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

net/publication/222011472

Milling machine spindle analysis using FEM and non-contact spindle

excitation and response measurement

Article in International Journal of Machine Tools and Manufacture · June 2007

DOI: 10.1016/j.ijmachtools.2006.10.004

CITATIONS READS

126 837

4 authors, including:

Matti Rantatalo Jan-Olov Aidanpää

Luleå University of Technology Luleå University of Technology
37 PUBLICATIONS 927 CITATIONS 85 PUBLICATIONS 1,469 CITATIONS

SEE PROFILE SEE PROFILE

Peter Mats Norman

Svetsdoktorerna
30 PUBLICATIONS 447 CITATIONS

SEE PROFILE

All content following this page was uploaded by Peter Mats Norman on 13 March 2022.

The user has requested enhancement of the downloaded file.

 ARTICLE IN PRESS

International Journal of Machine Tools & Manufacture 47 (2007) 1034–1045

www.elsevier.com/locate/ijmactool

Milling machine spindle analysis using FEM and non-contact spindle

excitation and response measurement
Matti Rantataloa,, Jan-Olov Aidanpääb, Bo Göranssonc, Peter Normand
a
Division of Sound and Vibrations, Luleå University of Technology, SE-971 87 Luleå, Sweden
b
Division of Computer Aided Design, Luleå University of Technology, SE-971 87 Luleå, Sweden
c
SKF Nova, HK1-6, 415 50 Göteborg, Sweden
d
Division of Manufacturing Systems Engineering, Luleå University of Technology, SE-971 87 Luleå, Sweden

Received 16 July 2006; received in revised form 2 October 2006; accepted 5 October 2006
Available online 8 December 2006

Abstract

In this paper a method for analysing lateral vibrations in a milling machine spindle is presented including finite-element modelling
(FEM), magnetic excitation and inductive displacement measurements of the spindle response. The measurements can be conducted
repeatedly without compromising safety procedures regarding human interaction with rotating high speed spindles. The measurements
were analysed and compared with the FEM simulations which incorporated a spindle speed sensitive bearing stiffness, a separate mass
and stiffness radius and a stiffness radius sensitive shear deformation factor. The effect of the gyroscopic moment and the speed
dependent bearing stiffness on the system dynamics were studied for different spindle speeds. Simulated mode shapes were
experimentally verified by a scanning laser doppler vibrometer. With increased spindle speed, a substantial change of the eigenfrequencies
of the bearing-related eigenmodes was detected both in the simulations and in the measurements. The centrifugal force that acted on the
bearing balls resulted in a softening of the bearing stiffness. This softening was shown to be more influential on the system dynamics than
the gyroscopic moment of the rotor. The study performed indicates that predictions of high speed milling stability based on 0 rpm tap-
test can be inadequate.
r 2006 Elsevier Ltd. All rights reserved.

Keywords: Machine tool spindle; Centrifugal; Gyroscopic; Non-contact measurement; Angular contact ball bearings

1. Introduction predict a specific chatter free depth of cut which is

governed by the transfer function of the tool tip (assuming
Turning operations such as milling are common in the a rigid work piece). The chatter free depth of cut is
automotive and aerospace industries where large metal calculated for different spindle speeds which are then
work pieces are reduced to a fraction of their original plotted as a stability lobe chart. The transfer function is
weights to create complex thin structures. It is important normally measured manually by tap tests of a non-rotating
that unwanted behaviours like tool vibrations be avoided spindle/holder/tool system. The tool tip is excited by an
during these operations. Especially important to avoid is impulse hammer and the response is measured by a
self-excited machine tool chatter caused by the waviness of vibration transducer. The assumption used in this proce-
the machined surface. This type of vibration results in poor dure is that the dynamics of a spindle/holder/tool system is
surface finishes and in some cases material or machine independent of spin speed. This is, however, not true
damage. The phenomena is a significant issue and has been for high speed milling operations where the effect of
addressed and modelled by numerous authors over the past gyroscopic moments and centrifugal forces must be
decades, e.g. [1–4]. The models developed by these authors considered [5–7].
To analyse the spindle speed dependency the machine
Corresponding author. Tel.: +46 920 49 21 24; fax: +46 920 4910 30. tool must be analysed in rotating states that span the whole
E-mail address: matti.rantatalo@ltu.se (M. Rantatalo). range of operating spindle speeds. Schmitz et al. [8]

0890-6955/$ - see front matter r 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijmachtools.2006.10.004
 ARTICLE IN PRESS
M. Rantatalo et al. / International Journal of Machine Tools & Manufacture 47 (2007) 1034–1045 1035

Nomenclature [MT] element translational mass matrix

{q} generalised coordinates
I area moment of inertia of a hollow shaft rm inner mass radius
Cxm calibration factor CDST rk inner stiffness radius
Fc centrifugal force xm, ym instantaneous displacement
Ixt, Iyt coil current quadrant v1 and w1 xm0, ym0 magnetic centre offset
Ixb, Iyb coil current quadrant v3 and w3 Fxm, Fym magnetic force
ai contact angle inner ring E modulus of elasticity
ae contact angle outer ring Rm outer mass radius
Qi contact force inner ring Rk outer stiffness radius
Qe contact force outer ring v poisson’s number
Jd diametral moment of inertia Jp polar moment of inertia
d effective gap between rotor and magnets Ax reference acceleration
[G] element gyroscopic mass matrix F shear deformation
l element length k shear deformation factor
m element mass per unit length Gs shear modulus
[MR] element rotational mass matrix {h} state vector
[K] element stiffness matrix

presented an experimental method for the prediction of shaft with increasing spindle speed. It was shown that
stable cutting regions which reflects the dynamic change the softening of the bearing radial stiffness due to speed
that a rotating system undergoes. The method uses impulse could be compensated for when there was a thermally
hammer excitation and capacitive probe response measure- induced preload. Cao and Altintas [7] presented a general
ment of a tool rotating during different spindle speeds. method for the modelling of a spindle bearing system
Stability lobes for a discrete number of spindle speeds are which included the axial coordinates and a corresponding
then calculated and the limit of stable cut corresponding to spindle speed and a preload-dependent five degrees
the actual spindle speed used is identified and then used to of freedom bearing stiffness matrix. In the spindle model,
draw up a spindle speed dependent stability lobe chart. a rotor related centrifugal force was modelled by subtract-
Experimental tests revealed a changing stable limit for ing a term (O2 multiplied by a radial version of the
stable cut above 16 000 rpm due to changing spindle translational mass matrix) from the stiffness matrix.
dynamics. Simulations for different spindle speeds were performed
An alternative method to analyse a spindle bearing but only verified for a non-rotating spindle. Simulations
system is through modelling. Wang and Chang [9] describe predicted that the centrifugal force of the rotor would
a spindle modelling method based on finite element model influence the eigen-frequency more than the gyroscopic
(FEM). The model, however, did not include rotation and moment of the rotor. This result was not verified
therefore no centrifugal forces and gyroscopic moments. In experimentally.
1976 Nelson and McVaugh [10] presented a FEM This paper describes a method for analysing lateral
formulation of a rotor bearing system based on the Euler machine tool spindle vibrations based on a FEM and a
Bernoulli beam theory where the effect of gyroscopic contact-less dynamic spindle testing equipment (CDST).
moments and centrifugal forces was included. Zorzi and The aim of this work was to study the effect of the
Nelson [11] later on added internal damping and in 1980 gyroscopic moment and the speed dependent bearing
Nelson [12] presented another formulation based on the stiffness on the system dynamics. The study was performed
Timoshenko beam theory which included the shear on a 5-axis Liechti Turbomill ST1200 equipped with a
deformation effects. Xiong et al. [6] presented a way Fischer spindle (MFWS-2305/24/8) with an integrated
of combining this FEM representation and the milling motor spindle capable of speeds up to 24 000 rpm
cutting force model formulated by Altintas [2]. The model, supported by two 251 angular contact ball bearings. The
which only consisted of the rotor, predicted that the spindle spring preload was achieved experimentally and a
gyroscopic moment would not affect the stability regions in speed dependent bearing stiffness was calculated. The
milling but change the real part of the eigenvalues. The FEM elements were based on separate stiffness and mass
model also predicted a change of the spindle resonance radius. Simulations were performed with and without the
frequencies of about 710 Hz. Lin et al. [13] integrated speed sensitive bearing stiffness together with a stiffness
a thermo-mechanical model into the Timoshenko FEM radius sensitive shear deformation factor. Tap tests with
description. Numerical and practical experiments found accelerometers and a scanning laser Doppler vibrometer
an increase in bearing stiffness with increasing bearing (LDV) were used to verify the CDST measurements and
preload. The work also predicted a softening of the spindle the FEM simulation.
 ARTICLE IN PRESS
1036 M. Rantatalo et al. / International Journal of Machine Tools & Manufacture 47 (2007) 1034–1045

2. CDST measurements (e.g., x-directions v1 and v3) on opposite sides of the rotor
were working together pulling and releasing the rotor to
The CDST measured the frequency response function form the excitation.
(FRF) of the tool tip by exciting the rotor with electro- The force Fxm applied to the rotor in the x-direction
magnets while inductive displacements sensors registered (analogous in the y-direction) is expressed in terms of the
the rotor position in the x- and y-directions. The use of stator coil current and the instantaneous air gap as [17]
electromagnets and non-contact displacement sensors is " 2  2 #
common in the field of active magnetic bearings (AMB). I xt I xb
F xm ¼ C xm  ,
An AMB uses a control system to keep the rotor in place ðd  ðxm0 þ xm ÞÞ ðd þ ðxm0 þ xm ÞÞ
which adjusts the coil current in the magnets according to
(1)
the measured rotor position and a desired location. In this
machine tool speciality AMBs are mainly used as rotor where Cxm is a calibration factor (Cxm ¼ 1 in this case), Ixt
support bearings; other applications have been developed. is the measured current of top quadrant (v1), Ixb the
Auchet et al. [14] developed a method for indirect cutting measured current of bottom quadrant (v3), d the effective
force measurement by analysing the command voltage of a gap of 150 mm between the magnets and the rotor, xm0 the
magnetic bearing in a milling machine spindle supported by magnetic centre offset and xm the instantaneous displace-
AMBs. Knospe [15] investigated the potential of active ment measured by the displacement sensors. The coil
chatter suppression by the use of AMB. Chen and Knospe current Ixt and Ixb are a superposition of the excitation
[16] developed an approach to estimate the cutting current, a bias current of 0.8 A and a compensation current
dynamics by both exciting the system and increasing the for static loads; which was set to zero in this case. For
damping of the lathe tool using an AMB. Fig. 1 shows a example, when measuring the y-direction at 0 rpm; a force
photo and a sketch of the experimental set-up used in the Fym with an amplitude of approximately 714 N was
study. The spindle/holder/tool system dynamic was mea- generated by coil currents (Iyt, Iyb) oscillating between 0.4
sured at the tool tip in the x- and y-directions separately by and 1.2 A. The amplitude of the displacement ym at 712 Hz
the use of a CDST. The excitation of the rotor was carried (second resonance) was approximately 720 mm.
out by electromagnets which were fed by frequency step The instantaneous displacement of the rotor in the x- and
vice sine sweep coil current—thereby introducing a y-directions was measured by two displacement sensors; each
magnetic force Fm(t) which acted on the rotor. The rotor which were coupled together for each coordinate with
consisted of a specially manufactured dummy tool with a opposite signs facing the rotor 1801 apart. This arrangement
laminated rotor part to reduce the energy losses due to enabled the cancellation of any changes in displacement due
eddy current effects. In each direction two electromagnets to thermal expansion of the rotor diameter.

-v1 +v1

+z -w3 +w1
+y
DT
+w3 -w1

Spindle Tool +v3 -v3

Tool (C)
& housing Ax(t) holder +x
holder

+z Dummy
Dummy
tool
tool

+y +x -w3 CDST

Iy(t) -v3
+w3
+v3
ym(t) +dy
+dx Ix(t)
xm(t)

CDST +y
+x
(A) (B)

Fig. 1. (A) Photo of the setup with the dummy tool in an elevated state. (B) See-through sketch of the CDST with dummy tool in place. +dx and +dy
denotes the displacement sensors. (C) Electromagnet setup with dummy tool (DT).
 ARTICLE IN PRESS
M. Rantatalo et al. / International Journal of Machine Tools & Manufacture 47 (2007) 1034–1045 1037

-12 -12
CDST

Magnitude [log(m/N)]
Tap test
-13 -13

-14 -14

-15 -15

-16 -16
500 1000 1500 2000 500 1000 1500 2000

2 Mode 1 2
Phase [m/N]

Mode 3
0 Mode 2 0
Mode 4

-2 -2

500 1000 1500 2000 500 1000 1500 2000

Frequency [Hz] Frequency [Hz]

Fig. 2. Tap test of an elevated dummy tool compared to CDST measurements. Black: CDST, gray: tap test.

The displacement response and the exciting force for each parts governed the mass together with an assumed physical
frequency component in the interval 400–2000 Hz were inner mass radius rm. The assumed inner mass radius rm
measured and a spectrum estimation (H1) of the transfer inside the spindle shaft models the spring package and the
function was calculated. This procedure was repeated for all drawbar inside the spindle used for connecting the tool holder
speeds in the interval [0:2000:24 000] rpm and for each radial to the spindle. The connection between the spindle, holder
direction x and y. A reference accelerometer was mounted and the dummy tool was not specially modelled. Damping,
on the CDST housing to ensure that the assumption of a gravity and the centrifugal effects of the rotor described by [7]
rigid CDST construction and a firm machine table mounting were not included. The axial load during free run was
were present. The CDST were verified experimentally by tap considered neglectable and therefore excluded in the study.
tests of the mounted dummy tool at 0 rpm. Except for the The radial displacement and rotation along the radial
third mode in the y-direction, the two different measurement coordinates of a single element is expressed in the general-
methods resulted in similar FRFs (see Fig. 2). The mode ised coordinates q ¼ fxi ; yi ; fx;i ; fy;i ; xiþ1 ; yiþ1 ; fx;iþ1 ; fy;iþ1 g
shapes for 0 rpm were analysed by a scanning LDV which where i is the node number. The area moment of inertia
performed a line scan along the z-axis of the visible part of which is used in the forming of the stiffness matrix was
the dummy tool, holder and rotor. based on the stiffness radii I ¼ ðR4k  r4k Þp=4. The polar
moment of inertia J p ¼ mðR2m  r2m Þ=2 and the diametral
3. Spindle modelling moment of inertia J d ¼ mðR2m  r2m Þ=4 were based on the
mass radius. The shear deformation was expressed as
A FEM described by [12] is used to simulate the mode F ¼ ð12EIÞ=ðkG s l 2 Þ, where E is the modulus of elasticity,
shapes and the eigen-frequencies of the rotor bearing system. Gs the shear modulus and k the shear deformation factor.
The simulations are used to analyse and identify any detected The shear deformation factor is normally determined
gyroscopic or centrifugally induced speed dependency in the experimentally and for a solid circular shape a usual value
frequency response measured by the CDST. The simulation is is approximately 0.9. However, the spindle studied in this
focused on the gyroscopic moment of the rotor compared to paper consists of a hollow circular shaft with a variable
the centrifugal force of the balls in the ball bearings. Fig. 3 stiffness radius. In 2001 Hutchinson [18] proposed an
illustrates the FEM and the element division of the rotor analytical expression for the shear deformation coefficients
bearing system plotted along the x–z plane. Each FEM of a hollow circular shaft expressed as
element of length l consists of two parts with an inner and  2
outer radius (see Fig. 4). The hollow shaft of the rotor with its 6 R2k þ r2k ð1 þ uÞ2
k¼ , (2)
inner stiffness radius rk and the outer stiffness radius Rk ð7r4k þ 34r2k R2k þ 7R4k þ uð12r4k þ 48r2k R2k
governed the stiffness properties of the spindle. The physical þ12R4k Þ þ u2 ð4r4k þ 16r2k R2k þ 4R4k ÞÞ
outer mass radius Rm which included the shaft, motor
package, inner rings of the bearings and other additional which has been used in this work.
 ARTICLE IN PRESS
1038 M. Rantatalo et al. / International Journal of Machine Tools & Manufacture 47 (2007) 1034–1045

Bearing position FEM node Stiffness & mass Mass stiffness were added into the complete rotor stiffness
0.06 matrixes at corresponding nodal coordinates marked by
0.04
triangles in Fig. 3.
0.02
3.1. Preload measurement
[m]

-0.02 To calculate the bearing stiffness the preload force of the

bearings has to be known. However, when dealing with a
-0.04
real milling machine spindle the bearing preload is
-0.06 normally not a known parameter and must therefore be
0 0.1 0.2 0.3 0.4 0.5 0.6
[m]
measured. On-line measurements of the bearing load would
be preferable and a method for this has been presented by
Fig. 3. Model and FEM element division of milling machine spindle. Chen and Chen [20]. For most spindles in operation, this is
not the case and the preload must therefore be measured or
provided by the spindle designer. The spindle used in this
study was designed using spring-loaded bearings as shown
φy,i φy,i+1
in Fig. 5. The front and the back bearings are hybrid angle-
contact bearings placed back to back. The preload was
measured by pulling the spindle towards z while the axial
φx,i+1 displacement was measured using a dial indicator with a
resolution of 1 mm and a total measuring range of 750 mm.
φx,i
Rm The force used to pull the spindle was measured by a static
force sensor with a range 0–5 kN with a resolution of about
20 N. The static force sensor was made out of strain gauges
and connected to a strain gauge amplifier and an
y rm l
oscilloscope. Data was collected manually according to
x z
predefined steps. The estimated spring preload was equal to
the force needed to unload the front bearing. The force
Rk rk
needed to unload the front bearing was estimated to be
1450 N.

Fig. 4. FEM element. 3.2. Bearing stiffness calculations

Using the result from the preload measurement, the

The homogenous equation of motion for the finite bearing stiffnesses within the speed intervals
element assembly used in this model is expressed as [0:2000:24 000] rpm were calculated. The bearing stiffness
      calculations were performed by SKF and an in-house
½M  q€ þ O½G q_ þ ½K  q ¼ f0g, (3)
developed software ‘‘Bearing Beacon’’ based on the theory
where [M], [G], and [K] are the system matrixes of a shaft described in [21] by de Mul et al.
element (see Appendix A). The assembled second order
homogenous differential equation was transformed into a 4. Results
first order differential equation, using the state vector
notation described in [19]. The equation of motion could 4.1. Bearing stiffness
then be rewritten as
" # " # With increasing speed the load conditions between the
O½G ½M   ½K ½0
h_ þ fvg ¼ f0g, (4) balls and the rings in the bearing changes because of the
½M ½0 ½0 ½M

where
Motor Preload spring
( )
fqg Front bearing Housing Rotor
fhg ¼ (5)
_
fqg Z+
F
is the state vector. Solving the obtained first order
homogenous differential equation gives complex eigenvec-
tors with corresponding complex eigenvalues of the Dial
indicator Front bearing Back bearing
displacement and the velocity of each node and its
generalized coordinates. The front and the back bearing Fig. 5. Principal spindle drawing and preload measurement setup.
 ARTICLE IN PRESS
M. Rantatalo et al. / International Journal of Machine Tools & Manufacture 47 (2007) 1034–1045 1039

centrifugal force Fc which acts on each ball (see Fig. 6). The Bearing stiffness 0-24000 rpm (Preload 1450 N)
110
centrifugal force induced by the rotating balls forces the Fx/x
Back bearing
rings to separate axially. A new equilibrium state is reached 100
Mzx/zx
Fy/yz
where the contact forces Qi and Qe balance the new force

Percentage of stiffness at 0 rpm [%]

Myz/y
condition which includes both external forces and the 90
centrifugal force Fc. When the ball reaches this new
asymmetric position it will act like a spring when the 80
external force changes. This spring is in serial with the
70 Front bearing
normal Hertz contact springs resulting in a decreased
bearing stiffness for high spindle speeds. Table 1 shows the 60
stiffness matrix of the back bearing and front bearing
calculated for 0 rpm. 50
The bearing stiffness for different spindle speeds related
to the stiffness at 0 rpm are plotted in Fig. 7. The plot 40
shows that the stiffness in the radial direction of the back
bearing decreases to a level of 62% of its original (0 rpm) 30
0 0.5 1 1.5 2
value when the speed increases to 24 000 rpm. The Spindle speed [rpm] x 104
corresponding value of the front bearing is 38%. The
angular stiffness shows approximately the same amount of Fig. 7. Changes in bearing stiffness with increasing spindle speed, 100%
represents the bearing stiffness at 0 rpm. The corresponding radial,
reduction for the same speed range. The reduction of the
angular and coupled stiffness in the other directions (see not marked cells
front bearing stiffness tends to diminish more than the in Table 1) shows an equal behaviour and are therefore not presented in
back bearing above 18 000 rpm, giving it a characteristic the plot.
shape (hereinafter denoted ‘‘s shaped’’).

4.2. Gyroscopic and centrifugal effects

Inner ring
z
Qi In a rotor bearing system the modes often appears as a
mix of rotor related flexural modes and rigid-body modes
αi governed by the bearing properties [17]. The modes will
furthermore be effected by the gyroscopic moment of the
rotor and split up into two mode shapes [22]. The eigen-
Fc
frequencies of these forward and backward modes will be
Outer ring
influenced by the driving frequency O. This can be seen in
Fig. 8 where the simulated eigen-frequencies with speed
αe independent bearing stiffness are plotted. The simulation
shows an increasing difference with increasing speed
Qe
between the two eigen-frequencies and their value at
Fig. 6. Contact angles together with the Hertzian and centrifugal forces 0 rpm. The deviation at 24 000 rpm compared to 0 rpm is
which act on one ball. 79, 73, 715 and 723 Hz for modes 1–4, respectively (see
Table 2). The gyroscopic effect of the rotor can also be seen
in the eigen-frequencies of the simulation with speed
dependent bearing stiffness (see Fig. 9). In this simulation
Table 1 the influence of the characteristic shape of the decreasing
Stiffness matrix of the front and the back angular contact bearing at 0 rpm
bearing stiffness can be seen added to the gyroscopic effect
0 rpm x (1/m) y (1/m) yz (1/rad) zx (1/rad) resulting in a deviation at 24 000 rpm of 204 and 190, 202
Back
and 192, 16 and 15 Hz and finally 100 and 55 Hz for the
Fx (N) 3.36e8 0.00 0.00 − 6.10e6 first four backward and forward modes.
Fy (N) 0.00 3.36e8 6.10e6 0.00
M yz (Nm) 0.00 6.10e6 1.15e5 − 0.00 4.3. Mode shape analysis
M zx (Nm) − 6.10e6 0.00 0.00 1.15e5
Front Analysis of the simulated mode shapes (see Fig. 10)
Fx (N) 3.38e8 0.00 0.00 − 6.76e6 shows that the first mode shape starts at 0 rpm as a mode
Fy (N) 0.00 3.38e8 6.76e6 0.00
governed by the back bearing with its node close to the
M yz (Nm) 0.00 6.76e6 1.39e5 0.00
M zx (Nm) − 6.76e6 0.00 0.00 1.39e5 front bearing position. When the speed reaches 24 000 rpm
the mode has transformed into a mode governed by
Grey cells are plotted in Fig. 7 while the other cells are left out from the the front bearing stiffness. According to the simula-
plot due to their similar appearance. tion the transformation is performed in the interval
 ARTICLE IN PRESS
1040 M. Rantatalo et al. / International Journal of Machine Tools & Manufacture 47 (2007) 1034–1045

12 000–16 000 rpm. The first mode transforms its appear- The LDV measurement confirmed the simulated mode
ance by sliding its front bearing node position towards the shapes at the visible part of the spindle Fig. 10.
back bearing position. The second mode shape goes
through a similar transformation by starting as a front 4.4. CDST measurements
bearing mode at 0 rpm, going through a cylindrical mode
shape governed by both bearings at around 14 000 rpm and The FRFs Hxx and Hyy of the spindle/holder/dummy
finally ending up as a mode shape governed by the back tool measured by the CDST are shown in Figs. 11 and 12
bearing at 24 000 rpm. The transformation area can be seen where overlaying black lines in Fig. 12 represents the
in Fig. 9 as a disturbance of the shape of the changing simulated values. The two measurements made in the
eigen-frequency. The third mode can be seen as a flexural x- and y-directions show a decreasing s-shaped pattern for
mode with its two node points close to the bearing the eigen-frequencies of the second mode (at 752 Hz) and
positions while the fourth mode can be seen as a mixed the fourth mode (at approximately 1500 Hz). The back
mode governed mainly by the front bearing and flexural bearing related mode (mode 1) can only be seen vaguely in
properties. the y-direction at 664 Hz while the front bearing mode

2000 2000

1800 1800

Mode 4 Mode 4
1600 1600
Frequency [Hz]

Frequency [Hz]

1400 1400

1200 1200

1000 1000 Mode 3

Mode 3
Mode 2 800 Mode 2
800
Mode 1 Mode 1
600 600
Transformation area
400 400
0 8000 16000 24000 0 8000 16000 24000
Spindle speed [rpm] Spindle speed [rpm]

Fig. 8. Eigen-frequencies of the four first modes with the effect of the Fig. 9. Eigen-frequencies of the four first modes with the effect of the
gyroscopic moment of the rotor. speed dependent bearing stiffness and the gyroscopic moment of the rotor.

Table 2
Eigen-frequencies of CDST measurement and FEM simulation for 0 and 24 000 rpm

FEM 1 FEM 2 CDST

Backward Forward Backward Forward x Y

Mode 1
0 rpm 711 711 711 711 — 664
24 000 rpm 507 521 702 719 — —
Frequency change (%) 40.2 36.5 1.3 +1.1 — —

Mode 2
0 rpm 783 783 783 783 752 752
24 000 rpm 581 591 780 786 544 568
Frequency change (%) 34.8 32.5 0.4 +0.4 38.2 32.4
Mode 3
0 rpm 912 912 912 912 912 892
24 000 rpm 896 927 897 927 892 900
Frequency change (%) 1.8 +1.6 1.7 +1.6 2.2 +0.9
Mode 4
0 rpm 1578 1578 1578 1578 1496 1520
24 000 rpm 1478 1523 1555 1601 1384 1424
Frequency change (%) 6.8 3.6 1.5 +1.4 8.1 6.7

FEM 1 is simulation with speed dependent bearing stiffness and FEM 2 is without. —: not detected.
 ARTICLE IN PRESS
M. Rantatalo et al. / International Journal of Machine Tools & Manufacture 47 (2007) 1034–1045 1041

Mode LDV 0 rpm Simulation 0 rpm Simulation 14000 rpm Simulation 24000 rpm

LDV measured mode shapes Simulated mode shape Bearing position

Fig. 10. Mode shape analysis of the four first eigenmodes. Where no complex conjugate (dashed shapes) is present in the simulation the mode shape is real
valued. Drawings at top illustrate the positions of the mode shapes. LDV measured mode shapes are displayed to the left.

Hxx [log(m/N)] Hyy [log(m/N)]

2000 2000
-12.5 -12.5
-13
Mode 4 -13
1600 Mode 4 1600
-13.5
-13.5
Frequency [Hz]

Frequency [Hz]

-14
-14
1200 1200
-14.5
-14.5
Mode 3 -15 Mode 3
Mode 2 -15.5 -15
800 800 Mode 2

-16 -15.5
Mode 1

-16.5 -16
400 400
0 8000 16000 24000 0 8000 16000 24000
Spindle speed [rpm] Spindle speed [rpm]

Fig. 11. CDST measurement frequency response function Hxx. Fig. 12. CDST measurement frequency response function Hyy. Overlaying
black lines represents the simulations plotted in Fig. 9.

(mode 2) can be seen in both directions of the measure- seen in the measurements. A possible transformation area
ments for all spindle speeds. The third eigen-mode at where the two rigid rotor bearing modes transforms can be
approximately 900 Hz shows a relatively constant eigen- seen in the CDST measurement at 12 000 rpm and above as
frequency even thought the spindle speed increases. A small a change of magnitude. Furthermore the amplitude of this
change of the eigen-frequency (0.9–2.2%) of this mode second mode increases with increasing spindle speed,
(forward and backward) was found in the simulation and in especially above 12 000 rpm. The CDST measurement also
the CDST measurements. The simulated separation of the shows a higher frequency for the fourth mode when excited
modes into a backward and a forward mode could not be in the y-direction compared to the x-direction.
 ARTICLE IN PRESS
1042 M. Rantatalo et al. / International Journal of Machine Tools & Manufacture 47 (2007) 1034–1045

5. Discussion also apply to the first eigen-mode. This tendency can be

seen in the simulation (see Fig. 9). Due to this, a correct
In this paper, a method for analysing lateral vibrations in reading and identification of the eigen-frequencies of these
a milling machine spindle has been presented which two modes above the transformation area is difficult. The
includes the use of a contact-less spindle dynamic mea- relatively constant eigen-frequency of the eigen-mode at
surement device supported by FEM simulation. The FEM 912 Hz is consistent with the simulation; this also indicated
formulation was based on Ref. [12] and was extended by that the third eigenmode was a flexural mode with little
including a separate mass and stiffness radius together with influence of the radial bearing stiffness. This was also
a stiffness radius dependent shear deformation factor. confirmed by the mode shape simulation (see Fig. 10). The
FRFs in the radial directions for speeds in the interval characteristic s-shaped pattern could also be seen in the
[0:2000:24 000] rpm were measured and compared to eigen- CDST measurement of the fourth mode. This indicates
frequency simulation. that the mode was in some way related to the front bearing.
This indication was confirmed by the mode shape analysis
5.1. Gyroscopic and centrifugal effects which showed that this mode was a mix of a front bearing
mode and a flexural mode.
The bearing stiffness was found to be sensitive to the
centrifugal force acting on the bearing balls. This effect 5.3. Accuracy and validation
resulted in a substantial decrease in bearing stiffness
(38–62%) and hence bearing-related eigen-frequencies up The CDST measurements were compared with tradi-
to 40%. This effect was confirmed by the CDST measure- tional tap tests at 0 rpm. The tap tests showed for example,
ments. According to the simulations, the centrifugal effect a slightly lower eigen-frequency for the third mode and
on the bearings had a more significant effect on the eigen- with the CDST measurement the first mode was not
frequencies of bearing-related modes than the gyroscopic detected in the x-direction. This deviation could be due to
moment of the rotor. For example, mode 2 shows a different properties of the two excitation methods (CDST
33–35% frequency change when the centrifugal effect is and impulse) which in different ways could invoke non-
present as compared to only 0.4% when one is not present. linear properties of the system. Different positions of
One could expect that this softening could be neutralised the spindle during the two tests could also be a source of
for high loads that force the balls back to their nominal this deviation. This will be a subject of further studies.
positions. Another way to reduce the axial separation force Despite this small difference the tendency shown in the
of the rings would be by using angular contact bearings CDST measurements was not compromised and the
with a smaller angle (e.g., 151). The softening of the rotor change of spindle dynamic could clearly be seen. The
with increasing spindle speed and hence the reduction of different eigen-frequencies noted between the two radial
the eigen-frequencies which was reported by previous directions indicated that there was a non-symmetrical
authors could not be detected. This is evident when mounting of the spindle. This asymmetry was not
analysing the third simulated and measured mode, considered in the model. A deviation of approximately
(flexural) at approximately 900 Hz, where the eigen- 0–7% between the simulated and measured eigen-frequen-
frequency only seems to be affected by the gyroscopic cies at 0 rpm could be seen in the result. A possible reason
moment. for the deviation could be the stiff connection between
dummy tool, holder and spindle in the model. The
5.2. Mode analysis modelled mass of the spindle content, motor package
stiffness and the absence of the housing and preload
The separation of the modes into a backward and a mechanism in the model could also have been a source of
forward mode could not be detected in the measurements; this deviation.
this was most likely due to damping which may have Further studies will include the conversion of CDST
smeared the signal energy from the two modes into a single measurements from a dummy tool setup to a real cutting
peak. The measured eigen-mode at 752 Hz for 0 rpm is tool set-up.
assumed to be governed by the front bearing. This
assumption is based on the mode shape analysis where 6. Conclusion
the front bearing mode normally would result in a larger
displacement at the tool tip than the back bearing mode. A safe and repeatable method for spindle analysis is
Individual damping and phase conditions of these two presented including non-contact excitation and response
modes could reduce and smear the amplitude peak of the measurements of radial vibrations. The measurements were
first mode. According to the simulation of the second substantiated with simulations using a reduced finite
mode, the shape of the frequency change would begin by element model and mode shape verification by LDV. The
following the ‘‘s shape’’ of the front bearing stiffness, and model included only a rotor and speed dependent bearing
then jump over to follow the shape of the back bearing stiffness and resulted in an eigen-frequency deviation of
stiffness. The same, but in the opposite order, would then about 0–7% when compared to the measurements. Tests in
 ARTICLE IN PRESS
M. Rantatalo et al. / International Journal of Machine Tools & Manufacture 47 (2007) 1034–1045 1043

a milling machine of a high-speed spindle showed that the ml

½M 2  ¼
centrifugal forces of the ball bearing had a more significant 420ð1 þ FÞ2
2 3
influence on the system dynamics than the gyroscopic 140
6 7
moment of the rotor. The study performed indicates that 6 0 140 7
6 7
6 7
predictions of high speed milling stability based on 0 rpm 6 0 17:5l 3:5l 2 Sym: 7
6 7
tap test can be difficult due to speed dependent system 6 7
6 17:5l 0 0 3:5l 2 7
6 7
dynamics.  6 7,
6 70 0 0 17:5l 140 7
6 7
6 7
6 0 70 17:5l 0 0 140 7
Acknowledgements 6
6
7
7
6 0 17:5l 3:5l 2
0 0 17:5l 3:5l 2 7
4 5
The study was financed by The Swedish Agency for 17:5l 0 0 3:5l 2 17:5l 0 0 3:5l 2
Innovation Systems (Vinnova). SKF Nova is acknowl-
edged for their technical support and for their provision of jd
their developed CDST equipment. The Kempe Foundation ½N 0  ¼
30l ð1 þ FÞ2
provided the financing for the LDV system. 2 3
36
6 7
Appendix A 6 0 36 7
6 7
6 7
6 0 3l 4l 2 Sym: 7
A.1. System matrixes 6 7
6 7
Translational mass matrix [MT]+[M0]+F[M1]+F2[M2] 6 3l 0 0 4l 2 7
6 7
Rotational mass matrix [MR]+[N0]+F[N1]+F2[N2]
6 7,
6 36 0 0 3l 36 7
Element mass matrix [M] ¼ [MT]+[MR] 6 7
6 7
Element gyroscopic matrix [G] ¼ [G0]+F[G1]+F2[G2] 6 0 36 3l 0 0 36 7
6 7
Element stiffness matrix [K] ¼ [K0]+F[K1] 6 7
6 0 3l l 2 0 0 3l 4l 2 7
4 5
where 3l 0 0 l 2 3l 2 0 0 4l 2

ml jd
½M 0  ¼ ½N 1  ¼
420ð1 þ FÞ2 30l ð1 þ FÞ2
2 3 2 3
156 0
6 7 6 7
6 0 156 7 6 0 0 7
6 7 6 7
6 7 6 7
6 0 22l 4l 2 Sym: 7 6 0 15l 5l 2 Sym: 7
6 7 6 7
6 7 6 7
6 22l 0 0 4l 2 7 6 15l 0 0 5l 2 7
6 7 6 7
 6 7, 6 7,
6 54 0 0 13l 156 7 6 0 0 0 15l 0 7
6 7 6 7
6 7 6 7
6 0 54 13l 0 0 156 7 6 0 7
6 7 6 0 15l 0 0 0 7
6 7 6 7
6 0 13l 3l 2 0 0 22l 4l 2 7 6 0 7
4 5 4 15l 5l 2 0 0 15l 5l 2 5
13l 0 0 3l 2 22l 0 0 4l 2 15l 0 0 5l 2 15l 2 0 0 5l 2

jd
½N 2  ¼
30l ð1 þ FÞ2
ml 2 3
½M 1  ¼ 0
420ð1 þ FÞ2 6 7
2 3 60 0 7
294 6 7
6 7 6 7
6 0
6 294 7
7
6 0 0 10l 2 Sym: 7
6 7 6 7
6 0 2 7 6 7
6 38:5l 7l Sym: 7 60 0 0 10l 2 7
6 7 6 7
6 38:5l 0 0 7l 2 7 6 7,
6
6
7
7,
60 0 0 0 0 7

6 126 7 6 7
6 0 0 31:5l 294 7 6 7
6 7 60 0 0 0 0 0 7
6 0 126 31:5l 0 0 294 7 6 7
6 7 6 7
6
6 0
7
7
6 0 0 5l 2 0 0 0 10l 2 7
4 31:5l 7l 2 0 0 38:5l 7l 2 5 4 5
31:5l 0 0 7l 2 38:5l 0 0 7l 2 0 0 0 5l 2 0 0 0 10l 2
 ARTICLE IN PRESS
1044 M. Rantatalo et al. / International Journal of Machine Tools & Manufacture 47 (2007) 1034–1045

Jp 2 3
0
½G 0  ¼
30l ð1 þ FÞ2 60 0 7
2 3 6 7
0 6 2 7
60 0 l Sym: 7
6 7 6 7
EI 6 7
6 36 7 2
0 60 0 0 l 7
6 7 ½K 1  ¼ 3 6 7.
6 7 l 6 0 0 0 0 0 7
6 3l 0 0 Skew sym: 7 6 7
6 7 60 7
6 7 6 0 0 0 0 0 7
6 0 3l 4l 2 0 7 6 7
6 7 40 0 l 2 0 0 0 l2 5
6 7,
6 0 36 3l 0 0 7
6
6
7
7 0 0 0 l 2 0 0 0 l2
6 36 0 0 3l 36 0 7
6 7
6 7
6 3l 0 0 l2 3l 0 7
4 5
2
0 3l l 0 0 3l 4l 2 0
References

[1] W.F. Tobias, A Theory of Regenerative Chatter, The Engineer, 1958.

Jp
½G 1  ¼ [2] Y. Altintas, E. Budak, Analytical prediction of stability lobes in
30l ð1 þ FÞ2 milling, Manufacturing Technology 44 (1) (1995) 357–362.
2 3 [3] J. Tlusty, M.Polacek, The stability of the machine-tool against self-
0
6 7 excited vibration in machining, in: Proceedings of the International
6 0 0 7 Research in Production Engineering Conference, Pittsburgh, 1963,
6 7
6 7 pp. 465–474.
6 15l 0 0 Skew sym: 7 [4] H.E. Meritt, Theory of self-excited machine tool chatter, Journal of
6 7
6 7 Engineering for Industry (87) (1965) 447–454.
6 0 15l 5l 2 0 7
6 7 [5] W.X. Tang, X. Ai, S. Zhang, H. Jiang, Dynamic modeling for high-
6 7, speed milling system with centrifugal force and gyroscopic effect, Key
6 0 0 15l 0 0 7
6 7 Engineering Materials 259–260 (2004) 848–852.
6 7
6 0 0 0 15l 0 0 7 [6] G.L. Xiong, J.M. Yi, C. Zeng, H.K. Guo, L.X. Li, Study of the
6 7 gyroscopic effect of the spindle on the stability characteristics of the
6 7
6 15l 0 0 5l 2
15l 0 7 milling system, Journal of Materials Processing Technology 138 (1–3)
4 5
(2003) 379–384.
2
0 15l 5l 0 0 15l 5l 2 0 [7] Y. Cao, Y. Altintas, A general method for the modeling of spindle-
bearing systems, Journal of Mechanical Design 126 (6) (2004)
1089–1104.
Jp [8] T.L. Schmitz, J.C. Ziegert, C. Stanislaus, A method for predicting
½G 2  ¼ chatter stability for systems with speed-dependent spindle dynamics,
30l ð1 þ FÞ2 in: Transactions of the North American Manufacturing Research
2 3
0 Institute of SME Conference, Charlotte, vol. 32, 2004, pp. 17–23.
6 7 [9] W.R. Wang, C.N. Chang, Dynamic analysis and design of a machine
60 0 7 tool spindle-bearing system, Journal of Vibration and Acoustics 116
6 7
6 7 (1994) 280–285.
60 0 0 Skew sym: 7
6 7 [10] H.D. Nelson, J.M. McVaugh, The dynamics of rotor-bearing systems
6 7 using finite elements, Journal of Engineering for Industry 93 (2)
6 0 0 10l 2 0 7
6 7 (1976) 593–600.
6 7,
60 0 0 0 0 7 [11] E.S. Zorzi, H.D. Nelson, Finite element simulation of rotor-bearing
6 7 systems with internal damping, Journal of Engineering for Power
6 7
60 0 0 0 0 0 7 (1977) 71–76.
6 7
6 7 [12] H.D. Nelson, A finite rotating shaft element using Timoshenko beam
60 0 0 5l 2 0 0 7 theory, Journal of Mechanical Design (102) (1980) 793–803.
4 5
[13] C.-W. Lin, J.F. Tu, J. Kamman, An integrated thermo-mechanical-
0 0 5l 2 0 0 0 10l 2 0 dynamic model to characterize motorized machine tool spindles
during very high speed rotation, International Journal of Machine
Tools and Manufacture 43 (10) (2003) 1035–1050.
2 3 [14] S. Auchet, P. Chevrier, M. Lacour, P. Lipinski, A new method of
12 cutting force measurement based on command voltages of active
6 0 12 7 electro-magnetic bearings, International Journal of Machine Tools
6 7
6 7 and Manufacture 44 (14) (2004) 1441–1449.
6 0 6l 4l 2 Sym: 7
6 7 [15] C.R. Knospe, Active magnetic bearings for machining applications.
EI 6
6 6l 0 0 4l 2 7
7 Control Engineering Practice, 2005, in press, corrected proof.
½K 0  ¼ 3 6 7, [16] C. Min, C.R. Knospe, A new approach to the estimation of cutting
l 6
6
12 0 0 6l 12 7
7 dynamics using active magnetic bearings, Journal of Manufacturing
6 0 12 6l 0 0 12 7
6 7 Science and Engineering—Transactions of the ASME 127 (4) (2005)
6 7
4 0 6l 2l 2 0 0 6l 4l 2 5 773–780.
[17] G. Genta, Dynamics of rotating systems, in: F.F. Ling (Ed.),
6l 0 0 2l 2 6l 0 0 4l 2 Mechanical Engineering, Springer, Berlin, 2005.
 ARTICLE IN PRESS
M. Rantatalo et al. / International Journal of Machine Tools & Manufacture 47 (2007) 1034–1045 1045

[18] J.R. Hutchinson, Shear coefficients for Timoshenko beam theory, [21] J.M. de Mul, J.M. Vree, D.A. Maas, Equilibrium and associated load
Journal of Applied Mechanics 68 (1) (2001) 87–91. distribution in ball and roller bearings loaded in five degrees of
[19] D.J. Ewins, Modal Testing: Theory, Practice and Applications, freedom while neglecting friction—part I: general theory and
second ed., Research Studies Press Ltd., Philadelphia, 2000. application to ball bearings, Journal of Tribology 111 (1) (1989)
[20] J.-S. Chen, K.-W. Chen, Bearing load analysis and control of a 142–148.
motorized high speed spindle, International Journal of Machine [22] T. Yamamoto, Y. Ishida, Linear and Nonlinear Rotordynamics:
Tools and Manufacture 45 (12–13) (2005) 1487–1493. A Modern Treatment with Applications, Wiley, New York, 2001.

View publication stats