Professional Documents
Culture Documents
Nataraj An 2011
Nataraj An 2011
DOI 10.1007/s00170-011-3343-1
ORIGINAL ARTICLE
Received: 5 October 2010 / Accepted: 11 April 2011 / Published online: 26 April 2011
# Springer-Verlag London Limited 2011
1 Introduction
Nowadays, due to the increasing demand of higher
precision components for its functional aspect, surface
roughness of a machined part plays an important role in
the modern manufacturing process. Turning is a machining
operation, which is carried out on lathe. The quality of the
surface plays a very important role in the performance of
turning as a good quality turned surface significantly
improves fatigue strength, corrosion resistance, or creep
life. Surface roughness also affects several functional
attributes of parts, such as, contact causing surface friction,
wearing, light reflection, heat transmission, ability of
distributing and holding a lubricant, load bearing capacity,
coating, or resisting fatigue. Therefore, the desired finish
surface is usually specified and the appropriate processes
are selected to reach the required quality [1]. To achieve the
desired surface finish, a good predictive model is required
for stable machining. Generally, these models have a
complex relationship between surface roughness and
operational parameters, work materials, and chip breaker
types. Artificial neural networks (ANNs) are information
processing systems, and since their inception, they have
been used in several areas of engineering applications. In
experimental studies, some of the operating conditions of a
system can be investigated. For this type of experimental
work, experts and special equipment are needed. It also
requires too much time and high cost.
ANNs have been trained to solve non-linear and complex
problems that are not exactly modeled mathematically. ANNs
1044
2 Literature survey
Since turning is the primary operation in most of the
production processes in the industry, surface finish of
turned components has greater influence on the quality of
the product. Surface finish in turning has been found to be
influenced in varying amounts by a number of factors such
as feed rate, work material characteristics, work hardness,
unstable built-up edge, cutting speed, depth of cut, cutting
time, tool nose radius. According to these parameters, a
detailed literature survey is carried out as follows. Adeel H.
Suhail et al. [2] focused on an experimental study to
optimize the cutting parameters using workpiece surface
temperature and surface roughness by employing Taguchi
techniques. Al-Ahmari [3] developed empirical models for
tool life, surface roughness, and cutting force for turning
operations. Two important data mining techniques used
were response surface methodology and neural networks.
Avisekh Banerjee et al. [4] conducted a study of feasibility
of on-line monitoring of surface roughness in turning
operations using a developed opto-electrical transducer.
Regression and neural network (NN) models were
exploited to predict surface roughness and compared to
actual and on-line measurements. Azouzi et al. [5] proposed
an on-line prediction of surface finish and dimensional
deviation in turning using neural network-based sensor
fusion. Bajic et al. [6] focused on modeling of machined
surface roughness and optimization of cutting parameters in
face milling and examined the influence of cutting
parameters on surface roughness in face milling. David et
al. [7] described an approach to predict surface roughness
in a high-speed end-milling process and used artificial
neural networks and statistical tools to develop different
surface roughness predictors. Franic Cus [8] used binary
coded genetic algorithm (BCGA) for the optimization of
cutting parameters. This genetic algorithm optimizes the
cutting conditions having an influence on production cost,
time, and quality of the final product. Groover [9] depicted
the impact of three factors, namely, the feed, nose radius,
1045
3 Problem definition
Most of the measurement techniques have limitations to
their in-process use. The purpose of the analysis is to
develop techniques to predict the surface roughness of a
1046
Unit
rpm
mm/rev
mm
Levels
1
2,500
0.05
0.2
3,250
0.12
0.3
3,500
0.15
0.4
Cutting
speed (v)
(rpm)
Training set
1
2
3
4
5
6
7
8
9
10
11
12
13
Feed rate
(f) (mm/
rev)
Depth of
cut (d)
(mm)
Actual
roughness
(Ra) (m)
2,500
2,500
2,500
2,500
2,500
2,500
2,500
2,500
2,500
3,250
3,250
3,250
3,250
0.05
0.05
0.05
0.12
0.12
0.12
0.15
0.15
0.15
0.05
0.05
0.05
0.12
0.2
0.3
0.4
0.2
0.3
0.4
0.2
0.3
0.4
0.2
0.3
0.4
0.2
0.7933
0.8433
0.9467
1.3267
1.1633
1.4467
0.96
1.0033
0.9733
0.81
0.9033
0.9967
0.96
14
3,250
15
3,250
16
3,250
17
3,250
18
3,250
19
3,500
20
3,500
21
3,500
22
3,500
23
3,500
24
3,500
25
3,500
26
3,500
27
3,500
Test readings
28
2,750
29
2,750
30
2,750
31
3,000
0.12
0.12
0.15
0.15
0.15
0.05
0.05
0.05
0.12
0.12
0.12
0.15
0.15
0.15
0.3
0.4
0.2
0.3
0.4
0.2
0.3
0.4
0.2
0.3
0.4
0.2
0.3
0.4
1.24
1.4867
0.9633
1.0967
1.1633
0.94
0.9533
1.11
0.8
1.23
1.09
1.01
1.06
1.0467
0.08
0.08
0.08
0.1
0.15
0.25
0.35
0.15
1.07
1.0667
1.0533
1.2833
32
33
34
35
36
0.1
0.1
0.13
0.13
0.13
0.25
0.33
0.15
0.25
0.35
1.41
1.0567
1.4433
1.1233
1.2933
3,000
3,000
3,300
3,300
3,300
1047
K
1X
dk:p ok:p
2 k1
where dk,p =desired output for the pth pattern. The weights
of the links are updated as
2
1 XNP XK
d
o
kp
kp
p1
K1
NP
1048
Cutting
Speed (v)
(rpm)
Feed rate
(f) (mm/
rev)
Depth of
cut (d)
(mm)
Training set
1
2,500
2
2,500
3
2,500
4
2,500
5
2,500
6
2,500
7
2,500
8
2,500
9
2,500
0.05
0.05
0.05
0.12
0.12
0.12
0.15
0.15
0.15
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
3,250
3,250
3,250
3,250
3,250
3,250
3,250
3,250
3,250
3,500
3,500
3,500
3,500
3,500
3,500
3,500
3,500
3,500
Test readings
28
2,750
29
2,750
30
2,750
31
3,000
32
3,000
33
3,000
34
3,300
35
3,300
36
3,300
Actual
roughness
(m)
Difference
% Error
Trial
1
Trial
2
Trial
3
Average
0.2
0.3
0.4
0.2
0.3
0.4
0.2
0.3
0.4
0.95
1.01
1.07
1.44
1.33
1.56
1.13
1.17
1.12
0.93
0.96
1.08
1.47
1.3
1.59
1.1
1.14
1.09
0.92
0.98
1.11
1.19
1.28
1.61
1.07
1.12
1.13
0.93333333
0.98333333
1.08666667
1.36666667
1.30333333
1.58666667
1.1
1.14333333
1.11333333
0.7933
0.8433
0.9467
1.3267
1.1633
1.4467
0.96
1.0033
0.9733
0.140033333
0.140033333
0.139966667
0.039966667
0.140033333
0.139966667
0.14
0.140033333
0.140033333
17.65200218
16.60539942
14.78469068
3.012487123
12.0375942
9.674892284
14.58333333
13.95727433
14.38747902
0.05
0.05
0.05
0.12
0.12
0.12
0.15
0.15
0.15
0.05
0.05
0.05
0.12
0.12
0.12
0.15
0.15
0.15
0.2
0.3
0.4
0.2
0.3
0.4
0.2
0.3
0.4
0.2
0.3
0.4
0.2
0.3
0.4
0.2
0.3
0.4
0.94
1.01
1.14
1.13
1.36
1.62
1.1
1.24
1.33
1.06
1.11
1.23
0.93
1.35
1.24
1.14
1.18
1.19
0.95
1.05
1.15
1.08
1.38
1.63
1.08
1.26
1.29
1.08
1.07
1.25
0.93
1.35
1.24
1.14
1.18
1.19
0.96
1.07
1.12
1.09
1.4
1.63
1.13
1.21
1.29
1.1
1.1
1.27
0.93
1.39
1.21
1.15
1.21
1.18
0.95
1.04333333
1.13666667
1.1
1.38
1.62666667
1.10333333
1.23666667
1.30333333
1.08
1.09333333
1.25
0.93
1.36333333
1.23
1.14333333
1.19
1.18666667
0.81
0.9033
0.9967
0.96
1.24
1.4867
0.9633
1.0967
1.1633
0.94
0.9533
1.11
0.8
1.23
1.09
1.01
1.06
1.0467
0.14
0.140033333
0.139966667
0.14
0.14
0.139966667
0.140033333
0.139966667
0.140033333
0.14
0.140033333
0.14
0.13
0.133333333
0.14
0.133333333
0.13
0.139966667
17.28395062
15.50241706
14.0430086
14.58333333
11.29032258
9.414587117
14.53683518
12.76253001
12.0375942
14.89361702
14.6893248
12.61261261
16.25
10.8401084
12.8440367
13.20132013
12.26415094
13.3721856
0.08
0.08
0.08
0.1
0.1
0.1
0.13
0.13
0.13
0.15
0.25
0.35
0.15
0.25
0.33
0.15
0.25
0.35
1.21
1.2
1.2
1.44
1.55
1.21
1.57
1.24
1.44
1.19
1.21
1.18
1.43
1.55
1.19
1.58
1.27
1.42
1.23
1.21
1.2
1.4
1.55
1.19
1.6
1.28
1.44
1.21
1.2067
1.1933
1.4233
1.55
1.1967
1.5833
1.2633
1.4333
1.07
1.0667
1.0533
1.2833
1.41
1.0567
1.4433
1.1233
1.2933
0.14
0.14
0.14
0.14
0.14
0.14
0.14
0.14
0.14
Average
13.08411215
13.12458986
13.29155986
10.90937427
9.929078014
13.24879341
9.699993071
12.46327784
10.82502126
12.93591353
the test readings are then given and the network is trained. The
target value is then obtained and compared with actual output.
The predicted Ra values are compared with the actual Ra values
and the predicted Ra values obtained from the present study
show minimal in variations. The parameters taken could be
1049
1.4
Ra
1.2
1.0
3500
0.8
3000
0. 10
Feed
Speed
0.05
2500
0.15
1.4
Ra
1.2
1.0
0.15
0.8
0.10
2500
3000
S peed
0.05
3500
Feed
1050
1.4
Ra
1.4
1.2
Ra
1.0
1.0
0.4
0.8
0.3
3000
0.4
0.8
DOC
2500
Speed
1.2
0.3
DOC
0.05
0. 2
0. 10
3500
0. 2
0.15
Feed
1.4
1.4
Ra
1.2
Ra
1.0
3500
0.8
3000
1.2
1.0
0.15
0.8
0.10
Speed
0.2
0.2
0.3
DOC
2500
0.4
0.3
DOC
0.05
0.4
Feed
6 Conclusions
In this work, 36 specimens which are made up of the brass
C26000 material have been machined in a CNC turning
machine and then a TIME TR 100 surface roughness tester was
used to measure the roughness average (Ra) values of all the
specimens. The model was developed to predict the surface
roughness of Brass C26000 through Artificial Neural Networks technique by utilizing feed-forward back-propagation
method. This type of model has been evaluated by means of
the percentage deviation between the predicted Ra values and
the actual Ra values. The important conclusions drawn from
the present research are summarized as follows:
1. The surface roughness could be effectively predicted by
using spindle speed, feed rate, and depth of cut as the
input variables.
2. Considering the individual parameters, feed rate has
been found to be the most influencing parameter,
followed by spindle speed and depth of cut.
3. Model (including interaction terms), considering the
interaction between the individual parameters, could
achieve an accuracy of 87.07%.
4. The average actual roughness Ra value has been
obtained as 1.1999 m and the corresponding predicted
surface roughness value is 1.1859 m.
5. As the spindle speed increases, for lower feed rates, the
surface roughness decreases, for higher feed rates, the
surface roughness changes considerably.
6. The depth of cut influences the surface roughness
considerably for a given feed rate.
7. The increase in feed rate causes the surface roughness
to increase and then decrease. For lower depth of cut,
as the feed rate increases, surface roughness decreases
and then increases.
References
1. Lou MS, Chen JC, Li M (1998) Surface roughness prediction for
CNC End milling, materials and processes quality control
manufacturing. J Ind Technol 15(1):26
2. Suhail AH, Ismail N (2010) Optimization of cutting parameters
based on surface roughness and assistance of workpiece surface
temperature in turning process. Amer J Eng Appl Sci 3(1):102108
3. Al-Ahmari AMA (2007) Predictive machinability models for a
selected hard material in turning operations. J Mater Process
Technol 1(190):305311
1051
4. Banerjee A, Bordatchev EV, Choudhury SK (2009) On-line
monitoring of surface roughness in turning operations with optoelectrical transducer. Int J Manuf Res 4(1):5773
5. Azouzi R, Guillot M (1997) On-line prediction of surface finish
and dimensional deviation in turning using neural network based
sensor fusion. Int J Mach tools Manuf 37(9):12011217
6. Bajic D, Lela B, Zivkovic D (2008) Modeling of machined
surface roughness and optimization of cutting parameters in face
milling. Metalurgija 47(45):331334
7. Villaseor D, Morales R, Menndez C, Rodrguez J, Alique R (2006)
Neural networks and statistical based models for surface roughness
prediction. International Association of Science and Technology for
Development. Proceedings of the 25th IASTED International
Conference on Modeling, Identification and Control, pp. 326331
8. Cus F, Balic J (2003) Optimization of cutting process by GA
approach. Robot Comput-Integr Manuf 19:113121
9. Groover M (1996) Fundamentals of modern manufacturing. Prentice
Hall, Upper Saddle River (now published by Wiley, New York)
10. Hossain MI, Amin AKM, Patwari AU (2008) Development of an
artificial neural network algorithm for predicting the surface
roughness in end milling of Inconel 718 alloy. Int Conf Comput
Commun Eng ICCCE 13(15):13211324
11. Kadirgama K, Noor MM (2008) Optimization of surface roughness in end milling on mould aluminum alloys using response
surface method and radian basis function network. Jordan Journal
of Mechanical and Industrial Engineering 2(4):209214
12. Huang L, Chen JC (2001) A multiple regression model to predict
in-process surface roughness in turning operation via accelerometer. J Ind Technol 17(2):18
13. Mahdavinejad RA, Sharifi Bidgoli H (2009) Optimization of
surface roughness parameters in dry turning. J Achiev Mater
Manuf Eng 37(2):571577
14. Nalbant M, Gokkaya H, Toktas I (2007) Comparison of regression
and artificial neural network models for surface roughness prediction
with the cutting parameters in CNC turning. Modelling and
Simulation in Engineering. Hindawi Publishing Corp, New York
15. Oktem H, Erzurumlu T, Kurtaran H (2005) Application of
response surface methodology in the optimization of cutting
conditions for surface roughness. J Mater Process Technol
170:1116
16. Davima JP, Gaitondeb VN, Karnik SR (2008) Investigations into the
effect of cutting conditions on surface roughness in turning of free
machining steel by ANN models. J Mater Process Technol 205:1623
17. Parappagoudar MB, Pratihar DK, Datta GL (2008) Forward and
reverse mappings in green sand mould system using neural
networks. Appl Soft Comput 8:239260
18. Saad Kareem Shather,Abbas Fadhel Ibrheem (2007) Prediction of
surface roughness in end milling with multiple regression model.
Engineering and Technology 26(3)
19. Sakir Tasdemir, Sleyman Neseli, Ismail Saritas, Sleyman Yaldiz
(2008) Prediction of surface roughness using artificial neural
network in lathe. In: Proceedings on CompSysTech '08 Proceedings of the 9th International Conference on Computer Systems
and Technologies and Workshop for PhD Students in Computing
20. Srikanth T, Kamala V (2008) A real coded genetic algorithm for
optimization of cutting parameters in turning. Int J Comput Sci
Net Secur 8(6):189193
21. Suresh PVS, Venkateswara RP, Deshmukh SG (2002) A genetic
algorithmic approach for optimization of surface roughness
prediction model. Int J Mach Tools Manuf 42:675680
22. Uros Z, Franci C (2003) Optimization of cutting conditions during
cutting by using neural networks. Robot Comput-Integr Manuf
19:189199
23. Yang WH, Tarng YS (1998) Design optimization of cutting
parameters for turning operations based on the Taguchi method. J
Mater Process Technol 84:122129