Professional Documents
Culture Documents
Vibration and Noise Analysis of A Gear Transmissio
Vibration and Noise Analysis of A Gear Transmissio
Vibration and Noise Analysis of A Gear Transmissio
net/publication/4697309
CITATIONS READS
6 1,289
4 authors, including:
James Zakrajsek
NASA
37 PUBLICATIONS 882 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by James Zakrajsek on 04 March 2014.
_0 F..
and
[#c]
T [Kb] [#¢]
Nomenclature
Kbx,Kby bearing stiffness
in x and y directions
casingx-displacement
[cJ casing structuredamping matrix xe
[#jT[%][#¢l Yc casingy-displacement
[#JIce]
[#el g rotor z-displacement
1
angle of orientation In response to the importance of the applicationand
the limited amount of work performed in the area of
rotor orthonormal mode shape
gearbox noise and vibrationresearch,a program was
_c casing orthonormal mode shape startedwith the main objectiveof developing the meth-
odology required to predict gear transmission system
_x rotor translational mode shape
noise and vibration.This paper describesthe global dy-
_xe rotor rotational mode shape namic model that was developed to simulate the noise
and vibrationof a multistage gear system. Results pre-
Ox,#y,O, rotor rotational motion
dicted by the numerical-model axe compared to experi-
Ocx,Ocy, casing rotational motion mental resultsobtained from the gear noise test rig at
Ocz NASA Lewis Research Center.
(d critical
speed of rotor
critical
speed of casing AnalyticalProcedure for Gear System Vibration
= [#.]{Be} (2)
{we} 4. Calculate nonlineargear mesh forcesdue to rela-
[@cy0]
{Be} tivemotion between th_ two gears.
Experimental Studies
{A}
Test resultsfrom the gear noisetestrig(Fig. l(a)) at
{B}
(s) NASA Lewis were used in thisstudy to rei'meand vali-
date the globaldynamic model developed.Three phases
{z} = {D}[ {Z_}
of experimentalstudieswere performed: (1) the determi-
[[Dc} nation of staticmodal characteristics
using the impulse-
!Dt}l
response technique; (2)the transientvibration of both
rotor and the gearbox structureduring operation; and
The modal equationsfor the rotor system (Choy, 1991, (8)measurement of the noise generation at the gearbox
1993) can be written-in ffiodalcoordinatesas: surfaces.An outlineof the experimental studiesis pre-
sented below:
if]
{_,}
= [_.]
{_}- [_A]
{Z}- [_b]
{_'}
Evaluation of Experimental Modal Characteristics
+ [_:b--KAl
{Z}--[#]T
[Cbl[#j{_,
_}
The gear transmission system used in this test is
- [J]{z}+ [_I
T [KB]
[_,¢1
{Zc} shown in Fig. l(b).The gear system consistsof a p_r of
identicalparallelaxis spur gears supported by rolling
+ [#IT
{F(t)+FG(t)+L(t)
} element bearings.The top and four sidesof the gearbox
were dividedintoa gridof 116 node points.The transfer
(4) functionsof the system were found from the impulse-
response technique by using an impact hammer. Three
and forthe casing
accelerometers,one on the surface in the x-direction,
another in the y-direction, and the last in the
_direction,were used to obtain the responsesfrom the
[I]
{_._} {z_}- [_1{z_}- _0 (z_}- [Ub]
= [U_] {_,_}
impacts. Natural frequenciesand their corresponding
mode shapes excitedduring the experiment were evalu-
+ [#c] T [Kb] [@] {Z} + [#tIT [Cb] 1@]{Z} -t- [_¢]T {F¢(t)} ated using a two-channel dynamic signalanalyzer with
PC-based modal software.The modal characteristics
(s) found during the experiment were compared with those
generated by the numerical model for verification
Using the appropriate initialconditionsfor the modal purposes.
velocity{7,} and displacement {Z}, the modal accelera-
tions {Z} can be solved.The solutionprocedure can be
summarised as follows: Experimental Investigationsof System Dynamics
1. Input initialconditions to calculate modal The testrig is powered by a 150-kW (200-hp) vari-
accelerations. ablespeed electric
motor. An eddy-currentdynamometer
is used to absorb power at the opposite end. During this mental measurements confn'ms the accuracy of the dy-
test, accelerometers were placed on the top surface of the namic representation of the test gearbox using only a
gearbox to monitor the vibrations of the gearbox struc- limited number of modes. In addition, this correlation of
ture. In addition, two sets of noncontacting eddy-current the experimental modal characteristics with the numeri-
proximitors monitored motion of the gear shafting. The cal model helps to reduce the number of modes being
vibration signals of the gearbox surface and the shafts used in the numerical simulation to achieve an accurate
were recorded over a range of operating speeds from result.
3000 to 5000 rpm. A two-channel dynamic signal anal-
yzer computed the vibration frequency spectra which are A slow roll of the rotor-gear assembly revealed that
displayed in a waterfall diagram. The lateral vibration a substantial residual bow was present in the rotor as
frequency spectra of the gears were also computed. The shown by its large orbital motion given in Fig. 4. Fig-
results of this phase of the experimental work were corn, ure 4(a) represents the orbit of the driver rotor at the
pared with those from the model for verification of the gear location and Fig. 4(b) represents the orbit of the
numerical procedure and improvement of the modeling driven rotor. The circular orbit in the driver rotor at
of the gear transmission system. slow roll represents the residual bow deformation of the
rotor.The elliptical orbitin the driven rotor is due to a
combination of the residualbow effectsand the vertical
Investigations of Gearbox Noise gear forcefrom the torque of the drivinggear. In order
to analyticallysimulate the influenceof this effect,a
Two acousticmicrophones were used in th_sexperi- residual bow of 0.05 mm (2 mil) was incorporated into
ment. Both microphones were placed over the top cover the numerical model. Figure 5 represents the frequency
of the gearbox with one in the verticaldirectionand the spectra of the driven gear vibration generated from both
other at a 45° angle from the vertical. While rotor and the experimental and the numerical procedures. The ex-
gearbox vibration are recorded as describedin the last ceUent agreement in both results further verifies the
section,noise was obtained by the microphone system validity of the numerical study.
at various running speeds. A correlationbetween the
gearbox vibrationsand the noiselevelwas performed in In order to predict the noise of the transmission
the frequency domain using lineartransferand cross- system during operation, the vibrationof the gearbox
correlation functions. The nonlinear hypercoherence must be accuratelymodeled. Figures 6 and 7 presentthe
function (Jong, 1986) was alsoused to generate a rela- time domain vibration signal and corresponding fre-
tionship between the noise and the vibration data in quency spectraof the signalfor the gearbox top surface
order to develop a noise predictionalgorithm for the from both experimental and numerical studies. Fig-
global dynamic model. ure 6(b) shows the analyticallypredictedvibration sig-
nal (acceleration)
at the top of the casingstructureat an
operatingspeed of 3000 rpm while the corresponding ex-
Correlation and Benchmarking of .System Dynamics By perimental resultsare given in Fig. 6(a).Note that the
Numerical Procedures with Experimental Results amplitudes and general shapes of the two time signals
are very similar.The comparison of their frequency
The modal hammer excitedmany vibrationmodes of contents are given by the frequency spectra in Fig. 7.
the gearbox structure.Eight of these modes in the fre- Figure 7(a) shows the frequency components of the
quency range 0 to 3000 Hz seem to dominate the trans- measured vibrationat the top of the gearbox. Note that
ferfunctionspectra.These eightmodes, shown in Fig. 2, the major vibrationcomponents occur at the mesh fre-
were chosen to representthe vibrationof the system. In quency of 1250 Hz and at 3-times (3750 Hz) and 4-times
orderto produce an accurateanalyticalsimulationofthe (5000 I-Is)
the mesh frequency.The very largeamplitude
testgearbox, a similarsetof modes was generatedusing in the frequency components at 3750 and 5000 I-Isare
a NASTRAN finiteelement model of the gearbox struc- due to the excitationof the gearbox naturalfrequencies
ture. Out of 25 modes in the analyticalmodel in the near the 3-times and 4-times multiplesof the mesh fre-
0 to 3000 Hs frequency range, the eight dominant modes quency. The fundamental and the 2-times mesh fre-
(similar to those from the experimental studies) were quency component are substantially smaller due to the
used to represent the gearbox dynamic characteristics. lack of any gearbox natural frequencies near 1250 and
These analytically simulated modes are shown in Fig. 3. 2500 Hz. In addition, frequency components of substan-
Table I lists the natural frequencies of the simulated tial magnitude are also noticed at 700 and 2900 Hz,
modes. The calculated natural frequencies are within which are also close to the natural frequencies of the
5 percent of the measured modes. The three-dimensionai gearbox. Figure 7(b) depicts the results from the numer-
analytical mode shapes (Fig. 3) are very similar to the ical simulations of the system. The numerical results
experimental mode shapes (Fig. 2). The correlation of correspond closely to those from the experimental study
the results between the analytical model and the experi- with the exception of the 4-times mesh frequency
4
component. The magnitude and the generaldistribution The hypercoherence function (Jong, 1986) establishesa
of the frequency content of the spectrum isvery similar nonlinearrelationshipbetween the basicfrequency com-
to the experimental results. ponent and itsmultipleharmonics. Figure 10 shows the
hypercoherence functions of the noise signals for the
3000-rpm case.The magnitudes of the harmonic compo-
Correlationof Noise and Vibration nents form the nonlinearrelationshipof each component
to the fundamental gear mesh frequency.These relation-
In order to predict the noise generated in a gear ships provide a pictureofthe relativecontributionofthe
transmission system, a two-phase procedure was devel- higher harmonics with i_pect to the fundamental fre-
oped to correlatethe noise and vibrationdata obtained quency. These hypercoherence functionswere used on
during experimental studies: (1)linear correlation the numerical simulation data to examine the accuracy
through transferfunctions,and (2)nonlinearcorrelation of the noise prediction.
through hypercoherence functions. The relationships
derived in thisstudy were used to predictgearbox noise In order to predictthe noise levelof the transmission
from the numerical simulationof the gearbox vibrations. system during operation,the vibrationsof the gearbox
A descriptionof the procedures used ispresentedin the structure were simulated using the modal synthesis
following paragraphs: method. The transferfunctionformicrophone number 1
(verticalto the plate),as given in Fig. 9(a), was used to
predict the noise spectrum of the system (Fig.11(b)).
Correlationof Experimental Noise and Vibration Data Figure 11 shows the comparison of the resultsof the
experimental noise (Fig.11(a))with the predictednoise
In this experiment, a new set of two identical25- (Fig. 11(b)) for the verticalmicrophone. Although all
tooth gears were installedin the gearbox shown in the m_jor frequency components excited in both cases
Fig. l(b). The vibration of the gearbox surfaceswas are very similar,one can see that the predicted noise
monitored by accelerometers,and the noise data was spectrum (Fig.11(b)) has a somewhat higher amplitude
obtained by microphones, as describedin the previous at mesh frequency (lx) and.a very small 4-times mesh
section.The frequency and power spectra of both the frequency (4.x)component as in the vibrationspectrum.
vibration and noise signalsat 3000 rpm running speed
axe shown hx Fig.8. Note that in the vibration spec- Another way of examining the validityof the pre-
trum, a small frequency component existsat the mesh dicted noisesignalisthrough the nonlinearrelationships
frequency of 1250 Hz with two very largecomponents at derivedby the hypercoherencefunction.Figure 12 shows
3-times the mesh frequency(3750 Hz), and 4-timesmesh the comparison of the hypercoherence functions from
frequency (5000 I-Is). In the noise spectrum, while fre- both experimental and numerical solutionsfor micro-
quency components of considerable magnitudes are phone number 1. Note that both the experimental and
observed at the mesh frequencyof 1250 Hz, and at 2 and numerical resultsshow a very small 2-times component,
S-times the mesh frequency (2500 and 3750 Hz), _ large with a substantial amplitude at $, 4, and 5 times funda-
amplitude frequency component is also noticed at mental frequency. However, a 50-percent coherence is
400 I-Is. This 400-I_ frequency component does not coin- found in the 3, 4, and 5 times components from numeri-
cide with any of the major vibration frequencies at the cal predictions (Fig. 12(b)) while only about a 25-
top surface of the gearbox. A closer examination of the percent coherence is seen from the experimental results
system dynamic characteristics reveals that the 400-Hz (Fig. 12(a)). This difference is possibly due to the
frequency is an excitation of one of the natural frequen- limitations of the numerical model in simulating various
cies of the rotor system, which has a very small ampli- details in the gearbox transmission system. An explana-
tude in the gearbox vibration spectrum. A logical tion of the much higher (nearly double} coherence ampli-
explanation is that noise is being produced at the rotor- tudes in the numerical study is that only the major
gear system at 400 I-Is and is not being transferred to frequency-related vibration factors axe included in the
the vibration of the gearbox surface. This phenomenon model, which results in a predicted vibration signal con-
is further verified by the very large amplitude of the sisting mostly of frequency components that are closely
vibration-to-noise transfer function at 400 I-Iz as given related to the fundamental frequency.
in Fig. 9(a). A linear coherence function between the
noise and vibration signals is also given in Fig. 9(b).
Conclusions and Summary
From the experimental data, it can be observed that
in the frequency domain, most of the noise components, A methodology was developed using the modal syn-
other than the 400-Hz contributions, exist at multiples thesistechnique to simulate the global dynamics of a
of the mesh frequency. The nonlinear hypercoherence gear transmission system. Good correlationhas been
function was used for the second phase of this study. obtained using resultsfrom the global dynamic model
andmeasured
dynamiccharacteristics
from the NASA 6. Choy, F.K., Tu, Y.K., Savage, M., and Townsend,
gear noise testrig.A numerical relationshipwas also D.P., "Vibration Signature Analysis of Multistage
developed from the experimental data to correlatenoise Gear Transmission,_ Journal of the Franklin Insti-
and vibrationsin the frequency domain. This relation- tute,Vol. 328, No. 2/3, 1991, pp. 281-299.
ship was appliedto the vibrationsimulationsto predict
the noise generated by the gearbox. Results of the pre- 7. Choy, F.K., Ruan, Y.F., Zakrajsek, J.J.,
dicted noisewere compared to the experimentallymeas- Oswald, F.B., and Coy, J.J., "Analytical and
ured data linearlyusing the frequency spectra and Experimental Study of Vibrations in a Gear Trans-
nonlinearly using the hypercoherence function. Some mission,r AIAA Paper 91-2019, 1991.
conclusionsof thisstudy can be summarized as follows:
8.
Choy, F.K., Ruan, Y.F., Zakrajsek, J.J.,
I. The dynamics of a gear transmission system Oswald, F.B., "Modal Simulation of Gear Box
coupled with the vibrations of the gearbox structure Vibrations with Experimental Correlation,
_ to be
during operation can be simulated within 5 percent published in AIAA Journal of Propulsion and
of the modal frequencies using the modal synthesis Power, 1993.
technique.
9o David, J.W., Mitchell, L.D., and Daws, J.W.,
2. Predicted vibrationsat the housing surfaceusing _Using Transfer Matrices for Parametric System
the global dynamic model are in good agreement with Forced Response,_ Journal of Vibration_ Acoustics_
measured values. Stressand Reliabilityin Design, Vol. 109, No. 4,
Oct. 1987, pp. 356-360.
3. Both the frequenciesand magnitudes of the pre-
dicted noise components are similarto those obtained 10. EI-Bayoumy, L.E.,Akin, L.S.,Townsend, D.P., and
from experimental study. Choy, F.C., _The Role of Thermal and Lubricant
Boundary lJayersinthe TransientThermal Analysis
4. The hypercoherence function was found to be a of Spur Gears,_ NASA TM-I01435, 1989.
valuabletoolin comparing the nonlinearrelationshipsin
the predictednoise to the corresponding nonlinearrela- 11. Jong, J.,and Coffin,T., _DiagnosticAssessment of
tionshipsin the measured noise. Turbomachinery by the Hyer-Coherence Method,"
Advanced Earth-to-Orbit Propulsion Technology
1986, Vol. 1, NASA CP-2436, VOL-1, pp. 45-64,
References 1986.
1. Akin, L.S., and Townsend, D.P., _Lubricant Jet 12. Krantz, T.L., "Gear Tooth StressMeasurements of
Flow Phenomena in Spur and Helical Gears with Two Helicopter Planetary Stages,_ NASA
Modified Addendums--for Radially Directed Indi- TM-105651, 1992.
vidual Jets,_ NASA TM-101460, 1989.
13. Lewicki, D.G., Decker, H.J., and Shimski, J.T.,
2. August, R., and Kasuba, R., _Torsional Vibrations _Full-Scale Transmission Testing to Evaluate
and Dynamic Loads in a Basic Planetary Gear Advanced Lubricants,_ NASA TM-I05668, 1992.
System," Journal ofVibration7Acoustic_Stress_and
Reliabilityin Design, Vol. 108, No. 3, July 1986, 14. Lira, T.C., Signh, R., and Zakrajsek, J.J.,_Modal
pp. 348-353. AnalysisofGear Housing and Mounts," Proceedinss
of the 7th InternationalModal AnalysisConference,
3. Boyd, L.S.,and Pike,J.A., "EpicyclicGear Dynam- Vol. 2, Society of Experimental Mechanics, Bethel,
ics," AIAA Journal, Vol. 27, No. 5, May 1989, CT, 1989, pp. 1072-1078.
pp. 6O3-6O9.
15. Mark, W.D., "The Transfer Function Method for
4. Choy, F.K., Townsend, D.P., and Oswald, F.B., Gear System Dynamics Applied to Conventional
"Dynamic Analysis of Multimesh-Gear Helicopter and Minimum Excitation Gear Design," NASA
Transmissions,_ NASA TP-2789, 1988. CR-3626, 1982.
5. Choy, F.K., Townsend, D.P., and Oswald, F.B., 16. Oswald, F.B., "Gear Tooth StressMeasurements on
"Experimental and Analytical Evaluation of Dy- the UH-60A Helicopter Transmission,_ NASA
namic Load and Vibration of a 2240-kW Rotorcraft TP-2698, 1987.
Transmission," Journal of The Franklin Institut%
Vol. 326, No. 5, 1989, pp. 721-735. -
17. Oswald,F.B., Zakrajsek,J.J., Townsend,D.P., 19. Rebbechi, B., Oswald, F.O., and Townsend, D.P.,
Atherton,W., and Lin, H.H., '_Effectof Operating _Dynamic Measurements of Geax Tooth Friction
Conditions on Geaxbox Noise," NASA TM-105331, and Load," NASA TM-103281, 1991.
1992.
20. Seybert, Wu, T.W., and Wu, X.F., "Acoustical
18. Ozguven, H.N., and Houser, D.R., _Mathematical Analysis of Gear Housing Vibration,_ NASA
Models Used in Gear Dynamics-A Review," Journal TM-103691, 1991.
of Sound and Vibration,Vol. 121, No. 3, Max. 22,
1988, pp. 383-411.
658 658 0
1049 1006 -4.1
1710 1762 3.0
2OO0 2051 2.6
2276 2336 2.6
2536 1536 0
2722 2752 1.1
2962 3012 1.7
Dynamometer-=,
_... rSoft coupling ,_Test gearbox_
8
9
• /
7" ..L j.
z z
II
Mode: 1 Mode: 2
z z
Mode: 3 Mode: 4
z z
Mode: 5 Mode: 6
!
1 z
Mode: 7 Mode: 8
Figure 3.--Analytica_ determined mode shapes of the gear box (0 to 3000 Hz).
10
Shaft #1 Shaft #2
.OO6 D
o
I I I J
-.006 0 .006 -.006 0 .006
Displacement, X-direction, In.
11
Rotational
speed,
rpm
6O0O
5500
50OO
4500
4O00
35OO
3000
25OO
2000
I I 15oo
(a) Experimental
6OO0
55O0
500O
450O
4OOO
35O0
30OO
,o]
0.5
0
25OO
2OO0
I , I I I 1500
0 500 1000 1500 2000 0 5OO 1000 1500 2OOO
Frequency, Hz
Co)Ana_ical.
12
4X
20
2.5
2.0
f 3x
r" Mesh
10 1.5 - I frequency
I
-10
,I | I
-2o I l i i I ._ o _ =_ '
20 2.0 - Mesh
-10
o
0 0,1 0.2
1.0
0.5
Figure 6.ITime vibration signal at top of gear box for operating Iqgum 7._Vibration frequency spectra at top of gear box for
speed of 3000 rprn. operating speed of 3000 rpm.
13
.7O
3°E
2.5
2.0
.60
.50
+_.+o
CL
1.0
0.5 ,20
0 _ ++ .10
(a) Frequency spectrum for vibratiorL I
0
(a) Transfer function.
.15 i
r 400 I-tz
f
1.00
@
,+..,
.10
.75
o¢
"5.
E
<
8
.25
0 .. I
(b) Frequency spectrum for noise from micro-
phone number 1. I
0 2000 4000 6000 8000 10000
2000xl 03 Frequency, Hz
1.00
1 I I I I I
0 I J,, t,, l i_ I
¢D
0 .50 --
2
_3000
40005000
i.._400 Hz
i -r
I
I
.25 -- !
0
L .
2000
.. I
"4000
Frequency, Hz
[ I
6000 8000
I
ill I
(d) Power spectnJm for noise from microphone o 1 2 3 4 5 6 7
number 1. Harmonics of mesh frequency
Figure 8,--Frequency and power spectrum at Figure 10.--Hypercoherence function for the noise signal from
the operating speed of 3000 rpr_ microphone number 2 at 3000 rpm.
14
I I I I I I I
I
.50
I
.25
.14
•12 frequency
.10 3x I
i 0, (a) Experimental data.
.08 o
2
.06 2x 1.00 I I I I I I
2:
4X
.02
I I .75 -
.50 -
r- Mesh
.30 ii frequency
.10
| , ,,
2 3 4 5 6 7 8
0 2O00 4000 600O 80O0 10O0O Frequency harmonic
Frequency, Hz (b) Numerical predictions.
(b) Predicted noise.
IRgure 11 .--Frequency spectra results from microphone number 1 F'K3um12.--Hypercoherence function of the noise signal from
at the operating speed of 3000 rpm. microphone number I at 3000 rpm.
15
Form Approved
REPORT DOCUMENTATION PAGE OM_ No. 0704-0188
Public reporting burden fo'r this collection Of information Is estimated to average 1 hour per response, including the time for reviewing instruCtiOns, searching existing data sources,
gathering and maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this
collection Of information, including suggestions for reducing this burden, to Washington Headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson
Davis Highway. Suite 1204, Arlington. VA 22202-4302. and to the OWce of Management and Budget. Paperwork Reduction Project (0704-0188). Washington. DC 20503.
1. AGENCY USE ONLY (Leave blank) REPORT DATE 3. REPORTTYPE AND DATES COVERED
June 1993 Technical Memorandum
r
4. TITLE AND SUBTITLE FUNDING NUMBERS
WU-505--62-10
6. AUTHOR(S)
Prepared for the 29th Joint Propulsion Conference and Exhibit cosponsored by the AIAA, SAE, ASME, and ASEE, Monterey, California,
June 28-L30, 1993. EK. Choy and W. Oian, The University of Akron, Department of Mechanical Engineering, Akron, Ohio 44325-3903;
and J.L Zakrajsek and F.B. Oswald, NASA Lewis Research Center. Responsible person, J.J. Zakrajsek, (216) 433--3968.
l:;a. DISTRIBUTIONIAVAJLABILn'Y STATEMENT 12b. DISTRIBUTION'CODE
Unclassified - Unlimited
Subject Category 37
This paper presents a comprehensive procedure to predict both the vibration and noise generated by a gear transmission
system under normal operating conditions. The gearbox vibrations were obtained from both numerical simulation and
experimental studies using a gear noise test rig. In addition, the noise generated by the gearbox vibrations was recorded
during the experimental testing. A numerical method was used to develop linear relationships between the gearbox
vibration and the generated noise. The hypercoherence function is introduced to correlate the nonlinear relationship
between the fundamental noise frequency and its harmonics. A numerical procedure was developed using both the linear
and nonlinear relationships generated from the experimental data to predict noise resulting from the gearbox vibrations.
The application of this methodology is demonstrated by comparing the numerical and experimental results from the gear
noise test rig.