Professional Documents
Culture Documents
1.feasibility of Tool Condition Monitoring On Micro-Milling
1.feasibility of Tool Condition Monitoring On Micro-Milling
2011)
Abstract
This paper reviewed existing studies in the area of tool condition monitoring. The
review revealed the substantial amount of works that have been done in the area of tool
condition monitoring in machine tool. However, it was observed that investigating the
feasibility of using the main spindle and feed motor currents for tool condition
monitoring in micromachining has not been carried out despite their advantage of low
cost and simplicity. Therefore this study went further to investigate spindle and feed
motor current signals as to their ability for sensing and monitoring tool wear
progression on a micro milling machine. This initial investigation revealed some
promising process signals for monitoring tool wear progression in micro milling.
Keywords: Micro machine tool, micro machining, tool condition monitoring,
spindle motor, feed motor, current signal.
has chipped during a machining cycle, and thus In general tool condition monitoring
might continue with the machining, wasting comprises three steps, namely sensory signal
precious production time. Also, damage to a acquisition, feature extraction and decision
workpiece by a broken tool may not be making (see Fig. 1). The first step is to capture
observable without special instruments. the signals. A variety of sensors can be used to
The most appropriate strategy available capture information that may reveal the state of
to manufacturers and researchers to reduce the the tool during the manufacturing process.
possibility of tool failures is to introduce some Such monitoring signals may be force,
form of feedback representing the condition of vibration, sound or acoustic emissions,
the tool during the machining process. This temperature, power and motor current. The
feedback system for the machine tool is second step is to process the signals, from
referred to as tool condition monitoring which a set of monitoring indices, also called
(TCM). According to Donnell et al. (2001), the features or attributes, are obtained.
main goal of TCM is to increase productivity
and competitiveness by maximizing tool life,
minimizing machine down time, reducing
scrappage and damage. The traditional ability
of the operator to determine the condition of
the tool based on his/her experience and senses,
i.e. vision and hearing, is now the expected role
of the TCM system.
At very high spindle speeds, which are
usually used in micro machining, with the
miniaturization of the cutting tools, TCM is
critical and becomes an important issue (Tansel
et al. 2000; Chae et al. 2006). TCM can reduce Fig. 1. Illustration of the setup for tool condition
not only the production costs by reducing monitoring.
downtime and unnecessary tool changes, but it
can also improve the product quality. In Typical monitoring indices include (a)
general, TCM includes monitoring tool wear time-domain indices, such as mean, variance,
and tool breakage. root mean squares (RMS), threshold crossing
Tool wear monitoring methods are counts; (b) frequency-domain indices, such as
classified into direct and indirect methods (Lin the energy at a specific frequency band,
and Ting 1996; Ghani et al. 2002). The direct integral of amplitude peak, natural frequency,
method is implemented using a dedicated and damping ratio; (c) time-frequency domain
instrument, such as a tool maker optical indices; (d) higher order spectrum indices; and
microscope, to measure the size of the wear (e) wavelet indices (Du 1999). The last step is
directly, while the indirect method is based on the decision making step (which leads to
the acquisition of measured values of process classification), wherein monitoring indices are
variables (such as the change of cutting force, used to classify the present tool condition with
temperature, vibration, spindle motor current, respect to a predefined tool condition.
acoustic emission and surface roughness) and Classification is very important for the success
establishing relationships between these values of a tool condition monitoring system. Hence,
and tool wear (Lin and Ting 1996). Although tool wear classification algorithms should
indirect methods are more complex in terms of consider whether the features are
determining the actual tool wear, they are less discriminating enough, and avoid unnecessary
difficult to implement into the operation of a computation that may hinder the subsequent
machine. This is the reason why they are more classification process.
suitable for practical application and why they
are more widely used.
some neural network approaches could be In another study, the spindle motor
employed to efficiently select features that are current of a horizontal machining centre was
highly correlated with tool wear for TCM. measured using current hall sensor during
Al-Habaibeh and Gindy (2000) utilised drilling and the cutting torque was measured by
Taguchi's orthogonal arrays to design an a Kistler cutting force dynamometer, type
experimental programme for investigating the 9123B, (Oh et al. 2004). The cutting torque
number and type of sensors and signal was estimated from the measured spindle
processing methods most appropriate for the motor current and validated when it was
detection of fault conditions in milling compared with the measured cutting torque.
processes. In addition to recording AE and Results show that the drilling torque can be
cutting forces, six accelerometers were used to estimated accurately from the spindle motor
monitor the vibration of the spindle and the current with an average estimation error of
workpiece. The goal of the study was to reduce 6.6%. Subsequently, the cutting torque
the number of sensors needed to detect cutter measured indirectly from the spindle motor
deterioration and thus reduce the cost and time current was controlled by regulating the spindle
required for developing effective tool condition motor current through the machine controller.
monitoring systems for machining operations. Experimental results showed that the drilling
The study was able to compare 17 TCM torque was well-regulated at a given reference
systems with respect to the apparatus used, level and the risk of drill failure and drill flank
their cost and performance and it concluded wear were reduced remarkably through cutting
that it is essential to compromise between cost torque control.
and performance of the systems if a cheaper Xiaoli (1999) performed drilling cutting
system with good performance is needed. tests on quench steel workpiece material using
Haber et al. (2004) investigated tool- a Makino-FNC74-A20 machining center and
wear monitoring in high-speed milling based measured the ac servo motor current signals of
on the different signal signatures in the time the machine through the hall current sensor.
and frequency domain. Sensorial information The process signals were decomposed using
from a dynamometer, accelerometer and an wavelet transform and tool breakage features
AE-sensor were acquired, compared and were extracted from the decomposed signals.
analyzed. With respect to the time domain, the Subsequently, the study was able to detect tool
study concluded that the mean and peak values breakage in the drilling process for small
of cutting force and vibration exhibit the best diameter drill.
performance for tool-condition monitoring. Also, Kim et al. (2002) developed a drill
With respect to the frequency domain, the wear estimation model in terms of the drilling
study established that the AE-signals are torque. Drilling experiments were then
sensitive to changes in the tool’s condition, performed on a vertical machining centre and
with some peaks exhibiting a change in the spindle motor power was recorded using
magnitude depending on the sensor’s location. the load meter built into the machine controller.
However, in the case of a worn tool, the The drilling torque was estimated from spindle
amplitudes were found to increase irrespective power while tool wear was in turn estimated
of the sensor position. from the drilling torque. The wear estimation
Li et al. (2000) measured the feed cutting approach was found to be reliable with an
force using an inexpensive current sensor estimation error below 0.02 mm.
installed on the AC servo motor of a CNC The work described above is only a
turning centre. The study analysed the factors fraction of the work done on TCM at the
that affected a feed drive system in detail, and conventional large scale machining level. Most
presented a model of the feed drive system for of the research work consolidates the fact that
estimating feed cutting force with respect to the spindle motor current/power, axis drive
feed motor current. The feed cutting force was current/power, AE signal, vibration signal and
then estimated using the feed motor current cutting force signal have been widely used
with an error of less than 5%. within the indirect method to monitor tool
conditions during various machining operations and an optical vision system were used to
(Byrne et al. 1995; Jemielniak 1999; Prickett monitor tool conditions. The recorded signals
and Johns 1999; Sick 2002). were analyzed off-line and later applied to a
Despite more than a decade of intensive neuro-fuzzy method to fuse the signals together
research, the development of tool wear in a neural network environment. It was
monitoring systems is still on-going. Chae et reported that a comparison between the actual
al. (2006) reported that while tool wear tool wear and the simulated results from the
monitoring has been extensively studied at the neuro-fuzzy method showed good agreement,
conventional large scale machining level, very with 0.1222 minimum training error, when the
limited work has been conducted at the micro- resultant force, acceleration and AE signals
scale. An overview of studies at the micro were fused together. Another approach applied
machining level shows that Jemielniak et al. to TCM in micro milling operation investigated
(2008) based their strategy on an AE sensor an on-line laser system and an on-line tool-
and cutting force signals to monitor tool wear workpiece voltage monitoring system for tool
in micro-milling of cold-work tool steel. The condition monitoring (Gandarias et al. 2006).
cutting experiments were carried out on a high- These approaches involve connection of
precision milling machine, equipped with a electrical circuit and/or reflective laser sensors
50,000 rpm spindle and an HSK 25 tool holder. between the workpiece and the tool making the
The workpiece was clamped on a three-axis system expensive and more complex in terms
9256C1 mini-dynamometer alongside an AE- of additional sensors and mounting/removal of
sensor (type 8152B221). Signals from the two both the tool and workpiece.
sensors were acquired and the test was
regularly interrupted to measure tool wear
using an optical microscope. Results of the 3. The Challenge of Tool Condition
study revealed that the cutting forces and AE Monitoring on Micro-Machine Tool
signals provided robust and reliable
information for the tool condition monitoring From the review conducted on existing
in micro milling. In another study, Zhu et al. studies discussed in the second section, it is
(2008) extracted wavelet packets and time- clear that the apparatus used for the acquisition
domain features from cutting force signals and of process signals for TCM consists of AE
presented a discriminant feature selection sensors, accelerometers, dynamometers and
approach for the Hidden Markov Model power and/or current sensors attached to
(HMM) of tool conditions in micro-milling of various elements of a machine, cutting tool or
copper and steel. The cutting force was workpiece. At times it may involve fabrication
measured with a Kistler 9256A 3-channel of tailor-made fixtures for locating sensors as
dynamometer mounted under the workpiece, in Haber et al. (2004) depending on the nature
and tool wear was measured using a of machining operation which can make the
microscope. Though the working conditions process cumbersome, especially when many
were at micro level, a conventional large scale sensors are used on a micro machine tool.
machine tool, the MAKIMO V55 vertical Though the approach outlined within Al-
milling machine driven by 22KW spindle drive Habaibeh and Gindy (2000) could be
motor, was used for the experiments. Malekian implemented to reduce the number of sensors,
et al. (2009) examined factors affecting tool and hence the cost, without compromising the
wear and investigated a tool wear monitoring system's ability, the pair goes on to imply that
method using various accelerometers, force and simplicity and cost effectiveness is an
acoustic emission sensors in micro-milling. A important feature of any TCM system. It is
micro-CNC machining centre equipped with a essential that the system uses simple and cheap
300 W spindle was used in the experiment. sensors that are wear and maintenance free and
Several sensors, including a miniature Kistler do not obstruct the working space (Tonshoff et
3-axis force sensor (type 9017B), an AE al. 1988).
sensor, accelerometers and capacitance sensors,
Moreover, with the ever increasing using the feed and spindle motor current signals,
demand for reduced production costs and of a micro machine, for sensing tool wear
improved product quality, machining process progression in micro milling would be an interesting
condition monitoring, possibly without costly area of investigation for cost effective tool
sensor equipment, is an important aspect of wear monitoring. Thus, this study investigates
modern manufacturing. It has been reported and discusses the feasibility of utilizing the
that sensors to measure current or effective main spindle and feed motor current signals of
power of feed drives or the main spindle a micro milling machine for sensing tool wear
represent an alternative to measuring cutting progression in a milling operation.
forces using expensive dynamometers and An attempt was made in this study to find
force sensors. Another advantage is that current a method of extracting the signals without
sensors are easy to retrofit into the machine, hampering the cutting operations. This is
which is the reason why they have already been mainly possible through sensorless data
employed for tool condition monitoring acquisition approach. The investigation was
frequently(Byrne et al. 1995). carried out on a micro milling machine tool
The spindle motor current of a Tree 1000 developed in Ogedengbe (2010). Effort was
lathe was among the signals utilized by made during the development of the machine to
Dornfeld (1990) for successful monitoring of use controller supporting hardware modules
progressive tool wear in a single point turning built with digital signal processors (DSPs) so
operation. Also the feed cutting force which is that their capability could be harness for
a well known measurand in TCM has been acquisition of the current signals without using
estimated from the feed motor current using external sensors. Then the trend of the acquired
neuro-fuzzy technique, with 95% accuracy (Li signals was plotted against tool life to observe
et al. 2000). This approach was extended to how well the various signals correlate with the
tool condition monitoring by Li et al. (2004), progression of tool life.
wherein the feed-cutting force was estimated
using inexpensive current sensors installed on
the AC servomotor of a conventional large 4. Experimental Setup and Method
scale turning centre, with the results applied to
tool wear monitoring system. This shows that As mentioned earlier the experiments
the use of spindle motor current and feed motor were performed on a micro milling machine
current signals for tool condition monitoring tool having the X/Y/Z axis stroke as 50/50/50
has been successful on conventional large-scale mm. The maximum spindle speed is 100,000 rpm.
machine tools for macro mechanical machining A detail description of the machine is available
operations. However, a similar approach has in earlier study (Ogedengbe 2010). The
not been investigated in micro milling. One workpiece material was EN8 steel, supplied in
reason might be because some of the machines bar form with 25 mm × 25 mm edge length. It
used for the research in micro milling are was cut in pieces of about 30 mm length, which
conventional precision machine tools rather could be clamped on the machine’s XY table.
than small micro machines. The approach may The micro-tools used in this study were two-
not be feasible on such machines due to very fluted uncoated micro-grain tungsten carbide
low cutting forces involved in micro milling micro end mills, supplied by Fraisa
compared to the relatively high friction forces (www.fraisa.com). The specification of these
in the guideways and the spindle. However, on tools is given in Table 1. The tools were used
a micro machine tool with direct drive feed in a face milling operation with the tool
axes and an air bearing spindle, the friction moving in the X axis direction. Table 2 gives
components are likely to be relatively small. the experimental conditions employed in the
But the problem that could be encountered is tests. Based on the reviewed literature, both the
where and how to attach the current sensors to power consumed by the main spindle as well as
relevant elements of the micro machine. If this the feed drives was investigated.
problem could be solved, then the feasibility of
Table 1. Cutting tool specification. the X, Y, and Z feed motor and the spindle
motor current signals were collected for every
Flute Shank Helix Rake
Part No. Diameter
length diameter Angle Angle other pass of the tool on the workpiece.
M45710.100 1 mm 3 mm 3mm 30º 10º Customarily, crater wear (KT)) or flank
wear (VB) are usually used as the tool
Table 2. Experimental conditions. condition indicators in TCM. While crater wear
Axial Radial
Spindle
is the type of tool wear characterized by a
Operation depth of
cut
depth of
cut
Feed rate
speed concave depression in the rake face of the
Face
100 µm 250 µm
240 30,000 cutting tool adjacent to the cutting edge, flank
milling mm/m in rpm wear is the result of gradual wearing away of
the cutting edge. Flank wear, mostly caused by
The Automation 3200 soft motion abrasion, is the most desired form of tool wear
controller used on the machine tool can collect indicator. Hence, a total wear in flank VBmax of
real time information about the performance of 100 µm was used as the criterion for tool life.
the feed motors as well as the motor for the Pictures of each of the two flutes of the tool,
main spindle. This performance information taken in situ, during the cutting test were
which includes feedback current signals can be measured and the average wear in the flank of
acquired through the Ndrive CL. For this, a the two flutes was taken as the final flank wear
program was written to collect the feedback value as given in Eq. ((1), where VB1 and VB2
current signal of all the three linear axes as well are the flank wear of the first and the second
as the main spindle without using any external flute respectively. An example of tool wear
sensor. The data collected was saved to a file in measurement is shown in Fig. 3. The cutting
text format. At the spindle speed of 30,000 tests were stopped when a total flank wear of
rpm, using a two-fluted end mill, the tooth 100 µm was exceeded.
passing frequency was 1 kHz. Given due
regard to the Nyquist-Shannon sampling VB1 + VB2
theorem the signals were sampled at a rate of 8 VB = . (1)
kHz. Fig. 2 shows the experimental setup 2
employed in these tests.
part of tool life (i.e. between the first one and change seems to vary slightly from one signal
half to two minutes of starting the milling to the other. This observation can be used to
operations). Also shown in Fig. 4 are typical draw a conclusion that all the signals have a
photographs of progressive tool wear of Tool 1. strong correlation with the progression of tool
Next the feedback current signals life. The signals acquired during the lives of
acquired during the cutting tests were analysed Tools 2 and 3 exhibited a similar characteristic
both in the time and frequency domain to as the ones depicted in Figs. 5 to 8 and was
examine their usability for tool condition therefore not displayed.
monitoring. The three-dimensional curves of It has to be mentioned that the end mill
the raw process signals were generated in terms broke during the 135th pass, which caused the
of the number of cutting passes to assess the spindle to stall and the feedback system to
entire spectrum of acquired process signals receive abnormally high spindle power signals.
both in the time and frequency domain. As this was not recognised in time, the
machining was continued with the broken tool
and stalled spindle shaft for passes 136 to 138.
This explains the excessively high spindle
power recorded during these passes 136 to 138.
Unlike the spindle power signal, the feed
current signals of the X, Y and Z stages dropped
close to zero, as the tool after it had broken was
no longer in contact with the workpiece, which
means that only the internal friction of the
stages had to be overcome. This could be used
to further imply that the current signals can
identify not only when the tool has worn out
but also when the tool has broken.
Further analysis was conducted in the
frequency domain. Using the Fast Fourier
Fig. 4. Flanks wear curves for cutting tests and Transform (FFT), the intensity of the
pictures of Tool 1 at various stages. frequencies contained in the acquired signals
was calculated. In order to better visualise the
result, plotting the FFT results from all the
In respect of the number of passes of the passes was not done. Instead, only the
tool over the workpiece, the tool lives were 134 frequency intensity plots of every 10th pass, i.e.
passes, 124 passes and 135 passes for tools 1, 2 passes one, 1,11, 21, 31, and so on, were
and 3, respectively. As an example, Figs. 5, 6, plotted, see Error! Reference source not
7 and 8 depict the feedback current signals found. 9, 10, 11 and 12 (for Tool 1). To
sampled over the life of the first tool, with eliminate severe truncation of the overall
which 138 passes were machined. The spindle frequency intensity due to the high amplitude
and X-axis feed motor current signals are at around 0 Hz, the frequency axis starts at 100
shown in Fig. 5 and Fig. 6 while the Y-axis and Hz. As with the time domain graphs, Tools 2
Z-axis feed motor current signals are shown in and 3 exhibited a very similar behaviour in
Fig. 7 and Fig. 8, respectively. From the graphs terms of their frequency plots and are therefore
it becomes obvious that the four different not presented here.
signals showed a similar trend with progressing From Figs. 10, 11 and 12 it can be seen
tool life, so much so that all of them exhibited, that in the case of a new tool the feed currents
firstly, an increase in magnitude and, secondly, for the x, y and z stages all exhibit a distinctive
a strong increase in fluctuation within each peak at 1 kHz, which matches the cutting edge
pass. This increase takes place continuously engagement frequency for the two-fluted end
throughout the entire tool life with notable mill rotating at 30,000 rpm, and its harmonics.
variation in the rate of change between In the case of the X feed current, Figure 10, the
consecutive milling passes though the rate of
Regular Paper 168
AU J.T. 14(3): 161-172 (Jan. 2011)
intensity plot of the frequencies changes with The fact that in the case of the Z feed
progressing tool life. During the first 25 to 50 current signal the frequency peak at 500 Hz
passes of the tool over the workpiece, the exceeds the frequency at 1 kHz in terms of
frequency plot is characterised by distinctive magnitude could have been brought about by
peaks at 1, 2 and 3 kHz, as pointed out before, the cutting tool exhibiting wear which was
and a very low intensity between these peaks. more severe on one cutting edge than on the
As the tool begins to wear, however, the other. As a result, the cutting edge engagement
intensity plot dominated by the three peaks frequency of one edge (500 Hz) would be
turns into a plateau-like shape, spanning from dominant over the tool engagement frequency
around 500 to about 2,500 Hz, with the peaks of the two-fluted tool (1 kHz).
almost completely vanished. This change
cannot be observed for the Y and Z feed current
signals however, which retain the distinctive
peaks at 1 kHz and their harmonics throughout
the entire tool life. Unlike the frequency
spectra for the three linear axes, the frequency
spectrum of the spindle power, see Figure 9,
does not exhibit distinctive peaks at all. It
could be seen on Figures 9 to 12 that the
amplitude of the distinctive peaks exhibited by
the curves representing frequency domain
analysis of the current signals slightly changes
with tool progression. Even though these
changes are not very obvious in this case as
compare with the time domain analysis it could
be inferred that these frequency intensities
possess certain information that could represent
the condition of the tool.
Fig. 9. Change in amplitude peak of Spindle Fig. 12. Change in amplitude peak of Z-axis
current signal over tool life (Tool 1). feed current signal over tool life (Tool 1).
6. Conclusion
Four signals, namely spindle motor
current as well as the X-, Y- and Z-axis feed
motor currents, were recorded and examined
both in the time and frequency domain in order
to identify their feasibility for tool wear
progression monitoring in micro milling. The
results suggest that these signals possess
characteristic trends that would be useful for
monitoring tool wear progression in micro
milling. However, it is important that in order
to quantify the change in the characteristics of
the process signals brought about by tool wear
so that the same can be used for decision
making during tool condition classification, Ghani, A.K.; Choudhury, I.A.; and Husni, A.
some distinctive features would have to be 2002. Study of tool life, surface roughness
extracted from the recorded process signals in and vibration in machining nodular cast iron
relation to tool flank wear both in time and with ceramic tool. J. Mater. Process. Tech.
frequency domains. Hence this study will 127: 17-22.
continue in this direction with a view to Haber, R.E.; Jiménez, J.E.; Peres, C.R.; and
establish an effective tool condition monitoring Alique, J.R. 2004. An investigation of tool-
system for micro milling using spindle and wear monitoring in a high-speed machining
feed motor current signals. process. Sensors Actuators A, 116: 539-45.
Jemielniak, K. 1999. Commercial Tool Condition
Monitoring Systems. Int. J. Adv. Manuf.
7. References Tech. 15: 711-21.
Jemielniak, K.; Bombin, S.; and Aristimuno,
Al-Habaibeh, A.; and Gindy, N. 2000. A new P.X. 2008 Tool condition monitoring in
approach for systematic design of condition micromilling based on hierarchical
monitoring systems for milling processes. J. integration of signal measures. CIRP Ann.
Materials Proc. Technol. 107: 243-51. Manuf. Tech. 57: 121-24.
Byrne, G.; Dornfeld, D.; Inasaki, I.;Konig, W.; Jemielniak, K.; Kwiatkowski, L.; and Wrzosek,
and Teti, R. 1995. Tool Condition P. 1998. Diagnosis of tool wear based on
Monitoring (TCM) - The Status of Research cutting forces and acoustic emission
and Industrial Application. Ann. CIRP 44: measures as inputs to a neural network. J.
541-67. Intell. Manuf. 9: 447-55.
Chae, J.; Park, S.S.; and Freiheit, T. 2006. Jemielniak, K.; and Otman, O. 1998. Tool
Investigation of micro-cutting operations. failure detection based on analysis of
Int. J. Mach. Tools Manuf. 46: 313-32. acoustic emission signals. J. Mat. Proc.
Chen, S.L.; and Jen, Y.W. 2000. Data fusion Tech. 76: 192-97.
neural network for tool condition Kim, H.Y.; Ahn, J.H.; Kim, S.H.; and Takata,
monitoring in CNC milling machining. Int. S. 2002. Real-time drill wear estimation
J. Mach. Tools Manuf. 40: 381- 400. based on spindle motor power. J. Mat. Proc.
Dimla-Sr, D.E.; and Lister, P.M. 2000. On-line Tech. 124: 267-73.
metal cutting tool condition monitoring. I: Li, X.; Li, H.X.; Guan, X.P.; and Du, R. 2004.
force and vibration analyses. Int. J. Mach. Fuzzy estimation of feed-cutting force from
Tools Manuf. 40: 739-68. current measurement - A case study on
Donnell, G.O.; Young, P.; Kelly, K.; and intelligent tool wear condition monitoring.
Byrne, G. 2001. Towards the improvement IEEE Trans. Syst. Man Cybern. - Part C:
of tool condition monitoring systems in the Applic. Rev. 34: 506-12.
manufacturing environment. J. Mater. Li, X.; Venuvinod, P.K.; and Chen, M.K. 2000.
Process.Tech. 119: 133-39. Feed cutting force estimation from the
Dornfeld, D.A. 1990. Neural network sensor current measurement with hybrid learning.
fusion for tool condition monitoring. Ann. Int. J. Adv. Manuf. Tech. 16: 859-62.
CIRP 39: 101-5. Lin, S.C.; and Ting, C.J. 1996. Drill wear
Dow, T.A.; Miller, E.L.; and Garrard, K. 2004. monitoring using neural Networks. Int. J.
Tool force and deflection compensation for Mach. Tools Manuf. 36: 465-75.
small milling tools. Precis. Engin. 28: 31-45. Malekian, M.; Park, S.S.; and Jun, M.B.G.
Du, R. 1999. Signal understanding and tool 2009. Tool wear monitoring of micro-
condition monitoring Engin. Applic. Artif. milling operations. J. Mat. Proc. Tech. 209:
Intell. 12: 585-97. 4,903-14.
Gandarias, E.; Dimov, S.; Pham, D.T.; Ivanov, Malekiana, M.; Park, S.S.; and Jun, M.B.G.
A.; Popov, K.; Lizarralde, R.; and Arrazola, 2009. Tool wear monitoring of micro-
P.J. 2006. New methods for tool failure milling operations. J. Mat. Proc. Tech. 209:
detection in micromilling. Proc. IMechE 4,903-14.
Vol. 220 Part B: J. Engin. Manuf. 137-44.
Regular Paper 171
AU J.T. 14(3): 161-172 (Jan. 2011)