You are on page 1of 15

Prediction of surface roughness of end

milling operation using genetic algorithm

G. Mahesh, S. Muthu & S. R. Devadasan

The International Journal of


Advanced Manufacturing Technology

ISSN 0268-3768
Volume 77
Combined 1-4

Int J Adv Manuf Technol (2015)


77:369-381
DOI 10.1007/s00170-014-6425-z

1 23
Your article is protected by copyright and
all rights are held exclusively by Springer-
Verlag London. This e-offprint is for personal
use only and shall not be self-archived
in electronic repositories. If you wish to
self-archive your article, please use the
accepted manuscript version for posting on
your own website. You may further deposit
the accepted manuscript version in any
repository, provided it is only made publicly
available 12 months after official publication
or later and provided acknowledgement is
given to the original source of publication
and a link is inserted to the published article
on Springer's website. The link must be
accompanied by the following text: "The final
publication is available at link.springer.com”.

1 23
Author's personal copy
Int J Adv Manuf Technol (2015) 77:369–381
DOI 10.1007/s00170-014-6425-z

ORIGINAL ARTICLE

Prediction of surface roughness of end milling operation using


genetic algorithm
G. Mahesh & S. Muthu & S. R. Devadasan

Received: 9 December 2012 / Accepted: 22 September 2014 / Published online: 17 October 2014
# Springer-Verlag London 2014

Abstract In the present study, the predictive model is devel- 1 Introduction


oped to observe the effect of radial rake angle on the end
milling cutting tool by considering the following machining End milling is a process of generating a machined surface by
parameters: spindle speed, feed rate, axial depth of cut, and gradually removing a predetermined amount of material from
radial depth of cut. By referring to the real machining case the work piece with a minimum feed rate of a milling cutter
study, the second-order mathematical models have been de- rotating at a high speed.
veloped using response surface methodology (RSM). A num- While machining, the surface roughness plays a major role
ber of machining experiments based on statistical five-level for evaluating the work piece quality. Surface roughness is
full factorial design of experiments are carried out in order to influenced by controlled machining parameters, such as feed
collect surface roughness values. The direct and interaction rate, spindle speed, depth of cut, as well as by non-controlled
effects of the machining parameter with surface roughness are influences such as non-homogeneity of the work piece and the
analyzed using Design Expert software. The optimal surface tool, tool wear, machine-motion errors, formation of chips,
roughness value can be attained within the specified limits by and unpredictable random disturbances [1]. Surface roughness
using RSM. The genetic algorithm (GA) model is trained and is most influenced by the feed rate, whereas the vibrations
tested in MATLAB to find the optimum cutting parameters increase the prediction accuracy [2]. Waviness refers to irreg-
leading to minimum surface roughness. The GA recommends ularities on the surface produced by machine or work piece
0.25 μm as the best minimum predicted surface roughness deflections, structural reasons, vibrations, and other irregular-
value. The confirmatory test shows the predicted values which ities in the cutting process [3]. The cutting speed, feed rate,
were found to be in good agreement with observed values. and depth of cut are the important parameters affecting surface
finish. But these parameters are strongly dependent on the
dynamic properties of the tool as well as spindle [4].
Keywords Radial rake angle . RSM . GA . DoE . Vertical Researchers have proposed several approaches to predict
milling machine . Surface roughness surface roughness based on theoretical approach [5, 6]. Ana-
lytical models had been created to predict surface roughness
and tool life in terms of cutting speed, feed rate, and axial
G. Mahesh (*)
depth of cut in milling material [7, 8], experimental investi-
Department of Mechanical Engineering, Sree Sakthi Engineering
College, Karamadai, Coimbatore, Tamil Nadu 641104, India gations [9, 10], design of experiments [11, 12], and optimiza-
e-mail: msv240606@gmail.com tion techniques such as neural network, genetic algorithm,
fuzzy logic, and neural fuzzy approaches to predict surface
S. Muthu
roughness [13] [14–18].
Mechanical Department, NGP Institute of Technology, Coimbatore,
Tamil Nadu 641048, India The usage of genetic algorithm for the optimization of
e-mail: smuthu231155@gmail.com cutting conditions to predict surface roughness in end milling
process was very limited [19]. The integrated genetic pro-
S. R. Devadasan
gramming and genetic algorithm approach were presented to
Department of Production Engineering, PSG College of Technology,
Coimbatore, Tamil Nadu, India predict surface roughness based on cutting parameters (spin-
e-mail: devadasan_srd@yahoo.com dle speed, feed rate, and depth of cut) and on vibrations
Author's personal copy
370 Int J Adv Manuf Technol (2015) 77:369–381

between a cutting tool and work piece [20]. For the cutting angle may affect the quality of the machined surface, and an
conditions of feed rate, cutting speed, and axial depth of cut, increase in the rake angle improves the overall quality of the
GA was used to predict surface roughness (Ra) value which is machined edge [30]. Rake and clearance angles are the most
lower than the values of experimental results [21]. The surface important design factors. Particularly, rake angle affects stiff-
roughness value predicted in GA optimization technique de- ness of the cutting edge and rigidity of the tool. Convention-
creases with high cutting speed and very small feed rate [22]. ally, an end mill with a positive rake angle improves machin-
In order to reduce machining time and to achieve better ing ability. In case of a negative rake angle, the stiffness of the
surface roughness and metal removal rate, a combination of cutting edge is increased and chipping is suppressed [31].
high speed, low feed, with moderate depth of cut were to be Experiments were conducted to observe the effect of tool
selected for machining processes. GA approach was used to geometry on the quality of surface produced by considering
find the optimum machining conditions for corresponding the radial rake angle as 4°,10°, and 16°; nose radius as
given maximum and minimum values of surface roughness. 0.4°,0.8°, and 1.2°; and cutting conditions (cutting speed
The surface roughness increases with an increase in the depth and feed rate) on the machining performance in dry milling.
of cut [23]. Surface roughness is influenced by tool geometry, Mathematical models have been developed for surface rough-
feed rate, cutting conditions, and irregularities of machining ness prediction using RSM and validated through chi-square
operations such as tool wear, chatter, tool deflections, cutting test. Genetic Algorithm is used to find the optimum process
fluid, and work piece properties. Feed rate has the greatest parameters. The authors [32, 33] investigated the effect of
influence on surface roughness in the milling process when graphite molybdenum disulfide as solid lubricants at the con-
evaluated by the application of the GA optimization technique tact surface between tool and work piece during machining by
[24]. The high cutting speed values for the milling process was using end mill cutters of different tool geometries (radial rake
preferred for the low surface roughness value by the GA angle and nose radius) under different cutting speeds and feed
technique [25]. The effect of cutting conditions like feed rate, rates [34]. Experimental studies have been conducted to ob-
spindle speed, axial–radial depth of cut, and radial rake angle serve the effect of tool geometry (radial rake angle and nose
on surface roughness was discussed. GA reduces the Ra value radius) and cutting conditions (cutting speed and feed rate) on
in the mould cavity from 0.412 to 0.375 μm [26]. the quality of surface produced in dry milling with four fluted
solid TiAIN-coated carbide end mill cutters. Mathematical
1.1 Influence of radial rake angle in end mill cutter models have been developed to predict surface roughness by
using RSM. The optimization was carried out by using genetic
Rake angle is the angle between the leading edge of a cutting algorithm to obtain the best possible tool geometry and cutting
tool and perpendicular to the surface being cut. Rake angles conditions. An attempt has been made to predict the process
for milling cutters are specified in two directions, axial and parameters such as tool wear and vibration by considering the
radial rake angles. Axial rake is the cutting insert’s angle with significant parameters such as tool diameter, number of flutes,
respect to the central axis of the cutter/spindle assembly. rake and clearance angles, and work piece material [35].
Radial rake is the cutting insert’s angle with respect to the From the literature sources, it is found that the machining
periphery of the cutter. Common configurations include (a) of Al 6063 metal matrix composite is an important area of
positive in both directions, (b) negative in both directions, and research, but only very few studies have been carried out on
(c) positive in one direction and negative in the other. Exper- surface roughness in end milling. In the present study, an
imental studies had been conducted to predict the effects of attempt is made to investigate the effect of process parameters
rake angle (−4°, 0°, 4°, 8°, 12°, 16°, 20°) and helix angles on such as spindle speed, feed rate, radial depth of cut, axial depth
the cutting forces variations during milling of hard materials. of cut, and radial rake angle on surface roughness in end
Rake angle becomes higher; the cutting forces components milling using RSM approach. This methodology helps to
become lower [27]. An experiment had been conducted by obtain best possible cutting conditions and tool geometry for
using an end mill cutter to predict surface roughness by dry milling of Al-6063 using High-speed steel (HSS) end mill
considering the radial rake angle (6.2°, 7.0°, 9.5°, 13.0°, and cutters. The adequacy of the developed mathematical model
14.8°) combined with cutting speed and feed rate. The authors has also been tested by the analysis of variance test. Design
also suggested that the optimization of cutting conditions to Expert 6.0 package is used to analyze the data and develop the
predict surface roughness (Ra) in end milling involving radial model.
rake angle is still lacking [28]. Mathematical models have
been developed to predict surface roughness using regression. 1.2 Optimization by using GA
The genetic algorithm was used to predict the optimum pro-
cess parameters [29]. An integrated SA-GA approach was Genetic algorithm (GA) is a procedure used to find approxi-
used to optimize the surface roughness value by considering mate solutions to search problems through the application of
the same parameters. The authors also suggested that the rake the principles of evolutionary biology. Genetic algorithms use
Author's personal copy
Int J Adv Manuf Technol (2015) 77:369–381 371

biologically inspired techniques such as genetic inheritance, adopts both mathematical and statistical techniques which
natural selection, mutation, and sexual reproduction (recom- are useful for the modeling and analysis of problems in which
bination or crossover). Genetic algorithms are typically im- a response of interest is influenced by several variables. By
plemented using computer simulations in which an optimiza- conducting experiments and the posterior application of re-
tion problem is specified. For this problem, members of a gression analysis, a model of the response variable of interest
space of candidate solutions, called individuals, are represent- is obtained. The real relationship between the response and the
ed using abstract representations called chromosomes. The independent variables is unknown. For that reason, the first
GA consists of an iterative process that evolves a working step in RSM is to find an approximation of the true functional
set of individuals called a population toward an objective relationship between the response and the independent vari-
function or fitness function [36]. The evolutionary process ables [38]. The objective is to optimize the response. Similar-
of a GA is a highly simplified and stylized simulation of the ly, various researches have been done to predict surface
biological version. It starts from a population of individuals roughness theoretically through the equation
randomly generated according to some probability distribu-
tion, usually uniform and updates this population in steps Ra ¼ cN k1 Bk2 C k3 Dk4 γ k5 ð1Þ
called generations. In each generation, multiple individuals
are randomly selected from the current population based upon
some application of fitness, bred using crossover, and modi- where
fied through mutation to form a new population.
Ra is the predicted surface roughness (μm).
Crossover—exchange of genetic material (substrings) N is the spindle speed (m/min).
denoting rules, structural components, features of a ma- B is the feed rate (mm/rev).
chine learning, search, or optimization problem. C is the axial depth of cut (mm).
Selection—the application of the fitness criterion to D is the radial depth of cut (mm).
choose which individuals from a population will go on Γ is the radial rake angle (°).
to reproduce. k1, k2, k3, k4, are the model parameters (to be estimated from
Replication—the propagation of individuals from one k5 experimental data).
generation to the next. c is response error.
Mutation—the modification of chromosomes for single Equation (1) may be written as
individuals.
lnRa ¼ ln C þ k 1 lnN þ k 2 lnB þ k 3 ln C þ k 4 lnD þ k 5 lnγ
The solution of a problem that GA attempts to solve is coded ð2Þ
into a string of binary numbers known as chromosomes. Each
chromosome contains the information of a set of possible The linear model of Eq. (2) is
process parameters. Initially, a population of chromosomes is
formed randomly. The fitness of each chromosome is then y ¼ β 0 x0 þ β 1 N þ β 2 B þ β 3 C þ β 4 D þ β 5 γ ð3Þ
evaluated using an objective function after the chromosome
has been decoded. Selected individuals are then reproduced, where y is the true response of the surface roughness on a
usually in pairs, through the application of genetic operators. logarithmic scale x0 =1 (dummy variable) and N, B, C, D, and
The operators are applied to pairs of individuals with a given γ are the logarithmic transformations of spindle speed, feed
probability and result in new offspring. The offspring from rate, axial depth of cut, radial depth of cut, and radial rake
reproduction are then further perturbed by mutation. These angle, respectively.
new individuals then make up the next generation. These The general second-order model is
processes of selection, reproduction, and evaluation are repeat- 0
ed until some termination criteria are satisfied. y ¼ y‐ε ¼ β0 x0 þ β1 x1 þ β2 x2 þ β3 x3 þ β4 x4 þ
β5 x5 þ β11 x1 2 þ β22 x2 2 þ β 33 x3 2 þ β44 x4 2 þ
β 55 x5 2 þ β12 x1 x2 þ β13 x1 x3 þ β14 x1 x4 þ ð4Þ
β15 x1 x5 þ β23 x2 x3 þ β 24 x2 x4 þ β 25 x2 x5 þ
β34 x3 x4 þ β35 x3 x5 þ β 45 x4 x5
2 Surface roughness

There are various surface roughness amplitude parameters


such as roughness average (Ra), root–mean–square (RMS), where
and maximum peak-to-valley roughness (Rmax), which are y′ is the estimated response and y is measured as surface
used in industries [37]. Response surface method (RSM) roughness based on the logarithmic scale, ε is the
Author's personal copy
372 Int J Adv Manuf Technol (2015) 77:369–381

Table 1 Factor and their levels


Parameter (input variables) Units Factor level

−2 −1 0 1 2

Spindle speed (N) RPM 1500 2000 2500 3000 3500


Feed rate (B) mm/rev 0.04 0.06 0.08 0.1 0.12
Axial depth of cut (C) mm 0.5 0.6 0.7 0.8 0.9
Radial depth of cut (D) mm 0.5 0.6 0.7 0.8 0.9
Radial rake angle (γ) Degree 12 15 18 21 24

experimental error, and β0 is the free term of the regression 4. The objective function of GA leads to the minimum value
equation. The coefficients β1, β2, β3, β4, and β5 are linear of surface roughness (Ra). MAT LAB software with GA
terms. The coefficients β11, β22, β33, β44, and β55 are quadratic tool box is used to optimize the minimum value of surface
terms, and the coefficients β12, β13, β14, β15, β23, β24, β25, β34, roughness.
β35, and β45 are interaction terms [39]. 5. GA optimization solution is evaluated. The optimum
cutting condition which leads to achieve minimum sur-
face roughness values generated from GA are compared
2.1 Experimental procedure with experimented value and RSM result.

Five phases involved in the end milling operation are as


follows: 2.2 Identification of process parameter

1. The experiments are conducted on a vertical machining The independently controllable process parameters af-
center with high-speed steel end mill cutter under dry fecting the surface roughness were identified to carrying
condition by considering the process parameters such as out the experimental work. The parameters and its
spindle speed, feed rate, radial depth of cut, axial depth of ranges, spindle speed, feed rate, radial depth of cut,
cut, and radial rake angle. The surface roughness is mea- and axial depth of cut are selected based on suggestion
sured by using a surface roughness tester. given by [40, 41]. The ranges of machining conditions
2. The design matrix is selected for conducting the experi- and tool radial rake angle were selected as recommend-
ment at five levels, five factor central composite rotatable ed by the tool manufacturer and machining data hand-
designs. Thus, the 32 experimental runs allow the estima- book [42]. The surface roughness is the output
tion of the linear, quadratic, and two-way interactive response.
effects of the process parameters. The cutting parameters selected for minimizing the surface
3. A second-order quadratic model is developed for the roughness
prediction of surface roughness. The model is checked
for its adequacy using ANOVA. Minimize : Ra ðN ; B; C; D; γ Þ ð5Þ

Table 2 Chemical composition


and mechanical properties Chemical composition of Al 6063 alloy Mechanical properties of Al 6063 alloy

Chemical Element % Present Brinell hardness 73


Si 0.2–0.6 Vickers hardness 83
Fe 0.35 Ultimate tensile strength 241 MPa
Cu 0.10 Tensile yield strength 214 MPa
Mg 0.45–0.9 Modulus of elasticity 68.9 GPa
Mn 0.10 Ultimate bearing strength 434 Mpa
Cr Max 0.1 Bearing yield strength 276 Mpa
Zn Max 0.1 Poission’s ratio 0.33
Ti Max 0.1 Shear strength 152 MPa
Others Total Max 0.15
Al Bal
Author's personal copy
Int J Adv Manuf Technol (2015) 77:369–381 373

Table 3 Design matrix and response

Run Factor 1 Factor 2 Factor 3 Factor 4 Factor 5 Response 1


N/spindle speed B/feed rate C/axial depth D/radial depth γ/Radial rake angle Surface roughness (Ra)

rpm mm/rev mm mm Degree μm


1 1 (3000) −1 (0.06) −1 (0.6) −1 (0.6) −1 (15) 0.43
2 0 (2500) 0 (0.08) 0 (0.7) 0 (0.7) 0 (18) 0.45
3 1 (3000) 1 (0.10) −1 (0.6) 1 (0.8) −1 (15) 0.49
4 0 (2500) 0 (0.08) 0 (0.7) 0 (0.7) 0 (18) 0.48
5 −1 (2000) 1 (0.10) 1 (0.8) 1 (0.8) −1 (15) 0.51
6 0 (2500) 2 (0.12) 0 (0.7) 0 (0.7) 0 (18) 0.58
7 −1 (2000) −1 (0.06) −1 (0.6) 1 (0.8) −1 (15) 0.53
8 0 (2500) 0 (0.08) 2 (0.9) 0 (0.7) 0 (18) 0.45
9 1 (3000) −1 (0.06) 1 (0.8) −1 (0.6) 1 (21) 0.46
10 1 (3000) 1 (0.10) 1 (0.8) 1 (0.8) 1 (21) 0.50
11 1 (3000) 1 (0.10) 1 (0.8) −1 (0.6) −1 (15) 0.53
12 1 (3000) 1 (0.10) −1 (0.6) −1 (0.6) 0 (18) 0.50
13 1 (2000) 1 (0.10) −1 (0.6) −1 (0.6) −1 (15) 0.52
14 1 (3000) −1 (0.06) 1 (0.8) 1 (0.8) −1 (15) 0.48
15 −1 (2000) −1 (0.06) 1 (0.8) 1 (0.8) 1 (21) 0.60
16 1 (3000) −1 (0.06) −1 (0.6) 1 (0.8) 1 (21) 0.51
17 0 (2500) −2 (0.04) 0 (0.7) 0 (0.7) 0 (18) 0.55
18 −1 (2000) 1 (0.10) −1 (0.6) 1 (0.8) 1 (21) 0.61
19 0 (2500) 0 (0.08) 0 (0.7) 0 (0.7) 2 (24) 0.65
20 −1 (2000) −1 (0.06) 1 (0.8) −1 (0.6) −1 (15) 0.58
21 2 (3500) 0 (0.08) 0 (0.7) 0 (0.7) 0 (18) 0.41
22 −1 (2000) 1 (0.10) 1 (0.8) 0 (0.6) 1 (21) 0.53
23 0 (2500) 0 (0.08) 0 (0.7) 0 (0.7) −2 (12) 0.58
24 0 (2500) 0 (0.08) 0 (0.7) 0 (0.7) 0 (18) 0.49
25 0 (2500) 0 (0.08) −2 (0.5) 0 (0.7) 0 (18) 0.46
26 0 (2500) 0 (0.08) 0 (0.7) 2 (0.9) 0 (18) 0.51
27 0 (2500) 0 (0.08) 0 (0.7) 0 (0.7) 0 (18) 0.52
28 0 (2500) 0 (0.08) 0 (0.7) −2 (0.5) 0 (18) 0.46
29 −1 (2000) −1 (0.06) −1 (0.6) −1 (0.6) 1 (21) 0.53
30 −2 (1500) 0 (0.08) 0 (0.7) 0 (0.7) 0 (18) 0.53
31 0 (2500) 0 (0.08) 0 (0.7) 0 (0.7) 0 (18) 0.52
32 0 (2500) 0 (0.08) 0 (0.7) 0 (0.7) 0 (18) 0.45

2.3 Finding the limits of the process parameters

The working ranges of all selected factors were fixed by


conducting trial runs. This was carried out by varying one of
the factors while keeping the rest of them as constant values.
The upper limit of a given factor was coded as (+2) and the
lower limit was coded as (−2). The coded values for interme-
diate values were calculated using Eq. (1) [43]

X i ¼ 2f2X −ðX max þ X min Þg X max −X min ð6Þ

where Xi is the required coded value of a variable X and is


Fig. 1 Surface roughness tester (SURFTEST SJ-201) any value of the variable from Xmax and Xmin. The selected
Author's personal copy
374 Int J Adv Manuf Technol (2015) 77:369–381

Table 4 ANOVA table for the


prediction of Ra Source Sum of squares df Mean square F value P value
Prob>F

Model 0.088224 20 0.0044112 9.02604 0.0003 Significant


N-spindle speed 0.019824 1 0.01982351 40.5621 <0.0001
B-feed rate 0.001336 1 0.001336 2.73443 0.1264
C-axial depth 7.83E-08 1 7.82755E-08 0.00016 0.9901
D-radial depth 0.001602 1 0.0016022 3.27837 0.0976
γ-Radial rake angle 0.004862 1 0.0048621 9.94880 0.092
NB 0.004111 1 0.0041108 8.41155 0.0144
NC 0.000103 1 0.0001030 0.21089 0.655
ND 0.000399 1 0.0003985 0.81550 0.3858
Nγ 7.34E-05 1 7.3396E-05 0.15018 0.7058
BC 0.002949 1 0.0029489 6.03411 0.0319
BD 0.001203 1 0.0012027 2.46110 0.145
Bγ 0.000206 1 0.00020587 0.42125 0.5296
CD 0.000846 1 0.0008464 1.732048 0.2149
Cγ 0.003091 1 0.0030909 6.324573 0.0287
Dγ 0.002046 1 0.0020459 4.186304 0.0654
N^2 0.000568 1 0.0005684 1.163054 0.3039
B^2 0.010916 1 0.0109163 22.33658 0.0006
C^2 0.001945 1 0.0019450 3.979895 0.0714
D^2 1.28E-05 1 1.28153E-05 0.026222 0.8743
γ^2 0.029088 1 0.0290878 59.51844 <0.0001
Residual 0.005376 11 0.0004887
Lack of fit 0.000426 6 7.09857E-05 0.071702 0.997 Not significant
Pure error 0.00495 5 0.00099
Cor total 0.0936 31
Std. dev. 0.02210 R2 0.94256
2
Mean 0.5125 Adj R 0.83813

process parameter with their limits and notations are given in a half replication 24 = 16 factorial design plus 6 center
Table 1. points and 10 star points. The upper (+2) and lower (−2)
levels of all the five variables as shown in Tables 1 and 2
2.4 Development of design matrix are established by trial runs prior to the actual predefined
parameters to establish the exact predicted values. The
The design matrix chosen to conduct the experiments was a intermediate levels of −1, 0, +1 of all the variables had
five-factor central composite rotatable design (CCD) been calculated by interpolation. The selected design ma-
consisting of 32 sets of coded conditions and comprising trix is shown in Table 3.

Fig. 2 Direct effect of radial rake angle over surface roughness Fig. 3 Direct effect of spindle speed over surface roughness
Author's personal copy
Int J Adv Manuf Technol (2015) 77:369–381 375

Fig. 4 Direct effect of axial depth of cut over surface roughness


Fig. 6 Direct effect of feed rate over surface roughness
2.5 Experimentation
of determinant R2 indicate the goodness of fit for the models
The test plates of size 50 mm×50 mm×50 mm are prepared and are nearer to 1. The regression equation obtained from the
from aluminum alloy (Al 6063) square block. The chemical Design Expert software in terms of actual factors is given
composition and mechanical properties of the base material below.
are presented in Table 2. The experiment was conducted on
MAKINO CNC Vertical Machining Center (Model-S33) with 2.6.1 Surface roughness (Ra)
high-speed steel end mill cutter under dry condition. The
surface roughness is measured by using a surface roughness
tester (SURFTEST SJ-201) as shown in Fig. 1.
¼ 0:33209 þ 3:48082E−005  N − 4:67956  D
þ 3:25372  B þ 0:56205  −C − 0:11719  γ
2.6 Response surface model for the prediction of surface
þ 1:63368E−003  N  D − 5:17353E−005  N  B
roughness
– 1:01735E−004  N  C − 1:55774E−006  N  γ
A rigorous analysis is carried out with the experimental data − 6:91838  D  B − 4:41838  D  C þ 0:065223  D  γ
using DESIGN EXPERT V7 software of state case. A second-
order quadratic model is developed for the prediction of − 0:74132  B  C − 0:050545  B  γ þ 0:041122  C  γ
surface roughness. The model is checked for its adequacy
−1:76502E−008  N 2 þ 48:34360  D2 − 0:81626  B2
using analysis of variance (ANOVA).
Table 4 is the ANOVA table for the prediction of Ra. In −0:066256  C 2 þ 3:54175E−003  γ 2 ð7Þ
Table 4, a “model F value” of 9.02604 with a “model P value”
less than 0.0003 implies that the selected model is significant
and values greater than 0.10 indicate the model terms are not
significant. The P value <0.0001 represents that there is only a
0.01 % chance that such model could occur due to noise. To
illustrate the relationship between the response and the pro-
cess parameters, the fitted polynomial equations are expressed
in the form of surface plots. Table 4 also shows the other
adequacy measures R2, Adj R2, and Pred R2. The coefficients

Fig. 7 Surface interaction plot of spindle speed and feed rate over surface
Fig. 5 Direct effect of radial depth of cut over surface roughness roughness
Author's personal copy
376 Int J Adv Manuf Technol (2015) 77:369–381

Fig. 8 Surface interaction plot of spindle speed and axial depth over
surface roughness Fig. 10 Surface interaction plot of spindle speed and radial rake angle
over surface roughness

3 Results and discussion


is clear that the radial rake angle and spindle speed have a
3.1 Direct effect of variables significant effect on the surface roughness. The effect of
process parameters on the surface roughness is discussed
In this work, the effects of radial rake angle, spindle speed, below. Figure 2 shows that the increase in radial rake angle
feed rate, axial depth of cut, and radial depth of cut were resulted in an increase in the surface roughness. This is due to
experimentally investigated. From Figs. 2, 3, 4, 5, and 6, it the fact that the surface roughness increases. Figure 3 shows

Fig. 9 Surface interaction plot of spindle speed and radial depth over Fig. 11 Surface interaction plot of feed rate and axial depth over surface
surface roughness roughness
Author's personal copy
Int J Adv Manuf Technol (2015) 77:369–381 377

Fig. 12 Surface interaction plot of feed rate and radial depth of cut

that if the spindle speed increases, the surface roughness Fig. 14 Surface interaction plot of axial depth and radial depth over
surface roughness
decreases and vice versa. It is evident that the spindle speed
is between 3000 and 3500 RPM resulted in better surface
roughness. Figures 4, 5, and 6 clearly indicate that the increase
in radial and axial depths of cut and feed rate increases the
surface roughness. This is due to the contact area, metal being
cut, and the cutting force increases; this makes the tool and
work piece unstable which results in the increase in the surface
roughness.

Fig. 13 Surface interaction plot of feed rate and radial rake angle over Fig. 15 Surface interaction plot of axial depth and radial rake angle over
surface roughness surface roughness
Author's personal copy
378 Int J Adv Manuf Technol (2015) 77:369–381

depth on surface roughness. At a lower spindle speed, the


increase in axial depth has a significant influence on surface
roughness whereas at a higher spindle speed, the variation in
axial depth does not significantly affect the surface roughness.
This is due to the fact that the higher torque produced at a lower
spindle speed has distorted the material surface with axial
loading. Figure 9 shows the interaction effect of spindle speed
and radial depth on surface roughness. From the figure, it is
seen that both spindle speed and radial depth are influencing on
the effect of the surface roughness. However, it is found that the
interaction effect between spindle speed and radial depth has
not played a significant role. Figure 10 shows the interaction
effect of spindle speed and radial rake angle on surface rough-
ness. The surface roughness is significantly higher at rake angle
between 18° and 24°. The similar trend is observed at all
spindle speeds. The result showed that the industry preferring
a good surface roughness has to fix the radial rake angle
between 13 and 18.
Figure 11 shows the interaction effect of feed rate and axial
depth on surface roughness. It is evidenced that the interaction
effect between axial depth and feed rate has a greater influence
on surface roughness. It is noticed that at a higher feed rate
Fig. 16 Surface interaction plot of radial depth and radial rake angle over
along with axial depth of cut between 0.7 and 0.9 mm, the
surface roughness
surface of the material is distorted and produced higher surface
roughness. Interaction effect of feed rate and radial depth on
3.2 Interaction effect surface roughness is shown in Fig. 12; higher feed rate and
radial depth increase the surface roughness whereas lower feed
The interaction effect of process parameters on the surface rate and radial depth decrease the surface roughness. To obtain
roughness is discussed below. Figure 7 shows the interaction the quality surface of Ra between 0.37 and 0.28, feed rate less
plot between surface roughness and spindle speed with respect than 0.09 and radial depth less than 0.7 should be preferred.
to feed rate. If the feed rate increases gradually toward the Interaction effect of feed rate and radial rake angle on the
speed, the surface roughness value decreases proportionately. surface roughness are depicted in Fig. 13. The surface rough-
As the spindle speed increases from 1500 to 3500 RPM, the ness “Ra” is converged in all means from the maximum to a
surface roughness value is reduced. From the result, it is evi- minimum value at the region of feed rate (0.07–0.05) and radial
denced that at a lower spindle speed, the change in feed rate rake angle (18–15). Interaction effects of axial depth and radial
does not affect the roughness of the milling surface. But at a depth on the surface roughness are depicted in Fig. 14. In order
higher spindle speed, the change in feed rate has a significant to get a better surface finish, axial depth of cut and radial depth
effect. Hence, it can be concluded that at a higher spindle speed, of cut should be minimum as far as possible. Comparing the
a lower feed rate has to be chosen for quality surface finish. The factor axial and radial depth of cut, the axial depth of cut has an
inference can also be verified from the ANOVA (Table 4). exponential trend on the surface roughness.
Figure 8 shows the interaction effect of spindle speed and axial Interaction effects of axial depth and radial rake angle on
the surface roughness are depicted in Fig. 15. It is interesting
Table 5 GA parameters to observe that if the radial rake angle was in the region
Parameters Setting values
between 13 and 18, the milling process will produce a good
Population size 100 surface roughness even at all axial depth of cut. The higher as
Scaling function Rank well as lower radial rake angle produced maximum surface
Function Stochastic uniform roughness. Thus, the result showed an optimum rake angle
Mutation function Gaussian should be preferred for a good quality finish in end milling
Mutation rate 0.1 process. Interaction effects of radial depth and radial rake
Crossover function Scattered angle on the surface roughness are depicted in Fig. 16; as
Crossover rate 1.0 observed from the trend of axial and radial rake angles, the
Generations 1000 similar trend is also noticed between radial depth and radial
rake angle on surface roughness.
Author's personal copy
Int J Adv Manuf Technol (2015) 77:369–381 379

Table 6 Optimized process parameter predicted by GA

Sl. no A/spindle speed B/feed rate C/axial depth D/radial E/radial rake Confirmatory test % Error
(RPM) (mm/rev) (mm) depth (mm) angle (°)
Optimum value Experimental
(GA) (μm) value (μm)

1 2266 0.10 0.9 0.8 21 0.3231 0.3224 0.2


2 2296 0.10 0.9 0.9 18 0.3707 0.3710 −0.08
3 2058 0.09 0.8 0.7 18 0.2536 0.2533 0.1
4 2821 0.07 0.8 0.5 21 0.3625 0.3620 0.13

4 Evaluation of GA optimization results followed to recommend the correct combination of the para-
meters for the best optimal result. By solving the optimization
The selection of cutting parameters for optimization mini- problem, GA predicted the optimum roughness as 0.25 μm for
mizes the surface roughness. In this present study, an effort the machining of Aluminum 6063 in the selected cutting
has been made to determine the optimum values of cutting condition range. The GA-predicted surface roughness value
parameters to obtain the best possible surface quality within (best fitness function) is expected to be lower than the mini-
the specified range. The minimization of surface roughness by mum (smallest) Ra value of the experimental and regression
using GA can be expressed as models.

Minimize : Ra ðA; B; C; D; γ Þ ð8Þ 4.1 Validation of the model

Within ranges of cutting parameters, Table 3 shows that a regression model developed using
CCD by the RSM of DoE. Table 6 shows the compar-
1500 RPM≤A≤3500 RPM ison of predicted vs experimental value of surface
0.04 mm/rev≤B≤0.12 mm/rev roughness.
0.5 mm≤C≤0.9 mm The GA-predicted optimum conditions were further vali-
0.5 mm≤D≤0.9 mm dated with physical measurements. The percentage of error is
12°≤γ≤24° found to be within ±2 % which shows the validity of the
model. The experimental results of surface roughness with
To obtain the best optimal results, the major consideration the optimum cutting parameters (as predicted by GA) show
must be given to the number of the initial population size, the good agreement.
type of selection function, the crossover rate, and the mutation Figure 17 shows the performance of fitness value with
rate (Table 5). The author suggested [28] that no proper generation and the best individual performances of variables
guidelines are given by the researchers, which could be in coded form.

Fig. 17 The performance of fitness value with generation and the best individual performances of variables in coded form
Author's personal copy
380 Int J Adv Manuf Technol (2015) 77:369–381

5 Conclusion 6. Quintana G, Ciurana J, Ribatallada J (2010) Surface roughness


generation and material removal rate in ball end milling operations.
Mater Manuf Process 25(6):386–398. doi:10.1080/
The central composite designs are employed for developing 15394450902996601
second-order mathematical models for predicting surface 7. Mansour A, Abdalla H (2002) Surface roughness model for end
roughness parameter of end milling of Aluminum 6063 using milling: a semi-free cutting carbon casehardening steel (EN 32) in
dry condition. J Mater Process Technol 124(1–2):183–191. doi:10.
HSS end mill cutter with varying radial rake angles and
1016/S0924-0136(02)00135-8
process parameters viz. spindle speed, feed rate, axial depth 8. Alauddin M, El Baradie MA, Hashmi MSJ (1997) Prediction of tool
of cut, and radial depth of cut. The two-stage effort, obtaining life in end milling by response surface methodology. J Mater Process
a surface roughness model by surface response methodology Technol 71(3):456–465. doi:10.1016/S0924-0136(97)00111-8
9. Chang H, Kim J, Kim IH, Jang DY, Han DC (2007) In-process
and optimization of this model by genetic algorithms, has
surface roughness prediction using displacement signals from spindle
resulted in a fairly useful method of obtaining process param- motion. Int J Mach Tools Manuf 47(6):1021–1026. doi:10.1016/j.
eters in order to attain the required surface quality. ijmachtools.2006.07.004
10. Coker SA, Shin YC (1996) In-process control of surface roughness
due to tool wear using a new ultrasonic system. Int J Mach Tools
– The CCD model developed by RSM using the Design
Manuf 36(3):411–422. doi:10.1016/0890-6955(95)00057-7
Expert package is able to provide accurately the predicted 11. Gologlu C, Sakarya N (2008) The effects of cutter path strategies on
values of surface roughness close to actual values found surface roughness of pocket milling of 1.2738 steel based on Taguchi
in the experiments. The equations are checked for their method. J Mater Process Technol; 206(1–3):7–15. doi:10.1016/j.
jmatprotec.2007.11.300
adequacy with a confidence level of 95 %.
12. Dhokia VG, Kumar S, Vichare P, Newman ST (2008) An intelligent
– Feed rate has the greatest influence on surface roughness. approach for the prediction of surface roughness in ball-end machin-
– An increase in either the feed or the axial depth of cut ing of polypropylene. Robot Comput Integr Manuf; 24(6):835–842.
increases the surface roughness, while an increase in the doi:10.1016/j.rcim.2008.03.019
spindle speed decreases the surface roughness. 13. Lou SJ, Chen JC (1999) In-process surface roughness recognition
(ISRR) system in end-milling operations. Int J Adv Manuf Technol
– The higher as well as lower radial rake angle produced 15(3):200–209. doi:10.1007/s001700050057
maximum surface roughness. 14. Tsai Y, Chen JC, Lou S (1999) An in-process surface recognition
– An issue which can also be highlighted is related to the system based on neural networks in end milling cutting operations.
radial rake angle of the cutting tool used in the actual Int J Mach Tools Manuf 39(4):583–605. doi:10.1016/S0890-
6955(98)00053-4
experiment. The best minimum value of Ra is recom- 15. Benardos PG, Vosniakos GC (2002) Prediction of surface roughness
mended by the GA. It is clear from this study that the in CNC face milling using neural networks and Taguchi’s design of
best minimum Ra values came from 15° to 18° of radial experiments. Robot Comput Integr Manuf 18(5–6):343–354. doi:10.
rake angle in cutting tool. 1016/S0736-5845(02)00005–4
16. Brecher C, Quintana G, Rudolf T, Ciurana J (2011) Use of NC kernel
– The GA recommends 0.25 μm as the best minimum data for surface roughness monitoring in milling operations. Inter J
predicted surface roughness value. The confirmatory test Adv Manuf, Technol 53(9–12):953–962
showing the predicted values was found to be in good 17. Chen JC, Lou MS (2000) Fuzzy-nets based approach using an
agreement with observed values. accelerometer for in-process surface roughness prediction system in
milling operations. J Comput Integr Manuf Syst 13(4):358–368. doi:
10.1080/095119200407714
18. Ali Y, Zhang L (1999) Surface roughness prediction of ground
components using a fuzzy logic approach. J Mater Process Technol
89–90:561–568. doi:10.1016/S0924-0136(99)00022–9
References 19. Zain AM, Haron H, Sharif S (2008). An overview of GA technique
for surface roughness optimization in milling process. In IEEE proc
inter symp on infor technol, IT Sim, 4: 1–6 doi: 10.1109/ITSIM.
1. Jang DY, Choi Y-G, Kim H-G, Hsiao A (1996) Study of the corre- 2008.4631925
lation between surface roughness and cutting vibrations to develop an 20. Brezocnik M, Kovacic M (2003) Integrated genetic programming
on-line roughness measuring technique in hard turning. Int J Mach and genetic algorithm approach to predict surface roughness. Mater
Tools Manuf 36(4):453–464. doi:10.1016/0890-6955(95)00074-7 Manuf Process 18(3):475–491. doi:10.1081/AMP-120022023
2. Ming-Yung W, Chang H -Y (2004) Experimental study of surface 21. Oktem H, Erzurumlu T, Erzincanli F (2006) Prediction of minimum
roughness in slot end milling AL2014-T6. Int J Mach Tools Manuf surface roughness in end milling mold parts using neutral network
44:51–57. doi:10.1016/j.ijmachtools.2003.08.011 and genetic algorithm. J Mater Des 27(9):735–744. doi:10.1016/j.
3. Arias ER, Mecanico (1983) Analysis of surface roughness for end matdes.2005.01.010
milling operations. M.S thesis, Texas Tech University, Lubbock, 22. Tansel IN, Ozcelik B, Bao WY, Chen P, Rincon D, Yang SY
Texas, United States (2006) Selection of optimal cutting conditions by using
4. Palanisamy P, Rajendran I, Shanmugasundaram S (2007) GONNS. Int J Mach Tools Manuf 46(1):26–35. doi:10.1016/
Optimization of machining parameters using genetic algorithm and j.ijmachtools.2005.04.012
experimental validation for end-milling operations. Int J Adv Manuf 23. Suresh PVS, VenkateswaraRao P, Deshmukh SG (2002) A genetic
Technol 32:644–655. doi:10.1007/s00170-005-0384-3 algorithmic approach for optimization of surface roughness predic-
5. Martellotti ME (1941) An analysis of the milling process. Trans tion model. Int J Mach Tools Manuf 42(6):675–680. doi:10.1016/
ASME 63:667 S0890-6955(02)00005-6
Author's personal copy
Int J Adv Manuf Technol (2015) 77:369–381 381

24. Brezonick M, Kovavic M, Ficko M (2004) Prediction of surface and surface quality in end milling. Int J Mach Tools Manuf 46:189–
roughness with genetic programming. J Mater Process Technol 198. doi:10.1016/j.ijmachtools.2005.04.008
157–158:28–36. doi:10.1016/j.jmatprotec.2004.09.004 34. Suresh Kumar Reddy N, Venkateswara Rao P (2007) A genetic
25. Colak O, Kurbanoglu C, Kayacan MC (2007) Milling surface rough- algorithmic approach for optimization of surface roughness predic-
ness prediction using evolutionary programming methods. J Mater tion model in dry milling. Mach Sci Technol Int J 9(1):63–84. doi:10.
Des 28(2):657–666. doi:10.1016/j.matdes.2005.07.004 1081/MST-200051263
26. Oktem H, Erzurumlu T, Kurtaran H (2005) Application of response 35. Yesilyurt I (2006) End mill breakage detection using mean frequency
surface methodology in the optimization of cutting conditions for analysis of scalogram. Int J Mach Tools Manuf 46(3–4):450–458.
surface roughness. J Mater Process Technol 170(1–2):11–16. doi:10. doi:10.1016/j.ijmachtools.2005.03.014
1016/j.jmatprotec.2005.04.096 36. Goldberg DE (1989) Genetic algorithms in search, optimization, and
27. Bissey S, Poulachon G, Lapujoulade F (2007) Modelling of tool machine learning. Addison-Wesley Longman Publishing Co, Inc,
geometry in prediction of cutting force during milling of hard materials. Boston
Mach Sci Technol Int J 9(1):101–115. doi:10.1081/MST-200051376 37. Bernardos PG, Vosniakos GC (2003) Predicting surface roughness in
28. Zain AM, Haron H, Sharif S (2010) Application of GA to optimize machining: a review. Int J Mach Tools Manuf 43(8):833–844. doi:10.
cutting conditions for minimizing surface roughness in end milling 1016/S0890-6955(03)00059-2
machining process. Expert Syst Appl 37(6):4650–4659. doi:10.1016/ 38. Montgomery DC (2001) Design and analysis of experiments. Wiley,
j.eswa.2009.12.043 New York
29. AzlanMohdZain HH, Sharif S (2011) Integration of simulated an- 39. Cochran WG, Cox GM (1957) Experimental designs. Wiley, New
nealing and genetic algorithm to estimate optimal solutions for min- York
imizing surface roughness in end milling Ti-6AL-4 V. Int J Comput 40. Sivasakthivel PS, VelMurugan V, Sudhakaran R (2011) Prediction of
Int Manuf 24(6):574–592. doi:10.1080/0951192X.2011.566629 vibration amplitude from machining parameters by response surface
30. Kaneeda (1991) TCFRP cutting mechanism. Trans N Am Manuf Res methodology in end milling. Int J Adv Manuf Technol 53(5–8):453–
Inst SME 19:216–221 461. doi:10.1007/s00170-010-2872-3
31. Kim Y-H, Ko S-L (2002) Development of design and manufacturing 41. Sivasakthivel PS, Sudhakaran R (2013) Optimization of machining
technology for end mill in machining hardened steel. J Mater Process parameters on temperature rise in end milling of Al 6063 using
Technol 130–131:653–661. doi:10.1016/S0924-0136(02)00728-8 response surface methodology and genetic algorithm. Int J Adv
32. Suresh Kumar Reddy N, Venkateswara Rao P (2005) Selection of Manuf Technol 67(9–12):2313–2323. doi:10.1007/s00170-012-
optimum tool geometry and cutting conditions using a surface rough- 4652-8
ness prediction model for end milling. Int J Adv Manuf Technol 42. Hindustan Machine Tools (HMT) (2001) Production technology,
26(11–12):1202–1210. doi:10.1007/s00170-004-2110-y Tata McGraw-Hill Education
33. Suresh Kumar Reddy N, Venkateswara Rao P (2006) Experimental 43. Kannan T, Murugan N (2006) Prediction of ferrite number of duplex
investigation to study the effect of solid lubricants on cutting forces stainless steel clad metals using RSM. Weld J (AWS) 85(5):91–100

You might also like