Mechanical Systems and Signal Processing 93 (2017) 225240

Contents lists available at ScienceDirect

Mechanical Systems and Signal Processing

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

Development and testing of an integrated smart tool holder for

four-component cutting force measurement
Zhengyou Xie, Yong Lu , Jianguang Li
School of Mechatronics Engineering, Harbin Institute of Technology, Harbin, China

a r t i c l e i n f o a b s t r a c t

Article history: Cutting force measurement is a significant requirement for monitoring and controlling the
Received 18 September 2016 machining processes. Hence, various methods of measuring the cutting force have been
Received in revised form 21 January 2017 proposed by many researchers. In this study, an innovative integrated smart tool holder
Accepted 28 January 2017
system based on capacitive sensors is designed, constructed and tested, which is capable
of measuring triaxial cutting force and a torque simultaneously in a wireless environment
system. A standard commercial tool holder is modified to make itself be the force sensing
Keywords:
element that has advantages of simple structure and easy machining. Deformable beams
Smart tool holder
Rotating dynamometer
are created in the tool holder, and the tiny deformations of which used to calculate the
Cutting force measurement four-component cutting force are detected by six high precision capacitive sensors. All
Milling process the sensors and other electronics, like data acquisition and transmitting unit, and wireless
Deformable beam power unit, are incorporated into the tool holder as a whole system. The device is intended
Capacitive sensor to be used in a rotating spindle such as in milling and drilling processes. Eventually, the
static and dynamic characteristics of the smart tool holder have been determined by a ser-
ies of tests. Cutting tests have also been carried out and the results show it is stable and
practical to measure the cutting force in milling and drilling processes.
2017 Elsevier Ltd. All rights reserved.

1. Introduction

Nowadays, manufacturing enterprises face the growing demands of increased product quality, reduced cost and higher
processing efficiency, so cutting process monitoring becomes particularly important [1,2]. As one of the most significant
machining process variables, the information of cutting force generated in metal cutting can be used to understand the prin-
ciple of chip formation [3], optimize cutting parameter [4], design tool geometry [5], monitor tool wear or failure [6,7], and
detect and suppress chatter vibration [8]. Thus, various methods to measure cutting force in turning, milling and drilling
have been proposed by researchers.
Commonly, table dynamometers are the most generally used equipments in machining operations, since they can provide
highly accurate and effective cutting force measurement, and are easy to design and manufacture. Many types of table
dynamometers are developed based on strain gauge sensor [911], especially piezoelectric sensor [12,13]. However, when
used, the dynamometer is clamped onto the machine table and a workpiece is mounted on top of the dynamometer, which
undoubtedly limits the geometry and dimension of the workpiece. Besides, considering table dynamometers complicated
installation and high cost, they are more suitable for laboratory or experimental use rather than for practical application
on production machines.

Corresponding author.
E-mail address: luyhit@163.com (Y. Lu).

http://dx.doi.org/10.1016/j.ymssp.2017.01.038
0888-3270/ 2017 Elsevier Ltd. All rights reserved.
226 Z. Xie et al. / Mechanical Systems and Signal Processing 93 (2017) 225240

Due to the limitations of current commercial sensor systems and the significant challenges for flexibility and reconfigura-
bility, sensor integration techniques are being accepted. Byrne [14] proposed an integrated force sensor solution for drilling
operations by integrating two piezoelectric force sensing rings into a direct driven motor spindle. Similarly to Jun et al. [15],
who integrated a force ring to the spindle housing to measure triaxial cutting force, and finished the evaluation work on the
sensor performance. Totis et al. [16] developed a rotating dynamometer for milling process by installing commercial piezo-
electric triaxial force sensors between the modular cartridge and the cutter body. Also, Ma et al. presented several sensor-
integrated methods for measuring the feed and transverse forces [17] and the torque [18] in milling process, a thin film
polyvinylidene fluoride (PVDF) piezoelectric strain sensor was mounted to the shank of a cutting tool, and all the electronics
and batteries were located in a metal housing that was mechanically attached to the tool holder.
There are already some kinds of commercial piezoelectric rotating dynamometers developed by several companies, like
Kistler Instrument Corporation [19], and some of these products can measure four-component cutting force simultaneously
with good dynamic performance. However, these products are mostly very expensive. Thus, some other kinds of sensors are
adopted for cutting force measurement. Albrecht et al. [20] presented a method of measuring just radial cutting force by
integrating capacitive sensors into the spindle to measure the static and dynamic variations of the gap between the sensor
head and the rotating spindle shaft. Rizal et al. [21] developed a force sensing element on which strain gauges were mounted,
and then integrated it on a tool holder to measure three cutting forces in milling. Similar work was done by Suprock [22] and
Nichols [23], who designed a strain gauge-based tool holder for torque measurement, which was defined as smart tool
holder in their studies. All the rotating dynamometers developed by these researchers were tested and confirmed to have
good performance for cutting force measurement. Nevertheless, to authors best knowledge, the most present devices based
on capacitive sensors or strain gauges can only measure two or three-axis cutting force, or torque. Too many strain gauges
are needed if they are used to measure more axises forces, just like the dynamometer developed by Rizal et al. [21], in which
three full Wheatstone bridge circuits were designed and 24 strain gauges in total were integrated, which doubtless increases
the system complexity and reduces the reliability.
This present study makes a further contribution in addressing the issues, dealing with the design and construction of an
integrated smart tool holder system, which is capable of measuring triaxial cutting force and torque simultaneously in
milling and drilling operations. A standard commercial tool holder is modified to form deformable beams and make itself
be the force sensing element. Six capacitive sensors are used and integrated to the force sensing tool holder to detect the
beams deformations, and then calculate the four-component cutting force. The rotating dynamometer is designed to mea-
sure forces not higher than 1000 N of cutting and 100 N
m of torque with up to 3000 rpm of spindle speed when using a two-
inserts tool. Moreover, the device has advantages of simple structure and low cost, which make it easy to manufacture and
suitable for industrial use. A series of tests have been carried out, and the results show it is stable and practical to measure
the cutting force in milling and drilling processes.

2. Force component in milling operation

When the cutters cut the workpiece material, the cutting force acts on the milling cutters directly, and then transmits to
the tool holder. The force on each cutter which consists of main cutting force, thrust force and perpendicular cutting force is
complicated and time-varying. However, to facilitate detection and analysis, the cutting force can be decomposed into three-
dimensional orthogonal forces and a torque at a certain time as illustrated in Fig. 1. A standard right-handed Cartesian coor-
dinate is established. The axis of rotation is set as Z-axis with the positive direction towards machine spindle. Thus, this work
is to design and construct an integrated smart tool holder system as a rotating dynamometer for measuring these four com-
ponent cutting forces that are an axial force (Fz), two mutually perpendicular radial forces (Fx and Fy), and a torque (T, same as
Mz). By assembling different types of cutter arbors and cutters, the smart tool holder can be applied in not only milling but
drilling operations.

3. Design and construction

3.1. Design and model of force sensing tool holder

The smart tool holder for cutting force measurement is regarded as an integrated rotating dynamometer. The structure
form and dimensions of the force sensing element have great influence on the dynamometers stiffness, sensitivity and accu-
racy of measurement, and hence they were taken into consideration to ensure the performance. Fig. 2 illustrates the geomet-
rical structure of the force sensing tool holder. The prototype of the tool holder used in this paper is BT50SLN32-150 (Harbin
Measuring & Cutting Tool Group Co. Ltd.) with inner diameter (d) 32 mm, outer diameter (D) 70 mm and height of cylindrical
part (H) 113 mm. And the material is 40Cr. Four horizontal deformable beams and four vertical deformable beams which can
make deformation concentrated are created by machining grooves along the radius on the tool holder. The deformations in
the proper positions of the beams can be detected by high precision capacitive sensors, and then the cutting forces and tor-
que can be calculated by these deformations. A previous similar prototype, which was studied by the authors in a prelim-
inary work [24], was proven to fulfill the force-measuring needs. Compared to the previous structure, the grooves were
made along a line parallel to the diameter in this paper instead of along the radial direction to make the structure much
 Z. Xie et al. / Mechanical Systems and Signal Processing 93 (2017) 225240 227

Fig. 1. Components of the cutting force in milling.

Fig. 2. Model of force sensing tool holder: (a) perspective view (b) enlarged view and (c) dimensions.

easier to manufacture. The sizes of the sensing tool holder were determined, including width (w1) 20 mm and height (h1)
13 mm of upper grooves, width (w2) 23.5 mm and height (h2) 2 mm of under grooves, gap between upper and under grooves
(t) 4.5 mm and distance from these grooves to the grasping ring (s) 18 mm.
Considering the decoupling needs for three forces and torque, six deformation detection positions (P1P6) were set on the
beams as shown in Fig. 3. Each deformation is detected by one capacitive sensor of six in total. Four sensors were mounted in
the middle of the horizontal deformable beams to measure these positions deformations along the Z-axis, while two sensors
were mounted on the vertical deformable beams to measure the torsional deformations of these positions.
Fig. 4 is the diagram of the tool holder under each force respectively. As a result of the structure symmetry, four defor-
mations on positions P1, P2, P3 and P4 have the same magnitude and direction under the action of axial force, while two tor-
sional deformations on positions P5 and P6 are almost zero for that vertical beams are only under compression. When there is
only a radial force, take Fx for instance, acting on tool holder, which is similar to the effect that a bending moment acts on
deformable beams, one side of tool holder is under compression and the other is under tension, and their magnitude are
equal. As a consequence, the deformations on positions P1 and P3 are equal in magnitude and opposite in direction. When
a torque acts on the tool holder, it is obvious that the positions P5 and P6 have equal deformations in the opposite direction.
Therefore, every force and torque at any time can be solved out based on six sets of deformation data by using (1) due to
the structure symmetry.
228 Z. Xie et al. / Mechanical Systems and Signal Processing 93 (2017) 225240

Fig. 3. Deformation detection positions on deformable beams.

Fig. 4. Diagrams of (a) axial force (b) radial force and (c) torque action on tool holder.

8
Dd1 Dd2 Dd3 Dd4
>
> F z k1
>
> 4
>
< F k Dd3 Dd1
x 2 2
1
>
> F y k3 Dd4 2 Dd2
>
>
>
:
T k4 Dd6 2 Dd5

where Ddi (i = 16) is the deformation on each position Pi (i = 16), ki (i = 14) is the coefficient related to structure
sensitivity.
Theoretically, if the structure of force sensing tool holder and positions of these deformation detection are completely
symmetric, Fx, Fy and T calculated by Eq. (1) are zeros when there is only an axial force Fz acting on tool holder and so to
other forces. It shows great force decoupling performance using this kind of force sensing structure to measure multi-
directional forces.

3.2. Finite element analysis

3.2.1. Static analysis

In order to determine the deformation distribution of horizontal and vertical deformable beams subjected to each force,
ANSYS Workbench has been used to perform the static analysis. Fig. 5(a) shows the structural finite element mesh. The
model was established and meshed to tetrahedrons with 192,805 nodes and 132,177 elements. Considering the real con-
straint and load conditions, the taper shank part of the tool holder was selected to fix and the axial force Fz, radial force, take
Fx for instance, and torque T acted at the tool holder individually. According to the magnitude ranges of cutting forces in
 Z. Xie et al. / Mechanical Systems and Signal Processing 93 (2017) 225240 229

Fig. 5. Finite element analysis (a) mesh model; deformation nephograms under (b) axial force Fz; (c) radial force Fx; and (d) torque T.

common milling and drilling processes, the smart tool holder was designed to measure cutting force up to 1500 N in axial
direction, 1000 N in radial direction, and torque up to 100 N
m. Thus, the forces were set to the maximum values in ANSYS.
Then, the deformation nephograms of beams under each force were achieved as shown in Fig. 5(bd).
Fig. 5(b) shows all the four horizontal beams deformations have the same magnitude and direction due to axial force Fz.
Moreover, under the action of radial force Fx, the deformations of the two horizontal beams along the force direction are
equivalent while the directions are opposite, and other two horizontal beams deformations are near to zero as shown in
Fig. 5(c). Fig. 5(d) shows the deformation distribution of the vertical beams due to the torque (T). All the analysis results
fit the model well. The structural sensitivities of the force sensing tool holder were obtained based on static analysis as sum-
marized in Table 1.
Moreover, to determine whether the tool holder would break if the measuring ranges are exceeded, the structural
strength of the tool holder has been analyzed using ANSYS Workbench. The tool holder was under the maximum Fx of
1000 N, Fy of 1000 N, Fz of 1500 N and T of 100 N
m simultaneously, and the result showed the maximum equivalent stress
was 40.885 Mpa, which was far less than the yield strength of the 40Cr material of 785 Mpa. It means the tool holder is safe
enough when the forces exceed the measuring range.

3.2.2. Free modal analysis

Milling is a complex dynamic process with a changing cutting force in size and direction. Its necessary to analyze the
dynamic characteristics of the force sensing tool holder. Vibration modal is one of the intrinsic characteristics of elastic struc-
tures, and the natural frequencies and vibration modes can be obtained by modal analysis. To ensure the dynamic perfor-
mance, the first order natural frequency (fn, Hz) of the force sensing tool holder should be at least four times the tooth
passing frequency (fe, Hz) which is highly relevant to spindle speed (n, rpm) [25]. The relationship can be expressed as
follows:

nZ
fn P 4
fe 4
2
60

where Z is the number of teeth. Thus, the tooth pass excitation frequency would be 100 Hz if the rotating dynamometer was
designed to be applied in machining process with a spindle speed up to 3000 rpm when using a two-inserts tool, and the
natural frequency should be not less than 400 Hz.
Free modal analysis of the force sensing tool holder has been done using ANSYS Workbench without regard to any con-
straints and loads. Table 2 shows the nine orders natural frequencies, where the former six are rigid body modes that are
230 Z. Xie et al. / Mechanical Systems and Signal Processing 93 (2017) 225240

Table 1
Structural sensitivities of force sensing tool holder.

Axial direction (lm/N) Radial direction (lm/N) Torsional direction (lm/N
m)

Sensitivity:Dd=DFT 4.299  104 2.743  103 1.322  102

Table 2
Natural frequencies of force sensing tool holder based on free modal analysis.

Order Natural frequency (Hz)

16
7 3845.4
8 3862.7
9 5942.6

related to six degrees of freedom. The first natural frequency of the tool holder without considering rigid body modes is
3845.4 Hz that is high enough.

3.3. Construction and fabrication of the smart tool holder

3.3.1. Design and mounting of capacitive sensors

Each capacitive sensor used to measure the deformation of beams can be considered as a parallel plate capacitive dis-
placement sensor, which is composed of a capacitance detection circuit and two parallel conductive plates, one of which
is mounted on deformable beams as the movable plate and the other is glued on a mounting block as the fixed plate. Then
the mounting block is installed on the tool holder through bolts as illustrated in Fig. 6(a). The capacitance (C) would be given
by [26]
e0 er A
C 3
d
where e0 8:854  1012 F=m is the permittivity of free space, er is the relative permittivity of the dielectric material
(1.000585 of the air), A is the effective area of the plates, and d is the distance between the plates.The movable plate would
move with the deformable beams when a force acts on the tool holder, thus the plate distance changes and so would the
capacitance. Assuming that the movable plate moves Dd, thus the distance changes to d  Dd, the increment of capacitance
(DC) can be calculated by
e0 er A e0 er A
DC  4
d  Dd d
    1
DC e0 er S e0 er S e0 er S Dd Dd
 1 5
C d  Dd d d d d

Fig. 6. Photograph of (a) capacitance plates installation scheme; (b) mounting block and (c) installation of capacitive sensors on tool holder.
 Z. Xie et al. / Mechanical Systems and Signal Processing 93 (2017) 225240 231

Table 3
Parameters and sensitivities of the capacitive sensors.

Axial direction Radial direction Torsional direction

Area of the plates: A (mm2) 165.046 110
Distance between the plates: d (lm) 50 40
Initial capacitance: C (pF) 29.24 24.36
Sensitivity:DC=Dd(pF/lm) 0.585 0.609
Sensitivity:DU=DFT 1.01 mV/N 6.44 mV/N 32.17 mV/N
m

Then Eq. (5) can be written as following by using Taylor expansion.

"  2  3 #
DC Dd Dd Dd Dd
1


6
C d d d d

The sensitivity of the capacitance with distance can be described as Eq. (7), when Dd is small enough compared to d.
DC C
7
Dd d
Then, the sensitivity of each capacitive sensor under the cutting force can be attained as
DU DU DC Dd


8
DFT DC Dd DFT
where DU=DC is the sensitivity of the capacitance detection circuit which was designed to be 4.0 V/pF, and Dd=DFT is
the structural sensitivity summarized in Table 1.
Fig. 6(b) shows a fabricated prototype of the mounting block with two capacitance plates which were made based on
printed circuit board (PCB) technology. The plates for axial deformation detection are circular, 14.5 mm in diameter, and
the others for torsional deformation detection are rectangular, 10  11 mm2 in dimension. The plate distances of these
two kinds of capacitive sensors are set to 50 lm and 40 lm respectively, and the values are guaranteed by using feeler gauge
when assembling. Fig. 6(c) shows the installation of capacitive sensors on the tool holder. The capacitance detection circuits
were constructed based on CMOS technology, and the same interface circuit was used to develop a closed-loop accelerom-
eter [27], which showed low noise and high resolution. Moreover, the circuit was designed to have a frequency bandwidth of
3.0 kHz that was high enough for force measurement. The parameters and sensitivities of the capacitive sensors are summa-
rized in Table 3.

3.3.2. Construction and fabrication

The construction of the model rotating dynamometer integrated in the tool holder is illustrated in Fig. 7(a and b) shows
the photograph of the complete developed smart tool holder. The components of the complete system comprise a force sens-
ing tool holder, electrical system mounting bracket, mounting block for capacitance plates, cover and wireless power unit.
The electrical system mounting bracket is used to mount power management circuit, signal acquisition and wireless trans-
mission circuit. The materials of tool holder and mounting block are 40Cr and structural steel, and the electrical system
mounting bracket and cover are made of plastic. Since the device was designed based on a commercial tool holder type
BT50SLN32-150, it can be equipped with different standard tools that the original one could, thus providing a flexible
and reconfigurable machining process.
The data acquisition and wireless transmission system were developed based on STM32 and AR9331. The force data were
collected and transmitted to the host computer via Wi-Fi at a sampling rate of 5 kHz for further analysis and processing.
Based on the LabVIEW, PC system has been designed and realized real-time data display, storage and processing.

4. Testing

4.1. Static calibration test

To determine the relationship between the input force and output voltage static calibration test is essential. A calibration
platform was setup based on a vertical machining center (Zhongjie VMC1600B) and several static calibration tests were
done. According to design analysis of force sensing tool holder, the calibration loads are based on full scale output (FSO) with
an incremental step of 100 N in force and 10 N
m in torque. The output voltage values of six capacitive sensors were all
recorded and the increments were calculated in each load interval.
The output data of six sensors were recombined and calculated by using Eq. (9), and the calibration curves of the axial
force Fz, the radial forces Fx and Fy, and the torque T were achieved as shown in Fig. 8(ad).
232 Z. Xie et al. / Mechanical Systems and Signal Processing 93 (2017) 225240

Fig. 7. Photograph of (a) exploded view (b) a completely smart tool holder.

Fig. 8. Static calibration curves of (a) axial force Fz, (b) radial force Fx, (c) radial force Fy and (d) torque T.

8
>
> DU z DU1 DU2 4 DU3 DU4
>
>
< DU x D U 3  D U 1
2
9
> DU DU
> DU y 4 2 2
>
>
:
DU t DU6 2 DU5
The calibration curves show that the smart tool holder had high sensitivities with 0.74 mV/N, 6.32 mV/N, 7.64 mV/N and
19.32 mV/N
m in Fz, Fx, Fy and T directions respectively. However, actual dimensions and initial distance between two capac-
itance plates cannot be guaranteed exactly the same as designed due to machining and assembly errors, which caused
changes in sensitivity of each sensor. Besides, the structure of sensing tool holder was not symmetric strictly. These factors
 Z. Xie et al. / Mechanical Systems and Signal Processing 93 (2017) 225240 233

resulted in the errors of sensitivities between the designed values and the experimental results. Also, the cross interference
inevitably arose, which means the output voltages in one direction were affected not only by the force in this direction but
also by forces in other directions as expressed in Eq. (10).
8
> DU z DU zz DU zx DU zy DU zT
>
>
< DU x DU xz DU xx DU xy DU xT
10
>
> DU y DU yz DU yx DU yy DU yT
>
:
DU T DU Tz DU Tx DU Ty DU TT
where DU ij i; j x; y; z; t is the output voltage increment in Fi direction under force Fj. And it can be described as Eq. (11)
due to the linear system.

DU ij cij F j bij i; j x; y; z; t 11
where cij is the output sensitivity in F i direction under force Fj, and bij is the intercept. Then Eq. (10) can be written as
following by substituting Eq. (11).
2 3 2 3 2 3 2 3
DU z czz czx czy czt Fz bz
6 DU 7 6 c cxx cxy cxt 7 6 7 6 7
6 x7 6 xz 7 6 F x 7 6 bx 7
6 76 7
6 76 7 12
4 DU y 5 4 cyz cyx cyy cyt 5 4 F y 5 4 by 5
DU T ctz ctx cty ctt T bt

bi biz bix biy bit i z; x; y; t 13

Therefore, each force in different direction can be calculated as Eq. (14). Where all the coefficients were obtained by static
calibration test. When output voltages of six capacitive sensors under cutting forces are achieved, these forces can be decou-
pled by using Eq. (15).

F41 C1
44
DU41  b41 14

2 3 2 3 02 3 2 31
Fz 1270:096 114:899 150:091 40:511 DU z 0:067
6 F 7 6 30:186 156:331 7
0:158 7 B6 B 7 6 7C
6 x7 6 3:727 6 DU x 7 6 0:633 7C
6 76 7
B6 76 7C 15
4 F y 5 4 51:565 9:918 125:620 15:743 5 @4 DU y 5 4 0:158 5A
T 0:142 8:532 0:056 51:827 DU T 0:039
To assess the effect of decoupling, some tests were done. Some forces were applied to the smart tool holder in four direc-
tions respectively, and outputs of capacitive sensors were achieved and then substituted to Eq. (15) to calculate the forces. By
comparing the force values between applied and calculated, the decoupling errors can be calculated as follows.
8 F F
< iout
F
jload
 100%; i j
eij jload
i; j x; y; z; T 16
: F iout  100%; ij
F jload

where Fj-load is the force applied in j direction, Fi-out is the force calculated in i direction, and eij is the measurement error
when i = j, while the cross talk error when i j.
The results of decoupling test were described in Table 4. The results showed the measurement errors were small with a
maximum of 1.11%. The maximum cross talk error was 11.63% occurred on torque T to axial force Fx, and all the others did
not exceed 6.4%. It means that the smart tool holder is acceptable to use for cutting force measurement, and the results were
in agreement with Jun et al. [15] in whose work a spindle-based force sensor was evaluated and confirmed to have a cross
talk error maximum of 12%.

Table 4
The results of decoupling test.

Axes Load Output after decoupling Error (%)

Fz (N) Fx (N) Fy (N) T (N
m) Fz Fx Fy T
Fz 1500 N 1505.79 3.16 0.20 0.59 0.39 (N/N) 0.21 (N/N) 0.013 (N/N) 0.039 (N
m/N)
Fx 1000 N 63.95 1002.73 1.97 3.07 6.40 (N/N) 0.27 (N/N) 0.20 (N/N) 0.31 (N
m/N)
Fy 1000 N 42.19 13.36 988.94 6.71 4.22 (N/N) 1.34 (N/N) 1.11 (N/N) 0.67 (N
m/N)
T 100 N
m 11.63 2.73 2.69 99.52 11.63 (N/N
m) 2.73 (N/N
m) 2.69 (N/N
m) 0.48 (N
m/N
m)
234 Z. Xie et al. / Mechanical Systems and Signal Processing 93 (2017) 225240

4.2. Dynamic calibration test

4.2.1. Pulse test under free state

The dynamics identification of the smart tool holder under free state was performed by means of pulse test. The device
was hung up along Z-axis to simulate a free state in X and Y directions, as shown in Fig. 9, and then excited by using a modal
impact hammer (Longke type LK1427, sensitivity of 4.19pC/N). An accelerometer (Donghua type DH186, sensitivity of
10 mV/m
s2 and measurement bandwidth of 0.55 kHz) was connected to the smart tool holder. The input (hammer) and
output (accelerometer) signals were acquired by a dynamic signal collection system (Donghua type DH8302), and the modal
analysis was performed in order to derive the frequency response function of the smart tool holder. The result of free modal
test shows the first two resonance peaks located at 3710.9 Hz and 3867.1 Hz, see Fig. 10. Compared to that from finite ele-
ment analysis, the relative errors of these test results are 3.6% and 0.1%. It means that the simulation and testing have a good
agreement.

4.2.2. Step test on machine

Free modal analysis and experiment results show that the structure of smart tool holder itself had a high natural fre-
quency. However, the sensing and transmission system, and the actual use condition were all not considered. Thus, the step
response calibrations were done by mounting the smart tool holder on spindle of a vertical machining center (VMC1600B),
and the step force was generated by fracture of the marble strip. Similar method was used and proved effective by Xu [28], in
whose work the step force was generated by cutting the string connected with a weight.
Fig. 11 illustrates the experimental setup. The smart tool holder under test was mounted on the spindle. A marble strip
was put on the upholder which was fixed on the machine table in Fig. 11(a and b) or connected with the tool holder in Fig. 11
(c). Move the spindle or the sliding table slowly to press the marble strip. A negative step force was generated once the mar-
ble strip was fractured. The output voltages were collected and frequency response functions were obtained by using power
spectrum analysis.
Figs. 12 shows the frequency response functions for each direction. It is observed that the natural frequencies of the four
directions are approximately 466.7 Hz, 402.2 Hz, 404.8 Hz and 435.3 Hz, which are much lower than the results of free
modal experiments. When the tool holder is mounted on the machine spindle, they are considered as a whole system. It
is probable that the low test results are caused by the spindle system. Ma et al. [17] and Albrecht et al. [20] have imple-
mented several impact hammer tests on the tool holder-spindle system in their papers, respectively, and the results were
both around 400 Hz. However, the smart tool holder system can still be applied in machining process up to 3000 rpm of spin-
dle speed when using a two-inserts tool. Some schemes can be used, like a Kalman filter [20], if a higer spindle speed is
needed.

4.3. Cutting test

To evaluate the devices performance of cutting force measurement in real cutting process, two kinds of cutting tests, end
milling and drilling, were carried out. A 32 mm diameter cutter arbor (BAP400R-35-160-C32) with two PVD coated carbide
inserts (Mitsubishi APMT1604PDER) was selected and assembled in the smart tool holder. All these experiments were per-
formed by cutting 45 steel under dry cutting condition using a vertical machining center (VMC1600B) as illustrated in
Fig. 13. The cutting forces measured by the smart tool holder were transmitted to the upper computer #1 and recorded.
A reference table dynamometer (Kistler 9257B) was mounted between the workpiece and the machine table, and the cutting
forces measured by which were recorded by upper computer #2 and considered as reference forces.

Fig. 9. Experimental setup for pulse tests under free state.

 Z. Xie et al. / Mechanical Systems and Signal Processing 93 (2017) 225240 235

Fig. 10. Frequency responses of smart tool holder under free state.

Fig. 11. Dynamic calibration experimental setup for (a) axial force, (b) radial force, and (c) torque directions.

Fig. 12. Frequency responses for (a) axial force Fz, (b) radial force Fx, (c) radial force Fy and (d) torque T directions.
236 Z. Xie et al. / Mechanical Systems and Signal Processing 93 (2017) 225240

Fig. 13. Photography of cutting test.

4.3.1. End milling

Two mill tests were performed with the machining parameters as follows: axial depth of cut (ap) of 0.5 mm, radial depth
of cut (ae) of 10 mm, feed speed (f) of 100 mm/min, and spindle speed (n) of 1000 rpm and 1800 rpm. The cutting forces in
each test were measured by the smart tool holder and the reference dynamometer simultaneously, and the comparison anal-
ysis of force data in three rotations was done as shown in Figs. 1417. Considering the measuring coordinate system of the
smart tool holder is rotating during cutting process, while the one of the reference dynamometer is fixed, the radial forces (Fx
or Fy) cannot be compared directly. The resultant force in radial direction (Fr) was calculated by using Eq. (17) and the com-
parison results are shown in Fig. 14(b) and 15(b).

q
Fr F 2x F 2y 17

The results show that the cutting forces caused by each insert were measured by the smart tool holder reliably and the
amplitude values are very close to those measured by the reference dynamometer. To evaluate the accuracy of the force
measurements, the relative errors were calculated by using Eq. (18).

jF av gST  F av gREF j
eF  100% 18
jF av gREF j
where Favg-ST and Favg-REF are the average forces measured by smart tool holder and reference dynamometer, respectively.
The relative errors of axial force and radial resultant force measurement are 9.3% and 4.6% at 1000 rpm, 8.9% and 6.8% at
1800 rpm.
Fig. 15 is the comparison diagram of amplitude-frequency characteristics between two force-measuring devices at a spin-
dle speed of 1000 rpm. It shows the force signals measured by the smart tool holder have the spindle frequency of 16.5 Hz,
and the tool passing frequency of 33.0 Hz. When the spindle speed increases to 1800 rpm, the spindle frequency and tool
passing frequency also increase to 29.6 Hz and 59.2 Hz, respectively, which are all almost equal to those measured by the
reference dynamometer. The torque signals in end milling were not analyzed and compared since Kistler type 9257B
dynamometer can only detect the torque on a fixed processing position.

Fig. 14. Comparison of (a) axial force Fz and (b) radial resultant force Fr measurement result in milling (n = 1000 rpm).
 Z. Xie et al. / Mechanical Systems and Signal Processing 93 (2017) 225240 237

Fig. 15. Comparison of amplitude-frequency characteristic in milling (n = 1000 rpm).

Fig. 16. Comparison of (a) axial force Fz and (b) radial resultant force Fr measurement result in milling (n = 1800 rpm).

Fig. 17. Comparison of amplitude-frequency characteristic in milling (n = 1800 rpm).

4.3.2. Drillng
To perform the drilling test, a workpiece was preliminarily machined with a hole of 32 mm in diameter and 10 mm in
depth on it. The cutter diameter of the tool the smart tool holder used was 35 mm, which means the radial depth of cut
in one side was 1.5 mm. The feed speed was set to 30 mm/min until feeding 6 mm depth, and the spindle speed was
1000 rpm. Figs. 1820 show the comparisons of torque, axial force and radial resultant force measurement results between
smart tool holder and reference dynamometer, respectively. The data curves are similar and the amplitude values are very
close in each diagram. Detailed view of the cutter cuts in workpiece is shown in Fig. 18, which indicates that the torques
measured by the two devices have a good agreement.
The force data in 46 s were selected and calculated the average values. Then the relative errors of torque, axial force and
radial resultant force measurement were calculated out which were 8.07%, 9.05% and 4.6%. Both the two sensing systems had
the tool passing frequency of 16.8 Hz as illustrated in Fig. 21.
The test results of the end milling and drilling clearly show the developed smart tool holder has a good cutting force mea-
suring performance. The relative errors of each force measurement did not exceed 10%, which can be accepted.
238 Z. Xie et al. / Mechanical Systems and Signal Processing 93 (2017) 225240

Fig. 18. Comparison of torque T measurement result in drilling.

Fig. 19. Comparison of axial force Fz measurement result in drilling.

Fig. 20. Comparison of radial resultant force Fr measurement result in drilling.

Fig. 21. Comparison of amplitude-frequency characteristic in drilling.

 Z. Xie et al. / Mechanical Systems and Signal Processing 93 (2017) 225240 239

5. Conclusion

In this work, a smart tool holder based on capacitive sensors for measuring three orthogonal force components (Fx, Fy and
Fz) and the torque (T) was designed and developed, which can be used as a rotating dynamometer in milling or drilling pro-
cess. A standard commercial tool holder was modified to make itself be the force sensing element, which has advantages of
simple structure and easy machining, then six capacitive sensors and other electronics, like data acquisition and transmitting
unit, and wireless power unit, were incorporated into the tool holder as a whole system. The device is compatible with dif-
ferent standard cutting tools, thus providing a flexible and reconfigurable machining process.
A series of tests have been carried out to evaluate the performance of the developed smart tool holder. The results are
summarized as following:

 The static calibration tests showed that the new device can measure cutting forces up to 1500 N of axial force, 1000 N of
radial force and 100 N
m of torque with high sensitivities of 0.74 mV/N, 6.32 mV/N, 7.64 mV/N and 19.32 mV/N
m in Fz, Fx,
Fy and T directions respectively. Moreover, the cross sensitivity errors are low with a maximum of 6.40% except that the
torque T to axial force Fz.
 The pulse tests showed the natural frequency of the tool holder itself under free state is 3.7 kHz, which has a good agree-
ment with the result of finite element analysis. And the step tests showed the natural frequencies of the tool holder-
spindle system in all the force orientations are approximately 466.7 Hz, 402.2 Hz, 404.8 Hz and 435.3 Hz, which means
its dynamic range is suitable in machining process up to 3000 rpm of spindle speed when using a two-inserts tool.
 The tests of end milling and drilling were performed and the results showed the smart tool holder system is capable of
measuring cutting force timely and accurately with relative errors not exceeding 10% compared to a commercial table
dynamometer.

In general, the device could be successfully applied in milling and drilling processes for cutting force measurement. The
force data measured can be used, for example, to monitor tool wear or breakage, thus the developed smart tool holder can
significantly benefit machining processes.

Acknowledgments

This work was supported by the National High Technology Research and Development Program of China (863 Program)
under Grant no. 2013AA041107.

References

[1] S.Y. Liang, R.L. Hecker, R.G. Landers, Machining process monitoring and control: the state-of-the-art, J. Manuf. Sci. Eng. 126 (2004) 297310.
[2] R. Teti, K. Jemielniak, G.O. Donnell, D. Dornfeld, Advanced monitoring of machining operations, CIRP Ann. Manuf. Technol. 59 (2010) 717739.
[3] S. Sun, M. Brandt, M.S. Dargusch, Characteristics of cutting forces and chip formation in machining of titanium alloys, Int. J. Mach. Tools Manuf 49
(2009) 561568.
[4] U. Yalcin, A.D. Karaoglan, I. Korkut, Optimization of cutting parameters in face milling with neural networks and Taguchi based on cutting force, Surf.
Rough. Temp. Int. J. Prod. Res. 51 (2013) 34043414.
[5] F.A. Almeida, F.J. Oliveira, M. Sousa, A.J.S. Fernandes, J. Sacramento, R.F. Silva, Machining hardmetal with CVD diamond direct coated ceramic tools:
effect of tool edge geometry, Diam. Relat. Mater. 14 (2005) 651656.
[6] P. Bhattacharyya, D. Sengupta, S. Mukhopadhyay, Cutting force-based real-time estimation of tool wear in face milling using a combination of signal
processing techniques, Mech. Syst. Sig. Proc. 21 (2007) 26652683.
[7] S.C. Lin, R.J. Lin, Tool wear monitoring in face milling using force signals, Wear 198 (1996) 136142.
[8] D.H. Chin, M.C. Yoon, Cutting force monitoring in the end milling operation for chatter detection, Proc. Inst. Mech. Eng. Part B J. Eng. Manuf. 219 (2005)
455465.
[9] S. Yald, F. nsa Ar, H. Sa Lam, H. Isik, Design, development and testing of a four-component milling dynamometer for the measurement of cutting force
and torque, Mech. Syst. Sig. Proc. 21 (2007) 14991511.
[10] Y. Li, Y. Zhao, J. Fei, Y. Zhao, X. Li, Y. Gao, Development of a tri-axial cutting force sensor for the milling process, Sensors 16 (2016) 405.
[11] S. Karabay, Analysis of drill dynamometer with octagonal ring type transducers for monitoring of cutting forces in drilling and allied process, Mater.
Des. 28 (2007) 673685.
[12] G. Totis, O. Adams, M. Sortino, D. Veselovac, F. Klocke, Development of an innovative plate dynamometer for advanced milling and drilling applications,
Measurement 49 (2014) 164181.
[13] N. Tounsi, A. Otho, Dynamic cutting force measuring, Int. J. Mach. Tools Manuf. 40 (2000) 11571170.
[14] G. Byrne, G.E.O. Donnell, An integrated force sensor solution for process monitoring of drilling operations, CIRP Ann. Manuf. Technol. 56 (2007) 8992.
[15] M.B. Jun, O.B. Ozdoganlar, R.E. Devor, S.G. Kapoor, A. Kirchheim, G. Schaffner, Evaluation of a spindle-based force sensor for monitoring and fault
diagnosis of machining operations, Int. J. Mach. Tools Manuf 42 (2002) 741751.
[16] G. Totis, G. Wirtz, M. Sortino, D. Veselovac, E. Kuljanic, F. Klocke, Development of a dynamometer for measuring individual cutting edge forces in face
milling, Mech. Syst. Sig. Proc. 24 (2010) 18441857.
[17] L. Ma, S.N. Melkote, J.B. Morehouse, J.B. Castle, J.W. Fonda, M.A. Johnson, Thin-film PVDF sensor based monitoring of cutting forces in peripheral end
milling, J. Dyn. Syst. Measur. Contr. Trans. ASME 134 (2012) 51014.
[18] L. Ma, S.N. Melkote, J.B. Castle, PVDF sensor-based monitoring of milling torque, Int. J. Adv. Manuf. Technol. 70 (2014) 16031614.
[19] Kistler RCD Type 9171A Rotating Dynamometer for High-Performance Cutting, 2016. <https://www.kistler.com/us/en/application/sensor-
technology/cutting-force-measurement/milling/kistler-rcd-type-9171a-rotating-dynamometer-for-high-performance-cutting/> (accessed
2016.11.28).
[20] A. Albrecht, S.S. Park, Y. Altintas, G. Pritschow, High frequency bandwidth cutting force measurement in milling using capacitance displacement
sensors, Int. J. Mach. Tools Manuf 45 (2005) 9931008.
240 Z. Xie et al. / Mechanical Systems and Signal Processing 93 (2017) 225240

[21] M. Rizal, J.A. Ghani, M.Z. Nuawi, C.H. Che Haron, Development and testing of an integrated rotating dynamometer on tool holder for milling process,
Mech. Syst. Sig. Proc. 5253 (2015) 559576.
[22] C.A. Suprock, J.S. Nichols, A low cost wireless high bandwidth transmitter for sensor-integrated metal cutting tools and process monitoring, Int. J.
Mechatron. Manuf. Syst. (2009) 441454.
[23] J.S. Nichols, Design and Application of a Wireless Torque Sensor for CNC Milling, ProQuest, UMI Dissertations Publishing, 2009, p. 112.
[24] Z. Xie, J. Li, Y. Lu, Design and analysis of a new tool holder with force sensing element, in: S. Yingying, C. Guiran, L. Zhen (Eds.), Advances in Intelligent
Systems Research, 2015, pp. 247251.
[25] M.C. Shaw, Metal Cutting Principles, second ed., Oxford University Press, New York, 2005.
[26] D.S. Nyce, Linear Position Sensors: Theory and Application, John Wiley & Sons, 2004.
[27] Y. Liang, L. Xiaowei, C. Weiping, Z. Zhiping, High resolution interface circuit for closed-loop accelerometer, J. Semicond. 32 (2011) 45005.
[28] K.J. Xu, C. Li, Z.N. Zhu, Dynamic modeling and compensation of robot six-axis wrist force/torque sensor, IEEE Trans. Instrum. Meas. 56 (2007) 2094
2100.