Está en la página 1de 5

Proceedings of the 2009 IEEE

International Conference on Mechatronics and Automation


August 9 - 12, Changchun, China
Motor Noise Source Identification
Based on Frequency Domain Analysis
Qingnian Wang, Kun Yan and Hongyan Li
Department ofinformation engineering
University ofNanchang
Nanchang, Jiangxi Province, China
sophia_ncu@126.com
Abstract - The problems on motor noise source identification
are discussed in this paper. The reasons of motor vibration and
species of motor noises are introduced. The advantages and
disadvantages of some usual noise source identification methods
are also clarified in this paper. As an emphasis, the basic theory
and methods about frequency domain analysis are studied. Use
MATLAB software to process collected data and do frequency
domain analysis. Do an experiment to confirm the frequency
domain analysis method of this paper, and the experiment shows
that this method is effective in motor noise source identification,
and can use to practical situation.
Index Terms - noises source detection, frequency domain
analysis, spectrum analysis.
I. INTRODUCTION
Motor is the main source of power presently, and it is also
the important equipments in people's daily life. But motor
noise takes people a lot of inconvenience during production.
So, motor noise source detection and motor noise eliminating
are becoming a pop problem. The motor noise is a admixture
of different frequency and different intensity noise, and
sometimes can go upwards to 110 decibel[I]. This affects
operators' health seriously. To control motor noise effectively,
detecting and searching main noise source are very important.
II. MOTOR NOISES DETECTION
A. Reasons ofMotor Vibration
Motor is a kind of energy transforming setting. Its
vibration comes down to every parts of it, because of the
complex configuration. The research of motor vibration relates
to a lot of aspects, such as electromagnetism, energy
transforming, mechanism vibration etc.. Generally, the reasons
of motor vibration includel";
1) Electromagnetism Force: Electromagnetism force acts
in the air gap which between stator and rotor of the motor, and
its force wave is revolving or pulsant in the air gap. The
magnitude of the force is related to electromagnetism load,
some configuration of motor's effective parts and calculation
parameter. The magnitude of the vibration which engendered
by electromagnetism force is related to the stator which in the
motor's surge system. the frequency of most motors'
electromagnetism vibration is in a range of 100""4000 hertz.
2) Bearing: The vibration which engenders by bearing is
related to a lot of factors, such as the quality of the bearing,
the machining precision of bearing copulate surface and the
978-1-4244-2693-5/09/$25.00 2009 IEEE 2373
Meiling Yuan
Department ofScience
University ofNanchang
Nanchang, Jiangxi Province, China
yankun 198404196666@163.com
machining precision of bearing cap's surface which towards
motor base etc..
3) Mechanism Imbalance of Rotor: The mechanism
imbalance of rotor can bring distinct vibration, especially
some high speed motors with 3000 revolutions per minute or
higher are more distinct.
B. Species ofMotor Noises
When motor running, there are kinds of noises source
exist at the same time. Generally, they can divide them into
three kinds: mechanism noises, electromagnetism noises and
air motivity noises. These noises are ineluctability.
1) Mechanism Noises: Rotor running unbalance is the
usual causation of engendering mechanism noises. When
motor fix not well, the connatural frequency of rotor and stator
is same with circumgyration frequency, then cause mechanism
noises. Mechanism noises take about 5% in total noises.
2) Electromagnetism Noises: When motor running,
electromagnetic force act on air gap between stator and rotor.
This engender circumgyration force wave, the wave cause
stator vibrate to engender noise. Electromagnetism noises take
about 20% in total noises.
3) Air Motivity Noises: There are three main components
of air motivity noises: circumgyration noises, vortex noise and
fluting noises. Air motivity noises is the main component of
motor noise, it take about 75% in total noises.
C. The Methods ofNoise Source Detection
Noise source detection is an analysis on every sound
source which exist on motor, to know why engender vibration
and noise; Then analyze the characteristic of sound source,
arrange the noises from big to small; Finally, confirm the
place where the vibration source exist and which is the main
noise source. There are several methods of noise source
detection in common use, such as surface vibration speed
measurement, sound intensity method, vibration
measurement, signals disposal method and so on. The
precision of sound intensity method is lowness, because
other noises affect it a lot in reverberation situation. So we
use sound intensity method to evaluate approximately.
Vibration measurement uses digital signals disposal
technology to work out the radiant efficiency of different
surfaces. But it is still hard to confirm the radiant efficiency
of different parts of motor accurately in theory. Surface
vibration speed measurement uses vibration measurement
to detect noise source. There is no requiring to any
qJ(f) is the function of phase'".
We can get every frequency component, distributing
domain of frequency and the amplitude or energy of each
frequency component via spectrum analysis. Accordingly,
we can get the main frequency and its amplitude which affect
the motor running. Base on a lot of experience and system
analysis, Table I collect motor vibration noise sspectrum
analysis. Table I provide favourable basis for examining
motor noise and vibration.
TABLE I
MOTOR VIERATION NOISES FREQUENCY SPECTRUM ANALYSIS
Suppose a noise vibration signal x(t) turns into X(f) by
Fourier Transform, namely:
X (/) = [ x(t)e- j2Jtji dt . (3-2)
Commonly, X(f) is a complex function,
make X(f) = R(f) +iI(f) , namely:
IX(f)1 = .JR(f)2 + I (f)2 . (3-3)
IX(f)1 is a amplitude spectrum function. So the estimate
value of power density (Sx) is:
Sx = IX(!)1
2

qJ(f) =arctg[I(f)IR(f)IJ
acoustics environment, so it is convenience and
practicality. Signals disposal method use sound source
signal time-domain values which are recorded in noise
measurement. Then use one certain signals dispose method to
analyze out main sound source signal according to a certain
characteristic of sound source signal time-domain. Frequency
domain analysis method, concern function method and
inverted spectrum method all belong to signals disposal
method, and they all have the strongpoint of exactitude. So
this paper uses frequency domain analysis to detect motor
noise source.
III. FREQUENCY DOMAIN ANALYSIS
Most circumgyration mechanism always engender
periodic vibration signal. This signal is not a single
frequency signal, but contains kinds of frequencies. These
frequencies components always have relationship with some
parts of physical characteristic of the motor. Take frequency
(F) as abscissa, amplitude (A) as longitudinal coordinates;
make a relationship fig between frequency and amplitude.
Analysing this relationship fig is called frequency domain
analysis. The frequency domain analysis used in this paper is
mainly about spectrum analysis. The aim of spectrum
analysis is using Fourier Transform to analyse the
complicated time course wave, and tum this wave into some
single harmonic wave. Use spectrum analysis, we can obtain
frequency configuration of the complicated signal and
amplitude or Phase information of every single harmonic
wave.
Spectrum analysis in noise source detection is to know
further about the composing of signal base on time-domain
analysis, and find the relationship between motor noise
signal and electromagnetism field or rotor, stator
configuration, to diagnose the place and reasons of
vibration. According to the character of signals and the
differences of the transform methods, Spectrum analysis
can be expressed by power spectrum, holospectrum,
amplitude spectrum, phase spectrum, Refinement spectrum,
inverted spectrum and so on. Power spectrum and
holospectrum are analysed in this paper.
A. Power Spectrum Analysis
Power spectrum is a description of signal energy or power
distributing. It not only describes some diagnostic frequency,
but also can describe the energy distributing status in a certain
frequency domain. Power spectrum contains self frequency
spectrum and frequency spectrum. Self frequency spectrum
contains the same information with amplitude spectrum, but it
is clearer than amplitude spectrum in the same condition. Self
frequency spectrum can be obtained by Fourier Transform of
interrelated function or calculating from amplitude spectrum.
Suppose a flash time-domain signal x(t) , its period is
infinite, and its math expression by Fourier Transform is:
F(f) = (2Jrfl/2 [ x(t)e-
JJtji
dt. (3-1)
2374
Mainly
frequency(Hz)
f inside 2kHz '""
5kH have obvious
peak value
f inside 1kHz '""
1.5kHz have obvious
peak value
f inside 500Hz '""
1kHz have obvious
peak value
f inside 50Hz '""
400Hz have obvious
peak value
Vibration parts
Noise and
vibration of
bearing itself
Noise and
vibration of
bearing's axis
orientation
Noise and
vibration of motor
shell syntony
Noise and
vibration of axial
fleeing
(3-4)
(3-5)
Reason analysis
l.Bad quality of bearing
itself
2.Improper cooperation of
bearing and bearing Room
I.No axial flexible settings
or the settings are badness
2.Moter techniques quality
axis orientation is locked
3.Radius of curvature of
bearing rolling road is small
Syntony with
electromagnetism
vibrational harmonic wave
I.No axial flexible settings
or the settings are badness
2.Big warp of techniques
quality
3.Eccentricity inside, outside
pellet is irregular or have
scar
Fig I Measuring points positioning

Time(s)
Obtain
Data
Motor Noise
-
collection
signal
1
Main frequency-
Signal
noise
-
domain
-
source analysis
disposal
N
<II
E0.1
C
0
e
..
0; \).1
u
u
<l: 02
o
Fig 2 Flowchart of experiment system
0.2
Holospectrum analysis gathers amplitude, phase and
frequency of dynamic signal in the space, integratedly reflect
the complete movement of research object inside its plane.
This describes the signal's movement character of different
frequency straight as well as exactly.
IV. EXAMPLE VALIDATIONS
In order to detect noises source, we analyze an
asynchronism motor with noise overproof. The type of this
motor is Y100L-2, rated power is 3 kilowatt, and rotate
speed is 2880 revolutions per minute. Firstly, put thre e
acceleration sensors which type is Le0103 on motor
bearing separately. The concrete places are axis direction,
radial level and radial plumb'' which are shown in Fig 1.
Point 3, 4 and 5 shown in Fig 1 are the concrete places.
Signals from sensors are collected, and then dispose the
signals to distill the parameters. Do frequency-domain
analysis on vibration signal to obtain the main noise source
finally. Fig 2 is the flowchart of the entire experiment
system. Analyze the acceleration signals which collect by
the sensors, and read the vibration signals of three points:
axis direction, radial level and radial plumb of motor
bearing. As shown by Figs 3, 4 and 5.
B. Holospectrum Analysis
Holospectrum analysis fully uses the phase information
which is neglected by power spectrum analysis, and dispose
the signal integratedly which transmits by sensors, that the
sensors are 90 degree apart in the space. This solves the
problem omitted by power spectrum analysis effectively.
Process of constructing holospectrum:
I) Vibration signal which transmits by single sensor is
decomposed into harmonic component by improved Fast
Fourier Transform Algorithm (FFT) arithmetic.
2) Dispose the harmonic signals integratedly , that the
signals are the same frequency of uprightness and level in the
same bearing surface. Then put that into a movement tract.
The figure of holospectrum changes into ellipse from
traditional figure, and the ellipse figure have some
characteristics of Eccentricity, obliquity and veer besides size.
So the ability of trouble distinguish ability is improved a lot.
This paper uses planar holospectrum (PH). PH bases on
spectrum. Its abscissa is frequency of vibration, and its
ordinate is rev. On PH, one beeline is composed by level
parameter and apeak parameters which phase difference is 0
degree or 180 degree. The obliquity of this beeline depends on
the amplitude ratio of the two parameters. When the
amplitudes of the two parameters is equal and phase difference
is 90 degree or 270 degree, this PH is a circle. In the rest
situation, there are ellipses of different excentricitys. The
eccentricity direction of ellipse means the action of axes
vibration.
/=fa,
Noise and
vibration of Stator or rotor is eccentrical
10
--Power Grid
unilateral or syntony
frequency
magnet ic pull
Noise and
The intensity of Motor shell
/=2 10 ,
vibration of pole
radial magnetic
or Stator is not strong
pull or pulse
enough
lThree-phase is asymmetric
of rotor, such as out of slot,
/=2sfo , Noise and
Empty slot, bars broken,
shrinkage cavity,
S-Slip frequency
vibration of slip
eccentricity
2.Rotor is not strong enough
3.Fixing problem of bearing
f =%0 'n-- rotate
Noise and
vibration of Rotor imbalance
speed imbalance
f=mxz%Ohave
I .Problems number of
obvious peak value
on
vane, ventilation road and
n-- rotate speed;
thermal tendons
m- number of vane, Resonance(fluting)
2.Problems on cooperat ion
ventilation road or
of and ventilation
thermal tendons;
vane
z-time of harmonic
settings
wave
Fig 3 Acceleration wave of axial detected point
2375
Fig 7 Power spectrum of radial level detected point
50 tOO 150 200 250 300 350 450 500
FrequencylHZ)
1

o
5r
x
...:..
10
:... --r- --r- """'T- --.,.- ---r- ---,- ----,r-- .--- .------,
.1
&3
:c
2
a.
03
5
Time!sl
Fig 4 Acceleration wave of radial level detected point
0.3 ';== 'F"'= "i"""<=-"F'= "-F= "'F= =r= 9 -=-=r= = r=91
N O.2 .. .. .. J ; .) .)..: ..: ;..:.; ;.:..:.: :.: [ i
01
e-
.g 0
1!
-0.1 f
t: ! .' .' I , i , , I. . ' I . I
"'02 ! "1"" ! ! ! ! "!"" "!" : .
J 5
.. -' Ti "" m"' el"" s ,_, ....
Fig 5 Acceleration wave of radial plumb detected point
2
Ii
o 50 100 150 200 250 300 350 400 450 500
FrequencylHZj
Fig 8 Power spectrum of radial plumb detected point
The three acceleration waves above show that radial
vibration is stronger than axial vibration, and radial level
vibration is stronger than radial plumb vibration. In order to
find the reason of the high noise, we analyze the vibration
signals of the motor. Firstly, do power spectrum analysis on
the vibration signals of those three points, from the main peak
frequency segment of spectrum to [find reasons of vibration
preliminary. Fig 6, 7 and 8 show the power spectrums of
three points.
The rotate speed of this motor is 2880 revolutions per
minute, so its frequency is 48 hertz. From the power
spectrums of three detected points above, we can find that the
main wave energy of the three detected points fasten on radix
frequency, and accompany with twice and triple of radix
frequency. From Table I we can know that obvious peak
value of noise signal in radix frequency segment means
rotor imbalance. But power spectrum is calculated by
ignoring phase information. When there is a situation that
vibrates strong in radix frequency segment but the phase
difference between radial level and radial plumb is 0 degree
or nearly 180 degree, this means there are other reasons of
the vibration, such as eccentricity. In order to estimate this,
we do holospectrum analysis on radial level detected point
and radial plumb detected point. Fig 9 shows the
holospectrum analysis result.
0.1 r---r- ---, - -r- -,- ...,..- .,-- ,-- ,-- ,----,
z- 0.08
ij 0,06
0,0.1
c.
100 150 50 250 300 350 400
FrequencylHZ)
----
Fig 6 Power spectrum of axial detected point
450 500
From Fig 9 we can see that the ellipse is bigger in radix
frequency segment, the other component is smaller.
So combine with time- domain wave, power spectrum and
holospectrum analysis of motor vibration, we can ascertain
that the reason of abnormal vibration is the imbalance of
motor rotor. After investigation, the judgment above is right.
V. CONCLUSIONS
Frequency domain analysis is a common and effective
method of noise source detection. Do power spectrum analysis
to the vibration signal of motor, ean detect noise source
effectively. But if combine power spectrum analysis with
holospectrum analysis, we can find the reason of abnormal
vibration of motor more exactly.
The experiment validates that frequency domain analysis
is simple and effective, it provides technology basis to motor
noise controlment, also provide a simple and effective way of
motor diagnosis by noise source detection.
ACKNOWLEDGMENT
Qingnian Wang, Kun Yan, Hongyan Li and Meiling Yuan
thank the sponsors of ICMA 2009. Thank Institute of
Electrical and Electronics Engineers, IEEE Robotics and
Automation Society, Changchun University of Science and
Technology and Kagawa University.
REFERENCES
[I] Jian Huang, "Discussion of Reducing Motor Noises", Technology of
Machines and Electric Apparatus , vol. 2, pp. 43-46, 1996.
[2] Baoku Li, "Reasons of Miniature Magnetic DC Motor Noises", Small and
Special Elctrcal Machines, vol. 2, pp, 36-37, 1996.
[3] Shanghai Electrical Science Graduate School , "Measurement Method and
Limited Value of Rotating Electrical Machine", china Rotating Electrical
Machine Standardized Technique Committee Press, pp. 1-4, 1988.
[4] Fang Zheng, Mingxing Xu, Signal Processing Principle, Publish
Company of Qinghua University, pp. 16-17, 2000.
[5] Yingjing Wei, "Application of LabVIEW on Motor Fault Diagnoses
Expert System", Explosion-proof Electric Machine, pp. 19-21, March
2005.
2376
[6] Biaozheng Shen, Electric Machine Fault Diagnoses Technique, Publish
Company of Machine industry, PP. 52-56, 1996.
[7] Yongxiao Chen, Electric machine Noise Analysis and Control. Publish
Company ofZhejiang University, pp. 116-128,1987.
[8] Yayong Chen, Application of MATLAB in Signals Processing and
explainin detail, Publish Company of People Posts and
Telecommunications, pp. 180-187,2001.
2377
[9] Zhangsui Xu, Liqing Fang, Xiwu Wang, Xianzhang Zuo, Fault
Information Diagnoses Theory and Application, Publish Company of
National Defense Industry, pp. 11-57, 2000.
[10]Yujun Li, Jianguo Yang, "Diesel Engine Radicalization Noises
Measurement and Analysis Base on Surface Vibration Method", Noises
and Vibration Control, vol. 2, pp. 71-74,2007.

También podría gustarte