See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/325235724

Machining Parameters Optimization of AA6061 Using Response Surface

Methodology and Particle Swarm Optimization

Article  in  International Journal of Precision Engineering and Manufacturing · May 2018

DOI: 10.1007/s12541-018-0083-2

CITATIONS READS

2 115

5 authors, including:

Rashmi Malghan Karthik Rao M C

National Institute of Technology Karnataka Manipal Academy of Higher Education
8 PUBLICATIONS   21 CITATIONS    7 PUBLICATIONS   21 CITATIONS   

SEE PROFILE SEE PROFILE

Mervin A. Herbert
National Institute of Technology Karnataka
90 PUBLICATIONS   563 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Optimization View project

Ni-Ti shape memory alloys View project

All content following this page was uploaded by Rashmi Malghan on 01 June 2018.

The user has requested enhancement of the downloaded file.

 INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 19, No. 5, pp. 695-704 MAY 2018 / 695

REGULAR PAPER DOI: 10.1007/s12541-018-0083-2

ISSN 2234-7593 (Print) / 2005-4602 (Online)

Machining Parameters Optimization of AA6061 Using

Response Surface Methodology and Particle Swarm
Optimization

Rashmi LMalghan1, Karthik Rao M C2,#, ArunKumar S3, Shrikantha S Rao1, and Mervin A Herbert1
1 Department of Mechanical Engineering, NITK, Surathkal 575025, India
2 Department of Mechanical Engineering, Debre Markos University, Post box-269, Ethiopia
3 Department of Mechatronics Engineering, Manipal Institute of Technology, Manipal Academy of Higher Education, India
# Corresponding Author / E-mail: [email protected], TEL: +91-9986087148,+251970771639
ORCID: 0000-0002-3082-9309

KEYWORDS: Face milling, Regression, Particle swarm optimization, Desirability, Response surface methodology, Design of experiment

The influence of cutting parameters on the responses in face milling has been examined. Spindle speed, feed rate and depth of cut
have been considered as the influential factors. In accordance with the design of experiments (DOE) a series of experiments have
been carried out. The paper exemplifies on the optimizing the process parameters in milling through the application of Response
surface methodology (RSM), RSM-based Particle Swarm Optimization (PSO) technique and Desirability approach. These aforesaid
techniques have been applied to experimentally establish data of AA6061 aluminium material to study the effect of process parameters
on the responses such as cutting force, surface roughness and power consumption. By adopting the multiple regression techniques,
the interaction between the process parameters are acquired. The optimal parameters have been found by adopting the multi-response
optimization techniques, i.e. desirability approach and PSO. The performance capability of PSO and desirability approach is
investigated and found that the values obtained from PSO are comparable with the values of desirability approach.

Manuscript received: March 15, 2017 / Revised: December 15, 2017 / Accepted: February 4, 2018

NOMENCLATURE R-sq (adj) = Adj R-Squared

RGA = Real Parameters Genetic Algorithm
AMMC's = Aluminium Matrix Composites RPD = Relative percentage deviation
ANFIS = Adaptive Fuzzy Interface System HGA = Hybrid Genetic Algorithm SQP
ANOVA = Analysis of variance SGA = Simple Genetic Algorithm
ANN = Artificial Neural Network SR = Surface roughness, µm
CCFCD = Central Composite Face Centred Design
CNC = Computer Numerical Control
DOE = Design of Experiment
FX = Cutting Force, N 1. Introduction
GA = Genetic Algorithm
gbest = Global best Present scenario, aluminium alloys has expanded their attention of
pbest = Particle best many industrialists, researchers, engineers and designers as promising
GSA = Genetic simulated annealing structural materials for aerospace applications or the automotive
MMC = Metal Matrix Composite
industry. Particularly, aluminium (6 series) alloys have been considered
PSO = Particle Swarm Optimization
widely because of their benefits such as medium strength, formability,
RSM = Response Surface Methodology
weldability, corrosion resistance, and low cost, comparing to other
R-sq = Pre R-squared
aluminium alloys.1 Mithilesh2 combined RSM and teaching learning

© KSPE and Springer 2018

 696 / MAY 2018 INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 19, No. 5

based algorithm to optimize surface roughness for milling aluminium The results indicated that the proposed model could efficiently describe
alloy Al2016-T6. Bhopalen3 applied the RSM to study the effect of the performance indicators within the boundary of the factors that are
cutting parameters and inclination angle on the surface roughness for being considered in the study. Gupta10 anticipated a Hybrid Genetic
milling Inconel 718. A considerable number of studies have researched Algorithm (HGA) to optimize the non-productive tool path in which
the effects of the spindle speed (A), feed rate (B) and depth of cut (C) the initial seed solution is generated by heuristic and collectively with
on responses, i.e. cutting force (FX), surface roughness (SR) and power an initial solution created by a simple genetic algorithm (SGA). The
consumption. In recent research Tandon4,5 and Conceicao6 the models results were analyzed by using Relative percentage deviation (RPD)
have been developed and identified the outcome of some factors on the and it was derived that the HGA is more superior compared to SGA for
SR and FX. A central task in science and engineering practice is to a same computation time limit (stopping norm). Basker11 incorporated
develop models that give an adequate description of the physical Tabu search, GA, Ant Colony Algorithm and Particle swarm optimization
systems being observed. The main goal of this study is to attain a algorithm for optimizing machining parameters in the milling operation.
mathematical model that relates the responses to the three cutting Mainly the work was concentrated on the development and utilization
parameters in face milling, precisely to the spindle speed, feed rate and of the mentioned optimization techniques and the optimization system
depth of cut. In this work two different approaches have been adopted which helps in identifying the optimum machining parameters for milling
to attain the mathematical models. The first approach is DOE together operation. Wang12 integrated Genetic simulated annealing (GSA) for
with analysis of variance (ANOVA) and regression analysis. The second identifying the optimal machining parameter in case of multi-pass
method is by means of the PSO technique. Generally, the machining milling. The comparative study was made and the results signified that
parameters are chosen based on the machine data hand book, trial and the GSA was effective over GA. Reddy13 implemented the mathematical
error method or by literature. But adopting such trials is not a precise model, based on the concept of Response Surface Methodology (RSM)
way for selecting the appropriate machining parameter as it leads to to establish the cutting conditions and effect of tool geometry related to
wastage of time and cost. Hence, to overcome the intricacy, it is machining performance and even optimize the surface roughness
necessary to develop a technique to predict the appropriate machining response GA technique was utilized and the respective optimal condition
parameters. In the present study desirability and PSO techniques are was determined. Savas14 used GA for optimization of surface roughness
incorporated to identify the optimal process parameters. and concluded that surface roughness increases with the increase of the
machining parameters feed rate and depth of cut. Abburi15 used RGA
(real-parameters Genetic Algorithm) in turning operation for optimal
2. Literature Survey process parameters, thus the concept leading in minimal product time,
which in turn acts as the base for the SQP (Sequential Quadratic
Though a lot of research in milling operation has been attempted by Programming) code and results in an increase in the performance. The
a few researchers and papers relevant to optimization and the issues overall indication and suggestion were towards the usage of the
related to milling operation have been discussed. Tandon4 implemented numerical approach for attaining the optimum solution in case of
particle swarm optimization for optimizing multiple machining unequal depth of cut involvement. Mukherjee16 addressed the proposed
parameters and results indicated in a reduction in machining time by 35%. a generic framework in metal cutting processes in order to opt and attain
Tandon5 incorporated the ANN approach to develop a comprehensive an efficient approach and showed a path to identify the optimal cutting
model for critical machining parameters. The developed model was conditions or near optimal cutting conditions in various categories of
tested and validated for specific pocket milling scenario originated in metal cutting process. Onwubolu17 proposed an optimization concept
the industry. From the testing and validation, it was concluded that there based on Tribes for determination of the machining parameters in multi-
was an excellent agreement between the experimental and simulated pass plain milling and face milling operations. The machining parameters
forces. Jinhua7 developed integrated multi objective optimization are decided based on the strategy of maximizing and minimize the
technique and applied to Inconel 718 in the milling process for to attain production rate and derived that developed Tribes based approach is
minimum surface roughness and maximum compressive residual stress. efficient. Developed a model for predicting the surface roughness in
Conceicao6 developed model for the multi-pass cutting parameter in face face milling for aluminium material by adopting the PSO technique.
milling by incorporating the genetic search. Furthermore, the novel The conclusion was made that PSO is in good agreement with current
approach based on substituting the depth of cut with a sequence of surface roughness values. Ship-peng19 developed an adaptive fuzzy
depths of cut was developed. The performance of the developed model interface system (AFIS) to predict the surface roughness in the milling
was compared with rest of multi-pass models. Saffari8 used a Genetic process and proved that greater prediction (nearly of 96%) accuracy
Algorithm (GA) to obtain a minimal tool deflection in the milling was achieved. Julie20 implemented Taguchi approach to optimize the
process. In the study, the tool deflection was considered as the objective surface roughness in the milling operation. The authors analyzed the
while tool-life and surface roughness were the constraints. A comparative experiments using analysis of variance (ANOVA) and concluded that the
study was made to validate the performance of the optimization, the Taguchi approach was successful in optimizing the surface roughness.
attained results illustrate that optimized parameters are proficient of Xain21 considered and specified the significance of Artificial Neural
machining the workpiece more precisely with better surface finish. Network (ANN) for predicting the surface roughness in milling operation
Patwari9 illustrated mathematically the effect of machining parameters and concluded a better surface finish is achieved at a high rake angle,
on response surface roughness in the milling of Medium Carbon Steel low feed rate and high speed. Astilturk22 developed full factorial design
using the TiN tool in dry condition and included a genetic algorithm. of experimented (FFD). The FFD was adapted to reliability during
 INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 19, No. 5 MAY 2018 / 697

turning operation and authors concluded that the ANN model yielded
better results as compared to the statistical model and even the results
indicated that by varying the number of layers and nodes in the hidden
layer the prediction accuracy of respective response can be increased.
Benardos23 developed an ANN model for predicting the surface
roughness in the milling process and indicated that the ANN model was
able to predict the surface roughness with a mean error of 1.86%.
A sufficient amount of research articles is available on identifying
the optimal machining parameters by incorporating the soft computing Fig. 1 AA6061 material with the tool
techniques precisely fuzzy logic, GA, and ANN. A sufficient amount
of research articles is available for identifying the optimal machining
parameters by incorporating the soft computing techniques precisely
Fuzzy logic, GA, and ANN. Therefore, this work explores the feasibility
for multiobjective optimization with the RSM-PSO method. And it is
employed to optimize cutting force, surface roughness and power
consumption for milling AA6061. The paper gives the information
regarding conducting the experiments, performing RSM technique to
attain the response equations. Further, these response equations are
used as fitness function in PSO. The experimental results are compared
with RSM predicted results and later on the desirability and PSO
results are compared.
Fig. 2 Data acquisition

3. Conditions of Experiment Table 1 Composition of AA6061

Element Al Si Cu Mg Cr
Test samples made up of AA6061 with dimensions: length of 100 Weight (%) 97.9 0.60 0.28 1.0 0.20
mm length, 60 mm breadth and 12 mm thickness. The face milling
experiments were carried out by a tool SDMT 1205PDR-HQ-M IC28 Table 2 The Cutting Parameters and their levels
as depicted in Fig. 1. The cutting tool has 5 inserts. The details of the Spindle Speed Feed Rate Depth of Cut
Symbol Levels
insert, it is square type insert, side clearance angle 15°, tolerance 0.08 (rpm) (mm/min) (mm)
mm, type T, cutting edge length 12 mm, thickness 5 mm, type of mount A L-I 1000 300 1
90°, lead 15°, the radius of nose 0.4 mm. The tool holder selected is B L-II 2000 400 2
C L-III 3000 500 3
having the BT 30 taper type (Tool Holder: F90SD D50, 2F2, 12)
produced by Iscar. The type of machine utilized for the milling test was
CNC Vertical Milling machine (Spark DTC 250) by AMS. The the central composite design (CCD). The main goal of this work is to
experiments were performed with dry run machining condition of the identify the mathematical models that describe the dependence of
selected material. The cutting force was calculated via indirect approach responses on the three machining parameters: spindle speed (A) rpm,
based on the current consumption by each axis. These current values feed rate (B) mm/min, depth of cut (C) mm. The CCD models the
were fetched through Ethernet cable provided by FANUC as depicted responses by using the empirical second-order polynomial equation. In
in Fig. 2. The SR values of finish - milled work pieces were measured the present study, the RSM technique was incorporated in the analysis
by the Mitutoyo Surface Roughness Tester. The SR values were acquired and design of the experiments. RSM technique helps with its strategies
at a minimum of three different locations, later on; the measured surface to overcome the analysis quandaries thus leading to a better result. It
roughness was obtained by averaging the surface roughness values. The usually identifies the significance of the process parameters on the
chemical composition of the selected material is depicted in Table 1. The responses and the main purpose of RSM is to optimize the response.
cutting parameters considered were spindle speed, feed rate, and depth The Central Composite Face Centered Design (CCFCD) was used to
of cut. The experiments were designed and conducted based on the design implement the response models using RSM. A total of 20 experiments
and analysis of the experiment. In the present study, Design of Experiment was performed which incorporates of 8 cube points, 6 centre points in
(DOE), Response Surface Method (RSM) three factors of cutting a cube, 6 Axial points, and the alpha value is 1. The range of the
parameters and three levels have been considered as shown in Table 2. process parameters was set by taking into consideration of the tool or
inserts specification and even by performing the trial experiments in
order to achieve the desired responses. RSM has been used for
4. Statistical and Optimization Concepts: Design of mathematical modeling of Fx, SR and power consumption, The next
Experiment, Desirability, and PSO step is, the experiments were performed using the multiple regression
equations in order to identify the interaction effect between the process
In the current study, the design of the experiment was realized using parameters and the responses. Later on, The Desirability and PSO
 698 / MAY 2018 INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 19, No. 5

techniques were employed to determine the optimal parameters. The to renovate the overall best solution. The supplementing equations are
validation step is carried out by conducting the experiments in order to adopted to change the individual particle’s position to reach an overall
verify the established model. best possible solution in the search space.

n+1 n o o
4.1 Methodology on desirability and PSO km = w ⋅ km + c1 ⋅ Q1 ⋅ ( pbestm – Rm ) + c2 ⋅ Q2 ⋅ ( gbest – Rm ) (1)
The desirability Function approach is a multiple-response optimization
n
method. This approach was first introduced in 1980 by Suich and Where: km = ‘mth’ particle momentum at ‘nth’ iteration, w = inertia
Deringer. The method finds operating conditions “targeted” that the weight, c1, c2 = learning rates, Q1, Q2 = random numbers between 0-
most desirable response value. The general approach first converts each 1, pbestm = pbest position of mth particle, gbest = gbest position of
o o o o
response x1 into an individual desirability function di that varies over swarm, Rm = [ Rm1 , Rm2 , …, RmN ], ‘mth’ particle current position at
the range 0 < di < 1.The desirability functions are categorized into three ‘oth’ iteration in N-dimensional search space.
sectors based on the response characteristics. Once the momentum is calculated, the next position of ‘mth’ particle
1. If the target for the response is a maximum value / “Higher is is calculated using the following Eq. (2).
Better”.
o+1 o n+1
R m = R m + km (2)
⎧ 0 ri ≤ ri*
⎪
⎪ ri – ri*
- a
⎨ ---------------- ri* < ri < ri ′ Inertia weight can be selected any random value or it can be
⎪ ri ′ – ri* determined by opting the following Eq. (3).
⎪ 1 ri ≥ ri ′
⎩
[( wmax – wmin ) × iterationcurrent ]
Where: ri* is the minimum adequate value of ri, ri' is the maximum W = wmax – -----------------------------------------------------------------------
- (3)
iterationtotal
adequate value of ri and a describes
2. If the target for the response is a minimum value / “Smaller is Where: wmax = upper limit inertia weight, wmin = lower limit inertia
better” weight, iterationcurrent = current iteration, iterationtotal = total number of
iteration.
⎧ 1 ri ≤ ri″
⎪
⎪ ri* – ri
------------------ b
⎨ ri* ri″ < ri < ri*
⎪ – ri″
⎪ 5. Results and Discussion
⎩ 0 ri ≥ ri*

Where: ri" is the minimum value of ri, ri* is the maximum adequate Response surface methodology (RSM) is a collection of the
value of ri mathematical and statistical technique used for analyzing problems in
which several independent variables influence a dependent variable or
4.2 Concept of PSO response and the goal is to optimize the response. Second-order
PSO was developed by Eberhart and Kennedy in the year 1995. polynomial equation with interaction terms was fitted to the experimental
This approach is an evolutionary computational method that has been results to develop a mathematical model, which will help to predict the
based on the swarm intelligence of a flock of birds. One difference from extraction efficiency of different sets of combinations of four process
other evolutionary algorithm is that the particle does not use selection variables on the responses. Three empirical models were developed from
criterion in the iterative procedure. Therefore, population members will this study to predict the cutting force, surface roughness and power
survive from the beginning to the end in the optimization process. This consumption. The attained regression equations are depicted in Table 3.
algorithm has been widely successfully applied to solve many engineering The considered experiments with their respective input parameters and
problems,24-27 by imitating the seeking behavior of a swarm of birds, outcome of the responses are shown in Table 4. The experimental V/
the individual particles in an algorithm look for the overall best result S predicted values for all the considered experiments are represented in
of the fitness function. The particles get updated with their momentum Table 4.
and act as per the situation and particles usually update their momentum Generally, ANOVA comprises the sum of square, the degree of
based on the gained previous experience and global effort put up by freedom (DF), mean square, F-value, and P-value.38 The inclusion of
other particles in the search space. There are three constraints, social, ANOVA is essential to find out the influential parameters on the
cognitive and inertia that are responsible for the updating of the responses.39,40 The performance of the model was validated with
momentum of the particles.34-37 The social constraint is accountable to ANOVA. From the Tables 5-7, it can be observed that spindle speed
move faster the particle to the best position followed by another swarm has a major contribution to all the responses followed by the feed rate
so far, known as the gbest position. The cognitive factor rushes the and depth of cut. As depicted in Table 5 spindle speed has 91.183%
particle towards its individual best location proficient till then, known contribution on the response FX, 6.299% of the feed rate and followed
as the pbest. The inertia factor is used to maintain the stability between by the depth of cut with a 0.89267% contribution. The remaining
the current and overall investigation capabilities among the search space. contribution is contributed by the interaction of the process parameters.
The gbest position is determined to be changed on progressive era. In Similarly, In Table 6, it can be identified that spindle speed plays a vital
the progressive iteration if it is found that the pbest position is better role with the contribution of 40.526%, 14.210% of feed rate and
than gbest position, then the pbest position will be traded by the gbest 22.1053% of the depth of cut. The rest of the contributions are by the
 INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 19, No. 5 MAY 2018 / 699

Table 3 Regression Equations

SL. No Responses Regression Equations
1 FX -81.78776 + 0.081353*A + 0.18242*B + 4.42441*C - 3.24420E-005*A*B - 5.91517E-006*A*A
0.15850 - 2.80000E-005*A + 2.22000E-003*B + 0.13500*C - 2.50000E - 007*A*B + 2.00000E-005*A*C - 6.00000E-
2 SR
004*B*C
Power -0.030842 + 4.72657E-005*A - 2.33913E-004*B + 6.10045E-003*C + 1.15949E-007*A*B -8.91940E-009*A*A +
3
Consumption 6.31394E-007*B*B

Table 4 Experimental V/S RSM predicted results

Experimental RSM Prediction Error (%)
SL Spindle Depth Power FX SR Power FX SR Power
Feed rate FX SR
No. speed of cut consumption predicted predicted consumption error error consumption
(mm/min) (N) (μm)
(rpm) (mm) (kW) (N) (μm) predicted (kW) (%) (%) error (%)
1 1000 300 1 45.49 0.7 0.037 43.068 0.697 0.035 5.325 0.500 6.438
2 3000 300 1 143.3 0.52 0.132 138.987 0.531 0.128 3.003 -2.019 3.184
3 1000 500 1 73.26 0.96 0.108 73.063 0.971 0.112 0.268 -1.094 -4.035
4 3000 500 1 158.3 0.71 0.249 156.006 0.705 0.252 1.443 0.775 -0.902
5 1000 300 3 56.69 0.63 0.048 51.971 0.647 0.047 8.420 -2.619 2.037
6 3000 300 3 150.5 0.56 0.139 147.836 0.561 0.140 1.777 -0.089 -0.976
7 1000 500 3 83.56 0.68 0.125 81.912 0.681 0.125 1.972 -0.074 0.645
8 3000 500 3 164.2 0.48 0.262 164.855 0.495 0.264 -0.399 -3.021 -0.705
9 1000 400 2 63.42 0.76 0.074 64.290 0.749 0.074 1.466 1.513 0.219
10 3000 400 2 153.3 0.58 0.191 151.921 0.573 0.189 0.893 1.293 0.655
11 2000 300 2 100 0.61 0.090 101.367 0.609 0.096 -1.357 0.246 -6.919
12 2000 500 2 134.2 0.71 0.205 124.874 0.713 0.197 6.963 -0.352 3.737
13 2000 400 1 110.2 0.73 0.134 108.696 0.726 0.134 1.373 0.616 -0.297
14 2000 400 3 119.8 0.62 0.148 117.545 0.596 0.147 1.907 3.952 0.685
15 2000 400 2 115 0.66 0.141 113.121 0.661 0.140 1.651 -0.076 0.224
16 2000 400 2 115 0.66 0.141 113.121 0.661 0.140 1.651 -0.076 0.224
17 2000 400 2 114.8 0.65 0.139 113.121 0.661 0.140 1.480 -1.615 -1.388
18 2000 400 2 113 0.67 0.143 113.121 0.661 0.140 -0.089 1.418 1.622
19 2000 400 2 115 0.66 0.141 113.121 0.661 0.140 1.651 -0.076 0.224
20 2000 400 2 114 0.66 0.137 113.121 0.661 0.140 0.789 -0.076 -2.448

Fig. 3 Normal Probability plots of residuals for (a) FX, (b) SR, (c) Fig. 4 Plot of residuals v/s predicted: (a) FX, (b) SR, and (c) power
power Consumption for AA6061 consumption

interactions of the process parameters. As depicted in Table 7, it can be rate of 40.983% and followed by the depth of cut 0.6101%. The other
concluded that spindle speed has 55.737% major contribution, the feed contributions are contributed by the interactions occurred among the
 700 / MAY 2018 INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 19, No. 5

Table 5 ANOVA analysis for cutting forces (FX)

Sources Sum of squares DF Mean square F-value p-value Prob > F Remarks P (%)
Model 21831.51 5 4366.3 630.33 < 0.0001
A-A 19995.16 1 19995.2 2886.56 < 0.0001 significant 91.1835
B-B 1381.45 1 1381.45 199.43 < 0.0001 significant 6.2998
C-C 195.75 1 195.75 28.26 0.0001 significant 0.89267
AB 84.2 1 84.2 12.16 0.0036 significant 0.00384
A2 174.95 1 174.95 25.26 0.0002 significant 0.79782
Residual 96.98 14 6.93
Lack of fit 93.64 9 10.4 15.61 0.0037
Pure error 3.33 5 0.67
Cor total 21928.49 19

Table 6 ANOVA analysis for surface roughness (SR)

Sources Sum of squares DF Mean square F-value p-value Prob > F Remarks P (%)
Model 0.18 6 0.031 225.54 < 0.0001
A-A 0.077 1 0.077 570.38 < 0.0001 significant 40.5263
B-B 0.027 1 0.027 199.16 < 0.0001 significant 14.2105
C-C 0.042 1 0.042 311.19 < 0.0001 significant 22.1053
AB 5.00E-03 1 5.00E-03 36.83 < 0.0001 significant 2.63158
AC 3.20E-03 1 3.20E-03 23.57 0.0003 significant 1.68421
BC 0.029 1 0.029 212.12 < 0.0001 significant 15.2632
Residual 177E-03 13 1.36E-04
Lack of fit 1.57E-03 8 1.96E-04 4.89 0.0485
Pure error 2.00E-04 5 4.00E-05
Cor total 0.19 19

Table 7 ANOVA analysis for Power Consumption

Sources Sum of squares DF Mean square F-value p-value Prob > F Remarks P (%)
Model 0.061 6 0.01 749.4 < 0.0001
A-A 0.034 1 0.034 2492.41 < 0.0001 significant 55.7377
B-B 0.025 1 0.025 1877.39 < 0.0001 significant 40.9836
C-C 3.72E-04 1 3.72E-04 27.6 0.0002 significant 0.61016
AB 1.08E-03 1 1.08E-03 79.78 < 0.0001 significant 1.76393
A2 2.55E-04 1 2.55E-04 18.88 0.0008 significant 0.41738
B2 1.28E-04 1 1.28E-04 9.46 0.0088 significant 0.20918
Residual 1.75E-04 13 1.35E-05
Lack of fit 1.55E-04 8 1.94E-05 4.87 0.0489
Pure error 1.99E-05 5 3.99E-06
Cor total 0.061 19

process parameters. From the Tables 5-7, in the remarks column, the increases this is due to vanishing of the formed built up edge at the
process parameters and interactions of process parameters are assigned cutting edge tip leading to better surface finish.41 On the other side as
significant, this assigning of the values is entirely based on the condition the feed rate increases the value of SR increases this is due to increase
of “Prob > f <5% or 0.05”.28 If the aforesaid condition is satisfied, then in axial movement of the cutting tool as it will not completely remove
the values are assigned as significant else not significant. The significance the required material because of alteration of the tool to a new position.
of the regression fitted model is determined by incorporating the R2 In the case of influence of depth of cut on SR the value of SR decreases
coefficient correlation. The attained R-sq and R-sq (adj) for the model as the depth of cut reduces this is due to the rigidity effect of the
is 99.65 and 99.33 respectively. From the Fig. 3, it can be observed that machine occurred during the machining process.
there is a normal distribution of errors takes place as residuals for all Similarly, as represented in Figs. 5(b) and 5(c) the influence of
the models falls on the straight line. Fig. 4 exhibits the plot of residuals spindle speed, feed rate and depth of cut on FX and power consumption
v/s predicted values of responses. From Fig 4, it can be observed that signify the directly proportional relationship, i.e., as the spindle speed,
all the points of the experimental runs were spotted randomly inside the feed rate and depth of cut increase the responses FX and power
reliable range of residuals over the graph. consumption increases. This direct proportional relationship occurs due
to the higher chip tool interface area.42 From the Fig. 6, it can be
5.1 Effect of Process Parameters on responses SR, FX and Power recognized that the optimal values attained for all the responses by the
consumption influence of process parameters through the desirability approach are as
Fig. 5(a) represents that the SR value decreases as the spindle speed follows: spindle speed (3000 rpm), feed rate (500 mm/min) and depth
 INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 19, No. 5 MAY 2018 / 701

Fig. 6 Results of Desirability Approach: (a) All power consumption

Fig. 5 Effect of process parameters on (a) SR (b) FX (c) Power Responses-Desirability Approach

Table 8 RSM Experimented V/S Predicted (Validation Data sets)

Spindle Depth Power FX SR Power FX SR Power
Test Feed rate FX SR
speed of cut consumption predicted predicted consumption error error consumption
No. (mm/min) (N) (μm)
(rpm) (mm) (kW) (N) (μm) predicted (kW) (%) (%) error (%)
1 1200 340 1.2 47.62 0.69 0.038 48.41 0.72 0.06 -1.67 -4.88 -5.26
2 1800 340 1.8 103.28 0.57 0.094 98.61 0.6 0.1 4.52 -5.26 1.1
3 2400 340 2.6 108.44 0.59 0.103 113.44 0.59 0.14 -1.61 0.42 -0.97
4 1200 390 1.2 59.96 0.88 0.082 55.59 0.82 0.08 7.29 6.4 -3.66
5 1800 390 1.8 114.38 0.67 0.14 104.82 0.69 0.12 8.36 -2.24 7.14
6 2400 390 2.6 120.23 0.63 0.148 128.85 0.59 0.16 -7.17 6.27 -1.35
7 1200 460 1.2 74.26 0.97 0.109 69.63 0.93 0.11 6.23 4.36 -0.92
8 1800 460 1.8 136.32 0.72 0.21 139.5 0.73 0.16 -2.33 -1.85 -4.76
9 2400 460 2.6 140.87 0.61 0.23 134.99 0.59 0.2 4.17 2.51 4.35
10 2800 340 1.2 144.22 0.51 0.14 140.07 0.51 0.14 2.88 -0.59 0.71
11 2800 390 1.8 152.93 0.6 0.18 146.31 0.6 0.17 4.33 0.75 0.56
12 2800 460 2.6 163.96 0.49 0.28 164.13 0.5 0.22 -0.11 -2.45 3.57
13 1200 340 1.8 52.27 0.65 0.039 51.07 0.7 0.06 2.3 -7.18 -2.56
14 1200 390 1.2 57.86 0.87 0.069 55.59 0.89 0.08 3.93 -2.72 -1.45
15 2800 460 1.8 160.02 0.58 0.27 149.72 0.58 0.06 6.44 0.59 3.7

of cut (3 mm) on the responses 0.4945 (m), 166.85 (N) and 0.263 characteristics of PSO and indicates the best operating parameters
(kW). recommended for milling process of AA6061. These parameters play
a significant role in obtaining good convergence characteristics of PSO.
5.2 Desirability approach If the number of parameters increases, the learning rate increases. In turn,
Table 8 represents the experiments carried out by using the RSM the number of iterations increases in the search space. The outcome
technique for validation purpose and also it can be observed that the % leads to a probability of getting a global optimum solution and leading
of error attained between the experimental and predicted values carried the convergence to be accomplished in a smaller number of iterations.
out by using RSM technique. The RSM technique to predict individual Table 10 summarizes the optimal parameters attained through adopting
response it needs an individual equation, thus at a single go only one different techniques. The deviations between the PSO predicted results
response from its respective equation can be predicted, thus the multi and experimental results are marginal. However, the experimental error
objective optimization technique is suitable if more than one response is quite considerable and lies within the acceptable range of ±5%.
is incorporated. Table 9 clearly indicates projected model and signifies Moreover, the comparative results presented in Table 10 have indicated
the parameters that play a vital role in obtaining finer convergence that PSO predicted values have good agreement with the experimental
 702 / MAY 2018 INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 19, No. 5

results. It can be concluded that PSO technique gives fairly accurate with the PSO yields better results compared to that of desirability
values as compared to that of the desirability approach and thus PSO and are in good concurrence with the experimental results.
has a better computational efficiency.

5.3 Numerical elucidation of PSO ACKNOWLEDGEMENT

Table 9 Parameters of PSO
The authors would like to sincerely thank the Department of
Number of parameters 3
Mechanical Engineering, National Institute of Technology Karnataka,
Number of particles 100
Surathkal, for providing research facilities.
Number of iterations 100
Learning rate
C1 max = C2 max 1.4
C1 min = C2 min 1.8 REFERENCES
C1=C2=Cmin+R*(Cmax-Cmin) Where R = total iterations
Xulim [1000, 300, 1] 1. Demir, H., and Gündüz, S., “The Effects of Aging on Machinability
Xllim [3000, 500, 3] of 6061 Aluminium Alloy,” Materials & Design, Vol. 30, No. 5, pp.
1480-1483, 2009.
Table 10 Optimal process parameters
Spindle Depth 2. Dikshit, M. K., Puri, A. B., and Maity, A., “Empirical Modelling of
Feed rate SR FX Power
Parameters speed of cut Dynamic Forces and Parameter Optimization Using Teaching-
(mm/min) (μm) (N) (kW)
(rpm) (mm) Learning-Based Optimization Algorithm and RSM in High Speed
Desirability 3000 500 3 0.520 167.86 0.264 Ball-End Milling,” Journal of Production Engineering, Vol. 19, No.
PSO 3000 500 3 0.494 166.85 0.265 1, pp. 11-21, 2016.
Experimental 3000 500 3 0.480 164.20 0.262
3. Bhopale, N. N., Joshi, S. S., and Pawade, R. S., “Experimental
Investigation into the Effect of Ball End Milling Parameters on
6. Conclusions Surface Integrity of Inconel 718,” Journal of Materials Engineering
and Performance, Vol. 24, No. 2, pp. 986-998, 2015.
The optimization issue of machining parameters is dealt with as a
multi-objective optimization problem. Keeping in mind the end goal to 4. Tandon, V., El-Mounayri, H., and Kishawy, H., “NC End Milling
influence manufacturing engineers to have more control on the machining Optimization Using Evolutionary Computation,” International Journal
operations, the optimization strategy is to simultaneously reduce of Machine Tools and Manufacture, Vol. 42, No. 5, pp. 595-605,
production time and decrease production cost and enhance profitability 2002.
in the interim subject of fulfilling the imperatives on the spindle speed, 5. Tandon, V. and El-Mounayri, H., “A Novel Artificial Neural Networks
feedrate, depth of cut, cutting force, surface roughness and power Force Model for End Milling,” The International Journal of Advanced
consumption. A proficient multi-objective PSO to optimize the Manufacturing Technology, Vol. 18, No. 10, pp. 693-700, 2001.
machining parameters is developed to solve such multi-objective
optimization problem in optimization of face milling. The proposed 6. Conceicao A. C. A., Castro, C., and Davim, J., “Optimisation of
PSO algorithm does not have any difficulty in accomplishing optimal Multi-Pass Cutting Parameters in Face-Milling Based on Genetic
solutions with good convergence for multi-objective optimization Search,” The International Journal of Advanced Manufacturing
problems and the significant upgrades are made in contrast with the Technology, Vol. 44, Nos. 11-12, pp. 1106-1115, 2009.
outcomes by desirability approach. 7. Zhou, J., Ren, J., and Yao, C., “Multi-Objective Optimization of
Based on the experimental studies the following findings are found: Multi-Axis Ball-End Milling Inconel 718 Via Grey Relational
• The developed second order equation has shown good co-relation
Analysis Coupled with RBF Neural Network and PSO Algorithm,”
between the predicted and experimental values.
Measurement, Vol. 102, pp. 271-285, 2017.
• The results of ANOVA has revealed that the effect of spindle
speed is much more pronounced than the effects of feed rate and 8. Saffar, R. J., Razfar, M., Salimi, A., and Khani, M., “Optimization of
depth of cut, on the surface roughness, power consumption and Machining Parameters to Minimize Tool Deflection in the End Milling
cutting force. The validation test of RSM model has shown ±7% Operation Using Genetic Algorithm,” World Applied Sciences Journal,
of error. Vol. 6, No. 1, pp. 64-69, 2009.
• The results procured through PSO are likewise compared with the
9. Patwari, M. A. U., Amin, A., and Arif, M. D., “Optimization of
customary desirability approach and it was found that PSO
Surface Roughness in End Milling of Medium Carbon Steel by
technique displays extensive favorable position contrasted with
Coupled Statistical Approach with Genetic Algorithm,” Special
outcomes accomplished with the desirability approach.
Issue of the International Journal of the Computer, the Internet and
• Test data set demonstrate that the anticipated PSO model results
Management, Vol. 19, No. SP1, pp. 41.1-41.5, 2011.
matches well with experimental results. The verification experiments
show that the optimal level combination of machining parameters 10. Gupta, A. K., Chandna, P., and Tandon, P., “Hybrid Genetic Algorithm
 INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 19, No. 5 MAY 2018 / 703

for Minimizing Non Productive Machining Time during 2.5 D and Multiple Regression Method,” Expert Systems with Applications,
Milling,” International Journal of Engineering, Science and Vol. 38, No. 5, pp. 5826-5832, 2011.
Technology, Vol. 3, No. 1, pp. 183-190, 2011.
23. Benardos, P. G. and Vosniakos, G. C., “Prediction of Surface
11. Baskar, N., Asokan, P., Prabhaharan, G., and Saravanan, R., Roughness in CNC Face Milling Using Neural Networks and
“Optimization of Machining Parameters for Milling Operations Taguchi's Design of Experiments,” Robotics and Computer-Integrated
Using Non-Conventional Methods,” The International Journal of Manufacturing, Vol. 18, Nos. 5-6, pp. 343-354, 2002.
Advanced Manufacturing Technology, Vol. 25, Nos. 11-12, pp.
24. Malghan, R. L., Rao, K. M., Shettigar, A. K., Rao, S. S., and
1078-1088, 2005.
D’Souza, R., “Application of Particle Swarm Optimization and
12. Wang, Z., Rahman, M., Wong, Y., and Sun, J., “Optimization of Response Surface Methodology for Machining Parameters
Multi-Pass Milling Using Parallel Genetic Algorithm and Parallel Optimization of Aluminium Matrix Composites in Milling
Genetic Simulated Annealing,” International Journal of Machine Operation,” Journal of the Brazilian Society of Mechanical Sciences
Tools and Manufacture, Vol. 45, No. 15, pp. 1726-1734, 2005. and Engineering, Vol. 39, No. 9, pp. 3541-3553, 2017.

13. Reddy, N. S. K. and Rao, P. V., “Selection of Optimum Tool Geometry 25. Yusup, N., Zain, A. M., and Hashim, S. Z. M., “Overview of PSO
and Cutting Conditions Using a Surface Roughness Prediction for Optimizing Process Parameters of Machining,” Procedia
Model for End Milling,” The International Journal of Advanced Engineering, Vol. 29, pp. 914-923, 2012.
Manufacturing Technology, Vol. 26, Nos. 11-12, pp. 1202-1210,
26. Zhou, J., Ren, J., Feng, Y., Tian, W., and Shi, K., “A Modified
2005.
Parallel-Sided Shear Zone Model for Determining Material
14. Savas, V. and Ozay, C., “The Optimization of the Surface Roughness Constitutive Law,” The International Journal of Advanced
in the Process of Tangential Turn-Milling Using Genetic Algorithm,” Manufacturing Technology, Vol. 91, Nos. 1-4, pp. 589-603, 2017.
The International Journal of Advanced Manufacturing Technology,
27. Rashid, M. F. F., Hutabarat, W., and Tiwari, A., “A Review on
Vol. 37, Nos. 3-4, pp. 335-340, 2008.
Assembly Sequence Planning and Assembly Line Balancing
15. Abburi, N. R. and Dixit, U. S., “Multi-Objective Optimization of Optimisation Using Soft Computing Approaches,” The International
Multipass Turning Processes,” The International Journal of Advanced Journal of Advanced Manufacturing Technology, Vol. 59, Nos. 1-4,
Manufacturing Technology, Vol. 32, Nos. 9-10, pp. 902-910, 2007. pp. 335-349, 2012.

16. Mukherjee, I. and Ray, P. K., “A Review of Optimization 28. Yu, C., Xu, T., and Liu, C., “Design of a Novel UWB Omnidirectional
Techniques in Metal Cutting Processes,” Computers & Industrial Antenna Using Particle Swarm Optimization,” International Journal
Engineering, Vol. 50, Nos. 1-2, pp. 15-34, 2006. of Antennas and Propagation, Vol. 2015, Article ID: 303195, 2015.

17. Onwubolu, G. C., “Performance-Based Optimization of Multi-Pass 29. Chandrasekaran, M., Muralidhar, M., Krishna, C. M., and Dixit, U.,
Face Milling Operations Using Tribes,” International Journal of “Application of Soft Computing Techniques in Machining
Machine Tools and Manufacture, Vol. 46, Nos. 7-8, pp. 717-727, Performance Prediction and Optimization: A Literature Review,” The
2006. International Journal of Advanced Manufacturing Technology, Vol.
46, Nos. 5-8, pp. 445-464, 2010.
18. Raja, S. B. and Baskar, N., “Application of Particle Swarm
Optimization Technique for Achieving Desired Milled Surface 30. Raja, S. B. and Baskar, N., “Application of Particle Swarm
Roughness in Minimum Machining Time,” Expert Systems with Optimization Technique for Achieving Desired Milled Surface
Applications, Vol. 39, No. 5, pp. 5982-5989, 2012. Roughness in Minimum Machining Time,” Expert Systems with
Applications, Vol. 39, No. 5, pp. 5982-5989, 2012.
19. Lo, S. P., “An Adaptive-Network Based Fuzzy Interference System
for Predicting of Work Piece Surface Roughness in End Milling,” 31. Rao, R. V. and Pawar, P. J., “Parameter Optimization of a Multi-Pass
Journal of Materials Processing Technology, Vol. 142, No. 3, pp. Milling Process Using Non-Traditional Optimization Algorithms,”
665–675, 2003. Applied Soft Computing, Vol. 10, No. 2, pp. 445-456, 2010.

20. Zhang, J. Z., Chen, J. C., and Kirby, E. D., “Surface Roughness 32. Yang, W.-A., Guo, Y., and Liao, W., “Multi-Objective Optimization
Optimization in an End-Milling Operation Using the Taguchi Design of Multi-Pass Face Milling Using Particle Swarm Intelligence,” The
Method,” Journal of Materials Processing Technology, Vol. 184, Nos. International Journal of Advanced Manufacturing Technology, Vol.
1-3, pp. 233-239, 2007. 56, Nos. 5-8, pp. 429-443, 2011.

21. Zain, A. M., Haron, H., and Sharif, S., “Prediction of Surface 33. Farahnakian, M., Razfar, M. R., Moghri, M., and Asadnia, M., “The
Roughness in the End Milling Machining Using Artificial Neural Selection of Milling Parameters by the PSO-Based Neural Network
Network,” Expert Systems with Applications, Vol. 37, No. 2, pp. Modeling Method,” The International Journal of Advanced
1755-1768, 2010. Manufacturing Technology, Vol. 57, Nos. 1-4, pp. 49-60, 2011.

22. Asiltürk, I. and Çunkaş, M., “Modeling and Prediction of Surface 34. Yang, W.-A., Guo, Y., and Liao, W.-H., “Optimization of Multi-Pass
Roughness in Turning Operations Using Artificial Neural Network Face Milling Using a Fuzzy Particle Swarm Optimization Algorithm,”
 704 / MAY 2018 INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 19, No. 5

The International Journal of Advanced Manufacturing Technology,

Vol. 54, Nos. 1-4, pp. 45-57, 2011.

35. Razfar, M., Asadnia, M., Haghshenas, M., and Farahnakian, M.,
“Optimum Surface Roughness Prediction in Face Milling X20Cr13
Using Particle Swarm Optimization Algorithm,” Proceedings of the
Institution of Mechanical Engineers, Part B: Journal of Engineering
Manufacture, Vol. 224, No. 11, pp. 1645-1653, 2010.

36. Zheng, L. Y. and Ponnambalam, S., “Optimization of Multipass

Turning Operations Using Particle Swarm Optimization,” Proc. of
7th International Symposium on Mechatronics and its Applications
(ISMA), pp. 1-6, 2010.

37. Raja, S. B. and Baskar, N., “Optimization Techniques for Machining

Operations: A Retrospective Research Based on Various Mathematical
Models,” The International Journal of Advanced Manufacturing
Technology, Vol. 48, Nos. 9-12, pp. 1075-1090, 2010.

38. Montgomery, D. C., “Design and Analysis of Experiments,” John

Wiley & Sons, 2017.

39. Phadke, M. S., “Quality Engineering Using Robust Design,” Prentice

Hall PTR, 1995.

40. Bement, T. R., “Taguchi Techniques for Quality Engineering,” Taylor

& Francis, 1989.

41. Rai, R., Kumar, A., Rao, S., and Shriram, S., “Development of a
Surface Roughness Prediction System for Machining of Hot
Chromium Steel (AISI H11) Based on Artificial Neural Network,”
ARPN Journal of Engineering and Applied Sciences, Vol. 5, No. 11,
pp. 53-59, 2010.

42. Reddy, N. S. K., Shin, K.-S., and Yang, M., “Experimental Study of
Surface Integrity during End Milling of Al/SiC Particulate Metal-
Matrix Composites,” Journal of Materials Processing Technology,
Vol. 201, Nos. 1-3, pp. 574-579, 2008.

View publication stats