ARTICLE IN PRESS

International Journal of Machine Tools & Manufacture 50 (2010) 1928

Contents lists available at ScienceDirect

International Journal of Machine Tools & Manufacture

journal homepage: www.elsevier.com/locate/ijmactool

Investigation of variable optimum preload for a machine tool spindle

Shuyun Jiang , Hebing Mao
School of Mechanical Engineering, Southeast University, Nanjing 211189, China

a r t i c l e in fo

abstract

Article history:
Received 30 June 2009
Received in revised form
6 October 2009
Accepted 6 October 2009
Available online 14 October 2009

Angular contact ball bearings have been widely used in machine tool spindles, and the bearing preload
plays an important role on the performance of the spindle. With the development of high speed
machining, especially for high speed milling, heavy cutting at a low speed and light cutting at a high
speed are often performed on a single machine tool spindle, thus, high stiffness at low speed and low
temperature rise at high speed are required. The traditional constant pressure preload method cannot
meet the technical requirement of this kind of spindle any more. The variable preload technology is
systematically investigated in this paper. At high speed range, FEM method is used to analyze the
temperature distribution of the spindle, and the variable spindle preload is determined according to the
constraint of temperature rise of bearings. At low speed range, the spindle preload is resolved by the
fatigue life of bearings. The dynamic stiffness of the variable preload spindle is analyzed utilizing the
Transfer Matrix Method (TMM) and a nonlinear bearing model that includes the centrifugal force and
gyroscopic effects. An experimental set-up for the variable preload spindle is developed using hydraulic
pressure that can automatically adjust the bearing preload. The proposed method to determine variable
preload is veried experimentally by measuring the dynamic stiffness of the spindle and the
temperature rise of the test bearing. The results show that the variable preload spindle gives
outstanding behavior that the temperature rise at high speed is lower than that of the constant pressure
preload spindle, and the dynamic stiffness at low speed range is signicantly increased.
& 2009 Elsevier Ltd. All rights reserved.

Keywords:
Machine tool spindle
Variable bearing preload
Temperature rise
Fatigue life
Dynamic stiffness

1. Introduction
Initial preload of rolling bearings is widely applied to obtain
high stiffness of machine tool spindles, restrain vibration and
enhance rotational accuracy [13]. Proper preload of the angular
ball bearing is important to the speed, rigidity, accuracy and life of
the spindle. The proposed bearing load mechanisms mainly
include hydraulic instruments [3,4,7] and piezoelectric actuators
[8]. The controller using hydraulic pressure is a most popular
method for applying an automatic variable preload due to its
simple structure and maintenance.
Currently, with the development of high speed machining,
especially for high speed milling, heavy cutting at low speed and
light cutting at high speed are often performed on a single
machine tool spindle. For conventional spindle, the constant
pressure preload method is preferred because it is simple and easy
to maintain. However, the constant pressure preload is not
favored for a high speed machining spindle.
Literature review shows that some researchers have investigated the effect of the preload on the dynamic characteristics of
the spindle system [27], and found that high preload of the
bearing can enhance the stiffness and the natural frequency of the

 Corresponding author. Tel.: + 86 25 52090533.

E-mail address: jiangshy660118@yahoo.cn (S. Jiang).

0890-6955/$ - see front matter & 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijmachtools.2009.10.001

spindle. Other works have focused on the effects of bearing

preload on the thermal behaviors of the spindle [711], and
revealed that when the spindle speed exceeds a certain value,
there exists a variable bearing preload that produces minimum
temperature rise [8,9].
Hagiu [12] has studied the optimum preload for a spindle
supported by a pair of angle contact bearings (type of 7207) with a
rotational speed of 20,000 rpm. In the paper, the effect of axial
preload on the service life and dynamic stability was analyzed,
and the following conclusions were given:
(1) There are certain preload values that ensure higher service life
for the tested bearings and lower vibration levels of the test
spindle. Moreover, for these preload values stable functioning
thermal regimes for the test spindle were obtained.
(2) The natural frequencies of test spindle, also, the corresponding
vibration levels increase with the increase of bearing preload.
It should be pointed out that the optimum preload of the
spindle under a certain constant speed was discussed by Hagiu,
and the aim was to lengthen the service life of the bearing and
improve the dynamic behavior of the spindle at this certain speed.
In general, at low speed range, due to heavy cutting, a higher
preload is required to obtain high spindle stiffness and reduce the
vibration of spindle; at high speed range, due to heat generation of
bearing, a lower preload is required to reduce the heat amount

ARTICLE IN PRESS
20

S. Jiang, H. Mao / International Journal of Machine Tools & Manufacture 50 (2010) 1928

Nomenclature
A
a
a1
a2
a3
b
Cr
C0
cair
d
dm
Fa
Fr
Hf
hr
h0
kair
L10
L10a
M
M1
M2
N
n
Pr
P0
Q

cylindrical outside surface area of outer ring

semi-major axe
failure probability life adjustment factor
material life adjustment factor
operation condition life adjustment factor
semi- minor axe
basic rating dynamic radial load
basic rating static load
specic heat capacitance of air
diameter of the ow cross-section
bearing pitch diameter
axial load
radial load
heat generation of bearing
thickness of the outer ring
initial clearance t
thermal conductivity of air
fatigue life that 90% of a group of bearings will endure
adjusted basic rating life
amount of bearing friction torques
load torque
viscous friction torque
number of balls
spindle speed
equivalent dynamic radial load
equivalent static radial load
ball load

and lengthen the use life, and meanwhile to maintain necessary

stiffness for light cutting at a high speed. Therefore, for a high
speed machining spindle, a variable preload is needed to enhance
the performance of the entire speed range.
The main objective of this paper is to propose a method to
determine the variable preload of high speed machine tool spindle
at the entire speed range, so as to ensure that the spindle has
outstanding thermal and dynamic performances. In order to
obtain the optimum preload effectively, this study suggests that
the entire range of speed should be divided into low and high
speed sections; a lower preload is applied for the high speed
section within the limit of the temperature rise by the machine
tool industry standard, and a higher preload is applied for the low
speed section within the limit of bearing fatigue life by the general
requirement of the user.
Specically, in high speed section, the bearing preload can be
determined according to the constraint of temperature rise of
bearings, and the temperature distribution of the spindle has been
predicted by aid of FEM. In the low speed section, the preload can
be resolved by the bearing fatigue life. The dynamic stiffness of
the variable preload spindle has been studied by utilizing the
TMM and a nonlinear bearing model. A variable bearing preload
control system has been developed by using hydraulic transmission, which on-line supplies different preload levels by adjusting
the hydraulic pressure. An experimental set-up for the variable
preload spindle has been developed to verify the proposed
method by measuring the temperature rise of the bearing and
the dynamic stiffness of the spindle.

R
rh
Th
Tr
T0
X
Y
Xs
Ys

p
ah
ar
Z
l
la
lr

mair
n
nair
n0
n1
x

thermal contact resistance

inner diameter of housing mating bearing
temperature of housing inner face
temperature of outer ring
initial temperature
radial load factor of P0
axial load factor of P0
radial load factor of Pr
axial load factor of Pr
thermal permissivity
linear thermal expansion coefcient of the housing
linear thermal expansion coefcient of the outer ring
convection coefcient
thermal conductivity of the half space
thermal conductivity of air
thermal conductivity of outer ring
dynamic viscosity of air
viscosity of lubricant
kinematic viscosity of air
kinematic viscosity of lubricant
standard viscosity of lubricant
complete elliptic integral of the second kind

Subscripts
air
h
r

refers to air
refers to housing
refers to bearing outer ring

driven by a motorized high speed spindle. The shaft of the spindle

is supported by two pairs of angular contact ball bearings, which
are preloaded by a hydraulic chamber. The hydraulic pressure
induces displacement in the axial direction, then pushing the
outer ring of the rear bearings, and eventually the displacement
will be transferred to the inner ring of front bearings through the
shaft of the spindle. Therefore, the displacement caused by
hydraulic pressure is converted into an axial force, which
increases the preload of bearings, and the preload of each
bearing is equal to one half of the hydraulic pressure. The
bearings are grease lubricated in this study, while the oilair
lubrication nozzle is devised in advance for further study. A disk
with an unbalanced mass is mounted at the location of spindle
nose to generate a self-exciting force when it rotates, which can be
used to measure the dynamic stiffness of spindle.
The operating principle of the variable preload spindle system
is as follows. The rotational speed of the spindle is detected
through a revolution transmitter. According to the detected
spindle speed, a control signal is sent to the D/A conversion.
Subsequently, the current of proportional electromagnetic is
adjusted to change the hydraulic pressure in the chamber. Finally,
the variable preload is applied to the bearings for different spindle
speeds. As shown in Fig. 1, the temperature rise of the front
bearings is measured by thermocouple, the displacement of the
spindle nose is detected by an eddy current sensor and a signal
acquisition unit.

3. Analysis of bearing preload

2. Variable preload spindle system

3.1. Preload for high speed

Fig. 1 shows a schematic of the overall concept of the variable

preload spindle system developed in this study. The test spindle is

The Finite Element Method has been successfully applied to

analyze the temperature distribution of the integrated spindle

ARTICLE IN PRESS
S. Jiang, H. Mao / International Journal of Machine Tools & Manufacture 50 (2010) 1928

signal
acquisition unit
driving motor
(motorized spindle)

power
amplifier

D/A
conversion

21

proportional
electromagnetic

hydraulic
station

variable preload spindle

(test spindle)
thermocouple

hydraulic
chamber
revolution
transmitter

disk
unbalanced
mass

coupling
eddy current
sensor
oil-air lubrication nozzle

rearing bearings

front bearings
signal
acquisition unit

computer
Fig. 1. Schematic of the variable preload spindle system.

[1317]. In this section, the effects of the ball bearing preload on

the temperature rise at high speed is analyzed by aid of FEM (the
commercial software ANSYS is selected). The bearing preloads
corresponding to rotational speeds were determined according to
the constraint of temperature rise of bearings provided by
machine tool industry.
3.1.1. Heat sources of spindle
Without considering the windage losses in this spindle, which
is the viscosity shear friction of the air between the shaft and
housing, heat is mainly generated at bearing raceways and balls
due to the friction inuenced by speed, preload and lubricant.
Bearing temperature would rise substantially due to the heat
generated by friction losses and rolling resistance. The empirical
heat generation in the bearing is given by [1,18]
Hf 1:047  104 nMW

A reasonable estimate of the total friction of a given ball

bearing under moderate preload, lubricant and speed conditions
is the sum of the load torque and viscous friction torque. But when
a ball bearing operates at a high rotation speed, the effects of
centrifugal force and gyroscopic moment should be considered,
which has been expressed by a spinning friction item [1], if any,
that is,
M M1 M2 Ms

where M1 is the load torque due to all mechanical friction

phenomena except for uid friction, and is the function of applied
load. The following equation is given to describe this torque:
M1 f1 p1 dm Nmm

in which parameter f1 is a factor depending on bearing design and

relative bearing preload, p1 depends on the magnitude and
direction of the applied load.
For angular contact ball bearings,
f1 0:0013P0 =C0 0:33

p1 Fa  0:1Fr

Table 1
Values of Xs and Ys for angular contact ball bearings.
Contact angle

151

201

251

301

351

401

Xs
Ys

0.5
0.47

0.5
0.42

0.5
0.38

0.5
0.33

0.5
0.29

0.5
0.26

where P0 Xs Fr Ys Fa ; the values of Xs and Ys for the single-row

angular contact ball bearings with different contact angles are
given in Table 1[1].
M2 in Eq. (2) is the viscous friction torque, and can be
empirically expressed as follows [1]:
M2 107 f0 n0 n2=3 d3m N mm;
M2 160  107 f0 d3m N mm;

n0 n 4 2000
n0 n r 2000

6
7

where f0 is a factor that depends on bearing type and lubrication

type, and for angular contact ball bearing f0 = 2, n0 is the
kinematics viscosity of lubricant under an operation temperature.
Ms in Eq. (2) is the spinning friction moment, and can be
expressed as follows [1]:
Ms

3mQax
8

where m is the coefcient of friction with value from 0.01 to 0.06.

The parameters of the variable preload spindle (as shown in
Fig. 1) and its bearings are listed in Tables 2 and 3, respectively.
The bearing is lubricated with IsotexTopasL30 grease, and Table 4
gives its base oil viscositytemperature relationship.
The room temperature is controlled as constant 20 1C.
Generally, the steady state temperature rise of the bearing outer
ring is limited to no more than 25 1C, according to machine tool
industry standard. Thus, the outer ring is limited to 45 1C. Usually,
the inner ring is 510 1C higher than the outer ring, and the ball is
10 1C higher than the inner ring [19]. In this paper, the base oil
viscositytemperature characteristics of the grease are considered
for the heat generation.

ARTICLE IN PRESS
22

S. Jiang, H. Mao / International Journal of Machine Tools & Manufacture 50 (2010) 1928

Table 2
Parameters of the test spindle.
Parameter

Value

Rotational speed (rpm)

Bearing span (mm)
Max diameter of shaft (mm)
Length of shaft (mm)

07000
256
85
435

Table 3
Parameters of ball bearings.
Bearing

Front bearing

Rear bearing

Type
Material
Inner diameter (mm)
Outer diameter (mm)
Width (mm)
Ball diameter (mm)
Number of balls
Contact angle (deg.)

7015
Steel
75
115
20
12.3
20
25

7012
Steel
60
95
18
11.0
18
15

Table 4
The base oil viscositytemperature relationship of IsotexTopasL30.
Temperature (1C)
25
40
60

Kinematics viscosity (mm2/s)

3. The bearing balls are supposed as equal volume elements,

which share only one node in contact with the housing and the
shaft, respectively.
4. Since ball bearing rotates at a high speed, heat has been
distributed evenly, and consequently temperatures are the
same among every ball.
5. Temperatures gradually transfer from heat sources to the shaft
and housing. Thus, temperature maps of the model are
continuous.
6. Thermal-induced load of the shaft is neglected because of
hydraulic preload mechanism [21].
7. Heat radiation to ambient air is ignored.
Heat transfer conditions of main spindle components and heat
dissipation boundary conditions include:
1. Convection of the air gap between the shaft including mounted
components and the housing.
2. Convection of the shaft nose and ambient air.
3. Free convection of ambient air around stationary surfaces (the
housing, etc.).
4. Conduction of the initial clearance t between the outer
bearing ring and the housing.
5. Conduction from balls to the inner ring and outer ring.
6. The initial temperatures of bearings and room temperature are
measured and given.

35
18
10

0.020

3.1.2.1. Convection of the air between shaft and housing. For this
experiment condition, the air owing through the spindle system
is assumed to be forced turbulent air, and the convection coefcient can be calculated by [20]

0.016

0.012

where Z is the convection coefcient between forced air and

surfaces of components, Nu is the Nusselt number, kuid is the
thermal conductivity of the uid, and d is the diameter of the ow
cross-section of the cylinder.
Air in this study is highly turbulent, and hence the Nusselt
number is given according to [22]

Ms / M

0.024

0.008
0.004
0.000
0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07

Fig. 2. The ratio of Ms to M (rotational speed, 7000 rpm; axial preload, 1000 N).

Nu kair
d

Nu 0:133 Re2=3 Pr1=3

10

Eq. (10) proposed is valid for

Re o4:3  105

This test spindle is designed with the maximum rotation speed

of 7000 rpm. Fig. 2 shows the variation in the ratio of Ms to M
under the maximum speed of 7000 rpm and the preload of 1000 N
with the different friction coefcient m. The analytic calculation
indicates that the value of spinning friction is only 2.1% of the total
friction M. Thus, the spinning friction is not the main heat source
for the test spindle in this paper.

3.1.2. Heat sinks in spindle

Some assumptions are made for this spindle model, such as
[10,20]
1. The model is axisymmetric without considering some vent
holes, oil-leaking holes, nuts and some other micro-structures.
2. The spindle including housing and shaft system is meshed as a
whole system.

0:7 oPr o 670

where Re is the Reynolds number, and Pr the Prandtl number.
Two equations are used for Reynolds number and Prandtl
number, respectively:
Re uair d=nair

11

Pr cair mair =kair

12

where uair is the mean velocity of the uid owing through the
shaft surface superimposed by axial and tangent speeds.
Fig. 3 shows the heat convection coefcients of different
diameters of the multi-diameter shaft with respect to spindle
speed, and those of the housing inner surface are listed in Table 5.

3.1.2.2. Convection of ambient air. A free convection coefcient can

be assumed for ambient air around stationary surfaces like the

ARTICLE IN PRESS
S. Jiang, H. Mao / International Journal of Machine Tools & Manufacture 50 (2010) 1928

where c is a geometric factor depending on the size of the contact

area.
c is a function of k 1  a2 =b2 and is dened as

d=27mm
d=34mm
d=50mm
d=65mm
d=85mm
d=100mm

100

Rb

50

4000

5000
Spindle speed (rpm)

Spindle speed (rpm)

3000
4000
5000
6000
7000

6000

7000

Convection coefcient (W/m2K)

12.5
15.2
17.6
19.9
22.1

spindle housing. The coefcient can be obtained as a = 9.7 W/

(m2 K) according to [10].

3.1.2.3. Conduction between outer bearing rings and housing.

Conduction through the clearance t of the outer ring and the
housing is temperature-related. The equivalent permissivity can
be obtained as [10]
1
A
hr =lr hg =la

13

Here lr =60.5 W/m K and la =0.0283 W/m K at 45 1C.

Considering both thermal induced lengths of the housing and
the outer ring, the average air gap of the clearance t hg around
the perimeter is [10]
hg h0  Tr  T0 ar  Th  T0 ah rh

c
4l1 a

17

4l2 a

Rb Rb =N

3.1.2.4. Conduction between balls and raceways. The following

calculations of contact resistances between the balls and the raceways are based on the work by Harris [1], Nakajima [23] and
Bossmanns [10]. The thermal contact resistance R for a half-space
is
15

18

Similarly, the contact resistance from balls to the inner

raceway can be determined. Thermal permissivity as the inverse
of thermal contact resistance can be further determined. The
thermal permissivities of the front and the rear bearings are
plotted in Fig. 3 as functions of preload. Specially, the value of
preload in this paper refers to the total preload of the spindle
system, i.e. the pressure Pd in the hydraulic chamber, and each
bearing preload is one half of the spindle preload Pd.
Fig. 4 shows that the thermal permissivity of the outer ring is
greater than that of the inner one. It can be attributed to the
rolling bearing analysis theory by Harris [1]; a concave surface
will conform to contacting bodies, thus reducing the curvature.
Conversely, convex surfaces increase the curvature. Thus, the
contact area of outer ring raceway is larger than that of inner ring
raceway under a xed axial preload, and subsequently, the
thermal permissivity between the ball and the outer ring
raceway is larger than that between the ball and the inner ring
raceway.

3.2. Preload for low speed

The appropriate preload range for low speed is obtained based
on the Fatigue Life Theory for calculating bearing basic rating life
in this study. The Fatigue Life Theory was rst proposed by
Lundberg and Palmgren [24], and nally adopted by International
Standards Organization Standard ISO 281/1 [25]
L10a a1 a2 a3 L10

14

Here h0 = 15 mm, Th = 43 1C, Tr = 45 1C, T0 = 20 1C, ar = 1.17 

10  5 1/K for outer ring of GCr15 and ah = 2.3  10  5 1/K for
housing of 45# steel.
The thermal permissivity for the front bearing (7015) is
obtained as 5.36 W/K by using Eq. (13) together with Eq. (14)
with a ring thickness hr of 5.5 mm, rh of 115 mm, and A of
0.00723 m2. The thermal permissivitiy for the rear bearing (7012)
is obtained as 4.88 W/K with a ring thickness hr of 4.65 mm, rh of
95 mm, and A of 0.00537 m2.

16

Combining all the balls, the combined thermal resistance from

the ball to the outer ring becomes

Table 5
Convection of the housing inner surface.

4la

dy
p
1  k2 sin2 y

For a bearing ball, its contact resistance to the outer ring

becomes

Fig. 3. Convection of the shaft.

Zp=2

150

0
3000

19

1.2
1.0
Thermal Permissivity (W/K)

Convection coefficient (W/m2K)

250

200

23

0.8
0.6
Inner Ring/Ball Contact, Rear
Outer Ring/Ball Contact, Rear
Inner Ring/Ball Contact, Front
Outer Ring/Ball Contact, Front

0.4
0.2
0.0
0

1000

2000

3000 4000 5000

Preload (N)

6000

7000 8000

Fig. 4. Thermal permissivity of the front and the rear bearings.

ARTICLE IN PRESS
S. Jiang, H. Mao / International Journal of Machine Tools & Manufacture 50 (2010) 1928

L10

 
106 Cr e
60n Pr

20

where Pr XFr YFa ; the values of X and Y for the single-row

angular contact ball bearing with a contact angle of 151 are listed
in Table 6[1]. And the parameter e = 3 for ball bearings was
substantiated by the US National Bureau of Standards [26].
Since a2 and a3 are closely interrelated and interacted, an
integrated correction factor a23 is applied to calculate the life of
bearing. The ratio of viscosity is used to determine the value of a23

v
v1

21

where k is the ratio of the viscosity of the lubricant and the

standard viscosity of the lubricant, n is the viscosity of the
lubricant, and n1 is the standard viscosity of the lubricant
subjected to rotating velocity, bearing type and size.
a1 and a23 can be obtained in detail according to [27] under a
variety of preloads and lubrication conditions. And then the life of
the front bearing in this spindle can be acquired.

4. Calculations of the variable preloads

It has been conrmed by our previous test studies that a higher
rotating speed of bearing will lead to a higher temperature rise,
especially for this spindle lubricated with grease. This is because
Table 6
Values of X and Y for angular contact ball bearing.
Contact angle

151

Fa
C0

0.015
0.029
0.058
0.087
0.12
0.17
0.29
0.44
0.58

Fa
iZD2

0.172
0.345
0.689
1.03
1.38
2.07
3.45
5.17
6.89

Fa
Fr

4 e FFar r e

0.44

1.47
1.40
1.30
1.23
1.19
1.12
1.02
1.00
1.00

the frictional torque of the bearing includes the spinning moment

in addition to the lubricant viscosity and applied load terms. The
test has also shown that the steep increase in the temperature rise
occurs as the rotational speed is within a certain range of rotation
speed (about 30003500 rpm for this test spindle), and when the
rotation speed is high than this range, the increment is extremely
large. Furthermore, this type of spindle is designed with the
maximum rotation speed of 7000 rpm, according to the machining process, the working speeds higher than about 3000 rpm are
considered as the range of light cutting at high speed. Based on
the above discussion, the rotational speed of 3000 rpm is adopted
to divide the sections for low and high speed roughly.
4.1. Preload for high speed
Because the spindle is axisymmetric in geometry and thermal
boundary conditions without considering some micro-structures,
the thermal model is an axisymmetric two dimensional
model, and then it was meshed with the element PLANE 55. The
thermal parameters of the materials used in the spindle are given
in Table 7.
Fig. 5 shows the continuous simulation temperature
distribution of the spindle after achieving steady state under a
certain condition. It can be seen that the maximum temperature
55.7 1C is located at bearing ball because the bearing is the single
heat source of this system, and the balls heat capacity is relatively
small. It can also be observed that the temperature drops
gradually to housing and shaft. The shaft temperatures are
shown to be higher than the housing since a high heat
conduction resistance exists between bearing outer rings and
the housing.

100
0.38
0.40
0.43
0.46
0.47
0.50
0.55
0.56
0.56

Table 7
Thermal parameters of materials.
Material

45#Steel

40Cr

GCr15

Density (kg/m3)
Modulus of elasticity (GPa)
Poisson ratio
Specic heat capacity (J/kg K)
Conduction coefcient (W/m K)

7.85
205
0.28
480
50.2

7.85
205
0.29
460
60.5

7.85
208
0.29
460
50

90

Temperatures (C)

24

80
70

1000N
2000N
3000N
4000N
5000N
6000N
7000N
8000N

60
50
40
30
3000

4000

5000
6000
Spindle speed (rpm)

7000

Fig. 6. Steady state temperatures for different preloads and spindle speeds.

Fig. 5. Temperature distribution of the spindle (rotational speed, 7000 rpm; preload, 1000 N).

ARTICLE IN PRESS
S. Jiang, H. Mao / International Journal of Machine Tools & Manufacture 50 (2010) 1928

Speed (rpm)

Variable preload (N)

r 4000
r 5000
r 6000
r 7000

2000
1000
1000
1000

Table 9
Lives of the front bearing (h).

1000
2000
3000
4000
5000
6000
7000
8000

Rotational speed (rpm)

4000

5000

6000

7000

579,137
72,392
21,449
9049
4633
2681
1688
1131

463,309
57913
17159
7239
3706
2144
1350
904

386,091
48,261
14,299
6032
3088
1787
1125
754

330935
41366
12256
5170
2647
1532
964
646

Table 10
Lives of the front bearing (h).
Preload (N)

1000
2000
3000
4000
5000
6000
7000
8000

Speed (rpm)

Variable preload (N)

r 1000
r 2000
r 3000

8000
7000
6000

40

30

20

10

0
1000

constant preload (simulated)

variable preload (simulated)

2000

3000
4000
5000
Spindle speed (rpm)

6000

7000

Fig. 7. Comparisons of temperature rises between variable preload spindle and

constant preload spindle.

Table 8
Variable preloads for high speed range.

Preload (N)

Table 11
Variable preloads at low speed range.

Temperature rise (C)

Fig. 6 illustrates the relationships between bearing

temperatures and rotational speed under different preloads. It
can be seen that all temperatures increase almost linearly with
the increases of rotational speed and preload. Specically, the
effects of preloads on the temperature rise increase as the
rotational speed increases. When the preload increases from
1000 to 8000 N at speed of 3000 rpm, the temperature increases
from 40.4 to 55.6 1C, and the percent of increment is only 37.6%.
When at 7000 rpm, the temperature increases from 55.1 to
80.0 1C, and the percent is up to 45.2%.
For a xed rotation speed, when the temperature rise of the
ball bearing is limited to 25 1C by machine tool industry standard,
the maximum corresponding preload in Fig. 6 can be chosen as
the variable preload. When the spindle speed reaches 6000 rpm,
the temperature rise will exceed 25 1C while the preload is
reduced to a lower level. From the rolling bearing analysis theory
[1], the spinroll ratio is increased by decreasing the preload. As
the spinroll ratio increases, the bearing produces more friction
and hence heat generation. Meanwhile, it is observed from our
experiment that the preload less than 1000 N cannot eliminate
clearance effectively and thus decrease the rigidity and increase
the friction and wear resulting from excessive sliding. Therefore,
the preload of 1000 N can be determined for high speed range
over 6000 rpm. The variable preloads with the rotation speed
during high speed range (30007000 rpm) can be obtained, as
listed in Table 8.

25

Rotational speed (rpm)

500
1000

2000

3000

5,575,215
808,860
273,130
128,980
72,656
45,535
30,627
21,663

2,555,320
370,730
125,190
59,115
33,301
20,870
14,037
9929

1,935,833
280,850
94,838
44,784
25,228
15,811
10,634
7522

3,949,136
572,940
193,470
91,359
51,465
32,254
21,694
15,345

4.2. Preload for low speed

Fatigue life theory (Eq.(19)) is applied to analyze and
determine the preload of the spindle at low speed range. The
rotation speeds from 500 to 3000 rpm and the preloads from 1000
to 8000 N are used to estimate fatigue life. The lives of the front
bearing under various preloads and speeds are listed in Table 10.
Considering the actual operating conditions, two shifts a day
with 8 h per shift, the rated life of the spindle is about 2 years, i.e.
the expected life of the bearing is about 12,000 h. For a xed
rotation speed, when the life of calculation is larger than 12,000 h,
the maximum corresponding preload is the variable preload.
Thus, the variable preload at low speed is obtained from the
calculated results, as listed in Table 11.

5. Comparison of spindle behavior between variable preload

and constant preload
5.1. Temperature rises
Based on the analysis above, the variable preload for the entire
speed range can be obtained from Tables 8 and 11. A preload of
2000 N is selected as the constant preload according to the
medium preload. The temperature rises of bearing under variable
preload and the constant pressure preload were simulated by aid
of FEM. The temperature rises corresponding to the rotational
speeds are shown in Fig. 7.
For constant pressure preload, the temperatures of the bearing
increase almost linearly with the spindle speed. For variable
preload, the bearing temperatures increase when the speed is less
than 3000 rpm, and decrease as the spindle speed exceeds
3000 rpm. When the spindle speed exceeds 4000 rpm, the bearing
temperatures are lower than the constant pressure preload, by
3 1C at 7000 rpm.

ARTICLE IN PRESS
26

S. Jiang, H. Mao / International Journal of Machine Tools & Manufacture 50 (2010) 1928

5.2. Dynamic stiffness

The variable preload for different spindle speeds has been
presented in the above sections. This section focuses on the effects
of the variable preload on dynamic stiffness of the spindle; the
dynamic behavior of a spindle including critical speed and
unbalance response can be simulated by utilizing TMM [2830]
and a nonlinear bearing model [1] that includes the centrifugal
force and gyroscopic effects. Recently TMM has been applied to
analyze the dynamic stiffness of the machine tool spindle in our
laboratory, and computer program was developed [31]. The
dynamic stiffness of the variable preload spindle was systematically studied by using TMM in this paper.
The parameters of the variable preload spindle are listed in
Table 3, and the variable preloads for different rotational speeds
have been obtained as listed in Tables 8 and 11. To illustrate the
outstanding dynamic behavior of the variable preload spindle, the
dynamic stiffness of the spindle is predicted and compared with
that of the constant pressure preload spindle with a preload of
2000 N, as illustrated in Fig. 8. It can be seen that at low speed
range, the stiffness of the variable preload spindle is signicantly
enhanced. However, at high speed range, the spindle stiffness is
lower than that of the constant pressure preload spindle. In detail,
at static steady, the static stiffness of the variable preload spindle
can reach 182.52 N/mm, while that of the constant pressure
preload spindle is only 149.21 N/mm. With the rotation speed
increasing, the preload for variable spindle decreases; in turn, the
stiffness of the spindle also decreases. At a speed of 4000 rpm, the
variable preload of the variable preload spindle is equal to that of
the constant pressure preload spindle, the dynamic stiffness of the
two spindles are equal. When the rotation speed exceeds
4000rpm, the dynamic stiffness of the variable preload spindle
is lower than that of the constant pressure preload spindle. As the
spindle reaches the maximum speed of 7000 rpm, the dynamic
stiffness of the variable preload spindle is 129.08 N/mm, lower
than that of the constant pressure preload spindle by 12.8%.

5.3. The bearing service life in the high speed section

Fatigue life theory is applied to determine the preload at low
speed range in this paper. It is well known that the service life will
be shortened as rotational speed increases. In the high speed
section, a lower preload has been applied to the bearing to

Dynamic stiffness of spindle (N/m)

200

decrease the temperature rise within a limit, but it is necessary to

pay attention to the service life and ensure whether it is higher
than the rating life by machine tool industry standard.
A comparison in service lives of the front bearing with a series
of preloads under different high speeds is list in Table 9, where, it
can be seen that the life is shortened apparently as rotational
speed increases, while it is shortened sharply with the increase of
preload. For the variable preload spindle, the preload of 1000 N is
for the operating speed ranging from 5000 to 7000 rpm, and
2000 N for the speed from 4000 to 5000 rpm. It is conrmed from
results that the estimated service lives corresponding to the
working conditions are much higher than the expected life of
12,000 hours.
A lower preload has been adopted for the high speed section
within the limits of a temperature rise by machine tool industry
standard and a necessary dynamic stiffness for maintaining light
cutting. The above discussion illustrates that the spindle with
variable preload can reach a much higher working speed, and the
lower preload at high speed can be greatly helpful to lengthen the
service life of the bearing.

6. Experimental verication
6.1. Temperature rise
To verify the proposed method for variable preload in this
paper, an experimental set-up for the spindle was developed, as
shown in Fig. 9. In this section, the temperature rises were
detected by two thermocouples.
The temperature rises of both the variable preload and the
constant preload spindle were measured during high speed range,
as shown in Fig. 10. It can be seen that the simulation results agree
with the experimental measurement, and the bearing
temperature of the variable spindle is decreased in the high
speed range compared with that of the constant pressure preload
spindle. Furthermore, the experimental temperatures were all
lower than the theoretical under the constant pressure preload
and the variable preload due to some assumptions introduced into
the thermal model. The difference of temperatures between
theory and experiment is not signicant for constant pressure
preload spindle. As to the variable preload spindle, the
experimental temperatures are slightly lower than the
theoretical from 3000 to 5000 rpm, but agree well from 6000 to
7000 rpm.
6.2. Dynamic stiffness

190
Experiments were carried out to verify the predicting dynamic
stiffness of the spindle according to the rotation speed. As shown
in Fig. 1, an unbalanced mass is mounted on the disk located at
spindle nose to generate an exiting force during the rotation of the
spindle. The displacement of the spindle nose is detected by an
eddy current sensor and the displacement signal is sampled by a
signal acquisition unit (AZ308R), then the sampled signal is
analyzed by the CRAS V7.0 software and the out-of-balance
response at the spindle nose is obtained. The dynamic stiffness at
the spindle nose is as follows:

180
170
160
150
140
130
constant preload (simulated)
variable preload (simulated)

120

110
100
0

1000

2000 3000 4000 5000

Spindle speed (rpm)

6000 7000

Fig. 8. Comparison of the dynamic stiffness between two spindles.

me o2
9x9

where m is the mass of the unbalanced mass, 9x9 is the amplitude

of the displacement at the spindle nose.
The dynamic stiffness of the spindle under the constant
pressure preload and variable preload as stated above are

ARTICLE IN PRESS
S. Jiang, H. Mao / International Journal of Machine Tools & Manufacture 50 (2010) 1928

27

40
35
Temperature rise (C)

30
25
20
15
Constant preload (theoretical)
Constant preload (experimental)
Variable preload (theoretical)
Variable preload (experimental)

10
5
0
3000

Dynamic stiffness of spindle (N/m)

Fig. 9. The architecture of test rig.

200
190
180
170
160
150
140
130

contant preload (simulated)

contant preload (experimental)
variable preload (simulated)
variable preload (experimental)

120
110
100

4000

5000
Spindle speed (rpm)

6000

7000

Fig. 10. Comparison of experimental temperature rises and the theoretical.

measured, respectively. The comparison of experimental and

simulation results is shown in Fig. 11, where it can be seen that
the simulation results are essentially coincident with the
experimental measurement, and the stiffness of the variable
preload spindle is signicantly enhanced at low speed range. At
high speed range, the stiffness of the variable preload spindle is
lower than that of the constant pressure preload spindle, and the
dynamic stiffness of the variable preload spindle is 86.0% of that of
the constant pressure preload spindle at rotation speed of
7000 rpm. The variable preload technology can be applied to
machine tool spindle, which can not only decrease the

1000

2000 3000 4000 5000

Spindle speed (rpm)

6000

7000

Fig. 11. Comparison of experimental and simulated results.

temperature rises of rolling bearing at high speed range but also

improve the dynamic stiffness of spindle at low speed range. The
above results are promising considering the fact that heavy
cutting at low speed and light cutting at high speed are demanded
on a single machine tool spindle in current high machining
environment.
The variable preload corresponding to the rotation speed is
investigated in this paper, while, in practice, variable preload also
depends on cutting forces, especially for the heavy cutting at lower
speed. It should be pointed out that the variable preload with the
rotation speed proposed in this paper has laid a solid foundation for

ARTICLE IN PRESS
28

S. Jiang, H. Mao / International Journal of Machine Tools & Manufacture 50 (2010) 1928

further investigation on the variable preload with cutting forces.

The cutting forces can be converted to equivalent axial preload
based on bearing theory, and furthermore a closed loop control
system can be introduced into a machining center to automatically
adjust preload considering both the rotation speed and the cutting
force. This issue will be further studied in the future.

7. Conclusions
This paper has developed a method to analyze the variable
preload with rotation speed for a high speed machining spindle. The
entire range of speed is divided into low and high speed sections; a
lower preload is applied for the high speed section within the limit
of the temperature rise and a higher preload is applied for the low
speed section within the limit of bearing fatigue life.
In the high speed section, the temperature distribution of the
spindle has been analyzed by aid of FEM, and the spindle preload
was determined according to the constraint of temperature rise of
bearings. In the low speed section, the spindle preload has been
resolved by the fatigue life of bearings. The dynamic stiffness of
the variable preload spindle was studied by utilizing TMM and a
nonlinear dynamic model of bearing. The proposed method for the
variable bearing preload has been veried by experiment study.
Compared with the traditional constant pressure preload
technology, the temperature rise of the variable preload spindle
bearing at high speed range is reduced due to a lower preload, and
the lower preload can be greatly helpful in lengthening the service
life of the bearing. Meanwhile, the dynamic stiffness of the
variable preload spindle is signicantly enhanced at low speed
range due to a larger preload. Furthermore, the dynamic stiffness
of the variable preload spindle is 86.0% of the constant pressure
preload spindle at a maximum speed of 7000 rpm, though it
decreases slightly at high speed. The above results are promising
considering the fact that heavy cutting at low speed and light
cutting at high speed are demanded on a single spindle of high
speed machining.
The hydraulic actuator chosen for the preload applied is based
on the fact that the hydraulic system is an essential subsystem of
a machining center, for example, the automatic tool changer is
usually actuated by a hydraulic cylinder installed at the rear end
of the spindle. So we think it is simple and cheap to apply the axial
preload by using the hydraulic system employed in the machine
tool.

Acknowledgements
The authors gratefully wish to acknowledge the supports of
National Science Foundation through Grant nos. 50475073,
50775036 and Jiangsu High Technology Research and Development Project through Grant nos. BK2002059, BG2006035,
BK2009612.
References
[1] T.A. Harris, in: Rolling Bearing Analysis, third ed., John Wiley and Sons, New
York, 1991.

[2] Mohammed A. Alfares, Abdallah A. Elsharkawy, Effects of axial preloading of

angular contact ball bearings on the dynamics of a grinding machine spindle
system, Journal of Materials Processing Technology 136 (2003) 4859.
[3] Yuzhong Cao, Yusuf Altintas, A general method for the modeling of spindlebearing systems, Journal of Mechanical Design, Transactions of ASME 126
(2004) 10891104.
[4] Yuzhong Cao, Y. Altintas, Modeling of spindle-bearing and machine tool
systems for virtual simulation of milling operations, International Journal of
Machine Tools & Manufacture 47 (2007) 13421350.
[5] S.A. Spiewak, T. Nickel, Vibration based preload estimation in machine tool
spindles, International Journal of Machine Tools & Manufacture 41 (2001)
567588.
[6] Chi-Wei Lin, Jay F. Tu, Model-based design of motorized spindle systems to
improve dynamic performance at high speeds, Journal of Manufacturing
Processes 9 (2007) 94108.
[7] Chi-Wei Lin, Jay F. Tu, Joe Kamman, An integrated thermo-mechanicaldynamic model to characterize motorized machine tool spindles during very
high speed rotation, International Journal of Machine Tools & Manufacture 43
(2003) 10351050.
[8] Jenq-Shyong Chen, Kwan-Wen Chen, Bearing load analysis and control of a
motorized high speed spindle, International Journal of Machine Tools &
Manufacture 45 (2005) 14871493.
[9] Y.-R. Jeng, C.-C. Gao, Investigation of the ball-bearing temperature rise under
an oilair lubrication system, Proceedings of the Institution of Mechanical
Engineers 215 (Part J) (2001) 139148.
[10] Bernd Bossmanns, Jay F. Tu, A thermal model for high speed motorized
spindles, International Journal of Machine Tools & Manufacture 39 (1999)
13451366.
[11] Cheng-Hsien Wu, Yu-Tai Kung, A parametric study on oil/air lubrication of a
high-speed spindle, Precision Engineering 29 (2005) 162167.
[12] G. Hagiu, Reliable high speed spindles by optimum bearings preload,
International Journal of Applied Mechanics and Engineerings 8 (2003) 5770.
[13] V.P. Raja, P.R. Pillai, Thermal analysis of spindle units under high speed
machining, IE (I) Journal-MC 81 (2001) 155159.
[14] Jin Kyung Choi, Dai Gil Lee, Thermal characteristic of the spindle bearing
system with a gear located on the bearing span, International Journal of
Machine Tools & Manufacture 38 (1998) 10171030.
[15] H.J. Pahk, S.W. Lee, Thermal error measurement and real time compensation
system for the CNC machine tools incorporating the spindle thermal error and
the feed axis thermal error, International Journal of Advanced Manufacture
Technology 20 (2002) 487494.
[16] Xu Min, Jiang Shuyun, Ying Cai, An improved thermal model for machine tool
bearings, International Journal of Machine Tools and Manufacture 47 (2007)
5362.
[17] Sun-Min Kim, Sun-Kyu Lee, Prediction of thermo-elastic behavior in a
spindle-bearing system considering bearing surroundings, International
Journal of Machine Tools & Manufacture 41 (2001) 809831.
[18] A. Palmgren, in: Ball and Roller Bearing Engineering, S.H. Burbank,
Philadelphia, PA, 1959.
[19] Ono Fan, in: Application of Rolling Bearing Design, River Press, Tokyo, 1979.
[20] Hongqi Li, Yung C. Shin, Integrated dynamic thermo-mechanical modeling of
high speed spindles, Part 1: model development, Journal of Manufacturing
Science and Engineering, Transactions of the ASME 126 (2004) 148158.
[21] Hongqi Li, Yung C. Shin, Analysis of bearing conguration effects on high
speed spindles using an integrated dynamic thermo-mechanical spindle
model, International Journal of Machine Tools & Manufacture 44 (2004) 347
364.
[22] F. Kreith, Convection heat transfer and ow phenomena of rotating spheres,
International Journal of Heat and Mass Transfer 6 (1963) 881.
[23] K. Nakajima, Thermal contact resistance between balls and rings of a bearing
under axial, radial and combined loads, Journal of Thermophysics and Heat
Transfer 9 (1) (1995) 8895.
[24] G. Lundberg, A. Palmgren, Dynamic capacity of rolling bearings, Acta
Polytechnica (7) .
[25] ISO 281/1, Dynamic Load Ratings and Rating Life, Part 1, Calculation Methods,
1977.
[26] J. Lieblein, M. Zelen, Statistical Investigation of Fatigue Life of Ball Bearings,
National Bureau of Standards, Report No. 3996, March 28, 1955.
[27] E.V. Zaretsky, E. Bamberger, T. Harris, et al., Life adjustment factors for ball
and roller bearings, ASME Engineering Design Guide, 1971.
[28] J.S. Rao, in: Rotor Dynamics, J. Wiley, New York, 1983.
[29] L.T. Yan, Z.G. Zhu, Z.H. Song, Q.H. Li, F.J. Liu, in: Aynamic Analysis of Structural
System, Beihang University Press, Beijing, 1989.
[30] Y.E. Zhong, Y.Z. He, Z. Wang, F.Z. Li, in: Rotor Dynamics, Tsinghua University
Press, Beijing, 1987.
[31] M. Qian, S.Y. Jiang, Dynamic optimization of a high speed motorized spindle,
China Mechanical Engineering 16 (10) (2005) 864868.