Journal of Cleaner Production 100 (2015) 325e332

Contents lists available at ScienceDirect

Journal of Cleaner Production

journal homepage: w w w . e l s e v i e r . c o m / l o c a t e / j c l e p r o

Application of compressed cold air cooling: achieving multiple performance

characteristics in end milling process
Sonja Jozi'c*, Dra~zen Baji'c, Luka Celent
Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, University of Split, RuCera Bo~skovi'ca 32, 21000 Split, Croatia

a r tic l e in f o
a b s t r a c t
Article history:
This paper applies Taguchi's design of experiment methodology and grey relational analysis for multi objective optimization of process parameters
Received 7 January 2015
Received in revised form 23 and cutting environments during machining operation. In addition to the machining using cutting fluid, this research includes machining with
March 2015 compressed cold air cooling and dry machining. These environment-friendly cutting techniques are considered to be two practical ways to the
Accepted 23 March 2015 cleaner manufacturing in the context of the sustainable production since the reduction of environmental harmful impact is an important topic in
Available online 20 April 2015 sustainable production. Experiments were performed based on L18 standard orthogonal array design by five input process parameters, namely
cutting speed, feed per tooth, radial depth of cut, cutting time and cutting environment. Cutting envi- ronment is considered as machining using
Keywords: cutting fluid, machining with compressed cold air cooling and dry machining. The objective of process parameters optimization is to achieve at the
Milling process same time the maximum volume of removed material and minimum surface roughness, minimum flank wear and minimum cutting force
Compressed cold air cooling Taguchi components. Based on grey relational analysis, the optimal setting of process parameters were identified and analysis of variance indicated that
method cutting time, feed per tooth and cutting environment are the most influential parameters on machining performances.
Grey relational analysis Multi- © 2015 Elsevier Ltd. All rights reserved.
objective optimization

1. Introduction
Sangwan, 2014; Oektem et al., 2006; Reddy and Rao, 2006). Today, the
intention is to improve production while alleviating the manufacturing impact on the
The most of metal parts which are embedded in different ma- chines and devices get environment. The correct choice of cutting environment as well as the suitable machining
their final form by different processing techniques such as machining processes. Machining technique are decisive factors for obtaining better performance characteris- tics. Cutting fluids
processes make about 60 %e80 % of the total manufacturing and represent the most often used are used in metal machining for a variety of reasons such as improving tool life, reducing
metal processing in manufacturing. Moreover, machining is in single and small serial workpiece thermal deformation, improving surface finish and flushing away chips from
production irreplaceable and it represents the only possible economically viable way of pro- the cutting zone. Identification of oil-based cooling/lubrica- tion fluids (CLF) as a main non-
duction. Out of the three principal machining processes such as turning, drilling and sustainable element of the machining has invited environmental concerns on calling for the
milling, milling still remains the most impor- tant operation because of its efficiency and its reduction of cutting fluids in industry and finding alternative types of cooling. Most out of the
possibility of creating complex geometrical shapes. Milling is also commonly used as a large number of research conducted in the area of machining is focused on its activities and
secondary process to add or refine features on parts that were manufactured using some other improvement of the process itself. Some of these improvements, such as opti- misation of
process such as casting or forging. Most of the articles published in the field of the cutting parameters machining, minimizing use of cutting fluids and reduction of energy consumption necessary
optimization are mainly based on technological and economic considerations without the to remove the material from workpiece, have important positive consequences for the
environmental dimension, (Kant and environment. For example, usage of cutting fluids which causes serious health and
environmental problems arising from their application often opens new research fields for
the purpose of solving the above mentioned problems. Strict regulations on use
* Corresponding author. Tel.: þ385 (0)21305 914, þ385 (0)914305914 (mobile); fax: þ385
(0)21305 918.
E-mail address: [email protected] (S. Jozi'c).

http://dx.doi.org/10.1016/j.jclepro.2015.03.095
0959-6526/© 2015 Elsevier Ltd. All rights reserved.
 S. Jozi'c et al. / Journal of Cleaner Production 100 (2015)

and disposal and higher purchase price, costs of cutting fluids which in the 80's of the last
quality, the best cutting conditions are nano-coated tool and small quantity of oil sprayed in a
century made about 3% of machining costs now reach the level of 17% (Weinert et al.,
mist form as a lubricant. Da Silva et al. (2011) presented a comparative study of the
2004). For the sake of comparison, costs of purchase and replacement of tools repre- sent a
influence of two machining environments (dry and wet) in tool wear and surface quality
share of 2 %e4 % of production costs. Also some studies showed that out of the total power
during end milling of AISI 1047 steel with carbide tools. Fluids were directed to the
consumption necessary for machining, almost 60% was used for procurement and supply of
cutting zone by three different tech- niques: flooding, reduced flow rate and minimum
cutting fluids, which made a serious problem considering the trend of permanent growth of
quantity of lubrication (MQL). The results showed that longer machined length values and
energy resources in the previous decade (King et al., 2001; Rah€auser et al., 2011). higher material machining removal volume were ob- tained when machining was performed
Consequently, rapid tool wear and achieving a certain quality of the machined surface do not by means of reduced flow rate system.
exclusively represent the most significant obstacle in economic sustainability of machining, but Rahman et al. (2003) investigated how different types of cooling affect the machining
it is also important to take into consideration the aspect of application of cutting fluids. performances of end milling operation of mould steel using uncoated tungsten carbide inserts.
Previously mentioned cost and a series of negative effects on humans and the environment They applied chilled air, conventional overhead flood coolant and dry cutting. The relative
explain why modern production industry gives more and more attention to the careful performance of these modes is assessed in terms of surface finish, tool wear, cutting force and
selection of socially and envi- ronmentally acceptable coolants. The results of research on nega- tive quality of the chips. Cold air is more favourable in terms of tool wear when cutting param-
effects of cutting fluids as well as multiplication of these effects in recent times bring into question eters have lower values. Cold air is also more favourable in terms of surface roughness if the
the sustainability of such an approach of cooling applications in machining. The use of inap- cutting conditions are set to the higher values. Cutting forces are lower in many cases when using
propriate type, quantity and technique of cutting fluids application which is not in proportion chilled air, than when using flood coolant.
with a particular machining process brings to the reduction in efficiency of positive function of
Environmentally conscious machining of difficult-to-machine materials was reported by
cutting fluids, reduction of tool life, increase the surface roughness and consequently to the
Shokrani et al. (2012). Different types of coolant/lubricants currently in use in machining
decline in process productivity. Many contemporary researches in the field of cutting fluid
industries were reviewed and the drawbacks of using conventional cutting fluids were
application in machining suggest the abandoning of the use of conventional and transfer to the use
defined. The major drawbacks were the environmental and health impacts with the costs
of alternative cooling techniques and eventually dry machining. The best approach to reduce the
associated with their use, mainte- nance and disposal. Due to the difficulties in machining
usage and costs of using cutting fluids is not to use them at all (Sreejith and Ngoi, 2000).
difficult-to- machine materials, there were no techniques that were found to be a complete
The concept of dry machining has many advantages, such as non-pollution of the alternative for cutting fluids. As a result, further research on cooling techniques, cutting tool
atmosphere or water, no residue or the swarf resulting in the reduction of disposal and cleaning materials, cutting pa- rameters and tool geometries was identified as essential and had
cost, no danger to health, such as skin rupture or allergy. The elimination of coolant also imposes the potential to provide significant advantages.
loss of its positive effects, namely lubrication, cooling and chip flushing. Consequently, the
Analysis of single performance characteristics of machining processes has been carried
mechanical and ther- mal loads of the cutting tool are increased, (Jianxin et al., 2005; Schulz
out by many researchers. However, single objective approaches constitute only adequate
et al., 2001). To analyse the cutting process in a dry machining operation, Haddag and
simplifica- tions of the real problem. Machining processes are complex in nature and require
Nouari (2013) developed multi- steps of 3D finite elements modelling including prediction of
often optimizing of various different and con- flicting objectives. The Taguchi method together
tool wear and heat transfer in the tool. Zhong et al. (2010) presented a comparative study on
with grey rela- tional analysis (GRA) has been used in dealing with optimization of multi-
dry milling and a newly proposed method called little quantity lubrication in milling. Their
objective criteria (Siddique et al., 2010; Balasubramanian and Ganapathy, 2011; Sadasiva
study is based on vibration signals that are dependent on the use of cutting fluid. In order to
et al., 2012). Taguchi's parameter design offers a simple and systematic approach to
reduce vibration and cutting fluid usage, process param- eters should be considered while
optimizing design for performance, quality and cost of production (Camposeco-Negrete,
deciding whether and how to apply cutting fluid. Fratila and Caizar (2011) used
2013; Rama Rao and Padmanabhan, 2012). Sreenivasulu and Srinivasa (2012)
Taguchi optimi- zation methodology to optimize the cutting parameters in face milling in
have used GRA to find optimal cutting parame- ters for the drilling process in order to
order to get the best surface roughness and the minimum power consumption. Their
minimize surface roughness and roundness error.
investigation includes conventional flood lubrication, minimal quantity lubrication and dry
Literature clearly indicates that comprehensive investigations have to be performed in
milling.
order to obtain more effective and environment-friendly machining conditions. Therefore, this
Sun et al. (2010) developed a new cooling approach with cryo- genic compressed air in paper deals with multi-objective optimization problem of the cutting parameters and cutting
order to cool the cutting tool edge during turning of Tie6Ale4V alloy. The cutting forces, chip environments (dry, wet and air-cooled) during end milling process. Surface roughness, flank
morphology and chip temperature were measured and compared with those measured
wear, cutting force components in X and Y direction, Fx and Fy, and volume of removed
during machining with compressed air cooling and in dry cutting conditions. They
material were considered as response variables. Cutting force component in Z direction, has
observed flank wear and results showed that application of both compressed air and
not been used into consider- ation because of the negligible small value of the Fz in comparison
cryogenic compressed air reduced flank wear.
to components in X and Y direction. The Taguchi design approach is utilized for experimental
Liew (2010) investigated the performance of TiAlN/AlCrN nano- multilayer coated, planning during end milling with coated carbide inserts. Additionally, GRA was used to
TiAlN single-layer coated and uncoated carbide tools in low-speed milling of stainless steel find the optimal machining parameters satisfying the multiple characteristics of the process, i.e.
under flood and mist lubrication. Observed from the view point of tool wear and surface lower values of surface roughness, flank wear, cutting force and higher value of the volume of
removed material. Finally,
 S. Jozi'c et al. / Journal of Cleaner Production 100 (2015)

Table 1
The chemical composition and mechanical properties of specimens. Chemical composition % Table 2
Insert and end mill geometry.
Insert length, la 11 mm
Insert width, iW 6.8 mm
C Si Mn P S Cr Ni Mo Cu
Insert thickness, s 3.59 mm
0.430 0.278 0.77 0.018 0.028 1.0 0.08 0.185 0.08 Edge radius, rε 0.8 mm
o
9 Clearance angle 21
Mechanical properties x
Number of inserts 3
Yield Tensile Elongation Notch impact Hardness Nominal cutter diameter 20 mm
strength [MPa] strength [MPa] [%] energy [J] [HRC]

1128 1223 14.4 42 35 done by using toolmaker's microscope with 100 magnification and USB camera. The
system for measuring the cutting force con- sists of dynamometer Kistler 9257A mounted
the analysis of variance was carried out to obtain significant pa- rameters influencing multi between the work- piece and the table as well as multi-channel charge amplifier Kistler 5007A
performance characteristics of the process and confirmation test was performed. that amplifies the electrical charges delivered from dyna- mometer into corresponding voltage
and then forwards the voltage signals to the A/D interface board (BMC USB-AD16f).
2. Experimental setup Milling force components Fx, Fy and Fz, are the sum of projections of the tangential, radial
and axial forces acting on the cutting edges during milling. Milling force components present the
The end milling experiments were carried out on CNC vertical machining centre, mean values of the maximum cutting force on each insert. Measurements and analysis
Spinner VC560, equipped with a 12,000 rpm electro spindle and the SK 40 tool holder. equipment are shown in Fig. 2.
Workpiece material was steel 42CrMo4 and was prepared for milling operations in the form 250
mm 110 mm 110 mm blocks and adapted to the experi- ment needs. Chemical 2.2. Cutting conditions
composition and mechanical properties of test samples are given in Table 1.
x x
The cutting tool was end mill CoroMill R390-02A20-11M pro- duced by Sandvik. Cutting speed vc, feed per tooth ft, radial depth of cut ae, cutting time, i.e. insert's
The inserts were carbide coated with highly resistant TiN coating, Grade Coromant engagement time and cutting environment were considered as controllable variables. Axial
GC1025. Insert geometry varied depending on the depth of cut in order to obtain better depth of cut ap was constant, 5 mm. The adopted values of cutting parameters corre- spond
condition of the chips flow. The closest point for the clearance angle to the operational limits recommended by the toolmaker together with the machine tool capabilities.
o
is 21 and before that angle there is one facets ~0.2 mm with the The cutting parameters and their levels are given in Table 3. Every experiment was per-
o
angle of 0 . The end mill and insert are shown in Fig. 1. Tool spec- ifications are given formed with non-test inserts and down milling method was adopted. Dry cutting,
in Table 2. conventional wet cooling (flood coolant) and compressed cold air cooling were used in the
experiments. Vortex tube system was applied for cold air producer.
2.1. Measurement and analysis of surface roughness, flank wear
and cutting force components 2.3. Air cooling with cold air gun using vortex tube system

The profilometer Mitutoyo Surftest 301 was used for surface roughness measuring. Cut- Employing compressed cold air for cooling in machining oper- ations is a relatively new
off length and the sampling length for the surface roughness measurements were selected to be technique. Air that rotates around an axis is called a vortex. The vortex tube, also known as
0.8 mm and 5.6 mm, respectively. Every measurement was repeated five times and the the Ran- queeHilsch vortex tube, is a mechanical device that creates cold and hot air by forcing
average value was considered. Flank wear mea- surements were performed in accordance compressed air through a generation chamber, which spins the air at a high rate of
with the International Standard ISO 8688-1 and periodically according to the experi- speed (1,000,000 rpm) into the vortex. The vortex tube consists of an inlet nozzle, vortex
mental plan. The inserts were removed from the tool holder after a given interval, and the flank chamber, cold-end orifice, hot-end control valve and a tube. It has no moving parts. High pressure
wear of the all three inserts was measured wherein the average flank wear value was used. air stream enters into the vortex tube tangentially, and there it splits into two lower pressure
This was streams, one hot and one cold temperature stream. Cold gas stream leaves the tube through a
centre orifice near the entrance nozzle, while the hot gas stream flows toward the control valve and
leaves the tube there, Fig. 3. The control valve located in the hot exhaust end can be used to
adjust the temperature drop and rise (Kumar Rao et al., 2014; Chengming, 2005).
In this research the cooling of cutting tools and the workpiece is performed by means of Cold
Air Guns which use vortex tube technology and filtered compressed air to produce sub-
freezing air
o
with temperature of 34 C, - Fig. 4. With no moving parts to wear
out, Cold Air Guns require no electricity at the target, just a com- pressed air source. The
effective cooling from a Cold Air Gun can eliminate heat-related parts growth while improving
parts toler- ance and surface finish quality. The air is a free available natural resource and
compressed air is available on the regular basis at the

Fig. 1. End mill and insert used in the cutting tests.

 S. Jozi'c et al. / Journal of Cleaner Production 100 (2015)

Fig. 2. Measurement and analysis equipment.

Table 3
The experimental factors and their levels.
Levels of exp. runs Cutting speed, vc [m/min] Feed per tooth ft [mm/tooth] Radial depth of cut, ae [mm] Engagement time of inserts, ct [min] Cutting conditions, ce

A B C D E

1 100 0.05 1 10 Dry machining (DM)

2 125 0.08 1.5 16 Cutting fluid (CF)
3 150 0.11 2 22 Air cooling (AC)

shop floor for other purposes. It has no adverse effect on the health of the operator.
are weighted equally. This means that every factor appears on the same number of levels and that
every factor on any level will be in all combinations with other factors. This allows testing only a
2.4. Experimental design based on Taguchi method limited collection of factors combinations instead to check all possible combinations like
factorial design of experiment. Such method allows determination of factors which most affect
The experimental design proposed by Taguchi method uses the orthogonal arrays to organize process with a minimum amount of experiments, thus saving time and resources. The
the factors affecting the process. An orthogonal array means the design is balanced so that factor experimental design for five controllable variables with three levels, presented in Table 3, is
levels organized by the Taguchi's L18 orthogonal array. The L18 orthogonal array has eight columns
and eighteen rows, so it has 17 degrees of freedom to manipulate one parameter with two
levels and seven parameters with three levels. Since it was taken into account five parameters,
two columns are remained empty. However, it still guaranteed orthogonality, even if two
columns of the array remained empty. Experimental runs with results are given in Table 4.
Four process variables are measured. These included: surface roughness Ra [mm], flank
wear VB [mm], cutting force components Fx [N] and Fy [N] while the total

Fig. 3. The schematic representation of vortex tube working principle (Kumar Rao et al., 2014).

Table 4
Experimental layout using L18 orthogonal array and performance results.

Fig. 4. Photographic view of the experimental set-up.

Runs A B C D E Ra [mm] VB [mm] Fx [N] Fy [N] VRM [mm3]

1 1 1 1 1 1 0.18 0.04 361.98 767.96 11,943.68

2 1 2 2 2 2 0.31 0.07 301.34 823.69 45,859.87
3 1 3 3 3 3 0.62 0.07 486.13 1192.40 11,5605.1
0
4 2 1 1 2 2 0.37 0.07 266.59 739.71 23,885.35
5 2 2 2 3 3 0.44 0.08 399.39 1108.63 78,821.66
6 2 3 3 1 1 0.64 0.07 465.14 S. Jozi'c et 65,684.71
1163.79 al. / Journal of Cleaner Production 100 (2015)
7 3 1 2 1 3 0.19 0.05 246.46 664.14 26,871.02
8 3 2 3 2 1 0.36 0.11 424.53 1102.73 91,719.75
9 3 3 1 3 2 0.62 0.12 658.58 1337.98 86,703.82
10 1 1 3 3 2 0.45 0.09 319.28 959.69 52,547.77
11 1 2 1 1 3 0.27 0.041 381.0 775.62 19,108.28
1
12 1 3 2 2 1 0.93 0.12 506.89 1087.89 63,057.32
13 2 1 2 3 1 0.60 0.12 350.92 1008.51 49,263.54
14 2 2 3 1 2 0.33 0.05 329.36 955.12 47,770.70
15 2 3 1 2 3 0.47 0.06 486.44 960.99 52,547.77
16 3 1 3 2 3 0.21 0.10 375.67 919.43 57,324.84
17 3 2 1 3 1 1.10 0.16 556.89 1267.84 63,057.32
18 3 3 2 1 2 0.32 0.06 408.38 933.45 59,116.24
 S. Jozi'c et al. / Journal of Cleaner Production 100 (2015)

volume of removed material VRM [mm3] is calculated as shown below. analysis, the complex multiple response optimization problem can be simplified into an
Highly efficient production can be achieved by increasing the volume of removed optimization of a single response grey rela- tional grade. The procedure for determining the
material per time unit. The material removal rate, MRR [mm3/min] in milling process, grey relational grade is discussed below.
Fig. 5, is calculated as follows:

vc$1000 3.1. Data pre-processing

MRR ¼ apaevf ¼ apaeftnN ¼ apaeft N (1)
Dp
Data pre-processing is performed in order to transfer original sequence to a comparable
where ap [mm] is axial depth of cut, ae [mm] radial depth of cut, vf [mm/min] feed rate, ft sequence. Numerical data are normal- ized between zero and one. Depending on
[mm/tooth] feed per tooth, D [mm] mill diameter, n [rpm] spindle speed and N number characteristics of data sequence, various methodologies of data pre-processing are avail- able. In
of inserts. this study, the normalized value of original sequence for surface roughness, cutting force
Total volume of removed material VRM [mm3] is calculated as follows: components and flank wear which are smaller-the-better performance characteristic can be
expressed as:
VRM ¼ MRRt (2) ( )
In this study, further processing procedure should find optimal max yij - yij
x ¼ ( ) ( ) (3)
controllable variables for multi objective function using the GRA ij
method. Five quality objectives of the process were chosen, max y ij - min y ij
including the surface roughness, flank wear, cutting force compo- nents Fx, Fy and volume of
removed material. Typically, the small values of surface roughness, flank wear and cutting where yij are original data.
force compo- nents Fx, Fy and large value of volume of removed material are desirable for The normalized value of original sequence for volume of removed material which is
the milling operations. larger-the better performance charac- teristic can be expressed as:
( )
The influence of the controllable process factors on the multi
objective function was studied. This was based on the correlation yij - min yij (4)
x
between the process factors. By analysing the grey relational grade ¼ ( ij) ( )ij
matrix, the most influential factors can be selected. ij
max y - min y
The normalized values of surface roughness, flank wear, cutting force components and
3. Grey relational analysis volume of removed material removal rate are shown in Table 5.
For a response j of experiment i, if the value xij which has been processed by data pre-
processing procedure is equal to 1 or nearer to 1 than the value for any other experiment, then the
A grey system presents a system in which a part of information is known and the other
performance of experiment i is considered as the best for the response j. The reference
part is unknown (Venkata Rao, 2011). Because of the uncertainty that always exists, grey
systems will give a variety of available solutions. Grey relational analysis, based on this sequence X0 is defined as (x01, x02, …, x0j, …, x0n) ¼ (1, 1, …, 1, …, 1), where x0j is
theory, can be further effectively adapted for solving the complicated interrelationships among the reference value for jth response and it aims
the designated performance characteristics. Through this analysis, the grey relational grade to find the experiment whose comparability sequence is the closest to the reference sequence [29].
(GRG) is favourably defined as an indicator of multiple performance characteristics for Table 6 shows deviation sequences. Next, the grey relational coefficient is used for determining
evaluation. In recent years, the grey relational analysis has become the powerful tool to analyse how close xij is to x0j. The larger the grey relational coefficient, the closer
the processes with multiple performance characteristics. In grey relational xij and x0j.
The grey relational coefficient can be determined as:

Table 5
Normalized experimental results.

Fig. 5. Parameters for determining the material removal rate.

Ra [mm] VB [mm] Fx [N] Fy [N] VRM [mm3]
 S. Jozi'c et al. / Journal of Cleaner Production 100 (2015)
1 1 1 0.7197 0.8459 0
2 0.8587 0.7500 0.8668 0.7632 0.3272
3 0.5217 0.7500 0.4184 0.2160 1
4 0.7935 0.7500 0.9512 0.8878 0.1152
5 0.7174 0.6667 0.6289 0.3404 0.6452
6 0.5000 0.7500 0.4694 0.2585 0.5184
7 0.9891 0.9167 1 1 0.1440
8 0.8403 0.4167 0.5679 0.3491 0.7696
9 0.5217 0.3333 0 0 0.7212
10 0.7065 0.5833 0.8233 0.5614 0.3917
11 0.9022 0.9917 0.6735 0.9346 0.0691
12 0.1848 0.3333 0.3681 0.3711 0.4931
13 0.5435 0.3333 0.7465 0.4889 0.3600
14 0.8370 0.9167 0.7989 0.5682 0.3456
15 0.6848 0.8333 0.4177 0.5595 0.3917
16 0.9674 0.5000 0.6865 0.6211 0.4378
17 0 0 0.2467 0.1041 0.4931
18 0.8478 0.8333 0.6071 0.6003 0.4551
 S. Jozi'c et al. / Journal of Cleaner Production 100 (2015)

Table 6
Deviation sequences. Table 7
Grey relational coefficients and grey relational grade.

Ra [mm] VB [mm] Fx [N] Fy [N] VRM [mm3] Ra [mm] VB [mm] Fx [N] Fy [N] VRM [mm3] Grade Grey order

1 0 0 0.2803 0.1541 1 1 0.7797 0.6667 0.7897 0.6789 0.3333 0.7477 2

2 0.1413 0.2500 0.1332 0.2368 0.6728 2 0.511 0.6667 0.4623 0.3894 0.4263 0.6682 6
1
3 0.4783 0.2500 0.5816 0.7840 0 3 0.7077 0.6667 0.9110 0.8168 1 0.6059 11
4 0.2065 0.2500 0.0488 0.1122 0.8848 4 0.6389 0.6000 0.5740 0.4312 0.3611 0.6926 8
5 0.2826 0.3333 0.3711 0.6596 0.3548 5 0.5000 0.6667 0.4851 0.4027 0.5849 0.5658 13
6 0.5000 0.2500 0.5306 0.7415 0.4815 6 0.9787 0.8571 1 1 1 0.5128 7
7 0.0109 0.0833 0 0 0.8560 7 0.7188 0.4615 0.5364 0.4344 0.3687 0.8409 1
8 0.1957 0.5833 0.4321 0.6509 0.2304 8 0.511 0.4286 0.3333 0.3333 0.6845 0.5671 10
1
9 0.4783 0.6667 1 1 0.2788 9 0.6301 0.5455 0.7389 0.5327 0.6420 0.4497 16
10 0.2935 0.4167 0.1767 0.4386 0.6083 10 0.8364 0.9836 0.6050 0.7514 0.4511 0.5797 12
1 0.0978 0.0083 0.3265 0.1654 0.9309 11 0.3802 0.4286 0.4417 0.4429 0.3497 0.7052 5
1
12 0.8152 0.6667 0.6319 0.6289 0.5069 12 0.5227 0.4286 0.6636 0.4945 0.4966 0.4380 17
13 0.4565 0.6667 0.2535 0.5111 0.6400 13 0.7541 0.8571 0.7131 0.5366 0.4386 0.5096 15
14 0.1630 0.0833 0.2011 0.4318 0.6544 14 0.6133 0.7500 0.4620 0.5316 0.4331 0.6588 3
15 0.3152 0.1667 0.5823 0.4405 0.6083 15 0.9388 0.5000 0.6146 0.5689 0.4511 0.5616 14
16 0.0326 0.5000 0.3135 0.3789 0.5622 16 0.3333 0.3333 0.3990 0.3582 0.4707 0.6186 9
17 1 1 0.7533 0.8959 0.5069 17 0.7667 0.7500 0.5600 0.5558 0.4966 0.3841 18
18 0.1522 0.1667 0.3929 0.3997 0.5449 18 0.7797 0.6667 0.7897 0.6786 0.4785 0.6222 4

ðDmin
þxx0jD; max
xijÞ ¼ for i ¼ 1; 2;…; m and j ¼ 1; 2;…;
n The means of the grey relational grade for each level of
g
Dij þxDmax
controllable parameters were calculated from Table 7 and sum- marized in Table 8. The
(5) larger the grey relational grade, shown in bold in Table 8, the better the multiple performance
characteristics.
where
y(x0j,xij) is the grey relational coefficient between xij and x0j, Therefore, the optimal values of the controllable parameters are as follows: cutting speed of 100
m/min (level 1), feed per tooth of
0.05 mm/tooth (level 1), radial depth of cut of 2 mm (level 3), cutting time of 10 min
Dij ¼ x0j - xij ; (6)
(level 1) and cutting environments e air cooled machining (level 3). Fig. 6 shows the
{ milling parameters in relation to the grey relational grade.
Dmin ¼ min Dij; i ¼ 1; 2; …; m; j ¼ 1; 2; …; n (7)

{ 4. Analysis of variance for grey relational grade

Dmax ¼ max Dij; i ¼ 1; 2; …; m; j ¼ 1; 2; …; n (8)
The analysis of variance (ANOVA) is a method which is used in this study to find
x is distinguishing coefficient, x2(0,1]. In the present study the value of x is
out which controllable parameter significantly affects the performance characteristic. This is
assumed as 0.5. Distinguishing coefficient (x) is the index for distinguishability. The accomplished by separating the total variability of the grey relational grades, which is measured
smaller x, the higher its distinguishability. by the sum of the squared deviations from the total mean of the grey relational grade into
The measurement formula for quantification in grey relational space is called the grey contributions by each controllable parameter and the error [29]. The percentage contri- bution
relational grade. A grey relational grade is a weighted sum of the grey relational coefficients and by each of the process parameters in the total sum of the squared deviations was used to
it is calculated using equation: evaluate the importance of the controllable parameter change on the performance characteristic.
The results of ANOVA for the grey relational grade values are shown in Table 9. The
X
n
results indicate that the percentage contri- bution of the cutting time, feed per tooth and cutting
GðX0 ; Xi Þ ¼ wj g x0j ; xij for i ¼ 1; 2; …; m (9) condition are 37%, 24% and 21%, respectively. These three parameters signifi- cantly
j¼1
influenced the grey relational grade, and cutting time is the most significant parameter for
P multiple performance characteris- tics. Table 9 shows that the cutting speed and depth of cut do
where n w j¼jj ¼ 1 not have statistically significant effect on the multiple performance
G(X0,Xi) 1is the grey relational grade between comparability
sequence Xi and reference sequence X0.
The weight of response j is wj and usually depends on decision makers' judgement.
The grey relational grade indicates the degree of similarity be- tween the comparability Table 8
sequence and the reference sequence. If an experiment gets the highest grey relational grade with Response table for grey relational grade.
the reference sequence, it means that comparability sequence is most similar to the reference Parameter Level 1 Level 2 Level 3 Rank [max/min]
sequence and that experiment would be the best choice.
vc 0.6246 0.5835 0.5804 4 (0.0442)
Table 7 shows the grey relational coefficients and grade for each experiment. The highest ft 0.6649 0.5921 0.5317 2 (0.1332)
grey relational grade is the order of 1. The experiment number 7, highlighted in grey, is the ae 0.5904 0.6074 0.5905 5 (0.0168)
nearest optimum controllable parameters combination: cutting speed of 150 m/min (level 3), ct 0.6818 0.5910 0.5158 1 (0.1660)
feed per tooth of 0.05 mm/tooth (level 1), radial depth of cut of 1.5 mm (level 2), cutting time ce 0.5266 0.6119 0.6502 3 (0.1237)
of 10 min (level 1) and cutting environments e air cooled machining (level 3). Total mean value of the grey relational grade ¼ 0.5962.
 S. Jozi'c et al. / Journal of Cleaner Production 100 (2015)

Fig. 6. Grey relational grade graph.

Table 9
Results of the analysis of variance. from 0.04 mm to 0.035 mm, respectively, cutting force component Fx decreased and
volume of removed material increased from 11,943.68 mm3 to 23,885.35 mm3. Finally,
df Seq SS Adj MS F-ratio Contribution
it is shown clearly that the multiple performance characteristics in end milling were
vc 2 0.00713 0.003565 0.785152 3% significantly improved by 0.1195. A good agreement between the actual and
ft 2 0.053369 0.026684 5.877163 24%
predicted GRG was obtained.
ae 2 0.001175 0.000587 0.129345 1%
ct 2 0.082364 0.04118 9.070219 37%
2
ce 2 0.047731 0.023865 5.256328 21% 6. Conclusions
Error 7 0.031782 0.00454
Total 17 0.22355
In this paper, the controllable parameters influencing the mul- tiple performance
df e Degree of freedom, SS e Sum of square, MS e Mean square. characteristics of end milling were studied on the basis of Taguchi's experimental design
method. The optimum milling parameters were determined for the better multiple per- formance
characteristics (surface roughness, cutting force compo- nents, flank wear and volume of
characteristics. It may be noted that cutting speed might have an effect on some performance removed material) in the milling process by using the grey relational analysis. This research
characteristics individually and its ef- fect is probably not significant, when all responses are pro- posed the orthogonal array combined with the grey relational analysis to optimize multiple
considered together as it has been observed in presented research. performances of the end milling process when three different machining conditions are
used: dry machining, conventional machining with cutting fluid and novel technique known
5. Predicting optimal value and confirmation test as air cold machining. The cutting time has also been used as a controllable value in order to track
the flank wear. All process values were measured at the same point of time.
The optimal grey relational grade is predicted at the selected optimal setting of controllable The conclusions were summarized as follows:
variables. The most significant parameter with optimal level was already selected as D1
(cutting time) followed by E3 (cutting environment) and B1 (feed per tooth). The estimated 1 It can be concluded from the grey relational grade and the response table for the
grey relational grade can be calculated as: grey relational grade that the larger the grey relational grade, the better the multiple
performance characteristics. So, the optimal levels of milling parameters for the desired
X
p performance characteristics is the combination labelled as A1B1C3D1E3. With this
b¼ g þ
g ðgi - gt Þ (10) combination it is possible to decrease the surface roughness, cutting force components and
t
i¼1 flank wear and maximize the volume of removed material.
2 Based on the ANOVA of the GRG results, it is observed that cutting time, cutting
where gt is the total mean of the grey relational grade, gi is the mean of the grey relational grade environment and feed per tooth exerted a significant influence on multiple responses.
at the optimal level, and p is the number of the controllable variable that significantly affect the 3 As a result, the effectiveness of this approach is verified by the test experiment. The grey
multiple performance characteristics. relational grade of the multiple
The A1B1C3D1E3 is an optimal parameter combination of the end milling process
obtained by means of GRA and was considered as a confirmation test. Based on Eq (10) the
estimated GRG can be obtained. Table 10 presents the results of the confirmation test using the
optimal controllable variables. Surface roughness and flank wear were slightly improved
from 0.18 mm to 0.16 mm, and

Table 10
Results of the confirmation test.

Setting level Ra [mm] VB [mm] Fx [N] Fy [N] VRM [mm3] GRG Improvement in GRG

Initial controllable parameters A1B1C1D1E 0.18 0.04 361.89 767.97 11,943.68 0.7477
1
Optimal controllable parameters Prediction A1B1C3D1E3 0.7937
Experiment A1B1C3D1E3 0.16 0.035 321.42 889.35 23,885.35 0.8672 0.1195
 S. Jozi'c et al. / Journal of Cleaner Production 100 (2015)

performance characteristics can be significantly improved by 0.1195 through this Oektem, H., Erzurumlu, T., Coel, M., 2006. A study of the Taguchi optimization method for surface roughness
method. in finish milling of moulds surfaces. Int. J. Adv. Manuf. Technol. 28, 694e700.
Rah€auser, R., Pflüger, T., Regenfelder, M., 2011. Economic potentials of improve- ments of energy
4 This paper confirms that the air cold technique can be suc- cessfully applied when it efficiency in the supply of cooling lubricants for metal cutting machine tools. In: Proceedings of the MOTSP 2011
is necessary to fulfil previously established multiple performance characteristics. International Conference on Management of Technology e Step to Sustainable Production. Brac, Croatia.
Rahman, M., Senthil Kumar, A., Manzoor-Ul Salam, Ling, M.S., 2003. Effect of chilled air on machining
performance in end milling. Int. J. Adv. Manuf. Technol. 21, 787e795.
Rama Rao, S., Padmanabhan, G., 2012. Application of Taguchi methods and ANOVA in optimization of
References process parameters for metal removal rate in electrochemical machining of Al/5%SiC composites. Int. J.
Eng. Res. Appl. 2, 192e197.
Balasubramanian, S., Ganapathy, S., 2011. Grey relational analysis to determine optimum process parameters Reddy, N.S.K., Rao, P.V., 2006. Selection of an optimal parametric combination for achieving a better surface
for wire electro discharge machining (WEDM). Int. J. Eng. Sci. Technol. 3, 95e101. finish in dry milling using genetic algorithms. Int. J. Adv. Manuf. Technol. 28, 463e473.
Camposeco-Negrete, C., 2013. Optimization of cutting parameters for minimizing energy consumption in Sadasiva, R.T., Rajesh, V., Venu Gopal, A., 2012. Taguchi based grey relational anal- ysis to optimize face
turning of AISI 6061 T6 using Taguchi methodology and ANOVA. J. Clean. Prod. 53, 195e203. milling process with multiple performance characteristics. In: International Conference on Trends in Industrial
Chengming, G., 2005. Experimental Study on the RanqueeHilsch Vortex Tube. and Mechanical Engi- neering (ICTIME’2012). March 24 e 25, Dubai.
Technische Universitet Eindhoven. Schulz, H., Doerr, J., Rass, I.J., Schulze, M., Leyendecker, T., Erkens, G., 2001. Perfor- mance of oxide PVD-
Da Silva, R.B., Vieira, J.M., Cardoso, R.N., Carvalho, H.C., Costa, E.S., Machadoa, A.R., De coatings in dry cutting operations. Surf. Coat. Technol. 146/ 147, 480e485.
A' vila, R.F., 2011. Tool wear analysis in milling of medium carbon steel with Shokrani, A., Dhokia, V., Newman, S.T., 2012. Environmentally conscious machining of difficult-to-machine
coated cemented carbide inserts using different machining lubrication/cooling systems. Wear 271, materials with regard to cutting fluids. Int. J. Mach. Tools Manuf. 57, 83e101.
2459e2465. Siddique, A.N., Khan, Z.A., Mallick, Z., 2010. Grey relational analysis coupled with principal component
Fratila, D., Caizar, C., 2011. Application of Taguchi method to selection of optimal lubrication and cutting analysis for optimisation design of the process parameters in in-feed centreless cylindrical grinding. Int. J. Adv.
conditions in face milling of AlMg3. J. Clean. Prod. 19, 640e645. Manuf. Technol. 46, 983e992.
Haddag, B., Nouari, M., 2013. Tool wear and heat transfer analyses in dry machining based on multi-steps Sreejith, P.S., Ngoi, B.K.A., 2000. Dry machining: machining of the future. J. Mater.
numerical modelling and experimental validation. Wear 302, 1158e1170. Process. Technol. 101, 287e291.
Jianxin, D., Tongkun, C., Lili, L., 2005. Self-lubricating behaviours of Al2O3/TiB2 ceramic tools in dry Sreenivasulu, R., Srinivasa, Ch., 2012. Application of grey relational analysis for surface roughness and
high-speed machining of hardened steel. J. Eur. Ceram. Soc. 25, 1073e1079. roundness error in drilling of Al 6061 alloy. Int. J. Lean Think. 3, 67e78.
Kant, G., Sangwan, K.S., 2014. Prediction and optimization of machining parameters for minimizing power Sun, S., Brandt, M., Dargusch, M.S., 2010. Machining Tie6Ale4V alloy with cryogenic compressed air
consumption and surface roughness in machining. cooling. Int. J. Mach. Tools Manuf. 50, 933e942.
J. Clean. Prod. 83, 151e164. Venkata Rao, R., 2011. Advanced Modelling and Optimization of Manufacturing Pro- cesses International
King, N., Keranen, L., Sutherland, J., 2001. Wet Versus Dry Turning: A Comparison of Machining Costs, Research and Development. Springer-Verlag London Limited.
Product Quality and Aerosol Formation. SAE Paper, pp. 1579e1588. Weinert, K., Inasaki, I., Sutherland, J.W., Wakabayashi, T., 2004. Dry machining and minimum
Kumar Rao, K.K., Pradeep, G.R.C., Ramesh, A., Prabhakar, K., 2014. Fabrication and quantity lubrication. CIRP Ann. e Manuf. Technol. 53, 511e537.
experimental analysis of adiabatic vortex tube. J. Mech. Civ. Eng. 11, 57e63. Zhong, W., Zhao, D., Wang, X., 2010. A comparative study on dry milling and little quantity lubricant
Liew, W.Y.H., 2010. Low-speed milling of stainless steel with TiAlN single-layer and TiAlN/AlCrN nano- milling based on vibration signals. Int. J. Mach. Tools Manuf. 50, 1057e1064.
multilayer coated carbide tools under different lubrication conditions. Wear 269, 617e631.