Accepted Manuscript

An optical measurement technique for dynamic stiffness and damping of pre-

cision spindle system

ChaBum Lee, Abolfazl Zolfaghari, Gyu Ha Kim, Seongkyul Jeon

PII: S0263-2241(18)30788-7
DOI: https://doi.org/10.1016/j.measurement.2018.08.049
Reference: MEASUR 5828

To appear in: Measurement

Received Date: 27 April 2017

Revised Date: 6 February 2018
Accepted Date: 23 August 2018

Please cite this article as: C. Lee, A. Zolfaghari, G. Ha Kim, S. Jeon, An optical measurement technique for dynamic
stiffness and damping of precision spindle system, Measurement (2018), doi: https://doi.org/10.1016/
j.measurement.2018.08.049

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
An Optical Measurement Technique for Dynamic Stiffness and Damping of Precision Spindle System

ChaBum Lee1,*, Abolfazl Zolfaghari1, Gyu Ha Kim2 and Seongkyul Jeon1

1
Department of Mechanical Engineering, Tennessee Technological University
1 William L Jones Dr., Cookeville, TN, 38505, USA
Corresponding: clee@tntech.edu, Tel. +1-931-372-6169
2
TRI-N Co. Ltd
88, Dalseong2cha 3-ro, Guji-myeon, Dalseong-gun, Daegu, Korea

Abstract

The dimensional measurement technique utilizing curved-edge diffraction is applied for the dynamic
system identification of precision spindle system. Both static and dynamic behavior including stiffness,
damping ratio, and parasitic motion of a precision ball bearing spindle system is characterized by impact
response, while the spindle displacement is measured with the curved-edge-type sensors (CES). The CES
effects of spindle radius and surface quality were theoretically and empirically investigated. The capacitive-
type sensors (CS) were used for a baseline comparison with the CES outputs. Unlike CS, CES is not
sensitive to the spindle radius, surface quality, and coupling motions, shows high bandwidth and low noise,
and allows for the dimensional measurement at a localized area. These results indicate that CES can be a
good alternative to CS for spindle metrology. In the experiment, natural frequency (56 Hz (linear) and 680
Hz (angular)), stiffness (460~790 N/µm) and damping ratio (0.04~0.08) of precision spindle system were
measured by CS and CES at the same time. The dynamic model of ball bearing spindle system was also
discussed.

Keyword: Spindle metrology; Curved edge sensor; Edge diffraction; System identification

1. Introduction

Precision ball bearing, hydrostatic or aerostatic spindles represent a key part of machine tools because they
determine the quality of the final product produced and the overall manufacturing productivity and
efficiency. Therefore, dynamic system identification and condition monitoring of precision spindles is a
key factor in increasing the availability of the machine tool and achieving a more robust machining process
[1-4]. Precision spindle metrology, i.e., utilizing a cylindrical or spherical target artifact with non-contact
sensors to measure a machine tool spindle’s “axis of rotation” errors, including radial, axial, and tilt errors,
is basic science and technology for sustainable manufacturing [5,6].

Significant research has been devoted to the development of measurement and analysis techniques of
precision spindles. Currently, a few approaches in measuring and evaluating precision spindle systems by
using capacitive-type sensors (CS) or eddy current-type sensors (ECS) exist [7,8]. However, these sensors
are typically designed for flat target surface measurement. CS and ECS are the most commonly used in
spindle measurements. It is standard practice for commercial CS and ECS to be factory calibrated with flat-
target surfaces. Because the probes are calibrated to a flat target, measuring a target with a curved surface
will cause errors. Because the probe will measure the average distance to the target, the gap at zero volts
will be different than when the system was calibrated [9,10]. From the previous work [11,12], the CS shows
spindle radius effects on the displacement measurement. As depicted in Figure 1, errors will also be

1
introduced because of the different behavior of the electric fields between sensing electrode and the curved
surface and between guard electrode and the curved surface. In cases where a non-flat target must be
measured, the system can be factory calibrated to the final target shape or the probe has to be custom-made
for a certain shape. CS or ECS targeting spindle surfaces will lead to four sources of error if they are
calibrated with flat targets: (1) the sensitivity of the sensor increases, resulting in exaggerated displacement
measurements; (2) the sensing range is both decreased and shifted towards the target; (3) the otherwise
linear output of the sensor system becomes nonlinear for targets of decreasing diameter; and (4) the sensing
noise issues also become significant due to interference between probes in the case of multi-probe methods.
Since these effects lead to critical measurement error in precision metrology for manufacturing applications
requiring the highest accuracy, it is urgent that a new rigorous sensing methodology for precision
dimensional metrology of curved target surfaces be investigated.

In this study, a curved-edge sensor (CES) is adapted for dynamic system identification of precision ball
bearing spindle system. In the previous work [13-15], knife-edge sensors (KES) was firstly introduced, and
the result indicates that the KES can be a good alternative to CS in nanopositioning applications, e.g.,
microscopy, surface profilers, or coordinate measuring machines. In addition to KES, CES was introduced
[11,12]. CES was compared with CS in terms of linearity, measuring range, bandwidth and long-term
stability. As a result, it is expected that CES can be one of alternative sensors applicable to spindle
metrology. Here a CES model that can describe curved-edge diffraction phenomenon on a curved surface
is established by using electromagnetic wave propagation theory. The curved-edge diffraction effects
according to surface quality, shaft radius, and materials are empirically characterized, and dynamic system
identification of ball bearing spindle system is discussed.

Figure 1. A schematic of dimensional measurement from the curved target surface by using
capacitive-type sensor.

2. Measurement Principle

Diffraction is the bending of wavefronts around obstacles, and its effects are typically only noticeable for
waves where the wavelength similar to the size of the diffracting obstacle [16-18]. Diffraction occurs with
all propagating waves, including sound waves, water waves, electromagnetic (EM) wave. Diffraction over

2
a knife-edge, so-called knife-edge diffraction, can be understood through using Huygens construction [16].
This effect is initially an oscillation then as the direct path is cut off, a signal loss. Unlike knife-edge
diffraction, an analytical approach to diffraction by a curved edge is very difficult because curved edge
diffraction must be understood from the asymptotic solution of several canonical problems, which involve
the illumination of the edge by different wave fronts [18]. There exist few convenient and efficient
computational methods to look at curved edge diffraction problems in the transition region adjacent to
shadow and reflection boundaries.

Figure 2. A schematic of curved-edge diffraction.

As illustrated in Figure 2, an EM wave incident on the curved edge gives rise to an incident wave (Ei),
reflected wave (Er), and edge diffracted wave (Ed), and those electric fields may be expressed as [13]

2
x0 , j  y 0 , j
2
 
(1a)
Ei ( x0, j , y0, j , z j )  E0  e  .
2

 1  jk r  jk r
Er ( xd , j , yd , j , z j )  
k  space (2 ) 2 aperture
i e
o s0
E e o 0d dydxdkydkx, (1b)
  
Ed ( xd , j , yd , j , z j )  Ei  e  j (  ko n( R cos  S ( y , z ))) e  jko ( rs0  S ( y , z )) . (1c)


where, Eo is the amplitude of the incident field, Ei is the electric field of the light source in the absence of
 
the edge; E r is the electric field reflected from the surface before the curved edge, and Ed is the diffracted
electric field. The regions, RB, SB and DSB, indicate reflection boundary, shadow boundary, and deep
shadow boundary, and R indicates the radius of spindle shaft. The S(y,z) is the surface geometry of spindle

shaft. The ko is the wavevector, α is the beam radius, ri  j is a position vector from i to j in the coordinate
system. Such waves that propagate along a surface ray may be excited at the curved edge shadow
boundaries similar to knife edge diffraction. Such surface ray fields may also be excited at the transition
region and can be separated into three different regions [18]. The total electric field E at each region may
be represented as

3
   
ERB  Ei ui  Er ur , (2a)
  
ESB  Ei ui  Ed ud , (2b)
 
EDSB  Ed ud . (2c)

The functions ui, ur, and ud are unit step functions. This sum predicts the scattering of the optical large
platform that involves the use of spatial domain Fresnel integrals for the scattered or diffracted fields 41-44.
If the detector size is smaller than SB region and the distance between the curved edge and the detector is
much larger than the detector size, the reflection field can be exclusive to the total electric field calculation.
At his condition, thus, the total electric field from the curved edge can be as same as that of knife edge
diffraction. In this study, the detector (ϕ250 µm) is used to neglect the reflection field from the curved edge.

3. Experiments

Curved edge diffraction fringes of two test samples with different radius (ϕ3, 5, 10, 20, 30, 40 and 50 mm)
conditions, surface conditions and different material (Cu, Al) conditions were measured as shown in Figure
3. The Cu sample was machined by ultraprecision machine tool (Nanoform 200) that is used for optical
component machining, and the Al sample was machined by general machine tool that can be found in any
machine shop. The He-Ne laser (633 nm wavelength) was incident on the edge of the sample, the fringes
that can be collected from RB, SB and DSB regions were obtained by scanning the curved edges of each
sample as illustrated in Figure 2. The distance between the curved edge and the detector was kept 150 mm.
As a result, any significant changes according to materials, radius curvature were not found in given
experiment conditions. However, a slight difference in fringe amplitude between curved edge and knife
edge was observed. It could be understood that curved edge is less efficient to block the incident light to
the edge in the small region of the shaft that the light is in contact with the shaft edge. A small portion of
reflection field could have an influence in the total field even though the detector with small active area is
used. Therefore, it can be expected that the total field becomes lower as the shaft radius goes down to
submillimeter or micrometer level. This radius effect of curved edge diffraction will be investigated in
detail in the future study.

Figure 3. Curved-edge diffraction effects of materials, radius and surface quality: (a) ultraprecision-
machined Cu, (b) generally-machined Al and (c) comparison (ϕ20 mm).

4
 Figure 4. Experiment setup.

In the experiment, the precision ball bearing spindle system was chosen. Dynamic system identification of
the spindle system was investigated by the curved edge sensor as illustrated in Figure 4. The He-Ne laser
with beam diameter 500 µm is 4 splitted by beam splitters (BS), and each beam is incident on the edge of
the spindle shaft. The four photodiodes (PD) collect the intensity of incident beams, and are amplified by
operational amplifiers. The electronics bandwidth was set 10 kHz. The light intensity to be measured at i-
th detector can be expressed by
 
Ii  Ei , SB  Ei*, SB , (3)

The displacement, r0 and r90, of the spindle shaft can be expressed by

I PD 2  I PD1
ro  K o , (4a)
I PD 2  I PD1

I PD 4  I PD 3
r90  K90 . (4b)
I PD 4  I PD 4

where, K0 and K90 are the sensitivity of CES at 0 and 90 degree, respectively.

5
 10
r0
r90

Sensor Outputs [V]

5

0
Sensitivity
-5 r0 167.5mV/m
r90 163.4mV/m

-10
0 20 40 60 80 100 120
Displacement [m]
Figure 5. Calibration results: linearity errors (calculated as standard deviation of residual errors) were 0.10%
and 0.14% for r0 and r90 in full scale (95% confidence).

Figure 6. Comparison of CS and CES: Spindle runouts with respect to spindle velocity; (a) 1000 rpm, (b)
3300 rpm and (c) 8600 rpm.

To properly calibrate CES with CS, the flat target surface (flat area in the spindle housing) was used for
CS. The CES was calibrated by moving the spindle housing along two directions. As a result of calibration,
sensor outputs showed high linearity (0.1% and 0.14% for r0 and r90 in full scale approximately 100 µm).

The CS and CES outputs were compared with respect to spindle velocity up to 8700 rpm (maximum rpm).
The two CS was placed around spindle shaft (ϕ34 mm) along the r0 and r90 directions. As seen in Figure 6,
the maximum spindle runout was found at a 3300 rpm condition because of the natural frequency of spindle
system (56 Hz). Also, the spindle runout was significantly reduced at 8700 rpm condition. However, there
are large discrepancy between two outputs in micrometer regions. From the previous studies [11], it is well-
known that CS is sensitive to lateral motions and curvature of the measuring target surface. In addition to
lateral motion effect, CS is sensitive to surface roughness of the spindle shaft, while CES is not sensitive to
the roughness effect (Figure 3). Thus, it was considered that the curved target surface of the spindle shaft

6
could be associated with CS measurement error. Also, it could be alignment error associated with two CS
or CES setup with 90 degree interval.

Figure 7. Impact response at spindle velocity 0 rpm (stationary) condition: (a) r0 and (b) r90.

Figure 8. Impact response at spindle velocity 1000 rpm condition: (a) r0 and (b) r90.

7
Figure 9. Impact response at spindle velocity 3300 rpm condition: (a) r0 and (b) r90.

Figure 10. Impact response at spindle velocity 8600 rpm condition: (a) r0 and (b) r90.

8
4. System Identification

The impact test using hammer with force sensor along the radial direction was performed. The CS and CES
measured the displacement along the r0 and r90 directions and the applied force was measured at the same
time, while the impact was applied to the r0 direction under stationary condition (Figure 7) and rotating
conditions (Figure 8-10), respectively. Two sensors showed the similar results that the amplitude of
oscillation decreases with time. The CES results in the graphs from Figure 7 to Figure 10 look noisier than
the CS results, but it is not a noise but the signals that represent the motion of the spindle. Because the CS
with a certain effective sensing area (here ϕ5 mm) averages the displacement between the CS probe and
spindle shaft surface as discussed in the Section of Introduction, it is difficult for the CS to measure the
sub-micrometer scale spindle motion in a localized point. This difficulty becomes more significant as the
effective sensing area of the CS is larger.
Two frequencies, 56 Hz and 680 Hz, are presented in the oscillation curves measured by CES in all
conditions, and 680Hz oscillations are only observed for approximately 40 ms after impact. While, 680Hz
small oscillation was not observed in CS output because of 200Hz bandwidth limit. To see the difference
in two sensor outputs in details, the fast Fourier transformation (FFT) was performed with two sensor output
data in two difference spindle conditions (stationary, 1000 rpm) as shown in Figure 11. Here ω n, ωa, ωo,
and ωh are the natural frequency (56 Hz), natural frequency (680 Hz) due to angular motion, operating
frequency (16.7 Hz) and harmonic frequency (33.4 Hz), respectively. Similarly, 680 Hz signal was not
found in CS outputs.

Figure 11. Results of FFT of two sensor outputs: (a) at spindle velocity 0 rpm (stationary) and (b) at spindle
velocity 1000 rpm.

9
 Figure 12. Experiment results: (a) stiffness and (b) damping ratio.

Static and dynamic stiffness and damping ratio of precision ball bearing spindle was measured by CS and
CES as shown in Figure 12. The stiffness was calculated by dividing the applied force with the displacement
and the damping ratio was calculated from logarithmic decrease method in the oscillation [19]. It can be
seen that the dynamic stiffness is slightly higher than the static stiffness and the dynamic damping ratio is
slightly lower than the static damping ratio.

Figure 13. Dynamic model of spindle system: MS, CL, KL, CA, KA, F(t), M(t), L are mass of the spindle
system, linear damper, linear stiffness, angular damper, angular stiffness, force, moment, and axial distance
between the ball bearing and measurement position.

Two frequencies, 56Hz and 680Hz, found in the impact response were related with linear and angular
displacement. It is expected that the linear and angular stiffness and damper in the ball bearing support are
key parameters to determine the dynamic motion of the spindle system. Thus, dynamic model of the spindle
system can be expressed by

M S rL (t )  CL rL (t )  K L rL (t )  F (t ), (5a)

M S(t )  CA(t )  K A (t )  M (t ), (5b)
rtotal  rL  L . (5c)

Where, rL and θ are the linear and angular displacement. Each stiffness can be obtained from the natural
frequencies, 56Hz (linear) and 680Hz (angular), and each damping ratio can be obtained as a result of

10
logarithmic oscillation decrease trend. Based on this system identification process, a dynamic model of this
spindle system was built in Matlab Simulink environment. The similar dynamic responses (1000 rpm
condition, Figure 8(a)) between measurement result and Simulink model result were achieved as seen in
Figure 14.

Figure 14. Comparison of spindle system responses.

5. Conclusion

A novel sensing method based on edge diffraction technique applicable to spindle metrology and spindle
dynamic system identification was investigated. Contrast to CS, (1) CES is not sensitive to is not sensitive
to materials, radius and surface quality of the measurement target, that is, the sensor sensitivity does not
change with respect to those parameters; (2) Because CES is an optical sensor, it has no parasitic electrical
noise issue between the probe and the target surface compared to CS; (3) the CES sensing bandwidth and
measuring range can be higher and longer than CS; (4) CES allows for a localized motion measurement by
reducing the beam width; and (5) CES is a low-cost sensor showing high sensing performance. The
precision ball bearing spindle system was characterized by impact test, and its dynamic model was
established based on the experiment. Thus, this technique can be used to five degree-of-freedom spindle
error measurements, spindle health monitoring, and dynamic characterization of the spindle system in a
near future. The curved edge diffraction model will be in-depth studied to fundamentally understand edge
diffraction behavior in reflection boundary, shadow boundary, and deep shadow boundary and to
characterize how this behavior affects CES performance.

Acknowledgements
The research was supported by National Science Foundation (Award Number: CMMI 1463502) through
the Tennessee Technological University. Similarly, this work was supported by the Center for
Manufacturing Research.

References

1. W. Gao, Precision nanometrology: Sensors and measuring systems for nanomanufacturing, Springer (2010).
2. K. Liu, M. Sun, T. Zhu, Y. Wu, Y. Liu, Modeling and compensation for spindle’s radial thermal drift error
on a vertical machining center, Int. J. Mach. Tools Manuf., 105, 58-67 (2016).

11
3. J. Kim, S. Zhao, G. Kim and S-K. Lee, Rolling bearing-suspended spindle run-out control using repetitive
control and adaptive feedforward cancelation, Int. J. Prec. Eng. Manuf., 14(12), 2171-2178 (2013).
4. J. Kim and S-K. Lee, Micro-patterning technique using a rotating cutting tool controlled by an
electromagnetic actuator, Int. J. Mach. Tools Manuf., 101, 52-64 (2016).
5. M. Madden, M. Aketagawa, T. Kumagai, Y. Maeda and E. Okuyama, Concurrent measurement method of
spindle radial, axial and angular motions using concentric circle grating and phase modulation
interferometers, Meas. Sci. Technol. 25, 094005 (2014).
6. Eric R. Marsh, Precision spindle metrology, DEStech Publications, Inc., 2017.
7. J. Fleming, A review of nanometer resolution position sensors: Operation and Performance, Sens. Actuators
A: Physical, 190, 106-126 (2013).
8. K. Liu, M. Sun, T. Zhu, Y. Wu, Y. Liu, Modeling and compensation for spindle’s radial thermal drift error
on a vertical machining center, Int. J. Mach. Tools Manuf., 105, 58-67 (2016).
9. Philip T. Smith, R. Ryan Vallance and Eric R. Marsh, Correcting capacitive displacement measurements in
metrology applications with cylindrical artifacts, Precision Engineering, Vol. 29, No. 3, pp. 324-335 (2005).
10. R. Ryan Vallance, Eric R. Marsh and Philip T. Smith, Effects of targeting spherical surfaces with capacitive
displacement sensors, Proceeding of American Society for Precision Engineering Annual Meeting (2004).
11. C. Lee, S. M. Mahajan, R. Zhao, and S Jeon, A curved edge diffraction-utilized displacement sensor for
spindle metrology, Rev. Sci. Instrum. 87, 075113 (2016).
12. C. Lee, R. Zhao, S. Jeon, A simple optical system for miniature spindle runout monitoring, Measurement,
102, 42-46 (2017).
13. C. Lee, S-K. Lee and J. A. Tarbutton, Novel design and sensitivity analysis of displacement measurement
system utilizing knife edge diffraction for nanopositioning stages. Rev. Sci. Instrum. 85, 095113 (2014).
14. ChaBum Lee, Large-range nano-scanning devices based on optical sensing technology, Book chapter,
Advanced Mechatronics and MEMS Devices II, Springer: 10.1007/978-3-319-32180-6_22 (2017).
15. C. Lee, J. A. Tarbutton and S-K. Lee, Positioning control effectiveness of optical knife edge displacement
sensor-embedded monolithic precision stage, Sen. Actuators A, 233, 390-396 (2015).
16. E. Hecht, Optics, 4th ed. (Addison-Wesley Longman Inc., 1987).
17. B. A. Davis and G. S. Brown, Diffraction by a randomly rough knife edge, IEEE Trans. Antennas Propag.
50(12), 1769–1778 (2002).
18. G. L. James, Geometrical Theory of Diffraction for Electromagnetic Waves, 3rd Ed., (IET, 1979).
19. Katsuhiko Ogata, System Dynamics, Pearson Prentice Hall 2004.

12