Sie sind auf Seite 1von 9

Int J Adv Manuf Technol (2006) 30: 10211029

DOI 10.1007/s00170-005-0156-0
ORIGINAL ARTI CLE
E. Daniel Kirby
.
Zhe Zhang
.
Joseph C. Chen
.
Jacob Chen
Optimizing surface finish in a turning operation
using the Taguchi parameter design method
Received: 24 February 2004 / Accepted: 11 May 2005 / Published online: 1 December 2005
# Springer-Verlag London Limited 2005
Abstract This paper presents an application of the Taguchi
parameter design method to optimizing the surface finish in
a turning operation. The Taguchi parameter design method
is an efficient experimental method in which a response
variable can be optimized, given various control and noise
factors, and using fewer experimental runs than a factorial
design. The control parameters for this operation included:
spindle speed, feed rate, depth of cut, and tool nose radius.
Noise factors included varying roomtemperature, as well as
the use of more than one insert of the same specification,
which introduced tool dimension variability. A total of 36
experimental runs were conducted using an orthogonal
array, and the ideal combination of control factor levels was
determined for the optimal surface roughness and signal-to-
noise ratio. A confirmation run was used to verify the
results, which indicated that this method was both efficient
and effective in determining the best turning parameters for
the optimal surface roughness.
Keywords Taguchi parameter design
.
Turning
operations
.
Surface roughness
1 Introduction
1.1 Background
A typical turning operation produces parts which have
critical features requiring a specified surface roughness.
Applications include: bearing surfaces on axles, bearings,
and races; ultra-clean surfaces in contaminant-sensitive
components; and sealing surfaces on bores and pistons.
With this much reliance placed on turning operations, lathe
operators are expected to make use of their own experience
as well as published machining guidelines to achieve spec-
ified surface roughness values, with few or no rejects, in a
timely manner to keep up with production schedules. Un-
fortunately, due to inadequate knowledge of the complexity
and number of factors affecting surface roughness in turn-
ing operations, machine shops tend not to achieve these
goals efficiently, usually using what knowledge they have
to just get by. Proper selection of cutting tools, parameters,
and conditions for optimal surface quality (as well as tool
life) requires a more methodical approach by using exper-
imental methods and mathematical and statistical models.
Not only does this require considerable knowledge and
experience to design experiments and analyze data, but
traditional design-of-experiment (DOE) techniques require
a large number of samples to be produced [1]. Therefore,
a more efficient method is needed to effectively optimize
cutting parameters for surface roughness in turning
operations.
1.2 Introduction to the Taguchi method
One method presented in this article is an experimental
design process called the Taguchi approach or the Taguchi
method. Similar to DOE, the Taguchi method is a technique
for optimizing a process or design using multiple input
parameters. Cesarone [2] gives a basic outline of areas in
which these two methods differ, as illustrated in Table 1.
Due to the nature of these differences, Cesarone [2] recom-
mends determining the ideal method to use for a study on a
case-by-case basis, and to adapt components of each tech-
nique if necessary. A researcher should always fully under-
stand the various experimental methods in order to properly
apply them to individual studies to maximize both the ef-
ficiency and the results of a study.
The complete Taguchi methods are actually comprised of
three main phases, which are all intended to be conducted
offline. These three phases include system design, param-
eter design, and tolerance design. The Taguchi parameter
design stage, which is the phase used in this study, is com-
monly referred to as the Taguchi method, and will be so
E. D. Kirby
.
Z. Zhang
.
J. C. Chen (*)
.
J. Chen
Department of Industrial Education and Technology,
Iowa State University,
Ames, IA 50011-3130, USA
e-mail: cschen@iastate.edu
Tel.: +1-515-2948040
Fax: +1-515-2941123
referred to here. This phase requires that the factors are
known and that production should be in progress. The
major goal of this phase is to increase the performance of
the production process by adjusting the controlled factors
[3].
1.3 Application of the Taguchi method
to turning operations
The implication of fundamental knowledge of the turning
process for applying the Taguchi method to this problem
warrants a review of past studies involving machining pa-
rameters and conditions, and their effect on surface rough-
ness. Feng and Wang [4] found that many published studies
include spindle speed and feed rate, and a few included the
depth of cut. Decreased feed rate has been found to gen-
erally reduce surface roughness; however, the effects of the
spindle speed and depth of cut on surface roughness seemto
have different interpretations by different authors [46].
Tool nose radius has also been shown to have a direct effect
on surface roughness, in that a larger radius can reduce
surface roughness [7]. Variation in edge radius fromtool-to-
tool in a batch of inserts of the same specifications has been
found to be as high as 0.001 in, and has been shown to have
an effect on cutting forces [8]. Another possible variable is
the room temperature; it has been suggested that the tem-
perature of the room plays a part in the analysis of the
thermal dynamics of turning operations [9], although the
effect is rather unexplored. All of these factors, whether
controlled or uncontrolled, may be considered when op-
timizing a turning system for surface roughness.
Numerous studies have been conducted on the subject of
parameter optimization of turning operations, each focus-
ing on a specific methodology and parameters. There are
some excellent examples of published studies available
which have been conducted using the Taguchi method for
the purpose of optimizing turning parameters. Davim [5]
conducted such a study using a steel workpiece, with con-
trol parameters of spindle speed, depth of cut, and feed rate,
and the response parameter being the surface roughness.
Davim [10] also conducted a more elaborate study using a
composite workpiece; spindle speed, feed rate, and cutting
time control variables; and tool wear, spindle power, and
surface roughness response variables. Kopac et al. [6] re-
searched the optimization of parameters for turning cold-
drawn steel bars, using spindle speed, cutting tool material,
workpiece condition, depth of cut, and cutting sequence
(first or second cut); and surface roughness as a response
parameter. Vernon and zel [11] presented a Taguchi meth-
od analysis of the results of a recent DOE study using a steel
workpiece; spindle speed, workpiece length, and cutting
tool material control parameters; and a surface roughness
response parameter. Yang and Tarng [1] studied a similar
workpiece and the same control parameters as Davim [5],
but included both surface roughness and tool life as re-
sponse parameters. All of these studies successfully pro-
duced well-defined and useful correlations between their
control and response parameters. Clearly, numerous param-
eters can be included in these types of studies. A unique
combination of parameters can, therefore, be tailored to suit
a researchers facilities and distinct objectives.
Table 1 Comparison of the DOE method to the Taguchi method
Factor DOE method Taguchi method
Process
knowledge
Assumes no fundamental understanding
of the process being investigated
Requires knowledge of the process and the interactions
likely to exist between inputs
Number
of tests
Requires all combinations of inputs Requires a much smaller number
of combinations
Noise
factors
Traditionally ignores noise factors, but they
can be added to the plan (which cause the
number of tests to multiply)
Includes the noise factors in the design
Variability Ignores variability and assumes a deterministic
nature of the system to find the most effective
combinations of inputs to maximize or mini-
mize output
Assumes a stochastic system, looking at both levels and variability
of the output to allow the selection of input variable combinations
to optimize output or minimize variability
Confirming
experiment
Does not require a confirming experiment,
since all input combinations were tested
Should include a confirming experiment, because the selected input
combinations were probably not part of the original experimental plan
Table 2 The basic Taguchi L
9
(3
4
) orthogonal array
Run Control factors and levels
A B C D
1 1 1 1 1
2 1 2 2 2
3 1 3 3 3
4 2 1 2 3
5 2 2 3 1
6 2 3 1 2
7 3 1 3 2
8 3 2 1 3
9 3 3 2 1
1022
1.4 Purpose of the study
The purpose of this study is to efficiently determine the
optimal turning operation parameters for the surface rough-
ness of aluminum bar stock under varying conditions using
Taguchis parameter design method. The questions which
this study will address include:
1. What are the relationships between the control
parameters and the response parameter?
2. What are the effects of two noise parameters on the
response parameter?
3. What are the optimal turning operation parameters for
surface roughness, given two possible noise factors?
2 Experimental design
This experiment will make use of the Taguchi method in
order to address the issues and questions being studied. The
standardized Taguchi-based experimental design used in
this study was an L
9
(3
4
) orthogonal array, as described in
Fowlkes and Creveling [12], and as shown in Table 2. This
basic design uses up to four control factors, each with three
levels. A total of nine runs must be carried out, using the
combination of levels for each control factor (Athrough D),
as indicated in Table 2. The addition of noise factors is
optional, and requires each run to be conducted once for
each combination of noise factor.
The selected parameters, discussed in the introduction,
are listed in Table 3, along with their applicable codes and
values for use in the Taguchi model. The control and noise
factors are independent variables, and the response variable
is the dependent variable.
The control factors are the basic, controlled parameters
used in a turning operation. The feed rate and depth of cut
were selected from within the range of parameters for finish
turning [13]. The spindle speeds, while slightly lower than
those normally used for turning aluminum workpieces [13],
were chosen for safety reasons, due to the nature of the
machine being used. The tool radii were chosen from the
available aluminum cutting inserts from VNE Corporation
(Janesville, WI) [14].
The noise factors listed in Table 3 are often uncontrolled
variables in machine shops which may affect the surface
roughness of a turning operation. The selected factors were
considered because they could be controlled in an exper-
imental setup. The temperature ranges include both a nor-
mal (6575F) and a high (95100F) shop temperature
range. The normal range is based on a common room tem-
perature range, and includes the temperature settings of
most heating and air conditioning systems. The high tem-
perature range is what a machine shop without air con-
ditioning in the El Paso, Texas/Ciudad Juarez, Mexico area
would usually experience as a high temperature during the
summer [15]. The second noise factor is the use of two
inserts with identical specifications to introduce edge radius
variations into the cutting process.
A modified orthogonal array, Table 4, was created using
the basic Taguchi orthogonal array and the selected param-
eters from Table 3. In this array, the basic array with the
control factors are shown as the inner control factor array,
and the added noise factors are shown as the outer noise
array. Also indicated here is the required repetition for each
run for each combination of noise factors (N1, N2, N3, and
N4). This brings the total number of runs to 36 for this
experiment.
Table 3 Parameters, codes, and level values used for the orthogonal
array
Parameter Code Level 1 Level 2 Level 3
Control factors
Spindle speed (rpm) A 1,500 2,250 3,000
Feed rate (ipr) B 0.004 0.008 0.012
Depth of cut (in) C 0.010 0.020 0.030
Tool radius (in) D 0.008 0.016 0.032
Noise factors
Tool number X 1 2
Temperature range (F) Y 6575 95100
Response variable
Surface roughness (in R
a
)
Table 4 Modified Taguchi L
9
(3
4
) orthogonal array
Outer noise array
X 1 1 2 2
Y 1 2 1 2
Inner control factor array
Run A B C D N1 N2 N3 N4
1 1 1 1 1
2 1 2 2 2
3 1 3 3 3
4 2 1 2 3
5 2 2 3 1
6 2 3 1 2
7 3 1 3 2
8 3 2 1 3
9 3 3 2 1
1023
This study was conducted using samples cut from a
single length of 1-in diameter 6061-T6 aluminum alloy rod.
Standardized material per ASTM B221 was selected to
ensure consistency of the 6061 alloy, which is a common
wrought alloy used in industry [16]. These samples were
turned and measured for surface roughness, and the results
analyzed using the setup and procedures described in the
next section.
3 Experimental setup and procedure
This experiment was conducted using the hardware listed in
Table 5. The listed items were chosen based on both ne-
cessity for this study, as well as availability. The workpiece
samples were randomly selected, 2.25-in lengths cut from
the aluminum rod described in the previous section. The
CNC lathe was set up with the tool holder, with the ther-
mometer probe mounted near the workpiece location. The
heater was placed so that air would circulate in the enclosed
workspace, creating a hot breeze in the workspace. All var-
iable parameters were run in accordance with the selected
parameters (Table 3) and the orthogonal array (Table 4).
Table 6 shows the setup sheetthe individual sample num-
bers and their parameters, listed in order, according to the
orthogonal array run and noise numbers. The spindle speed,
feed rate, and depth of cut parameters were programmed
into the CNCprogram, which created a straight cut of 0.5 in
in the z axis. The tool inserts were changed as needed to
obtain the prescribed radius, and to alternate between the
two tools in each set. A set of two tool inserts was used for
each radius, and marked 1 or 2 to make identification
easier in accordance with the setup sheet. The room tem-
perature of 6575F was maintained with the room heater
in the CNC laboratory, while the higher temperature of
95100F was maintained (when needed) with the space
heater. These temperatures were monitored during the cut-
ting process with the thermometer, to ensure that the tem-
perature did not go out of range during that time.
The surface roughness measurement process included
the use of a profilometer, as well as all the necessary hard-
ware to align the stylus motion perpendicular to the axis of
the turned part during the measurement. Blocks were also
used to align the profilometer vertically with the top surface
of the workpiece, in accordance with the manufacturers
specifications. The maximum surface roughness measure-
ment (maximum R
a
) of each cut was recorded for analysis.
The maximum R
a
was chosen because it was possible that
control and noise factor combinations would produce var-
iance in the surface quality across the cut, and using the
maximumR
a
would allowany sporadic occurrences of high
surface roughness to affect the study. This was done with
the goal of selecting parameters that would stand out more
as producing an optimal surface finish across the entire cut.
This study required the use of two types of software: a
spreadsheet package for the tabulation and simple calcula-
tions of the results, and a statistical analysis package for
analyzing the results. The software titles selected for these
tasks were Microsoft Excel and SAS Institute Inc.s JMP
package.
Table 5 Hardware list
Item Specifications
CNC lathe Clausing/Colchester Storm A50 slant bed lathe
Spindle speed range: 1,000 rpm minimum for full motor power,
4,000 rpm maximum
Feed rate range: 0.0000014.000000 in per rev (ipr)
Least input movement increment: X 0.00005 in, Z 0.0001 in
Surface roughness measurement
device
Federal Pocketsurf stylus profilometer
Measures R
a
in in; stylus travel 0.1 in
Space heater Honeywell quick heat ceramic heater
Small forcedair space heater with thermostat
With thermal protection devices for safety
Thermometer Taylor digital thermometer #1420
Digital thermometer with probe, indoor/outdoor readings
Range 50120F
Cutting tool inserts (two of each
radius)
VNE Versa Turn CCGT 430.5-AF (0.008 inch nose radius)
VNE Versa Turn CCGT 431-AF (0.016 inch nose radius)
VNE Versa Turn CCGT 432-AF (0.032 inch nose radius)
Tool holder VNE Versa-Turn SCLCR 124B (for above inserts)
Surface table Polished granite surface
V-block Sufficient size to hold 1-in-diameter1.25 in-long bar stock
V-block clamp Clamps workpiece to v-block
Clamps (two) To hold v-block and profilometer alignment block
Misc. blocks and plates Various sizes of flat machined blocks and plates
1024
4 Results and analysis
4.1 Completion and analysis of the orthogonal array
The results of the surface roughness measurements (in R
a
)
of each sample are shown in Table 7, along with the ad-
ditional parameters of the expanded orthogonal array. The
individual surface roughness measurements are noted as N1
through N4 for each run in the array. A final column has
been added to this array, to indicate the signal-to-noise (S/
N) ratio, calculated as follows:
10 log
1
n
X
y
2
i


(1)
where is the S/N ratio, y
i
are the individual surface
roughness measurements in columns N1 through N4, and n
is the number of noise factors; in this case, n=4.
Also added to this array are the standard deviation (s),
variance (s
2
), and the mean (R
a
) of the surface roughness
measurements, which are used to verify the performance of
the calculated S/Nratio. This type of experiment, in which a
smaller response variable is desirable, should produce S/N
ratios that increase as the variance and means decrease. As
revealed in Table 7, this is the case here, which indicates
that no anomalies were introduced in the measurements or
calculations.
The high variance for each run also seems to indicate that
the noise factors affected the resulting surface roughness
values of the samples. The effect of changing between in-
serts of the same specifications (noise factor X) appears to
Table 6 Experiment setup sheet
Run Spindle speed Feed rate Depth of cut Tool radius Noise Tool # Temp F Sample ID#
1 1,500 0.004 0.010 0.008 N1 1 6575 1
N2 1 95100 2
N3 2 6575 3
N4 2 95100 4
2 1,500 0.008 0.020 0.016 N1 1 6575 5
N2 1 95100 6
N3 2 6575 7
N4 2 95100 8
3 1,500 0.012 0.030 0.032 N1 1 6575 9
N2 1 95100 10
N3 2 6575 11
N4 2 95100 12
4 2,250 0.004 0.020 0.032 N1 1 6575 13
N2 1 95100 14
N3 2 6575 15
N4 2 95100 16
5 2,250 0.008 0.030 0.008 N1 1 6575 17
N2 1 95100 18
N3 2 6575 19
N4 2 95100 20
6 2,250 0.012 0.010 0.016 N1 1 6575 21
N2 1 95100 22
N3 2 6575 23
N4 2 95100 24
7 3,000 0.004 0.030 0.016 N1 1 6575 25
N2 1 95100 26
N3 2 65759 27
N4 2 5100 28
8 3,000 0.008 0.010 0.032 N1 1 6575 29
N2 1 95100 30
N3 2 6575 31
N4 2 95100 32
9 3,000 0.012 0.020 0.008 N1 1 6575 33
N2 1 95100 34
N3 2 6575 35
N4 2 95100 36
1025
be fairly consistent in either increasing or decreasing the
surface roughness values for every insert of the same nose
radius (control factor D). For example, every level 1 of
control factor D experiences larger surface roughness val-
ues with level 2 of noise factor X. Levels 2 and 3 of control
factor D generally experienced larger surface roughness
values with level 1 of noise factor X. This would seem to
indicate a possible variation in nose radius between the 2
tools of each set. The application of a hot room temperature
(noise factor Y), however, did not have a consistent no-
ticeable effect on the surface roughness values for any of
the control factors. These effects were only based on visual
interpretation of the orthogonal array; however, a simple
statistical test would provide a more valid analysis of this
variance. Therefore, a t-test for each noise factor was per-
formed in order to determine the validity these perceived
effects. The results of these tests, shown in Tables 8 and 9,
resulted in a very high p-value, indicating that a statistically
significant difference could not be detected between the two
levels of each noise factor. This would seemto indicate that,
while the noise factors could possibly have an effect on the
surface roughness, this effect may be negligible and can
only be completely determined with more experimental re-
search. This is beyond the scope of the type of study con-
ducted here, in which noise factors are included for the sole
purpose of determining a combination of control factors
that results in a response variable which is the most immune
to noise factors [3].
The R
a
(mean response variable) effect table under the
array in Table 7 indicates the mean of the response variable
means for each level of each control factor. This specifies
the mean surface roughness value that each level of each
control factor produced during this experiment.
The S/N effect table under the array in Table 7 indicates
the mean of the S/N values for each level of each control
factor. Since we are looking for the mean and variance of
the surface roughness values to be as small as possible, the
ideal S/N effects should be as large as possible. This can be
shown graphically as well; Fig. 1 shows plots of the re-
sponse and S/N ratio effects from Table 7. These graphs
reveal the level to be chosen for the ideal cutting parameters
(the level with the highest point on the graph), as well as the
relative effect each parameter has on the S/N ratio (the
general slope of the line).
As seen in the response and S/N ratio effects graphs
(Fig. 1), all four parameters had at least some effect on the
Table 7 Completed orthogonal array
Outer noise array
X 1 1 2 2
Y 1 2 1 2
Inner control factor array
Run A B C D N1 N2 N3 N4 R
a
s s
2

1 1 1 1 1 47 47 61 64 54.75 9.03 81.58 34.86


2 1 2 2 2 137 137 124 130 132.00 6.27 39.33 42.42
3 1 3 3 3 145 147 117 116 131.25 17.06 290.92 42.42
4 2 1 2 3 32 31 26 36 31.25 4.11 16.92 29.95
5 2 2 3 1 159 162 178 171 167.50 8.66 75.00 44.49
6 2 3 1 2 230 214 207 210 215.25 10.24 104.92 46.67
7 3 1 3 2 44 49 50 40 45.75 4.65 21.58 33.24
8 3 2 1 3 64 69 52 46 57.75 10.59 112.25 35.34
9 3 3 2 1 244 240 248 247 244.75 3.59 12.92 47.78
Level A B C D
R
a
effects
level 1 106.00 43.92 109.25 155.67
level 2 138.00 119.08 136.00 131.00
level 3 116.08 197.08 114.83 73.42
effects
level 1 39.90 32.68 38.95 42.37
level 2 40.37 40.75 40.05 40.78
level 3 38.79 45.62 40.05 35.90
Table 8 t-test for effect
on response of alternating
between two tools of the same
specifications
t df p-value Mean difference Std. error difference 95% C.I.
Lower Upper
Equal variance 0.020 34 0.984 0.5 25.373 51.064 52.064
Unequal variance 0.020 33.98 0.984 0.5 25.373 51.066 52.066
1026
surface roughness and S/N ratios in this study. The feed rate
and tool nose radius are shown to have a strong effect on
surface roughness and their S/N ratios, which was expected
from the literature review. The spindle speed and depth of
cut had a smaller effect, as evidenced by the shallow slope
of the lines. The effect of spindle speed appears to be non-
linear, with the middle level being the least desirable for
surface roughness.
4.2 Determining the ideal parameters
The S/N ratio and R
a
effects data plotted in these graphs
can be used to determine the optimal set of parameters from
this experimental design. The arrows in the graphs indicate
the levels at which the S/N ratio and R
a
effects are at their
optimal magnitudes; that is, the S/N ratio effect is at its
highest magnitude, and the R
a
effect is at its lowest mag-
nitude. A conflict appears in the graph of spindle speed
effects; the R
a
effect is optimized at level 1, while the S/N
ratio effect is optimized at level 3. This conflict could be
considered relatively insignificant, since the difference in
each effect across the three levels for spindle speed is
relatively small in comparison with the feed rate and tool
nose radius effects. However, since a confirmation run was
required for this experiment, a single level must be chosen.
Since the spindle speed affects the linear speed of the feed
rate (which is in in per rev [ipr]), a higher spindle speed was
chosen to maximize productivity. Additionally, selecting
the level based on the S/N ratio is ideal, since the S/N ratio
effect can be considered as resulting in the best response
given the noise in the system [12]. The optimized levels of
each of the other three parameters were included in the
confirmation run. As seen in Table 10, the selected spindle
speed was at level 3 (3,000 rpm), the optimal feed rate was
at level 1 (0.004 ipr), the optimal depth of cut was at level 1
(0.010 in), and the optimal tool nose radius was at level 3
(0.032 in). Therefore, the optimized combination of levels
for the four control factors from the analysis so far was A3-
B1-B1-B3. The final confirmation run would be used to
verify that this combination was ideal.
A Spindle Speed
S/N
R
a
-50.00
-45.00
-40.00
-35.00
-30.00
Level
S
/
N
40.00
80.00
120.00
160.00
200.00
R
a
1 2 3
B Feed Rate
S/N
R
a
-50.00
-45.00
-40.00
-35.00
-30.00
Level
S
/
N
40.00
80.00
120.00
160.00
200.00
R
a
1 2 3
C Depth of Cut
S/N
R
a
-50.00
-45.00
-40.00
-35.00
-30.00
Level
S
/
N
40.00
80.00
120.00
160.00
200.00
R
a
1 2 3
D Tool Nose Radius
S/N
R
a
-50.00
-45.00
-40.00
-35.00
-30.00
Level
S
/
N
40.00
80.00
120.00
160.00
200.00
R
a
1 2 3
Fig. 1 Response and S/N ratio effects
Table 9 t-test for effect on the
response of normal (6575F)
and high (95100F) room
temperature
t df p-value Mean difference Std. error difference 95% C.I.
Lower Upper
Equal variance 0.164 34 0.870 4.167 25.363 47.377 55.711
Unequal variance 0.164 34 0.870 4.167 25.363 47.377 55.711
1027
4.3 Confirmation run
While the relative values for each of these parameters were
consistent with the literature review, one can see in Table 6
that the combination of selected control parameter values
was not actually carried out in the experiment. Since the
Taguchi method selects only a small number of combina-
tions of control variables (rather than all combinations as
per the DOE method), this situation was expected. Due to
this situation, a confirmation run was necessary in order to
validate the results.
The objective of the confirmation run was to determine
that the selected control parameter values would produce
better surface finishes than those produced in the first part
of the experiment. To create this comparison, the research-
ers compared the surface roughness mean of products
produced using the selected control parameter values to the
surface roughness mean of products produced in the first
part of the experiment, which are found in Table 7. A sam-
ple of 15 workpieces of the same material and dimensions
described earlier was turned using the selected control
parameter values. The surface roughness was then mea-
sured using the setup described earlier. The response var-
iable used in the confirmation run was the mean R
a
, in in,
of measurements taken across the length of the cut, as rec-
ommended by BabusHaq et al. [17].
Table 11 indicates the results of the confirmation run,
including the mean Ra of each workpiece cut, as well as the
overall mean Ra
(R
a
), overall standard deviation, and 99%
confidence intervals for the mean R
a
. Based on these results,
it can be concluded with 99% confidence that, by turning
samples using the setup described in this study, and the
parameters indicated in Table 10, the resulting average sur-
face roughness (R
a
) will range from 19.93 in to 23.94 in.
This result can be compared with the measurements found
in Table 7, of which, the lowest measurement was 26 R
a
in. This comparison confirms that the selected control pa-
rameter values using the Taguchi parameter design process
produced the lowest surface roughness for the control pa-
rameter values used in this experiment.
5 Conclusions and discussions
This study presented an efficient method for determining
the optimal turning operation parameters for surface finish
under varying conditions through the use of the Taguchi
parameter design process. This process was applied using a
specific set of control and noise parameters, and a response
variable of surface roughness. The use of the L
9
(3
4
) or-
thogonal array, with four control parameters and two noise
factors, allowed this study to be conducted with a sample of
36 workpieces.
The study found that the control factors had varying
effects on the response variable, with feed rate and tool nose
radius having the highest effects. The noise factors, on the
other hand, were found to not have a statistically noticeable
effect. The study led to the selection of a combination of
levels for each control parameter, which were used to create
an additional sample of 15 workpieces. The measurement
of the workpieces in this confirmation run led to the con-
clusion that the selected parameter values from this process
produced a surface roughness that was much lower than the
other combinations tested in this study. The use of the
Taguchi parameter design technique was considered suc-
cessful as an efficient method to optimize surface roughness
in a turning operation.
Table 11 Results of the confirmation run
Sample # R
a
1 22.5
2 23.25
3 16.5
4 20.75
5 20.75
6 21
7 23.25
8 26.5
9 23.25
10 18
11 22.25
12 20.5
13 22
14 25
15 23.5
R
a
21.93
s 2.52
99% Confidence Interval (R
a
)
19.93 23.94
Table 10 Selected control
parameter values
Parameter A: spindle speed B: feed rate C: depth of cut D: tool nose radius
S/N ratio effect 38.79 32.68 38.95 42.37
Level 3 1 1 3
Value 3,000 rpm 0.004 ipr 0.010 in 0.032 in
1028
References
1. Yang WH, Tarng YS (1998) Design optimization of cutting
parameters for turning operations based on the Taguchi method.
J Mater Process Technol 84:122129
2. Cesarone J (2001) The power of Taguchi. IIE Solutions 33
(11):3640
3. Roy RK(2001) Design of experiments using the Taguchi
approach: 16 steps to product and process improvement. Wiley,
New York
4. Feng C-X, Wang X-F (2003) Surface roughness predictive
modeling: neural networks versus regression. IIE Trans 35:1127
5. Davim JP (2001) A note on the determination of optimal
cutting conditions on the surface finish obtained in turning
using design experiments. J Mater Process Technol 116
(2/3):305308
6. Kopac J, Bahor M, SokoviC M (2002) Optimal machining
parameters for achieving the desired surface roughness in fine
turning of cold pre-formed steel workpieces. Int J Mach Tools
Manuf 42(6):707716
7. Thomas M, Beauchamp Y, Youssef AY, Masounave J (1995)
Effect of tool vibrations on surface roughness during lathe dry
turning process. Comput Ind Eng 31(3/4):637644
8. Schimmel RJ, Manjunathaiah J, Endres WJ (2000) Edge radius
variability and force measurement considerations. ASME J
Manuf Sci Eng 122:590593
9. Huang Y, Liang SY (2003) Cutting forces modeling consider-
ing the effect of tool thermal propertyapplication to CBN
hard turning. Int J Mach Tools Manuf 43:307315
10. Davim JP (2003) Design of optimisation of cutting parameters
for turning metal matrix composites based on the orthogonal
arrays. J Mater Process Technol 132:340344
11. Vernon A, zel T (2003) Factors affecting surface roughness in
finish hard turning. In: Proceedings of the 17th International
Conference on Production Research (ICPR-17), Blacksburg,
Virginia, August 2003
12. Fowlkes WY, Creveling CM (1995) Engineering methods for
robust product design: using Taguchi methods in technology and
product development. Addison-Wesley, Reading, Massachusetts
13. Harig H (1978) Basic precision machining for metalworking
trainees. National Tooling and Machining Association, Wash-
ington, DC
14. VNE Corporation (1999) Versa-Turn Catalog. Janesville,
Wisconsin
15. National Weather Service (2003) Temperature, degree days, and
precipitation normals in El Paso, Texas for June. National
Weather Service Forecast Office, El Paso Area, Santa Teresa,
New Mexico Available online at http://www.srh.noaa.gov/elp/
climat/elpclimate/elpjunnorm.shtml
16. Avallone EA, Baumeister T III (eds) (1996) Marks standard
handbook for mechanical engineers (10th edn). McGraw-Hill,
New York
17. BabusHaq RF, Probert D, Snaith B, OCallaghan PW, George
HE (1990) Perceived and real roughness variations across
machined surfaces. Int J Mater Prod Technol 5(1):1224
1029

Das könnte Ihnen auch gefallen