Professional Documents
Culture Documents
RSM Basic PDF
RSM Basic PDF
RSM Basic PDF
Mathematical models of tool life and surface roughness for turning operation
through response surface methodology
Hari Singh1,* and Pradeep Kumar2
1
Mechanical Engineering Department, National Institute of Technology, Kurukshetra 136 119
2
Department of Mechanical and Industrial Engineering, Indian Institute of Technology, Roorkee 247 667
Mathematical models of tool life and surface roughness have been developed for turning En24 steel with titanium
carbide coated tungsten carbide inserts. Response surface methodology (RSM) has been applied for developing the models
in the form of multiple regression equations correlating dependent parameters, tool life and surface roughness, with cutting
speed, feed rate and depth of cut, in a turning process. The second order response surface was found suitable for present
work. The central composite rotatable design was used to plan the experiment.
Keywords: Carbide inserts, En24 steel, Experimental design, RSM, Surface roughness, Tool life
IPC Code: G09B25/02
Response Surface Methodology (RSM) Table 1—Components of central composite second order
In RSM15, it is possible to represent independent rotatable design
factors in quantitative form. Then these factors can be Number of Factorial Star Center Total Value
thought of as having a functional relationship or variables points points points N α
response: K 2k 2k n
Table 2—Values of selected variables at different levels to machine En24 steel (0.4%C) of 220 BHN. ISO
Variable Levels designation of insert is SPUN 120308 and tool holder
symbols -1.682 -1 0 +1 +1.682 is CSBPR 2525 H 12. Observations (20) of TL, by
measuring flank wear (width of wear land), were
A, m/min 190 209.82 242.69 280.71 310 recorded using Large Tool Maker’s microscope. For
B, mm/rev 0.12 0.13 0.16 0.18 0.20 TL assessment, flank wear width was measured at an
C, mm 0.70 0.75 0.84 0.93 1.00 interval of 1 min and the flank wear criterion of
0.45 mm against 1 TL was applied. SR was measured
Table 3—Experimental design values and average values of for each run using a Philips roughness tester having a
responses least count of 1 ru (0.025 µ). The levels at which the
Sr. Randomized A B C TL SR parameters were tested are given in Table 2.
No. Experimental design values of process variables along
with average values of TL and SR are given in
1 19 210 0.13 0.75 25 80 Table 3.
2 18 280 0.13 0.75 17 70
3 17 210 0.18 0.75 23 110 Results and Analysis
4 16 280 0.18 0.75 15 100 Design Expert 7.0 software (dx7) was used for
5 15 210 0.13 0.93 24 80 regression and graphical analysis25. For analysis of the
6 14 280 0.13 0.93 16 80
data, checking of goodness of fit of model is very
much required. Model adequacy checking includes
7 13 210 0.18 0.93 22 120
test for significance of regression model and on model
8 12 280 0.18 0.93 15 110
coefficients, and test for lack of fit26. Analysis of
9 1 190 0.16 0.84 27 100 variance (ANOVA) is performed for this purpose.
10 2 310 0.16 0.84 13 90
11 3 240 0.12 0.84 21 70 Analysis of Tool Life
12 4 240 0.20 0.84 18 120 The fit summary recommended that linear model is
13 20 240 0.16 0.70 20 90 statistically significant for analysis of TL. Model F-
14 11 240 0.16 1.00 19 100
value of 274.56 implies the model is significant
(Table 4). There is only a 0.01% chance that a "Model
15 5 240 0.16 0.84 19 100
F-Value" this large could occur due to noise. Values
16 6 240 0.16 0.84 20 110
of "Prob > F" (< 0.0500) indicate model terms are
17 7 240 0.16 0.84 19 100 significant. In this case A, B, C are significant model
18 8 240 0.16 0.84 19 90 terms. Values more than 0.1000 indicate the model
19 9 240 0.16 0.84 21 90 terms are not significant. "Lack of Fit F-value" of
20 10 240 0.16 0.84 20 100 0.15 implies the Lack of Fit is not significant relative
to the pure error. There is a 99.51% chance that a
TL - Tool life, min; SR – Surface roughness, ru "Lack of Fit F-value" this large could occur due to
Standard deviation = 0.53, Mean = 19.65, Coefficient of variation % = 2.69, Predicted residual error of sum of
squares (PRESS) = 5.79, R2 = 0.9809, R2 adjusted = 0.9774, Predicted R2 = 0.9753, Adequate precision = 56.837
224 J SCI IND RES VOL 66 MARCH 2007
noise. Non-significant lack of fit is good. Predicted R2 Model F-value of 46.93 implies the model is
of 0.9753 is in reasonable agreement with the adjusted significant (Table 5). There is only a 0.01% chance
R2 of 0.9774. Adequate precision measures signal to that a "Model F-Value" this large could occur due to
noise ratio. A ratio more than 4 is desirable; ratio of noise. Values of "Prob > F" (< 0.0500) indicate model
56.837 indicates an adequate signal. This model can terms are significant. In this case, A, B, C are
thus be used to navigate design space. Normal significant model terms. The "Lack of Fit F-value" of
probability plot of residuals for TL (Fig. 3) shows that 0.22 implies the Lack of Fit is not significant relative
residuals are falling on a straight line, which means to the pure error. There is a 98.39% chance that a
the errors are normally distributed. Final response "Lack of Fit F-value" this large could occur due to
equation in terms of coded factors is given as: noise. Non-significant lack of fit is good. The
TL = +19.65 -3.99* A -0.88* B -0.34 * C … (4) predicted R2 of 0.8709 is in reasonable agreement
with adjusted R2 of 0.8788. Adequate precision
The final response equation in terms of actual measures the signal to noise ratio. A ratio more than 4
factors is given as: is desirable; ratio of 23.072 indicates an adequate
TL = +55.95470 -0.11268* Cutting speed -35.28010* signal. This model can thus be used to navigate design
Feed rate -3.80907 * Depth of cut … (5) space. Normal probability plot of the residuals for SR
The developed response surface, within the (Fig. 4) indicates that residuals are falling on a
experimental region, is as: straight line, indicating that the errors are normally
distributed. The final response equation in terms of
TL = 26914 v -1.424 f -0.299 d -0.160 ± 0.1687×10-2 … (6) coded factors is given as:
where TL = tool life, min; v = cutting speed, m/min; SR = +95.50 -3.43 * A +15.68 * B +3.43 * C … (7)
f = feed, mm/rev; d = depth of cut, mm.
The final response equation in terms of actual
Analysis for Surface Roughness (SR)
factors is given as:
The fit summary recommended that the linear SR= -9.96807 -0.096718 * Cutting speed + 627.05404
model is statistically significant for analysis of SR. * Feed rate+38.09070 * Depth of cut … (8)
Fig. 3—Normal probability plot residuals for tool life Fig. 4—Normal probability plot residuals for surface roughness
SINGH & KUMAR: MATHEMATICAL MODELS OF TOOL LIFE AND SURFACE ROUGHNESS 225
Standard deviation = 5.11, Mean = 95.50, Coefficient of variation % = 5.35, Predicted residual error of sum of
squares (PRESS) = 528.74, R2 = 0.8980, R2 adjusted = 0.8788, Predicted R2 = 0.8709, Adequate precision =
23.072
The developed response surface within the experi- machining of AISI 4340 steels, Trans ASME, J Eng Ind, 104
mental region is as: (1982) 121-131.
4 Nakayama K & Shaw M C, Machining High Strength
SR = 3112 v-0.248 f1.112 d0.34 ± 0.5815×10-2 … (9) Materials, Annals of CIRP, 15 (1967) 456-459.
5 Sarmah B P, Some investigations into performance of CVD-
where SR = surface roughness, ru; v = cutting speed, coated indexable carbide inserts in machining of difficult to
m/min; f = feed, mm/rev; and d = depth of cut, mm. machine materials, Ph D Thesis, Mech. and Ind Engg Dept,
The predicted values of TL (24.8688 min) and SR University of Roorkee, Roorkee, India, 1988.
(79.8236 ru) in this case study are similar to reported 6 Lo K C & Chen N N S, Prediction of tool life in hot
machining of alloy steels, Int J Prod Res, 15 (1977) 47-63.
values27 of TL (20.19 min) and SR (81.38 ru) at 95 %
7 Kalnth G S, Sachdeva S C & Choudhary A P, An
confidence level. The results thus compare well with Experimental investigation into hot machining of En 24 steel,
the present case study through RSM. Proc AIMTDR Conf (IIT Bombay, Mumbai) 1978, 285-289.
8 Dearnley P A & Trent E M, Wear Mechanism of Coated
Limitations of the Case Study Carbide Tools, Metal Technology, 9 (1982) 60-74.
The results of this case study are valid for En 24 9 Dearnley P A, Rake and flank wear mechanism of coated and
steel and selected process variables and their specified uncoated carbide tools, ASME Trans, J Engg Mat & Technol,
range. Any extrapolation must be confirmed through 107 (1985) 68-82.
further experimentation. 10 Bose et al, Improved performances of coated carbide tools
with positive rake and in-built chip breaker, Power
Metallurgy Int, 17 (1985) 78-82.
Conclusions 11 Lambert B K & Taraman K, Development and utilization of a
Models developed in this study can be used to mathematical model of a turning operation, Int J Prod Res, 11
predict tool life of titanium carbide coated tungsten (1973) 69-81.
carbide inserts and surface roughness obtainable for 12 Taraman K, Multi-machining output multi-independent
turning En 24 alloy steel provided process variables variable turning research by response surface methodology,
Int J Prod Res, 12 (1974) 232-245.
are within the specified range. The predicted values of
13 Hassan G A & Suliman S M A, Experimental modelling and
the tool life and surface roughness are 24.8688 min optimization of turning medium carbon steel, Int J Prod Res,
and 79.8236 ru respectively. The effect of process 28 (1990) 1057-1065.
variables on tool life and surface roughness can be 14 El Baradie M A, Surface roughness model for turning grey
further ascertained by plotting the developed models. cast iron (154 BHN), Proc Instn Mech Engrs, 207 (1993) 43-
50.
References 15 Cochran G & Cox G M, Experimental Design (Asia
Publishing House, New Delhi) 1962, 27-32.
1 Mottram R A & Woolman J, The Mechanical and Physical
Properties of the British Standard EN Steels, 1st edn, vol 2, 16 Box G E P & Hunter J S, Multifactor experimental design, J
EN 21-EN 39 (Pergamon Press, Oxford) 1966, 72-73. Ann Math Stat, 28 (1957) 195-241.
2 Robert R B, Metallic Materials, Specification Hand Book 17 Box G E P, Hunter W G & Hunter J S, Statistics for
(McGraw-Hill, New York) 1968, 60-64. Experimenters (Wiley, New York) 1978, 45-55.
3 Komanduri R, Schroeder T, Hazra J, von Turkovich B F & 18 Draper N R, Center points in second order response surface
Flom D G, Catastrophic shear instability in high speed designs, Technometrics, 18 (1982) 127-133.
226 J SCI IND RES VOL 66 MARCH 2007
19 Singh H & Kumar P, Multi-criteria optimization of machining 23 Singh H & Kumar P, Effect on power consumption for turned
parameters using the concept of fuzzy set logics, Proc 16th parts using Taguchi technique, Productivity, 45 (2004) 231-
National Conv Mech Eners (University of Roorkee, Roorkee) 238.
2000, 516-521. 24 Singh H, Optimization of machining parameters for turned
20 Singh H, Kumar P & Singh I P, Effect of process parameters parts through Taguchi’s technique, Ph D Thesis, Kurukshetra
on surface roughness of turned parts using response surface University, Kurukshetra, India, 2000.
methodology, Proc Int Conf Math Modelling (University of 25 Design expert software, Version 7, Users Guide, Technical
Roorkee, Roorkee) 2001, 483-489. manual, Stat-Ease Inc. Minneapolis, MN, 2004.
21 Singh H & Kumar P, Quality optimization of turned parts 26 Montgomery D C, Peck E A & Vining G G, Introduction to
(EN24 Steel) by Taguchi Method, Productivity, 44 (2003) 43- Linear Regression Analysis, 3rd edn (Wiley, New York) 2003,
49. 70-89.
22 Singh H & Kumar P, Tool wear optimization in turning 27 Singh H & Kumar P, Optimizing multi-machining charac-
operation by Taguchi method, Indian J Eng Mater Sci, 11 teristics through Taguchi’s approach and utility concept, J
(2004) 19-24. Manuf Technol Manage, 17 (2006) 255-274.