CIRP Annals - Manufacturing Technology 64 (2015) 365–368

Contents lists available at ScienceDirect

CIRP Annals - Manufacturing Technology

jou rnal homep age : ht t p: // ees .e lse vi er . com /ci r p/ def a ult . asp

Evaluation of dynamic stiffness of machine tool spindle by non-contact

excitation tests
Atsushi Matsubara (2)*, Shota Tsujimoto, Daisuke Kono
Department of Micro Engineering, Kyoto University, Japan

A R T I C L E I N F O A B S T R A C T

Keywords: This paper presents a non-contact excitation method for evaluating the dynamic stiffness of a rotating
Spindle
spindle. The spindle response to an excitation force is measured, and frequency response functions (FRFs)
Stiffness
are obtained. Based on the measured FRFs, dynamic uncertainty and its effect on cutting stability are
Chatter
investigated. Regenerative forces are generated using displacement feedback with a time lag element,
and a closed-loop excitation test is executed automatically. The stability map obtained from the closed-
loop test and the stability charts calculated from the FRFs are compared, and the uncertainty of the
spindle dynamics during operations is clarified.
ß 2015 CIRP.

1. Introduction domain has been conducted using FRFs and cutting models [13]. In
general, the analytical predictions are confirmed by the results of
Machine tool spindles must rotate cutting tools precisely and cutting tests. However, cutting processes have various effects, such
provide sufficient energy for material removal [1]. For high- as the multiple regenerative effect [14,15], process damping [16],
productivity machining, spindles should be designed to have multiple frequencies, and multiple delay [17]. These effects make it
sufficient stiffness to resist cutting forces. The basic components of difficult to isolate the dynamic behavior of the spindle for analysis.
a spindle, such as its rotating shaft, support bearings, and contact In this research, the uncertainty of the dynamic stiffness of a
parts, have elasticity, and their contribution to stiffness at the tool rotating spindle was investigated using a non-contact excitation
point is complex. State-of-art design methods consider the method. Two excitation tests were combined. One test used a
nonlinearity of preloaded bearings [2], interface dynamics [3], conventional open-loop excitation, and the other used a closed-
and the thermal distribution of the components [4,5]. Measure- loop excitation. In the closed-loop test, the regenerative force was
ment methods are important in the evaluation of the designed generated on the basis of the monitored spindle displacement.
stiffness and its achievement in the manufacturing process. Chatter behavior was investigated in conjunction with the
Rantatalo et al. investigated the vibration characteristics of uncertainty of the FRFs measured in the open-loop test.
rotating spindles, including the rotor dynamics, using non-contact
magnetic excitation [6]. Abele et al. investigated the speed 2. Measurement system
dependency of the natural frequency and damping of a hybrid
spindle using the active magnetic bearing (AMB) component 2.1. Configuration of measurement setup
[7]. Ozturk et al. demonstrated the effect of spindle bearing preload
on dynamic stiffness and chatter stability [8]. Abele and Fiedler Fig. 1 shows an overview of the test device used. The device
proposed the state-space identification of the frequency response consists of a magnet loader, a dummy tool (a load target),
functions (FRFs) of spindle compliance during milling operations measurement sensors, and a personal computer (PC) to control the
[9]. Suzuki et al. investigated a different identification method process. The loader has coils that generate magnetic forces, which
based on chatter information [10]. Tobias discussed the nonline- are controlled by coil currents. In this research, one coil set was
arity of spindle stiffness and its effect on chatter [11]. The Author used to emulate the cutting forces of side milling or boring. The coil
investigated the influence of spindle temperature on stiffness current was supplied by an amplifier, to which a current command
using a non-contact measurement method and a magnet loading was sent from the PC through a digital-to-analog (D/A) converter.
device [12]. Displacement sensors were installed in the head stock of the
The influence of machine dynamics on the occurrence of spindle to measure the spindle displacement at the holder. The
vibration has been analyzed using measured and calculated displacement signal was filtered using a high-pass filter and
machine dynamics. Regenerative chatter analysis in the frequency captured by an analog-to-digital (A/D) converter. This high-pass
filter filters out the machine vibration other than the spindle
vibrations. A tool dynamometer was installed between the
* Corresponding author. machine table and the magnet loader to measure the attractive
E-mail address: [email protected] (A. Matsubara). force. The force signal was also captured by the A/D converter. Using

http://dx.doi.org/10.1016/j.cirp.2015.04.101
0007-8506/ß 2015 CIRP.
366 A. Matsubara et al. / CIRP Annals - Manufacturing Technology 64 (2015) 365–368

The relationship between the coil current and magnetic force is

as follows:

f ¼ K i i2c (1)

where Ki is the current force coefficient, which is inversely

proportional the square of the gap distance. A sine wave with an
offset is given to the current as follows:

ic ¼ I0 þ Ia sinðvtÞ (2)

where I0 is the current offset, Ia is the current amplitude, and v is

the excitation frequency. Substituting (2) into (1) and referring to
the fundamental harmonic term, the relationship between the
force and command current can be estimated as follows:
Fig. 1. Configuration of measurement setup.
f r ¼ K ft Ia sinðvtÞ ¼ K ft ir (3)

where Kft = 2KiI0. The nominal depth of cut determines the current
offset.

3. Nonlinearity in spindle stiffness

To evaluate nonlinearity in the spindle stiffness, the current

offset I0 and amplitude Ia were changed, and open-loop FRFs
were measured. The spindle speed was set at 3000 min 1 . Using
the measured data, the FRFs from f i to d f (Gdi) and the FRFs from f h
to d f (G df) were estimated. The coherence in each measurement
was over 0.9 except for the vicinity of the natural frequency,
where the minimal coherences are 0.75 for Gdi and 0.45 for
Gdf . This reason is that the vibration components other than
the force-excitation response also increase at the resonance
Fig. 2. Block diagram of excitation system. frequency.
Fig. 3 shows the estimated FRFs for the variable offset and
this device, the spindle stiffness of a multitasking machine was constant amplitude. The natural frequencies decreased as the
measured. The spindle’s key specifications were as follows: the current (force) offset increased, which reflects the soft spring
maximum speed was 12,000 min1, the preload type was constant- characteristic in the force–displacement relationship. Gdf
position preload, the front bearing employed angular contact exhibits a sharper change in gain and a greater lag in phase
bearing, the rear bearing employed cylindrical roller bearing, and than Gdi. Note that the force measurement unit has a natural
the diameter of the front bearing was 60 mm. frequency near 300 Hz, which provides uncertainty to both the
gain and the phase.
2.2. Control block and excitation test

Fig. 2 shows the block diagram of the excitation force

generation. The blocks from ir to df correspond to the analog
components, where ir is the command current sent to the current
amplifier, ic is the coil current of the magnet loader, f is the magnet
force, fh is the measured force, d is the spindle displacement
measured at the toll holder, and df is the filtered displacement. The
transfer functions of the current amplifier, magnet loader, spindle
compliance, high-pass filter, and force measurement are Gamp, Gfi,
Gdf, Ghpf, and Gm, respectively. The cut-off frequency of the high-
pass filter was set at 300 Hz because structural vibrations of the
machine occur at frequencies below this threshold.
The blocks from df to ir are in the digital control system, which
represents the cutting force feedback. To simulate regenerative
vibration, a time lag element with a spindle rotation period T was
inserted in parallel to the feedback loop. In Fig. 2, h0 is the nominal
cut depth, Kf is the specific cutting force, and a is the cut width.
Multiplication of a and Kf yields the cutting stiffness. The term fi
refers to the emulated cutting force, and the term Kft refers to the
transferring gain from force to coil current. Fig. 3. Measured FRFs for variable offset and constant amplitude: Ia = 0.3 A, Gdi is FRF
In the open-loop test, the feedback pass was cut, and a swept- from fi to df, Gdf is FRF from FRF from fh to df.

sine signal was given directly to the command current ir. From the
measured ir and df, an FRF was estimated. This FRF is the serial
product of Gamp, Gfi, Gdf, and Ghpf and is referred to as the open-loop Fig. 4 shows the estimated FRFs for the constant offset and
FRF f. The force signal was also measured, and the FRF from fh to df variable amplitude. The change of the natural frequency could be
was estimated to check the open-loop FRF. seen in Gdf where Ia is less than 0.6 A. However, it could not be seen
In the closed-loop test, T was set to the real rotation period of for larger amplitudes, where the gain decrease could be seen in
the spindle, and aKf was set. The feedback loop was then closed to Gdi. This indicates that the soft spring effect is weak at this force
check the oscillation. All tests were performed automatically under offset and the gain in the current–force relationship is decreased
spindle rotating conditions. with the increase of the current amplitude.
 A. Matsubara et al. / CIRP Annals - Manufacturing Technology 64 (2015) 365–368 367

Fig. 5. Measured temperature change.

Fig. 4. Measured FRFs for variable amplitude and constant offset: I0 = 1.5 A, Gdi is FRF
from fi to df, Gdf is FRF from FRF from fh to df.

4. Measurement and evaluation

4.1. Test procedure and results

Spindle stiffness is known to be sensitive to the spindle speed

and spindle temperature. The stability of the closed loop was
evaluated for several combinations of spindle speed and cutting
stiffness aKf. The test procedure is as follows.

(1) Rotation of the spindle was started, and the spindle was
allowed to warm up for 40 min. The rotation speed was set at
3000 min1. Fig. 6. Example measurements from closed-loop test: N = 2900 min1,
(2) After the warm up, FRFs were measured in the open-loop test. aKf = 1.05  105 N/m.
To investigate the dependency of the spindle stiffness on the
speed, three speed levels were considered. I0 and Ia were set at
constant values of 1.5 A and 0.3 A, respectively. C: The vibration increased and decreased alternately.
(3) The stability limit of aKf for each spindle speed N was calculated
from the measured FRF f, using the following relations [18]: An example of pattern B is shown in Fig. 6.
K ft
aK f ¼ (4) 4.2. Comparison of chatter map and estimated lobe charts
2Re½f

60 Im½f The occurrence of chatter in the closed-loop test was examined

vc ¼ 2kp þ 3p  2tan1 (5) using the vibration level. To clarify whether the instability arose
N Re½f
from the current limit or from the spindle damping, the maximum
where vc is the chatter frequency in rad/s, N is the spindle speed in current amplitudes were checked and shown in a vibration map.
min1, and k is the number of vibration waves per revolution, and The stability lobes calculated from the open-loop FRFs were also
Re and Im are the real and imaginary parts, respectively, of the FRF. shown in the map.
(4) Based on the calculated lobes, the spindle speed N and aKfs Figs. 7 and 8 show the vibration levels of the test results and the
were selected for the closed-loop test. In the test case, the stability lobe charts calculated from the open-loop FRFs measured
speed range was set to vary about a target speed of 3000 min1 before and after the closed-loop test, respectively. Three lines were
to cover k = 15 in increments of 25 min1. The closed-loop tests
were conducted with a current offset of 1.5 A.
(5) To check the temperature dependency, open-loop FRFs were
measured at the middle and end of the test. During the entire
test, the temperature was measured with a thermocouple
installed near the outer race of the front bearing.

Fig. 5 shows the temperature and spindle speed over the course
of the test. As this figure shows, the temperature increased
gradually, and the total increase was approximately 2 8C. During
the closed-loop test, several vibration patterns were observed. The
vibrations were analyzed in both the time and frequency domains.
The power spectral densities (PSDs) of the vibrations were
calculated using Fourier transform and short-time Fourier
transform (STFT). The measured vibrations were categorized
according to the following patterns:

A: The vibration amplitude remained at the initial level.

B: The oscillation started and then was sustained at a certain Fig. 7. Vibration map and stability lobes: limit lines are given from FRFs measured
amplitude. before the closed-loop test.
368 A. Matsubara et al. / CIRP Annals - Manufacturing Technology 64 (2015) 365–368

saturated because of the current limit. Speed dependency of the

stability lobes is illustrated in Fig. 7 but not in Fig. 8. The reason for
this is that the FRF became insensitive to spindle speed as the
spindle temperature increased.
To investigate the nonlinearity discussed in Section 3, stability
lobes were calculated using the FRFs shown in Fig. 4. Fig. 9 shows
the calculated lobes for different current aptitudes. As could be
seen in Fig. 9, the effect of the soft spring characteristic on the
stability is small. The stability zone expands, when the current
amplitude is larger than 0.6 A, which could tell the oscillation
dwell. This comes from the gain decrease in the current–force
relation and the current limit of the amplifier. The calculated lobes
in Figs. 8 and 9 could not explain the stability at a speed of
3075 min1. Fig. 10 shows the measured signals at this speed. The
first oscillation starts at 803 Hz. This oscillation disappears, and the
second oscillation starts at 765 Hz, which makes it difficult to
assess the stability. A similar oscillation pattern is described as
Fig. 8. Vibration map and stability lobes: limit lines are given from FRFs measured bifurcation in [19].
after the closed-loop test.

5. Conclusions

Non-contact excitation tests of a machine tool spindle were

conducted to predict chatter conditions and monitor oscillation
under regenerative force feedback. The uncertainties associated
with spindle stiffness due to speed and vibration amplitude were
assessed quantitatively. In the case of measurement of a constant-
position-preload spindle, the soft spring characteristics was
observed in the open-loop test, but its effect on chatter in the
closed-loop test was found to be small. The speed effect was found
to be influenced by the spindle temperature.

References

[1] Abele E, Altintas Y, Brecher C (2010) Machine Tool Spindle Units. CIRP Annals –
Manufacturing Technology 59(2):781–802.
[2] Altintas Y, Cao Y (2005) Virtual Design and Optimization of Machine Tool
Spindles. CIRP Annals – Manufacturing Technology 54(1):379–382.
Fig. 9. Vibration map and stability lobes: limit lines are given from FRFs in Fig. 4. [3] Ertürk A, Özgüven HN, Budak E (2007) Effect Analysis of Bearing and Interface
Dynamics on Tool Point FRF for Chatter Stability in Machine Tools by using a
New Analytical Model for Spindle–Tool Assemblies. International Journal of
Machine Tools and Manufacture 47:23–32.
[4] Lin CW, Tu JF, Kamman J (2003) An Integrated Thermo-mechanical-dynamic
Model to Characterize Motorized Machine Tool Spindles during Very High
Speed Rotation. International Journal of Machine Tools and Manufacture
43(10):1035–1050.
[5] Holkup T, Cao H, Kolář P, Altintas Y, Zeleny J (2010) Thermo-mechanical Model
of Spindles. CIRP Annals – Manufacturing Technology 59(1):365–368.
[6] Rantatalo M, Aidanpaa JO, Goransson B, Norman P (2007) Milling Machine
Spindle Analysis using FEM and Non-contact Spindle Excitation and Response
Measurement. International Journal of Machine Tools and Manufacture
47:1034–1045.
[7] Abele E, Schiffler A, Rothenbücher S (2007) System Identification during
Milling via Active Magnetic Bearing. Production Engineering 1(2):219–226.
[8] Ozturk E, Kumar U, Turner S, Schmitz T (2012) Investigation of Spindle Bearing
Preload on Dynamics and Stability Limit in Milling. CIRP Annals – Manufactur-
ing Technology 61:343–346.
[9] Abele E, Fiedler U (2004) Creating Stability Lobe Diagrams during Milling. CIRP
Annals – Manufacturing Technology 53(1):309–312.
[10] Suzuki N, Kurata Y, Kato T, et al (2012) Identification of Transfer Function by
Inverse Analysis of Self-excited Chatter Vibration in Milling Operations.
Precision Engineering 36:568–575.
[11] Tlusty J, Livingston IV R, Teng YB (1986) Nonlinearities in Spindle Bearings and
their Effects. CIRP Annals – Manufacturing Technology 35(1):269–273.
[12] Matsubara A, Sawamura R, Asano K, Muraki T (2014) Non-contact Measure-
Fig. 10. Measurement results for closed-loop test: N = 3075 min1, ment of Dynamic Stiffness of Rotating Spindle. Procedia CIRP 14:484–487.
aKf = 2.40  105 N/m). [13] Altintas Y, Kilic ZM (2013) Generalized Dynamic Model of Metal Cutting
Operations. CIRP Annals – Manufacturing Technology 62(1):47–50.
calculated from the measured FRFs for the three different spindle [14] Tlusty J, Ismail F (1981) Basic Non-linearity in Machining Chatter. CIRP Annals –
Manufacturing Technology 30(1):299–304.
speeds considered. The discontinuities in the lines reflect
[15] Kondo O, Kawano Y, Sato H (1981) Behaviour of Self-excited Chatter due to
measurement error. The marker size corresponds to the maximum Multiple Regenerative Effect. ASME Journal of Engineering for Industry 103:324–329.
vibration amplitude of the current during the test period. The cross [16] Budak E, Tunc LT (2010) Identification and Modeling of Process Damping in
Turning and Milling using a New Approach. CIRP Annals – Manufacturing
mark indicates that the current reached the amplifier limit. The
Technology 59(1):403–408.
test run at 3075 min1 was repeated, and the results are shown as [17] Bachrathy D, Steppn G (2013) Improved Prediction of Stability Lobes with
the ‘‘second set.’’ As these figures show, the vibration amplitude Extended Multi Frequency Solution. CIRP Annals – Manufacturing Technology
increased significantly near the stability limit, as indicated by the 62(1):411–414.
[18] Altintas Y (2000) Manufacturing Automation, Cambridge University Press.
circle radius becoming larger. Beyond the limit, the amplitude [19] Stepan G (2001) Modelling Nonlinear Regenerative Effects in Metal Cutting.
remained constant for a certain range of aKf and then increased and Philosophical Transactions of the Royal Society London 359(A):739–757.