Construction of Semianalytical

Song He
Acoustics and Dynamics Laboratory,
Solutions to Spur Gear Dynamics
Department of Mechanical Engineering,
The Ohio State University,
Given Periodic Mesh Stiffness
Columbus, OH 43210
e-mail: he.81@osu.edu and Sliding Friction Functions
Todd Rook Gear dynamic models with time-varying mesh stiffness, viscous mesh damping, and slid-
Goodrich Aerospace,
ing friction forces and moments lead to complex periodic differential equations. For
101 Waco Street,
example, the multiplicative effect generates higher mesh harmonics. In prior studies,
Troy, OH 45373
time-domain integration and fast Fourier transform analysis have been utilized, but these
e-mail: todd.rook@goodrich.com
methods are computationally sensitive. Therefore, semianalytical single- and multiterm
harmonic balance methods are developed for an efficient construction of the frequency
Rajendra Singh responses. First, an analytical single-degree-of-freedom, linear time-varying system
Fellow ASME,
model is developed for a spur gear pair in terms of the dynamic transmission error.
Acoustics and Dynamics Laboratory,
Harmonic solutions are then derived and validated by comparing with numerical inte-
Department of Mechanical Engineering,
gration results. Next, harmonic solutions are extended to a six-degree-of-freedom system
The Ohio State University,
model for the prediction of (normal) mesh loads, friction forces, and pinion/gear dis-
Columbus, OH 43210
placements (in both line-of-action and off-line-of-action directions). Semianalytical pre-
e-mail: singh.3@osu.edu
dictions compare well with numerical simulations under nonresonant conditions and
provide insights into the interaction between sliding friction and mesh stiffness.
关DOI: 10.1115/1.2988478兴

1 Introduction 2 Problem Formulation

Sliding friction acts as an excitation to spur and helical gear
dynamics, as described by Houser et al. 关1兴, Velex and Cahouet
关2兴, Velex and Sainsot 关3兴, and Lundvall et al. 关4兴. Earlier, Vaishya
and Singh 关5–7兴 developed a single-degree-of-freedom 共SDOF兲
spur gear model with rectangular mesh stiffness k共t兲 and sliding
friction profiles; they solved the dynamic transmission error
共DTE兲 ␦共t兲 response by using the Floquet theory 关5兴 and multi-
term harmonic balance method 共MHBM兲 关6兴. Their work was re-
cently refined and extended to helical gears in our papers 关8,9兴
where closed-form solutions of ␦共t兲 for a SDOF system are de-
rived under friction excitation. The equal load sharing assumption
关5–9兴 yields simplified expressions and analytically tractable so-
lutions, but these do not describe realistic conditions. This particu-
lar deficiency was partially overcome in our articles 关10,11兴 where
we proposed a multidegree-of-freedom 共MDOF兲 model with real-
istic time-varying k共t兲 and sliding friction functions. However, we
utilized numerical integration and fast Fourier transform 共FFT兲
analysis methods in prior studies 关8–11兴 that are often computa-
tionally sensitive. Hence, a semianalytical algorithm based on
MHBM is highly desirable for an efficient construction of fre-
quency responses without any loss of generality. Recently, Velex
and Ajmi 关12兴 implemented a similar harmonic analysis to ap-
proximate the dynamic factors in helical gears 共based on tooth
loads and quasistatic transmission errors兲. Their work, however,
does not describe the multidimensional system dynamics or in-
clude the frictional effect, which may lead to “multiplicative”
terms as described earlier. The prime objective of this article is
thus to extend prior publications 关5,10兴. In particular, we intend to
develop semianalytical harmonic balance solutions to the 6DOF
spur gear model 关10兴 with realistic mesh stiffness k共t兲 and sliding
friction functions.
Contributed by the Power Transmission and Gearing Committee of ASME for
publication in the JOURNAL OF MECHANICAL DESIGN. Manuscript received October 16,
2007; final manuscript received June 10, 2008; published online October 7, 2008.
Review conducted by Philippe Velex.
Gear dynamic models are usually described by periodic differ-
ential equations 关13–15兴 due to significant variations in mesh stiff-
ness k共t兲 and mesh damping c共t兲 within the fundamental period tc
共one mesh cycle兲. Additionally, dynamic friction force F f 共t兲 and
torque M f 共t兲 also undergo periodic variations, with the same pe-
riod tc, due to changes in normal mesh loads and coefficient of
friction ␮, as well as a reversal in the direction of F f 共t兲 at the pitch
point 关5–10兴, especially in spur and helical gears. For the sake of
illustration, a unity ratio NASA spur gear 共with tip relief and
contact ratio ␴ around 1.6兲 is used as an sample case for this
study; refer to Table 2 of Ref. 关10兴 for its parameters. Major pe-
riodic excitations, namely, the mesh stiffness k共t兲 of the two mesh-
ing tooth pairs and the frictional functions f共t兲, are shown for the
sample case in Figs. 1共a兲 and 1共b兲, respectively. Here, points A
and D are the starting and ending points of one complete mesh
event; points B and C are the lowest and highest points of single
tooth contact; P is the pitch point, where friction force changes
direction. Refer to Fig. 2共a兲 for geometric locations of the points
along the line-of-action 共LOA兲 direction, while the off-line-of-
action 共OLOA兲 direction is the frictional direction. Define tooth
pairs 1 and 0 as the pairs rolling along lines AC and CD, respec-
tively. The effective stiffness function ki共t兲 of each meshing tooth
pair can be computed by ki共t兲 = Ns,i共t兲 / 关rb␪s,i共t兲兴, 共i = 0, 1兲 关10兴,
where rb is the base radius and Ns,i共t兲 and ␪s,i共t兲 are the static
normal contact force and angular deflection computed by a static
finite element/contact mechanics 共FE/CM兲 analysis 关16兴. Modifi-
cations to the tooth profiles, such as the linear tip relief, are char-
acterized by changes in the effective Ns,i共t兲 and ␪s,i共t兲 profiles.
Hence the resulting ki共t兲 includes the effect of the tip relief as
well. Refer to Ref. 关10兴 for the calculation procedure of ki共t兲.
Derivations of f共t兲 will be explained later in Eqs. 共4a兲–共4e兲 along
with key definitions of the system model.
The fundamental nature of the linear time-varying 共LTV兲 sys-
tem is illustrated in Fig. 2共b兲, and the system model is described in
our paper 关10兴. Key assumptions include the following: 共i兲 The

Journal of Mechanical Design Copyright © 2008 by ASME DECEMBER 2008, Vol. 130 / 122601-1

Downloaded 03 Dec 2008 to 164.107.168.20. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Fig. 1 „a… Periodic mesh stiffness functions of the spur gear
pair example „with tip relief… given nominal pinion torque T̄p
= 550 lb in. Key: blue dashed line, k0„t…; red solid line, k1„t…. „b…
Periodic frictional functions. Key: blue dashed line, f0„t…; green
dashed-dotted line, f1„t…; red solid line, f2„t….

Fig. 2 „a… Snap shot of the contact pattern „at t = 0… for the
pinion and gear are rigid disks. 共ii兲 Vibratory motions are small in sample spur gear pair. „b… Normal „mesh… and friction forces of
comparison to the nominal motion; this would lead to a LTV the analytical spur gear system model.
model. Only nonresonant conditions are considered since signifi-
cant dynamic motions at resonance may not only change the con-
tact pattern 共such as the effective center distance兲 but also cause
teeth separation, which would introduce nonlinear interactions and X pi共t兲 = LXA + 共S − i兲␭ + mod共⍀̄ prbpt,␭兲, i = 0, . . . ,S = floor共␴兲
further complicate the formulation. 共iii兲 The Coulomb friction is 共1e兲
assumed with a constant ␮ although the sign is reversed at the
pitch point. Xgi共t兲 = LYC + i␭ − mod共⍀̄grbgt,␭兲, i = 0, . . . ,S = floor共␴兲
The governing SDOF equation in terms of DTE ␦共t兲 = rbp␪ p共t兲
− rbg␪g共t兲 is given below, where subscripts p and g correspond to 共1f兲
the pinion and gear, respectively; ␪ is the vibratory component of S S
the rotation; T̄ and ⍀̄ are the nominal torque and rotation speed; ␭
is the base pitch; and J is the moment of inertia. Note that Je is
T共t兲 = Jb 兺
i=0
关ci共t兲␧˙ 共t兲 + ki共t兲␧共t兲兴 + 兺 sgn关mod共⍀̄ r
i=0
p bpt,␭兲

defined differently from that in Ref. 关10兴 for a compact formula-

tion of Eq. 共1a兲, + 共S − j兲␭ − LAP兴␮关ci共t兲␧˙ 共t兲 + ki共t兲␧共t兲兴关X pi共t兲Jgrbp
+ Xgi共t兲J prbg兴 共1g兲
S S

Je␦¨ 共t兲 + Jb 兺 关c 共t兲␦˙ 共t兲 + k 共t兲␦共t兲兴 + ␮兺 sgn关mod共⍀̄ r

i=0
i i
i=0
p bpt,␭兲
Here, the jth tooth pair passes though the pitch point P during the
meshing event, and the reversal at P is characterized by the sign
function “sgn” with a constant coefficient of Coulomb friction ␮
+ 共S − j兲␭ − LAP兴关ci共t兲␦˙ 共t兲 + ki共t兲␦共t兲兴关X pi共t兲Jgrbp 关5兴. The modulus function, defined as mod共t , tc兲 = t − tc · floor共t / tc兲,
is used to formulate the periodic friction force F f 共t兲 and the mo-
+ Xgi共t兲J prbg兴 = Te + T共t兲 共1a兲
ment arm X共t兲 at the fundamental period of tc. The “floor” func-
tion rounds off the contact ratio ␴ to the nearest integers toward a
J e = J pJ g 共1b兲 lower value. Also, T共t兲 of Eq. 共1g兲 represents the time-varying
component of the forcing function due to the unloaded static
Jb = Jgr2bp + J pr2bg 共1c兲 共manufacturing兲 transmission error ␧共t兲. Finally, geometric lengths
LXA, LYC, and LAP of Fig. 2共a兲 could be found as follows, where
u is the gear ratio, l is the center distance, ␣wt is the transverse
Te = T̄ pJgrbp + T̄gJ prbg 共1d兲 operating pressure angle, and ra is the outside radius,

122601-2 / Vol. 130, DECEMBER 2008 Transactions of the ASME

Downloaded 03 Dec 2008 to 164.107.168.20. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 LXA = l sin ␣wt − 冑r2ag − r2bg 共2a兲 Each periodic function, ki共␶兲, ci共␶兲, f i共␶兲, or ␦共␶兲, now has a pe-
riod of T = 1. Figure 1共b兲 shows typical f 0共␶兲, f 1共␶兲, and f 2共␶兲
LYC = 冑r2ag − r2bg + ␭ 共2b兲 functions, which describe the periodic moment arm and sliding
friction excitations for the sample case 关10兴.
ul sin ␣wt
LAP = 冑r2ag − r2bg − 共2c兲 3.1 Direct Application of Multiterm Harmonic Balance.
u+1 Define the Fourier series expansions of the periodic ki共␶兲 and ci共␶兲
Observe that Eqs. 共1a兲–共1g兲 significantly differs from the clas- in Eqs. 共4a兲–共4e兲 up to N mesh harmonics as follows, where the
sical Hill’s equation 关13兴 in several ways. First, the periodic ki共t兲 angular frequency ␻n = 2␲n 共in rad/s兲 and n is the mesh order:
is not confined to a rectangular wave assumed by Vaishya and N N
Singh 关5–7兴 or a simple sinusoid as in Mathieu’s equation 关1兴.
Instead, Eqs. 共1a兲–共1g兲 should describe realistic, yet continuous, ki共␶兲 = Aki0 + 兺
n=1
Akin cos共␻n␶兲 + 兺B
n=1
kin sin共␻n␶兲 共5a兲
profiles of Fig. 1共a兲 resulting from a detailed FE/CM analysis
关16兴. Hence, multiple harmonics of ki共t兲 should be considered.
冕
1
Second, the periodic viscous ci共t兲 term should dissipate vibratory
Aki0 = ki共␶兲d␶ 共5b兲
energy due to the sliding friction besides its kinematic effect.
0
␦i共t兲ci共t兲 and 兺i=1
S ˙
Third, the 兺i=1 S
␦i共t兲ki共t兲 terms of Eqs. 共1a兲–共1g兲

冕
incorporate combined 共but phase correlated兲 contributions from 1
all 共yet changing兲 tooth pairs in contact. Consequently, the relative Akin = 2 kin cos共␻n␶兲d␶ 共5c兲
phase between neighboring tooth pairs should play an important 0
role in the resulting response ␦共t兲. Fourth, multiplicative effects

冕
between ki共t兲, ci共t兲, Xi共t兲, and ␦共t兲 should result in higher mesh 1
harmonics, which poses difficulty in constructing closed-form so- Bkin = 2 kin sin共␻n␶兲d␶ 共5d兲
lutions. Lastly, the coupling between T共t兲 and ␧共t兲 in Eq. 共1g兲 0
indicates that the frictional forces and moments reside on both
sides of Eq. 共1a兲 as either periodically varying parameters or ex- N N
ternal excitations, thus posing further mathematical complications. ci共␶兲 = 2␨冑ki共␶兲Ie = Ac i0
+ 兺A cin cos共␻n␶兲 + 兺B cin sin共␻n␶兲
n=1 n=1

共6a兲
3 Semianalytical Solutions to the SDOF Spur Gear
Dynamic Formulation
冕
1

Consider the sample case with only the mean load Te, i.e., Adi0 = ci共␶兲d␶ 共6b兲
T共t兲 = 0 or ␧共t兲 = 0. Equations 共1a兲–共1g兲 can be rewritten over one 0

complete mesh cycle 0 艋 t 艋 tc, as follows for the sample case

冕
1
with two mesh tooth pairs, where me and Fe are the effective mass
and force; E1, E2, and E3 are gear constants, Acin = 2 cin cos共␻n␶兲d␶ 共6c兲
0
me␦¨ 共t兲 + 关c0共t兲␦˙ 共t兲 + k0共t兲␦共t兲兴关1 + E1 + E3t兴 + 关c1共t兲␦˙ 共t兲

冋 冉 冊册 冕
1
LAP cin sin共␻n␶兲d␶ 共6d兲
+ k1共t兲␦共t兲兴 1 + 共E2 + E3t兲sgn t − = Fe 共3a兲 Bcin = 2
⍀̄ prbp 0

The f i共␶兲 functions could be expanded explicitly as shown be-

E1 = ␮关共LXA + ␭兲Jgrbp + LYCJ prbg兴/Jb 共3b兲 low,
E2 = ␮关LXAJgrbp + 共LYC + ␭兲J prbg兴/Jb 共3c兲 N

兺
1 1
f 0共 ␶ 兲 = − sin共␻n␶兲 共7a兲
E3 = ␮⍀̄ prbp共Jgrbp − J prbg兲/Jb 共3d兲 2 n=1 n␲

me = Je/Jb 共3e兲 N
4 sin共␻n␶ P兲
Fe = Te/Jb 共3f兲
f 1共 ␶ 兲 = 1 − 2 ␶ P − 兺
n=1 ␻n
cos共␻n␶兲

Next, reformulate Eqs. 共3a兲–共3f兲 in terms of the dimensionless N

4关cos共␻n␶ P兲 − 1兴
time ␶ = t / tc, such that ␦⬘共␶兲 = d␦ / d␶ = tc␦共t兲 and ␦⬙共␶兲 = d2␦ / d2␶ + 兺 ␻n
sin共␻n␶兲 共7b兲
= t2␦¨ 共t兲, n=1

再兺
c

me␦⬙共␶兲 + tc关c0共␶兲␦⬘共␶兲 + tck0共␶兲␦共␶兲兴关1 + E1 + tcE3 f 0共␶兲兴 N

1 4
+ tc关c1共␶兲␦⬘共␶兲 + tck1共␶兲␦共␶兲兴关1 + E2 f 1共␶兲 + tcE3 f 2共␶兲兴 = t2c Fe f 2共␶兲 = − ␶2P + 2 关1 − ␻n␶ P sin共␻n␶ P兲
2 ␻n n=1
共4a兲
N

f 0共␶兲 = mod共␶,1兲 共4b兲 − cos共␻n␶ P兲兴cos共␻n␶兲 + 兺 关␻ ␶ n P cos共␻n␶ P兲 − sin共␻n␶ P兲

冎
n=1
f 1共␶兲 = sgn关mod共␶,1兲 − ␶ P兴 = sgn关f 0共␶兲 − ␶ P兴 共4c兲

f 2共␶兲 = mod共␶,1兲sgn关mod共␶,1兲 − ␶ P兴 = f 0共␶兲f 1共␶兲 共4d兲 − 0.5␻n兴sin共␻n␶兲 共7c兲

␶ P = LAP/␭ 共4e兲 Finally, assume that the periodic dynamic response ␦共␶兲 is of

Journal of Mechanical Design DECEMBER 2008, Vol. 130 / 122601-3

Downloaded 03 Dec 2008 to 164.107.168.20. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 the following form:
N N

␦ 共 ␶ 兲 = A ␦0 + 兺 A␦
n=1
n cos共␻n␶兲 + 兺 B␦
n=1
n sin共␻n␶兲 共8兲

Substitute Fourier series expansions of Eqs. 共5a兲–共5d兲, 共6a兲–

共6d兲, 共7a兲–共7c兲, and 共8兲 into Eqs. 共4a兲–共4e兲 and balance the mean
and harmonic coefficients of sin共␻n␶兲 and cos共␻n␶兲. This converts
the linear periodic differential equation into easily solvable linear
algebraic equations 共as expressed below兲 where K = h is a square
matrix of dimension 共2N + 1兲 consisting of known coefficients of
ki共␶兲, ci共␶兲, and f i共␶兲. By calculating the inverse of K = h, the 2N
+ 1 Fourier coefficients of ␦共␶兲 could be computed at any gear
mesh harmonic 共n兲,

冤 冥冤冥
A ␦0 Fe
A ␦1 0
B ␦1 0
= h兴
关K = 共9兲
... ...
A ␦N 0
B ␦N 0

3.2 Semianalytical Solutions Based on One-Term HBM.

Next, we construct one-term HBM 关17兴 solutions to conceptually
illustrate the method. Set the harmonic order N = 1 共only the fun-
damental mesh, in addition to the mean term兲 in Eqs. 共5a兲–共5d兲,
共6a兲–共6d兲, 共7a兲–共7c兲, 共8兲, and 共9兲 and balance the harmonic terms
Fig. 3 Semianalytical versus numerical solutions for the
in Eqs. 共4a兲–共4e兲. This leads to a K = h matrix of dimension 3. Three
of its typical coefficients are given as follows and the rest could be SDOF model, expressed by Eq. „3…, given T̄p = 550 lb in., Ω̄p
= 500 rpm, and ␮ = 0.04. „a… Time-domain responses; „b… mesh
found in a similar manner: harmonics in frequency domain. Key: blue solid line and 䉭,
1
numerical simulations; black dashed line and 䊐, semianalytical
Kh11 = 2 共tcAk11A f21E3 + Bk11B f11E2 − 共tcBk01E3/␲兲 + 2Ak00 + 2Ak10 solutions using one-term HBM; red dashed-dotted line and 䊊,
semianalytical solutions using five-term HBM.
+ Ak11A f11E2 + 2Ak00E1 + tcAk00E3 + 2Ak10A f10E2
+ 2tcAk10A f20E3 + tcBk11B f21E3兲 共10a兲

MHBM algorithm 关19,20兴. First, define mean and vibratory speed

Kh21 = Ak01 + Ak11 + A01E1 + Ak10A f11E2 + tcAk10A f21E3
variables ⍀̃ and ␽, where ␷ is the subharmonic index,
+ Ak11A f10E2 + tcAk11A f20E3 + 共tcAk01E3/2兲 共10b兲
⍀̃ = 2␲/共␷tc兲 共11a兲
Kh31 = Bk11A f10E2 + Bk11 + tcBk11A f20E3 + Ak10B f11E2
␽ = ⍀̃t 苸 关0,2␲兲 共11b兲
tc tcBk01E3
+ tcAk10B f21E3 + Bk01E1 + Bk01 − Ak00E3 +
␲ 2 ¯ = mod共␷␽/2␲,1兲
␽ 共11c兲
共10c兲 Also, define the differential operator “ˆ” as
The Fourier series coefficients of ␦共␶兲 are then obtained by
d␦
inverting K = h. Figure 3 shows that the one-term HBM solution ␦ˆ = = ⍀̃−1␦˙ 共12a兲
predicts the overall tendency 共mean and first harmonic兲 fairly well d␽
when compared with numerical simulations at T̄ p = 550 lb in. and
␦˙ = ⍀̃␦ˆ 共12b兲
⍀̄ p = 500 rpm. This confirms that the one-term HBM 共and likewise
the MHBM兲 approach coverts the periodic differential Eqs. Equations 共3a兲–共3f兲 are then converted into the following form,
共4a兲–共4e兲 with multiple interacting coefficients into simpler alge- where E4 = ⍀̃−1E3:
braic calculations that are computationally more efficient than nu-
merical integrations and subsequent FFT analyses. Thus, the semi- ⍀̃2me␦ˆ 共␽兲 + 关⍀̃c0共␽
¯ 兲␦ˆ 共␽兲 + k 共␽
0
¯ 兲␦共␽兲兴关1 + E + E ␽
1
¯
4 兴
analytical solution provides an effective design tool. Also, most
¯ 兲␦ˆ 共␽兲 + k 共␽
+ 关⍀̃c1共␽ ¯ 兲␦共␽兲兴关1 + 共E + E ␽
¯ ¯
coefficients of Eqs. 共10a兲–共10c兲 show sideband effects that are 1 2 4 兲sgn共␭␽/LAP
introduced by k共t兲 共or c共t兲兲 and f i共t兲 functions.
− 1兲兴 = Fe 共13兲
3.3 Iterative MHBM Algorithm. When N 艋 5, we can utilize or expressed more compactly as
a symbolic software 关18兴 to balance multiple harmonic terms and
= h. However, the computational cost involved with each ˆ
calculate K ⍀̃2me␦共␽兲 + ⍀̃C共␽兲␦ˆ 共␽兲 + K共␽兲␦共␽兲 = Fe 共14a兲
element of K = h increases by N3 due to the triple multiplication of
periodic coefficients in Eqs. 共4a兲–共4e兲. Consequently, for higher N ¯ 兲共1 + E + E ␽
¯ ¯ ¯ ¯
C共␽兲 = c0共␽ 4 兲 + c1共␽兲关1 + 共E2 + E4␽兲sgn共␭␽/LAP
共say, ⬎5兲, a direct computation of K = h becomes inefficient and 1

thus inadvisable. Instead, we apply a matrix-based iterative − 1兲兴 共14b兲

122601-4 / Vol. 130, DECEMBER 2008 Transactions of the ASME

Downloaded 03 Dec 2008 to 164.107.168.20. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 ¯ 兲共1 + E + E ␽
K共␽兲 = k0共␽ ¯ ¯ ¯ ¯
1 4 兲 + k1共␽兲关1 + 共E2 + E4␽兲sgn共␭␽/LAP

− 1兲兴 共14c兲
For the MHBM, the discrete Fourier transform 共DFT兲 is applied to
transform from the time domain to the frequency domain, and a
discretized Fourier expansion matrix F= transforms the frequency
domain back to the time domain,

冤 冥
1 sin共␽1兲 cos共␽1兲 . . . sin共N␽1兲 cos共N␽1兲
1 sin共␽2兲 cos共␽2兲 . . . sin共N␽2兲 cos共N␽2兲
F= =
] ] ]
] ]
1 sin共␽ M 兲 cos共␽ M 兲 . . . sin共N␽ M 兲 cos共N␽ M 兲
共15a兲

冦 冧
␦ 共 ␽ 1兲
␦ 共 ␽ 2兲
␦គ ⬅ = F= ⌬គ 共15b兲
]
␦共␽ M 兲

冦 冧
␦ˆ 共␽1兲
Fig. 4 Semianalytical versus numerical solutions for the
␦ˆ 共␽2兲 SDOF model as a function of pinion speed with ␮ = 0.04. „a…
␦គ̂ ⬅ =⌬
= F= D គ 共15c兲 Mesh order n = 1, „b… n = 2, „c… n = 3, and „d… n = 4. Key: red ⴛ,
] numerical simulations; blue solid line, semianalytical solutions
␦ˆ 共␽M 兲 using five-term HBM.

冦冧
␦ˆ 共␽1兲
observed in the numerical simulations. The semianalytical solu-
tions are efficiently used in Fig. 4 for parametric studies of ␦共␶兲 at
␦ˆគ ⬅ ␦ˆ 共␽2兲 = 2⌬គ
= F= D 共15d兲
the gear mesh harmonics over a range of ⍀̄ p, and these calcula-
]
tions are indeed achieved with reduced computational cost.
␦ˆ 共␽M 兲
Here, ␽i = i2␲ / M and M 艌 2N + 1, ⌬ គ is a vector of 2N + 1 Fourier 4 Semianalytical Solutions to 6DOF Spur Gear Dy-
coefficients, and the Fourier differentiation matrix D= is given as namic Formulation

冤 冥
0 0 0 ... 0 0 A careful examination of the 6DOF model reported in Ref. 关10兴
0 0 − 1 ... 0 0 reveals that the dynamic bearing displacement y共␶兲 in the OLOA
direction depends on the LOA dynamics since the friction force
0 1 0 ... 0 0 ␮k共␶兲␦共␶兲 acts in the OLOA direction. Conversely, the dynamic
= =
D 共15e兲
] ] ]
0 0 bearing displacement x共␶兲 in the LOA direction is not influenced
0 0 0 ... 0 − N by 共i.e., is decoupled from兲 the OLOA motion y共␶兲. This is be-
cause of the following reasons: First, no off-diagonal term exists
0 0 0 ... N 0
in the effective shaft-bearing stiffness matrix that couples the

冤 冥
LOA and OLOA dynamics 关8兴. Second, the friction force de-
0 0 0 ... 0 0
scribed by the Coulomb model is independent of y共␶兲. Third, re-
0 −1 0 ... 0 0 call our assumption that the dynamic responses 共such as y共␶兲兲
0 0 − 1 ... 0 0 have no influence on the kinematics 共such as the center distance兲.
=2=
D 共15f兲 Such “one-way coupling” between LOA and OLOA dynamics im-
] ] ]
0 0
plies that we can reduce the 6DOF model into a simpler 3DOF
0 0 0 . . . − N2 0 model 关21兴 in terms of x p共␶兲, xg共␶兲, and ␦共␶兲. Dynamic responses
0 0 0 ... 0 − N2 of the 3DOF model could further be used to determine the friction
Applying the DFT to the equation of motion yields the following forces that excite the OLOA dynamics.
MHBM equations where F= + is the Moore–Penrose or pseudoin- For the 3DOF system, when the composite DTE ␦共␶兲
verse of the DFT matrix, = rbp␪ p共␶兲 − rbg␪g共␶兲 + x p共␶兲 − xg共␶兲 can be approximated by the
semianalytical solution ␦共␶兲 = rbp␪ p共␶兲 − rbg␪g共␶兲 of the SDOF
关⍀̃2meD
= 2 + ⍀̃F= +C = F= 兴⌬គ = F= +Fគ e
= + F= +K
= F= D 共16a兲 model, a similar one-way coupling is created where ␦共␶兲 is decou-
pled from x p共␶兲 and xg共␶兲, while x p共␶兲 and xg共␶兲 are excited by
= ⬅ diag共兵K共␽1兲
K K共␽2兲 ¯ K共␽ M 兲其兲 共16b兲 ␦共␶兲 in terms of the normal loads. This suggests that when the
excitation 共mesh兲 frequencies do not excite any coupled
= ⬅ diag共兵C共␽1兲
C C共␽2兲 ¯ C共␽ M 兲其兲 共16c兲 transverse-torsional modes in the LOA direction, the harmonic
solutions of the SDOF system could be extended to predict trans-
Fគ e = 关Fe 0 ¯ 0 0兴T 共16d兲 lational responses in both LOA and OLOA directions by utilizing
Figure 3 shows that the five-term HBM solutions compare well the one-way coupling effects. Note that the semianalytical method
with numerical simulations. Likewise, an increase in N captures analyzes the 6DOF system 关10兴 as a 5DOF model as it calculates
higher frequency components around the tenth mesh harmonic, as the ␦共t兲 rather than absolute angular displacements ␪ p共t兲 or ␪ p共t兲.

Journal of Mechanical Design DECEMBER 2008, Vol. 130 / 122601-5

Downloaded 03 Dec 2008 to 164.107.168.20. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 In order to quantify the nonresonant condition under which the
one-way coupling exists in the LOA direction, we use a 6DOF
linear time-invariant spur gear model of Fig. 5共a兲 and focus on its
subset of a unity gear pair 共3DOF model兲 to study the natural
frequency distribution. We define the following parameters for the
关kmM + 共2km + KB兲me兴 ⫾ 冑关kmM + 共2km + KB兲me兴2 − 4MmekmKB
⍀N1,N3
2
= 共17a兲
2Mme
KB
⍀N2
2
= 共17b兲
M
Mode 1 共⍀N1兲 and Mode 3 共⍀N3兲 correspond to the first and
second coupled transverse-torsional modes, respectively, and
Mode 2 共⍀N2兲 is the purely transverse mode. Also, the natural
frequency of the corresponding SDOF torsional system can be
estimated as ⍀NS = 冑km / me. Figure 5共b兲 compares the natural fre-
quencies 共⍀N兲 of the 3DOF and SDOF models as a function of the
stiffness ratio KB / km. For an easy comparison with the excitation
共mesh兲 frequency ␻n, which is determined by the nominal pinion
rpm, all natural frequencies are converted in Fig. 5共b兲 from rad/s
into rpm units.
Fig. 5 „a… 6DOF spur gear pair model and its subset of unity
gear pair „3DOF model… used to study the natural frequency
distribution. „b… Natural frequencies ΩN as a function of the
stiffness ratio KB / km. Key: black solid line, ΩSN of SDOF sys-
tem „torsional only, in terms of DTE…; blue dashed line, ΩN1 of
3DOF system; green dotted line, ΩN2 of 3DOF system; red
dashed-dotted line, ΩN3 of 3DOF system.
122601-6 / Vol. 130, DECEMBER 2008
Downloaded 03 Dec 2008 to 164.107.168.20. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
sample case: mass of pinion 共gear兲 m p = mg = M; moment of inertia
J p = Jg = J; basic radius rbp = rbg = R; the equivalent mass me
= J / 共2R2兲; time-averaged mesh stiffness km 共T p = 550 lb in.兲; shaft-
bearing stiffness KBp = KBg = KB. The natural frequencies of the
3DOF system are found as follows 关21兴:
Observe that ⍀N of the SDOF model asymptotically approaches
only one of the natural frequencies of the 3DOF model. Conse-
quently, when the excitation frequency coincides with either of the
two resonances not predicted by the SDOF, larger errors are ex-
pected when extending the solutions of the SDOF model into
other dimensions. Such nonresonant restriction of the proposed
algorithm is, however, consistent with our assumption that vibra-
tory motions are small in comparison to the mean motion. When
KB / km ⬍ 1 共or KB / km ⬎ 10兲, ⍀NS asymptotically approaches ⍀N3
共or ⍀N1兲. Moreover, ␻n does not excite any resonance in Zone I
共␻n Ⰶ ⍀N1兲 and Zone III 共␻n Ⰷ ⍀N3兲. Additionally, nonresonant
Zone II could be found for both “soft” 共KB / km ⬍ 1, ⍀N2 Ⰶ ␻n
Ⰶ ⍀NS兲 and “stiff” 共KB / km ⬎ 10, ⍀NS Ⰶ ␻n Ⰶ ⍀N2兲 shaft-bearing
cases. In these nonresonant zones, the semianalytical solution ␦共␶兲
of the SDOF model could be extended to the 6DOF system. Fig-
ure 6 compares the five-term HBM prediction of ␦共␶兲 for the
SDOF system with numerical simulation for a 6DOF model 关10兴
for two limiting values of KB / km. When KB / km = 100, prediction
matches well with numerical simulation. However, when KB / km
= 0.37 共i.e., nominal case of Ref. 关10兴兲, a good correlation is ob-
served only away from system resonances ⍀N in Zones I–III.
The dynamic normal 共mesh兲 loads Ni共␶兲 of the pinion and gear
are equal in magnitude but opposite in direction 关10兴. To account
Fig. 6 Semianalytical versus numerical solutions for the 6DOF
model as a function of Ω̄p with ␮ = 0.04. „a… Mesh order n = 1, „b…
n = 2, „c… n = 3, and „d… n = 4. Key: blue solid line, predictions
using five-term HBM; green 䊊, numerical simulations †4‡ with
nominal KB „KB / km = 0.37…; red ⴛ, numerical simulations with
stiff KB „KB / km = 100….
Transactions of the ASME
 for interactions between ki共␶兲 and ␦共␶兲 as well as between ci共␶兲
and ␦˙ 共␶兲, Fourier series is expanded to find the N 共␶兲 terms up to i
2N mesh harmonics,
2N

N 共␶兲 = k 共␶兲␦共␶兲 + c 共␶兲␦˙ 共␶兲 = A

i i i Ni0 + 兺A
n=1
Nin cos共␻n␶兲

2N

+ 兺B
n=1
Nin sin共␻n␶兲 共18兲

Similarly, the dynamic friction forces F fi共␶兲 are expanded in Eqs.

共19a兲 and 共19b兲. Since tooth pair 1 is associated with a periodic
change of friction force at the pitch point, F f1共␶兲 is expanded up to
3N mesh harmonics due to a multiplication of k1共␶兲, ␦共␶兲, and
f 1共␶兲,
2N 2N

F f0共␶兲 = ␮N0共␶兲 = AF00 + 兺

n=1
AF0n cos共␻n␶兲 + 兺B
n=1
F0n sin共␻n␶兲

共19a兲
3N Fig. 7 Semianalytical versus numerical solutions of the LOA
F f1共␶兲 = ␮N1共␶兲f 1共␶兲 = AF10 + 兺A
n=1
F1n cos共␻n␶兲 displacement xp for the 6DOF model as a function of Ω̄p with
KB / km = 100, ␮ = 0.04. „a… Mesh order n = 1, „b… n = 2, „c… n = 3, and
3N
„d… n = 4. Key: red ⴛ, numerical simulations; blue solid line, pre-
dictions using five-term HBM.
+ 兺B
n=1
F1n sin共␻n␶兲 共19b兲

In the LOA 共or x兲 direction, the transfer function 共at frequency 2N

␻n兲 with N p共␶兲 as the input and x p共␶兲 as the output is found by
using the corresponding linear time-invariant model as follows
y p共␶兲 = M py共0兲AF00 + M py共0兲AF10 + 兺M
n=1
py 共␻n兲AF0n cos关␻n␶

关10兴, where K pBx and ␨ pBx are the shaft-bearing stiffness and 2N

兺M
damping terms:
+ ␣ py共␻n兲兴 + py 共␻n兲BF0n sin关␻n␶ + ␣ py共␻n兲兴
Xp t2c n=1
共␻兲 = 共20a兲
Np 共− ␻2m p + K pBxt2c 兲 + j2␻tc␨ pBx冑K pBxm p 3N

Magnitude M pxn and phase ␣ pxn at the nth mesh order, ␻n

+ 兺M
n=1
py 共␻n兲AF1n cos关␻n␶ + ␣ py共␻n兲兴
= 2␲n␶ 共rad/ s兲, are
3N

M px共␻n兲 =
t2c
冑共− ␻2nmp + KpBxt2c 兲2 + 4␻2nt2c ␨2pBxKpBxmp 共20b兲 + 兺M
n=1
py 共␻n兲BF1n sin关␻n␶ + ␣ py共␻n兲兴 共23兲

Equations 共21兲 and 共23兲 confirm that multiplications between pe-

2␻ntc␨ pBx冑K pBxm p riodic coefficients ki共␶兲 共or ci共␶兲兲, ␦共␶兲 共or ċi共␶兲兲, and f i共␶兲 lead to
␣ px共␻n兲 = tan −1
共20c兲
␻2nm p − K pBxt2c higher mesh harmonic components, which are commonly ob-
served in spur gears 关1兴.
Thus the pinion bearing displacement x p共␶兲 in the LOA 共or x兲 Figure 7 compares the semianalytical x p共␶兲 with numerical pre-
direction is expanded as
diction 关10兴 as a function of ⍀̄ p with KB / km = 100. Good correla-
2N
tions are observed up to 16,000 rpm 共in Zone 1 with high KB / km兲
x p共␶兲 = M px共0兲共AN00 + AN10兲 + 兺M
n=1
px共␻n兲共AN0n + AN1n兲cos关␻n␶ including the LOA shaft-bearing resonances at n = 3 and n = 4.
Likewise, the semianalytical y p in the OLOA direction compares
2N well with numerical simulations in Fig. 8, where the shaft-bearing
+ ␣ px共␻n兲兴 + 兺M
n=1
px共␻n兲共BN0n + BN1n兲sin关␻n␶ + ␣ px共␻n兲兴 resonances are also observed at n = 3 and n = 4. Although good
correlations are usually expected for stiff shaft bearings, it is
worthwhile to examine the case where KB is comparable or less
共21兲
than km. By using the nominal parameters of Ref. 关10兴, where
In the OLOA 共or y兲 direction, the magnitude and phase of the KB / km = 0.37, semianalytical predictions are compared with nu-
transfer function from friction force F f 共␶兲 to pinion displacement merical simulations in Figs. 9 and 10 for LOA and OLOA re-
y p共␶兲 could be found at ␻n as sponses, respectively. Observe that semianalytical x p matches nu-
merical simulations only from system resonances ⍀N, as
t2c explained earlier in Fig. 5. Nonetheless, good correlation is ob-
M py共␻n兲 = 共22a兲
冑共− ␻2nmp + KpByt2c 兲2 + 4␻2nt2c ␨2pByKpBymp served in the OLOA direction over the operating speed range in
Fig. 10 since the bearing resonance dictates the y p dynamics.
2␻ntc␨ pBy冑K pBym p Finally, semianalytical predictions are used to examine the in-
␣ py共␻n兲 = tan−1 共22b兲 teractions between the sliding friction and tip relief 共embedded in
␻2nm p − K pByt2c the realistic mesh stiffness function k共t兲兲. Predicted displacement
Thus the pinion displacement y p共␶兲 in the OLOA 共or y兲 direction ratios of y p to x p for the first four harmonics are displayed in Fig.
is expanded as 11 for the sample case 共with long tip relief兲 关10兴 as compared to

Journal of Mechanical Design DECEMBER 2008, Vol. 130 / 122601-7

Downloaded 03 Dec 2008 to 164.107.168.20. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Fig. 8 Semianalytical versus numerical solutions of the OLOA
displacement yp for the 6DOF model as function of Ω̄p with
KB / km = 100, ␮ = 0.04. „a… Mesh order n = 1, „b… n = 2, „c… n = 3, and
„d… n = 4. Key: red ⴛ, numerical simulations; blue solid line, pre-
dictions using five-term HBM.
the perfect involute profile case 共without any tip relief兲 as a func-
tion of the mean torque T̄ p. Observe that for the perfect involute
profile, the y p / x p ratios remain almost constant over the entire
torque range, while applying tip relief significantly alters the
y p / x p ratios. A peak ranging from 0.8 up to 15 is observed for the
long tip relief case around the “optimal” load of 550 lb in., imply-
ing that an amplification due to the sliding friction takes place.
This can be explained by the fact that the application of tip relief
not only minimizes x p at the design load but also amplifies the
harmonic of y p due to an out-of-phase relationship between the
dynamic normal load N共␶兲 and dynamic friction forces F f 共␶兲 关10兴.
Fig. 9 Semianalytical versus numerical solutions of the LOA
displacement xp for the 6DOF model as a function of Ω̄p with
KB / km = 0.37, ␮ = 0.04. „a… Mesh order n = 1, „b… n = 2, „c… n = 3, and
„d… n = 4. Key: green 䊊, numerical simulations; blue solid line,
predictions using five-term HBM.
122601-8 / Vol. 130, DECEMBER 2008
Downloaded 03 Dec 2008 to 164.107.168.20. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
Fig. 10 Semianalytical versus numerical solutions of the
OLOA displacement yp for the 6DOF model as a function of Ω̄p
with KB / km = 0.37, ␮ = 0.04. „a… Mesh order n = 1, „b… n = 2, „c… n
= 3, and „d… n = 4. Key: green 䊊, numerical simulations; blue
solid line, predictions using five-term HBM.
5 Conclusion
An application of the harmonic balance method to the gear
dynamic models with multiple periodic coefficients leads to an
efficient algorithm to evaluate the effect of sliding friction on spur
gear dynamics 共as compared with the numerical integration
method 关10兴兲. Multiterm harmonic balance solutions based on the
SDOF linear time-varying model are successfully extended to a
6DOF system model for the predictions of 共normal兲 mesh loads,
friction forces, and bearing displacements in the LOA and OLOA
directions. A simplified 3DOF linear time-invariant model is used
Fig. 11 Semianalytical predictions for the ratio of OLOA dis-
placement xp to the LOA displacement xp for the 6DOF model
as a function of T̄p with KB / km = 0.37, ␮ = 0.04. „a… Mesh order
n = 1, „b… n = 2, „c… n = 3, and „d… n = 4. Key: red dashed line with
ⴛ, perfect involute gear; blue solid line with 䊊, gear with tip
relief.
Transactions of the ASME
 to explain the nonresonant operating conditions under which the
proposed algorithm is applicable. Semianalytical solutions also
provide new insights into the multiplicative interactions between
time-varying mesh stiffness 共say, with contributions from tip re-
lief兲 and sliding friction 共including friction force excitation and
varying torque兲. Such dynamic interactions explain the higher har-
monic components and the phase relationship between sliding
friction and mesh stiffness function. These are the main contribu-
tions of our paper. The methods of this article could be extended
to multimesh spur gear dynamics.
Acknowledgment
This article is based on a three-year study that was supported by
the U.S. Army Research Office 关22兴. We acknowledge the Gradu-
ate School of The Ohio State University for awarding the Presi-
dential Fellowship to the first author.
References
关1兴 Houser, D. R., Vaishya, M., and Sorenson, J. D., 2001, “Vibro-Acoustic Ef-
fects of Friction in Gears: An Experimental Investigation,” SAE Paper No.
2001-01-1516.
关2兴 Velex, P., and Cahouet, V., 2000, “Experimental and Numerical Investigations
on the Influence of Tooth Friction in Spur and Helical Gear Dynamics,” ASME
J. Mech. Des., 122共4兲, pp. 515–522.
关3兴 Velex, P., and Sainsot, P., 2002, “An Analytical Study of Tooth Friction Exci-
tations in Spur and Helical Gears,” Mech. Mach. Theory, 37, pp. 641–658.
关4兴 Lundvall, O., Strömberg, N., and Klarbring, A., 2004, “A Flexible Multi-Body
Approach for Frictional Contact in Spur Gears,” J. Sound Vib., 278共3兲, pp.
479–499.
关5兴 Vaishya, M., and Singh, R., 2001, “Analysis of Periodically Varying Gear
Mesh Systems With Coulomb Friction Using Floquet Theory,” J. Sound Vib.,
243共3兲, pp. 525–545.
关6兴 Vaishya, M., and Singh, R., 2001, “Sliding Friction-Induced Non-Linearity and
Journal of Mechanical Design
Downloaded 03 Dec 2008 to 164.107.168.20. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
Parametric Effects in Gear Dynamics,” J. Sound Vib., 248共4兲, pp. 671–694.
关7兴 Vaishya, M., and Singh, R., 2003, “Strategies for Modeling Friction in Gear
Dynamics,” ASME J. Mech. Des., 125, pp. 383–393.
关8兴 He, S., Gunda, R., and Singh, R., 2007, “Inclusion of Sliding Friction in
Contact Dynamics Model for Helical Gears,” ASME J. Mech. Des., 129, pp.
48–57.
关9兴 He, S., and Singh, R., 2008, “Dynamic Transmission Error Prediction of He-
lical Gear Pair Under Sliding Friction Using Floquet Theory,” ASME J. Mech.
Des., 130共5兲, 052603.
关10兴 He, S., Gunda, R., and Singh, R., 2007, “Effect of Sliding Friction on the
Dynamics of Spur Gear Pair With Realistic Time-Varying Stiffness,” J. Sound
Vib., 301, pp. 927–949.
关11兴 He, S., and Singh, R., 2007, “Dynamic Interactions Between Sliding Friction
and Tip Relief in Spur Gears,” DETC 2007-34077, Tenth International Power
Transmission and Gearing Conference, Las Vegas, Sept. 4–7.
关12兴 Velex, P., and Ajmi, M., 2007, “Dynamic Tooth Loads and Quasi-Static Trans-
mission Errors in Helical Gears—Approximate Dynamic Factor Formulae,”
Mech. Mach. Theory, 42共11兲, pp. 1512–1526.
关13兴 Richards, J. A., 1983, Analysis of Periodically Time-Varying Systems,
Springer, New York.
关14兴 Jordan, D. W., and Smith, P., 2004, Nonlinear Ordinary Differential Equations,
3rd ed., Oxford University Press, New York.
关15兴 Thomsen, J. J., 2003, Vibrations and Stability, 2nd ed., Springer, New York.
关16兴 EXTERNAL2D, CALYX software, 2003, “Helical3D User’s Manual,” ANSOL
Inc., Hilliard, OH, www.ansol.www
关17兴 Kim, T. C., Rook, T. E., and Singh, R., 2005, “Super- and Sub-Harmonic
Response Calculations for a Torsional System With Clearance Non-Linearity
Using Harmonic Balance Method,” J. Sound Vib., 281共3–5兲, pp. 965–993.
关18兴 MAPLE 10, symbolic software, 2005, Waterloo Maple Inc., Waterloo, ON.
关19兴 Padmanabhan, C., Barlow, R. C., Rook, T. E., and Singh, R., 1995, “Compu-
tational Issues Associated With Gear Rattle Analysis,” ASME J. Mech. Des.,
117, pp. 185–192.
关20兴 Duan, C., and Singh, R., 2005, “Super-Harmonics in a Torsional System With
Dry Friction Path Subject to Harmonic Excitation Under a Mean Torque,” J.
Sound Vib., 285共4–5兲, pp. 803–834.
关21兴 Kahraman, A., and Singh, R., 1991, “Error Associated With A Reduced Order
Linear Model of Spur Gear Pair,” J. Sound Vib., 149共3兲, pp. 495–498.
关22兴 Singh, R., 2005, “Dynamic Analysis of Sliding Friction in Rotorcraft Geared
Systems,” Army Research Office, Grant No. DAAD19-02-1-0334.
DECEMBER 2008, Vol. 130 / 122601-9