Tribology Internahonal Vol. 29, No. 1, pp.

71-83, 1996
Copyright @ 1996 Elsevier Science Ltd
Printed in Great Britain. All rights reserved
ELSEVIER 0301-679X/%/$15.00 +O.Otl
SClENCEi 0301-679X(95)00037-2

Analysis of the effects of

surface pitting and wear on the
vibration of a gear
transmission system
F. K. Choy*, V. Polyshchuk*, J. J. Zakrajsek+, R. F. Handschuh* and
D. P. Townsendi

A comprehensive procedure to simulate and analyse the vibrations

in a gear transmission system with surface pitting, wear, and
partial tooth fracture of the gear teeth is presented. An analytical
model was developed where the effects of surface pitting and
wear of the gear tooth were simulated by phase and magnitude
changes in the gear mesh stiffness. Changes in the gear mesh
stiffness were incorporated into each gear-shaft model during the
global dynamic simulation of the system. The overall dynamics of
the system were evaluated by solving for the transient dynamics
of each shaft system simultaneously with the vibration of the
gearbox structure. In order to reduce the number of degrees-of-
freedom in the system, a modal synthesis procedure was used in
the global transient dynamic analysis of the overall transmission
system. An FFT procedure was used to transform the averaged
time signal into the frequency domain for signature analysis. In
addition, the Wigner-Ville distribution was also introduced to
examine the gear vibration in the joint time-frequency domain for
vibration pattern recognition. Experimental results obtained from a
gear fatigue test rig at NASA Lewis Research Center were used to
evaluate the analytical model.

Introduction cost and downtime required make such programs

inefficient and uneconomical.
Over the last two decades, problems arising from
excessive wear and gear tooth surface pitting in gear Vibration signature analysis methodologies are being
transmission systems have been of increasing concern developed to examine non-intrusively the health and
for a variety of gear users. At increased power and wear of gear transmission systems. Using spectral
higher speeds, gear wear and fatigue failures are major analysis, the amplitude of the frequency spectrum of the
concerns in aerospace applications. Although regular measured vibration signal is calculated and displayed in
visual inspections and preventive maintenance can a continuous manner. However, the spectral analysis
help to reduce the failure rate of gear systems, the technique is difficult to apply in a highly complex
system where the large number of spectral lines often
makes it difficult to detect significant changes in
*The University of Akron, Akron, USA
National Aeronautics and Space Administration, USA
the spectrum. Another methodology is the joint
Vehicle Propulsion Directorate. U.S. Army Research Laboratory, time-frequency approach which applies the Wigner-
Lewis Research Centre. USA Ville distribution (WVD)- on the time vibration
Tribology International Volume 29 Number 1 1996 77
Effects of surface pitting and wear: F. K. Choy et al.

signal of the system. Unlike the regular Fourier

transform process, the WVD provides an instantaneous
frequency spectrum of the system at any instant = {Fbi(t)) + {Fgi(t)) + {Fui(t)) (1)
throughout the sampling period (while FFT provides where [M] and [K,] are respectively the mass and
an averaged frequency spectrum of the total sampling shaft stiffness matrices of the rotor, { Wi} is the general
period). The spectral density of the fundamental displacement vector of the ith rotor in its local
exciting frequency and its sidebands change as the coordinate system, and { Fbi( t)} , { F,J t)} , and { Fui( t) }
shaft rotates through a complete revolution. Some are respectively the force vectors acting on the
success has been achieved in applying the WVD ith rotor system due to bearing forces, gear mesh
concept in the health monitoring of gear transmission interactions, and mass-imbalances.
systems 3- 6. However, a complete vibration signature
database is needed for development of an effective The equations of motion of the gearbox with p rotor
pattern recognition scheme. In order to populate such systems can be expressed as
databases, the development of an accurate analytical
procedure to predict vibrations in gear systems due to
wear and fatigue failure is necessary. i=l

The objective of this paper is to develop a comprehen- where [ Tci] represents the coordinate transformation
sive procedure to simulate and analyse the vibration between the ith rotor and the gearbox.
in a gear transmission system with effects of surface
pitting and wear of the gear teeth under normal Evaluation of bearing forces
operating conditions. The effects of changes in magni-
tude and phase of the mesh stiffness at one particular The bearing forces { F,i(t)} for the ith rotor can be
tooth or a number of consecutive teeth were evaluated evaluated as:
in order to simulate the effects of surface pitting and
wear. The effects of these localized changes in the {Fbi(t)I = [C,il({@il - [Ticl{*cil)
gear mesh were incorporated into each gear-rotor +[Kbil({WiI - [TicI{iVcil) (3)
model for the dynamic simulation7-9. The dynamics
of each gear-rotor system were coupled with each where [ Cbi] and [Kbi] are respectively the damping
other through the gear mesh interacting forces. The and stiffness of the bearing, [ Tj,] is the coordinate
coupling between the rotors and the casing structure transformation matrix for the gearbox with respect to
was generated through the bearing support forces. the ith rotor, and Wci are the casing displacements at
The global vibrations of the system were evaluated by the rotor locations.
solving the transient dynamics of each rotor system
simultaneously with the vibration of the casing. In Evaluation of gear forces
order to minimize the computational effort, the number
of degrees-of-freedom of the system were reduced by The gear forces generated from the gear mesh interac-
using a modal synthesis procedure7,8. Experimental tion12 can be written as:
vibration results obtained from a gear fatigue test rig
at NASA Lewis Research CenteP were used to verify {Fg,(t)) = {Fri(t)) + {Fti(t)) (4)
the analytical procedure. where { Fri( t)} is the vector containing the gear forces
and moments resulting from the relative rotation
Solution procedures between the two mating gears and { Fti(t)} is the
vector containing gear forces and moments due to the
Dynamics of the gear-shaft configuration and translational motion between the two gears.
the gearbox system
The dynamics of the ith individual gear-shaft system Modal synthesis procedure
can be evaluated through the equations of motion for
the vibrations of an individual rotor-bearing-gear In order to calculate the transient and steady state
system as shown in Fig 17s8, given in matrix form, as: dynamics of the system, the coupled rotor and casing
equations of motion must be solved simultaneously.
To minimize the computational effort, the modal
transformation7,8 procedure will be applied to reduce
the degrees of freedom of the global equations of
motion. Using m undamped mode shapes of the ith
rotor system [&i, $i2, $i3, . . . , &,,J and m, undamped
mode shapes of the gearbox [&, &i, +c3r . . . ,
&,J, the rotor displacement for the ith rotor can be
written as:
Bearing forces Bearing forces
{WiI =,$lAij{+ijl (5)
ft
Gear forces and, similarly, the gearbox displacements as:
Fig 1 Schematic of the rotor-gear bearing system {WC>= [&I{&) (6)
78 Tribology International Volume 29 Number 1 1996
 Effects of surface pitting and wear: F. K. Choy et a/.

where {Ai} and {A=} are the model time functions of S(t) =x(t) + jH[x(t)] (12)
the ith rotor and the gearbox respectively. Using the where H[x(t)] is the Hilbert transform of x(t).
expansion in Equation (5), the equations of motion
for the ith rotor in Equation (1) can be written as: In order to avoid a repetition in the time domain
WVD, a weighting function4 is added to the time data
[Ml[+cl{Ai> + [KsI[+iI{AiI before the evaluation process. Such a process may
= {Fbi(t)) + lFgiCt)) + IFuiCt>l (7) decrease the resolution of the distribution, but it will
eliminate the repetition of peaks in the time domain,
Premultiplying by [ +ilT and using the orthogonality and thus the interpretation of the result will be
conditions of the mode shapes, the ith rotor equations substantially easier.
of motion can be written as:
{Ai} + [A]{Ai} = {Fbi} + {Fgi} + {Fui} (8) Description of experimental study
where [A*] is the diagonal matrix of the squares of
the natural frequencies of the system. The experiment was performed on the spiral bevel
gear fatigue test rigl, as illustrated in Fig 2, at the
For the gearbox system, a similar transformation is NASA Lewis Research Center. The primary purpose
carried out as Equation (2) can be written as: of this rig is to study the effects of gear tooth design,
gear materials, and lubricants on the fatigue strength
{&I + [CI{A,) = @cd (9) of aircraft quality gears. Because spiral bevel gears
For a system of k rotors, Equation (8) can be repeated are used extensively in helicopter transmissions to
k times and solved with the casing equation (9) transfer power between non-parallel intersecting shafts,
simultaneously for the modal accelerations {Ai} and the use of this fatigue rig for diagnostic studies
{A,}. A numerical integration scheme is used to is practical. Vibration data from an accelerometer
integrate the accelerations to obtain velocities and mounted in the pinion shaft bearing housing was
displacements at each time step for transient calcu- captured by an analogue to digital conversion board.
lations. The 1Ztooth test pinion, and the 36tooth gear have
a 35 spiral angle, a 1 in. (25.4 mm) face with a 90
shaft angle, and 22.5 pressure angle. The pinion
Signature analysis of vibration signal transmits 720 hp at a nominal speed of 14400 rpm.
The test rig was stopped several times during the test
Frequency domain analysis
for gear damage inspection. The test was concluded
The frequency spectrum analysis is used by applying at 17.8 operational hours when a broken tooth was
a discrete Fourier transform on the average time signal detected visually during one of the shutdowns.
x(r) such that the spectral components are: Pictures of tooth damage on the pinion at various
stages in the test are shown in Fig 3. At the first rig
X(k) exp( Fj
= Ttx(r) (10) shut-down, at about 5.5 h into the test, a small pit
i=o was observed on one of the teeth on the test pinion,
where x(t) is the time average of the vibration signal as illustrated in Fig 3A. The test was stopped again
B(t) and T is the sampling interval. The frequency at approximately 12 h and the pitted area spread to
components are examined in the frequency domain cover approximately 75% of the face of the pinion
and compared with those obtained at various stages tooth, as seen in Fig 3B. In addition, pitting started
of the fault development in the experimental gear test to appear on the adjacent teeth. Figure 3B shows the
rig. pinion at the end of the test, at 17.8 h. It was found
that one of the three heavily pitted pinion teeth had
Joint time-frequency analysis: the Wigner-Ville experienced a tooth breakage, losing one third of the
distribution tooth, as shown in the figure.

To examine the vibration signal in a joint time-

frequency domain, the Wigner-Ville method1-3 is used
in this study. The Wigner-Ville distribution will provide
an inter-domain relationship between time and fre-
quency during the period of the time data window.
The WVD (Wigner-Ville distribution) can be written
as:

WV(t,f) = /;(t + ;jx* (i - ;jedi (11)

where WV(t,f) is the Wigner-Ville distribution in both

the time domain t and the frequency domain f. To
allow sampling at the Nyquist rate and eliminate the
concentration of energy around the frequency origin
due to the cross product between negative and positive
frequency 1.2, the analytic signal was used in evaluating Fig 2 Picture of the bevel gear test rig in NASA Lewis
the WVD. The analytic signal s(t) is defined as: Research Center
Tribology International Volume 29 Number 1 1996 79
Effects of surface pitting and wear: F. K. Choy
Fig 3 Pictures of the damaged pinion teeth: (a) 5.5 h; (6) 12 h; (c) 17.8 h
Discussion of results
To study the effects of gear tooth pitting and wear on
the dynamics of the rotor system, the numerical
simulation procedure described above was used to
model the vibrations of the pinion gear in the test rig.
During the experimental study, vertical direction
vibration signals from the pinion gear are time
synchronously averaged for spectral analysis and analy-
sis using the joint time-frequency distribution (WVD).
In order to perform an accurate comparison, the
averaged time signal from the vertical vibration of the
pinion gear is also generated using the numerical
model. During these simulations, approximate gear
mesh stiffness models are developed to simulate the
effects of wear and pitting of the pinion tooth on the
dynamics of the system.
As has been established, the changes due to gear
tooth wear or failure can be represented by the
amplitude and phase changes in vibration, which, in
turn, can be represented by magnitude and phase
changes in mesh stiffness5,6. To demonstrate the effects
of mesh stiffness change on gear vibration, the variation
of the mesh stiffness model used for this study is given
in Fig 4. The undamaged configuration of the mesh
stiffness is given by 0 phase change (Fig 4A), and
0% amplitude reduction (Fig 4B). During the wear
and pitting process, two types of stiffness changes are
6 (4 :
0 20
Angle of rotation
Fig 4 Stiffness changes in the gear mesh model
80 Tribology International Volume 29 Number
et al.
examined, i.e. the phase changes, shown in Fig 4A,
and the amplitude changes, shown in Fig 4B. Figures
5 and 6 show the time, frequency, and joint
time-frequency analysis (WVD) of the pinion gear
vibration signals with the approximated changes in
gear mesh stiffness.
Figure 5 shows the effects of mesh stiffness phase
changes in the WVD representation of the predicted
vibration signal. As seen in Fig 5, a phase change in
the mesh stiffness at the sixth tooth of the 12-teeth
pinion resulted in a temporary increase of amplitude
and phase of the pinion vibration time signal during
the sixth tooth pass location. As the phase shift in the
mesh stiffness increases, from 1.5 to 4.5, the changes
in amplitude and phase in the vibration signal become
more pronounced. In the frequency spectra, this
change in mesh stiffness will result in the increase of
the amplitude in the sideband frequencies. However,
as discussed earlier, although the frequency spectrum
provides good indications of the existence of the non-
synchronous components, it cannot distinguish the
time locations of their occurrences. The joint time-
frequency analysis using WVD shows the existence of
various frequency components as the pinion rotates
through a complete revolution of 360. Note that, in
this case, the WVD shows a continuous excitation of
the mesh frequency (12 x rotational speed) throughout
25%
2-
Zi- 50%
E. I
Gj:
(bl :
I
40 0 20 40
Angle of rotation
1 1996
 Effects of surface pitting and wear: F. K. Choy et al.

360 dea 360 deg

1 12.30 Hz) l! [Hz)

0.4 OS1
v 0:

0 10 20 30 40 50 0 10 20 30 40 50
(4 Order of running speed (a) Order of running speed

360 360 deg 360 I60 deg

12.30 (kHz) 12.30 (Hz)

r I
0 1 A h
1
0 10 20 30 40 50 0 10 20 30 40 50
lb) Order of running speed @I Order of running speed

360 360 dea 360 360 deg

I
12.30 Hz) 12.30 w
V va
VVL A h
0
I I
0 10 20 30 40 50 0 10 20 30 40 so
(cl Order of runningspeed (c) Order of runningspeed

Fig 5 Simulated pinion vibration signature due to phase Fig 6 Simulated pinion vibration signature due to
change in the gear mesh stiffness: (a) 1.5; (b) 3.0; amplitude change in the gear mesh stiffness: (a) 0%
(c) 4.5 reduction; (b) 25% reduction; (c) 50% reduction

the complete 360 revolution while subsynchronous Figure 6 shows the effects of reductions in mesh
components of eight times, four times, and one times stiffness at the sixth gear location. With the reduction
rotational speed are occurring at the sixth and seventh of mesh stiffness. a substantial change in the vibration
tooth pass locations. at the sixth tooth pass location (150-180) is observed.
Tribology international Volume 29 Number 1 1996 81
 Effects of surface pitting and wear: F. K. Choy et al.

360 360 deg 360 360 deg

0 10 20 30 40 50 0 10 20 30 40 50
Order of running speed lb) Order of running speed

Fig 9 Numerically simulated pinion vibration signature due to damage on pinion teeth due to wear and pitting:
(a) single tooth; (b) three teeth

large cross pattern at the sixth tooth pass location due References
to tooth break-off. However, some discrepancies have
been detected between the experimental and the 1. Boashash B. and Black P.J. An efficient real-time implementation
of the Wigner-Ville distribution. IEEE Trans. Acoust. Speech
numerical time signal at the fourth and fifth tooth Sig. Proc., November 1987, ASSP-35. 11
pass locations. The experimental time signal consists
of some higher frequency, smaller amplitude vibration 2. Claasen T.A.C.M. and Mecklenbrauker W.F.G. The Wigner
distribution-A tool for time-frequency signal analysis, Part I.
modulation, which is not being numerically simulated. Philip J. Res. 1980, 35
This additional modulated signal resulted in the
excitation of the 14 times rotational speed component, 3. Shin Y.S. and Jeon J.J. Pseudo Wigner-Ville time-frequency
as shown in the frequency spectrum in Fig 7B, and, distribution and its application to machinery condition monitor-
ing. J. Shock Vibr. 199.311994, 1, I. 65
also, in turn, is responsible for the small differences
created in the WVD. 4. Mcfadden P.D. Examination of a technique for the early
detection of failure in gears by signal processing of the time
domain average of the meshing vibration. Mech. Syst. Sig.
Summary and conclusions Proc. 1987, 1, 2, 173-183
A numerical procedure has been developed to simulate 5. Mcfadden P.D. and Wang W.J. Time-frequency domain analysis
the dynamics of gear transmission systems with the of vibration signal for machinery diagnostics (II) the weighted
effects of gear tooth damage due to wear and pitting. Wigner-Ville distribution. Report No. OUEL 1891, University
The work presented in this paper can be summarized of Oxford, 1991
as follows: 6. Forrester, B.D. Analysis of gear vibration in the time-frequency
domain. Proc. 44th Meeting of the Mechanical Failure Prevention
(1) A modal synthesis methodology has been
Group, February 1990
developed to simulate the dynamics of gear
transmission systems. While the computational I. Choy F.K., Tu Y.K., Savage M. and Townsend D.P. Vibration
effort has been greatly reduced by modal trans- signature analysis of multistage gear transmission. J. Franklin
Inst. 1991, 328, 213
formation, the numerical results generated main-
tain good accuracy. 8. Choy F.K., Ruan Y.F., Zakrajsek J.J. and Oswald F.B. Modal
(2) Gear tooth damage due to wear and pitting can simulation of gearbox vibration with experimental correlation.
be simulated by amplitude and phase changes in AIAAJ. Propul. Power, March 1993. 9, 2
the gear mesh stiffness model. The gear mesh 9. Kabraman A., Ozguven H.N., Houser D.R., and Zakrajsek J.J.
model developed can easily be incorporated Dynamic analysis of geared rotors by finite element. NASA
into the global transmission system for dynamic TM-102349. AVSCOM-TM-89-C-006, 1990
predictions. 10. Stewart R.M. Some useful data analysis techniques for gearbox
(3) Using the time averaging technique, frequency diagnostics. Paper MHM/R/10/77 Institute of Sound and
spectrum analysis, and the Wigner-Ville distri- Vibration Research, 1977
bution, a signature analysis scheme can be 11. Zakraisek J.J., Handschub R.F. and Decker H.J. Application
developed to examine and characterize the of fault detection techniques to spiral bevel gear fatigue data.
vibration signal of the gear system. NASA TM-106467. nresented in the 48th Mechanical Failures
Prevention Group Meeting, Wakefield, MA. April 19-21 1994
(4) A parametric study of the effects on the vibration
signal due to various changes in the gear mesh 12. Cboy F.K., Townsend D.P. and Oswald F.B. Dynamic analysis
stiffness model, simulating various degrees of of multimesh-gear helicopter transmissions. NASA TP-2789.
pitting and wear damage, could provide a compre- 1988
hensive database for gear fault detection and 13. Boyd L.S. and Pike J.A. Epicycle gear dynamics, AIAA J. May
damage estimation research. 1989. 27. 5

Tribology International Volume 29 Number 1 1996 83