Development of a criteria for squeal noise detection applicable to on-board noise monitoring
Typ
Examensarbete för masterexamen
Master's Thesis
Master's Thesis
Program
Sound and vibration (MPSOV), MSc
Publicerad
2024
Författare
Lennartsson, Jesper
Vedin, Axel
Modellbyggare
Tidskriftstitel
ISSN
Volymtitel
Utgivare
Sammanfattning
Railway vehicles may generate a high tonal noise commonly referred to as curve
squeal during navigation through curves. Typically, this noise is characterized by
a single high pitch frequency accompanied by a few overtones. Its tonal character
makes it more irritating for a listener than a broad-band noise at the same sound
pressure level. This is a contributing factor to why curve squeal is perceived as being
so disturbing.
Curve squeal is produced as a result of self-excited wheel vibrations which are trig gered by large magnitude lateral forces developed in the contact between the inner
wheel and the low rail. Another type of squeal noise is called flange squeal which
instead originates from the outer wheel. In comparison to the tonal curve squeal,
flange squeal has a more broadband character.
This study is based on noise data recorded by an onboard monitoring system during
one month of traffic on the Green line of the Stockholm metro. Five different meth ods of detecting curve and flange squeal are evaluated. The methods accounted for
are: (A) an implementation of the algorithm for curve squeal detection in operation
at the Stockholm metro today, (B) an algorithm taken from literature developed by
a research group at the university of Wollongong, Australia, (C) the application of a
tonality module available in the commercial software ArtemiS SUITE, (D) a method
devised by the authors. Finally, (E) a Hybrid method that combines method A and
B. Method A is based on the sound pressure level difference between outer and inner
wheels while B evaluates the spectrum of the inner wheel and looks for the highest
1/24 octave band to compare against a set criterion. C calculates the tonality of the
inner wheel and marks when it is above a set limit. D calculates the sound pressure
level and compares it against a threshold. All methods except C are implemented
in Matlab whereas the software Artemis SUITE is used to evaluate method C.
The results showed that most of the methods did not produce accurate results rather
containing many false positives.
Method A performed best with respect to curve squeal detection whereas the so called Hybrid algorithm, containing criteria from both methods A and B, showed
best abilities to detect flange squeal.
Beskrivning
Ämne/nyckelord
Squeal noise , curve squeal , flange squeal , detection algorithms