You are on page 1of 6

Mathematical modelling of gearbox nonlinear vibration

V. Zeman
M. Byrtus
University of West Bohemia in Pilsen University of West Bohemia in Pilsen
Pilsen, Czech Republic
Pilsen, Czech Republic
Abstract The contribution presents the original
method of the mathematical modelling of gear drive nonlinear vibrations using modal synthesis method with DOF
number reduction. The model respects nonlinearities
caused by gear mesh interruption and by contact forces acting between journals of the rolling-element bearings and
the outer housing. As a result of nonlinear couplings, gearbox vibrations in case of small static load and in resonances
are accompanied by impact motions, bifurcation of solution
and chaotic motions. The presented method is applied to
the real car gearbox.
Keywords: gearbox, nonlinear vibration, nonlinear couplings,
modal synthesis method

I. Introduction
The gear drive and gearboxes vibrations are commonly
investigated with the assumption of linearized coupling
forces ,  or for an isolated gear pair with nonlinear
meshing coupling . These simplifying assumptions are
acceptable for the system with gears only if dynamic meshing deformations are smaller than their static deformations
and in cases when the housing is relatively rigid and the
interaction between single pairs of gears is neglectable.

B2

s=1
G1

G2

G5

G3

s (t) + (Bs + s Gs )q s (t) + Ks qs (t) =

Ms q
s = 1, 2, . . . , S,
= fsE (t) + fsB + fsG ,

4
s=2
B4

B3

G6
1

2
s=3

kL

B5

IL

Fig. 1. Scheme of the real gearbox

E-mail:
E-mail:

zemanv@kme.zcu.cz
byrtus@kme.zcu.cz

Presented approach to the gearbox modelling uses the

modal synthesis method based on suitable system decomposition into subsystems and on separate modelling of couplings among subsystems.

Generally, the subsystems rotating with angular velocity s are described in local generalized coordinates qs (t)
with system of ns ordinary differential equations in matrix
form 

G4
3

II. Modelling of gearboxes

A. Modelling of subsystems

s=4

B1

In last years the subject of modelling of large rotating

systems has been studied in our department. The result of
this effort is a creation of modal synthesis method allowing to model rotating linearized systems with complicated
structures . Our first results gained from the modelling of
nonlinear vibrations applied to a simple test-gearbox were
presented in .
This contribution is focused on modelling of car gearbox
vibration (figure 1) which is included into the car drive by
means of discrete torsional couplings characterized by torsional stiffness k of the twin-disk clutch and torsional stiffnesses kL , kR of the half-axles. The driving I and driven
wheels IL , IR rotate by the constant angular speeds and the
the initial torque M acting on the driving shaft. The car
gearbox vibrations are excited by kinematic transmission
errors of gear pairs transmitting the power.

B6

kR

IR

(1)

where Ms , Bs and Ks are symmetrical mass, damping

and stiffness matrices of the uncoupled subsystems of order ns and Gs is skew symmetrical matrix of the gyroscopic effects of the same order for s = 1, 2, . . . , S 1
only. These matrices are usually created by means of finite
element method combined with discrete parameters representing masses of rigid gear discs. External forced excitation is described by vector fsE (t). Vector fsB expresses
the coupling forces in rolling-element bearings and vector
fsG represents the forces in spur helical gear couplings. All
force effects described in vectors above are acting on the
subsystem s.

y
\$ i, j

#
ax
(
ei,j Fi,j + eax
fSB =
i,j Fi,j ),

e i, j

\$i

H i, j

ax

F i, j

ax

F i, j

&i

ri
vi

wi
"i

Fi

Si

e i, ax

!i

ui
!i

Si

"i

where vectors ei,j , eax

i,j describe the geometry of contact
points at the stator. Bearing indices i are governed by the
same conditions as above.
In the general coordinate space
q = [ qT1 , qT2 , . . . , qTS ]T Rn ,

Fig. 2. Scheme of a bearing coupling

B. Bearing model
The bearing model, used in our approach to gearbox
modelling, respects real number of rolling elements uniform distributed between the inner and outer race (see figure 2). Let us suppose that the rolling-element j of the bearing i touches the outer race at the contact point Hi,j . The
force Fi,j trasmitted at this point depends nonlinearly on
the rolling-element deflection i,j according to the Hertzs
contact theory in this way
Fi,j
ax
Fi,j

"n
i,j
=
H(i,j ),
C
! ax "n
i,j
=
H(ax
i,j ).
C

(2)

The designation ax belongs to axial deflections of rollingelements and corresponding axial forces that moreover
arise in the radi-axial rolling bearings. The deflections originate from the deformation of the rolling elements and the
races at the contact points. Parameters n and C depend on
rolling-elements geometry, elastic modulus and Poissions
ratio of the bearing components, as mentioned in . Heaviside function H corrects the contact forces when the deflections are negative which means that the rolling elements
The calculation of the deflections i,j and ax
i,j supposing the rigid inner and flexible outer race, is in detail described in .
Then the vector fsB in equation (1) (model of rotating
parts) can be expressed in following form
fsB =

#
i,j

ax
(ti,j Fi,j + tax
i,j Fi,j ), s = 1, . . . , S 1, (3)

where vectors ti,j ,

describe the bearing geometry.
Among bearing indices i belong only these which correspond to bearings coupled with the subsystem s. Similarly,
the vector fSB in equation (1) (model of a stator housing)
tax
i,j

(4)

i,j

&i

vi
z

H i, j

n=

S
#

ns

(5)

s=1

of the whole system the bearing model is described by the

global coupling vector in the form
B
f1
f2B #

ax
(ci,j Fi,j + cax
(6)
fB = . =
i,j Fi,j ).
..
fSB

i,j

Vectors ci,j and cax

i,j describe global geometrical properties
of each bearing contact point j in bearing i and have this
structure
*
+T
ci,j = tTi,j eTi,j ,
(7)
*
+T
T
ax T
cax
eax
.
i,j = (ti,j ) (
i,j )

To stabilize the numerical simulation of the nonlinear

model it is efficient to separate the linear part of nonlinear bearing force characteristic. The linear part is described
by stiffness and damping matrices in the general coordinate space (5) and the bearing force vector (6) can be then
rewritten in following form
#
ax
f B = KB q BB q +
(ci,j fi,j + cax
(8)
i,j fi,j ),
i,j

where KB and BB are global stiffness and damping bearing matrices. Their structure depends on the number of
rolling elements and on the nodal points to which they are
fixed on the shafts (for details see ). The bearing damping matrix is supposed to be proportional to the stiffness
matrix
BB = B KB
(9)
and functions fi,j have form
,!
"n
i,j
ki,j i,j H(i,j )
fi,j (i,j ) =
C
ki,j i,j H(i,j ).

(10)

ax
In the same way can be expressed the function fi,j
(ax
i,j ).
The parameters ki,j represent linearized rolling-elements
stiffness that is calculated in dependence on an external
shown in .

C. Gearing model
The force effect of the spur helical gear coupling Gz (see
figure 1) in (1) is expressed by vector
#
fsG =
(11)
z,i Fz (t, dz , dz ),
z

where sign - (minus) corresponds to driving gear and sign

+ (plus) corresponds to driven gear (see figure 3). VecT
tor z,i = [. . . , z,i
, . . . ]T is the ns -dimensional extended
vector given by geometrical parameters (normal pressure
angle), (angle of inclination of the teeth), (angle of position vector), ri (rolling radius of the gear). The driving
(driven) gear is fixed on the shaft at the nodal point i (j).
Details are shown in . The resultant force Fz transmitted
by gearing z can be approximately expressed in the form
Fz (t, dz , dz ) = kz (t)dz + bz dz + fz (t, dz ),

(12)

where kz (t) is time dependent meshing stiffness and bz is

coefficient of viscous damping of gearing on gear mesh
line. Nonlinear function
fz = kz (t)dz H(dz ) + kz (t)(dz + uz )H(dz uz )
(13)
of gearing deformation dz corrects the linear elastic part
of the force Fz in the phases of the gear mesh interruption. The parameter uz is tooth backlash and H is Heaviside function.

qi (t) = [. . . ui vi wi i i i . . . ]T and qj (t) having similar form describe displacements of nodal points i and j.
The function z (t), defining kinematic transmission error
of gearing z, can be expressed by Fourier series
z (t) =

K
#

S
(C
z,k cos kz t + z,k sin kz t),

where z are meshing frequencies.

Analogous to the bearing model, we can express the
global gear coupling vector in general coordinate space (5)
in following way
G
f1
f2G
Z
#

+ fG (t),
f G = ... =
cz Fz (t, q, q)
(16)

f G z=1
S1
0

where fG (t) is vector of internal kinematic excitation generated in gear meshing that can be expressed in form
fG (t) =

Z .
#
z=1

/
z (t) cz .
kz z (t) + bz

rj

uj
wj

f G = KG q BG q +

y
i

A
vi

Gearing deformation
dz (t) =

(19)

It is advantageous to assemble condensed mathematical model of the system with reduced degrees of freedom
(DOF) number, because, in particular, the housing subsystem could have too large DOF number and this can hinder
from consecutive performing of various dynamical analyses
and optimization. The modal transformations

ri

Fig. 3. Scheme of a gearing coupling

T
z,i
qi (t)

cz fz (t, q) + fG (t),

z=1

ui
wi

Z
#

where KG and BG are global linearized stiffness and

damping matrices of gear couplings whose structure is in
detail described in .

driving gear

(18)

For the same reason as above, it is efficient to separate

up the linear part of nonlinear gearing force characteristic.
Equation (16) can be then rewritten to this form

vj

(17)

The global vector of geometrical parameters of the gearing

z in general coordinate space (5) has following structure
+T
*
T
T
cz = z,i
z,j
0T
.

driven gear

(15)

k=1

+ z,j qj (t) + z (t),

qs (t) = m Vs xs (t),
(14)

of gears in mesh fixed on shafts at nodal points i

and j expresses the relative motion of theoretical contact point of teeth on the gear mesh line. Vectors

s = 1, 2, . . . , S,

(20)

are introduced for this purpose. Matrices m Vs Rns ,ms

are modal submatrices obtained from modal analysis of the
mutually uncoupled, undamped and non-rotating subsystems, whereas ms (ms ns ) is the number of the chosen

master modes of vibration. The new configuration space of

the dimension m is defined by vector
x = [ xT1 , xT2 , . . . , xTS ]T ,

m=

S
#

ms .

(21)

s=1

The model (1) can be then rewritten using terms (8) and
(19) in the global condensed form
/
.
(t) + B + 0 G + VT (BB + BG ) V x(t)+

x
/
.
+ + VT (KB + KG ) V x(t) =
.##
(22)
ax
= VT
(ci,j fi,j (q) + cax
i,j fi,j (q))+
+

Z
#
z=1

/
cz fz (t, q) + fG (t) + fE (t) ,

where fE (t) = [ (f1E (t))T , (f2E (t))T , . . . , (fSE (t))T ]T is

the global vector of external excitation,
0
1
B = diag m VsT Bs m Vs ,
!
"
s m T m
(23)
G = diag
V s G s Vs ,
0
V = diag ( m Vs )
are block diagonal matrices (S = 0 holds for the stator subsystem) and = diag ( m s ) is diagonal matrix assembled of spectral submatrices m s Rms ,ms of the subsystems.
III. Analysis of car gearbox vibration
The mathematical model of gear drives is strongly nonlinear due to the possibility of gear mesh interruption and in
consequence of nonlinear bearing couplings respecting loss
of contact in some contact points in dependance on position
of journal centre. To perform the dynamical analysis the
condensed mathematical model (22) has to be transformed
into the state space to use the time integration method. The
time integration is started from the initial state
x(0) = ( + VT (KB + KG )V)1 VT fE (0),

x(0)
= 0

(24)

to minimize the startup transient motions. In general, the

vector fE (0) can describe an arbitrary external excitation at
the start of numerical integration.
The presented methodology was applied to the real car
gearbox (figure 1). The gearbox can be disassembled into
four subsystems drive shaft with gears and twin-disk
clutch (s = 1), driven shaft with gears (s = 2), total differential with divided driving axle and wheels (s = 3) and
housing (s = 4) wired in several nodal points with the fixed
engine. Subsystems are joined by discrete couplings gear
meshings (z = 1, 2, . . . , 6) and rolling-element bearings

(B1 to B6 ) considering four rolling-elements only. The

initial number of DOF of the uncoupled subsystems after
discretization by FEM using MATLAB code for rotating
subsystems and software package ANSYS for housing was
n1 = 198 (drive shaft), n2 = 140 (driven shaft), n3 = 36
(differential) and n4 = 166816 (housing). Numerical experiments show that the reduced (condensed) model (22)
of the complete system of the order m = 180 (m1 = 60,
m2 = 50, m3 = 20, m4 = 50) is acceptable in the frequency range up to 5000 Hz.
The gearbox vibration was investigated for second transmission degree, when gearing z = 2 and z = 6 transmit
the power. The other gearing pairs rotate freely about the
deformed shaft axles. The vibration is caused by the internal kinematic excitation in gearing having form (17). Only
three first amplitudes of the Fourier series (15) were taken
into account and have following values
S2,1 = 2.05 106 m,

S6,1 = 3.14 106 m,

S2,1
S2,1
, S2,3 =
,
2
3
(25)
S6,1
S6,1
S
, 6,3 =
,
=
2
3
C
C
C
C
= C
2,2 = 2,3 = 6,1 = 6,2 = 6,3 = 0.

S2,2 =
S6,2
C
2,1

A. Constant gear mesh

The linearized condensed model (22) (for fi,j (q) 0,
ax
fi,j
(q) 0 and fz (t, q) 0) was used for determination of the constant gear mesh regions. Using the linearization we neglect the possibility of gear mesh interruption
of each rolling-element in all bearings. Supposing the internal kinematic excitation in gearing defined by (15), (17)
and (25), we can investigate for which operational parameters the gear mesh is constant and we can find a boundary
of interrupted gear mesh.
From knowledge of the time response of the system in
configuration space x(t) we can formulate the condition of
constant gear mesh z
min dz (t) = min{cTz Vx(t) + z (t)} > 0.
tT

tT

(26)

This condition is generally true for both linearized and full

nonlinear model of the gearbox.
Figures 4 and 5 show areas of interrupted gear mesh
(white area) and of constant gear mesh (grey area) of the
second and sixth gear mesh transmitting the power in dependence on two operational parameters on revolutions
per minute of the driving shaft and on external static torque
acting on driving shaft and on discs IL and IR , which
model driven car wheels (reduced mass of the car). Results were gained from the linearized model by computing
the amplitudes of steady state motion. The steady gearing deformations are then expressed by Fourier series with

00
6

15

10

0
500

revolutions per minute

Fig. 4. Map of constant gear mesh of second gearing G2 .

above mentioned harmonic components with the fundamental frequency equal to the meshing frequency. There
are many peaks corresponding to resonance states of the
gearbox on the boundary between constant and interrupted
gear mesh. The gear box shows greater displacements and
thence greater gearing deformations and greater tendency
towards gear mesh interruption in all resonance states.

20

6e-006

15

3e-006

10

500

revolutions per minute

Fig. 5. Map of constant gear mesh of sixth gearing G6 .

B. Nonlinear car gearbox vibration

Further, we are concerned with the qualitative analysis of
the car gearbox nonlinear vibrations. The aim is to investigate the influence of gear mesh interruption and contact
loss of the rolling-elements on the behaviour of the whole
system. It is efficient to compare the acceptance of the linearized model and the full nonlinear model with respect to
the boundary of constant gear mesh.
The nonlinear model includes such nonlinear phenomena like impacts in gearing and nonlinear contact forces
transmitted by rolling-elements of bearings. These phenomena, which are shown for systems with several DOF
number in , are sources of nonlinear effects in solution

of the model. The time response of such systems is accompanied by bifurcation of solution in dependence on chosen
operational parameters. Vibro-impact systems are characterized by period doubling scenario, when the period number of the time response increases unexpectedly twice for
a certain values of operational parameters. This scenario
could repeat till the motion becomes chaotic. Moreover,
the time response of nonlinear model is influenced by impacts in gear meshings with one freely rotating gear. These
motions are neglected in the linearized model used before.
All results introduced here are gained for the condensation level of the car-gearbox and for kinematic transmission
errors mentioned in the chapter before.
minima and maxima of gearing deformation d2 [m]

006
6e-

3e
-

20

!6

x 10

6
4
2
0
!2
!4
!6
!8
2000

2500

3000

revolutions per minute

3500

Fig. 6. Bifurcation diagram for minima (black dots) and maxima (grey
dots) of gearing deformation of the second gear mesh transmitting the
power corresponding to the static torque of 5 [Nm], see figure 4.

Figure 6 shows the bifurcation diagram of gearing deformation of gearing z = 2 for a chosen revolution range
of the gearbox, which is inscribed with the bold black line
in figure 4. The corresponding statical external torque has
than value of M = 5 Nm. In the bifurcation diagram a line
designating the zero gearing deformation is plotted. Each
dot plotted under this line corresponds to gear mesh interruption during one period of motion. Grey dots agree with
local maxima of gearing deformation and the black ones
agree with local minima, respectively. In the chosen operational area exist periodical solutions which follow the period doubling scenario. This nonlinear phenomenon mostly
accompanies the crossing of the area of constant gear mesh,
as  shows. Since the time course of the gearing deformation has a quasi-periodical structure, it is very difficult to
detect points of period doubling in the diagram 6.
Comparing figures 6 and 4, the point of loss of the constant gear mesh agrees and it happens at 2500 rpm. Increasing revolutions the motion overcomes according the nonlinear model to chaotic motion in the range of 2900-3000
rpm and 3250-3500 rpm. The contact loss endures then for

x 10

x 10

-6

3
3

1.5

1
0.5

(a)

x 10

0.005 0.01

(b)
-0.02

-6

x 10

-0.01

0.01

0.02

-6

5
0

4
3

-2

2
1
0

!6

-6

2.5

dz [m]

the rest of operational states, while according the map of

constant gear mesh gained from the linearised model in the
case of steady-state vibration the operational states between
3200 and 3450 rpm (see the bold black line in figure 4) lie
in the area of the constant gear mesh.
In figure 7 the bifurcation diagram corresponding to the
static unloaded gearing z = 5 with one freely rotating gear
is shown. It is obvious, the rotational motion of this gear
is chaotic in the whole operational area, but its impacts do
not too much influence the motion of the whole system. We
have chosen a low level of external statical torque to show
distinctly the nonlinear phenomena in gear meshings.
x 10

-4

(c)

-6
-0.02

(d)

0.02

-0.02

0.02

dz [ms1 ]

Fig. 8. Phase trajectories of second gear mesh G2 deformation d2 for

static torque M = 5 [Nm] for different values of revolutions per minute
of driving shaft: (a) 2100 rpm, (b) 2500 rpm, (c) 3100 rpm. Picture
(d) shows phase trajectory of the fifth gear mesh G5 deformation d5 for
3100 rpm.

0
!1
!2
!3
!4
!5
!6
!7
2000

2500

3000

revolutions per minute

3500

Fig. 7. Bifurcation diagram for minima (black dots) and maxima (grey
dots) of gearing deformation of the fifth gear mesh with freely rotating
driven gear corresponding to the static torque of 5 [Nm].

By increasing the level of external torque M and by decreasing the revolutions per minute of the driving shaft the
motion becomes more stable and the areas of chaotic motion disappear, as figures 4 and 6 shows. Moreover, the
number of impacts per one period of motion decrease. Figure 8 shows the evolution of phase trajectories of the second
and fifth gear meshes for different values of revolutions of
the driving shaft. The picture (a) shows a quasi-periodic solution of gearing deformation d2 corresponding to the motion with no impacts in gearing. Pictures (b) and (c) display
a quasi-periodic solution of the gear mesh G2 deformation
with contact loss in gearing. The case (d) corresponds to a
chaotic solution of the gear mesh deformation of the meshing G5 , which includes a freely rotating gear transmitting
no power.
IV. Conclusion
The paper describes the methodology of the large coupled rotating systems modelling and the analysis of their
nonlinear vibrations. The models of these systems suppose
a flexible stator and nonlinear gear couplings between rotor subsystems and nonlinear rolling-element bearings. The

whole system model is created by means of the modal synthesis method which allows to reduce number of degrees of
freedom of the mathematical model. The methodology is
applied to the real gearbox. Nonlinear vibrations excited
by kinematic transmission error accompanied by nonlinear
phenomena like bifurcation, periodic and quasi-periodic solutions and chaos are investigated. These regions of motions and changes among the regions are very interesting
from the theoretical and practical point of view.
This work was supported by the research project MSM
4977751303 of the Ministry of Education, Youth and Sports
of the Czech Republic.
References












Dresig H. Schwingungen mechanischer Antriebsysteme. SpringerVerlag, 2001.

Hajzman M. and Zeman V. Noise analysis and optimisation of gearboxes. Engineering Mechanics, Vol. 13, No. 2, pages 117-132, 2006.

Hortel M. and Skuderov

a A. Influence of linear and nonlinear damping on the stability of motion of kinematic pairs of gears. In Proc. of
Conf. Engineering mechanics 2006, pages 104-105, Svratka, Czech
Republic, May 15-18, 2006 (full text on cd-rom).
Zeman V. and Hajzman M. Modelling of shaft system vibration with
gears and rolling element bearings. In Proc. of Colloquium Dynamics of Machines 2005, pages 163-170, Prague, February 8-9, 2005.
Byrtus M. and Zeman V. Modelling and analyses of gear drive
nonlinear vibration. In Proc. of the Fifth EUROMECH Nonlinear
Dynamics Conference, pages 66-67, Eindhoven, August 7-12, 2005
(full text on cd-rom).
Zeman V. and Hlavac Z. Mathematical modelling of vibration of
gear transmissions by modal synthesis method. In Proceedings of the
Ninth World Congress on the Theory of Machines and Mechanisms,
Vol. 1, pages 397-400, Politechnico di Milano, Italy, 1995.
Kramer E. Dynamics of Rotors and foundations. Springer-Verlag.
Berlin, 1993.
Thompson J. M. and Stewart H. B. Nonlinear dynamics an chaos.
John Wiley & Sons, Chichester, 2002.