I'll have a look tomorrow.

I don't much like the 'harmonics of pitch mode' explanation, FWIW.



Cheers

Greg Locock


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

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

 

[BrianE22]

Just curious - I'm used to seeing spectral densities. You have, what, one frame of an FFT? An average of FFTs?

 

[GregLocock]

I think these are averages.

The peaks in the seat track response don't line up with the wheelbase filtering effect (5.16 Hz and integer multiples).
I never see zero at several frequencies on a body response. I haven't seen a road spectrum that looks like that- I'm also guessing this must be a very smooth road.

There doesn't seem to be much tire first order (fair enough) on the spindle.

wheelhop (15.7 hz?) on the spindle seems too high in frequency and too well damped, but isn't coming through into the body.

Sorry I haven't got anything very positive to add.




Cheers

Greg Locock


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

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

 

[NP32]

Yes you're right, the FFTs are averaged over 3 runs, apologies for not mentioning that before.

This is a highly damped 2 seater sports car running on a relatively smooth road.

Thanks for your help.

 

[GregLocock]

Incidentally brian's point about units is very valid, I can guess what you mean, but until it is in a psd, it is guess.

Cheers

Greg Locock


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

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

 

[BrianE22]

ok., still curious about your FFT calculation. Generally, the FFT calculators take in only a limited, factors of 2, number of points at a time, say 1024 points, 2048 points, etc. For your maximum frequency of interest of 25 hertz you'd probably want to sample at a minimum of 100 hertz and assuming the run was at least 10 seconds or so you'd have over 10000 points for your FFT. I guess that's do-able, just not what I'm use to seeing.

The reason I ask is to rule out that that the bumps in your FFT plots aren't a result of an FFT anomaly. I see you used a square top window, did you try a hanning window to see if the results looked different? What is the frequency resolution of your FFT?

 

[GregLocock]

He's almost certainly using ensemble averages, each frame of 4 seconds length. I'd guess 50-99% overlap, with about 30-60 seconds of data.





Cheers

Greg Locock


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

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

 

[GregLocock]

Here's some data from the body side of a car. It isn't running at constant speed, so that might explain why there are no dips to zero amplitudes, but I don't think so.

The accelerometer is high pass filtered at 3 Hz. I think the vertical scaling is right but it isn't my data.

This shows the sort of behaviour I am used to seeing, jagged (this is 1000 seconds of data), one lump of response at primary ride (0-3 Hz), another at engine bounce (say 6-9 hz), and then wheelhop at 10-12 Hz.





Cheers

Greg Locock


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

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