Correlation of tyre/road noise measurements via machine learning algorithms

Tyre/road interaction is the main cause for noise emission
when considering road vehicles travelling faster than
approx. 60km/h. A multitude of aspects play a major
role regarding the overall observed sound pressure
level (SPL), i.e. speed, vehicle type, driving style and
pavement type/condition. The latter, however, facilitates
the possibility of independent tuning of acoustic
properties, e.g. via diamond grinding of concrete pavements
or open-graded asphalts. Investigations regarding
possible emission reductions are often performed using
trailers, wherein installed microphones record the rolling
noise close to the tyre/road contact. Another option is
the analysis of isolated, statistical pass-bys of cars and
trucks. This paper investigates the relationships between
different trailer-based measurement techniques according
to RVS 11.06.64 (RVS-method) and to ISO 11819-2
(CPX-method), as well as between the CPX-method and
the Pass-by-measurements following ISO 11819-2 (SPBmethod)
[1–3]. Comparisons and correlations of pavement
dependent overall SPL and analyses of respective
third-octave bands will be presented. This includes the
use of unsupervised machine learning algorithms allowing
the determination of relevant frequency bands of either
method.