Professional Documents
Culture Documents
in the JOURNAL OF VIBRATION AND ACOUSTICS. Manuscript received August 2000; The stiffness matrix is represented by a mean value K0 and a
Revised September 2001. Associate Editor: L. A. Bergman. variational part Kv (t) as
68 Vol. 124, JANUARY 2002 Copyright 2002 by ASME Transactions of the ASME
k v i t 2k ai a
s1
s
i sin s i tb i s cos s i t , i1,2 (3)
b i s cos s c i 2p i sin s c i (4)
k g1 k 0 k g1 0 s
or four Fourier terms reasonably approximate the mesh stiffness
k v1 k v1 0 variation.
For the time-invariant case, the eigenvalue problem associated
Kv t k v 1 k v 1 k v 2 k v2 (2) with 1 is
0 k v2 k v2
K0 i i2 M i (5)
where k gi and k v i (t) are the mean and time-varying components
of the mesh stiffnesses, k i (t)k gi k v i (t) Fig. 2. where i are the natural frequencies. The vibration modes i are
The variational parts k v i (t) are periodic at the mesh frequency normalized as T MI with 1 , 2 , 3 . Applying the
i and expressed in Fourier series as modal transformation qu and using 3, Eq. 1 becomes
3
u n 2n u n 2 D
r1 s1
1
s
nr sin s 1 tE nrs cos s 1 t
s
D k g1 T
a s1
a s1
0
a s1
a s1
0
0
0
0
s
E k g1 T
0
0
0
0
b s1
b s1
0
b s1
b s1
s
F k g2 T
a s2
a s2
0
a s2
a s2
0
0
0
0
s
G k g2 T
0
0
0
0
b s2
b s2
0
b s2
b s2
(7)
Equation 6 is a set of coupled Hills equations subjected to
multi-frequency parametric excitations from two gear meshes.
Parametric excitations give rise to instabilities when harmonics
of the excitation frequencies are close to particular combinations
of the natural frequencies. Three types of instability are of most
interest: 1 primary instability i 2 p , 2 secondary instabil-
ity i p , and 3 combination instability i p q , where
p , q are different natural frequencies. Plots of the excitation
mesh frequency i versus the amplitude of the stiffness varia-
tion k ai illustrate the instability regions under operating conditions
for example, Fig. 3. Perturbation methods 8,17,18 are used in
Fig. 2 Modeling of mesh stiffnesses k i t k gi k vi t . c i are this study to determine the boundaries separating the stable and
contact ratios, k gi are average mesh stiffnesses, and p i T i are unstable regions. Floquet theory and numerical integration are
phasing angles. used to validate the analytical findings.
e i n s 1 t iE nrs e i n s 1 t e i n s 1 t
F nrs e i n s 2 t e i n s 2 t iG nrs e i n s 2 t
e i n s 2 t cc n1,2,3 (12)
Three different mesh conditions are examined.
D psp k g1 1 p 2 p 2 a 1s , F psp K g2 2 p 3 p 2 a 2s
E psp k g1 1 p 2 p 2 b 1s , G psp k g2 2 p 3p 2 b 2s
(19)
where 1 p , 2 p , 3 p are the element rotations in mode p . 1 p
1 p 2 p is the relative tooth deflection of the first mesh in
mode p . Similarly, 2 p 2 p 3 p represents the modal deflec-
tion in the second mesh. For the primary instability boundary Fig. 5 Comparison of instability regions for various contact
around 2 p , insertion of Eq. 19 into 16 yields ratios and mesh phasing. The parameters are in Table 1. ---c 1
c 2 1.5, h 0.5; c 1 1.1, c 2 1.9, h 0.9; -"-"-"- c 1 c 2 1.5,
p1p k g1 21 p a 11 k g2 22 p a 21 2 h 0.
k g1 21 p b 11 k g2 22 p b 21 2 / 2p (20)
The mesh deflections 1 , 2 in each mode can be observed from
the mode shapes Fig. 4. The two meshes in 1 are both in phase of its primary instability regions is between that of 1 and 3
and have smaller 1 , 2 than those of 3 , where the two meshes (1)
( 33 (221 ) (111 ) ). In addition, the mesh deflections in a vibra-
are both out of phase. Thus, 33(1)
(111 ) and the instability bound- tion mode are related to the modal strain energy U 1 k 1 21 /2,
aries around 2 3 have larger slope than those around 2 1 . Mode U 2 k 2 22 /2. Examination of Eq. 20 shows that vibration modes
2 has one mesh in phase and the other out of phase, so the size with more strain energy in the meshes have larger instability re-
gions and are more susceptible to parametric excitations. The
above results apply for mesh stiffness variations of arbitrary
shape.
For mesh stiffnesses having rectangular waveforms, one can
clearly identify the effects of contact ratios and mesh phasing on
the instability regions. Use of Eq. 4 in 20 yields
2
p
2
(21)
tively. Figure 6a shows the instability regions for R3/5. The
4 boundaries associated with 1 instabilities near 2 3 , 2 3 ,
2h (23)
p q
2
1 3 , and 3 are determined by Eq. 26. The primary insta-
bility associated with 2 2 3 2 occurs at 1 R 2
Depending on the sign of the product 1 p 2 p 1q 2q , the extreme R(2 3 2 ) in Fig. 6a.
regions of a combination instability can be achieved by adjusting When Rm or 1/m for integer m, the parametric excitations
c 1 , c 2 , and c 1 c 2 2h. from the two meshes interact. Consider the case with R1/m,
Unfortunately, the primary, secondary, and combination insta- where the sm instabilities caused by 1 overlap with the pri-
bility regions cannot be minimized at the same time. The condi- mary instabilities caused by 2 . Considering instability of the
tions reducing the primary instability regions dashed lines, Fig. p-th mode where 1 2 p /m 1 and 2 m 1 2 p
5 result in large combination instability regions, and vice versa m 1 , the terms leading to unbounded response in Eq. 12
dash-dot lines. Depending on specific applications, a trade-off vanish for
may be made to reduce multiple instability regions, though none
are true minima solid lines. Adjusting contact ratios and mesh p D pmp F p1p i E pmp G p1p e im 1 0
2i p A p / A
phasing is clearly an effective means to minimize instability re- (29)
gions and avoid resonances under operating conditions.
Bounded solution of Eq. 29 is ensured for
3.2 Four-Gear Systems: Equal Mesh Stiffness Variations. 21 D pmp F p1p 2 E pmp G p1p 2 / m p 2 (30)
Two-stage countershaft systems Fig. 1a have two different For example, when 2 2 1 (R1/2), the boundaries for 1
mesh frequencies 1 R 2 , which means more instability re- secondary instabilities overlapping with 2 primary instabilities
gions than three-gear systems. We consider the case where the are
gear facewidth and material are such that the mesh stiffness am-
plitudes are identical at the two meshes ( 1 2 ), although
1 p D p2p F p1p 2 E p2p G p1p 2 1/2 (31)
the contact ratios and phasing are not restricted. Depending on the 2p
ratio RZ 2 /Z 4 , the parametric instability regions associated with
1 and 2 may overlap each other. For Rm/ j m, j are inte- Figure 6b shows instability regions in the ( 1 ,) plane for R
gers, the s j instabilities single mode and combination of 1 1/2. Note the instability at 1 3 couples with the instability
and the sm instabilities of 2 occur simultaneously. Because at 2 2 3 , and the combined instability region is much larger
their instability regions are typically the largest, the interactions than the case without interaction Fig. 6a. Using Eqs. 4 and
involving either m or j1 are of most interest. 19 in 31, the slopes of these boundaries are
When Rm, 1/m for integer m, the s1 instabilities from one 1 k 2g1 41 p sin2 2c 1 k 2g2 42 p sin2 c 2
mesh decouple from the sm instabilities of the other mesh. In
this case, instability occurs when s 1 or s 2 is close to p 2k g1 k g2 21 p 22 p sin 2c 1 sin c 2
q , but these instability boundaries can be calculated indepen-
dently. For s 1 p q 1 , the terms leading to unbounded 2
response in Eq. 12 are eliminated for cos 2c 1 c 2 2h 1/2 (32)
p
q D pqs iE pqs e i 1 0
2i p A p / A (24) Minimization of 1 requires c 1 1,1.5,2 and c 2 1,2 for 1
c 1,22. The instability region can also be reduced by adjusting
p D qsp iE qsp e i 1 0 the phasing h according to cos(2c1c22h)1 with sign the
2i q A q / A (25)
same as sin(2c1 ). The primary instability regions under 1 (s
Expressing the solutions as A p a p e i , A q a q e i ( ) and 1) do not coincide with any other instability regions, so these
examining , the condition separating bounded and unbounded instability boundaries are calculated from Eq. 26 with pq.
solutions of Eq. 24 and 25 is Other overlap situations are possible, such as the 2 secondary
instability (s2) overlaps with the 1 fourth instability (s4),
1
1
s
p q D pqs 2 E pqs 2
p q
(26)
but the interaction between these higher instabilities is typically
weak and the instability regions are much smaller. Combination
instabilities can be analyzed similarly.
Similarly, the boundaries for instabilities associated with s 2
p q 2 are 3.3 Three and Four-Gear Systems: Unequal Mesh Stiff-
ness Variations. This general case allows all parameters of the
1 F pqs 2 G pqs 2 two mesh stiffnesses to differ. In contrast with prior cases, the
2 p q (27)
s p q gears may have differing facewidths and material properties such
that the amplitudes of stiffness variation at each mesh vary inde-
Use of Eqs. 4 and 19 in 26 and 27 yields pendently ( 1 2 ). The contact ratios and mesh phasing are un-
Fig. 6 Instabilities regions when 1 R 2 , 1 2 . a R tained by presuming a linear variation of the boundaries in the
35, b R 12. The parameters are in Table 1 and c 1 c 2 ( 1 , 1 ) plane for given 2 . To construct this linear approxima-
1.5, h 0.*** denotes numerical solutions. tion, one point is calculated under the condition 1 0, 2 C and
a second point is obtained at 1 2 C. From Eq. 27, the pri-
mary stability boundary limits for 1 0, 2 C are
restricted. The design of one mesh must account for dynamic in- C
1 2 p F p1p 2 G p1p 2 1/2 (33)
teractions with the mesh stiffness variation of the other. p
When Rm, 1/m for integer m, there is no interaction between
the parametric excitations from the two meshes. The 1 instabili- From Eq. 18, the stability boundary limits for 1 C, 2 C are
ties are only affected by 1 and their boundaries in the ( 1 , 1 ) C
plane are s 1 p q 1 1 with 1 determined by Eq. 26. 1 2 p D p1p F p1p 2 E p1p G p1p 2 1/2 (34)
p
The 2 instabilities are only affected by 2 and their boundaries
in the ( 2 , 2 ) plane are s 2 p q 2 2 with 2 deter- An example is for the primary instability when R1, C0.3.
mined by Eq. 27. Connecting the two points obtained from Eqs. 33 and 34 yields
When Rm or 1/m for integer m, a mode may be simulta- the instability boundaries, which agree well with the numerical
neously driven to instability by both mesh excitations. In this case, solution Fig. 7a. Assembling the ( 1 , 1 ) planes for various
the first mesh instability regions can be significantly affected by 2 C generates three-dimensional plots of 1 versus 1 , 2
the presence of 2 and vice versa. Closed-form boundaries of the Fig. 7b. The parametric excitation in the second mesh dramati-
form for primary instability 1 2 p 1 1 2 2 for inde- cally changes the shapes of the instability regions. Notice that the
pendently varying 1 , 2 are cumbersome. Alternatively, simple second parametric excitation widens the primary instability region
yet accurate approximations for the instability boundaries are ob- for small 1 compared to monofrequency excitation Fig. 3. In
4 An Example
The two-stage gear system Fig. 1a studied by Tordion and
Gauvin 14 and Benton and Seireg 15 is used as an example.
These two papers come to markedly different conclusions as dis-
cussed below. The system parameters are given in Table 1 and
c 1 1.47, c 2 1.57. In keeping with the published work, the
double-tooth contact mesh stiffness k max 1 is kept constant, so
the average mesh stiffnesses k gi decreases as k ai increases Fig.
2b.
Tordion and Gauvin assumed that k v 1 and k v 2 have the same
amplitude and frequency but different contact ratios and phasing.
They applied an infinite determinant method 20 to plot the
boundaries of primary and secondary instabilities dashed lines in
Fig. 8. Their results deviate significantly from the numerical so-
lution as a result of analytical errors. In addition, the Fourier ex-
pansion they derived for rectangular waveforms Eqs. 11 and
12 in 14 is incorrect. Nevertheless, they conclude that The
phase displacement between the meshing stiffnesses has a great
influence on the width of the instability regions.
Benton and Seireg 15 considered the same system. They de-
coupled the equations using the modal transformation and ne-
glected the off-diagonal terms of the transformed time-varying
stiffness matrix that is, T Kv (t). These treatments reduce Eq.
6 to three uncoupled Mathieu equations. The average value of
two contact ratios was used to make the stiffness variations k v 1
and k v 2 identical. With these approximations, they conclude that
the instability regions are independent of the mesh phasing, that
is, the normal mode technique . . . without considering the phase
variations . . . provides a relatively simple means of predicting
the instability regions with sufficient accuracy for practical pur-
poses. This conflicts directly with Tordion and Gauvin. In fact,
the mode uncoupling method does not provide satisfactory results
when the mesh phasing is non-zero Fig. 8b.
The perturbation results resolve the discrepancy: Mesh phasing
strongly impacts the mesh stiffness variation instabilities. In fact,
mesh phasing, along with contact ratios, plays a key role in mini-
mizing instability regions as discussed earlier. The excellent
agreement of analytical and numerical stability boundaries con-
firms this finding Figs. 3, 6, 7, 8.
To further validate the stability conditions, free responses under
nontrivial initial conditions are calculated numerically Fig. 9 for
the parameters at point A of Fig. 8 4.2, k a 0.3. For
point A in Fig. 8a, the responses are unstable Fig. 9a, as Fig. 8 Comparison of instability regions from different analy-
ses. The parameters are from Table 1, c 1 1.47, c 2 1.57, and
identified by perturbation and numerical methods. This point, phasing a h 0, b h 0.4. Perturbation method;*** Nu-
however, is stable according to Tordion and Gauvin Fig. 8a. merical method; ---Tordion and Gauvin 1977; """ Benton and
When the phasing h0.4 at point A Fig. 8b, stable responses Seireg 1980.
occur Fig. 9b. This is consistent with the perturbation and nu-
merical solutions but conflicts with both Tordion and Gauvins
and Benton and Seiregs results. in two-stage gear chains. However, the effects of contact ratio and
mesh phasing on instability conditions derived in the study are
5 Discussions generally applicable to suppress instability and hence contact loss.
Rectangular waveforms are close approximations of the mesh The instability analysis can be reduced to single mesh gears
stiffness in spur gears with involute teeth. For helical gears or spur with one natural frequency n . Setting k g2 0 in Eqs. 21 and
gears with tooth modification, mesh stiffness deviates from the 22, it follows that primary and secondary instabilities vanish as
rectangular shape. Equation 4 is not valid for other functions, the contact ratio c 1 1,2 and c 1 1,1.5,2 , respectively. Maxi-
but the general Fourier expansion 6 can still be used in matrices mum primary and secondary instability occurs at c 1 1.5 and c 1
D,E,F,G to determine the instability boundaries. 1.25,1.75 , respectively. Kahraman and Blankenship 16 ex-
If damping is considered, the system stability improves and the perimentally studied a pair of spur gears under mesh stiffness
instability regions shift to the right in the , plane 10,13. excitation for various contact ratios. They showed that the ampli-
Furthermore, damping and nonlinearity e.g., tooth separation tude A 1 in the first mesh frequency harmonic of the response is
must be considered to determine the limit cycle amplitude of the minimized when the contact ratio c 1 1,2 . This is because para-
dynamic response when operating conditions cause instability. metric excitations are eliminated for integer contact ratios. When
Additional study is needed to quantitatively examine these effects the mesh frequency n , their measured A 1 reaches maximum
Acknowledgment
The authors thank Timothy L. Krantz of the Army Research
Lab at NASA Lewis Research Center for his support and advice
on the project. This material is based upon work supported by the
NASA Glenn Research Center under grant NAG-1979 and the
U.S. Army Research Office under grant DAAD19-99-1-0218.