Journal of Cleaner Production 94 (2015) 321e329

Contents lists available at ScienceDirect

Journal of Cleaner Production

journal homepage: www.elsevier.com/locate/jclepro

Single and multiple goal optimization of structural steel face milling

process considering different methods of cooling/lubricating
Katica Simunovic*, Goran Simunovic, Tomislav Saric
Mechanical Engineering Faculty in Slavonski Brod, University of Osijek, Trg I. Brlic Mazuranic 2, HR 35000 Slavonski Brod, Croatia

a r t i c l e i n f o a b s t r a c t

Article history: With the main aim of studying the reduced use of cooling/lubricating fluid and thereby contributing to
Received 16 July 2014 cleaner manufacturing and environmental protection, the present study investigates the influence of four
Received in revised form input variables, namely methods of cooling/lubricating (dry cutting conditions, cooling/lubricating
4 February 2015
through the tool and flood cooling/lubricating), depth of cut, number of revolutions and feed rate, on the
Accepted 5 February 2015
surface roughness of face-milled structural steel. Through response surface methodology, the optimal
Available online 14 February 2015
machining parameters for obtaining the least surface roughness and the optimal parameters for two
objective functions, namely minimum surface roughness and minimum standard deviation of the vari-
Keywords:
Cooling/lubricating
ability in the response transmitted from the factors, i.e., the propagation of error, were determined. The
Face milling robustness study revealed that the obtained optimal parameters (depth of cut, 1.04 mm; feed rate,
Structural steel 100 mm/min; number of revolutions, 800 rev/min; and dry machining) will yield a minimum roughness
Surface roughness of 0.62 mm with a minimum variability in the response transmitted from the factors. This study also
Optimization demonstrated that nearly the same surface roughness of machined area was achieved for dry machining
and cooling/lubricating through the tool, while the surface roughness higher values were obtained for
flood cooling/lubricating. Due to the short machining time (small surface), the tool heating did not occur
in dry machining while in cooling/lubricating through the tool, positive influence of high pressure was
emphasized due to suppressing the formed chips from the tool and protecting the tool edge. Therefore, in
both processes better quality of surface was obtained than in low pressure flood cooling/lubricating in
which suppressing chips on the tool edge occurs as well as vibrations and increase of surface roughness.
© 2015 Elsevier Ltd. All rights reserved.

1. Introduction Analysis, or shorter Life Cycle Assessment, LCA, developed in the

1990s, points to the need to save material and energy by using more
Not long ago, the investigations of machinability and surface recycled materials and to the development and application of
roughness were mostly based on studying the influence of alternative (sustainable) machining technologies and sustainable
machining parameters to obtain the best possible quality of the production (Pusavec et al., 2010a, 2010b). All of the above ap-
surface with lower costs and reduced tool wear and energy con- proaches contribute to sustainable development through
sumption. However, one of the most important aims of in- improvement of environmental performances.
vestigations conducted today is cleaner production, i.e., application The previously mentioned approaches point to the fact that in
of “environmentally friendlier” technologies. As a result, a number machining, the principles of sustainable development and sus-
of different approaches and tools have recently been developed tainable environment need to be applied and fulfilled. To achieve
aimed at the prevention of environmental pollution, at the possi- this, one of the goals of greater importance is to use lower quan-
bility of products being recycled and reused, at saving of energy tities of cutting fluid (and of other fluids as well) and alternative
consumption and at the responsibility to future generations. Thus, methods of cooling/lubricating while machining, as this reduces
for example, the approach Life Cycle Engineering e Assessment and environment pollution. Along with reduced quantities, the aim is
also to use bio-degradable fluids and thus contribute to environ-
ment protection. The influence of cutting fluids on human health
* Corresponding author. Tel.: þ385 35 493 430.
(toxicity) must not be neglected either. Therefore, one of the
E-mail addresses: katica.simunovic@sfsb.hr (K. Simunovic), goran.simunovic@ important factors to be varied is also the method, way or strategy of
sfsb.hr (G. Simunovic), tomislav.saric@sfsb.hr (T. Saric).

http://dx.doi.org/10.1016/j.jclepro.2015.02.015
0959-6526/© 2015 Elsevier Ltd. All rights reserved.
322 K. Simunovic et al. / Journal of Cleaner Production 94 (2015) 321e329

cooling/lubricating. The type of cutting fluid and the method of cutting speed, depth of cut and feed rate on specific energy, tool life
cooling/lubricating in terms of the pollution of environment, the and surface roughness during end milling of AISI 304 stainless steel
effect on human health, the heating and wear of tool, as well as the was investigated in Kuram et al. (2013). They concluded that the
economic issues, is therefore discussed in many studies. machining performance could be improved by the use of vegetable
It can be noticed that different ways of wet or MQL cooling/ based oil emulsions (sunflower and canola) which could be
lubricating are applied. These include cryogenic cooling, high- considered as an alternative to semi synthetic coolants/lubricants.
velocity and high pressure of cutting fluid, narrow, pulsed jet, A similar subject was investigated by Cetin et al. (2011) and similar
micro-closed loop recirculation of coolant through the tool holder conclusions were also made but at examining turning AISI 304L
e indirect cooling, micro-flood technology, micropooled lubri- stainless steel by the application of the Taguchi method. Lawal et al.
cating, and cooling/lubricating through the tool (in the present (2014) also evaluated vegetable oil-in-water emulsion cutting
paper). In the paper of Aggarwal et al. (2008a) it is demonstrated fluids (as in Cetin et al., 2011; Kuram et al., 2013) and mineral oil-in-
that the cryogenic environment exerts the strongest influence on water emulsion in turning AISI 4340 steel with coated carbide tools.
reducing power consumption, followed by the cutting speed and A conclusion was drawn that vegetable oil-in-water emulsion
the depth of cut while feed rate and nose radius showed to be non (palm kernel and cottonseed) could be considered as an alternative
significant in CNC turning of AISI P-20 tool steel. Those authors, but to mineral oil. At end milling of Inconel 718, Zhang et al. (2012)
in Aggarwal et al. (2008b) studied the same material by varying combined the afore-mentioned vegetable based oil emulsions
cutting speed, feed rate, depth of cut and nose radius in CNC and cryogenic environment (they used cryogenic compressed air
turning under cryogenic cutting environment. Thepsonthi et al. of 30  C), and called it minimum quantity cooling lubrication-
(2009) pointed out considerable advantages of high-velocity, nar- MQCL.
row, pulsed jet of cutting fluid in high-speed milling of hardened Nano-cutting fluids have also been used, and their properties
steel (reduced tool wear, reduced forces, reduced quantity of have been studied. Sayuti et al. (2014) applied the Taguchi method
coolant, improved roughness). As economical, this method is for optimization of cooling/lubricating technique in turning of
applied in Pusavec et al. (2010a, 2010b) as well and compared with hardened steel AISI4140 by the application of nanofluid (a mixture
conventional and cryogenic machining by the application of the of SiO2 nanoparticles with an average size of 5e15 nm in the
LCA approach. These authors have concluded that the two alter- mineral oil and pressured air is used). The following parameters
native cooling/lubricating methods can be successfully applied at were studied: nanoparticles concentration, nozzle angle and air
lower speeds of machining and that they are superior at high carrier pressure and their influence on surface roughness and tool
machining speeds, compared with conventional machining (flood wear.
cooling/lubricating). Minton et al. (2013) dealing with the It can be noticed that many investigators have used design and
machining of titanium proved that the heat resulting from friction analysis of experiment (DOE) methodology, i.e., statistical methods
during machining was successfully taken away by a specially to model and optimize the machining process. Response surface
designed, diamond coated insert via closed loop recirculation of methodology and appropriate experimental designs, mostly central
coolant, through the tool holder (indirect cooling). composite design and Taguchi method is frequently used by many
Many authors have studied machining under dry, minimal authors. Factorial (full or fractional) and D e optimal design has also
quantity lubrication (MQL) or near dry machining (NDM) and wet been found.
conditions. In nearly all investigations, it has been concluded that Apart from the design of experiment methodology, application of
MQL is a good alternative to wet machining, regarding the tool the methods of artificial intelligence in surface roughness investi-
wear, costs, health and environment protection. A number of au- gation has become more and more common. Using the neural
thors have concluded that in wet or MQL conditions, better surface network and genetic algorithm to predict minimum surface
roughness can be achieved due to lower friction. However, reduced roughness in end-milling molded parts is presented in Oktem et al.
consumption of fluid is marked at NDM but also dangerous to (2006). Neural networks are used in Simunovic et al. (2013) to pre-
health because of the application of atomized mist or spray at dict surface roughness of face-milled aluminum alloy, while in Saric
micro-flood technology (Marksberry, 2007). Turning of AA5083-O et al. (2013) this method of artificial intelligence is used to predict
wrought aluminum alloy with a high Mg content (4.5%) in dry and simulate the surface roughness of S235JRG2 structural steel.
and wet machining conditions by the use of coated carbide tool is Finite element method (FEM) is applied to predict the stress of
investigated by Davoodi and Tazehkandi (2014). It is concluded that tools with microholes (Lei et al., 2009). These authors used solid
the investigated Al alloy can be machined in dry conditions with lubricant (tungsten disulfide).
high cutting speeds. In Sarikaya and Gullu (2014) Taguchi design Also applied is the LCA approach in which the effect of cooling/
and response surface methodology in CNC turning of AISI 1050 steel lubricating method is analyzed more extensively, in a broader way.
(the same material is investigated by Yalcin et al., 2009) are carried In Fratila (2010), workpiece material, scrap processing, use of
out to study the effect of cooling condition (dry, wet and MQL), lubrication, and energy consumption are considered. Energy, water
cutting speed, feed rate and depth of cut on surface roughness. A and land use, acidification, solid waste and global warming po-
conclusion is made on effectiveness of MQL and increasing the tential in cutting fluid production are studied in Pusavec et al.
quality of cutting operations. In Tosun and Pihtili (2010), optimi- (2010a, 2010b), taking into consideration usage amounts, con-
zation was performed, but in that case by the application of gray sumption rate and machine tool usage.
relational analysis while examining the face milling process of 7075 Based on the references to the reviewed literature published in
aluminum alloy, by varying spindle speed, feed rate, cooling tech- different journals, it can be concluded that this subject can be
nique, and cutting tool material, by measuring surface roughness approached from different standpoints, i.e., in a multidisciplinary
and material removal rate. Lower values of surface roughness were way, while not disregarding sustainable development. However, in
obtained for air cooling than for dry procedure, but higher than for this case, an excellent synergy is needed between educational and
fluid cooling, in investigating surface roughness and tool wear in research institutions and industry or government. It is noticed that
CNC milling of annealed AISI 1050 steel as demonstrated in Yalcin the field of sustainable development has to be introduced into the
et al. (2009). universities (Lozano et al., 2013).
Vegetable, mineral and synthetic oils as cutting fluids have been In view of the ever-increasing possibility of different machining
compared and used. The effect of vegetable based oil emulsions and centers to also use cooling/lubricating through the tool in addition
 K. Simunovic et al. / Journal of Cleaner Production 94 (2015) 321e329 323

to flood cooling/lubricating, the authors of this paper studied the

influence and practical importance of this input variable for the
surface quality of face-milled structural S235JRG2 steel, along with
the usual parameters of machining (depth of cut, number of revo-
lutions and feed rate). Only a limited number of papers addressing
the methods of cooling/lubricating through the tools were found,
i.e., the industrial application of a new technology is demonstrated
in the present paper.

2. Experimental work

2.1. Experimental details

In this paper, the examined material is structural steel EN 10025

S235JRG2 (the average surface hardness amounts to 127.3 HB,
whereas the average cross section hardness amounts to 123.7 HB),
whose chemical composition is presented in Table 1.
Samples of dimensions 100  70  20 mm (Fig. 1) were used in
the experiment.
The following equipment was used for the experiment:

- CNC vertical milling center with chip conveyor and automatic

tool changing, with the possibility of cooling/lubricating
through the tool, Fig. 2. Technical data are as follows: slideway
longitudinal (X axis) 2540 mm; slideway cross (Y axis) 815 mm;
slideway vertical (Z axis) 800 mm; maximal power 42 kW;
maximal spindle speed 5000 rev/min; palette size
3140  850 mm; control unit Siemens 840 D.
- angle milling head of 80-mm diameter of producer ISCAR (with
cutting inserts H490 ANKX) (Fig. 3)
Fig. 1. S235JRG2 steel samples used for the experimental work.
- roughness of machined surfaces measuring device e a contact
diamond stylus profiler Talysurf Surtronic duo by Taylor Hobson
- oil-water emulsion for cooling, chip flushing and lubricating
(soluble oil Schell Adrana D407 with 30% of mineral oil is used,
which is intended for the machining of steel, cast iron,
aluminum and brass; the pressure for cooling/lubricating
through the tool amounts to 1.3 MPa; the pressure for flood
cooling/lubricating amounts to 0.3 MPa).

2.2. Design of experiment

In the pre-experimental planning phase, trial runs were con-

ducted on which occasion the experimental technique, measuring
device, selection of the levels and range of the factors could be
tested. Acknowledging the afore mentioned relevant background in
the Introduction section, having in mind the aim of the experiment
(regression models and optimization) and considering the previous
experience in machining using the mentioned machine as well as
the performed pre-experiments, a decision was made to carry out
face-centered central composite design (CCD) e an experimental
design within response surface methodology for fitting response
surfaces (Montgomery, 2009), Fig. 4. For the face-centered CCD,
parameter a, the distance of the axial runs from the design center, is
1, so that every numeric factor has three levels.
The levels and ranges for the three numeric (quantitative) fac-
tors e depth of cut, feed rate and the number of revolutions and one
categoric (qualitative) factor e cooling/lubricating, Fig. 5 are shown Fig. 2. Machining setup for the cooling/lubricating through the tool.
in Table 2.

Table 1 In conducting the experiment, all of the three basic principles of

Chemical composition of steel grade S235JRG2. experimental design, i.e., randomization, replication and blocking
Element C Si Mn P S Cr Ni Mo
(Montgomery, 2009) were taken into consideration. Because the
experiment could not be performed in one day but in three days,
wt. % 0.21 0.287 0.64 0.021 0.018 0.116 0.093 0.018
the experiment was designed observing the blocking principle, i.e.,
324 K. Simunovic et al. / Journal of Cleaner Production 94 (2015) 321e329

Table 2
Levels and ranges of input variables (factors) used in experimental design.

Name Factor Units Low Center level High

level level

1 0 þ1

Depth of cut, a A mm 0.5 1 1.5

Feed, f B mm/min 100 300 500
Number of revolutions, n C rev/min 400 600 800
Cooling/lubricating D Dry Through the tool Flood

every day of the experiment represented one block. Blocks are used
to reduce, or eliminate, the variability caused by nuisance factors
that can affect the response, but do not concern us directly, as
design factors (Montgomery, 2009). In the investigation, the same
machine with the same operator was used; response measuring
was conducted by the same person, and the same charge of ma-
terial was applied so that there was no need for further blocking.

3. Results and discussion

3.1. Regression models

Fig. 3. Angle-milling head of 80-mm diameter used for the experimental work. Tables 3, 4 and 5 show the results of measured roughness for
conducted experiments in all three blocks. The last column repre-
sents average response for five repeated measurements of rough-
ness Ra.
The principle of randomization was also observed, which means
that the runs and allocation of the experimental material were both
randomly determined. In this way, the observations (or errors) are
independently distributed random variables (Montgomery, 2009),
and some impact of undesirable factors is reduced to a minimum
(Montgomery, 2009). It can be seen from Tables 3, 4 and 5, which
display the results of the conducted experiments, that the samples
were not machined in standard order (Std.) but randomly.
For this experiment, six replicates are made at the center point
for every level of the categoric factor, i.e., two replicates by every
level of the categoric factor in every block. Standard operating
conditions are applied for the center point in the design.
Included in this section of the paper are the empiric models
obtained on the basis of experimental results, as well as the sta-
Fig. 4. Scheme of face-centered central composite design for the three factors, which tistical hypothesis testing methodology aimed at comparing mean
are designated as x1, x2 and x3. surface roughness for three ways of cooling/lubricating.
Statistical analysis of the measured response e roughness Ra
(Tables 3, 4 and 5), the generation of run order, analysis of variance
and numerical optimization is performed by the licensed software

Fig. 5. Levels of categoric (qualitative) factor e cooling/lubricating.

 K. Simunovic et al. / Journal of Cleaner Production 94 (2015) 321e329 325

Table 3 Table 5
Experimental design and results obtained for day 1 (block 1). Experimental design and results obtained for day 3 (block 3).

Std Run Type Factor 1 Factor 2 Factor 3 Factor 4 Response Std Run Type Factor 1 Factor 2 Factor 3 Factor 4 Response
A: Depth of B: Feed, C: Number of D: Cooling/ Roughness, A: Depth B: Feed, C: Number D: Cooling/ Roughness,
cut, mm mm/min revolutions, lubricating mm of cut, mm mm/min of revolutions, lubricating mm
rev/min rev/min

44 1 Factorial 1.5 500 400 Flood 1.75 35 37 Axial 1 100 600 Through 0.65
50 2 Center 1 300 600 Flood 2.28 the tool
4 3 Factorial 1.5 500 400 Dry 1.2 60 38 Center 1 300 600 Flood 2.25
47 4 Factorial 0.5 500 800 Flood 1.48 33 39 Axial 0.5 300 600 Through 1.04
1 5 Factorial 0.5 100 400 Dry 0.46 the tool
41 6 Factorial 0.5 100 400 Flood 0.75 13 40 Axial 0.5 300 600 Dry 1.19
46 7 Factorial 1.5 100 800 Flood 0.69 56 41 Axial 1 500 600 Flood 2.03
30 8 Center 1 300 600 Through 1.35 17 42 Axial 1 300 400 Dry 1.89
the tool 19 43 Center 1 300 600 Dry 1.33
21 9 Factorial 0.5 100 400 Through 1.13 18 44 Axial 1 300 800 Dry 0.97
the tool 38 45 Axial 1 300 800 Through 1.03
24 10 Factorial 1.5 500 400 Through 0.97 the tool
the tool 20 46 Center 1 300 600 Dry 1.29
6 11 Factorial 1.5 100 800 Dry 0.59 40 47 Center 1 300 600 Through 1.11
9 12 Center 1 300 600 Dry 1.43 the tool
7 13 Factorial 0.5 500 800 Dry 0.49 58 48 Axial 1 300 800 Flood 1.97
10 14 Center 1 300 600 Dry 1.18 16 49 Axial 1 500 600 Dry 1.24
26 15 Factorial 1.5 100 800 Through 0.81 37 50 Axial 1 300 400 Through 1.41
the tool the tool
29 16 Center 1 300 600 Through 1.24 14 51 Axial 1.5 300 600 Dry 0.88
the tool 15 52 Axial 1 100 600 Dry 1.44
27 17 Factorial 0.5 500 800 Through 0.99 55 53 Axial 1 100 600 Flood 1.82
the tool 34 54 Axial 1.5 300 600 Through 0.98
49 18 Center 1 300 600 Flood 1.97 the tool
36 55 Axial 1 500 600 Through 1.12
the tool
39 56 Center 1 300 600 Through 1.41
Table 4 the tool
Experimental design and results obtained for day 2 (block 2). 53 57 Axial 0.5 300 600 Flood 1.73
54 58 Axial 1.5 300 600 Flood 2.38
Std Run Type Factor 1 Factor 2 Factor 3 Factor 4 Response
57 59 Axial 1 300 400 Flood 2.1
A: Depth of B: Feed, C: Number of D: Cooling/ Roughness,
59 60 Center 1 300 600 Flood 1.88
cut, mm mm/min revolutions, lubricating mm
rev/min

2 19 Factorial 1.5 100 400 Dry 0.84

22 20 Factorial 1.5 100 400 Through 1 Table 6
the tool Descriptive statistics parameters derived from the experimental results.
12 21 Center 1 300 600 Dry 1.15
Minimum roughness, mm 0.38
8 22 Factorial 1.5 500 800 Dry 0.47
Maximum roughness, mm 2.38
11 23 Center 1 300 600 Dry 1.27
Mean roughness, mm 1.28
32 24 Center 1 300 600 Through 1.59
Standard deviation of roughness, mm 0.52
the tool
25 25 Factorial 0.5 100 800 Through 0.63
the tool
51 26 Center 1 300 600 Flood 2.16 - cooling/lubricating
23 27 Factorial 0.5 500 400 Through 0.87
the tool
- interaction of a linear number of revolutions with cooling/
45 28 Factorial 0.5 100 800 Flood 0.63 lubricating
42 29 Factorial 1.5 100 400 Flood 0.81 - quadratic effects of depth of cut and feed rate.
5 30 Factorial 0.5 100 800 Dry 0.38
31 31 Center 1 300 600 Through 1.17
For these terms, p values for the F statistic (16.56; 7.9; 35.85;
the tool
3 32 Factorial 0.5 500 400 Dry 1.67 3.38; 11.40 and 10.17) are lower than the probability of type I error
52 33 Center 1 300 600 Flood 2.19 0.05, i.e., the significance level.
43 34 Factorial 0.5 500 400 Flood 1.37 The improved reduced regression model, in terms of coded
48 35 Factorial 1.5 500 800 Flood 1.74 factors (1, 0, 1), is represented as follows:
28 36 Factorial 1.5 500 800 Through 1.22
the tool
Ra ¼ 1:57 þ 0:05A þ 0:2B  0:14C  0:22D½1  0:2D½2
 0:06BD½1  0:1BD½2  0:18CD½1 þ 0:07CD½2  0:3A2
Design Expert (version DX8, 8.0.7.1, Stat-Ease, Inc. Minneapolis,
 0:3B2
2010). Some descriptive data are given in Table 6.
A significant second-order regression model (Equation (1)), with (1)
a non-significant lack of fit (p value of 0.0744 of lack of fit is higher
For this reduced regression model, which also includes non-
than 0.05, but it would be better if the difference was greater) is
significant terms to support hierarchy, the coefficient of multiple
derived with the following significant terms:
determination R2, which represents the portion of explained vari-
ability in total variability, is 0.77. The adjusted coefficient of
- feed rate
determination R2adj, which does not depend on the number of
- number of revolutions
variables added to the model, is 0.72. The ordinary and the adjusted
326 K. Simunovic et al. / Journal of Cleaner Production 94 (2015) 321e329

coefficients of determination do not differ considerably; therefore, Table 7

it can be concluded that the non-significant terms are probably not Average values of surface roughness (mm) calculated by the levels of factors.

included in the model. The prediction error sum of squares PRESS is Low level Center level High level
6.25; consequently, R2 for prediction in our experiment amounts to Depth of cut, mm 0.99 1.53 1.11
0.56. These two values are the measures of predicting the response Feed rate, mm/min 0.84 1.53 1.24
in a new experiment (Montgomery, 2009). The coefficient of vari- Number of revolutions, rev/min 1.21 1.49 0.94
ation C.V. % is the portion of the standard deviation in the mean and Cooling/lubricating 1.07 1.09 1.7

is equal to 20.9%.
For dry machining conditions, roughness Ra, in dependence on
parameters, is represented by Equation (2) in terms of natural At center levels of a particular numerical factor there is the highest
factors, the same as for cooling/lubricating through the tool number of parameter combinations with average (24 in all) and
(Equation (3)) and for flood cooling/lubricating (Equation (4)). high feed rate (three in all) and with small (three in all) and average
number of revolutions (24 in all).
Ra ¼ 0:16 þ 2:5a þ 0:005f  0:0016n  1:2a2  0:000007f 2 As the maximum average values of surface roughness are ob-
(2) tained by these combinations, otherwise possible by high feed rates
and smaller number of revolutions, the authors thus prove that the
center level of feed rate (300 mm/min) functions as a high level of
Ra ¼ 0:5 þ 2:5a þ 0:005f  0:0004n  1:2a2  0:000007f 2 feed rate (500 mm/min), i.e. it affects the increase of roughness.
(3) However, for the center level of number of revolutions (600 rev/
min), considered as an important factor, it can be concluded that it
behaves as a small number of revolutions (400 rev/min), thus
Ra ¼ 0:42 þ 2:5a þ 0:006f  0:00014n  1:2a2  0:000007f 2
affecting the increase of roughness. At center levels of factors there
(4) is, simultaneously, a small number of combinations with the min-
Fig. 6 shows the relationship between the model calculated imum feed rate (three in all) and the highest number of revolutions
values (predicted) and those obtained by the experiment (actual). (three in all) to exert any considerable influence on decreasing the
Model diagnostics (normal probability plot of residuals; the average value of roughness.
relationship between residuals and input factors, run order and After the analysis of average values of roughness for center
fitted values) are also performed and these diagnostics indicate that levels of factors, indicative of surface roughness increase, the re-
the residuals are structureless, that is, they are not affected by the sults for low and high levels of numerical factors will also be
previously mentioned variables. For the sake of brevity, these plots discussed.
are not presented in the paper. Additionally, the influence of a There is no considerable difference among average values of
particular run on the fitted value is tested, and none of the data roughness for the low and high level of the “depth of cut” factor. In
exert a strong influence on the regression model. other words, surface roughness does not depend on depth of cut, as
In order to even better explain the study results, it is useful to already proven earlier (factor “depth of cut” is not important).
show the average responses at each level of a particular factor At a low level of important factor “feed rate” (100 mm/min),
(Montgomery, 2009), i.e. the marginal response averages at the average roughness is lower than average roughness for high level
levels of the four factors. This is displayed in Table 7 in which (500 mm/min), i.e. better quality of machined surface is obtained
average values of surface roughness are given when the study re- by lower feed rates.
sults are sorted by the levels of a particular factor. Contrary to that, at a low level of the “number of revolutions”
It can be noticed that the highest average values of surface factor (400 rev/min), average roughness is higher than average
roughness (bolded) appear for center level (intermediate values) of roughness for high level (800 rev/min), i.e. higher number of rev-
numerical factors (depth of cut, feed rate and number of revolu- olutions result in better quality of machined surface.
tions). Average values of roughness at center level are even higher It is obvious that the average roughness values for dry
than the average levels at low and high level of the factors, thus machining and cooling/lubricating through the tool are nearly the
proving that the introduction of center point is useful for discov- same, and that the maximum value of average roughness is for
ering the curvature in response surface (Montgomery, 2009). Au- flood cooling/lubricating. By the application of the hypothesis
thors of this paper explain this fact physically in the following way. testing procedure which includes not only average values of
roughness but also standard deviation and number of data,
methods of cooling/lubricating will be additionally compared
(Section 3.2.).

3.2. Statistical hypothesis testing

It is evident from the previous section that cooling/lubricating is

a significant factor, i.e., that the change of levels of this factor has a
considerable effect on surface roughness. In this section, the pro-
cedure called statistical hypothesis testing is applied for comparing
the mean surface roughness of the different ways of cooling/
lubricating.
As already mentioned, the average values of roughness for dry
machining and cooling/lubricating through the tool do not differ,
and the highest value of average roughness is for flood cooling/
lubricating. For these three methods of machining standard de-
Fig. 6. Values of roughness Ra calculated by the model e predicted (Equation (1)) viations of data were also calculated, being 0.41 mm, 0.24 mm and
versus the values obtained by measuring e actual. 0.55 mm for dry machining, cooling/lubricating through the tool
 K. Simunovic et al. / Journal of Cleaner Production 94 (2015) 321e329 327

Table 8
Results of the two-sample t-tests obtained by the hypothesis testing procedure.

Comparison of means t-test t statistic t critical Decision

Dry e cooling/lubricating Assuming unequal 0.16 2.04 t statistic value is within the interval [t critical, þt critical]; the null hypothesis can
through the tool variances be accepted
Dry e flood cooling/lubricating Assuming equal 4 2.02 t statistic value is out of the interval [t critical, þt critical], the alternative hypothesis
variances can be accepted
Cooling/lubricating through Assuming unequal 4.45 2.05 t statistic value is out of the interval [st critical, þt critical], the alternative hypothesis
the tool e flood cooling/ variances can be accepted
lubricating

and flood cooling/lubricating, respectively. The largest deviations - depth of cut 1.04 mm
are for flood cooling/lubricating, and the lowest for cooling/lubri- - feed rate 100 mm/min
cating through the tool. This shows that flood cooling is not a - number of revolutions 800 rev/min
uniform procedure, opposite to cooling/lubricating through the - dry machining.
tool where more uniform results were obtained.
The null hypothesis is established, according to which, the mean Achieved roughness with the least transmitted variability of
surface roughness for two methods of cooling/lubricating is the 0.27 mm is 0.62 mm (Fig. 9).
same, which is the opposite of the alternative hypothesis. Finally, the confirmation runs are performed to test the optimal
Table 8 depicts the two-sample t-test results. To compare the parameters, that is, to predict a new response observation at the
dry conditions-cooling/lubricating through the tool, the same as point x0. The confirmation runs are performed with the following
the cooling/lubricating through the tool-flood cooling/lubricating, optimal levels of factors:
the t-test was applied assuming unequal variances, because the F-
test proved that the sample variances were different. - depth of cut e 1.04 mm
Based on Table 8, it can be concluded that the mean surface - feed rate e 100 mm/min
roughness obtained in dry cutting conditions and in cooling/ - number of revolutions e 800 rev/min
lubricating through the tool does not differ, whereas there is a - cooling/lubricating e dry conditions.
considerable difference between the mean surface roughness in dry
conditions and in flood cooling/lubricating, which is the same According to the Equation (5) (Montgomery, 2009), a 100 (1-a) %
result as in the cooling/lubricating through the tool and flood prediction interval for a new observation y0 can be calculated.
cooling/lubricating.
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 ffi
b b 2 1 þ x0 ðX 0 XÞ1 x0  y0
0
3.3. Single- and multiple-goal optimization of parameters y ðx0 Þ  ta=2;mr s
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 ffi
The following is performed in this section: b s 2 1 þ x0 ðX XÞ1 x0
0 0
y ðx0 Þ þ ta=2;mr b (5)

- calculation of standard deviation of the variability in response

transmitted from the factors, i.e., propagation of error (POE) where
- single goal numerical optimization (aimed at obtaining optimal
parameters for achieving minimum surface roughness) b
y ðx0 Þ e the value of response predicted by the regression model
- multiple goal numerical optimization (aimed at obtaining at new observation point x0
optimal parameters for achieving simultaneously both mini- ta/2,mr e the value of variable of t distribution with m-r degrees
mum surface roughness and minimum standard deviation of of freedom, for defined probability a of the type I error
the variability in response transmitted from the factors, i.e., m e the number of runs
propagation of error). r e the number of regression model terms
s 2 e unbiased estimator of variance
b
Based on the pre-experimental research and the practical op- X e the matrix of the levels of the independent variables
0

erator's experience and knowledge in the area of machining, the X e the transpose of the matrix X.
estimated standard deviations for the numeric factors are as
follows:

- estimated standard deviation of depth of cut: 0.01 mm

- estimated standard deviation of feed rate: 0.5 mm/min
- estimated standard deviation of the number of revolutions:
2.5 rev/min.

Fig. 7 shows the response surface plot of the factor variation

transmitted to response (for flood cooling/lubricating and for the
depth of cut, which is 1 mm).
With the depth of cut 0.5 mm, feed rate 100 mm/min and
number of revolutions 800 rev/min in dry machining, a minimum
surface roughness of 0.27 mm would be obtained; thus, for the
single optimization, these parameters would be optimal (Fig. 8).
With the simultaneous minimum propagation of error and Fig. 7. Response surface plot of the factor variation transmitted to response for flood
minimum roughness, optimal parameters would be: cooling/lubricating and depth of cut of 1 mm.
328 K. Simunovic et al. / Journal of Cleaner Production 94 (2015) 321e329

wear (just opposite to the conventional wisdom), while the wet

cutting conditions resulted in the worst surface finish (the highest
values of surface roughness) as in the present paper. Bruni et al.
(2006) investigated the effect of the type of coolant/lubricant,
cutting inserts technology and machine bed material (cast iron and
polymer concrete) on the quality of the treated surface and tool
wear in finish turning of AISI 420B martensitic tempered stainless
steel.
The present study demonstrated that the best surface quality
was achieved by dry machining (without cooling/lubricating).
However, the treated surface of the sample was sufficiently small
that the machining time was short and there was no significant
heating of the tool. In the case of longer-treated surfaces (or using
several passes of machining), this would likely lead to more
intensive heating of the tool. Consequently, the developed heat
would affect the cutting inserts and tool life, and the surface
Fig. 8. Response surface plot for minimum surface roughness for dry cutting and depth roughness may thus be higher. Hence, the machining of small
of cut of 0.5 mm. workpieces (short-term machining) can be considered an envi-
ronmentally friendly milling process. The same as the authors of
the present paper, Lei et al. (2009) also conclude that they will
include longer cuts and tool life in their further investigations. Tool
cooling from within, which means a specially designed tool, was
taken into consideration in investigations by those authors, where
the cutting inserts with microholes filled with liquid and solid
lubricant were compared with dry and flood cutting of mild steel.
The formation of patches on the tool edge, which positively af-
fects the surface roughness, can be prevented by cooling/lubri-
cating through the tool. Because the cooling lubricant is applied
under high pressure (1.3 MPa) in the present study, it suppresses
the formed chips from the tool and protects the tool edge. As a
result, the tool is always in contact with a clean surface (without
chips). Lei et al. (2009) also interconnect high-pressure of coolant/
lubricant and prevention of chip material suppressing. For better
performance, these authors suggested high pressure lubrication
through the tool's microholes to prevent or minimize blocking the
holes at the cutting edge caused by the chip material. They also
Fig. 9. Contour plot for minimum surface roughness and minimum propagation of conclude that conventional flood cooling has no such effects as
error for dry machining and number of revolutions of 800 rev/min.
micropool lubrication from within because it cannot penetrate to
the chipetool interface.
Finally, the worst surface quality is obtained by flood cooling/
A 95% interval, within which the new observation would be lubricating. The cooling is not uniform, and the cooling lubricant
expected to lie, is from 0.31 to 0.94 mm. The actual measured suppresses chips on the tool edge (flood cooling/lubricating pres-
average value of surface roughness in confirmation experiments sure amounts to at most 0.3 MPa and is often significantly less).
was 0.54 mm. It lies within the calculated interval. Thus, the removed chips are retained longer in the cutting zone,
It can be concluded that the best surface quality with minimum causing additional undesirable vibrations that affect the increase in
propagation of error can be achieved with the minimum feed rate surface roughness.
and the highest number of revolutions for dry cutting condition.
Similar conclusions are drawn in the work by Fratila and Caizar 4. Conclusions
(2011). These authors concluded that minimum depth of cut,
minimum feed rate, maximum cutting speed, and minimum Response surface methodology is applied for modeling and
lubricant flow rate imply a smaller Ra. Fratila and Caizar (2011, optimizing the face milling of S235JRG2 structural steel. Based on
2012) investigated face milling of AlMg3, varying cutting speed, experimental data and by the application of the least square
feed rate, depth of cut and the quantity of lubrication analyzing the method, the obtained empirical models indicate that out of the four
temperature variations in the cutting zone and the surface quality parameters, the feed rate, the number of revolutions and the
i.e. surface roughness. cooling/lubricating exert a significant effect on surface roughness.
The depth of cut is a non-significant factor, but it is found in
3.4. Discussion regression models due to the promotion of the hierarchy of the
models.
A number of authors have concluded that in wet or MQL con- It is established that almost the same surface roughness of the
ditions, better surface roughness can be achieved due to lower treated area is obtained for dry machining and cooling/lubricating
friction. However, the authors of the present paper have concluded through the tool. However, as this was a case of small treated sur-
that lower values of surface roughness are obtained in dry cutting, faces, it is pointed out that in dry machining heating of the tool in a
which is also supported by Bruni et al. (2006), and which is not short time did not occur. Therefore, further investigations could
quite unusual. In Bruni et al. (2006), it was demonstrated that the include longer-treated surfaces (or using several passes of
method of cooling/lubricating had no considerable effect on tool machining) and an analysis of tool wear in addition to the surface
 K. Simunovic et al. / Journal of Cleaner Production 94 (2015) 321e329 329

roughness measurement with the main aim of defining the optimal Fratila, D., 2010. Macro-level environmental comparison of near-dry machining and
flood machining. J. Clean. Prod. 18, 1031e1039. http://dx.doi.org/10.1016/
parameters to achieve minimum roughness, minimum transmitted
j.jclepro.2010.01.017.
variability in the response and minimum tool wear. Fratila, D., Caizar, C., 2011. Application of Taguchi method to selection of optimal
It is important to notice the role of the pressure applied in lubrication and cutting conditions in face milling of AlMg3. J. Clean. Prod. 19,
cooling/lubricating through the tool which was higher than the 640e645. http://dx.doi.org/10.1016/j.jclepro.2010.12.007.
Fratila, D.F., Caizar, C., 2012. Assessment of cooling effect and surface quality to face
pressures in flood/cooling lubricating for which the highest values milling of AlMg3 using several cooling lubrication methods. Mater. Manuf.
of surface roughness are achieved due to suppressing chips on the Process. 27, 291e296. http://dx.doi.org/10.1080/10426914.2011.577864.
tool edge and thus higher vibrations. Kuram, E., Ozcelik, B., Bayramoglu, M., Demirbas, E., Simsek, B.T., 2013. Optimization
of cutting fluids and cutting parameters during end milling by using D-optimal
So, as it is demonstrated that the center level of factor “feed rate” design of experiments. J. Clean. Prod. 42, 159e166. http://dx.doi.org/10.1016/
affects roughness as a high feed rate, while the center level of factor j.jclepro.2012.11.003.
“number of revolutions” behaves as a small number of revolutions, Lawal, S.A., Choudhury, I.A., Nukman, Y., 2014. Evaluation of vegetable and mineral
oil-in-water emulsion cutting fluids in turning AISI 4340 steel with coated
the range between the higher and lower levels for these two factors carbide tools. J. Clean. Prod. 66, 610e618. http://dx.doi.org/10.1016/
can be reduced in further investigations. j.jclepro.2013.11.066.
Because the experimental data were obtained in a real industrial Lei, S., Devarajan, S., Chang, Z., 2009. A study of micropool lubricated cutting tool in
machining of mild steel. J. Mater. Process. Technol. 209, 1612e1620. http://
environment and an impetus was given by industry experts dx.doi.org/10.1016/j.jmatprotec.2008.04.024.
(focused on reducing the cooling/lubricating fluid consumption and Lozano, R., Lozano, F.J., Mulder, K., Huisingh, D., Waas, T., 2013. Advancing higher
thus the effect on the environment), the authors of the current education for sustainable development: international insights and critical re-
flections. J. Clean. Prod. 48, 3e9. http://dx.doi.org/10.1016/j.jclepro.2013.03.034.
paper hypothesize that the present study and the insights yielded
Marksberry, P.W., 2007. Micro-flood (MF) technology for sustainable manufacturing
will be useful for companies with modern centers for the operations that are coolant less and occupationally friendly. J. Clean. Prod. 15,
machining of widely applicable structural steels. Applicability of 958e971. http://dx.doi.org/10.1016/j.jclepro.2006.01.006.
dry cutting is possible for small surfaces or short-term machining Minton, T., Ghani, S., Sammler, F., Bateman, R., Furstmann, P., Roeder, M., 2013.
Temperature of internally-cooled diamond-coated tools for dry-cutting tita-
while for longer-term machining, cooling/lubricating through the nium. Int. J. Mach. Tools. Manuf. 75, 27e35. http://dx.doi.org/10.1016/
tool can be applied for the investigated steel type. In this way the j.ijmachtools.2013.08.006.
application of flood/cooling lubricating is avoided as a technology Montgomery, D.C., 2009. Design and Analysis of Experiment, seventh ed. John
Wiley and Sons Inc, Hoboken.
that uses a great quantity of cooling lubricant and affects the Oktem, H., Erzurumlu, T., Erzincanli, F., 2006. Prediction of minimum surface
pollution of environment and the increase of costs. The technolo- roughness in end milling mold parts using neural network and genetic algo-
gists use these findings as a basis for support in deciding on which rithm. Mater. Des. 27, 735e744. http://dx.doi.org/10.1016/j.matdes.2005.01.010.
Pusavec, F., Krajnik, P., Kopac, J., 2010a. Transitioning to sustainable production e
tools and methods of cooling to select (according to the earlier part I: application on machining technologies. J. Clean. Prod. 18, 174e184. http://
generated technological documentation, i.e. according to the times dx.doi.org/10.1016/j.jclepro.2009.08.010.
of machining for particular technological processes). The present Pusavec, F., Kramar, D., Krajnik, P., Kopac, J., 2010b. Transitioning to sustainable
production e part II: evaluation of sustainable machining technologies. J. Clean.
study also presents a contribution to cleaner manufacturing Prod. 18, 1211e1221. http://dx.doi.org/10.1016/j.jclepro.2010.01.015.
because the parameters used in both the single and dual optimi- Saric, T., Simunovic, G., Simunovic, K., 2013. Use of neural networks in prediction
zations were obtained under dry cutting conditions. and simulation of steel surface roughness. Int. J. Simul. Model. 12, 225e236.
http://dx.doi.org/10.2507/IJSIMM12(4)2.241.
Sarikaya, M., Gullu, A., 2014. Taguchi design and response surface methodology
References based analysis of machining parameters in CNC turning under MQL. J. Clean.
Prod. 65, 604e616. http://dx.doi.org/10.1016/j.jclepro.2013.08.040.
Aggarwal, A., Singh, H., Kumar, P., Singh, M., 2008a. Optimizing power consumption Sayuti, M., Erh, O.M., Sarhan, A.A.D., Hamdi, M., 2014. Investigation on the
for CNC turned parts using response surface methodology and Taguchi's morphology of the machined surface in end milling of aerospace Al6061-T6 for
technique-A comparative analysis. J. Mater. Process. Technol. 200, 373e384. novel uses of SiO2 nanolubrication system. J. Clean. Prod. 66, 655e663. http://
http://dx.doi.org/10.1016/j.jmatprotec.2007.09.041. dx.doi.org/10.1016/j.jclepro.2013.11.058.
Aggarwal, A., Singh, H., Kumar, P., Singh, M., 2008b. Optimization of multiple quality Simunovic, G., Simunovic, K., Saric, T., 2013. Modelling and simulation of surface
characteristics for CNC turning under cryogenic cutting environment using roughness in face milling. Int. J. Simul. Model. 12, 141e153. http://dx.doi.org/
desirability function. J. Mater. Process. Technol. 205, 42e50. http://dx.doi.org/ 10.2507/IJSIMM12(3)1.219.
10.1016/j.jmatprotec.2007.11.105. Thepsonthi, T., Hamdi, M., Mitsui, K., 2009. Investigation into minimal-cutting-fluid
Bruni, C., Forcellese, A., Gabrielli, F., Simoncini, M., 2006. Effect of the lubrication- application in high-speed milling of hardened steel using carbide mills. Int. J.
cooling technique, insert technology and machine bed material on the work- Mach. Tools Manuf. 49, 156e162. http://dx.doi.org/10.1016/
part surface finish and tool wear in finish turning of AISI 420B. Int. J. Mach. j.ijmachtools.2008.09.007.
Tools Manuf. 46, 1547e1554. http://dx.doi.org/10.1016/ Tosun, N., Pihtili, H., 2010. Gray relational analysis of performance characteristics in
j.ijmachtools.2005.09.007. MQL milling of 7075 Al alloy. Int. J. Adv. Manuf. Technol. 46, 509e515. http://
Cetin, M.H., Ozcelik, B., Kuram, E., Demirbas, E., 2011. Evaluation of vegetable based dx.doi.org/10.1007/s00170-009-2118-4.
cutting fluids with extreme pressure and cutting parameters in turning of AISI Yalcin, B., Ozgur, A.E., Koru, M., 2009. The effects of various cooling strategies on
304L by Taguchi method. J. Clean. Prod. 19, 2049e2056. http://dx.doi.org/ surface roughness and tool wear during soft materials milling. Mater. Des. 30,
10.1016/j.jclepro.2011.07.013. 896e899. http://dx.doi.org/10.1016/j.matdes.2008.05.037.
Davoodi, B., Tazehkandi, A.H., 2014. Experimental investigation and optimization of Zhang, S., Li, J.F., Wang, Y.W., 2012. Tool life and cutting forces in end milling Inconel
cutting parameters in dry and wet machining of aluminum alloy 5083 in order 718 under dry and minimum quantity cooling lubrication cutting conditions.
to remove cutting fluid. J. Clean. Prod. 68, 234e242. http://dx.doi.org/10.1016/ J. Clean. Prod. 32, 81e87. http://dx.doi.org/10.1016/j.jclepro.2012.03.014.
j.jclepro.2013.12.056.