1

I've been trying to find established/novel methods for calculating ride comfort using accelerometers and recently I came across this paper which uses ISO 2631 standard to get an objective measure of ride comfort.

Upon reading the paper, it highlights that the standard objective ride measure does not correlate strongly with the subjective ride measures. Hence, the authors use NN to figure out a better correlation.

The objective measure the standard uses, as discussed in the paper(I did not buy the ISO standard yet), is the running root mean square(RMS) of acceleration to calculate weighted acceleration($a_w$). This method considers the occasional shock and transient vibration through the use of a short integration time constant.

enter image description here enter image description here

(I think the equation is wrong because the power outside of integration should be half)

My questions:

  1. Has anyone here used ISO 2631 to estimate ride comfort?
  2. When I perform running RMS on 20,000 acceleration data points, instead of getting smoothed 20,000 points, how do I get a single/cumulative RMS or $a_w$ to estimate ride quality using table 5?
MajorMajorMajorMajor
  • 31
  • 2
  • So Formula 1 or Rolls Royce? – Solar Mike May 03 '21 at 18:21
  • @SolarMike F1 hypothetically. A good follow up question – MajorMajorMajorMajor May 03 '21 at 18:29
  • Then it was not measuring ride comfort but maximum grip... – Solar Mike May 03 '21 at 18:44
  • @SolarMike No I'm measuring ride comfort – MajorMajorMajorMajor May 03 '21 at 18:46
  • (2) you should be able to just average all the windowed RMS points. Perhaps a trimmed mean. If any weighting is to be done, I'd consider applying a filter up front to weight for whichever frequency components humans find most annoying. – Pete W May 03 '21 at 21:04
  • @PeteW yes, the authors of the paper used a butterworth bandpass filter for that – MajorMajorMajorMajor May 04 '21 at 12:06
  • hmm interesting choice. butterworth is underdamped if I understand correctly. probably doesn't matter tho – Pete W May 04 '21 at 13:16
  • @PeteW The paper also mentions frequency weighting and I posted a question on signal processing stack on that. I do not understand what weightings mean. Can you check it out? link – MajorMajorMajorMajor May 07 '21 at 11:52
  • looks like you construct a bandpass filter which does the weighting, run your data through that, followed by the RMS or the analogous 4th-power expression. – Pete W May 07 '21 at 12:20
  • @PeteW Then Is filtering the data with weights possible in post processing the data meaning after collecting the data without any filters. And then write code, to do the filtering? or the filtering should be done real time while recording the data? – MajorMajorMajorMajor May 07 '21 at 13:15
  • yes of course, in fact it's easier in post, because you can use non causal filters (ie you can run time both forward and backward in successive passes, which makes it trivial to get filters with no phase lag) – Pete W May 07 '21 at 14:44

0 Answers0