Measurement 49 (2014) 164–181

Contents lists available at ScienceDirect

Measurement
journal homepage: www.elsevier.com/locate/measurement

Development of an innovative plate dynamometer for advanced

milling and drilling applications
G. Totis a,⇑, O. Adams b, M. Sortino a, D. Veselovac b, F. Klocke b
a
DIEGM – Dept. of Electrical, Management and Mechanical Engineering, University of Udine, Italy
b
Werkzeugsmachinenlabor – WZL der RWTH, Aachen, Germany

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

Article history: Tool geometry optimization, workpiece material characterization, process monitoring and
Received 28 June 2013 optimization are based on the measurement of cutting forces by using machining dyna-
Received in revised form 19 November 2013 mometers. Commercial dynamometers cover a wide range of machining applications, nev-
Accepted 25 November 2013
ertheless there is a lack of measuring devices suitable for investigating milling and drilling
Available online 5 December 2013
applications with relatively small cutters and high spindle speeds. In this work, the devel-
opment and testing of an innovative plate dynamometer designed for this purpose is dis-
Keywords:
cussed. The new measuring system was based on three high-sensitive triaxial piezoelectric
Milling
Drilling
force sensors arranged in a novel triangular configuration. Component design was opti-
Cutting forces mized by using FE numerical approaches, according to the general guidelines derived from
Dynamometer mathematical modeling of sensor dynamics. The prototype of the proposed device was
Piezoelectric manufactured and experimentally tested against two high-end commercial plate dyna-
Design optimization mometers by performing static calibration, modal analysis and cutting tests. Experimental
results proved the excellent characteristics of the new device and its effectiveness for
investigating advanced machining applications.
Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction principle used for cutting force estimate. Other important

Cutting force measurements are essential for under-
standing the physics of the cutting process [1] and for
developing reliable cutting force models [2], as well as
for workpiece material characterization, machinability
tests and tool comparison [3], tool geometry optimization
[4], cutting parameters optimization [5]. Also, cutting force
measurements can be successfully applied for tool
condition monitoring [6, 7] and detection of chatter
vibrations [8].
State of the art regarding cutting force measuring in
milling is presented in Table 1. The existing methods and
devices can be classified according to different criteria:
the location of the sensing element and the physical
⇑ Corresponding author. Address: Via delle Scienze, 206, I-33100 Udine,
Italy. Tel.: +39 0432 558258.
E-mail address: giovanni.totis@uniud.it (G. Totis).
features to be taken into account for a rough classification
are the force measuring range (which is also related to sen-
sitivity), the type and number of independent force com-
ponents measured by the device.
Regarding sensor location, the measuring device can be
installed in the feed drives of the linear axes or in the spin-
dle – spindle housing subsystem. Alternatively, it could be
embedded into the rotating spindle adapter – toolholder,
or it could be integrated into the machine tool table –
workpiece fixture.
It should be also recalled that cutting forces are always
estimated by using indirect methods, i.e. by measuring the
effects of cutting forces such as local deformations, dis-
placements or accelerations of mechanical elements com-
posing the machining system. For this purpose, several
physical principles were exploited, such as the conversion
of electro-mechanical power, the variation of electrical
resistance of a stretched conductor, the piezoelectric effect
and many others.

0263-2241/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.measurement.2013.11.049
 G. Totis et al. / Measurement 49 (2014) 164–181 165

Table 1
State of the art of cutting force measurement in milling (⁄ = with dynamics compensation; ⁄⁄
= nominal natural frequency, ideal clamping conditions, no
workpiece or light workpiece; NS = not specified).

Sensor location Sensor description Fx; Fy Axial force F z Cutting torque M z Ref.

Range Band- Range Band- Range Band-

(kN) width (kN) width (Nm) width
(kHz) (kHz) (kHz)
Linear drives, Cutting forces derived from servo-motor currents of feed >1 0:13 = = = = [9]
spindle, drives measured by amperometers (research device)
spindle Cutting forces measured from command voltages of electro- 5 <2 = = = = [10]
housing magnetic bearings of milling machine spindle (research
device)
Cutting forces derived from spindle vibrations measured by >5 0:5ð1 Þ = = = = [11]
displacement capacitance sensors (research device)
 
Cutting forces derived from spindle acceleration measured by >1 1ð2 Þ >1 1ð2 Þ = = [12]
triaxial accelerometer integrated in the toolholder (research
device)
Cutting forces derived from strain gauges attached to spindle >2 0:2 = = = = [13]
bearings or other spindle elements
Force sensors (Kistler 9135 (axial) and Kistler 9145 (shear) 3 0:5ð1:5 Þ 36 0:5ð1:5 Þ = = [14,15]
rings) integrated in the spindle
Spindle adapter, Torque rotating dynamometer based on strain gauges = = = = > 100 0:7ð< 2 Þ [16]
tool-holder (research device)
Rotating mechatronic toolholder, based on piezo-resistive = = 2 NS > 50 NS [17]
strain gauges or piezoelectric elements (research device)
Piezoelectric rotating dynamometer Kistler 9123C 0:5  5 0:5ð2 Þ 2  20 0:5ð1 Þ 20  200 1 [8,18]
(10,000 rpm max)
Piezoelectric rotating dynamometer Kistler 9125A = = 0:3  3 ð< 5 Þ 10  50 2:5 [19]
(25,000 rpm max)
Strain gauges mounted under each cutting insert, and >1 NS = = > 100 NS [20]
telemetry for data transmission (research device)
Piezoelectric triaxial rings under each cutting insert and 3 1.7 2 1.7 200 1.7 [21]
telemetry for data transmission (prototype of commercial
sensor)
Platform Plate dynamometer based on strain gauges attached on 5 1 5 1:5 = = [22]
(workpiece flexible octagonal rings (research device)
 
fixture) Piezoelectric plate dynamometer Kistler 9257 5 < 1ð2:3 Þ 5 < 1ð3:5 Þ = = [21]
Piezoelectric plate dynamometer Kistler 9272 (‘‘Drilling 5 1ð3 Þ 5 4 ð6 Þ 200 2:2ð4 Þ This
Dyn.’’) paper
Piezoelectric plate dynamometer Kistler 9256C1 (‘‘Mini Dyn.’’) 0.25 > 2ð5 Þ 0.25 > 3ð5 Þ = = [23–28],
This
paper
Proposed piezoelectric plate dynamometer (research device, > 1 >2 >1 >3 < 35 > 2:5 This
‘‘Delta Dyn’’) paper

For instance, Jeong et al. [9] extracted cutting force
information from the current signals of the servo motors
governing machine tool axes, while Auchet et al. [10]
adopted a similar approach based on control signals of
electro-magnetic spindle bearings. These approaches do
not alter machine tool architecture and dynamics, never-
theless a poor frequency bandwidth can be obtained in
the first case (130 Hz), while advanced and very expensive
machine tool elements such as the electro-magnetic bear-
ings are needed in the second case.
Another possibility is to derive the cutting forces from
the local deformations of flexible mechanical parts of the
machining system by using strain gauges [13,29,22]. Nev-
ertheless, when using strain gauges it is usually necessary
to weaken the mechanical structure where they are at-
tached in order to achieve a sufficient sensitivity. As a con-
sequence, the dynamics of the (weakened) mechanical
structure do considerably disturb the measured cutting
force signals, implying a modest frequency bandwidth.
Cutting forces may be also estimated from spindle or
cutter vibrations, by using accelerometers [12] or displace-
ment probes [11]. As before, one major drawback of these
solutions is the poor frequency bandwidth.
Generally, the best compromise between stiffness and
sensitivity can be achieved by using piezoelectric crystals
[30]. Piezoelectric rings can be installed into the spindle
housing [14,15], or rotating dynamometers can be clamped
between the tool and the spindle [8,19,17]. However, the
range of application of these devices is limited to the
low-medium spindle speed range. In addition, cutting force
signals acquired by these sensors are disturbed by the low-
frequency resonances of the spindle-tooling system.
Recently, an innovative rotating dynamometer capable
of measuring individual cutting edge forces in face milling
was developed by Totis et al. [21]. The device was based on
piezoelectric triaxial force cells installed behind each cut-
ting insert, which was inspired by the pioneering work of
Adolfsson et al. [20], who applied strain gauges for the
166 G. Totis et al. / Measurement 49 (2014) 164–181
same purpose. By using this advanced configuration, supe-
rior dynamic properties were achieved. However, the
range of application of this device was still limited to con-
ventional milling processes characterized by low to med-
ium spindle speeds and medium to large cutter diameters.
Nowadays there is an increasing industrial interest in
high speed machining performed with conventional cut-
ting tools at very high cutting speeds [31], as well as in mi-
cro milling and drilling processes requiring very small
cutter diameters [32]. In both cases, very high spindle
speeds are required, which imply rapidly changing cutting
forces characterized by high-frequency dynamics. Under
these extreme cutting conditions, neither force sensors in-
stalled in the spindle housing nor rotating dynamometers
integrated in the tooling system can be successfully ap-
plied, and the only feasible solution are plate dynamome-
ters placed between the workpiece and the machining
table [24–26,28].
Nevertheless, plate dynamometers (as well as the other
types) are generally affected by a limited frequency band-
width fdyn achievable in operating conditions. This should
be considerably higher than the tooth pass excitation fre-
quency fs characterizing the selected application, i.e.
v c zt
fdyn  fs ¼ ½kHz ð1Þ
60pD
where v c is the cutting speed expressed in m/min, zt is the
number of teeth, D is cutter diameter expressed in mm. An
idea of the required frequency bandwidth for different
milling and drilling applications, together with an estimate
of the actual limit of current dynamometers (including all
the considered types), can be derived from Fig. 1.
The actual dynamic limit of current dynamometers is
due to intrinsic dynamic properties of available sensors
and to the dynamic coupling between the machining sys-
tem and the sensor embedded into it. When considering
plate dynamometers, workpiece mass and stiffness as well
as machine tool table dynamics may considerably alter the
frequency response of the dynamometer with respect to
ideal workpiece characteristics and clamping conditions
usually taken as a reference by sensor manufacturer.
Accordingly, the resulting frequency bandwidth may be
significantly reduced with respect to the nominal value de-
clared in the sensor data-sheet.
By analyzing currently available devices, it was assessed
that conventional applications characterized by high cutting
forces and by low to medium spindle speeds can be success-
fully investigated by commercial plate dynamometers such
as the Drilling Dyn considered in this research for the sake of
comparison. Similarly, micro milling applications character-
ized by extremely small cutting forces and very high spindle
speeds can be effectively studied by using the Mini Dyn ta-
ken as main reference in this work, see Fig. 1.
Nevertheless, there are still some advanced milling and
drilling operations performed with small to medium cut-
ters in the medium to high spindle speed regime which
would require a special dynamometer characterized by
high sensitivity, an adequate measuring range and by a
wide frequency bandwidth, which is not available yet.
Accordingly, this research work was aimed at develop-
ing such an innovative plate dynamometer and comparing
it to state of the art measuring devices tested under real
operating conditions. In addition, a general methodology
for designing high-performance dynamometers will be
proposed.
The paper is structured as follows. Sensor technical
specifications will be first outlined. A simplified mathe-
matical model of sensor dynamics will be introduced in
order to provide general guidelines for sensor design opti-
mization, which will be carried out by using FE numerical
approaches. The results of static calibration and experi-
mental modal analysis performed on the dynamometer
prototype will be discussed and compared to the proper-
ties of the reference plate dynamometers. Eventually, the
accuracy, repeatability and effectiveness of the new
device will be evaluated in real cutting conditions by
performing milling and drilling tests with high spindle
speeds.
2. Sensor development
2.1. Sensor specifications
The new sensor was designed in order to fulfill the fol-
lowing requirements:
 suitable for measuring both conventional, high-
speed and micro milling and drilling applications
with cutter diameter D  0:5—20 mm, cutting
speed v c  50— 1000 m/min, zt  2—6 teeth;
 three cutting force components F x ; F y ; F z and cut-
ting torque (when drilling) M z should be available;
 measuring range of transversal components
F x ; F y  1 kN; measuring range of axial force
F z   2 kN; measuring range of cutting torque
M z  30 Nm;
 static errors should be smaller than about 1% FSO
and cross-talk static errors should be smaller than
about 2%, for all sensing directions;
 frequency bandwidth (in real operating conditions)
higher than 3 kHz, or at least comparable with best
state of the art commercial dynamometers evalu-
ated in the same conditions;
 compatible with lubricants;
 reference workpiece diameter Dw  40 mm;
 through holes should be possible within an area of
diameter Di  30 mm;
 modular design, compatible with different kinds of
CNC machine tools;
 relatively easy to realize and of minimum cost.
2.2. Simplified model of sensor dynamics
When considering piezoelectric force sensors embed-
ded into the machining system, a global mass, damping
and stiffness matrices would be necessary for modeling
the dynamic behavior of the measuring chain in operating
conditions. By adopting this approach, it is possible to rep-
resent the characteristics of the frequency response,
including its multiple modes of vibrations and dynamic
cross-coupling between different sensing directions.
 G. Totis et al. / Measurement 49 (2014) 164–181 167

Fig. 1. Classification of conventional and advanced milling and drilling applications in terms of typical cutter diameters and tooth pass excitation
frequencies, and comparison between application ranges of considered dynamometers (Mini Dyn and Drilling Dyn are high-end commercial
dynamometers, whereas Delta Dyn is the new device developed in this paper).

However, in order to give an insight into system k1 k2

¼ x2n1  x2n1;eff < x2n2;eff  x2n2 ¼ ð4Þ
dynamics each sensing direction can be roughly approxi- M1 M2
mated by a series of mass-spring-damper harmonic
The input force F in is applied to the workpiece, whereas the
oscillators, and dynamic cross-talk between different
measured output signal F out is derived from dynamometer
directions can be neglected. Moreover, in order to high-
internal deformation, which is approximately proportional
light the main ideas which inspired sensor development,
to the force exchanged at node D. The main goal is to deter-
a qualitative lumped-element mathematical model with
mine the frequency response of the measuring chain
only three degrees of freedom can be considered, as illus-
trated in Fig. 2. F out ðjxÞ
W ðjxÞ ¼ ð5Þ
In this model, an important low-frequency resonance F in ðjxÞ
arising from machine tool basement and table dynamics
which can be written as
is represented by an equivalent harmonic oscillator of
modal mass M T , stiffness k1 and damping c1 . The plate F out ðjxÞ F 2 ðjxÞ
W ðjxÞ ¼ ð6Þ
dynamometer is ideally decomposed into a base (mDA , F 2 ðjxÞ F in ðjxÞ
fixed on machine tool table) and a top plate (mDB ), on
which the workpiece fixture imagined as a flexible Under the above hypotheses, the first frequency response
mechanical structure characterized by modal masses on the right can be approximated by
mFA ; mFB , stiffness k2 and damping c2 is clamped. Eventu- F out ðmDB þ mFA Þ 1
W out;2 ¼ ffi1
ally, the workpiece is represented as a single rigid body F2 k1 2
½ðjx=xn1 Þ þ 2n1 ðjx=xn1 Þ þ 1
of mass mw attached to the fixture. ð7Þ
Other realistic hypothesis are:
Similarly, one can obtain
MT  mi 8i F2 2n2 jx=xn2 þ 1
ð2Þ W 2;in ¼ ffi
M1 ¼ MT þ mDA þ mDB þ mFA  M2 ¼ mFB þ mW F in ðk2 =k1 Þðjx=xn2 Þ2 ð2n2 ðjx=xn2 Þþ1Þ þ ½ðjx=x Þ2 þ 2n ðjx=x Þ þ 1
n2 2 n2
½ðjx=x Þ2 þ2nðjx=x Þþ1
n1 n1

i.e., the modal mass located behind the sensing element is
expected to be significantly smaller than that in front of it.
Also, it is reasonable to neglect the internal compliance of
the dynamometer with respect to machine tool and fixture
compliances, that is to say
kD  k1 > k2
Moreover, one can realistically assume that the second res-
onance xn2;eff (mainly due to the fixture-workpiece subsys-
tem considered separately) is higher than that associated
to machine tool table-dynamometer base dynamics
xn1;eff , i.e.
ð8Þ
In short, the overall frequency response of Eq. 5 has two
resonance peaks located at xn1;eff and xn2;eff , deriving from
the dynamic interaction with the mechanical subsystems
placed behind and in front of the sensor, respectively.
ð3Þ Even when the workpiece mass mW is small, the modal
masses associated to dynamometer top plate (mDB ) and
workpiece fixture (mFA and mFB ) are relatively high when
considering conventional plate dynamometers. When the
structure is excited by time-varying external forces, the
inertial effects arising from structure vibration cause
undesired amplification and phase delay of measured
168 G. Totis et al. / Measurement 49 (2014) 164–181

Fig. 2. 2 DoF lumped-elements model for a qualitative explanation of sensor dynamics.

signals. Accordingly, the frequency bandwidth is practically
limited to xn1;eff for conventional plate dynamometers.
Since low-frequency resonances due to machine tool
basement and table will always be present, the new dyna-
mometer concept was based on the optimization of the
mechanical subsystem placed in front of the sensing ele-
ments. Specifically, this can be achieved by minimizing
the modal masses of the dynamometer top plate and work-
piece fixture. By reducing mDB and mFA , Eq. 7 yields
ðmDB þ mFA Þ ! 0 ) W out;2 ! 1
perfectly elide the resonance peak at xn1;eff . This effective
pole-zero overlap is the key for attenuating the low-fre-
quency disturbances affecting the measuring chain.
On the contrary, it is not possible to eliminate the high-
frequency disturbances at xn2;eff , unless the sensing ele-
ment was ideally located in front of this second harmonic
oscillator, and the inertial mass in front of it was mini-
mized. Accordingly, sensor location should be as close as
possible to the input force, for achieving the best dynamic
performance.
ð9Þ

Besides, it can be easily demonstrated that Eq. 8 2.3. Sensor design

simplifies to
Sensor design was developed from geometrical and
2
ðjx=xn1 Þ þ 2n1 ðjx=xn1 Þ þ 1 functional specifications outlined in Section 2.1, as shown
W 2;in  2
; x
xn2
ðjx=ðtxn1 ÞÞ þ 2tn1 ðjx=ðtxn1 ÞÞ þ 1 in Fig. 3. Piezoelectric triaxial force cells Kistler type
ð10Þ 9016B4 were selected as main sensing elements, because
of their great sensitivity, satisfactory measuring range
where and high static compliance, see Table 3 for reference.
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi For a proper functioning of the device, load cells must
M1 x2n2 k1 be clamped between a basement (fixed at machine tool
t¼ ¼ ð11Þ
M1 þ M2 x2n2 k1 þ x2n1 k2 table) and a top plate (where the workpiece is fixed). A pre-
defined axial preload has to be applied to each force cell
It can be noticed that separately, in order to assure the transmission of both axial
M2 and tangential stresses, in every direction (i.e. both com-
! 0 ) t ! 1 ) W 2;in ! 1 for x
xn2 ð12Þ pressive and tractive).
M1
Conventional plate dynamometers are usually
In other words, when the modal masses in front of the composed of 4 triaxial force cells located at the vertexes
sensor are minimized, the zero located at xn1 does almost of a square or rectangular configuration. In this work, an
 G. Totis et al. / Measurement 49 (2014) 164–181
Fig. 3. Comparison between 3 and 4-sensors plate dynamometer configuration.
innovative configuration with only 3 load cells arranged at
the vertexes of an equilateral triangle was conceived, as
illustrated in Fig. 3. Since this configuration resembles
the Greek letter D, the new device was called Delta Dyn.
The proposed configuration is very appealing, since
 the cost associated to the sensing elements is about
25% less than that of the conventional configuration;
 it has a simple architecture easy to realize and
characterized by more isostatic clamping condi-
tions with respect to 4-sensors configuration;
 the total inertial mass above the sensing elements
is slightly smaller, in comparison with 4-sensors
configuration, for a fixed sensor distance from the
center of the device.
In order to preliminary compare different device config-
urations (based on 3 or 4 sensors) in terms of static and dy-
namic characteristics, FE models were realized and
analyzed. Specifically, static compliances in transversal
and axial directions, as well as the first natural frequencies
were estimated by assuming ideal clamping conditions at
dynamometer base. The calculations were carried out by
COSMOSWorks integrated into SolidWorks CAD environ-
ment. The system was discretized by a mesh of parabolic tet-
rahedron elements (10 nodes) of maximum size  5 mm.
In accordance with the obtained numerical results re-
ported in Table 2,, 4-sensors configuration is characterized
by higher stiffness and resonance frequencies. Neverthe-
less, in real operating conditions the vibrating subsystem
behind the dynamometer (machine basement, machine
tool table and dynamometer base) would give rise to
low-frequency resonances which would negatively affect
both the investigated configurations in a similar manner.
These low-frequency resonances would indeed cause dy-
namic disturbances which are approximately independent
from dynamometer high-frequency structural modes of
vibration, whereas they are mainly due to the inertial mass
169
placed in front of the sensing elements which is slightly
higher in the case of 4-sensors configuration. Accordingly,
4-sensors configuration could lose its significant dynamic
advantage with respect to that based on 3-sensors. Besides,
the lower static stiffness predicted for 3-sensors configura-
tion is still satisfactory for the considered applications. For
all the reasons outlined above, the 3-sensors design was
finally preferred.
The effect of different top plate materials was also
investigated, i.e. a generic alloy steel (Young’s modulus
E ¼ 210 GPa, density q ¼ 7:80 g=cm3 ), Ti6Al4V titanium
alloy (E ¼ 105 GPa, q ¼ 4:43 g=cm3 ) and Al7075 aluminum
alloy (E ¼ 72 GPa, q ¼ 2:81 g=cm3 ). Calculations were
repeated for a light workpiece made of Al7075 aluminum
alloy and a heavier one made of alloy steel. As evidenced
in Table 2, the first resonance frequency is only slightly
influenced by the selected top plate material. It is worth
noting, the first resonance frequency obtained with aluminum
top plate is even slightly higher than that obtained with
alloy steel top plate, when a light workpiece is fixed on
the device. On the other hand, an expected increase of
transversal and axial compliances was assessed when
considering less rigid top plate materials. Nevertheless,
compliance estimates obtained with top plate made of
Ti6Al4V or Al7075 are not drastically different from
the compliances of Mini Dyn, as reported in Table 3.
Accordingly, Al7075 aluminum alloy was selected as top
plate material, since it assures an acceptable stiffness and
a promising dynamic behavior by concurrently fulfilling
the principle of inertial mass minimization.
Special workpieces endowed with through holes could
be directly fixed by using bolts. Alternatively, workpieces
with simpler geometry can be easily fixed by using the
special clamps shown in Fig. 4. It can be observed that a
further optimization of top plate material geometry would
be theoretically possible by eliminating such clamps.
Nevertheless, this configuration was finally adopted in
order to reduce the raw workpiece geometry complexity.
170 G. Totis et al. / Measurement 49 (2014) 164–181

Table 2
Numerical results derived from FEM calculations for dynamometer design optimization.

Dynamometer Top plate Mass above load cells (g) Workpiece Transversal compliance Axial compliance First resonance
configuration material (without workpiece) material (mass hxx  hyy (103 lm/N) hzz (103 lm/N) frequency fn1 (Hz)
(g))
Square Alloy 597 Al7075(53) 2.60 2.72 7366
configuration – 4 steel
sensors
Ck45(145) 2.05 1.65 6745
Ti6Al4V 380 Al7075(53) 3.48 3.45 7643
Ck45(145) 2.92 2.36 6612
Al7075 275 Al7075(53) 3.94 4.62 7876
Ck45(145) 3.48 3.06 6470
Triangle (D) Alloy 578 Al7075(53) 4.68 3.60 4940
configuration – 3 steel
sensors
Ck45(145) 3.77 2.32 4659
Ti6Al4V 359 Al7075(53) 6.48 4.82 4994
Ck45(145) 5.43 3.46 4492
Al7075 253 Al7075(53) 7.92 5.86 5160
Ck45(145) 6.86 4.48 4416

Table 3
Static characteristics of the new dynamometer in comparison with commercial reference dynamometers (NS = considered Not Significant by stepwise
regression).

Property Symbol Measuring units Drilling Dyn New Delta Dyn Mini Dyn
Sensitivity Fx pC/N  7:8  26  26
Fy  7:8  26  13
Fz  3:5  11:5  26
Mz pC/Nm  160  740  456—1677
Range Fx ; Fy kN 5 1 0:25
Fz 5    þ 20 2 0:25
Mz Nm 200 35 8
Static compliance hxx ; hyy 103 lm/N  2:5 7 4
hzz  0:5 5 4
Static errors eFx 6 %FSO 1 1.30 0.5
eFy 1 1.48 0.5
eFz 1 1.27 0.5
eMz 1 1.29 0.5
Cross-talk jGxy =Gxx j 6% 2 0.32 2
jGxz =Gxx j 1 0.77 2
jGyx =Gyy j 2 0.48 2
jGyz =Gyy j 1 2.36 2
jGzx =Gzz j 2 0.56 2
jGzy =Gzz j 1 NS 2

It can be also noticed that the modular design of dyna-
mometer base allows to easily install the measuring device
both on lathes and milling machines.
After definition of dynamometer design and materials,
main mechanical components were manufactured and
assembled.
Each sensor was connected to a dedicated charge
amplifier Kistler type 5073A311 by using lubricant
resistant cables. Output signals derived from charge
amplifiers were transmitted to National Instruments
analog input modules NI9215 installed on a National
Instruments chassis cDAQ-9178, further connected via
USB to a PC. Sensor signals were sampled and recorded
by using a specific software developed in LabVIEW
environment. Sampling frequency was set to 25 kHz.
The working environment for data analysis was Math-
Works MATLAB.
Eventually, the whole device was assembled and each
load cell was preloaded up to 10 kN, as recommended by
sensor manufacturer.
The three components F x ; F y and F z of the resultant
force acting on the device and the main cutting torque
Mz were derived from linear combinations of the nine
available sensor signals, by considering the orientation of
each sensor with respect to the chosen reference frame.
The coefficients of this basic linear model were slightly
refined in order to take into account small sensor rotations
(caused by the assembly phase) with respect to their nom-
inal position, by direct comparison with the data collected
during the static calibration phase.
 G. Totis et al. / Measurement 49 (2014) 164–181 171

Fig. 4. (a) Exploded view of designed dynamometer; (b) FEM simulation of static behavior; (c) FEM simulation of dynamic behavior, under ideal clamping
conditions (triangle configuration based on 3 sensors, top plate and workpiece made of aluminum Al7075).

3. Static calibration cells. However, such static errors were greatly reduced
when machining, as evidenced in Section 5.
Static calibration was performed by using calibrated Static cross-talk is a signal disturbance caused by an
weights of known masses which were applied to a special external force acting along a direction different from that
workpiece fixed on the device, one direction at a time, corresponding to the signal. In order to estimate cross-talk
according to the standard procedure adopted for instance errors, the following procedure was adopted, as proposed
by Jin et al. in [33]. The quasi-static forces were compared by the authors in [21]. For each new sensor signal the step-
with the known input forces in order to evaluate the static wise regression [34] was applied in order to estimate the
characteristics of the dynamometer. By doing so, admissi- static gains Gij :
ble force ranges, as well as static relative errors and 2 3 2 32 3
cross-talk disturbances were evaluated. F x;out Gxx Gxy Gxz F x;in
6 7 6 76 7
Static and dynamic characteristics of the new device 4 F y;out 5 ¼ 4 Gyx Gyy Gyz 54 F y;in 5 ð13Þ
were compared to two high-end commercial dynamome- F z;out Gzx Gzy Gzz F z;in
ters, namely Kistler 9272 (Drilling Dyn) and Kistler
9256C1 (Mini Dyn). where F j;in (with j ¼ x; y or z) are the effective input forces
Results of static calibration are summarized in Table 3. and F j;out are the force components measured by the device.
It is worth noting that the sensitivity of the proposed The ratios of the cross-gains Gij to the direct-gains Gii were
device along cutting force components F x ; F y and F z is sim- computed and considered as cross-talk indicators. When
ilar to that of Mini Dyn, which is specifically designed for the ratio Gij =Gii is classified as not significant – NS – it
micro milling applications. However, the measuring range th
means that the i output direction was not significantly
of the new device is considerably wider than that of Mini th
disturbed by cross-talk deriving from the j input direction
Dyn, in accordance with outlined sensor specifications. In according to the stepwise regression results. As evidenced
addition, the new device allows cutting torque measure- in Table 3, the obtained cross-talk indicators are satisfac-
ment with intermediate sensitivity and measuring range tory, since they are not significantly greater than the
among those of Mini Dyn and Drilling Dyn, as desired. threshold of 2% declared for Mini Dyn.
Static relative errors were fairly small, although they
were slightly higher than the values declared by sensor
manufacturer for the reference dynamometers. This not 4. Dynamics identification
excellent behavior  measured under quasi-static loading
conditions  was likely due to the lower Young’s modulus The identification of dynamometers dynamics was
of the aluminum alloy chosen as top plate material, which accomplished by means of pulse tests [35]. Each dyna-
could be in contrast with the higher stiffness of the load mometer was excited by using an impact hammer Dytran
172 G. Totis et al. / Measurement 49 (2014) 164–181
Fig. 5. (a) Drilling Dyn (left), Delta Dyn (right) and respective reference frames during all experimental measurements. (b) Pulse testing procedure for
identification of the Empirical Transfer Function Estimates ETFEs of Delta Dyn, by hammering in the axial, radial and tangential directions. (c) Pulse testing
procedure on Mini Dyn and corresponding reference frame.
type 5800B4, (sensitivity 2.41 mV/N) connected to a Kistler
amplifier type 5134B.
Plate dynamometers were fixed on the machine tool ta-
ble of a 5-axis CNC milling machine Chiron FZ15S, both for
dynamic identification and cutting tests described in next
section. For the sake of comparison, the reference frames
of all the considered devices were aligned as shown in
Fig. 5. Samples with special geometry were realized and
fixed on each device in order to test both radial (F x ), axial
(F z ) and tangential (F y ; M z ) directions. In order to study
the influence of workpiece mass on dynamometer dynam-
ics, two samples with the same geometry but made of dif-
ferent materials were considered. Specifically, one sample
was made of Al7075 aluminum alloy (68 g) and the other
made of C45 steel (190 g). For a given sample material,
sample geometry was adapted from one device to the
other in order to allow proper fixing by contemporary
keeping a constant mass.
For each device and sensing direction, the Empirical
Transfer Function Estimate (ETFE) was computed by using
Eq. 5, in agreement with the standard modal analysis pro-
cedures described in [36]. By comparing the hammer sig-
nal F i;in applied along ith direction with the dynamometer
measured component F j;out along the jth direction, the
Empirical Transfer Function Estimates ETFEs of the dyna-
mometer were obtained, as follows:
F j;out ðjxÞ
W ji ðjxÞ ¼
F i;in ðjxÞ
Amplitude of direct ETFEs W ii ðjxÞ are given in Figs. 6
and 7, while cross-ETFEs W ji ðjxÞ were negligible within
the frequency range of interest. As expected, low-
frequency resonances below 3 kHz are visible in all the
diagrams, which basically derive from the low-frequency
vibration modes of the machine tool basement, machine
tool table and dynamometer base plus the residual part
of the dynamometer including the workpiece, which prac-
tically behaves as a rigid body in the low-frequency range.
Such resonances are partially compensated thank to the
physical principle described in Section 2.2. The efficacy of
compensation does mainly depend on the inertial mass
in front of the sensing elements, as demonstrated by the
worse dynamic behavior observed with the heavier sample
made of steel. In addition, several other important reso-
nance peaks are present above 4 kHz, which derive from
the dynamic behavior of the subsystem in front of the load
cells and from its complex dynamic interaction with the
rest of the machining system. It is worth noting that the
dominant resonance peaks of the new device are located
above 4 kHz, in fairly good accordance with the resonance
frequencies estimated by FEM and listed in Table 2.
The frequency bandwidth along a given direction can
be defined as the first relevant deviation of the ETFE
amplitude from the 3 dB bandwidth. When mounting
the lighter workpiece, the frequency bandwidth of the
new dynamometer (Delta Dyn.) was about 2.5 kHz along
ð14Þ transversal directions (F x ; F y ), 2.8 kHz along axial direction
(F z ), 2.7 kHz along cutting torque direction (M z ).
 G. Totis et al. / Measurement 49 (2014) 164–181 173

Fig. 6. Comparison between Drilling Dyn, Mini Dyn and new Delta Dyn dynamics with a light workpiece made of ERGAL (68 g). (a) ETFEs along X
direction; (b) ETFEs along Y direction; (c) ETFEs along Z direction; (d) ETFEs of cutting torque M z .

Fig. 7. Comparison between Drilling Dyn, Mini Dyn and new Delta Dyn dynamics with a relatively heavy workpiece made of C45 (190 g). (a) ETFEs along X
direction; (b) ETFEs along Y direction; (c) ETFEs along Z direction; (d) ETFEs of cutting torque M z .

From direct comparison with reference dynamometers
evaluated in the same conditions, the new device shows
a significantly better dynamic behavior than Drilling Dyn
along transversal and cutting torque directions (F x ; F y ; M z )
while its dynamic performance is slightly worse along
the axial direction F z , probably because of the greater stiff-
ness of Drilling Dyn.
On the other hand, even when mounting a light work-
piece on Mini Dyn, its frequency bandwidth gets consider-
ably narrower with respect to the natural frequencies
174 G. Totis et al. / Measurement 49 (2014) 164–181

Fig. 8. Experimental setup for cutting tests.

Table 4
List of tools used during cutting tests.

Tool Code Description

no.
1 SANDVIK R216.32 16025 Uncoated, fine grained cemented carbide end-mill with D ¼ 16 mm, z ¼ 2 flutes, helix angle ka ¼ 25, working
AP20A H10F side rake cf ¼ 13, working side clearance af ¼ 9.

2 SANDVIK R216.34 06050 Fine grained cemented carbide end-mill with thin TiAlN coating, D ¼ 6 mm, z ¼ 4 flutes, helix angle ka ¼ 50,
AK13P 1620 working side rake cf ¼ 10:5, working side clearance af ¼ 9.

3 SANDVIK R840-0330-50- Cemented carbide drill with TiN/TiAlN coating, D ¼ 3:3 mm, working cutting edge angle v ¼ 70.
A1A 1220

declared in sensor datasheet (above 5 kHz). As a conse-
quence, in real operating conditions Mini Dyn dynamics
are only slightly superior than those of the new device,
confirming the satisfactory characteristics of the proposed
configuration.
5. Cutting tests
In order to compare the characteristics of the new mea-
suring system with the reference plate dynamometers,
several cutting tests on special benchmark workpieces
with different cutting tools were carried out, as shown in
Fig. 8. Each cutting test was repeated on each dynamome-
ter by using the same cutting tool on a (nominally) identi-
cal workpiece, and by applying the same cutting
conditions. Obviously, it was not possible to apply several
dynamometers for simultaneous cutting force measure-
ments. This would have in fact required to fix one device
on the other, thus dramatically altering the dynamical
response of both devices.
Milling tests were carried out with different end-mill
cutters (described in Table 4) on special workpiece made
of ERGAL and C45. In order to test the capabilities of the
measuring systems on a wide variety of cutting conditions,
different cutter-workpiece radial engagements and cutting
parameters combinations were investigated, in accordance
with the Design of Experiments reported in Table 5.
Cutting tests were carried out on ERGAL aluminum
alloy using a 2-flute end-mill cutter and a small width of
cut. This was performed in order to cause an impulsive
behavior, such as the rotating pulse tests described in
[21], see Fig. 9. It is worth noting that Mini Dyn and Delta
Dyn correctly reproduced the effect of the very rapid
engagement between each cutting edge and the workpiece,
whereas Drilling Dyn signals slightly overestimated the
effective input forces and were affected by large oscilla-
tions disturbing the zero-signal level. These undesired sig-
nal features were due to sensor dynamics, i.e. to the typical
mass-spring-damper behavior illustrated in Section 2.2.
Complex interrupted cutting conditions were also mea-
sured on aluminum, as shown in Fig. 10. The effects of
intermittent cutting and variable uncut chip thickness
were accurately measured by Mini Dyn and Delta Dyn,
whereas Drilling Dyn measurements were characterized
by dynamic fluctuations overlapping the effective input
signals.
Eventually, face milling and low-immersion peripheral
milling tests were investigated on C45 by using a 4-flute
end-mill cutter, as illustrated in Figs. 11 and 12, respec-
tively. In such circumstances, the effects of teeth run-out
on process mechanics were clearly evidenced by all the
measuring devices.
It has to be remarked that Mini Dyn signals were the
most correlated with theoretical instantaneous uncut chip
 G. Totis et al. / Measurement 49 (2014) 164–181 175

Table 5
Design of experiments for cutting tests.

Test Tool Cutting speed vc Spindle speed n Feed per tooth fz Depth of cut ap Configuration Workpiece
no. no. (m/min) (rpm) (mm) (mm)
1 1 754 15,000 0.1 0.8 Rotating pulse test (Fig. 9) ERGAL
2 1005 20,000 (Al7075)
3 754 15,000 1 Complex interrupted cutting (Fig. 10)
4 1005 20,000
5 754 15,000 1.5
4 1005 20,000
7 2 151 8000 0.033 0.2 Rotating pulse test – Face milling C45
8 203 10,750 (Fig. 11)
9 254 13,500
10 151 8000 0.4
11 203 10,750
12 254 13,500
13 151 8000 0.8
14 203 10,750
15 254 13,500
16 151 8000 1.2
17 203 10,750
18 254 13,500
19 151 8000 2 Peripheral down milling (Fig. 12)
20 203 10,750 with ae =D ¼ 3:3%
21 254 13,500
22 151 8000 Peripheral down milling (Fig. 12) with
23 203 10,750 ae =D ¼ 6:7%
24 254 13,500
25 3 83 8000 0.05 Dead hole drilling ( 4 mm deep, Fig. 14)
26 166 16,000

Fig. 9. Comparison between dynamometers signals during a rotating pulse test executed on aluminum workpiece with 2-flutes endmill (test n.1 of
Table 5).
176 G. Totis et al. / Measurement 49 (2014) 164–181

Fig. 10. Comparison between dynamometers signals during a complex interrupted cutting operation executed on aluminum workpiece with 2-flutes
endmill (test n.5 of Table 5).

thickness, which further depended on cutter geometry X

þ1

(including teeth run-out), cutter-workpiece radial immer-

F out;p ðt Þ ¼ ch ejhxT t ð16Þ
h¼1
sion and cutting parameters. This was due to the superior
static and dynamic characteristics of Mini Dyn. where xT ¼ 2p=T is the main spindle pulsation.
However, an excellent qualitative agreement between Accordingly, the accuracy of the new device was evalu-
the new dynamometer (Delta Dyn) and Mini Dyn was gen- ated by comparing the Fourier coefficients extracted from
erally observed, especially along F x and F y directions. Delta Dyn (and Drilling Dyn) signals to the Fourier coeffi-
In order to assess the accuracy of the new device – cients extracted from Mini Dyn signals in the same cutting
closeness of the measured output to the effective conditions.
input – Mini Dyn signals were taken as main reference The following ratios were then calculated
for comparison, i.e. they were considered as representative    
 ch;iQ   F iQ ðjxh Þ 
of the effective input force F in .  ¼  i ¼ x; y; z; Q ¼ D; D ð17Þ
c  F ðjx Þ;
The measured signal F out can be decomposed into the h;iM iM h

sum
F out ðtÞ ¼ F out;p ðtÞ þ F out;a ðtÞ ð15Þ
where F out;p is the T-periodic component while F out;a is the
residual (aperiodic) component.
Under stationary and stable cutting conditions the
effective input signal F in is practically T-periodic, where
T ¼ 60=ð2pnÞ is the spindle revolution period. The mea-
sured signal is the output of a damped dynamic linear sys-
tem, thus it should be periodic with the same period T,
after the free oscillations have faded out.
Thus, the periodic term F out;p should be dominant, while
the aperiodic term F out;a should be low.
The periodic term was then written as a Fourier series,
as follows
for estimating Delta Dyn and Drilling Dyn dynamic accu-
racy (subscript D and D respectively).
In order to quantify the accuracy of the static term in a
simple way, the static relative errors - expressed with re-
spect to the static Fourier coefficient derived from Mini
Dyn c0;iM - were first computed, as follows
c0;iQ  c0;iM
e0;iQ ¼ ½% i ¼ x; y; z; Q ¼ D; D ð18Þ
c0;iM
Standard deviations of relative errors are reported in
Table 6. Such relative percentages correspond to an abso-
lute static error of less than 1.5 N along all directions, by
considering an average static term of about 20 N along x
and y directions and about 6 N along z direction, as ob-
tained from the performed cutting tests. It is worth noting
 G. Totis et al. / Measurement 49 (2014) 164–181 177

Fig. 11. Comparison between dynamometers signals during a face milling operation executed on C45 workpiece with 4-flutes endmill (test n.15 of Table 5).

Fig. 12. Comparison between dynamometers signals during a peripheral down milling operation executed on C45 workpiece with 4-flutes endmill (test
n.24 of Table 5).
178 G. Totis et al. / Measurement 49 (2014) 164–181

Table 6
Performance indicators derived from cutting tests.

Description Dir. i Mini Dyn Drilling Dyn Delta Dyn

 
Static term accuracy: st. dev. of relative error r e0;i (%) x 0(Ref.) 4.4 4.8
y 0(Ref.) 15.2 11.1
z 0(Ref.) 20.4 23.3
 
Average standard deviation of aperiodic signal component ( repeatability) r F i;a (N) x 3.12 3.55 2.74
y 2.94 3.44 2.59
z 1.08 1.01 1.32
 
Average absolute drift of zero-signal level DF i;0 =Dt  (N/s) x 0.024 0.044 0.038
y 0.033 0.040 0.036
z 0.057 0.115 0.089

Fig. 13. (a), (c), (e): Comparison in the frequency domain obtained from cutting tests between Drilling Dyn and Mini Dyn. (b), (d), (f): Comparison in the
frequency domain obtained from cutting tests between Delta Dyn and Mini Dyn.

that this static absolute error is considerably smaller than Specifically, the average standard deviation of the aperi-
the static error detected during the static calibration phase odic signal term was applied to measure the repeatability,
( 1:5% FSO). Thus, the static behavior of the new device in
real machining operations was considered satisfactory.
Although the static performance of the new device dur-
ing cutting was comparable to that of Drilling Dyn, the
greatest improvements were achieved in terms of dynamic
accuracy, as evidenced in Fig. 13. The new device shows in-
deed a good dynamic response up to about 3 kHz, whereas
some relevant deviations from the 3 dB bandwidth are
present at lower frequencies in the case of Drilling Dyn.
In order to characterize the repeatability of the new
measuring device – capability of producing the same
output for a given input, independently from its closeness
to the input – the aperiodic term F out;a was analyzed.
as shown in Table 6. However, it has to be stressed out that
this procedure tends to overestimate the signal variability,
since the aperiodic term is contaminated by both random
measurement noise and colored process noise due to
measuring chain dynamics. According to such an indicator,
Delta Dyn and Drilling Dyn were affected by similar errors,
which were significantly greater than those affecting
Mini Dyn.
Small deviations of the zero-signal level during machin-
ing were also examined. For each device and sensing
direction the average absolute drift detected during the
cutting tests was estimated and listed in Table 6. The
results are satisfactory: the performance of Delta Dyn
 G. Totis et al. / Measurement 49 (2014) 164–181
Fig. 14. Comparison between axial force and torque signals measured with Drilling Dyn and with Delta Dyn during a dead hole drilling operation with
carbide drill (test n.26 of Table 4, with spindle speed n ¼ 16; 000 rpm and feed f ¼ 0:1 mm/rev).
was indeed intermediate between that of Drilling Dyn and
that of Mini Dyn. Anyway, the removal of signal bias is nec-
essary from time to time: this is a minor drawback affect-
ing all dynamometers based on piezoelectric sensors.
Eventually, some drilling tests were also executed by
applying the new device and Drilling Dyn for comparison,
since the special equipment for measuring cutting torques
with Mini Dyn was not available. In general, signal analysis
evidenced a good correspondence between the measured
signals, as illustrated in Fig. 14. In detail, the typical ramps
due to entrance and exit of the drill from the workpiece are
clearly visible in both axial force F z and cutting torque M z
signals, as well as the high-frequency fluctuations due to
tool revolution. Both average trends and dynamic fluctua-
tions detected by Delta Dyn are very well correlated with
Drilling Dyn signal features. It has to be recalled that dril-
ling tests were performed at different times on the two
measuring devices separately, thus justifying minor dis-
crepancies among dynamometers’ signals.
In short, the new dynamometer exhibited outstanding
static and dynamic characteristics when applied for real
machining. Accordingly, it will allow accurate and repeat-
able cutting force and torque measurements of advanced
cutting processes such as high-speed machining and some
micro machining operations.
On the contrary, Drilling Dyn may not be suitable for
investigating advanced applications in the field of high-speed
machining or micromachining, which would require a wider
frequency bandwidth, i.e. a better dynamic response.
179
6. Conclusions
In this work, an innovative plate dynamometer for
advanced machining applications was conceived and
developed. The new device was designed for precise tool-
workpiece material characterization in milling and drilling
operations performed with relatively small cutters and
high spindle speeds. It is suitable for middle-range applica-
tions among those covered by existing commercial dyna-
mometers, usually limited to either conventional
machining or micro machining.
The device is capable of measuring the three main com-
ponents of the resultant cutting force (F x ; F y and F z up to
1–2 kN) and the main cutting torque (M z , up to 35 Nm,
only when performing drilling operations). It should be
pointed out that the measuring ranges of the new device
are intermediate between those of high-end commercial
dynamometers considered as a reference. In addition, the
dynamometer satisfied all the requirements outlined in
Section 2.1.
According to the developed mathematical model of sen-
sor dynamics, high-frequency resonances mainly due to
workpiece, clamping elements and dynamometer top plate
cannot be eliminated. On the contrary, low-frequency
dynamic disturbances deriving from dynamometer base,
machine tool table and basement can be effectively
compensated by minimizing the inertial mass in front
of the sensing elements. Thus, dynamometer design
optimization was based on these contrasting principles:
180 G. Totis et al. / Measurement 49 (2014) 164–181
maximization of static stiffness and first resonance
frequency of the device (imagined in ideal clamping
conditions), minimization of the inertial mass in front of
the load cells and minimum cost.
Three high sensitivity triaxial piezoelectric cells were
chosen as primary sensing elements, arranged according
to an innovative triangular configuration. Geometry and
materials of mechanical parts composing the device were
optimized by performing static and dynamic FEM simula-
tions. ERGAL aluminum alloy was selected as top plate
material, since it assured adequate static compliance
and satisfactory resonance frequency by concurrently
minimizing the inertial mass in front of the load cells. A
first prototype of the proposed device was realized and
assembled.
Static calibration was carried out, showing that static
errors were fairly small (below 1.5% FSO), although
slightly higher than those of the reference dynamometers
(Kistler ‘‘Drilling Dyn’’ and ‘‘Mini Dyn’’). Cross-talk indica-
tors were satisfactory, since they were not significantly
greater than the threshold of 2% characterizing the most
accurate among the reference dynamometers.
Dynamic behavior of the proposed dynamometer was
assessed by means of pulse testing. For the sake of compar-
ison, reference dynamometers dynamics were also evalu-
ated in the same clamping conditions and by mounting
the same workpiece mass. When mounting a light work-
piece, the frequency bandwidth of the new dynamometer
(‘‘Delta Dyn’’) was above 2.5 kHz along all sensing direc-
tions. As expected, the resonance frequencies below
2 kHz did not significantly affect the frequency response
of the proposed device.
The developed prototype showed a significantly better
dynamic behavior than Drilling Dyn along transversal
and cutting torque directions, while its dynamic perfor-
mance was slightly worse along the axial direction. On
the other hand, in real clamping conditions Mini Dyn
exhibited only slightly superior dynamic properties with
respect to the new Delta Dyn, confirming the satisfactory
results obtained by the developed measuring system.
However, by varying workpiece mass a further decrease
of frequency bandwidth was confirmed in all cases,
evidencing a limit of current piezoelectric dynamometers:
micro machining operations can be effectively measured
only when workpiece mass is sufficiently small.
Eventually, the behavior of the prototype during real
cutting processes was evaluated. For this purpose, different
cutter geometries, workpiece materials and cutting condi-
tions were tested on each dynamometer.
As expected, Drilling Dyn measurements in milling
were affected by dynamic disturbances due to its narrower
frequency bandwidth. On the contrary, a satisfactory
agreement between the new dynamometer and Mini Dyn
was observed, especially along the transversal directions
(F x ; F y ). The influence of cutter geometry and teeth
run-out, as well as the effect of cutter-workpiece radial
immersion and cutting parameters are clearly visible in
the acquired signals.
The static and dynamic accuracy, as well as the repeat-
ability and zero-signal level drift of the new dynamometer
were quantitatively estimated from the signals acquired
during the cutting tests. Signal analysis evidenced the
satisfactory performance of the new device, which was
intermediate among that of Mini Dyn and that of Drilling
Dyn, as expected.
Eventually, the effectiveness of the developed prototype
for measuring high-speed drilling operations was proved
by direct comparison with Drilling Dyn measurements.
In short, the new device extends the measuring range of
the most accurate commercial plate dynamometer (Mini
Dyn) by exhibiting similar dynamic properties. It would
be of further interest to apply it for investigating the chip
formation mechanism in novel advanced milling and dril-
ling operations.
Acknowledgements
This work was supported by Keymec, Centre for
Research, Innovation and Training in the field of Precision
Mechanics, San Vito al Tagliamento (PN), Italy, which is
gratefully acknowledged. The authors would also like to
thank Kistler company for the technical support.
References
[1] S. Sun, M. Brandt, M.S. Dargusch, Characteristics of cutting forces and
chip formation in machining of titanium alloys, Int. J. Machine Tools
Manuf. 49 (2009) 561–568.
[2] M. Fontaine, A. Moufki, A. Devillez, D. Dudzinski, Modelling of
cutting forces in ball-end milling with tool surface inclination. Part I:
Predictive force model and experimental validation, J. Mater.
Process. Technol. 189 (2007) 73–84.
[3] M. Sortino, G. Totis, F. Prosperi, Dry turning of sintered molybdenum,
J. Mater. Process. Technol. 213 (2013) 1179–1190.
[4] 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, Diamond Relat. Mater.
14 (2005) 651–656.
[5] F. Cus, M. Milfelner, J. Balic, An intelligent system for monitoring and
optimization of ball-end milling process, J. Mater. Process. Technol.
175 (2006) 90–97.
[6] G. Byrne, D. Dornfeld, I. Inasaki, G. Ketteler, W. Konig, R. Teti, Tool
condition monitoring (TCM) – the status of research and industrial
application, Ann. CIRP 44 (2) (1995) 541–567.
[7] C. Scheffer, H. Kratz, P.S. Heyns, F. Klocke, Development of a tool
wear monitoring system for hard turning, Int. J. Machine Tools
Manuf. 43 (10) (2003) 973–985.
[8] E. Kuljanic, M. Sortino, G. Totis, Multisensor approaches for chatter
detection in milling, J. Sound Vib. 312 (2008) 672–693.
[9] Y.-H. Jeong, D.-W. Cho, Estimating cutting force from rotating and
stationary feed motor currents on a milling machine, Int. J. Machine
Tools Manuf. 42 (2002) 1559–1566.
[10] S. Auchet, P. Chevrier, M. Lacour, P. Lipinski, A new method of cutting
force measurement based on command voltages of active electro-
magnetic bearings, Int. J. Machine Tools Manuf. 44 (2004) 1441–
1449.
[11] A. Albrecht, S.S. Parkb, Y. Altintas, G. Pritschow, High frequency
bandwidth cutting force measurement in milling using capacitance
displacement sensors, Int. J. Machine Tools Manuf. 45 (2005) 993–
1008.
[12] S.A. Spiewak, Acceleration based indirect force measurement in
metal cutting processes, Int. J. Machine Tools Manuf. 35 (1) (1995)
1–17.
[13] J. Tlusty, G.C. Andrews, A critical review of sensors for unmanned
machining, Ann. CIRP 32 (2) (1983) 563–572.
[14] 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. Machine Tools
Manuf. 42 (2002) 741–751.
[15] S.S. Park, Y. Altintas, Dynamic compensation of spindle integrated
force sensors with Kalman filter, ASME J. Dynam. Syst. Measure.
Contr. 126 (2004) 443–452.
 G. Totis et al. / Measurement 49 (2014) 164–181
[16] D.A. Smith, S. Smith, J. Tlusty, High performance milling torque
sensor, ASME J. Manuf. Sci. 120 (3) (1998) 504–551.
[17] H. Schulz, A. Versch, U. Fiedler, Process monitoring with mechatronic
tool holders, Product. Eng. (2001) 115–118. VIII(2).
[18] E. Kuljanic, M. Sortino, Tool wear estimation method – TWEM using
cutting forces in face milling, Int. J. Machine Tools Manuf. 45 (2005)
23–34.
[19] F. Klocke, R. Schmitz, D. Veselovac, A. Kirchheim, J. Stirnimann,
Experience with the new rotating high speed dynamometer, in:
Industrial Tooling Conference, Southampton, 2003.
[20] C. Adolfsson, J-E. Stahl, Cutting force model for multi-toothed cutting
processes and force measuring equipment for face milling, Int. J.
Machine Tools Manuf. 35 (12) (1995) 1715–1728.
[21] 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. Signal Process. 24 (2010)
1844–1857.
[22] S. Yaldiz, F. Unsacar, H. Saglam, H. Isik, Design, development and
testing of a four-component milling dynamometer for the
measurement of cutting force and torque, Mech. Syst. Signal
Process. 21 (2007). 1499-151.
[23] J. Stirnimann, A. Kirchheim, New Cutting Force Dynamometers for
High Precision-Machining, Industrial Tooling, vol. 97, Southampton
Institute, 1997.
[24] S. Filiz, C.M. Conley, M.B. Wasserman, O.B. Ozdoganlar, An
experimental investigation of micro-machinability of copper 101
using tungsten carbide micro-endmills, Int. J. Machine Tools Manuf.
47 (2007) 1088–1100.
[25] D. Biermann, A. Baschin, Influence of cutting edge geometry and
cutting edge radius on the stability of micromilling processes,
Product. Eng. Res. Develop. 3 (2009) 375–380.
181
[26] D. Biermann, A. Baschin, E. Krebs, J. Schlenker, Manufacturing of dies
from hardened tool steels by 3-axis micromilling, Product. Eng. Res.
Develop. 5 (2011) 209–217.
[27] F. Klocke, O. Adams, M. Garzon, D. Veselovac, A. Kirchheim, M.
Blattner, R. Thiel, High frequency force measurement of micro
milling operations in hardened steel, in: Proceedings of the 9th
International Conference on Advanced Manufacturing Systems and
Technology – AMST’11, Mali Losinj, Croatia, June 2011, pp. 611–622.
[28] M. Garzn, O. Adams, D. Veselovac, M. Blattner, R. Thiel, A. Kirchheim,
High speed micro machining processes analysis for the precision
manufacturing, in: 5th CIRP Conference on High Performance
Cutting 2012, Procedia CIRP 1, 2012, pp. 609–614.
[29] S. Karabay, Design criteria for electro-mechanical transducers and
arrangement for measurement of strains due to metal cutting forces
acting on dynamometers, Mater. Des. 28 (2007) 496–506.
[30] B. Bernhard, Messen mit Kristallen, Grundlagen und Anwendungen
der piezoelektrischen Messtechnik, ISBN 3-478-93269-6, Moderne
Industrie, Landsberg/Lech, 2002.
[31] H. Schulz, High speed machining, Ann. CIRP 41 (2) (1992) 637–643.
[32] D. Dornfeld, S. Min, Y. Takeuchi, Recent advances in mechanical
micromachining, CIRP Ann. 55 (2) (2006) 745–768.
[33] W.L. Jin, P.K. Venuvinod, X. Wang, An optical sensor based
cutting force measuring device, Int. J. Machine Tools Manuf. 35
(1995) 877–883.
[34] S. Weisberg, Applied Linear Regression, third ed., Wiley series in
probability and statistics, Wiley-Interscience, 2005.
[35] D.J. Ewins, Modal Testing: Theory, Practice and Application, second
ed., Research Studies Press LTD, England, 2000.
[36] L. Ljung, System Identification: Theory for the User, second ed.,
Prentice-Hall PTR, 1999.