Different Analysis Methods for Centrifugal Fan Noise

1
Different Analysis Methods for Centrifugal Fan
Noise

• A. Touhafia, G. Pispolab , M. Raadscheldersa and
  F. Asdrubalib
• a Erasmushogeschool Brussel dep. IWT, Belgium
• bDepartment of Industrial Engineering, University
  of Perugia, Italy

2
Outline

• Introduction and Goal
• The Studied Fan
• Used Measuring and Analysis method
• Measured Data
• Conclusion

3
Introduction and Goal

• Measuring Fan Noise is traditionally done using a
  pressure microphone on a distance of one meter
  from the fan.
• This technique gives the necessary measures that
  allows one to quantify the produced noise.
• However, the generality of this measuring
  approach doesnt result in information about the
  sources for the produced noise.

4
Introduction and Goal

• Achieve acoustical data of an Industrial Fan in
  order to retrieve a fingerprint that is related
  with the physiscal underlying phenomenon.
• Spectral analysis solely is not enough.
• Combine Spectral, temporal and directional
  information.
• Look at Pressure and Intensity measurements.
• Consider techniques like wavelets to analyse Self
  similarities, Discontinuities, Pure Tones and
  Short Term Evolution of the noise.

5
The Fan

• Centrifugal fan with forward-curved blades.
• Medium sized machine with-Double inlet-Nominal
  speed of 1260 rpm -Max volume flow
  rate 6670 m3/h

• Single-phase asynchronous integrated motor with
  variabel rotation speed
• Operational configuration for the measurements
  Free inlets - Free outlet
• Configuration that does not imply any ducts
  connected to the fan

6
Measuring and Analysis Methods

• Measuring Methods
• Traditional pressure measurements EUTERPE probe
  in a semi anechoic room
• (EUTERPE is a PC based class 1 Sound Level
  Meter)
• 2) Intensity meaurements Bruel Kjaer intensity
  analyser in a diffuse labroom

7
Measuring and Analysis Methods
Analysis Methods 1) The EUTERPE analyser to
calculate the pressure based acoustic properties
and spectral information. 2) The BruelKjaer
Noise Source Location (NSL) software to analyse
the measured intensity and to map it on a
grid. 3) The Mathlab wavelettoolbox
Waveletanalysis on measured soundfragments.
8
Measured data
Intensity map calculated by the software NSL
(BruelKjaer) from the measurements with a 50 mm
spacer
9
Measured data
Main Noise Sources -The inlets (red and purple)
-The housing (green and orange)
Intensity map calculated by the software NSL
(BruelKjaer) from the measurements with a 50 mm
spacer
Impossible to make measurements on the front
side a flow impinging on the probe influences
a lot the results
10
Measured data
The Effect of the Generated Air Flow in the Front
of the Fan
11
Measured data
The Effect of the Generated Air Flow in the Front
of the Fan
The lowerfrequency components are influenced.
-Very difficult to put a link between low
frequency noise created by the machine and the
measured data
12
Measured data
Grate with holes is put in front of the wind
generating parts.
-Measurement is less influenced by the airflow -
New frequency components become visible in
lowfrequency part- Largest among them (100 Hz)
is related to the motor rotational speed.
The effect of a grate in front of the fan
13
Measured data
Continuous Wavelet Transform with Morlet Wavelet
on -Pressure signal-1000 samples -Anechoic
chamber-sample frequency 48KHz
This is the results of Continuous Wavelet
Transform with Morlet Wavelet
14
Measured data
Continuous Wavelet Transform with Morlet Wavelet
on -1000 samples -Pressure signal-Anechoic
chamber-sample frequency 48KHz
Scale
Time
Amount of correlation with wavelet signal
15
Measured data
The self-similarities that one can see are
related to the fractal nature of the signal. Can
be related to main cause of fan noise which is
turbulence
Turbulent boundary layer on the blades causes
chaotic forces on the blades that become sound
sources.
16
Conclusions

• This research has been able to stress on the need
  of further investigation on centrifugal fan
  noise.
• firstly on a definitive reliable measurement
  method and then through innovative analysis
  techniques.
• As far as signal processing, it is interesting to
  notice that frequency domain is not the only
  possible research mean even time domain and
  time-scale domain can supply useful information.

17
Conclusions

• In contrast to other analysis methods, such as
  Short Time Fourier transform, wavelets are able
  to reveal self-similarities at different scales.
  
• This reveals the fractal or pseudo-fractal nature
  of this kind of phenomena.
• Actually, it is worth noticing that aerodynamic
  noise is substantially chaotic in nature, being
  caused by turbulence.