Professional Documents
Culture Documents
https://doi.org/10.1007/s00170-022-09950-2
ORIGINAL ARTICLE
Received: 31 March 2022 / Accepted: 11 August 2022 / Published online: 16 August 2022
© The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature 2022
Abstract
Tool wear monitoring has been an indispensable strategy for workshop operators to know the tool wear state accurately. Tool
wear monitoring system can illustrate the effects of various wear patterns by using analytical tool wear models. However,
the current tool wear monitoring methods rarely consider analytical modeling of tool wear patterns in milling process. In
this research, the indirect tool wear monitoring system with online measured cutting force and cutting temperature is pro-
posed. The proposed tool wear monitoring system is composed of five sections. These sections include local cutting force
prediction modeling, tool wear modeling for tool flank face, cutting force and cutting temperature sensing, cutting force
modifying, and tool wear width calculating. Firstly, the local cutting force model was modified as a function of flank wear
width VB to fit the method. Secondly, the analytical model of WC–Co carbide tool flank wear rate was proposed as a func-
tion of cutting force and cutting temperature in milling process. Finally, the flank wear width was calculated and modified
based on the flank wear rate model and the measured cutting force and cutting temperature. The system proposed for tool
wear monitoring was verified with interrupted face milling Inconel 718 experiments. The monitoring error and robustness
are also analyzed. This tool wear monitoring system can be extended to monitor the shape of tool flank wear zone and can
provide guidance for workshop application.
Keywords Tool wear monitoring · Theoretical modeling · Cutting force · Cutting temperature · Inconel 718
13
Vol.:(0123456789)
730 The International Journal of Advanced Manufacturing Technology (2022) 122:729–740
The tool wear patterns revealed by the analytical model can help cutting temperature data. The cutting parameters, workpieces
operator optimize cutting process. Some analytical models have and tool properties, tool geometry, and coefficients are regarded
been used for tool wear prediction, such as the empirical models as input parameters. Firstly, the local cutting force is predicted,
and analytical models. Capasso et al. [22] studied the tool wear based on the current flank wear width VB0. Secondly, the tool
mechanism in turning process of Inconel DA718. The proposed wear rate on flank face during the current spindle revolution is
empirical model included the cutting speed and cutting distance. calculated by the flank wear model considering different tool
Nouri et al. [23] proposed a real-time tool wear monitoring wear modes based on the predicted local cutting force and meas-
method by tracking the behavior of force model coefficients ured cutting temperature. Thirdly, the tool wear volume on flank
in milling process. This method could not reveal the tool wear face is obtained by integrating the tool wear rate over time and
modes. Some of the existing research on model-driven TWM corrected by the measured cutting force. Finally, the flank wear
methods is about turning process rather than intermittent cutting width VB* after the current spindle revolution is obtained by
of milling process. Nevertheless, these methods are utilized to the geometric relationship between VB0 and VB*. The VB* is
monitor the tool wear status in milling process. used as VB0 to predict the local cutting force in the next spindle
An indirect tool wear monitoring (TWM) system with revolution. The algorithm in the TWM system is realized by the
online measuring cutting force and cutting temperature programming software MATLAB R2019a.
is to be proposed. The proposed TWM system consists of
five sections. A local milling force model and a flank wear
model suitable for machining conditions are developed. The 2.1 Cutting force prediction model for cutting tool
accuracy of TWM is improved by the process of prediction- arc edge
calculation-measurement-modification. The high monitoring
accuracy is obtained by calibration data from one test group. The cutting force components along the X-axis (Fx), Y-axis
The TWM system can be extended to monitor the shape of (Fy), and Z-axis (Fz) in the machine coordinate system
flank wear zone. It shows a new thought to the application (MCS) are shown in Fig. 2a. The tangential (Ft) and radial
of analytical models in the TWM system which can provide (Fr) cutting force components in rotating tool coordinate sys-
guidance for workshop application. tem (TCS) is obtained by coordinate transformation accord-
ing to Eq. (1).
Fc = Fy cos 𝜑 + Fx sin 𝜑
2 Tool wear monitoring system during cutting Fr = Fy sin 𝜑 − Fx cos 𝜑 (1)
process Fz = Fz
The overall framework of the TWM system is shown in Fig. 1. where φ is the radial immersion angle.
The tool wear coefficients and cutting force coefficients are The cutting edge engagement shown in Fig. 1b is actually the
estimated by the pre-experimental tool wear, cutting force and tool arc edge in the small depth of cut conditions. The cut section
13
The International Journal of Advanced Manufacturing Technology (2022) 122:729–740 731
13
732 The International Journal of Advanced Manufacturing Technology (2022) 122:729–740
Table 1 The coefficients KH and Hardness ratio KH n assumed that the cutting temperature is evenly distributed
n [26] along the tool-workpiece interface. Based on Fick’s first
HT/HW ≤ 0.8 3 1 law, the average flux rate Jave can be calculated by Eq. (9).
0.8 ≤ HT/HW ≤ 1.25 5.3 3.5 √
1.25 ≤ HT/HW 2.43 7 vc D
Jave = −2C0 (9)
𝜋VB
respectively. The coefficients KH and n are selected in Table 1 where D is the diffusion coefficient, and C0 is the concentra-
depending on the hardness ratio of tool and workpiece material. tion of diffusing elements at the tool-workpiece interface.
The diffusive wear rate can be expressed as the product
2.2.2 Tool wear rate due to adhesive wear of the average flux rate Jave and the contact time tcontact
between materials [27]. The diffusive wear rate on the cut-
The high temperature and pressure on the tool-workpiece ting edge segment can be determined by Eq. (10).
interface lead to the formation of adhesion layer in processing √
Inconel 718. The tool material on the flank face is lost during dwT VB3 − 2R(T+273.15)
Q
�|
= KT |Jave ||Atcontact = KT e d𝜃 (10)
the adhesion layer falling off. dt vc (𝜃)
The adhesive wear has a great relationship with the nor-
mal stress on the contact surface. However, some areas on the where KT is the diffusive wear coefficient, Q is the process
contact surface are not adhered according to the research [27]. activation energy, and R is the gas constant.
It is presumed that p is the ratio of the adhesion area to the
contact area. The normal stress acting on the adhesion zone is
expressed as Eq. (7). 2.2.4 Other tool wear patterns caused by BUE and impact
in milling process
fh p
𝜎a = (7)
VB The BUE formed by adhesion on the tool-workpiece inter-
The adhesive wear rate on the cutting edge segment can be face is serious in the machining process of Inconel 718.
calculated by Eq. (8) by extending Childs’ [28] adhesive wear The stable BUE can increase the effective rake face angle
model to the milling process. and reduce tool wear [29]. However, the vibration and
impact in milling process make the BUE fall off leading
dwA K f (𝜃) to the formation of large and unstable BUE. The tool mate-
= A h vc (𝜃)d𝜃 (8)
dt HT VB rial on the tool-workpiece interface is taken away with the
unstable BUE falling off, causing tool wear.
where KA is the adhesive wear coefficient, and vc is the rela-
The flank wear rate on the cutting edge segment caused
tive sliding speed between contact surface, which is the cut-
by BUE and impact is proposed by the empirical model
ting speed in this study.
shown in Eq. (11) based on the form of cutting force
empirical model in Sect. 2.1.
2.2.3 Tool wear rate due to diffusive wear
( )
dwB ( ) vc (𝜃) nv
The diffusion of chemical elements between tool and workpiece = KB kcv ⋅ h(𝜃) + kev ⋅ d𝜃 (11)
dt v0
in metal cutting process will result in the tool material volume
and strength lost. The elements Ni and Fe in the workpiece where KB is the impact wear coefficient.
material have no traces of diffusion into the tool matrix, while
the elements in tool matrix is found to diffuse into the work- 2.3 Tool wear width prediction model
piece material when machining Inconel 718 with carbide tool.
The initial content of elements in the WC–Co cemented The wear volume on flank face wp can be written as Eq. (12) by
carbide tool is assumed to be constant C0. It is further integrating Eq. (4) over time and immersion angle θ.
t1 𝜃m ( )
dwH dwA dwT dwB
∫t0 ∫𝜃0
wp = + + +
dt dt dt dt
( √ ( )nv ) (12)
t1 𝜃m Hwn−1 K f (𝜃)vc (𝜃) ( ) vc (𝜃)
∫t0 ∫𝜃0
VB3 − 2R(T+273.15)
Q
= KH fh (𝜃)vc (𝜃) + A h + KT e + KB kcv ⋅ h(𝜃) + kev ⋅ d𝜃dt
HTn HT VB vc v0
13
The International Journal of Advanced Manufacturing Technology (2022) 122:729–740 733
The normal stress on the tool-workpiece interface is on former worn surface, and hc is the distance between the
obtained by predicting the local cutting force in Sect. 2.2. The former and fresh worn surfaces.
accumulated error of model will finally cause the theoreti- By combining Eqs. (14) and (15), the geometric relation-
cal results to deviate from the experimental results. The wear ship between the flank wear width VB and wear volume w
volume during one spindle revolution after modification (w) can be further expressed as Eq. (16).
can be modified by the measured cutting force to reduce the √
cumulative error as shown in Eq. (13). 2w(1 + tan 𝛼 tan 𝛾�)
VB∗ = VB20 + ( ) (16)
√ R0 𝜃m − 𝜃0 tan 𝛼
Fmeasure
w = wp (13)
Fpre The proposed model could be extended to different work-
piece materials and tool inserts by considering some proper-
where Fmeasure and Fpre are the resultant forces of measure- ties. The hardness Hw and the process activation energy Q
ment and prediction, respectively. were predefined mechanical and thermal properties when
The shape of the cutting edge segment is shown in Fig. 3a. the proposed model was applied to other workpiece materi-
The effective rake angle γ′ is actually the angle of negative als. The tool geometry needs to be taken into consideration
chamfer. The cross section through the tool flank surface is while applying the proposed model to other types of tool
shown in Fig. 3b. It is assumed that the effective rake angle and coated tools.
is constant during cutting. The geometric relationship of the
tool flank wear on cutting edge segment can be expressed as
Eq. (14). 3 Milling experiment and model verification
( )
w = R0 𝜃m − 𝜃0 Awear (14) 3.1 Experiment setup and measurement
where Awear is the cross-sectional area of the volume lost
The hardness of the workpiece material of Inconel 718 is
on flank face. Awear is assumed to be the average area of the
473 HV. The chemical compositions and physical–mechani-
cross-section to obtain the geometric relationship of the total
cal properties of Inconel 718 are listed in Tables 2 and 3,
wear volume.
respectively. The indexable face milling cutter used in experi-
The geometric relationship in Fig. 3b between the exist-
ment was FMP02-080-A27-SE09-08 (diameter D = 80 mm,
ing tool wear VB0 and tool wear VB* after one spindle revo-
maximum number of teeth z = 8). The milling insert was
lution can be obtained by Eq. (15).
SEET09T308PER-PM YBG202, which was provided by
( ∗ )
VB + VB0 hc ZCC·CT. The tool material was TiAlN-coated WC–Co-
Awear = (15) cemented carbide with 0.8-mm tool nose radius. The experi-
2
ment was carried out with one tooth cutter. When mounted in
where VB* is the flank wear width on fresh worn surface the tool holder, the effective rake, the clearance and inclina-
after one spindle revolution, V
B0 is the flank wear width tion angle are − 15°, 5°, and 0°, respectively.
13
734 The International Journal of Advanced Manufacturing Technology (2022) 122:729–740
The milling experimental setup of Inconel 718 is cutting temperature shown in Fig. 5 are recorded for further
shown in Fig. 4. The milling experiments was carried fitting. The model coefficients are calibrated with the data
out on a VMC0540d CNC machining center. The refer- of experimental group 1 by NLS method.
ence range of the cutting parameters (vc = 45 –75 m/min, Figure 6 shows the fitting result of cutting force and tool
fz = 0.05–0.15 mm/r) was recommended by the technical wear curve. The reduced error criterions 𝜒𝜈2 of cutting force
support of cutting tool supplier ZCC·CT, while the cutting is 0.9567 in Fig. 6a. The evolution trend of the average cut-
depth was kept constant at 0.2 mm. The experimental design ting force during one spindle revolution with cutting pass is
table used in this study is shown in Table 4. clearly shown in Fig. 6. A classic tool wear model proposed
The cutting force and cutting temperature measuring by Gavin is also fitted to further verify the model proposed
devices are shown in Fig. 4. The cutting force was meas- with the experiment result in Fig. 6b. The reduced error cri-
ured in-process in the MCS with 3-component dynamom- terions 𝜒𝜈2 of two models are 1.011 and 1.071, which indicate
eter (Kistler 9129AA, with the sensitivity of − 7.975pC/N) the curve is well fitted according to [33].
equipped with charge amplifier (5070A) and data acquisition The resultant cutting force of experimental and theoreti-
card (5697A, with the resolution per channel of 16 bit and cal results are shown in Fig. 7. It can be seen that the nota-
the sample rate of 1000 Hz). The DTC1401 infrared thermal ble errors occur in groups 3, 8, and 9. This is due to the cut-
imager with a range of − 20 to 900℃ was used to extract ting force changes slightly with the tool wear at the feed of
the maximum temperature in the infrared image as the aver- 0.15 mm/r, which need further investigation. The relative error
age cutting temperature of the cutting area. The Inconel 718 is 21% after removing the obvious outliers. Since the tool wear
emissivity was 0.2 without oxidation [30]. The flank wear monitoring model in Sect. 2.2 has error compensation accord-
width VB was measured with optical microscope every 20 s ing to the measured cutting force, the monitoring model is not
in each cutting condition until the 300-μm flank wear width sensitive to force error and the cutting force error is acceptable.
(ISO-8688).
3.2 Calibration of cutting force and tool wear 3.3 Verification of tool wear monitoring model
monitoring model
Figure 8 shows the comparison of the experimental results
The model calibration is the problem of estimation of the with the theoretical results of Luo’s model and the TWM
unknown distribution coefficients based on the curve fitting. system proposed in this study. The evolution trend of the
The nonlinear relationship between cutting force and tool theoretical value and the experimental value of the tool wear
wear makes the model very complex, which makes it hard has been observed to a satisfactory agreement under the
to solve the model coefficients. In this paper, model coef- experimental conditions in Fig. 8. Although the cutting force
ficients are estimated by the nonlinear least squares (NLS) error between the measured value and the theoretical value
method based on the Levenberg–Marquardt (LM) method has an effect on the tool wear calculation, the monitoring
[31] proposed by Marquardt [32] aiming at minimizing the error can be reduced through the cutting force modification
sum of square errors (SSE). in the TWM system. The monitoring errors of flank wear
The local cutting force model introduced in Sect. 2 has width in Fig. 8 are 9.4%, 17.1%, and 15.9%, respectively.
several coefficients to be estimated. The global cutting force Luo’s model has considerable accuracy when the cutting
in Eq. (3) is the average cutting force during one spindle speed and feed rate are low. However, serious errors occur
revolution. Consequently, the average value of cutting force with the increase of cutting parameters in Fig. 8b, c. The
obtained by the dynamometer is necessary in the coefficient RMSEs of TWM system are significantly smaller than Luo’s
estimation. The measurement results of flank wear and model at large cutting parameters in Fig. 9. The TWM system
proposed in this study can accurately monitor the tool wear
Table 3 Critical physical–mechanical properties of Inconel 718 in the process of milling 718 compared with Luo’s model.
The tool wear data of other test groups are also analyzed.
Elastic Poisson’s Tensile Yield Hardness
modulus ratio strength strength (HV)
The RSME and determination coefficient R2 between the
(GPa) (GPa) (GPa) TWM system and experimental data are shown in Fig. 10. The
results are less different from each other except for test groups
199.9 0.29 1394 1179 473 3 and 7. The noticeable error between the measured value and
13
The International Journal of Advanced Manufacturing Technology (2022) 122:729–740 735
the theoretical value of tool wear can be observed in the test in TWM system after small coefficients disturbance. The
groups 3 and 7 in Fig. 10. Experiments were repeated to rule variations of theoretical results for ± 5% and ± 10% errors
out chance results. The possible reasons are the following: (1) of the coefficients as small disturbance are summarized in
inconsistent tool quality causing tool to fail very quickly under Fig. 12 under the cutting conditions of group 1. The abra-
these conditions; (2) Inconel 718 material has uneven mechani- sive wear coefficient KH could be determined in Table 1
cal properties; (3) the tool wear rate changes sharply due to without being affected by the fitting results, thus the
the rapid change of the effective rake angle. The effective rake coefficient KH is not involved in the robustness analysis.
angle is considered to be constant in tool wear model. The ± 5% and ± 10% errors of the three coefficients leads
Of course, the error may also come from the inaccurate meas- to the maximum variations of 1.42% and 2.61% on the
urement of tool wear used in the coefficient calibration. Cutting monitoring value of flank wear width VB, respectively. It
force modification based on cutting force measurement men- is readily seen that the small disturbance of diffusion wear
tioned in Sect. 2 can reduce the cumulative error of tool wear. coefficient KT has the least effect on the theoretical results.
Figure 11 shows the comparison of the tool wear model According to the above results, the proposed TWM sys-
calculated with cutting force modification and without cut- tem with only one calibration group has reached over 70%
ting force modification in test group 2. The determination monitoring accuracy with high robustness in face milling
coefficient R2 is improved from 0.9051 to 0.9355 through Inconel 718. It can be seen that the TWM system proposed
cutting force modification. in this study requires less experimental data to obtain high
In addition, the robustness analysis of the proposed accuracy. The errors of this system could be attributed to
system is conducted to evaluate the stability of the model the cutting force model and the simplification of the tool-
workpiece contact condition. Besides, the increase of moni-
Table 4 Cutting conditions for tool wear monitoring test toring error was found in the experiment when the flank
wear width is over 300 μm. In our future study, detailed
Test index Cutting speed (m/min) Feed rate (mm/z)
reasons will be further investigated based on this work.
1 45 0.05
2 60 0.05
4 Case study‑application for evolution
3 75 0.05
of tool wear zone
4 45 0.1
5 60 0.1
The flank wear zone shape is directly reflected in the surface
6 75 0.1
topography of processed workpiece, causing the surface of
7 45 0.15
the workpiece to become rough [34]. Monitoring the evolu-
8 60 0.15
tion of flank wear zone shape can further control the work-
9 75 0.15
piece surface quality.
13
736 The International Journal of Advanced Manufacturing Technology (2022) 122:729–740
The evolution trend of the flank wear zone shape over results while it is the opposite after the turning point. These
cutting times is calculated based on the discretized cut- errors have two possible reasons. The first is that the mate-
ting edge according to Eq. (12) as shown in Fig. 13a. The rial hardness in the model is assumed to be uniform. How-
evolution trend of flank wear width at different positions ever, the surface hardness of the material is higher than the
of the cutting edge is similar. Obviously, the tool wear internal hardness due to work hardening. The assumption
zone shape has a turning point along the cutting edge. of uniform hardness leads to the larger theoretical value of
The instantaneous cutting thickness h(θ) in Fig. 1b has a theoretical boundary. The second is the chip plastic side flow
turning point along the cutting edge. The tool wear rate is in chip deformation. The impact of chips staying near the
positively correlated with h(θ) according to Eqs. (3) and cutting edge causes serious wear on the part after the turning
(12). Therefore, the changes of the instantaneous cutting point. The chip plastic side flow becomes more severe with
thickness h(θ) lead to the appearance of turning point in the increase of tool wear [35]. The scraping of chips on the
Fig. 13a. wear zone is more serious resulting in high wear rate.
The theoretical results of flank wear zone shape in To summarize, the TWM system can reduce the monitor-
Fig. 13b is compared with the experimental shape. The ing error by using cutting force modification. It has a con-
observable errors on both sides of the turning point are vincing result on tool flank wear evolution monitoring. The
quite different. The theoretical results of the flank wear zone TWM system proposed in this study can be extended to the
shape before the turning point is larger than the experimental monitoring of the flank shape.
13
The International Journal of Advanced Manufacturing Technology (2022) 122:729–740 737
13
738 The International Journal of Advanced Manufacturing Technology (2022) 122:729–740
Fig. 11 Tool wear of experimental value, theoretical value with and Fig. 12 Robust analysis of the TWM system under the cutting param-
without cutting force modification eters of test group 1
13
The International Journal of Advanced Manufacturing Technology (2022) 122:729–740 739
Fig. 13 Shape monitoring of flank wear zone: a evolution of flank wear zone shape with cutting time; b comparison between theoretical results
and experimental results
13
740 The International Journal of Advanced Manufacturing Technology (2022) 122:729–740
13