See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/226865448

Dynamic analysis of a linear motion guide having rolling elements for

precision positioning devices

Article  in  Journal of Mechanical Science and Technology · April 2008

DOI: 10.1007/s12206-007-1006-9

CITATIONS READS

38 1,636

7 authors, including:

Yong-Sub Yi Yoon Young Kim

Hyundai Motor Company Seoul National University
9 PUBLICATIONS   59 CITATIONS    373 PUBLICATIONS   4,739 CITATIONS   

SEE PROFILE SEE PROFILE

Jae Seok Choi Jeonghoon Yoo

Samsung Yonsei University
27 PUBLICATIONS   483 CITATIONS    138 PUBLICATIONS   1,178 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Design and Application of Meta-Structures in Microwave Ranges View project

Elastic metamaterials and the applications View project

All content following this page was uploaded by Jae Seok Choi on 23 December 2013.

The user has requested enhancement of the downloaded file.

 Journal of
Mechanical
Science and
Technology
Journal of Mechanical Science and Technology 22 (2008) 50~60
www.springerlink.com/content/1738-494x

Dynamic analysis of a linear motion guide having rolling elements for

precision positioning devices
Yong-Sub Yi1, Yoon Young Kim1,*, Jae Seok Choi2, Jeonghoon Yoo2,
Dong Jin Lee3, Suk Won Lee3 and Sung Jin Lee3
1
National Creative Initiatives Center for Multiscale Design, School of Mechanical and Aerospace Engineering,
Seoul National University, San 56-1 Shinlim-Dong, Kwanak-Gu, Seoul 151-742, Korea
2
Department of Mechanical Engineering, Yonsei University,134 Shinchon-Dong, Seodaemun-Gu, Seoul 120-749, Korea
3
Mechatronics Center, Samsung Electronics Co., Ltd.,416 Maetan 3-Dong, Yeongtong-Gu, Suwon-Si,
Gyeonggi-Do 443-742, Korea

(Manuscript Received October 9, 2006; Revised June 15, 2007; Accepted July 12, 2007)
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Abstract

Linear motion (LM) guides supported by rolling elements can be used to ultra-accurately position precision ma-
chines. For accurate positioning, the micro and macro dynamic behavior of the LM guide must be understood, but the
research on this subject is rather limited. In this investigation, experiments to reveal the dynamic characteristics of the
LM guide were performed and a simplified model to predict the observed dynamics of the LM guide was developed.
Several experiments conducted in the present research demonstrated the hysteretic behavior as well as frequency- and
force-dependent phenomena of the LM guide. The validity of the proposed modeling method was checked with theo-
retical and numerical analyses.

Keywords: Linear Motion (LM) guide; Micro-friction model; Contact analysis; Dynamic analysis
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

relatively good reliability, and low cost. However,

1. Introduction
nonlinear dynamic behavior of ball bearings needs to
Linear motion (LM) guides have been widely used be considered for accurate prediction of the motion of
for ultra-precise positioning devices to transport ma- LM guides.
chine parts through a linear path in machining centers The dynamic characteristics of the LM guide can
and X-Y tables etc. With machine parts becoming be largely classified into elastic deformation-induced
smaller and finer, the required order of precision has dynamics and friction-induced dynamics. The fric-
increased. This research presents some experimental tion-induced dynamics will be referred to as the dy-
results on ball bearing-supported LM guides up to namics of the LM guide in the moving direction,
micro-scale level and develops techniques to predict while the elastic deformation-induced dynamics as
the dynamic characteristics of LM guides, such as that in other directions.
their eigenfrequencies. For the friction-induced dynamics, some investiga-
Compared to LM guides supported by non-contact tions have employed micro-scale friction models.
bearings such as air bearings requiring complex and Micro-scale friction models can simulate nonlinear
expensive fabrication processes [1], LM guides using characteristics that cannot be described by classical
ball bearings have advantages such as high stiffness, friction models using the Coulomb friction, the
*
Stribeck effect and the others. Since the dynamic
Corresponding author. Tel.: +82 2 880 7154, Fax.: +82 2 872 5431
E-mail address: yykim@snu.ac.kr
model describing spring-like behavior in stiction state
DOI 10.1007/s12206-007-1006-9 (static friction) proposed by Dahl [2], several modi-
 Y. -S. Yi et al. / Journal of Mechanical Science and Technology 22 (2008) 50~60 51

fied models [3-7] have been proposed. Most of the

previous researches have investigated the rotor-arm
system [8, 9], but only few researches have applied
the micro-friction model to control the linear motor-
driven LM guide system [10]. Though a practical
control issue was considered in their work, the fric-
tion-induced dynamics of the LM guide has not been
fully investigated.
Ohta [11] and Ohta and Hayashi [12] performed
linear vibration analysis for the elastic deformation-
Fig. 1. Schematic of the LM (Linear Motion) guide.
induced dynamics. They carried out modal tests for a
ball bearing-supported LM guide system which
Ch 2,3,4 :Acceleration
Suspension
moved at a constant speed and determined the natural 3-axis
accelerometer wire

frequencies and the corresponding mode shapes in- Vertical

Steel Stinger
drection
cluding rigid-body modes and some flexural modes. block
Horizontal LM rail
Then they carried out simulations analytically and (Motion) direction LM block Shaker
Lateral
numerically. The eigenfrequencies of the LM guide direction

system for five fundamental modes (excluding the B& K

motion related to the degree of freedom in the moving Pulse
system
Ch 1 : Force

direction) were well predicted by their approach. Me- Function generation

(sine sweep )
kid [13] also carried out finite element analysis of the Power Amplifier

hydrostatic bearing-supported LM guide system. The

above-mentioned research has contributed substan- Fig. 2. The experimental setup.
tially to the understanding of the LM guide system
dynamics, but their modeling techniques are not fully Table 1. Experimental results for natural frequencies (Hz).
based on theoretical analyses.
Vertical excita-
Our study consists of two parts: the presentation of Mode No. Axial excitation Lateral excitation
tion
the experimental results for the ball bearing-supported
1st 0~74
LM guide, and the theoretical modeling technique of
the bearing ball stiffness for elastic deformation- 2nd 158 153 149
induced dynamic analysis. For the theoretical model-
ing, we newly developed a technique based on the 3rd 422 416 422
Hertz contact analysis of ball bearing to evaluate
4th 487 497 467
equivalent stiffness and predicted some natural modes
and frequencies by establishing an adequate finite 5th 1190
element model. Also, the development of a complete
6th 1460 1450 1590
theoretical model to predict friction-induced dynam-
ics of the LM guide by the LuGre model [3] was
tried; however, it was not successful. Nevertheless, located to prevent balls from interacting.
the experiments showed the interesting nonlinear The LM guide assembly including the block carry-
phenomena that occurred when the system moved ing the payload is shown in Fig. 2. The steel block
very slowly. was excited by a modal shaker in the three directions:
horizontal, lateral and vertical. The horizontal direc-
2. Dynamic behavior of LM guide tion was the LM guide’s direction of motion. A tri-
axis accelerometer was located at the corners of the
To study the LM guide dynamics experimentally, steel block. Fig. 2 shows the schematic setup of the
modal tests were conducted first. The LM guide used modal testing. The excitation frequency was varied
in this study was manufactured by THK Co. Ltd. It from 0 Hz to 3.2 kHz at the sweeping rate of 5 Hz/sec.
transports an LM block by use of 4 rows of recirculat- The excitation force level was also varied between
ing balls, as shown in Fig. 1. Between balls, cages are 0.1 N and 5 N.
52 Y. -S. Yi et al. / Journal of Mechanical Science and Technology 22 (2008) 50~60

Table 1 shows the eigenfrequencies of the LM Except the first eigenmode (i.e., the horizontally-
guide system, which were extracted from the meas- translational mode), the mechanics behind the every
ured frequency response functions shown in Fig. 3. rigid body modes is somewhat obvious; the elastic
Fig. 4 shows the mode shapes of the steel block that deformations of the bearing balls which lay between
correspond to the eigenfrequencies listed in Table 1. the LM block and the LM rail were responsible for
Four to six eigenfrequencies exist in each excitation the 2nd to 6th vibration modes. At a glance, however,
direction. To characterize the mode shapes, each the mechanism generating the first mode was not
mode shape is referred to as the (horizontally- clear. Its eigenfrequency considerably varied depend-
)translational, rolling, yawing, pitching, or bouncing ing on the experimental conditions, such as the mag-
and high-order rolling mode. Five modes from rolling nitude of the excitation force. Earlier investigations
to high-order rolling modes were also observed in [14, 15] reported the nonlinear behavior related to the
earlier researches [11, 12]. The flexural modes caused first mode, but the physics behind it has not been fully
by the elastic deformation of the LM block or the LM explained.
rail were not present in the frequency range con-
cerned, for they have relatively higher natural fre- 3. Elastic deformation-induced dynamics
quencies than the rigid body modes.
In this section, the theoretical analysis to predict the

(a)
(a) (b)

(b) (c) (d)

(c) (e) (f)

Fig. 3. Typical frequency responses obtained by excitations Fig. 4. Mode shapes: (a) translation, (b) rolling, (c) yawing,
in the (a) horizontal, (b) lateral and (c) vertical directions. (d) pitching, (e) bouncing and (f) high-order rolling.
 Y. -S. Yi et al. / Journal of Mechanical Science and Technology 22 (2008) 50~60 53

2nd up to the 6th eigenfrequencies and eigenmodes, a simplified model based on the elastic foundation
which are associated with the elastic deformation of concept [17] is used in this investigation.
the LM block, LM rail and bearing balls, will be pre- The contact plane between a bearing ball and the
sented. The key procedure in this analysis is the esti- LM rail, which is shown in part A in Fig. 5(a), is re-
mation of the equivalent stiffness and effective mod- placed by the elastic foundation of height, H , which
eling of the LM guide system. For these two tasks, rests on a rigid base, and is compressed by a rigid
contact analysis was carried out. Though some earlier indenter as in Fig. 5(c). The profile of the indenter
investigations used the contact mechanics approach h( x, y ) is taken as the sum of the profiles of the con-
for the LM guide systems, they, however, were tacting two bodies – h1 ( x, y ) of the LM rail and
mainly concerned with fatigue analysis and lifetime h2 ( x, y ) of the bearing ball as shown in Fig. 5(b),
estimate [16].
i.e.,
3.1 Evaluation of equivalent stiffness
h ( x, y ) h1 ( x, y )  h2 ( x, y ) . (1)
The Hertz contact model is most widely used for
contact analysis. However, it needs the solution of If the curved profile of the indenter is assumed to
complicated integral equations. To facilitate analysis, be continuous up to its second derivative in the con-
tact region, the profile may be written as:

1 2 1 2
h ( x, y ) x  y (2)
2 Rc 2 Rcc

where Rc and Rcc are the relative principal radii of

curvature at the origin.
If the penetration by the applied force P at the
origin is denoted by G , then the normal elastic dis-
placement component field of the foundation u z is
given by

(a) ­G  h( x, y ) G ! h
u z ( x, y ) ® (3)
¯ 0 G dh

Since the contact pressure at any point depends

only on the displacement at the point, one can write

K
(b) p ( x, y ) u z ( x, y ) (4)
H

where K is the elastic modulus of the foundation

and will be given later.
Using Eqs. (2)~ (4), the pressure p ( x, y ) inside
the contact area can be written as

KG § x2 y 2 ·
p ( x, y ) ¨1  2  2 ¸ (5)
H © a b ¹
(c)
where a 2G Rc and b 2G Rcc denote the semi-
Fig. 5. (a) Section view of the LM block and the rail, (b)
enlarged diagram of part A and (c) schematic diagram of axes of the ellipsoidal contact area.
the elastic foundation model. The integration of the pressure in Eq. (5) over the
54 Y. -S. Yi et al. / Journal of Mechanical Science and Technology 22 (2008) 50~60
1

contact area yields the total load, which is that the equivalent stiffness varies only with G 2 .
Therefore, the stiffness of the LM guide can be con-
K S abG
P (6) trolled by the preload, which is caused by the nega-
2H
tive clearance of the bearing balls.
K E*
It is well known that 0.60 should be
H r
used for the case of axisymmetric contact [17] where 3.2 Analysis results
r is the contact radius and E * is given by
Eq. (12) is used to compute the stiffness of a linear
1 1  Q 12 1  Q 2 2
 ( E : Young’s modulus, Q : spring in the analytic solution and finite element
E* E1 E2 analysis of the LM guide system, and these analyses
Poisson’s ratio). Thus, one may replace r with the check the validity of our proposition in explaining
1
1
correctly the modal behavior of the system from the
equivalent radius  ab  2  2G Rc 2G Rcc  2
,
2nd to the 6th eigenmode shown in Table 1 and Fig. 4.
1
1
In earlier works, the lowest five natural frequencies
2  RcRcc  4 G 2 and the following load-penetration
were obtained analytically as follows (for details, see
relation is obtained: Appendix).
1
0.60 E * S  RcRcc  4 3 3

P G2 CG 2 . (7)G 1 2 L( K1  K 2 sin 2 E )
2 f bouncing (13)
2S M
By representing the groove radius of the LM rail by 1 L3 ( K1  K 2 sin 2 E )
f pitching (14)
r1 and the ball radius by r2 , the rail profile h1 and 2S 6J y
the ball profile h2 can be written as
1 K 2 L3 cos 2 E
1 2 1 2 1 2 1 2 1 f yawing (15)
h1 x  y  x  y  y 2 (8) 2S 6J Z
2R1c 2R1cc 2uf 2r1 2r1
1 2 1 2 1 2 1 2 ­° f rolling Z1 / 2S
h2 x  y x  y . (9) ® (16)
c
2 R2 2 R2cc 2r2 2r2 °̄ f highrolling Z2 / 2S

Substituting Eqs. (8) and (9) into Eq. (1) yields the In Eq. (16) Z1 and Z2 satisfy the following
profile h of the rigid indenter: equation:
1 2 § 1 1 · 1 2 1 2
h h1  h2 x ¨  ¸ y2 x  y § ª p  MZ2 p2 º·
2r2 2
© 2r 2r1 ¹ 2Rc 2Rcc Det ¨ « 1 0 (17)
¨ 2»¸
©¬ p2 p3  J xZ ¼ ¸¹
(10)
p1 2 K 2 L cos 2 E
Therefore, the relative principal radii of the rigid
p2 K 2 L(2b cos 2 E  c 2 sin 2E )
indenter may be identified as
p3 2 L( K1c12  K 2 b 2 cos 2 E  K 2 c22 sin 2 E  K 2 bc2 sin 2E )
1 r1 r2 (18)
Rc r2 R '' (11)
1 1 r1  r2

r2 r1 In preceding equations K1 and K 2 denote the
equivalent stiffnesses of the upper and lower row of
Differentiating Eq. (7) with respect to G yields
the bearing balls, M the total mass of LM block and
the equivalent stiffness k due to the elastic defor-
steel block, J x , J y and J z the moment of inertia
mations of the bearing ball and the rail:
of the LM block and steel block about the x-, y- and z-
dP 3 § S  0.60 E * 1/ 4 ·
1
3 1 axes, respectively. The x-axis corresponds to the mo-
k ¨  RcRcc  ¸ G 2 { CG 2 (12) tion direction and the y- and z-axes correspond to the
dG 2© 2 ¹ 2
lateral and vertical direction, respectively, and they
Since E * is the material property and Rc and have the origin at the mass center. What L , b , c1 ,
Rcc are geometrical data, one can see from Eq. (12) c2 and E represent is shown in Fig. 6.
 Y. -S. Yi et al. / Journal of Mechanical Science and Technology 22 (2008) 50~60 55

Two analysis models are used for finite element ance between the balls and the rail, each of the spring
calculation. One model treats the LM block and the elements is compressed by an initial displacement of
steel block as a rigid body while the other, as solid 7 μm. This value is the mean value of the negative
brick elements. For the rigid body modeling, a con- tolerance estimate given by the manufacturer, THK
centrated mass is placed at the mass center with an Co. Ltd., which ranges from 4 μm to 10 μm. Though
equivalent moment of inertia, and then a rigid body is it was not possible to measure the exact tolerance
constructed by linking the springs and the mass center value, the validity of using it was checked indirectly
with massless rigid beams. For the solid brick ele- by the experiments conducted by Choi et al. [18].
ment modeling, 8-node brick elements are used, Analysis results for natural frequencies are presented
which were meshed by the Altair HyperMesh soft- in Figs. 7, 8 and Table 2. Analytical and FEM-based
ware. The bearing balls are replaced by equivalent numerical results showed good agreement. Especially,
linear spring elements, and therefore, 14 spring ele- the analytic solution and the FEM results with rigid
ments are located at each row, respectively. For mo- body modeling showed very similar results. Note that
dal analysis, the block Lanczos method of the the sa me concentrated mass and the mo-
ANSYS 7.0 is used. ment of inertia are used and the LM block is treated
To model the prestress effect by the negative toler as a rigid body for both methods. Because the flexural
modes of the LM block and the steel block are much
z
\ higher than the frequency range of concern, that is,
Steel block their stiffness are much higher than those of the bear-
ing balls, treating the blocks as rigid bodies is reason-
able.
I The predicted eigenfrequencies by the analytic and
O y
T numerical solution are somewhat different from the
a
c1
experimental eigenfrequencies for higher modes. This
LM block
D b might be because very simple one-dimensional
#4 #1
equivalent stiffness was used and the nonlinearity
#3 #2
shown in contact analysis was not applied. Consider-
E ing that the modal orders and mode shapes well agree
c2 with the experimental results and the errors for lower
Fig. 6. Geometry of the LM guide system. modes are less than 15%, however, the results appear
quite satisfactory.
1 1 1
DISPLACEMENT DISPLACEMENT DISPLACEMENT
Z Z
DEC 2 2004 DEC 2 2004 DEC 2 2004
STEP=1 STEP=1 STEP=1
Y X 13:42:22 13:43:03 Y X 13:44:40
SUB =2 SUB =3 SUB =4
FREQ=136.303 FREQ=502.508 FREQ=585.314
DMX =.496805 DMX =.485561 DMX =1.104

Y
Z X

(a) (b) (c)

1 1
DISPLACEMENT DISPLACEMENT
DEC 2 2004 Z
STEP=1 DEC 2 2004
Z 13:45:00 STEP=1
SUB =5 Y X 13:46:18
SUB =6
FREQ=1388 Y X FREQ=1890
DMX =.632962 DMX =1.107

(d) (e)
Fig. 7. Mode shapes obtained by FEM analysis using rigid elements: (a) rolling, (b) yawing, (c) pitching, (d) bouncing and (e)
high-order rolling.
56 Y. -S. Yi et al. / Journal of Mechanical Science and Technology 22 (2008) 50~60

1 1 1
DISPLACEMENT DISPLACEMENT DISPLACEMENT
STEP=1 DEC 2 2004 STEP=1 DEC 2 2004 STEP=1 DEC 2 2004
SUB =2 14:11:36 SUB =3 14:12:50 SUB =4 X Y 14:13:45
Y X
FREQ=132.981 FREQ=467.9 FREQ=561.943
Z
DMX =1.085 Z DMX =1.092 DMX =1.295

Y
X Z

(a) (b) (c)

1 1
DISPLACEMENT DISPLACEMENT
Y X DEC 2 2004 DEC 2 2004
STEP=1 STEP=1
SUB =5 14:14:18 SUB =6 Y X 14:14:42
Z
FREQ=1324 FREQ=1728
DMX =.657468 DMX =1.155 Z

(d) (e)
Fig. 8. Mode shapes obtained by FEM analysis using solid elements: (a) rolling, (b) yawing, (c) pitching, (d) bouncing and (e)
high-order rolling.

Table 2. Analysis results by the Euler-Lagrange equation and applied force is apparently smaller than the maximum
the finite element method (FEM I: rigid modeling, FEM II: static friction force, quite significant hysteretic mo-
solid modeling).
tions are observed. From the viewpoint of classical
Mode shape Experiment Theory FEM I FEM II friction, there should be no motion, that is, there
should be no stiffness for the motion.
Rolling 153 137.3 136.3 133.0
The mechanics behind these phenomena is not fully
Yawing 422 468.4 502.5 467.9 understood, but the micro-scale elastic and plastic
Pitching 487 545.3 585.3 561.9 deformations of the balls and the contacting regions
Bouncing 1190 1389 1388 1324 caused by friction are likely to be responsible for the
hysteretic force-displacement relation. Earlier re-
High rolling 1450 1889 1890 1728
searches reported similar phenomena between two
sliding plates or in the rotor-bearing system. However,
4. Friction-induced dynamics these researches considered quasi-static characteris-
tics more important than dynamic characteristics.
4.1 Hysteresis test
The mean slope of the hysteresis loop in Figs. 9(a)-
To understand the type of nonlinearity involved in (c) could be regarded as the linear stiffness. As the
the first eigenmode, a number of experiments were excitation force increases, the mean slope decreases
conducted in which the excitation force was varied and the inside area of the loop increases. This means
from 0.2 N to 4.2 N. The steel block was excited in that the apparent system stiffness decreases and the
the horizontal direction and the displacement of a dissipated energy increases as the force increases.
point in the block was measured by the laser Doppler This dependency on the excitation force level causes
vibrometer (LDV). In the present experiments, a very nonlinearity such as variable natural frequency for the
low excitation frequency (under 1 Hz) was selected to first translational mode.
minimize the inertial effect of the block and effec- The effect of the excitation frequency is shown in
tively show the static relationship between the excita- Fig. 10. The excitation force level is fixed, and then
tion. the system response is observed for various excitation
These experiments gave the following findings. frequencies. When the excitation frequency is below
When the excitation force exceeded approximately 20 Hz, the overall loop pattern is similar to that ob-
4.2 N, which was the maximum static friction force, tained in the earlier results shown in Figs. 9(a)-(c). If
the balls rolled, as shown in Fig. 9(d). Even when the we presume that the equivalent stiffness of the system
 Y. -S. Yi et al. / Journal of Mechanical Science and Technology 22 (2008) 50~60 57

(a) (b)

(c) (d)
Fig. 9. Force-displacement relation for various force levels (a), (b), (c) before rolling and (d) after rolling.
of-freedom mass-spring-damper system exhibits simi-
lar behavior, one may attempt to explain the phenom-
ena by an equivalent one-degree-of-freedom model
[14, 15].

4.2 Micro-friction model

In our investigation, the simulation of the phenom-

ena observed in Figs. 9 and 10 is tried through the
LuGre model [3], which is a popular modern friction
model. Even though other models such as the Leuven
model [4, 7] and the elastoplastic model [5] are avail-
able, they are equivalent to the LuGre model if the
Fig. 10. Force-displacement relation curve at various excita-
tion frequencies. reverse point is fixed. In the LuGre model, pre-rolling
displacement occurs due to the deformation of a large
as the average slope of the quasi-static experimental number of bristles at the contact surface. So, the mo-
results at low frequency and predict the natural fre- tion is described by average bristle displacement as
quency of the system by using a simple one-degree- follows:
of-freedom model, we can obtain 41.5 Hz from
F V 0 z  V 1 z  V 2 x (19)
1 k
f for system mass m=2.50 kg. This fre- x
2S m z x  z f ( x , z ) (20)
quency is very close to the experimental result of 38 g ( x )
Hz. When the excitation frequency approaches to the
resonance frequency, the mean slope becomes almost The symbols used in Eqs. (19) and (20) represent:
zero. For higher-frequency excitations, the sign of the F : is the friction force;
mean slope becomes negative. Since the one-degree- z : the average bristle deflection;
58 Y. -S. Yi et al. / Journal of Mechanical Science and Technology 22 (2008) 50~60

results from micro-scale friction in slowly moving

LM guide systems, was investigated. The nonlinearity
of the first translation mode was identified to result
from the hysteresis. The force-displacement relation
observed on the contact surface of the bearing balls
and the rail of LM guide was derived by using the
elastic foundation contact model, and the equivalent
stiffness was also obtained from contact analysis.
This relation was used for the dynamic analysis of the
LM guide system except for the first natural mode, a
translational mode. The dynamic analysis was con-
ducted analytically and numerically. For the analyti-
Fig. 11. Simulation results by using the LuGre model. cal solution, the Lagrange equation was used, and for
the numerical solution, the finite element method
x : the relative displacement between the two con- modeling the LM block and the steel block as a rigid
tacting surfaces; body and the bearing balls as linear spring elements
g ( x ) : the Stribeck curve for steady-state velocities; was used. The analytical and numerical solutions
V 0 : the bristle stiffness; agreed fairly well with the experimental results within
V 1 : the bristle damping; about 15% error. To simulate the first eigenmode
V 2 : the viscous damping coefficient. related to the hysteresis phenomenon, a micro-scale
The values of V 0 , V 1 and V 2 cannot be known friction model based on the LuGre theory was
before an experiment is carried out. Therefore, their adopted. Although the LuGre model was capable of
values were determined by the least-square method capturing the main physical phenomena, it should be
[19] which was applied to the test results obtained at improved to accurately describe the experimentally
the excitation force of 0.5 N. Then other cases were observed phenomena. The contact analysis based on
simulated by using Eq. (19) and (20). As may be seen the dynamic model developed in this investigation
in Fig. 11, the model coefficients matching experi- can be a useful initial tool for the LM guide system
mental results obtained for a specific condition (force design. However, the ultra-precise control of the sys-
level and frequency) may not simulate the experimen- tem requires an advanced hysteresis analysis model,
tal results obtained for other conditions. This is be- which can be a good research subject in the future.
cause the LuGre model has some defect in which the
stiffness at the reverse point is invariable regardless of Acknowledgments
the position or velocity of the block [8]. Though some
reports [3-7] have stated that the trajectory of the This study was supported by the Mechatronics
block is always the same from the point at which the Center of SAMSUNG Electronics.
direction of motion reverses, our experiments show
that each trajectory of the block moves along different References
paths as the excitation force changes. Perhaps, an [1] P. Nuij and M. Steinbuch, Two measurement tech-
improved model considering up to the physics not niques to determine higher order sinusoidal input
covered by the LuGre model such as rigid body roll- describing functions, Proceedings of ISMA 2004,
ing should be developed, and the development of Leuven, Belgium, (2004) 2145-2154.
such a model will be pursued in future investigations. [2] P. Dahl, A solid friction model, Aerospace Corp.,
El Segundo, CA, Tech. Rep. TOR-0158 (3107-18)-
1 (1968).
5. Conclusion
[3] C. Canudas de Wit, H. Olsson, K. J. Åström and P.
We have investigated the dynamic properties of the Lischinsky, A new model for control of systems
LM guide system, both experimentally and theoreti- with friction, IEEE Transactions on Automatic
cally. The natural frequencies of the system and the Control 40 (3) (1995) 419-425.
corresponding mode shapes have been observed [4] J. Swevers, F. Al-Bender, C. G. Ganseman and T.
through modal tests. In addition, hysteresis, which Prajogo, An integrated friction model structure with
 Y. -S. Yi et al. / Journal of Mechanical Science and Technology 22 (2008) 50~60 59

improved presliding behavior for accurate friction and S. J. Lee, Vibration analysis and its application
compensation, IEEE Transactions on Automatic of a linear motion guide supported by rolling ball
Control 45 (4) (2000) 675-686. bearings, Transactions of KSME A 29 (7) (2005)
[5] P. Dupont, V. Hayward, B. Armstrong and F. Alt- 955-963.
peter, Single state elastoplastic friction models, [19] C. Canudas de Wit and P. Lischinsky, Adaptive
IEEE Transactions on Automatic Control 47 (5) friction compensation with partially known dy-
(2000) 787-792. namic friction model, International Journal of
[6] C. Hsieh and Y. -C. Pan, Dynamic behavior and Adaptive Control and Signal Processing 11 (1997)
modeling of the pre-sliding static friction, Wear 242 65-80.
(2000) 1-17.
[7] V. Lampaert, J. Swevers and F. Al-bender, Modifi-
Appendix
cation of the Leuven integrated friction model struc-
ture, IEEE Transactions on Automatic Control 47 If the mass of the bearing ball is so small that it
(4) (2002) 683-687S. cannot affect the system behavior, the total contact
[8] R. H. A. Hensen, M. J. G. van de Molengraft and M. system can be easily expressed as a one degree-of-
Steinbuch, Frequency domain identification of dy- freedom oscillator. From Eq. (7), the new relationship
namic friction model parameters, Transactions on between the contact load P and the displacement
Control Systems Technology 10 (2) (2002) 191-196. q can be derived as
[9] T. -Y. Lin, Y. -C. Pan, and C. Hsieh, Precision-limit
1
positioning of direct drive systems with the exis- C 3/ 2 0.60 E * S RcRcc4 (A1)
tence friction, Control Engineering Practice 11 P q , C ,q 2G
2 2 2
(2003) 233-244.
[10] J. -S. Chen, K. -C. Chen, Z. -C. Lai and Y. -K. Then the linear spring stiffness k can be obtained
Huang, Friction characterization and compensation by differentiating Eq. (A1) for q.
of a linear-motor rolling-guide stage, International If the LM block moves in the vertical direction by
Journal of Machine Tools & Manufacture 43 dz, the spring of the upper surface is elongated by dz
(2003) 905-915. and that of lower surface shrinks by dz sin E as
[11] H. Ohta, Sound of linear guideway type recirculat- shown in Fig. A1. The LM system is symmetric about
ing linear ball bearings, Transactions of the ASME, the vertical direction, so the motion of equation in the
Journal of Tribology 121 (1999) 678-685. vertical direction can be derived as follows.
[12] H. Ohta and E. Hayashi, Vibration of linear
guideway type recirculating linear ball bearings, 3/ 2 3/ 2
Mz 2Ceq  z1  dz   2Ceq sin E  z2  dz sin E 
Journal of Sound and Vibration 235 (5) (2000) 847-
861. (A2)
[13] S. Mekid, High precision linear slide. Part I: de- C
where Ceq is a constant which multiplies by
sign and construction, International Journal of Ma- 2 2
chine Tools & Manufacture 40 (2000) 1039-1050. the number of balls in one row, and z1 denotes the
[14] Futami, A. Furutani, and S. Yoshida, Nanometer initial displacement of bearing balls of the upper rows
positioning and its micro-dynamics, Nanotechnol-
ogy 1 (1990) 31-37.
LM z
[15] J. Otsuka and T. Masuda, The influence of nonlin- block
ear spring behavior of rolling elements on ultrapre- ȕ
y
cision positioning control systems, Nanotechnology
9 (1998) 85-92. rail
[16] S. Shimizu, E. Saito and H. Uchida, Tribological
dz
studies of linear motion ball guide systems, Tribol-
ogy Transactions 41 (1) (1998) 49-59.
dzsinȕ
[17] K. L. Johnson, Contact Mechanics, Cambridge
University Press, Cambridge, UK, (1985). Fig. A1. Relationship between the displacement of the upper
[18] J. S. Choi, J. Yoo, Y. -S. Yi, Y. Y. Kim, D. J. Lee row and the lower row.
 60 Y. -S. Yi et al. / Journal of Mechanical Science and Technology 22 (2008) 50~60

caused by the minus tolerance and z2 the initial J xI  2L(K1c12  K2b2 cos2 E  K2c22 sin2 E  K2bc2 sin2E )I
displacement of lower rows. If the system is assumed  L(2K2b cos2 E  K2c2 sin2E ) y 0
to reach static equilibrium at dz 0 , the following (A9)
relation between the initial displacement of balls in
KL3
the upper and lower rows can be obtained: J yT  (1  sin 2 E )T 0 (A10)
6
§ KL3 · (A11)
§ z1 ·
3/ 2
J z\  ¨¨ cos 2 E ¸¸\ 0
¨ ¸ sin E (A3) © 6 ¹
© z2 ¹
Eqs. (A7)-(A11) represent the equations of motion
For simpler analysis, the arrays of the balls can be or high-order rolling, bouncing, rolling, pitching and
assumed to be continuous. Then, the spring constant yawing mode, respectively. From Eqs. (A8), (A10)
K per unit length can be written as and (A11), the natural frequencies are easily obtained
because z , T and \ are uncoupled.
Nk (A4)
K
L
1 2 L( K1  K 2 sin 2 E )
f bouncing (A12)
2S M
where N and L represent the average number of balls
in the load zone and the length of load zone on the 1 L3 ( K1  K 2 sin 2 E )
f pitching (A13)
circuit of the recirculating balls, respectively. 2S 6J y
Using the notation used in section 3.2, the kinetic
1 K 2 L3 cos 2 E
energy of the 5 degree-of-freedom system becomes f yawing (A14)
2S 6J Z
1 1 1 1 1
T My 2  Mz 2  J xI 2  J yT 2  J z\ 2 (A5)
2 2 2 2 2 Because y and I are coupled, the natural fre-
quencies for rolling and high order-rolling modes can
The potential energy by the elastic deformation of be obtained by solving the following equations:
the ball bearings can be formulated as the integration
for loading length. y Ye jZ t , I ) e jZ t
(A15)
ª p1  MZ 2 p2 º ­Y ½ ­0½
« 2 »® ¾ ® ¾
L/2
U
1
2  L³/ 2
>
K 1 ^ y  aI  l\  cos D   z  c1I  lT sin D `
2
(A6) ¬ p2 p3  J xZ ¼ ¯) ¿ ¯0¿
 K 2 ^ y  bI  l\  cos E   z  c 2I  lT sin E `
2

 K 2 ^  y  bI  l\  cos E   z  c 2I  lT sin E `
2 where
 K 1 ^  y  aI  l\  cos D   z  c1I  lT sin D ` dl
2
@ p1 2K2 L cos2 E
p2 K2 L(2b cos2 E  c2 sin2E )
where K1 denotes the equivalent stiffness of the p3 2L(K1c12  K2b2 cos2 E  K2c22 sin2 E  K2bc2 sin2E )
upper rows and K 2 the stiffness of the lower rows. (A16)
Using T and U , the following five equations of
motion can be derived from the Euler-Lagrange equa- To obtain a non-trivial solution, the determinant of
tions: the square matrix in Eq. (A15) must be zero. The
results are
My  (2K1L cos2 E ) y  K2 L(2b cos2 E  c2 sin 2E )I 0
(A7) ­° f rolling Z1 / 2S (A17)
Mz  2 L( K 1  K 2 sin 2 E ) z 0 (A8) ®
°̄ f high rolling Z 2 / 2S

View publication stats