1 Introduction

Face milling is the secondary process used for the cutting and finishing of parts. The process is employed for the fabrication of tooling for other processes with high accuracy in a suitable processing time [1] and also used in the parts of aerospace and automotive industries, where quality is the prime factor [2]. Surface roughness is the measure of quality in machined parts, highly depends upon controlling parameters: Fr, cutting speed, and Dc [3]. There is a need to optimize these controlling parameters to enhance the quality of parts as Ra is sensitive to Fr [4, 5]. It was observed that the parameters other than feed (axial and radial Dc and cutting speed) are also significant factors that affect the Ra [6]. Material removal rate is the second major measure, which need to be maximized without compromising the surface quality of the part [7]. The factors that contribute in achieving maximum MRR are Fr and Dc [8]. High Fr results in higher MRR but it also increase the cutting temperature which reduces the tool life [9] and surface quality [10]. Beside input parameters, cutting fluid has the prime importance in metal cutting processes for enhancing the quality of the workpiece by lubricating the tool. Workpiece interface lubrications (cutting fluid) improve the machining characteristics and workpiece’s surface quality [11]. But flooded cooling increases the manufacturing cost by 16% [12]. To minimize the coolant cost, MQL technique was introduced. Researchers focused on MQL technique as it reduces the consumption of lubricant by sprinkling the blend of air and lubricant and as environmental friendly [13]. In MQL, coolant mixed with compressed air is sprayed at the tool-workpiece interface at a low flow rate [14]. It is reported that the MQL reduces the coolant consumption 3 times than flooded cooling using flow rate 50–500 ml/h approximately. It is investigated that MQL produces superior surface finish than other conventional methods of lubrication [15, 16]. MQL not only improves the surface finish but also reduces cutting temperature and enhances the tool life by reducing the flank wear [17, 18] as it provides excess amount of oxygen at tool-workpiece interface forming the protective oxide layer [19]. But for material removal rate, it is evident that MQL did not show significant improvement for medium carbon steel [20].

To enhance the efficiency of MQL, nanoparticles are contaminated in the fluid. The addition of nanoparticles improves the surface quality [21]. It is claimed that 46% reduction in surface roughness was achieved as compared with the commonly used lubricant [12]. In nanofluid, particles form a lubrication film and fill the surface cavities which polish the surface and improve the surface quality [22, 23]. It is noticed that nanofluid exhibits better Ra as compared with base fluid [24]. However, increasing of nanoparticles concentration results in reduced Ra since the excess nanoparticles concentration enhances the viscosity of cutting fluid and fills the surface pores. Incoming nanoparticles shear off the existing ones and other ploughed off particles remain stuck on exfoliated film in tool-workpiece interface. Therefore, increased nanoparticles concentration may negatively influence the surface quality [25]. But the percentage improvement is different for soft and hard material while using MQL and nanofluid. And there is a need to study the trends of the input variables/parameters in different cutting conditions.

Number of statistical and mathematical techniques including response surface methodology, factorial design, Taguchi method, genetic algorithm, fuzzy logic, and artificial neural network have been used [26,27,28,29,30,31,32]. Among these techniques, genetic algorithm, fuzzy logic, and artificial neural network are the soft computing, while response surface methodology, factorial design, and Taguchi method are statistical techniques. Soft computing techniques also have the ability to predict the response measures, and repetitive hit and trial is used for prediction. However, statistical techniques require less number of experiments for the prediction and optimization. Therefore, Taguchi method has been observed with less number of experiments and also considered as cost-effective [33]. After the selection of suitable experimental design, various analysis techniques including multi-criteria decision-making (MCDM) analysis, grey relational analysis (GRA), and analysis of variance (ANOVA) have also been used by the researchers in order to optimize the response measures [34,35,36].

Industrial sector is still in efforts to obtain a proper combination of workpiece material, lubrication, and nanoparticles, which are highly efficient and inexpensive for the machining of specific material. For example, a lubrication and nanoparticles used for the machining of hard material can be expensive for the softer materials. The alloy chosen for this research is commonly used in machining application due to its high strength and corrosion resistance. However, little study is observed on investigating the effects of these cutting conditions on Ra along with MRR for this specific aluminum alloy in face milling process. Therefore, this study aims to analyze the impact of dry, MQL, and nanofluid cutting conditions on Ra and MRR. The influence of three effective input variables including Fr, Vs, and Dc has been investigated using Taguchi method.

The rest of this paper is organized as follows. Section 2 explains the experimental procedure for the machining of aluminum specimens with dry, MQL, and nanofluid medium and the Taguchi method used for the design of experiment. Investigated results, analysis using ANOVA, and optimization through contour plots have been presented in Section 3. Finally, conclusions of overall research and recommendations for the future study are discussed in Section 4.

2 Experimental procedure

This section gives a detail about material composition, experimental setup, sample preparation, and response measurements. Aluminum alloy 6082-T6 selected for the machining purpose having mechanical properties and composition is given in Tables 1 and 2, respectively. Optical emission spectrometer is operated to check the chemical composition of workpiece material used. Machining of workpiece material highly depends upon its mechanical properties. Machinability refers to the ease of cutting of metal and allowing the removal of material swiftly. Materials with superior machinability require less cutting power and time. Machinability of aluminum is considered to be excellent in terms of achieving minimum Ra and maximum MRR. To further improve the machinability of the selected workpiece material, different parameters and cutting conditions have been adopted. From the previous research, it has been observed that Fr, Vs, and Dc are the most effective input parameters for Ra and MRR [38,39,40]. Therefore, Fr, Vs, and Dc are used as process variables in three different cutting conditions: dry, MQL, and nanofluid. Fr is the linear motion of tool throughout the machining. Vs is the rotational motion of tool, and Dc is controlled by the vertical movement of the tool. Commercial soluble oil is used as lubricant in MQL machining. A system has been developed for the delivery of MQL between the tool and workpiece interface. Air pressure gun attached with a compressor has been employed to throw air-lubricant mist using 5-bar pressure. The oil mist has been produced inside the tank using air pressure. Flow rate of aerosol has been kept constant at 400 ml/h. In nanofluid, Al2O3 nanoparticles of size 80 μm mixed with soluble oil in 5% by weight ratio were used with the same flow rate.

Table 1 Properties of Al6082-T6 [37]
Full size table
Table 2 Chemical composition of aluminum alloy (Al-6082)
Full size table

Machining process is performed using NC MIKRON WF21C milling machine having maximum Vs of 4000 rpm as shown in Fig. 1a. Workpiece is clamped using fixtures to avoid any vibration and distortion during the machining. To avoid the effect of machine-tool fixture environment on the machining rate and quality, all workpieces are clamped with same type and number of fixtures. Use of proper machine tool fixtures enables experimental process to present the true effects of input parameters and cutting conditions. Cutting tool made of HSS with 16 mm diameter is employed. The specimens prepared for the face milling process along with dimensions are shown in Fig. 1 b and c, respectively. Surface roughness of the specimens was measured through surface roughness tester, shown in Fig. 2.

Fig. 1
figure 1

a Face milling process of aluminum specimen, b machined specimens, and c dimensions of specimen prepared for machining

Full size image
Fig. 2
figure 2

Surface roughness measuring apparatus

Full size image

2.1 Experimental design

Taguchi design is a robust technique used for the optimization of input variables and reduces the process variation. The technique uses signal-to-noise (S/N) ratio as quality characteristic measurement [41]. Using S/N ratios, Taguchi empirically found the two-stage optimization process indeed offers the optimum level combination while keeping mean on target, minimizing the standard deviation [42]. S/N ratio is beneficial in improvement of measurement and improving quality through variability reduction. Properties of S/N can be classified in three categories:

$$ \mathrm{Smaller}\ \mathrm{the}\ \mathrm{better}\ \mathrm{property};\mathrm{S}/\mathrm{N}=-10\log \frac{1}{n}\left(\sum {Y}^2\right) $$
(1)
$$ \mathrm{Nominal}\ \mathrm{the}\ \mathrm{best}\ \mathrm{property};\mathrm{S}/\mathrm{N}=10\ \log \frac{Y-}{S_y^2} $$
(2)
$$ \mathrm{Larger}\ \mathrm{the}\ \mathrm{better}\ \mathrm{property};\mathrm{S}/\mathrm{N}=10\log \frac{1}{n}\left(\sum \frac{1}{Y^2}\right) $$
(3)

where Y is mean of all the observed values, Sy2 is variance of y, y is observed data, and n depicts number of observed values. Smaller the better case was employed for Ra and larger the better for MRR. Fr, Vs, and Dc have been specified as input variables due to their remarkable impact on machining properties [43,44,45,46]. The input variables along with selected levels are shown in Table 3.

Table 3 Face milling input variables with levels
Full size table

Nine experiments were performed using Taguchi method L934 array for each cutting condition. Experiments were performed at each level of the input variables to measure the response variables included Ra and MRR. Three different readings of the machined surface have been taken for each specimen and their mean value is considered as final reading. Mean and standard deviation of the measured values against each experimental run have been presented in Table 4. MRR is calculated using Eq. 4 [47].

$$ \mathrm{MRR}=\frac{\left({W}_1-{W}_2\right)}{t\times \rho } $$
(4)
Table 4 Design matrix with observed responses
Full size table

Here W1 is initial weight of a specimen before machining, W2 is final weight after machining, ρ is density of workpiece material, and t is machining time.

3 Results and discussion

3.1 Graphical representation using signal-to-noise ratio approach

Taguchi’s S/N ratio represents the response or quality characteristics, and the largest S/N ratio value is desirable. S/N ratio was used for selecting the best combination of input variables to achieve optimum response. Average values of signal-to-noise ratio for Ra of different variables at their levels are shown in Table 5 and represented in graphical form in Fig. 3. The peak values of S/N ratio of control variables were selected representing the optimum conditions for the Ra. S/N ratio is also used to rank the input parameters on the basis of their contribution in Ra. Fr is the highest contributing factor followed by the Vs and Dc. At higher values of Fr and Dc and lower values of Vs, deformed chip cross section and volume and sharp and brittle fractures occur on the machining surface which increases the surface roughness [48, 49]. In dry condition, Ra is found to be minimum at low level of Fr and Dc and high level of Vs. Similar trends are observed for MQL and nanofluid cutting conditions as shown in Fig. 3 b and c, respectively.

Table 5 Average values of S/N ratios of Ra at different levels
Full size table
Fig. 3
figure 3

S/N ratio graph showing the effects of Fr, Vs, and Dc on Ra for a dry, b MQL, and c nanofluid

Full size image

Average values of signal to noise ratio for MRR are given in Table 6, and graph for the S/N ratio are presented in Fig. 4. Ranking of input parameters on the basis of their S/N ratio shows that Fr is the most contributing input parameter for MRR, while Dc and Vs are the second and third most contributing input parameters, respectively, and same can be seen in Fig. 4. MRR is maximum at high level of Fr and Dc and mid-level of Vs.

Table 6 Average values of S/N ratio of MRR at different levels
Full size table
Fig. 4
figure 4

S/N ratio graph showing the effects of Fr, Vs, and Dc on MRR for a dry, b MQL, and c nanofluid

Full size image

3.2 Analysis of results through analysis of variance

For the modeling of response variables, regression analysis is performed using statistical software Minitab. Analysis of variance (ANOVA) is employed to test the adequacy of the developed models.

3.2.1 ANOVA for Ra

ANOVA results declared that the effects of input variables, Fr, Vs, and Dc associated with Ra, were significant for dry conditions. Same input variables were obtained significant for MQL and nanofluid cutting. ANOVA results along with adequacy measures R2, adjusted R2, and predicted R2 values are provided in Table 4. The results demonstrate that the regression models are significant having p value less than 0.05. Adequacy measures R2, R2 (adjusted), and R2 (predicted) values for dry, MQL, and nanofluid conditions are close to one, indicating the adequacy of models. For the prediction of Ra, empirical models for dry, MQL, and nanofluid conditions are presented in Eqs. 5, 6, and 7, respectively.

$$ \mathrm{Ra}\ \left(\mathrm{Dry}\right)=0.801+0.001021\times \mathrm{Fr}-0.000363\times \mathrm{Vs}+0.684\times \mathrm{Dc} $$
(5)
$$ \mathrm{Ra}\ \left(\mathrm{MQL}\right)=0.561+0.000737\times \mathrm{Fr}-0.000267\times \mathrm{Vs}+0.540\times \mathrm{Dc} $$
(6)
$$ \mathrm{Ra}\ \left(\mathrm{Nanofluid}\right)=0.472+0.000653\times \mathrm{Fr}-0.000313\times \mathrm{Vs}+0.486\times \mathrm{Dc} $$
(7)

Percentage contribution of each factor in Ra for all cutting conditions extracted from the ANOVA tables has been presented as pie charts in Fig. 5a. Fr is the most contributing factor for Ra with percentage contribution of 60%. Percentage contribution of Vs and Dc are 30% and 7%, respectively.

Fig. 5
figure 5

Pie chart of percentage contributions for a Ra and b MRR

Full size image

3.2.2 ANOVA for MRR

The input parameters that significantly influence in MRR include Fr and Dc. Models developed for MRR are significant with p value less than 0.05 as shown in Table 7. Adequacy measures R2, R2 (adjusted), and R2 (predicted) are close to one, showing the adequacy of the models. For the prediction of MRR, empirical models for dry, MQL, and nanofluid are developed and presented in Eqs. 8, 9, and 10, respectively.

$$ \mathrm{MRR}\ \left(\mathrm{Dry}\right)=-116.2+0.1508\times \mathrm{Fr}-0.0200\times \mathrm{Vs}+302.2\times \mathrm{Dc} $$
(8)
$$ \mathrm{MRR}\ \left(\mathrm{MQL}\right)=-119.3+0.1515\times \mathrm{Fr}-0.0194\times \mathrm{Vs}+303.3\times \mathrm{Dc} $$
(9)
$$ \mathrm{MRR}\ \left(\mathrm{Nanofluid}\right)=-116.0+0.1507\times \mathrm{Fr}-0.0203\times \mathrm{Vs}+303.3\times \mathrm{Dc} $$
(10)
Table 7 Analysis of variance for Ra and MRR
Full size table

Pie chart in Fig. 5b showing the percentage contribution of different input variables for MRR has been obtained from the ANOVA table. Fr and Dc are the major contributing factors with percentage contribution of 46% each. Vs has very small contribution in achieving maximum MRR.

3.3 Optimization using contour plots

Contour plots are normally used for the optimization and prediction of the response variables. Optimization of milling process can be considered as multivariate and multi-criteria problems in which the objective is to maximize or minimize the single variable. Here the effects of input variables on Ra and MRR have been analyzed using contour plots. It is apt that the graphs represent the effects of two input variables at the middle level of all the other variables.

3.3.1 Contour plots for Ra

Figure 6a–c represents the effects of Fr and Vs on Ra for dry, MQL, and nanofluid, respectively. By comparing the Ra of dry, MQL, and nanofluid cutting conditioned parts, it is evident that the effect of Fr and Vs on Ra are similar. Ra is more sensitive to Fr as compared with Vs. Moreover, Ra decreases with increasing Vs and decreasing Fr. It is virtuous to state that minimum Ra is achieved in nanofluid cutting condition as compared with dry and MQL.

Fig. 6
figure 6

Contour plots of Ra vs Fr and Vs a dry, b MQL, and c nanofluid

Full size image

The effects of input variables Fr and Dc for all cutting conditions are shown in Fig. 7 a to c. The contour plot demonstrates that Ra is affected by Fr significantly, and it is less altered by Dc and same trend has been observed in all cutting conditions. Among three different types of cutting conditions, nanofluid machining yields better results.

Fig. 7
figure 7

Contour plots for Ra vs Dc and Fr a dry, b MQL, and c nanofluid

Full size image

While distinguishing the influence of Vs and Dc on Ra under dry, MQL, and nanofluid cutting conditions, identical trends have been observed (Fig. 8 a to c). Ra decreases with increasing Vs and decreasing Dc. Among different cutting conditions, nanofluid produces the improved surface quality.

Fig. 8
figure 8

Contour plots of Ra vs Dc and Vs a dry, b MQL, and c nanofluid

Full size image

3.3.2 Contour plots for MRR

Based on the previous discussions, Fr and Dc are significant factors for MRR. Therefore, it is no need to consider the Vs in MRR optimization. From Fig. 9 a to c, it is cleared that MRR is maximum at high level of Fr and Dc. And different cutting conditions have no effect on MRR.

Fig. 9
figure 9

Contour plots of MRR vs Dc and Fr a dry, b MQL, and c nanofluid

Full size image

3.4 Comparison of cutting conditions (dry, MQL, and nanofluid)

It is observed that MQL is better than dry cutting and nanofluid lubrication is better than MQL cutting for Ra. Observed responses of dry, MQL, and nanofluid conditions are compared in Fig. 10. The figure has been drawn from the design matrix given in Table 3. From the Fig. 10, it is evident that shifting from dry to MQL cutting, Ra improve from 26%~30% and further switching from MQL to nanofluid cutting, 13%~16% improvement has been observed. In MQL, lubricant penetrates in the machining zone between tool and workpiece with air pressure in a very effective way which reduces the surface roughness [50]. Moreover, in nanofluid, the particles present in the lubricant possess filling and polishing effect and rolled at tool-workpiece interface which reduces the Ra and frictional co-efficient [51, 52]. From Fig. 11, it can be clearly seen that variation in cutting conditions has no effect on MRR, while negligible variation has been observed which can be due to the equipment and human error.

Fig. 10
figure 10

Percentage improvement results of Ra

Full size image
Fig. 11
figure 11

Percentage improvement results of MRR

Full size image

4 Conclusion

The focus of this research is to analyze the effects of dry, MQL, and nanofluid cutting conditions in face milling process for Al-6082 alloy. The effects of Fr, Vs, and Dc on Ra and MRR are analyzed for dry, MQL, and nanofluid cutting conditions using Taguchi method.

  • The experimental results reveal that for Ra, (1) Fr is most significant input variables with percentage contribution of 60%; (2) Vs is significant with percentage contribution of 30%; and (3) Dc is less significant as compared with feed rate and Vs with percentage contribution of 7% for dry, MQL, and nanofluid cutting conditions.

  • For material removal rate, Fr and Dc are significant factors with percentage contribution of 46% each for all cutting conditions, while Vs is not as significant as Fr and Dc.

  • Comparative analysis for Ra shows that the 26~30% improvement will be achieved when shifted from dry to MQL and 13~16% improvement will be obtained when further move to nanofluid cutting condition.

  • Negligible effect of cutting conditions (dry, MQL, and nanofluid) is observed for material removal rate.

  • Other machining conditions including machine tool fixture environment are kept constant for all experiments to avoid their influence on surface roughness.

This research verified that the proposed nanofluid cutting condition for face milling process could be used by the practitioners to improve the quality of machined parts. Furthermore, the contour plots and developed empirical models for Ra and MRR will aid practitioners to select the optimum level of input variables for the desired Ra and MRR.

As aluminum is a softer material, therefore, it is necessary to use some soft nanoparticles in lubricant to keep surface roughness at its minimum level. Therefore, future study can be conducted on the comparative analysis of the performance of metallic and non-metallic or some soft nanoparticles in the lubricant. If non-metallic nanoparticles yield low surface roughness, then further studies will be conducted to optimize the particle’s size and concentration with respect to workpiece materials. The lowest surface roughness achieved in case of soft nanoparticles will eliminate the post-processing of workpiece material. Elimination of single process on industrial scale may reduce the consumption of resources. Moreover, the combination of such nanoparticles and some environmental friendly lubricants can be used to make the process healthy on industrial scale.