Beruflich Dokumente
Kultur Dokumente
discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/222837041
CITATIONS
READS
42
76
3 authors, including:
Yalcin Ertekin
Bill Tseng
Drexel University
SEE PROFILE
SEE PROFILE
887
Department of Architectural and Manufacturing Sciences, Western Kentucky University, Bowling Green, KY, 42101-3576, USA
b
Department of Mechanical and Industrial Engineering, The University of Iowa, Iowa City, IA, 52242-1527, USA
Received 18 December 2002
Abstract
The main focus of this study is to identify the most influential and common sensory features for the process quality characteristics
in CNC milling operationsdimensional accuracy (bore size tolerance) and surface roughnessusing three different material types
(6061-T6 aluminum, 7075-T6 aluminum, and ANSI-4140 steel). The materials were machined on a vertical CNC mill, retrofitted
with multiple sensors and data acquisition systems, to investigate the effects of variations in material types and machining parameters.
The sensor data include cutting force measurements, spindle quill vibration, and acoustic emission, each of which further divided
into measurable components, such as x, y, and z components in cutting force, x and y spindle quill vibration, DC, AC, and Count
Rate for acoustic emission signals. Those components were filtered and analyzed to determine the sensory features that best correlate
with process quality characteristics. Tool wear rate and machining characteristics appeared differently, depending on the material
types, yet some components of the sensory data were found to be significant with relation to the variations in bore size and surface
roughness for all three types of materials. This suggests that even under the varying cutting conditions involving different materials,
the identified sensory features can be used for the reliable and accurate control of milling operations.
2003 Elsevier Science Ltd. All rights reserved.
Keywords: Multiple sensor fusion in milling; Common sensory features; Varying cutting conditions
1. Introduction
The intense international and domestic market competition has driven the attention of manufacturers on automation of manufacturing systems as means for increasing the productivity and improving the quality of
products. This increasing trend towards automation has
created a need for the manufacturing industry to develop
a comprehensive approach to the problem of assuring
the quality [1]. To successfully obtain a component of
the correct size and surface finish the first time from an
automated system, a real-time machining process monitoring and diagnostic system is needed to integrate quality assurance into manufacturing processes. Such a
monitoring system must also replicate the overseeing
function and qualitative judgment of the expert machin-
0890-6955/03/$ - see front matter 2003 Elsevier Science Ltd. All rights reserved.
doi:10.1016/S0890-6955(03)00087-7
898
Y.M. Ertekin et al. / International Journal of Machine Tools & Manufacture 43 (2003) 897904
tiple regression analysis of multi-sensor data with surface roughness and bore tolerance. The listed topics are
presented in the subsequent sections and the findings and
analytical results are summarized in the conclusion.
Rai
Xi
Sj
Xi
j1
H
jL1
Fig. 1.
Experimental set-up.
Sj
Xi
Xi
Sj
jH1
Xi
Xi
jQ1
Sj
(1)
Xi
Xi
Y.M. Ertekin et al. / International Journal of Machine Tools & Manufacture 43 (2003) 897904
comprises variables that can be isolated from the experiment [5], which includes the flow of coolant, utilization
of chip breaker, work-holding devices, and selection of
tool types. By controlling those variables, one can isolate
or nullify their effects on the surface roughness. The
third group is a source of systematic uncertainty that is
neither measured nor isolated, yet their effects cannot be
neglected [5], including variations in machine repeatability, machine vibration and damping, cutting temperature, chip formation (chip geometry), chip exit speed,
variations in work material, cutting inserts, thermal
expansion of machine tool, and spindle motor power
consumption. The fourth group is also a source of systematic uncertainty, yet the effects here are assumed to
be very small or negligible [5], including room temperature, humidity, the amount of dust in the air, and fluctuations in the power source. Controlling and measuring
all the variables are infeasible, hence only the fraction
of the variables are ever measured and their relationships
with machining quality characteristics are derived. This
is referred to as a machining process model that represents a mapping of input and output variables for specific machining conditions [6].
In machining, the formulation of accurate mathematical models is very difficult, hence in most cases, empirical equations are obtained by performing statistical
analysis or through the training of neural network models to fit the experimental data. To collect the data, a
designed experiment is performed using predetermined
values of machining parameters. Even though empirical
equations do not consider the underlying principles and
mathematical relationships [7], empirical modeling does
have advantages of accuracy and ease of formulation [6].
One noticeable drawback of any empirical modeling is
the rigidity (narrow scope) that allows only limited variations in the machining conditions. That is, the cutting
conditions must be carefully controlled to remain similar
to those used in the cutting experiments. Cutting
materials from different batches may considerably
degrade the accuracy of the model since variations in
raw materials (e.g., hardness, strength, microstructure,
heat treatment conditions, and thermal properties) exist
between workpiece to workpiece or lot to lot [7]. Development of empirical models often leads to a large number of trials and a considerable amount of time and
effort, however, whenever variations occur, empirical
models need to be refitted to maintain the accuracy. To
avoid costly reproduction of cutting experiments under
the diverse production environment, while satisfying the
quality characteristics of the machined products, reliable
and accurate means of tool condition monitoring and
process control become necessary. Excellent reviews of
in-process monitoring techniques, classified as direct and
indirect techniques, have been made by many
researchers [8,9,10]. Indirect monitoring infers the tool
condition monitoring while machining is taking place so
899
that on-line implementation is possible. The main purpose is to prevent tool breakage. Various indirect monitoring techniques have been proposed, but most of them
have been limited to tool wear monitoring and tool
breakage detection using various process variables such
as cutting force [11,12,13,14,15,16], vibration
[17,18,19], temperature [20], acoustic emission
[21,22,23,24] and displacement [25]. The direct in-process monitoring techniques are of rather limited use and
sometimes unacceptable due to their slow down of the
manufacturing cycle.
In this study, all machining experiments were conducted on a Cincinnati Milacron Sabre-750 CNC vertical
machining center at Cincinnati Milacron Regional Technical Center in Arlington Heights, Chicago, IL. The
experimental set-up is shown in Fig. 1 and the drawing
of the machined components and its dimensions are
shown in Fig. 2. The components were machined from
6061-T6 Al, 7075-T6 Al, and ANSI-4140 steel blank
workpieces of size 130 mm 100 mm 63.5 mm. A
25.4 mm-diameter cobalt high speed steel (HSS) drill
was used to drill a starting hole. Then, a 25.4 mm-diameter, 2-flute cobalt HSS end mill was used to mill the
84 mm 84 mm square and the 70, 65, and 50 mm
bores. Fresh tools were assigned for each material and
the cutting continued until tools wore out. Up milling
was used to machine the square and down milling for
the three bores. Machining was done in the following
order: drill a 25.4 mm hole, rough mill the 84 mm 84
900
Y.M. Ertekin et al. / International Journal of Machine Tools & Manufacture 43 (2003) 897904
N
n
60
(2)
Table 1
Cutting parameters and material types used in the experiment
Material
Cutting speed
(m/min)
6061-T6 Al
7075-T6 Al
4140 Steel
95
150
200
168
84
28
2100
1050
350
Feed rate
(mm/min)
304.8
101.6
25.4
Radial depth of
cut (mm)
6
12.7
3
Axial depth of
cut (mm)
Cutter tooth
frequency (Hz)
12.7
30
3
70
35
23.33
Y.M. Ertekin et al. / International Journal of Machine Tools & Manufacture 43 (2003) 897904
Fig. 4.
901
Fig. 5.
902
Y.M. Ertekin et al. / International Journal of Machine Tools & Manufacture 43 (2003) 897904
Fig. 6.
Fig. 7.
Y.M. Ertekin et al. / International Journal of Machine Tools & Manufacture 43 (2003) 897904
to tool wear, the AEDC signal was selected as the sensory feature for multiple regression analysis.
In Sections 3.1 and 3.2, the case of 4140 steel is
presented, rather than providing a detailed description
for each material. It is because the steel is much harder
than the aluminums and the machining characteristics
appear to be more unstable (higher wear rate, rougher
surface finish, wider tolerance distribution due to larger
cutting force and bigger tool deflection, heavier
vibration, etc.) than those of the aluminums.
903
(3)
0.0146 AEDC3.
This indicates a very high correlation between Ra and
Fv and AEDC. The regression equation plot shows that
when there is a very low Fv in the machining process,
the highest surface roughness value is observed. Another
interesting observation is that AEDC response is insensitive to changes in surface roughness until the Fv reaches
approximately 163 N. Then, the AEDC starts becoming
sensitive. A simultaneous increase in both RMS values
of the Fv and AEDC indicate higher surface roughness
values, again showing the importance of monitoring
more than one process variable simultaneously in real
time. Their integration gives a more reliable and accurate
prediction of bore tolerance than when using a single
sensory feature.
3.2. Multiple regression of bore tolerance on Fv and
AEDC
Multiple regression analysis was done using bore tolerance (Tol) as the dependent variable. The quality
characteristic is the Tol of the 50 mm bore. Using the
forward elimination technique, the common sensory features for all three material types were again found to be
the Fv and AEDC. The best model is in the form of:
Tol 8.532880.1749 Fv 0.001163 105 F2v
(4)
904
Y.M. Ertekin et al. / International Journal of Machine Tools & Manufacture 43 (2003) 897904
[12]
[13]
[14]
[15]
[16]
Acknowledgements
The financial support of this work by the National
Science Foundations Manufacturing Machines and
Equipment program in the Division of Design and
Manufacturing Systems and Cincinnati Milacron Inc. is
gratefully acknowledged.
[17]
[18]
[19]
References
[1] K. Wright, K. Paul, D.A. Bourne, Manufacturing Intelligence,
Addison-Wesley Publishing Co. Inc, 1988.
[2] T. Matsumura, T. Obikawa, T. Shirakashi, E. Usui, Autonomous
turning operation planning with adaptive prediction of tool wear
and surface roughness, Journal of Manufacturing Systems 12 (3)
(1993) 253262.
[3] S. Krar, J. Oswald, Technology of Machine Tools, McGraw-Hill,
New York, 1990.
[4] C.M. Shaw, Metal Cutting Principles, Oxford University Press,
New York, 1986.
[5] Y. Ronen, Uncertainty Analysis, CRC Press Inc, Boca Raton,
FL, 1988.
[6] K. Park, S. Kim, Artificial intelligence approaches to determination of CNC machining parameters in manufacturing: a review,
Artificial Intelligence in Engineering. 12 (1998) 127134.
[7] M. Groover, Automation, Production Systems, and Computer
Integrated Manufacturing, Prentice-Hall International, Inc, Englewood Cliffs, NJ, 2001.
[8] G.F. Micheletti, W. Koenig, H.R. Victor, In-process tool wear
sensors for cutting operations, Annals of CIRP 25 (1976) 483
496.
[9] J. Tlusty, G.C. Andrew, A critical review on sensors for
unmanned manufacturing, Annals of CIRP 32 (2) (1983) 563
572.
[10] C.A. Lutterwelt, T.H.C. Childs, I.S. Jawahir, F. Klocke, P.K.
Venuvinod, Present situation and future trends in modeling machining operations, Annals of CIRP 47 (2) (1998) 587628.
[11] A. De Filippi, R. Ippolito, Adaptive control in turning: cutting
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
forces and tool wear relationship for P10, P20, P30 carbides,
Annals of the CIRP 17 (1969) 377385.
Y. Altintas, I. Yellowley, J. Tlusty, The detection of tool breakage in milling, in: Sensors and Controls for Manufacturing,
WAM of the ASME, Florida, November 1722, 1985, pp.
101110.
M.S. Lan, D.A. Dornfeld, Experimental studies of tool wear via
acoustic emission analysis, in: Proceedings of the 10th NAMRC,
SME, 1982, pp. 305311.
L.C. Lee, K.S. Lee, C.S. Gran, On the correlation between
dynamic cutting force and tool wear, Int. J. Mach. Tools Manufact 29 (3) (1989) 295303.
A.C. Okafor, Y.J. Chyou, Comparative investigation for on-line
end mill wear and breakage detection for unmanned machining,
in: Sensors and Controls for Manufacturing, WAM of the ASME,
Illinois, November 27December 2, 1988, pp. 105114.
P.K. Venuvinod, Sensor-based modeling of machining operations, in: CIRP International Workshop on Modeling of Machining Operations, Atlanta, GA, USA, 1998, pp. 397405.
V.A. Stewart, Nose and flank contact when machining under
chatter conditions, in: Proceedings of the International Conference on Production Technology, Institution of Engineers, Melbourne, 1974, pp. 167174.
N.K. Mehta, P.C. Pandy, G. Chakravarti, An investigation of tool
wear and the vibration spectrum in milling, Wear 91 (1983)
219234.
K.W. Yee, L. Evans, Drill-up an alternative for on-line determination of end mill wear, in: Proceedings of 13th NAMR Conference, May 1922, 1985, pp. 304309.
S.B. Rao, Tool wear monitoring through the dynamics of stable
turning, Journal of Engineering for Industry 108 (1986) 183190.
S. Kim, B.E. Klamecki, Milling cutter wear monitoring using
spindle shaft vibration, in: Monitoring and Control for Manufacturing Processes, Winter Annual Meeting of ASME, Dallas,
Texas, November 2530, 44, 1990, pp. 5774.
L.U. Colwell, Cutting temperature versus tool wear, Annals of
the CIRP 24 (1) (1975) 7376.
T. Moriwaki, Detection of cutting tool fracture by acoustic emission measurements, Annals of CIRP 29 (1980) 1.
E. Kannatey-Asibu Jr., D.A. Dornfeld, Quantitative relationships
for acoustic emission from orthogonal metal cutting, ASME,
Journal of Engineering for Industry 103 (3) (1981) 330440.
I.H. Kim, D.Y. Jang, In-process sensing of tool wear and process
states using a cylindirical displacement sensor in hard turning,
Proc. Instn., Mech. Engnrs 215B (2001) 16731682.
S.K. Dontamsetti, G.W. Fischer, Factors affecting surface roughness in finish turning of gray cast iron, Advanced Materials &
Manufacturing Process 3 (4) (1988) 689725.
V. Arshinov, G. Alekseev, Metal Cutting Theory and Cutting
Tool Design, Mir Publishers, 1976.
G. Boothroyd, W. Knight, Fundamentals of Machining and
Machine Tools, Marcel Dekker, Inc, 1989.
A.C. Okafor, M. Marcus, R. Tipirneni, Multiple sensor integration via neural networks for estimating surface roughness and
bore tolerance in circular end milling-Part 1: Time domain, Condition Monitoring and Diagnostic Technology 2 (2) (1991) 49
57.
V.A. Kudinov, Dynamics of Machine Tools, Mashinostroenic,
Moscow, 1967.