See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/301830801

On the design and analysis of an octagonal–

ellipse ring based cutting force measuring
transducer

Article in Measurement · August 2016

DOI: 10.1016/j.measurement.2016.04.055

CITATIONS READS

2 60

2 authors:

Mohammad Uddin Dong Songyi

University of South Australia University of South Australia
46 PUBLICATIONS 286 CITATIONS 2 PUBLICATIONS 58 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Biodegradable Mg alloy implants View project

All content following this page was uploaded by Mohammad Uddin on 06 June 2017.

The user has requested enhancement of the downloaded file.

 Measurement 90 (2016) 168–177

Contents lists available at ScienceDirect

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

On the design and analysis of an octagonal–ellipse ring based cutting

force measuring transducer
M.S. Uddin ⇑, Dong Songyi
School of Engineering, University of South Australia, Mawson Lakes, SA 5095, Australia

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

Article history: This paper presents study of design, modelling, and analysis of a modified octagonal ring based cutting
Received 6 January 2015 force measuring transducer. The shape of an octagonal ring was modified by altering the geometry of
Received in revised form 16 March 2016 the circle into an ellipse. The main objective is to increase the strain to the displacement ratio under a
Accepted 25 April 2016
given load so that the sensitivity is maximized.
Available online 26 April 2016
Analytical and three-dimensional (3D) finite element (FE) methods were deployed to estimate mechan-
ical responses such as strain, displacement, and stress of the transducer. Results showed that compared
Keywords:
to the octagonal ring with circle, the modified octagonal ring with ellipse had maximized the sensitivity
Octagonal–ellipse ring
Force transducer
by 15% and 25% in axial and tangential loading, respectively.
Strain gauge In order to assess the performance of the designed transducer, calibration tests were performed on an
Cutting force Instron machine under axial and tangential loading and cross-sensitivity of force measurements was
Sensitivity analysed. Following the ISO 376:2011 standard, an extensive uncertainty evaluation was performed to
Finite element analysis validate the measurement process used in the calibration.
Calibration Calibration results showed that the force transducer with the proposed octagonal–ellipse ring was able
Uncertainty to measure the force with average error less than 2%. The maximum average error in cross-sensitivity is
Cross-sensitivity
found to be less than 5% while the overall uncertainty in measurements was up to 1.78%.
Based on simulation and calibration test results, the position and the orientation of the strain gauges on
the transducer were determined, and a new and compact layout of the 3D cutting force measurement
system for milling was then designed and modelled. A frequency analysis was performed to assess the
vibrational response of the system. It is expected that the transducer integrated with octagonal–ellipse
rings would potentially be a sustainable and cheaper option for small-to-medium sized workshops to
accurately measure the milling forces.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction a compromise between the sensitivity and stiffness while achiev-

Force measurement is an essential tool to understand and
assess the behaviour and performance of a cutting process. Strain
gauge based transducers have been widely utilised in small scale
machining workshops, as economical alternatives. A strain gauge
is attached on a flexible mechanical structure to sense strain or
deformation under loading. It converts mechanical deformation
to the change in electrical resistivity which will then be read and
translated into equivalent force applied using data acquisition
and digital processing system [1].
It has been reported that dynamometers using force transduc-
ers typically tend to reduce the stiffness of machining system [2].
In the design of a dynamometer, therefore, it is important to accept
⇑ Corresponding author.
E-mail address: Mohammad.Uddin@unisa.edu.au (M.S. Uddin).
ing a wide range of frequency bandwidth of dynamometers to
measure the forces as accurate as possible. The property of sensi-
tivity refers to the ability of reacting responses under loading. A
sensitive dynamometer is expected to produce large deformation
when even a low load is applied while the structure of transducer
is well under the yield limit.
Typically circular, square, and octagonal shaped rings are
adopted as the fundamental geometric structures of a transducer.
A circular ring shaped transducer has been found to exhibit
excellent sensitivity under all directions of loads, in which, the
maximum strains are located at an angle of ±39.6° and ±90° from
a vertical plane, respectively under tangential and axis loads
applied (e.g. [3]. Despite of circular ring’s good sensitivity, its
implementation seems to be hard due to difficulties in mounting
ring itself with other surfaces and fixing strain gauges on the
curved surface of the ring [4]. As an alternative to circular ring,
the scenario of octagon–circle ring was developed, which resolved

http://dx.doi.org/10.1016/j.measurement.2016.04.055
0263-2241/Ó 2016 Elsevier Ltd. All rights reserved.
 M.S. Uddin, D. Songyi / Measurement 90 (2016) 168–177
problems of assembly and identifying the positions of stress/strain,
and compensated a significant portion of sensitivity. The rigidity of
an octagonal ring tends to be higher than a circular ring so that the
resonance frequency would be greater. However, the sensitivity
has been significantly reduced to compensate for the ease of imple-
mentation, while altering the outer surface into octagonal shape
[1]. In other words, the strain per unit deflection under a given
force decreases even at the most sensitive locations, which eventu-
ally limits the application of octagonal rings. Recently, an extended
octagonal ring (EOR) is designed and deployed in static force mea-
surement applications. However, the transducer becomes struc-
turally large and typically is suitable for measuring large forces
without increasing the measurement sensitivity.
Interestingly, among them, octagonal ring based transducers
are still widely employed for their higher flexibility in setting up
strain gauges. However, the accuracy of force measurements with
the octagonal rings is still limited by their poor sensitivity and they
may not be suitable in applications where higher accuracy is
required. In the past, many researches are focused on design, con-
struction and evaluation of force transducers to increase sensitivity
and flexibility according to the desired applications. Oraby and
Hayhurst [4] designed and developed a compact three-
component tool-shank dynamometer, in which, two holes are
made and positioned on the tool shank so that the maximum sur-
face strains are attained and recorded by the strain gauges. S ß eker
et al. [5] designed bending beam type load cells to measure and
assess the primary cutting forces in shaping operation. Korkut [6]
and Yaldiz and Unsacar [7], Yaldız et al. [8] developed and applied
octagonal–circle ring shaped transducers with strain gauges to
measure three-component milling and turning forces up to
4500 N with measurement sensitivity at any direction of ±5 N
and the cross-sensitivity between two directions perpendicular
to each other of 0.05–0.92%. Kumar et al. [9] studied simple square
ring shaped transducer for the ease of design, manufacture, and
measurement traceability. To measure static loads, a double
extended octagonal ring (DEOR) was adopted in a drawbar
dynamometer of agricultural equipment, showing a linearity of
0.99 with insignificant cross-sensitivity for horizontal and vertical
loads [10].
While piezoelectric material based dynamometers can exhibit a
better compromise between sensitivity and stiffness, providing
accurate measurements, they are highly expensive and may not
be afforded by even small to medium sized workshops [11]. There-
fore, there is an urgent need to design and develop a cost effective
dynamometer which will possess improved sensitivity without
reducing stiffness to measure cutting forces accurately in machin-
ing processes. This will eventually help workshops increase their
manufacturing productivity.
With this aim, this paper presents design and analysis of a new
modified octagonal–ellipse ring shaped force transducer which
forms the fundamental cell of a strain gauge based dynamometer.
Using both analytical and computation modelling, the sensitivity
including strain, deflection, stress and vibrational frequency of the
proposed design was estimated, analysed, and then compared with
that of an octagonal ring transducer. Calibration tests following ISO
376:2011 standard were performed to investigate the sensitivity of
force measurements in the axial and tangential directions. Further-
more, an extensive metrological uncertainty analysis was carried
out to verify the accuracy of force measurements using the proposed
octagonal–ellipse ring shaped transducer.
2. Proposed design – an octagonal–ellipse ring
Based on the analysis of existing transducers as outlined in
Section 1, a modified octagonal ring based transducer was pro-
posed. The main aim is to increase sensitivity while maintaining
169
an excellent rigidity of the transducer. Note again that the mechan-
ical sensitivity here refers to the ratio of strain against deflection or
displacement of ring under a given load applied. Keeping material,
width, equivalent thickness of the ring constant, one potential way
to increase sensitivity is to reduce the thickness of the ring at loca-
tions where the maximum strains occur. With this aim, the idea is
to alter geometry of the inner circle of an octagonal ring into
elliptical shape consisting of a major and a minor radius. Fig. 1
shows a schematic of the octagonal–ellipse ring. Due to lower
thickness along either major or minor axis, the expectation is that
the maximum strain under axial and tangential forces will
increase, thus improving overall sensitivity.
As can be seen in Fig. 1, the key geometric parameters of the
transducer are summarized as: rmajor = major radius of ellipse,
rminor = minor radius of ellipse, w = width, H = height and L = length
of the ring. According to [8], the width w is set to 30 mm as it
affects the sensitivity insignificantly. In order to compare the
performance, we derive the values of the following parameters of
the modified octagonal–ellipse ring from an octagonal ring as the
reference transducer. They are: rmajor = 34 mm, rminor = 30 mm,
tmajor = 6 mm, tminor = 10 mm, r = 31.94 mm, t = 8.19 mm, in which,
r and t are defined as the equivalent radius and effective thickness,
derived from combination of rmajor and rminor, and tmajor and
tminor, respectively. The details of determining the values of these
parameters are illustrated in Section 3.4. It has been reported in
[8] that for an octagonal ring, the sensitivity to rigidity ratio or
strain (2a ) per unit deflection (da ) under an axial force can be
expressed in terms of thickness (t) and radius (r) of the ring as:
2a t
¼ 0:61 ð1Þ
da r
From the above expression, it is clear that 2daa can be increased by
increasing t and decreasing r. However, too small r will cause
potential difficulties in mounting strain gauges on the curved sur-
face. As suggested by Ito et al. [12], a reasonably better compro-
mise between sensitivity and rigidity for an octagonal ring can
be achieved when rt ratio becomes 0.25 or greater. The ratio
between t and r of the octagonal–ellipse ring equals to approxi-
mately 0.25, thus offering correspondingly higher sensitivity
to rigidity ratio. This indicates that the choice of values for thick-
ness and radius are reasonably appropriate in the design of
octagonal–ellipse ring transducer. However, with reference geo-
metric parameters of an octagonal ring (shown above), we have
investigated the effect of rt ratio on mechanical sensitivity, enabling
us to determine an optimum geometry of the octagonal–ellipse
ring. The result of the investigation is illustrated in Section 3.4.
3. Design analysis
3.1. Analytical technique
Analytical modelling was employed to estimate strain, deflec-
tion and stress of the octagonal–ellipse ring. Circular ring theory
was applied to derive an equivalent analytical expression in deter-
mining location and magnitude of maximum strain and stress
under both tangential and axial loads.
It is assumed that cross-sectional dimension of the curved bar is
smaller than both minor and major radii of the ellipse, the octago-
nal–ellipse would be considered as a thin ring. Therefore, using the
theory of classical mechanics and the relationship between the
strain energy and the bending moment, for the given geometry
and material properties of the ring, the expressions for deflection
(d), strain (e) and stress (a) of the ring under axial (Fa) and tangen-
tial (Ft) forces can be written as follows. It has been found that the
170 M.S. Uddin, D. Songyi / Measurement 90 (2016) 168–177

tminor
w

rminor

ellipse
rmajor tmajor
H

L
Fig. 1. Illustration of the octagonal–ellipse ring.

maximum strain occurs at 90° and 47.5°, under axial and tangen-
tial forces, respectively.
Maximum strains,
0:71F a r
ea ¼ e90 ¼
Ewt2
2:83F t r
et ¼ e47:5 ¼
Ewt2
Maximum deflections,
0:55F a r 3
da ¼
Ewt3
7:5F t r 3
dt ¼
Ewt3
Maximum stresses,
ra ¼ E  ea
rt ¼ E  et
In this study, analytical expressions for other shapes such as cir-
cular, octagonal rings can be derived as well, which are directly
taken from existing literature [8]. Sensitivity was estimated using
analytical formulas and the results of the proposed octagonal–
ellipse ring were compared with those of the octagonal–circle shape.
3.2. Finite element modelling
As an alternative to analytical method, finite element (FE) mod-
elling was employed to validate the analytical method. Previously,
FE method, as an effective computational modelling tool, was used
to determine the optimum locations of strain gauge sensors on the
tool shank during turning process [13]. This method was imple-
mented over a finite element analysis (FEA) platform (integrated
and available on Solidworks Standard Student Edition 2013).
Fig. 2 shows a FE model of the octagonal–ellipse ring. The bound-
ary conditions have been applied to the model, according to the
analytical method. As can be seen in Fig. 2 the bottom flat surface
is fixed while axial and tangential forces are applied onto the top
flat surface. A constant force of 5000 N is considered at each direc-
tion during simulation. This force value is assumed to be within a
medium range cutting force value typically considered in small-to-
medium sized machine shops. FE analysis is assumed to be static
linear. A three-dimensional (3D) solid continuum sixteen-node
element with reduced integration was employed in meshing the
model. A mesh convergence study was performed by changing ele-
ment size of global and local meshing. Local meshing with fine ele-
ð2Þ ments was applied to locations of surface where critical stress/
strain occurs. It was found that global element size of 3 mm and
local element of 0.6–0.65 mm was optimum to achieve accurate
ð3Þ predictions. The material (of AISI 4340) for the ring is assumed lin-
early elastic and the corresponding properties are: elastic modulus
(E) = 205 GPa and Poisson ratio (#) = 0.3. The material offers advan-
tageous features such as high strength, ductility and toughness,
ð4Þ and is widely used in various machine components [14].
Upon given the details of computational parameters, finite ele-
ment analysis against the structure of octagonal–ellipse ring was
carried out. The key outputs from the static analysis are maximum
ð5Þ
strain, deflection, distribution of stress across the ring.
3.3. Comparison between analytical and FE analysis
ð6Þ
Using analytical formulas (Eqns. (2)–(7)) and FE modelling,
ð7Þ maximum strain, deflection, stress and sensitivity of the octago-
nal–ellipse ring under axial and tangential forces were estimated.
Table 1 shows a summary of the results obtained from both tech-
niques. It can be evidenced from Table 1 that analytical and FE pre-
dictions closely match to each other. For instance, the error in axial
and tangential sensitivity between analytical and FEA is 1.54% and
1.3%, which is significantly low and can be reasonably acceptable
in engineering design study. This also suggests that analytical
expressions of mechanical properties (e.g. strain, deflection) devel-
oped for the new and modified octagonal–ellipse ring were accu-
rate enough to predict the outcome of parameters.
3.4. Finding optimum geometry of octagonal–ellipse ring
By varying major and minor radii of the ellipse, the impact of
geometric size on the mechanical sensitivity of the ring was anal-
ysed by using FE study. Equivalent radius (r) and effective thick-
ness (t) were estimated for given major (rmajor) and minor (rminor)
radii. The effective thickness was estimated as the average of
thicknesses at two critical locations of the ring (i.e. about 90° and
47.5°) while the equivalent radius was defined as the average of
major and minor radii of the ring considered.
Fig. 3 shows the relationship between the ratio of rt and sensitiv-
ity for the octagonal–ellipse ring. Both major and minor radii were
 M.S. Uddin, D. Songyi / Measurement 90 (2016) 168–177 171

Axial loading
tangenal loading

fine meshing at crical surfaces

Fig. 2. FE model of the octagonal–ellipse ring.

Table 1 with further increase of rt . This eventually leads to an optimum size

Summary of results of octagonal–ellipse ring obtained from analytical and FEA.
of the octagonal–ellipse ring which is rmajor = 34 mm and
Loading Parameters Analytical FEA rminor = 30 mm. Note that width of the ring (w = 30 mm) remained
direction constant. These optimum geometric parameters were then used to
Axial Maximum strain, ea 4.85  104 4.50  104 fabricate and evaluate the proposed octagonal–ellipse ring with
Maximum deflection, da (mm) 0.069 0.063 the octagonal ring, as discussed in the following sections.
Maximum sensitivity, deaa (mm1) 7.03  103 7.14  103
Maximum von Mises stress, ra 98.9 92.19
(MPa) 3.5. Sensitivity improvement and critical stress/strain
Tangential Maximum strain, et 1.09  103 1.10  103
Maximum deflection, dt (mm) 0.361 0.360 Fig. 4 shows an improvement of sensitivity of octagonal–ellipse
Maximum sensitivity dett (mm1) 3.02  103 3.06  103 over an octagonal ring. Note that, in sensitivity comparison analysis,
Maximum von Mises stress, rt 223.45 225.3 the nominal diameter of inner circle of the octagonal ring was 32 mm
(MPa)
while keeping other geometries (e.g. width) constant. On the other
hand, the geometric dimensions of the octagonal–ellipse were the
optimum ones obtained from the previous analysis (Section 3.4).
varied at 30–34 mm while keeping one parameter constant at a Results indicate that compared to the octagonal ring, the modified
time and the corresponding thickness at two critical locations are octagonal–ellipse ring improves the maximum sensitivity by about
determined accordingly. It can be seen from Fig. 3 that the sensitiv- 15% and 26% in axial and tangential loading directions, respectively.
ity increases with the increase of rt . In other words, increasing radii It must be noted that the sensitivity could be further increased by
of the ellipse decreases the effective thickness, thus resulting in an increasing the radii of the ellipse. However, it will reduce the stiffness
increase in strain per deflection at critical locations. On the other of the ring which may cause breakage or failure of the transducer
hand, the results indicate that expanding the thickness heavily under both static and dynamic loading at a longer period.
reduces magnitude of strain even though it could prevent the Figs. 5 and 6 show maximum stress/strain of the octagonal–el-
transducer from large deflection. By considering both equivalent lipse ring under axial and tangential loading, respectively. It is
radius (r) and effective thickness (t), the maximum sensitivity is found that maximum stress and maximum strain are found at
achieved when rt reaches about 0.256, as can be seen in Fig. 3. After the same location of the ring. Under axial loading, maximum strain
this value, the sensitivity decreases slightly or remains unchanged is located at 90° from a vertical, whereas, it occurs at 47.5° under

3.10E-03 Highest sensivity

3.05E-03
3.00E-03 major radius = constant
Sensivity, mm-1

2.95E-03 minor radius = constant

2.90E-03
2.85E-03
2.80E-03
2.75E-03
2.70E-03
2.65E-03
2.60E-03 0.256
0.13 0.16 0.19 0.22 0.25 0.28 0.31
t/r
Fig. 3. Relationship between rt and sensitivity for an octagonal–ellipse ring.
172 M.S. Uddin, D. Songyi / Measurement 90 (2016) 168–177

8.00E-03 Axial Tangential off-sets. The orientations of the strain gauges depend on the direc-
7.00E-03 tions of the strain, which influence the performance of the load cell
Sensityvity (mm-1)

6.00E-03
5.00E-03
4.00E-03
3.00E-03
2.00E-03
1.00E-03
0.00E+00
Octagonal Octagonal-Ellipse
Fig. 4. Comparison of sensitivity between octagonal and octagonal–ellipse rings.
tangential loading. At locations of critical strains, inner surface
experiences compressive stress while outer surface tensile stress.
As can be seen in Fig. 6(a), quite large stress/strain is found around
bottom of the ring under tangential loading. This is due to can-
tilever beam effect at restraint of the ring. Furthermore, under
any given loading, maximum von Mises stress is found to be
225 MPa, which is well below the yield strength of the material
(of 710 MPa).
3.6. Location of strain gauges and Wheatstone bridge circuit
Based on the estimations of analytical and FEA studies, the loca-
tions of strain gauges were determined, including the angle and
[9]. Four locations of critical strains on the ring were identified for
measuring axial load and four locations for tangential load. Fig. 7
shows locations and orientations of the strain gauges on the surfaces
of the ring. For measuring axial load, four gauges were located at the
positions with 90°. Two of them were located at outer surfaces as
‘tensile gauges’ while the other two were located at inner surfaces
as ‘compressed gauges’. Likewise, for measuring tangential load, four
gauges were located at the positions with 47.5°. Two of them were
located at a couple of opposite diagonal surfaces as ‘tensile gauges’
while the other two were located at the other couple of opposite
diagonal surfaces as ‘compressed gauges’. Strain gauges located at
points – 1, 2, 3, and 4 will be sensitive under axial load (Fig. 7(a))
while tangential loads will activate strain gauges located at points
– 5, 6, 7, and 8 (Fig. 7(b)). Strain gauges shown in purple colour indi-
cate the compression experienced by the gauges while orange colour
exhibits the tension experienced by the gauges.
Finally, two full Wheatstone bridge circuits were made using
strain gauges to measure equivalent electrical voltage output
under both axial and tangential forces (see Fig. 7(c)). Note that it
is expected that the cross-sensitivity between voltage outputs
from two circuits would be negligible. This means that under axial
loading, circuit with strain gauges -1, 2, 3, and 4, will provide the
desired equivalent force measurement without any interference
from second circuit (i.e. its output voltage is very small). The

Fa

A B
A B

maximum
stress/strain

(a) (b)
Fig. 5. Locations of critical stress/strain of the ring under axial loading (a) front view (b) closed view of critical stress/strain.

Ft

A B

A
Maximum B
stress/strain

(a) (b)

Fig. 6. Locations of critical stress/strain of the ring under tangential loading (a) front view (b) closed view of critical stress/strain.
 M.S. Uddin, D. Songyi / Measurement 90 (2016) 168–177 173

Fa Ft
5 6

1 2 3 4

7 8
(a) (b)

1 3
Vo-axial
2 4

5 7
Vo-tan
6 8

Vin
(c)
Fig. 7. (a, b) Location of strain gauges on the transducer and (c) Wheatstone bridge circuit.

full Wheatstone bridge with four gauges on each circuit was imple-
mented and tested for load measurements, and the results of
which are presented and discussed in the following sections.
4. Experimental verification
4.1. Static calibration
In order to assess its force measurement performance, a single
octagonal–ellipse ring was fabricated. Strain gauges were attached
on the surfaces of the ring as per the identified critical locations (as
described in Section 3.6) and a full bridge circuit was constructed.
Calibration tests were performed on an Instron machine (a mate-
rial testing machine) and the ring was oriented in a way so that
axial and tangential loads can be applied. Fig. 8 shows the pho-
tographs of calibration test setup used in this study. For instance,
in the case of tangential loading scenario, the ring was rotated
90° from its normal position (i.e. axial loading case), as shown in
Fig. 8(b). The bridge circuit was interfaced with an amplifier, a
noise filter and digital display unit, enabling us to monitor and

(a) (b)

axial loading tangenal loading

octagonal-ellipse ring pin angle plate

Fig. 8. Calibration test setup for (a) axial loading and (b) tangential loading.
174 M.S. Uddin, D. Songyi / Measurement 90 (2016) 168–177

record the equivalent voltage output of the circuit when the force Table 2
transducer was subjected to either axial or tangential loading. The Comparison between applied and measured loads.

bridge circuit was balanced (i.e. zero output voltage) when no Loading Load Output Load Error between meas.
external load is applied. On the other hand, the bridge circuit direction applied (V) measured and appl. load (%)
becomes unbalanced when an external load is applied to the force (N) (N)

transducer and the voltage output is a measure of the load applied. Axial 100 0.206 100.55 0.55
Calibration tests were performed for axial and tangential load- 500 1.005 500.05 0.009
1000 2.005 1000.05 0.004
ing independently. During each test, a loading and unloading cycle
was followed, in which, load applied on the transducer was varied Tangential 100 1.095 101.42 1.4
500 5.455 497.78 0.44
up to 1000 N at an interval of 100 N and the corresponding voltage 1000 10.975 999.87 0.013
output of the circuit was recorded. Each test was repeated at least
three times and their average was taken as the final measurement.
Figs. 9 and 10 show force measurement response of the transducer
Figs. 9 and 10. The difference between measured and applied loads
in the form of voltage with respect to load applied in axial and tan-
was found to be always less than 2% under both axial and tangen-
gential directions, respectively. It can be seen from both figures
tial loading cases. This further validates the calibration accuracy of
that the voltage output shows a linear relationship with applied
force measurement by the proposed octagonal–ellipse ring based
load. Further, as shown in Figs. 9 and 10, the best fit curves depict-
transducer.
ing linear relationship between voltage output and applied load are
shown by equations with R2 ranging from 0.95 to 1, indicating a
precise measurement by the designed transducer. While it is 4.2. Cross-sensitivity analysis
expected that voltage output along axial direction would be pre-
dominant during axial loading test, there would be small amount The cross-sensitivity is defined as the strain measured on other
of force response in tangential direction (i.e. cross-sensitivity), as axes, which is normal to main axis or loading direction on the
is shown in Fig. 9. The same phenomenon was observed for tangen- transducer. For example, when the axial load is applied on the
tial loading test (see Fig. 13). Note that the cross-sensitivity of volt- transducer, strain gauges (attached on the surfaces) sensitive to
age output between two loading directions will be discussed in the only axial load will provide output voltage in response to load.
Section 4.2. However, it is possible that strain gauges sensitive to tangential
Table 2 illustrates measurement accuracy of the force trans- loading will be activated during test under axial loading. In this
ducer at applied loads of 100, 500, 1000 N in axial and tangential study, an average error (%) in cross-sensitivity was estimated as
directions. For a given applied load, true load measured by the the fraction of strain or voltage output of insensitive strain gauges
transducer was estimated using linear relationship obtained from with respect to voltage output of main strain gauges when the load
was applied in the main direction (either axial or tangential). As
long as the fraction of cross-sensitivity is small and within an
acceptable limit, the force measurement in the main loading direc-
2.5
tion would be considered accurate and reliable. Table 3 sum-
Axial response marises the cross-sensitivity of force measurements. Cross-
2 Tangenal Response sensitivity results are shown for 100, 500, and 1000 N loads. As
can be seen from Table 3, average error in cross-sensitivity for each
Output, Volt

1.5 load step is less than 5%. However, cross-sensitivity in tangential

y = 0.002x + 0.0049 loading as the main direction is larger than that in axial loading.
R² = 1 This can be partly due to potential errors in mounting the strain
1
gauges in the direction of measured forces.

0.5 y = 0.0002x + 0.0326

R² = 0.98 4.3. Measurement uncertainty

0 4.3.1. Uncertainty components

0 100 200 300 400 500 600 700 800 900 1000
No measurements are perfect in practice due to unavoidable
Load applied, N error sources during calibration tests. Therefore, an uncertainty
analysis had been conducted to validate the measurement accu-
Fig. 9. Load measurement sensitivity of the transducer under axial loading.
racy of the transducer. Uncertainty analysis followed a calibration
procedure based on ISO 376:2011 standard [15]. According to ISO
376-2011, in this study, six key uncertainty components are con-
14 Tangenal response
sidered and they are repeatability, reproducibility, resolution, zero
Axial Response
12
10
Output, Volt

Table 3
8 Cross-sensitivity results.
y = 0.011x - 0.0206
6 R² = 0.9992 Loading Load Output in axial Output in Average
direction applied direction (V) tangential error (%)
4 (N) direction (V)
y = 0.0002x - 0.0015
2 Axial 100 0.205 0.009 4.39
R² = 0.957
500 1.005 0.003 0.29
0 1000 2.005 0.075 3.74
0 100 200 300 400 500 600 700 800 900 1000
Tangential 100 0.049 1.095 4.47
Load applied, N 500 0.241 5.455 4.41
1000 0.302 10.975 2.75
Fig. 10. Load measurement sensitivity of the transducer under tangential loading.
 M.S. Uddin, D. Songyi / Measurement 90 (2016) 168–177 175

offset, hysteresis and interpolation. Combining these six individual 0.60%

Overall uncertainty, %
uncertainty components, an overall uncertainty of force measure-
ments, wc can be written by the expression shown below. 0.50%
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 0.40%
wc ¼ w2rpr þ w2rep þ w2res þ w2zer þ w2hys þ w2int ð8Þ
0.30%
In Eq. (8), wrpr = relative uncertainty due to reproducibility,
wrep = relative uncertainty due to repeatability, wres = relative uncer- 0.20%
tainty due to resolution of digital meter, wzer = relative uncertainty 0.10%
due to zero offset, whys = relative uncertainty due to hysteresis,
and wint = relative uncertainty due to interpolation. During calibra- 0.00%
tion tests, loading and unloading cycles were repeated five times. 0 100 200 300 400 500 600 700 800 9001000
Using six set of data at a given applied load, uncertainty compo- Load, N
nents were estimated. In this study, we used a digital meter of res-
olution of 0.001 volt to record output voltage and thus the Fig. 11. Overall uncertainty of measurement for axial loading.

uncertainty due to resolution (wres) was estimated by the ratio

between resolution of the meter and the average of the correspond-
ing measurements. The detailed mathematical expressions for esti- 2.00%
mating each uncertainty component are described in [15]. 1.80%

Overall uncertainty, %
1.60%
4.3.2. Uncertainty results 1.40%
Using Eq. (8), the overall uncertainty for each measurement 1.20%
starting from 100 N to 1000 N was calculated under axial and tan- 1.00%
gential loading cases. Tables 4 and 5 summarize the uncertainty 0.80%
results of the test under axial and tangential loading, respectively. 0.60%
Figs. 11 and 12 depict the change of uncertainty with respect to 0.40%
different loads. Uncertainty varies with the load applied on the 0.20%
transducer. At low loads, uncertainty is larger. However, as the 0.00%
applied load increases, uncertainty contribution to measurement 0 100 200 300 400 500 600 700 800 900 1000
decreases. In the case of both axial and tangential loading tests, Load, N
at each load step, overall uncertainty is found to be less than 2%.
Fig. 12. Overall uncertainty of measurement for tangential loading.
Furthermore, it can be seen from Fig. 12 that uncertainty under
tangential loading is slightly larger than that under axial loading
case. This can be due to the fact that in the case of tangential load-
ing test, the transducer was orientated at angle of 90° and an addi- 5. Design of milling dynamometer with octagonal–ellipse rings
tional angle plate and pin was used to fix the transducer on the
table of the Instron machine, causing further errors in the force 5.1. Layout design
measurements (see Fig. 8(b)).
Based on locations of maximum strains on the ring, a layout of
dynamometer with the modified octagonal–ellipse rings was
Table 4 designed for measuring milling forces. Fig. 13 shows a configura-
Uncertainty results of calibration tests under axial loading. tion of the dynamometer with the positions of strain gauges. Four
Force wrpr (%) wrep (%) wres (%) wzer (%) whys (%) wint (%) wc (%)
octagonal–ellipse rings – A, B, C and D were positioned to predom-
inantly sense and measure the forces. The orientations of strain
100 0.204 0.395 0.031 0.0003 0.1012 0.13 0.4739
gauges on the rings were designed based on the locations of max-
200 0.172 0.303 0.0268 0.0003 0.1315 0.33 0.4976
300 0.118 0.235 0.0238 0.0003 0.0194 0.15 0.3023 imum strain per deflection (as identified earlier) to obtain a high
400 0.151 0.211 0.0213 0.0003 0.1391 0.03 0.2972 measurement accuracy when connected with Wheatstone bridge
500 0.167 0.273 0.0385 0.0003 0.1574 0.11 0.3752 circuits. As can be seen in Fig. 13, strain gauges – 3, 4, 7, 8, 11,
600 0.144 0.249 0.0352 0.0003 0.0000 0.03 0.2915
12, 15, 16 are affected by the thrust force (Ft), and among them,
700 0.141 0.230 0.0326 0.0003 0.1329 0.31 0.4303
800 0.122 0 0.030 0.0003 0.0000 0.27 0.2973
strain gauges – 3, 7, 11, 15 will experience tensile stress while
900 0.000 0 0.028 0.0003 0.0000 0.14 0.1398 4,8, 12, 16 are subject to compressive stress. Cutting force (Fc) will
1000 0.113 0.185 0.0262 0.0003 0.1069 0.07 0.2548 be detected by rings C and D with strain gauges – 1, 2, 5, 6, while
feed force (Ff) by rings A and B along with strain gauges – 9, 10, 13,
14. Full Wheatstone bridge circuits can be developed connecting
Table 5
all strain gauges on four rings to measure all three forces simulta-
Uncertainty results of calibration tests under tangential loading.
neously. The bridge circuits can be designed to compensate for any
Force wrpr (%) wrep (%) wres (%) wzer (%) whys (%) wint (%) wc (%) changes in resistances due to temperature.
100 0.716 0.991 0.0017 0.0133 0.8511 0.66 1.6404
200 0.286 0.332 0.0011 0.0133 0.3817 1.69 1.7882
300 0.696 0.641 0.0008 0.0133 0.8062 0.31 1.2832 5.2. Dynamic analysis
400 0.517 0.517 0.0014 0.0133 0.6519 0.17 1.0043
500 0.509 0.777 0.0006 0.0133 0.1986 0.31 1.0024
600 0.375 0.520 0.0005 0.0133 0.3847 0.46 0.8813
It is to be noted that vibrational frequency of machine tools
700 0.389 0.587 0.0013 0.0133 0.2251 0.54 0.9241 must conform to the natural frequency of the dynamometer. In
800 0.437 0.175 0.0004 0.0133 0.2675 0.65 0.8440 order to assess the vibrational performance, we have estimated
900 0.169 0.314 0.0003 0.0133 0.0604 0.82 0.9013 the natural frequency of the proposed dynamometer including
1000 0.107 0.095 0.0003 0.0133 0.1374 0.87 0.8975
octagonal–ellipse rings with other structural frames. To do
176 M.S. Uddin, D. Songyi / Measurement 90 (2016) 168–177
top plate
side shields
boom plate
Fig. 13. Layout of dynamometer with octagonal–ellipse rings (left) and orientations of strain gauges (right).
this, the stiffness of a single ring is estimated using the expression
in Eq. (9).
Ewðt Þ3
Ka ¼
1:8ðr 3 Þ
Given the geometry and material properties of the ring for an
axial loading, the constant Ka is calculated as 57,800 N/mm.
Assuming the system as a spring mass system, the natural fre-
quency would be given by Eq. (10).
rffiffiffiffiffiffi
1 Ka
fd ¼
2p m
where Ka = spring constant (N/mm), m = mass of dynamometer
including additional objects on it (kg) and fd = natural frequency
of the dynamometer (Hz). For the given mass m of 5 kg, fd is esti-
mated to be 541 Hz. It is reported that, to avoid possible damage
due to resonance frequency, the natural frequency of a dynamome-
ter system should be at least four times greater than the frequency
of the machine which is generally related to spindle revolution. The
maximum spindle speed of the milling machine, on which the
dynamometer can mounted, would be 8100 rpm or approximately
135 rev/s (Hz). This speed is often used during rough and semi-
finishing in small machine shops, indicating that the designed
dynamometer would be robust and suitable in wider applications.
However, it must be noted that the design of the dynamometer with
the proposed octagonal–ellipse ring can easily be extended for the
measurement of larger cutting forces while the spindle will rotate
at a speed even higher than 8100 rpm.
Further, as the materials of mechanical structure and the strain
gauges would be easily available and fairly cheaper, the cost of the
whole force transducer would potentially be inexpensive and
affordable for small to medium size machining industries to under-
stand and analyse the behaviour of cutting process, which will
eventually enhance manufacturing productivity. It is to be noted
that, in addition to its use in machining processes, the designed
force transducer can be reconfigured and applied to many other
applications to measure applied forces or loads. They include – to
measure force in drilling process, tilling the soil by agricultural
machineries, applied load on the bridge due to vehicles passing
on it, carrying load in lifting cranes and so on. Therefore, it can
carefully be said that the proposed force transducer can be
regarded as a robust, affordable and reliable force measuring tool.
B
feed force (Ff)
C D
A cung force (Fc)
thrust force (Ft)
1 2 5 6 9 10 13 14
C D
3 4 7 8 11 12 A 15 16 B
6. Conclusions and outlooks
In this project, a modified octagonal–ellipse ring based force
transducer has been designed, modelled and analysed. Both analyt-
ical and FE methods are deployed to assess the sensitivity of the
ð9Þ
transducer. With respect to an octagonal ring, the sensitivity of
the octagonal–ellipse ring is found to increase by 15% and 25% in
axial and tangential loading directions, respectively.
Following the ISO 376:2011 standard, the designed transducer
was fabricated and calibrated. An uncertainty evaluation was per-
formed to validate the measurement process used in the calibra-
tion. Calibration results showed that the force transducer with
the proposed octagonal–ellipse ring was able to measure the force
ð10Þ with average error less than 2%. The maximum average error in
cross-sensitivity is found to be less than 5% while the overall
uncertainty was up to 1.78%.
Layout design of dynamometer with octagonal–ellipse rings
becomes compact and its natural frequency is conformed to the
vibrational frequency of the milling machine. Though the design
of dynamometer is shown for milling, it can be applied to measure
forces in other cutting process, e.g. turning and grinding.
As a future work, fabrication and actual dynamic testing of the
complete milling dynamometer would be the next step to evaluate
the efficacy of the proposed octagonal–ellipse ring based force
measuring transducer. Three-component force measurements
and vibrational performance would be performed on an actual
milling machine.
References
[1] D. Childs, ‘Machine Tool Dynamometers’, Machining Technology: Machine
Tools and Operations, Taylor & Francis Group, LLC, 2008.
[2] G. Totis, M. Sortino, Development of a modular dynamometer for triaxial
cutting force measurement in turning, Int. J. Mach. Tools Manuf. 51 (2011) 34–
42.
[3] M.J. O’Dogherty, The Design of Octagonal Ring Dynamometers, Cranfield
University Bedford, U.K., O’Dogherty Silsoe College, 1995.
[4] S.E. Oraby, D.R. Hayhurst, High-capacity compact-three component cutting
force dynamometer, Int. J. Mach. Tools. Manuf. 30 (4) (1990) 49–559.
[5] U. Sßeker, A. Kurt, I. Çiftçi, Design and construction of a dynamometer for
measurement of cutting forces during machining with linear motion, Mater.
Des. 23 (2002) 355–360.
[6] I. Korkut, A dynamometer design and its construction for milling operation,
Mater. Des. 24 (2003) 631–637.
[7] S. Yaldiz, F. Unsacar, Design, development and testing of a turning
dynamometer for cutting force measurement, Mater. Des. 27 (10) (2006)
839–846.
[8] S. Yaldız, S.F. Ünsaçara, 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) 1499–1511.
 M.S. Uddin, D. Songyi / Measurement 90 (2016) 168–177
[9] H. Kumar, C. Sharma, A. Kumar, The development and characterization of a
square ring shaped force transducer, Meas. Sci Technol. 24 (2013) 095007,
http://dx.doi.org/10.1088/0957-0233/24/9/095007.
[10] Y. Chen, N.B. McLaughlin, S. Tessier, Double extended octagonal ring (DEOR)
drawbar dynamometer, Soil Tillage Res. 93 (2007) 462–471.
[11] Kistler Instrument Corp. Force Sensors: High-resolution Measurement of
Force, Torque and Strain, US, 2004.
[12] S. Ito, S. Sakai, M. Ishikawa, Bull. Jpn. Soc. Precision Eng. 14–25 (1980).
View publication stats
177
[13] J.A. Ghani, P.S. Jye, C.H.C. Haron, M. Rizal, M.Z. Nuawi, Determination of sensor
loction for cutting tool deflection using finite element method simulation, J.
Mech. Eng. Sci. 226 (9) (2011) 2373–2377.
[14] W.S. Lee, T.T. Su, Mechanical properties and microstructural features of AISI
4340 high-strength alloy steel under quenched and tempered conditions, J.
Mater. Process. Technol. 87 (1–3) (1999) 198–206.
[15] EAL Expert Group, 2010. Uncertainty of Force Measurements. In: Calibration
Guide. s.l.:EURAMET Technical Committee, pp. 10–14.