Вы находитесь на странице: 1из 9

Song He

Acoustics and Dynamics Laboratory, Department of Mechanical Engineering, The Ohio State University, Columbus, OH 43210 e-mail: he.81@osu.edu

Todd Rook

Goodrich Aerospace, 101 Waco Street, Troy, OH 45373 e-mail: todd.rook@goodrich.com

Rajendra Singh

Fellow ASME, Acoustics and Dynamics Laboratory, Department of Mechanical Engineering, The Ohio State University, Columbus, OH 43210 e-mail: singh.3@osu.edu

Construction of Semianalytical Solutions to Spur Gear Dynamics Given Periodic Mesh Stiffness and Sliding Friction Functions

Gear dynamic models with time-varying mesh stiffness, viscous mesh damping, and slid- ing friction forces and moments lead to complex periodic differential equations. For example, the multiplicative effect generates higher mesh harmonics. In prior studies, time-domain integration and fast Fourier transform analysis have been utilized, but these methods are computationally sensitive. Therefore, semianalytical single- and multiterm harmonic balance methods are developed for an efficient construction of the frequency responses. First, an analytical single-degree-of-freedom, linear time-varying system model is developed for a spur gear pair in terms of the dynamic transmission error. Harmonic solutions are then derived and validated by comparing with numerical inte- gration results. Next, harmonic solutions are extended to a six-degree-of-freedom system model for the prediction of (normal) mesh loads, friction forces, and pinion/gear dis- placements (in both line-of-action and off-line-of-action directions). Semianalytical pre- dictions compare well with numerical simulations under nonresonant conditions and provide insights into the interaction between sliding friction and mesh stiffness. DOI: 10.1115/1.2988478

1 Introduction

Sliding friction acts as an excitation to spur and helical gear dynamics, as described by Houser et al. 1 , Velex and Cahouet 2 , Velex and Sainsot 3 , and Lundvall et al. 4 . Earlier, Vaishya and Singh 5–7 developed a single-degree-of-freedom SDOF spur gear model with rectangular mesh stiffness k t and sliding friction profiles; they solved the dynamic transmission error DTE t response by using the Floquet theory 5 and multi- term harmonic balance method MHBM 6 . Their work was re- cently refined and extended to helical gears in our papers 8,9 where closed-form solutions of t for a SDOF system are de- rived under friction excitation. The equal load sharing assumption 5–9 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 10,11 where we proposed a multidegree-of-freedom MDOF model with real- istic time-varying k t and sliding friction functions. However, we utilized numerical integration and fast Fourier transform FFT analysis methods in prior studies 8–11 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 12 implemented a similar harmonic analysis to ap- proximate the dynamic factors in helical gears based on tooth loads and quasistatic transmission errors . 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 5,10 . In particular, we intend to develop semianalytical harmonic balance solutions to the 6DOF spur gear model 10 with realistic mesh stiffness k t 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.

2 Problem Formulation

Gear dynamic models are usually described by periodic differ- ential equations 13–15 due to significant variations in mesh stiff- ness k t and mesh damping c t within the fundamental period t c one mesh cycle . Additionally, dynamic friction force F f t and torque M f t also undergo periodic variations, with the same pe- riod t c , due to changes in normal mesh loads and coefficient of friction , as well as a reversal in the direction of F f t at the pitch point 5–10 , especially in spur and helical gears. For the sake of illustration, a unity ratio NASA spur gear with tip relief and contact ratio around 1.6 is used as an sample case for this study; refer to Table 2 of Ref. 10 for its parameters. Major pe- riodic excitations, namely, the mesh stiffness k t of the two mesh- ing tooth pairs and the frictional functions f t , are shown for the sample case in Figs. 1 a and 1 b , 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 a for geometric locations of the points along the line-of-action LOA direction, while the off-line-of- action OLOA 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 k i t of each meshing tooth pair can be computed by k i t = N s , i t / r b s , i t , i =0, 1 10 , where r b is the base radius and N s , i t and s , i t are the static normal contact force and angular deflection computed by a static finite element/contact mechanics FE/CM analysis 16 . Modifi- cations to the tooth profiles, such as the linear tip relief, are char- acterized by changes in the effective N s , i t and s , i t profiles. Hence the resulting k i t includes the effect of the tip relief as well. Refer to Ref. 10 for the calculation procedure of k i t . Derivations of f t will be explained later in Eqs. 4 a 4 e along with key definitions of the system model. The fundamental nature of the linear time-varying LTV sys- tem is illustrated in Fig. 2 b , and the system model is described in our paper 10 . Key assumptions include the following: i 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

Fig. 1 aPeriodic mesh stiffness functions of the spur gear

¯

pair example with tip reliefgiven nominal pinion torque T =550 lb in. Key: blue dashed line, k 0 t ; red solid line, k 1 t . b

Periodic frictional functions. Key: blue dashed line, f 0 t ; green dashed-dotted line, f 1 t ; red solid line, f 2 t .

p

pinion and gear are rigid disks. ii Vibratory motions are small in comparison to the nominal motion; this would lead to a LTV model. Only nonresonant conditions are considered since signifi- cant dynamic motions at resonance may not only change the con- tact pattern such as the effective center distance but also cause teeth separation, which would introduce nonlinear interactions and further complicate the formulation. iii The Coulomb friction is assumed with a constant although the sign is reversed at the pitch point. The governing SDOF equation in terms of DTE t = r bp p t r bg g t is given below, where subscripts p and g correspond to the pinion and gear, respectively; is the vibratory component of

¯

¯

the rotation; T and are the nominal torque and rotation speed; is the base pitch; and J is the moment of inertia. Note that J e is defined differently from that in Ref. 10 for a compact formula- tion of Eq. 1 a ,

S

S

t + J b c i t t + k i t t + sgn mod p r bp t ,

¨

J e

˙

¯

i

=0

i

=0

˙

+ S j L AP c i t t + k i t t X pi t J g r bp

+ X gi t J p r bg = T e + T t

1 a

J e = J p J g

J b = J g r bp + J p r bg

2

2

1 b

1 c

¯

¯

T e = T p J g r bp + T

g J p r bg

1 d

122601-2 / Vol. 130, DECEMBER 2008

J p r b g 1 d 122601-2 / Vol. 130, DECEMBER 2008 Fig. 2 „

Fig. 2 aSnap shot of the contact pattern at t =0 for the sample spur gear pair. bNormal mesh and friction forces of the analytical spur gear system model.

 

¯

X pi t = L XA + S i + mod p r bp t , ,

i = 0,

, S = floor

 

1 e

 

¯

X gi t = L YC + i − mod g r bg t , ,

 

i = 0,

, S = floor

 
 

1 f

 

S

S

T t = J b c i t t + k i t t + sgn mod p r bp t ,

˙

¯

 

i

=0

i

=0

+ S j L AP c i t t + k i t t X pi t J g r bp

˙

 

+ X gi t J p r bg

 

1 g

Here, the j th 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

5 . The modulus function, defined as mod t , t c = t t c ·floor t / t c , is used to formulate the periodic friction force F f t and the mo- ment arm X t at the fundamental period of t c . The “floor” func- tion rounds off the contact ratio to the nearest integers toward a lower value. Also, T t of Eq. 1 g represents the time-varying component of the forcing function due to the unloaded static

manufacturing transmission error t . Finally, geometric lengths L XA , L YC , and L AP of Fig. 2 a could be found as follows, where

u is the gear ratio, l is the center distance, wt is the transverse

operating pressure angle, and r a is the outside radius,

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

L XA = l sin wt r ag

2

2

r bg

2 a

L YC = r ag

2

2

r bg

+

 

2

2

ul sin wt

L AP = r ag

r bg

u + 1

2

2

b

c

Observe that Eqs. 1 a 1 g significantly differs from the clas- sical Hill’s equation 13 in several ways. First, the periodic k i t is not confined to a rectangular wave assumed by Vaishya and Singh 5–7 or a simple sinusoid as in Mathieu’s equation 1 . Instead, Eqs. 1 a 1 g should describe realistic, yet continuous, profiles of Fig. 1 a resulting from a detailed FE/CM analysis 16 . Hence, multiple harmonics of k i t should be considered. Second, the periodic viscous c i t term should dissipate vibratory energy due to the sliding friction besides its kinematic effect.

t k i t terms of Eqs. 1 a 1g

Third, the

incorporate combined but phase correlated contributions from all yet changing tooth pairs in contact. Consequently, the relative phase between neighboring tooth pairs should play an important role in the resulting response t . Fourth, multiplicative effects between k i t , c i t , X i t , and t should result in higher mesh harmonics, which poses difficulty in constructing closed-form so- lutions. Lastly, the coupling between T t and t in Eq. 1 g indicates that the frictional forces and moments reside on both sides of Eq. 1 a as either periodically varying parameters or ex- ternal excitations, thus posing further mathematical complications.

S

˙

i

=1

i

t c i t and i =1

i

S

3 Semianalytical Solutions to the SDOF Spur Gear Dynamic Formulation

Consider the sample case with only the mean load T e , i.e., T t =0 or t =0. Equations 1 a 1 g can be rewritten over one complete mesh cycle 0 t t c , as follows for the sample case with two mesh tooth pairs, where m e and F e are the effective mass and force; E 1 , E 2 , and E 3 are gear constants,

¨

˙

˙

m e

t + c 0 t t + k 0 t t 1 + E 1 + E 3 t + c 1 t t

+ k 1 t t 1 + E 2 + E 3 t sgn t

bp = F e

L AP

¯

p

r

E 1 = L XA + J g r bp + L YC J p r bg / J b

E 2 = L XA J g r bp + L YC + J p r bg / J b

3

3

a

b

3 c

¯

E 3 = p r bp J g r bp J p r bg / J b

m e = J e / J b

3 d

3 e

F e = T e / J b

Next, reformulate Eqs. 3 a 3f in terms of the dimensionless time = t / t c , such that = d / d = t c t and = d 2 / d 2

3 f

= t c t ,

2

¨

m e + t c c 0 + t c k 0 1 + E 1 + t c E 3 f 0

2

+ t c c 1 + t c k 1 1 + E 2 f 1 + t c E 3 f 2 = t c

F e

 

4

a

f 0 = mod ,1

4

b

f 1 = sgn mod ,1 P = sgn f 0 P

4

c

f 2 = mod ,1 sgn mod ,1 P = f 0 f 1

4

d

P = L AP /

4

e

Journal of Mechanical Design

Each periodic function, k i , c i , f i , or , now has a pe- riod of T =1. Figure 1 b shows typical f 0 , f 1 , and f 2

functions, which describe the periodic moment arm and sliding friction excitations for the sample case 10 .

3.1 Direct Application of Multiterm Harmonic Balance. Define the Fourier series expansions of the periodic k i and c i in Eqs. 4 a 4 e up to N mesh harmonics as follows, where the angular frequency n =2 n in rad/s and n is the mesh order:

N

N

k i = A k i 0 + A k in cos n + B k in sin n

n

=1

1

A k i 0 = 0

n

=1

k i d

1

A k in = 2 0

k in cos n d

1

B k in = 2 0

k in sin n d

5 a

5

5

5

b

c

d

N

N

c i = 2 k i I e = A c i 0 + A c in cos n + B c in sin n

n

=1

n

=1

6 a

1

A d i 0 = 0

c i d

6 b

1

A c in = 2 0

1

B c in = 2 0

c in cos n d

c in sin n d

6

6

c

d

The f i functions could be expanded explicitly as shown be- low,

f 2 =

1

2

f 0 =

N

1

2

n

=1

1

sin n

n

f

N

1 =1−2 P

n

=1

4

sin n P

n

cos n

N

+

n

=1

4 cos n P − 1

n

sin n

2 +

P

4

2

n

N

n

=1

1 − n P sin n P

N

7 a

7 b

cos n P cos n + n P cos n P − sin n P

0.5 n sin n

n =1

7 c

Finally, assume that the periodic dynamic response is of

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

= A 0 + A n cos n + B n sin n

n

=1

n

=1

8

Substitute Fourier series expansions of Eqs. 5 a 5d , 6a 6 d , 7 a 7 c , and 8 into Eqs. 4 a 4 e and balance the mean and harmonic coefficients of sin n and cos n . This converts the linear periodic differential equation into easily solvable linear algebraic equations as expressed below where K= h is a square matrix of dimension 2 N +1 consisting of known coefficients of k i , c i , and f i . By calculating the inverse of K= h , the 2 N +1 Fourier coefficients of could be computed at any gear mesh harmonic n ,

K= h

A
A
B B N = F e

A N

0

1

1

0

0

0

0

9

3.2 Semianalytical Solutions Based on One-Term HBM.

Next, we construct one-term HBM 17 solutions to conceptually illustrate the method. Set the harmonic order N =1 only the fun- damental mesh, in addition to the mean term in Eqs. 5 a 5 d , 6 a 6 d , 7 a 7 c , 8 , and 9 and balance the harmonic terms in Eqs. 4 a 4 e . This leads to a K= h matrix of dimension 3. Three of its typical coefficients are given as follows and the rest could be found in a similar manner:

K h 11 =

1

2

+

+

t c A k 11 A f 21 E 3 + B k 11 B f 11 E 2 t c B k 01 E 3 / + 2 A k 00 + 2 A k 10

A k 11 A f 11 E 2 +

2 A k 00 E 1 + t c A k 00 E 3 + 2 A k 10 A f 10 E 2

2 t c A k 10 A f 20 E 3 + t c B k 11 B f 21 E 3

10 a

K h 21 = A k 01 + A k 11 + A 01 E 1 + A k 10 A f 11 E 2 + t c A k 10 A f 21 E 3

+ A k 11 A f 10 E 2 + t c A k 11 A f 20 E 3 + t c A k 01 E 3 / 2

10 b

K h 31 = B k 11 A f 10 E 2 + B k 11 + t c B k 11 A f 20 E 3 + A k 10 B f 11 E 2

+ t c A k 10 B f 21 E 3 + B k 01 E 1 + B k 01

c A k 00 E 3 + t c B k 01 E 3

t

2

10 c

The Fourier series coefficients of are then obtained by

inverting K= h . Figure 3 shows that the one-term HBM solution predicts the overall tendency mean and first harmonic fairly well

¯

when compared with numerical simulations at T p =550 lb in. and

¯

=500 rpm. This confirms that the one-term HBM and likewise the MHBM approach coverts the periodic differential Eqs. 4 a 4 e with multiple interacting coefficients into simpler alge- braic calculations that are computationally more efficient than nu- merical integrations and subsequent FFT analyses. Thus, the semi- analytical solution provides an effective design tool. Also, most coefficients of Eqs. 10a 10c show sideband effects that are

introduced by k t or c t and f i t functions.

3.3 Iterative MHBM Algorithm. When N 5, we can utilize

a symbolic software 18 to balance multiple harmonic terms and calculate K= h . However, the computational cost involved with each element of K= h increases by N 3 due to the triple multiplication of periodic coefficients in Eqs. 4 a 4 e . Consequently, for higher N say, 5 , a direct computation of K= h becomes inefficient and thus inadvisable. Instead, we apply a matrix-based iterative

p

122601-4 / Vol. 130, DECEMBER 2008

matrix-based iterative p 122601-4 / Vol. 130, DECEMBER 2008 Fig. 3 Semianalytical versus numerical solutions for

Fig. 3 Semianalytical versus numerical solutions for the

¯ ¯

SDOF model, expressed by Eq. 3, given T p =550 lb in., =500 rpm, and =0.04. aTime-domain responses; bmesh harmonics in frequency domain. Key: blue solid line and , numerical simulations; black dashed line and , semianalytical solutions using one-term HBM; red dashed-dotted line and , semianalytical solutions using five-term HBM.

p

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

˜

variables and , where is the subharmonic index,

˜

= 2 / t c

˜

=

t 0,2

¯

= mod / 2 ,1

11

11

11

a

b

c

Also, define the differential operator “ˆ” as

ˆ

= d

d

˜

=

˙

−1

12

a

b

Equations 3 a 3 f are then converted into the following form,

˙

= ˜ ˆ

12

˜

where E 4 =

1 E 3 :

˜

2

ˆ

ˆ

˜

¯ ˆ

¯

¯

m e

+ c 0 + k 0 1 + E 1 + E 4

˜

¯ ˆ

¯

¯

¯

+ c 1

− 1 = F e

+ k 1 1 + E 2 + E 4 sgn

/

or expressed more compactly as

L AP

13

˜

¯

 

ˆ

ˆ

˜

ˆ

2

m e

+ C + K = F e

 

14 a

 

¯

¯

¯

¯

C = c 0 1 + E 1 + E 4 + c 1 1 + E 2 + E 4 sgn

/

− 1

L AP

14 b

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

¯ ¯ ¯ ¯ ¯ K = k 0 1 + E 1 + E
¯
¯
¯
¯
¯
K = k 0 1 + E 1 + E 4 + k 1 1 + E 2 + E 4 sgn
/
L AP
− 1
14 c
For the MHBM, the discrete Fourier transform DFT 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,
sin 1
cos 1
sin N 1
cos N 1
1
sin 2
cos 2
sin N 2
cos N 2
F= = 1
]]
]
]
]
1
sin M
cos M
sin N M
cos N M
15
a
1
2
= F=
15
b
]
M
ˆ
1
Fig. 4 Semianalytical versus numerical solutions for the
ˆ
2
SDOF model as a function of pinion speed with =0.04. „ a…
ˆ
= F= D=
]
ˆ
15
c
Mesh order n =1, „ b… n =2, „ c… n =3, and „ d… n =4. Key: red ,
numerical simulations; blue solid line, semianalytical solutions
using five-term HBM.
M

ˆ

ˆ

= F= D= 2

ˆ

ˆ

1

ˆ

ˆ

2

]

ˆ

ˆ

M

15 d

Here, i = i 2 / M and M 2 N +1, is a vector of 2 N +1 Fourier coefficients, and the Fourier differentiation matrix D= is given as

D= =

0

0

0

0

0

1

]]

0

0

0

0

D= 2 =

0

0

0

0

−1

0

]]

0

0

0

0

0

−1

0

]

0

0

0

0

−1

]

0

0

0

0

0

0

0

N

0

0

0

0

N

0

0

0

0

0

N 2

0

N 2

0

0

0

0

0

15 e

15 f

Applying the DFT to the equation of motion yields the following MHBM equations where F= + is the Moore–Penrose or pseudoin- verse of the DFT matrix,

2 m e D= 2 + F= + C= F= D= + F= + K= F= = F= + F e

˜

˜

16

a

K= diag K 1

 

K 2

¯ K M

16

b

C= diag C 1

C 2

¯

C M

16

c

F e = F e

0

¯

0

0 T

16 d

Figure 3 shows that the five-term HBM solutions compare well with numerical simulations. Likewise, an increase in N captures higher frequency components around the tenth mesh harmonic, as

Journal of Mechanical Design

observed in the numerical simulations. The semianalytical solu-

tions are efficiently used in Fig. 4 for parametric studies of at

¯

the gear mesh harmonics over a range of p , and these calcula-

tions are indeed achieved with reduced computational cost.

4 Semianalytical Solutions to 6DOF Spur Gear Dy- namic Formulation

A careful examination of the 6DOF model reported in Ref. 10 reveals that the dynamic bearing displacement y in the OLOA direction depends on the LOA dynamics since the friction force k acts in the OLOA direction. Conversely, the dynamic bearing displacement x in the LOA direction is not influenced by i.e., is decoupled from the OLOA motion y . This is be- cause of the following reasons: First, no off-diagonal term exists in the effective shaft-bearing stiffness matrix that couples the LOA and OLOA dynamics 8 . Second, the friction force de- scribed by the Coulomb model is independent of y . Third, re- call our assumption that the dynamic responses such as y have no influence on the kinematics such as the center distance . Such “one-way coupling” between LOA and OLOA dynamics im- plies that we can reduce the 6DOF model into a simpler 3DOF model 21 in terms of x p , x g , and . Dynamic responses of the 3DOF model could further be used to determine the friction forces that excite the OLOA dynamics. For the 3DOF system, when the composite DTE = r bp p r bg g + x p x g can be approximated by the

semianalytical solution = r bp p r bg g of the SDOF

model, a similar one-way coupling is created where is decou- pled from x p and x g , while x p and x g are excited by in terms of the normal loads. This suggests that when the excitation mesh frequencies do not excite any coupled transverse-torsional modes in the LOA direction, the harmonic solutions of the SDOF system could be extended to predict trans- lational responses in both LOA and OLOA directions by utilizing the one-way coupling effects. Note that the semianalytical method analyzes the 6DOF system 10 as a 5DOF model as it calculates the t rather than absolute angular displacements p t or p t .

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 a and focus on its subset of a unity gear pair 3DOF model to study the natural frequency distribution. We define the following parameters for the

sample case: mass of pinion gear m p = m g = M ; moment of inertia

J p = J g = J ;

= J / 2 R 2 ; time-averaged mesh stiffness k m T p =550 lb in. ; shaft-

= K B . The natural frequencies of the

= r = R ; the equivalent mass m e

basic

radius

r bp

bg

bearing stiffness K Bp = K Bg

3DOF system are found as follows 21 :

= K B g 3DOF system are found as follows 21 : 2 N 1, N
= K B g 3DOF system are found as follows 21 : 2 N 1, N

2

N 1, N 3

= k m M + 2 k m + K B m e k m M + 2 k m + K B m e 2 − 4 Mm e k m K