Modeling Micro End Mill Operation

International Journal of Machine Tools & Manufacture 40 (2000) 21932211
Modeling micro-end-milling operations. Part III: influence of

tool wear
W.Y. Bao, I.N. Tansel
Mechanical Engineering Department, Florida International University, Center for Engineering and Applied
Sciences, 10555 West Flagler Street, Miami, FL 33174, USA
Received 28 February 2000; received in revised form 9 June 2000; accepted 15 June 2000
Abstract
The characteristics of the cutting forces were studied at different usage levels and the analytical model
of the micro-end-milling operations was modified to represent the tool wear. A new expression was derived
from the model to estimate the remaining tool life from experimental data. The parameters of the model
are estimated by using genetic algorithms. The difference between the simulated and experimental cutting
force profiles for new and worn tools was less than 8%. The remaining tool life was estimated with typically
10% error from the experimental data. Maximum error was 20%. The introduced analytical model and
genetic algorithm-based parameter estimation approach is very convenient for on-line tool wear monitoring
without extensive experimental study. 2000 Elsevier Science Ltd. All rights reserved.
Keywords: Manufacture; Micro-tools; End-milling; Modeling
1. Introduction
Many researchers and manufacturers have studied end-milling operations to design better tools,
to perform machining operations at the optimal cutting conditions, to avoid chatter, and to monitor
tool condition. Depending on the objectives of the study, different approaches were used. Analytical models were developed to design cutting tools better, and to understand the characteristics of
cutting forces [1,2]. To study the dynamic characteristics of the machining numerical methods
were widely used [3,4]. To investigate the influence of run-out and tool flexibility, numerical
methods were improved further. To detect tool breakage and estimate wear, characteristics of
various signals were studied [5,6], and the effectiveness of empirical models including the time
* Corresponding author. Tel.: +1-305-348-1932; fax: +1-305-348-3304.

E-mail address: [email protected] (I.N. Tansel).
0890-6955/00/$ - see front matter 2000 Elsevier Science Ltd. All rights reserved.
PII: S 0 8 9 0 - 6 9 5 5 ( 0 0 ) 0 0 0 5 6 - 0
2194
W.Y. Bao, I.N. Tansel / International Journal of Machine Tools & Manufacture 40 (2000) 21932211
Nomenclature
Fx
Fy
Fu
r
Z
b
ro
g
n
f
ft
a
b
q
qs
qe
p
Km
feed direction cutting force (N)

normal direction cutting force (N)
unit force (N)
tool radius (inch)
the number of tool teeth
tooth helix angle (rad)
run-out length (inch)
run-out angle (rad)
spindle speed (rpm)
feed rate (ipm)
feed per tooth (inch)
depth of cut (inch)
width of cut (inch)
tool cutting angle (rad)
integrating start angle (rad)
integrating end angle (rad)
proportional factor
material coefficient (N/cm2)
series analysis and neural networks was demonstrated [710]. Analytical and empirical models
were used with very different objectives. In this study, a recently proposed analytical model of
micro-end-milling operations (MEMO) [11,12] was modified to represent tool wear at any stage
of tool usage, and the application of genetic algorithms was proposed for estimation of the parameters of the model. Tool condition can be estimated from the values of these parameters. In this
paper, the concept is only introduced; however, the proposed integrated method can be used for
on-line monitoring of tool condition.
The increase of cutting forces with tool wear was observed in turning operations a long time
ago [13]. However, estimation of the tool wear from the cutting force is not easy since the cutting
forces continuously change even in typical turning operations when the cutting conditions change.
If identical parts are going to be machined, commercial tool monitoring systems warn the operator
if the monitored signals get out of an envelope while a part is turned. The number of parameters
increase in end-milling operations and estimation of the tool condition becomes much more complicated. Training of neural networks with simulated data to evaluate the tool condition is one of
the possible solutions. However, testing every possible overlap, depth of cut, and possible failure
requires working with thousands of training cases generated by simulation [10].
Modification of the analytical models of the end-milling operations to represent the tool wear
is straightforward. However, estimation of the parameters of these non-linear models and cutting
conditions could be extremely time consuming, if the conventional optimization techniques are
used. Genetic algorithms [14,15] allow the user to determine the resolution of each parameter of
the non-linear objective function and estimate them quickly.
2195
In the following sections the genetic algorithms are briefly discussed, the proposed model is
introduced, and the validity of the model is demonstrated by comparing the simulations with
experimental data obtained from the machining of metal (NAK-55 steel) and non-metal (EDM
POCO-C3 graphite) workpieces. The tool life can be estimated by using the proposed model from
the monitored cutting force data. The model can be easily simplified to represent the conventional
end-milling operations (CEMO).
2. Theoretical background
In this section, the developed analytical cutting force model is outlined and genetic algorithms
are briefly discussed.
2.1. Cutting force model of micro-end-milling operations
According to the developed analytical model [11,12] the cutting forces of MEMO can be calculated by using the following expressions:
1
ft
ft
ft
FxFu C3 sin3 qC4 cos3 q(1C5)sin2 q p(1C5)sin 2q C6 sin q
r
r
2
r
pC6 cos qp(1C5)q
qe
qs
1
ft
ft
ft
FyFu C4 sin3 qC3 cos3 qp(1C5)sin2 q (1C5)sin 2qp C6 sin q
r
r
2
r
(1)
(2)
qe
C6 cos q(1C5)q
qs
where:
Fu
ft
Kmrft
2 tan b
f
nZ
1
2
C3 1p
3
p
(3)
2196
1
2
C4 p
3
p
C5
2ro
sin g, for each cutting edge, respectively
pr
C6
4ro
cos g, for each cutting edge, respectively
ft
2.2. Genetic algorithms

Genetic algorithms are a class of dynamic computational models that mimic natural selection
and biological evolution to solve problems. Based on biological theories, the deoxyribonucleic
acid (DNA) molecule a tiny corkscrew built from simple chemicals and including a large
number of genes that define individual parts of an organisms blueprint is the most fundamental
piece of lifes mystery. The offspring inherit the characteristics from their parents through the
genes. The survivors, who are dependent on natural selection and survival of the fittest, determine the characteristics of the next generation, and therefore the evolution of the species[14,15].
The first application to solve problems by using biological evolution was proposed in 1975 by
John Holland [16]. Based on the biological DNA theory, a binary number was designated as an
individual chromosome to carry information of the data set, in which each bit represents a gene
(see Fig. 1). The individuals followed the biological DNA hereditary processes to mate with each
other and inherit a new generation through gene crossover and mutation. Natural selection as a
rule determined the survivals of the next generation. The evolution would continue until one
individual of the generation fitted the requested fitness level (see Fig. 2).
In the view of the optimization, the genetic algorithm is a powerful tool to find the optimal
solution of the problems with large data sets by using random search techniques. The natural
selection rule is the key to finding the optimal solution from the genetic evolution. The crossover
operation keeps the better information from the last generation and the mutation operation helps
to search the domain completely to avoid the search being trapped in a local maximum or mini-
Fig. 1.
A data chromosome with 20 genes presenting 3 DNA.
Fig. 2.
2197
Biological evolution cycle of species.
mum. The population size, mating pool size and the numbers of the children from each couple
are decided by the problem study.
The architecture of the genetic algorithms can be developed by using the following steps:
Selection of mating couples (parents) from the population pool.

Selection of hereditary chromosomes of the next generation from their parents.
Gene crossover.
Gene mutation.
Creation of the next generation (children).
Evolution.
The first generation can be randomly created, and repeating the above steps can evolve each
following generation. The evolution continues until the fitness is in a requested fitness level.
3. Cutting force modeling of micro-end-milling operations with tool wear
In this section, the analytical model of MEMO with tool wear representation capability is
introduced. Identification of the parameters of the proposed model by using the genetic algorithms
is presented. Tool life estimation is discussed.
3.1. Analytical cutting force model with tool wear representation
From the previous experimental study of tool wear of the MEMO, the following conclusions
can be drawn [17,18]:
2198
Both feed- and normal-direction cutting forces increase with tool wear.
The ratio of both feed- and normal-direction cutting forces between a new and a used tool
remains the same in the entire tool rotation. The cutting force characteristics of both directions
have similar performance.
Progress of tool wear can be studied in two stages. In the first stage, the tool wears gradually
and cutting force increases slightly. The second stage starts when the tool has been used for a
certain time and the cutting edges of the tool cannot perform their tasks effectively since they
get dull, or are damaged, or are covered with tiny chips. Cutting forces increase much faster in
the second stage than in the first stage and the tool breaks in a short time. Depending on the
characteristics of the tool, workpiece and cutting conditions, the first stage may last between a
few seconds to a few days. To represent the characteristics of the cutting force at different usage
levels the following wear coefficient is proposed:
KW1(C1L)C2
(4)
where: L is the tool life (inch), C1 (1/inch) and C2 are parameters.

The parameter C1 is the tool wear gradient. It depends on the tool, workpiece and cutting
conditions. If the tool wear gradient is small, the progress of the tool wear will take a long time.
It presents the cutting force linear growth with the tool wear. The parameter C2 shows the
exponential growth tendency of the cutting force; its influence is minimal for a new tool. However,
when the tool wear becomes significant, the parameter C2 presents how quickly the tool will fail.
This is related to the hardness of the workpiece material.
The new model uses the same expressions as the analytical model (Expression Eq. (1) and Eq.
(2)). However, the following term (Expression Eq. (3)) is modified to represent the effect of tool
wear to the characteristics of the cutting force:
Fu
KwKmrft
2 tan b
(5)
For a new tool, Kw is equal to 1. Theoretically, Kw is always bigger than 1 for a used tool and
equal to 0 when the corresponding edge of the tool is broken.
3.2. Determination of parameters of the proposed wear model
The wear parameters of the model are obtained from the experimental data by using a genetic
algorithms software called GATool (Genetic Algorithms Research Tool) that was developed in
1998 [19]. The objective function of the optimization program was the difference between the
experimental and simulated maximum cutting force in feed or normal direction. The objective
function is expressed as the following equation:
Min(E)
1
actual
(i)|
|Festimated(i)Fmax
Ni1 max
(6)
actual
(i) is the
where: E is the average absolute error, N is the number of the experimental data, Fmax
2199
experimental maximum cutting force in feed or normal direction at the ith tool life, Festimated
(i)
max
is the estimated maximum cutting force in feed or normal direction at the ith tool life.
In the genetic evolution procedures, 20-bit binary coding was used for two coefficients by
assigning 10 bits for each one. The population size was five. Mating pool size was two versus
two and one child from each couple. The uniform crossover (with 0.5 probability), jumping
mutation (with 0.1 probability), creeping mutation (with 0.05 probability) and elitism were chosen.
The proposed method of the optimization of the wear parameters is presented in Fig. 3.
3.3. Estimation of the remaining tool life
The expression to calculate the remaining tool life can be derived from the presented equations
in the previous section. If it is assumed that a tool would be broken when the force reached a
critical cutting force such as Fc, which can be obtained theoretically from the stress analysis
or experimentally by breaking the tool, the remaining tool life can be estimated by using the
following expression:
1
C
Lt
C1 2
Fc
1c2
new
Fmax
Fcurrent
max
1
Fnew
max
(7)
where: Lt is the remaining tool life, Fc is the critical cutting force in feed or normal direction,
new
is the maximum cutting force in feed or normal direction of a new tool estimated by the
Fmax
Fig. 3. Optimization of the wear parameters of the proposed model for micro-end-milling operations.
2200
current
analytical model, Fmax
is the maximum cutting force in feed or normal direction of the tool at
the certain usage level.
Theoretically, the maximum cutting force is smaller than the critical cutting force during the
machining so that the tool will be broken at any time if the maximum cutting force exceeds the
critical cutting force.
4. Experimental setup
The experiments of MEMO were performed at Florida International University and the Engineering Prototype Center of Radio Technology Division of Motorola Inc. Three different Fadal
CNC machines were used in the experiments. The test workpiece was attached to a Kistler 9257B
3-component piezoelectric dynamometer that was installed on the table of the machine tool. Two
components of the cutting force were collected by using Nicolet 310 and Nicolet Integra model
10 digital oscilloscopes through a Kistler 3-channel charge amplifier. The typical experimental
setup, equipment and contents were presented [11].
In the tool wear test of steel workpieces, the cutting force data were directly collected by using
the dynamometer and digital oscilloscope when the tool was machining the workpiece. Because
the cutting forces were almost at the same level with the experimental noise during the machining
of EDM POCO-C3 graphite workpieces, an aluminum test piece was attached to the top of the
EDM POCO-C3 workpiece and machined periodically. The cutting force data were collected
when the tool was cutting the aluminum test piece in order to have high signal-to-noise ratio (S/N).
5. Results and discussion
In this section, the relationship between the tool wear and cutting force is discussed. The cutting
force (including new and worn tools) and remaining tool life estimation accuracy of the proposed
model are presented.
5.1. Relationship between the tool wear and cutting force
The typical tool wear and increasing cutting force relationship has been known for a long time.
In this study this relationship was investigated by monitoring the cutting force characteristics of
carbide micro-end-mills while they were worn by cutting very hard metal (steel) and very soft
non-metal (graphite) workpieces.
Characteristics of the normal direction cutting force at different stages of tool life are presented
in Fig. 4 and Fig. 5 when two identical tools were used to machine NAK-55 steel workpieces.
Two-flute carbide end-mills with 0.030 diameter were used in the experiments. In test 1, the
cutting conditions were 30,000 rpm spindle speed, 2.5 ipm feed rate, 0.015 width of cut and
0.015 depth of cut. Fourteen sets of cutting force profiles were recorded at the different stages
of tool life and 42 tool life was observed. The cutting conditions of test 2 were 20,000 rpm
spindle speed, 1.25 ipm feed rate, 0.0225 width of cut and 0.015 depth of cut. Ten sets of
cutting force data were collected at the different tool usage levels and the tool had 33 of tool life.
2201
Fig. 4. Relationship of maximum cutting force and tool usage for machining of a steel workpiece with a carbide tool
in test 1.
Fig. 5. Relationship of maximum cutting force and tool usage for machining of a steel workpiece with a carbide tool
in test 2.
Cutting edges wear out very slowly when soft materials are cut. To evaluate the cutting force
characteristics when the soft materials are cut, the tool was worn by machining EDM POCO-C3
graphite workpieces and the cutting force data were collected periodically by cutting a slot on an
aluminum test piece. The cutting conditions for machining the POCO-C3 workpiece were 15,000
rpm spindle speed, 20 ipm feed rate, 0.030 width of cut and 1/32 depth of cut. The aluminum
test piece was cut at 15,000 rpm spindle speed, 5 ipm feed rate, 0.015 width of cut and 1/32
depth of cut. A two-flute carbide tool with 1/16 tool diameter was used in the tests. Nine data
2202
Fig. 6.
Relationship of maximum cutting force and tool usage for machining of a graphite workpiece with a carbide
tool.
sets of the cutting force were collected at different tool usage levels. The tool had 1350 of tool
life and did not break. Experimental data are presented in Fig. 6.
The characteristics of the cutting forces were studied during the one full rotation of the tool.
The new and the worn tools had exactly the same characteristics when the profile was normalized
at the identical cutting condition. However, the magnitude of the cutting forces changed when
the tool wore out. Based on this observation, use of a wear coefficient was satisfactory to include
the tool wear into the analytical MEMO model. The experimental normal direction cutting force
of the new and worn (pre-failure) tools when the NAK-55 steel workpiece is cut (test 1) is
presented in Fig. 7, and critical values are listed in Tables 1, 58. The average ratio is around 1.4.
The pictures of the cutting edges of new and worn tools are presented in Fig. 8 and Fig. 9.
Fig. 7. Comparison of the cutting force profiles between a new tool and pre-failure stage of the tool.
Table 1
Tool wear coefficient (the ratios of the cutting forces at the two extreme stages of tool life)
Tool rotation
Angle (degree)
0
18
36
54
72
90
108
126
144
162
180
198
216
234
252
270
288
306
324
342
360
Normal direction cutting force (N)

New tool
Pre-failure tool
17.40
25.80
32.28
29.44
20.20
12.80
7.00
0.00
0.00
0.00
4.36
18.28
27.04
28.40
19.20
3.92
0.00
0.00
0.00
0.00
10.04
13.84
33.20
44.12
43.44
31.40
12.00
0.00
0.00
0.00
0.00
5.48
25.92
37.84
36.80
25.36
6.04
0.00
0.00
0.00
0.00
5.80
Fig. 8. Cutting edges of a new tool.
Ratio
0.80
1.29
1.37
1.48
1.55
0.94
0.00
1.26
1.42
1.40
1.30
1.32
1.54
0.58
2203
2204
Fig. 9. Cutting edges of a worn tool.
The burr formation was different when the new and worn tools were used. The burr-free and burr
slots of an aluminum workpiece machined by a carbide end-mill from new to worn stages are
presented in Fig. 10.
Fig. 10. The burr-free and burr slots of an aluminum workpiece machined by a carbide end-mill from new to worn
stages (the scale of the ruler is in inches).
2205
5.2. Verification of the tool wear model

The proposed analytical model used ten parameters and two coefficients to represent the cutting
forces of MEMO and CEMO without or with tool runout, at different stages of the tool usage.
Three working condition variables: spindle speed (n), feed rate (f) and width of cut (b).
Two tool run-out variables: run-out (ro) and its angle (g).
Two cutting condition variables: tool cutting entry and exit angle, which presents depth of cut
(a), and up-and-down-milling.
Three tool geometry variables: tool diameter (2r), helix angle (b) and the number of tool
flutes (N).
Two coefficients: material coefficient (Km) and wear coefficient (Kw), which are related to the
tool and workpiece materials, and tool life.
These parameters and coefficients can be found experimentally.
The proposed model was tested on the experimental data of many MEMO cases and very good
agreement was observed between the theoretical and experimental results. The results of two
tools wear test data and the estimations of the developed model are presented in Fig. 11 and
Fig. 12. In both experiments, a two-flute carbide end-mill with 0.030 diameter was used to cut
a NAK-55 steel workpiece. 14 and 10 data sets of the cutting forces were recorded during the
tool life periods. These two tools were broken at 43.2 N and 38.2 N maximum cutting force in
normal direction, and had 42 and 33 of tool life, respectively.
The cutting force coefficient of the material was obtained from the single experiment with a
new tool. The wear-related C1 and C2 parameters of both wear models were determined by using
the genetic algorithms for test 1 and test 2 separately. The estimated C1 and C2 parameters are
listed in Table 2.
Fig. 11. Wear model of the NAK-55 steel wear test 1.
2206
Fig. 12. Wear model of the NAK-55 steel wear test 2.
Table 2
Accuracy of the cutting force estimations by using the proposed model
Test 1
Test 2
Average
Cutting Conditions
C1
C2
Error
%Error
30,000 rpm spindle speed

2.5 ipm feed rate
0.015 width of cut
0.015 depth of cut
20,000 rpm spindle speed
1.25 ipm feed rate
0.0225 width of cut
0.015 depth of cut
0.0192
4.352
1.1 N
3.3%
0.0240
4.287
2.0 N
7.1%
0.0216
4.320
2.2 N
8.0%
The experimental data was very close to the estimations of the proposed model. The average
deviations of both models were 3.3% and 7.1% for test 1 and 2, respectively. The averages of
C1 and C2 parameters were used to estimate the maximum cutting force at 24 different conditions.
The estimation error was less than 8% for both tests. These results indicate that wear progress
characteristics were very similar in different cutting conditions.
The average parameters C1 and C2 were used to calculate the wear coefficient Kw of the analytical model (Eq. (1) and Eq. (2)). The cutting forces were simulated at different usage levels of a
tool when NAK-55 steel was machined. The simulated and experimental cutting forces in normal
direction of a new and worn (pre-failure) tool of test 1 are presented in Fig. 13 and Fig. 14. They
are almost identical in both figures.
2207
Fig. 13. Simulated and experimental cutting forces during the micro-end-milling operations with a new tool.
Fig. 14.
Simulated and experimental cutting forces during the micro-end-milling operations with a pre-failure tool.
5.3. Estimation of remaining tool life

The remaining tool life estimation accuracy of Eq. (6) was tested on the experimental data
collected by using a carbide end-mill to machine a NAK-55 steel workpiece. The cutting conditions of the tests and estimated parameters are presented in Table 2. The critical cutting force
in normal direction was selected as 40 N by inspecting the cutting force variations just before
the breakage and tool life was estimated at the different stages of tool life by using different
parameter pairs (C1 and C2 obtained for test 1, test 2, average, all data together). Tool life esti-
2208
Table 3
Tool life estimation accuracy of tool 1 by using the C1 and C2 parameters obtained from the data of test 1
Usage of the tool
Actual remaining tool

life
21
24
27
30
33
36
39
Average
21
18
15
12
9
6
3
Estimated remaining
tool life
18.6
14.1
17.6
7.8
11.4
0
1.8
Error
% Error
2.4
3.9
2.6
4.2
2.4
6
1.2
3.2
5.7%
9.4%
6.3%
9.9%
5.7%
14.3%
2.8%
7.7%
mations and errors are presented in Tables 3 and 4. The average error was around 10%. The
results indicated that the proposed model could be used to estimate the remaining tool life.
The smallest estimation error was observed when the C1 and C2 parameters of the same cutting
conditions were used. The average error was around 7.7% and 9.2% for tools 1 and 2.
The average estimation error was very reasonable (around 9.1% and 10.0% for tools 1 and 2)
when the C1 and C2 parameters of the other test condition were used to estimate the remaining
tool life. The results are presented in Table 5 and Table 6.
The errors listed in Tables 36 indicate that operating conditions have very little influence on
the C1 and C2 parameters. It is possible to use the averaged wear parameters obtained from tests
at different operating conditions to estimate the remaining tool life. Estimation errors obtained
by using this approach are presented in Table 7 and Table 8. The average error was around 7.7%
and 8.6% for tools 1 and 2 when the average of the C1 and C2 parameters is used.
6. Conclusion
The characteristics of the cutting forces of the new and worn tools were studied and an analytical model is proposed to simulate the cutting forces of MEMO at different usage levels. The
parameters of the models were estimated by using the genetic algorithms.
Table 4
Usage of the tool

life
Estimated remaining
tool life
15
18
21
24
27
30
Average
18
15
12
9
6
3
13.2
12.5
15
8.8
8.7
6
Error
4.8
2.5
3.0
0.2
2.7
3.0
2.7
% Error
14.5%
7.6%
9.2%
0.6%
8.0%
9.1%
8.2%
Table 5
Usage of the tool

life
Estimated remaining
tool life
21
24
27
30
33
36
39
Average
21
18
15
12
9
6
3
15.0
11.3
14.2
6.3
9.2
0
1.5
Error
6.0
6.7
0.8
5.7
0.2
6.0
1.5
3.8
% Error
14.4%
15.9%
1.9%
13.5%
0.4%
14.3%
3.7%
9.1%
Table 6
Usage of the tool

life
Estimated remaining
tool life
15
18
21
24
27
30
Average
18
15
12
9
6
3
16.4
15.5
18.6
10.9
10.7
7.4
Error
1.6
0.5
6.6
1.9
4.7
4.4
3.3
% Error
4.9%
1.4%
20.1%
5.7%
14.3%
13.4%
10.0%
Table 7
Tool life estimation accuracy of tool 1 by using the C1 and C2 parameters obtained from both test data
Usage of the tool

life
Estimated remaining
tool life
Error
% Error
21
24
27
30
33
36
39
Average
21
18
15
12
9
6
3
18.6
14.1
17.6
7.8
11.4
0
1.8
2.4
3.9
2.6
4.2
2.4
6
1.2
3.2
5.7%
9.4%
6.3%
9.9%
5.7%
14.3%
2.8%
7.7%
2209
2210
Table 8
Tool life estimation accuracy of tool 2 by using the C1 and C2 parameters obtained from both test data
Usage of the tool

life
Estimated remaining
tool life
15
18
21
24
27
30
Average
18
15
12
9
6
3
14.6
13.8
16.6
9.7
9.6
6.6
Error
3.4
1.2
4.6
0.7
3.6
3.6
2.8
% Error
10.2%
3.6%
14.0%
2.2%
10.8%
11.0%
8.6%
The simulated cutting forces of the proposed model were compared with experimental data of
MEMO and very good agreement was observed between the simulated and experimental cutting
force profiles. The error of the cutting force simulations for both new and used tools was less
than 8%.
The proposed analytical model was modified to estimate the remaining tool life from the cutting
force data. The proposed expression estimated the remaining tool life with typically 10% error.
Maximum error was 20%.
It is possible to use the proposed model and genetic algorithm-based identification for on-line
monitoring of tool condition. All the parameters, which vary with the changing cutting conditions,
can be estimated instantaneously together with the cutting force coefficients. With this approach,
the change of depth and width of cut will be immediately recognized and cutting force coefficients
will be estimated accurately. The reference cutting force coefficients can be estimated when the
machining operation starts with a new tool. A monitoring system may update the cutting force
coefficients continuously and calculate the remaining tool life. The system needs only a critical
value of the cutting force that indicates the tool breakage. This procedure eliminates the need for
investigation of the characteristics of cutting forces at different usage levels of that specific tool,
workpiece, and cutting conditions.
References
[1] J. Tlusty, P. Macneil, Dynamics of cutting forces in end milling, Annals of the CIRP 24 (1) (1975) 2125.
[2] P.E. Gygax, Dynamics of single-tooth milling, Annals of CIRP 28 (1) (1979) 6570.
[3] J. Tlusty, F. Ismail, Special aspects of chatter in milling, Trans. of ASME, Journal of Eng. for Ind. 105 (1983)
2432.
[4] F. Ismail, A. Bastami, Improving stability of slender end mills against chatter, Trans. of ASME, Jour. of Eng.
for Ind. 108 (1986) 264268.
[5] Y. Altintas, I. Yellowley, J. Tlusty, The detection of tool breakage in milling operations, Trans. of ASME 110
(1988) 271277.
[6] M. Liu, S.Y. Liang, Monitoring of peripheral milling using acoustic emission, Trans. of NAMRI/SME (1990)
120127.
[7] S. Takata, M. Ogawa, P. Bertok, J. Ootsuka, K. Matushima, T. Sata, Real-time monitoring system of tool breakage
using Kalman filtering, Robotics and Computer-Integrated Manufacturing 2 (1) (1985) 3340.
2211
[8] S. Liang, D.A. Dornfeld, Tool wear detection using time series analysis of acoustic emission, Trans. of ASME,
J. of Engng for Ind. 111 (1989) 199205.
[9] I.N. Tansel, C. McLaughlin, Detection of tool breakage in milling operations: Part 1 The time series analysis
approach, International Journal of Machine Tools and Manufacturing 33 (4) (1993) 531544.
[10] I.N. Tansel, C. McLaughlin, Detection of Tool Breakage in Milling Operations: Part 2 The Neural Network
Approach, International Journal of Machine Tools and Manufacturing 33 (4) (1993) 545558.
[11] W.Y. Bao, I.N. Tansel, Modeling micro-end-milling operations. Part I: analytical cutting force model, International
Journal of Machine Tools and Manufacturing 40 (15) (2000) 21552173.
[12] W.Y. Bao, I.N. Tansel, Modeling micro-end-milling operations. Part II: tool runout, International Journal of
Machine Tools and Manufacturing 40 (15) (2000) 21752192.
[13] N.H. Cook, K. Subramanian, S.A. Basile, Survey of the state of the art of tool wear sensing techniques. Materials
Processing Laboratory, Department of ME, MIT, 1975.
[14] D.E. Goldberg, Genetic Algorithms in Search, Optimization and Machine Learning, Addison-Wesley, Reading,
MA, 1989.
[15] D.L. Carroll, Chemical laser modeling with genetic algorithms, AIAA Journal 34 (2) (1996) 338346.
[16] S.R. Ladd, Genetic Algorithms in C++, M and T Books, 1996.
[17] I.N. Tansel, T.T. Arkan, W.Y. Bao, N. Mahendrakar, B. Shisler, D. Smith, M. McCool, Tool wear estimation in
micro-machining I Tool usage cutting force relationship, International Journal of Machine Tools and Manufacture 40 (4) (2000) 599608.
[18] I.N. Tansel, T.T. Arkan, W.Y. Bao, N. Mahendrakar, B. Shisler, D. Smith, M. McCool, Tool wear estimation in
micro-machining II Neural network-based periodic inspector for nonmetals, International Journal of Machine
Tools and Manufacture 40 (4) (2000) 609620.
[19] I.N. Tansel, W.Y. Bao, B. Tansel, B. Shisler, D. Smith, J. Murray, Identification of cutting conditions by using
an analytical model and genetic algorithms for micro-end-milling operations, smart engineering system design:
Neural networks, fuzzy logic, rough sets and evolutionary programming, in: InIntelligent Engineering Systems
Through Artificial Neural Networks, vol. 8, ASME Press, 1998, pp. 779784.

Modeling Micro End Mill Operation

Uploaded by

Copyright:

Available Formats

Modeling Micro End Mill Operation

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Modeling Micro End Mill Operation

Uploaded by

Copyright:

Available Formats

International Journal of Machine Tools & Manufacture 40 (2000) 21932211

Modeling micro-end-milling operations. Part III: influence of

* Corresponding author. Tel.: +1-305-348-1932; fax: +1-305-348-3304.

feed direction cutting force (N)

2.2. Genetic algorithms

A data chromosome with 20 genes presenting 3 DNA.

Biological evolution cycle of species.

Selection of mating couples (parents) from the population pool.

where: L is the tool life (inch), C1 (1/inch) and C2 are parameters.

Normal direction cutting force (N)

Fig. 8. Cutting edges of a new tool.

Fig. 9. Cutting edges of a worn tool.

5.2. Verification of the tool wear model

Fig. 11. Wear model of the NAK-55 steel wear test 1.

Fig. 12. Wear model of the NAK-55 steel wear test 2.

30,000 rpm spindle speed

5.3. Estimation of remaining tool life

Actual remaining tool

Actual remaining tool

Actual remaining tool

Actual remaining tool

Actual remaining tool

Actual remaining tool

You might also like