- Email: [email protected]

ScienceDirect Materials Today: Proceedings 18 (2019) 2486–2495

www.materialstoday.com/proceedings

ICMPC-2019

Prediction of Machining Performances in Hardened AISI D2 Steel Ramanuj Kumar*, Ashok Kumar Sahoo, Purna Chandra Mishra, Amlana Panda, Rabin Kumar Das, Soumikh Roy School of Mechanical Engineering, KIIT Deemed to be University, Bhubaneswar-24, Odisha, India

Abstract

The current research focuses on predicting the surface roughness, the flank wear and (chip–tool) interface temperature when turning the hardened AISI D2 (55 ± 1 HRC) employing a mixed ceramic (Al2O3 + TiCN) tool in dry condition. Quadratic regression methodology has been introduced and their performances are compared on basis of mean error percentage among experimental and expected data. Normal probability plots are developed to understand the distribution of result data. ANOVA is implemented for all responses showing significant established regression models at 95 % of the level of confidence. 3D surface graph indicates that the feed is the most dominant term for Ra while cutting speed was more dominant for T and VBc. © 2019 Elsevier Ltd. All rights reserved. Selection and peer-review under responsibility of the 9th International Conference of Materials Processing and Characterization, ICMPC-2019 Keywords:Surface roughness; Hard part turning; Chip-tool interface temperature; Chip reduction coefficient; Flank wear; Regression modelling

1. Introduction In recent decades machining of hard to cut materials by turning mechanism is a popular topic for research development. Heat-treated steels are widely found its application in automobile, gear, bearing, die and press-tool industries. Hard turning is generally defined as the material removal process which is carried out on metal materials having hardness exceeds to 45 HRC or within the gamut of 45–68 HRC with the use of distinguish cutting tools such as PCBN, CBN, coated carbide inserts, cermet and ceramic inserts [1, 2, 3, 4]. Hard turning serves as a good alternative to the traditional grinding process, as hard turning is way more flexible, productive and environmental friendly machining operation of hardened materials. Even though the traditional grinding process produces the product of superior surface finish, but hard turning mechanism yields greater material removal rate (MRR), improved finish quality, decline processing costs, reduced manufacturing cycle time and minimizes environmental problems by avoiding the use of cutting fluids or with the application of minimal cutting fluids. * Corresponding author. Tel.: +0674 6540805 E-mail address: [email protected] © 2019 Elsevier Ltd. All rights reserved. Selection and peer-review under responsibility of the 9th International Conference of Materials Processing and Characterization, ICMPC-2019

R. Kumar et al. / Materials Today: Proceedings 18 (2019) 2486–2495

2487

Production of a finished hardened component involves sequential steps like forming, annealing, rough machining, the process of heat treatment and finish grinding. But hard turning successfully reduces the number of steps of manufacturing such as forming, rough machining, different heat treatment processes, and operation on finish grinding which ultimately reduces the machining cost and improves the productivity [1]. The main focus of this work is to understand the high speed hard turning and investigate parametric performances over cutting responses and establishment of empirical models for predicting the surface quality, wear at the flank surface, chiptool interface temperature. Various works are reported on modeling using RSM, ANN, genetic algorithms etc. some works have been highlighted as follows: Bensouilah et al.[5] employed L16Taguchi design and RSM statistical models to illustrate the impacts of input factors on output response variables. The ceramic insert generated surface quality is superior to that achieved by the uncoated. Das et al. [6-7] also used Taguchi as well as RSM concept for analyzing the variables. Quality of work surface highly dominated by cutting feed whereas it was improved with cutting speed till 170 m/min thereafter due to chatter it was reducing with speed. Bhemuni et al. [8] implemented the RSM based optimization and modeling in turning of heat treated D3 grade steel. The radial depth of cutting was traced to be the most critical term for tool flank wear. Pal et al. [9] formulated the mathematical expressions to correlate the variables. Determination coefficient was used to justify the reliability of the model and it was noticed to be very nearer to the predicted data of responses. Elbah et al. [10] applied the desirability function approach for the optimization of multiresponse data. RSM based empirical equations of second order and three-dimensional surface graphs were developed. Model ANOVA revealed the successful prediction of surface roughness with 95% of the confidence level. Mandal et al. [11] developed the mathematical relation based on regression of second order and its adequacy was checked by ANOVA. Point prediction optimization approach was implemented to select the optimal values of parameters. Rao et al. [12] formulated the regression model using RSM. R-square and R-square adjusted were found to be more than 75% however the model was moderately produced good agreement to fit. Makadia et al. [13] ascertained that from the equations based on RSM, the feed rate was the primary factor influencing the quality followed by cutting-tip radius. 3D surface plots were utilized to select the optimal values of parameters. Azizi et al. [14] developed the multiple regression mathematical equations of the first order to correlate the input parameters (depth of cut, cutting speed, cutting feed) with specific outputs i.e. machined surface quality and cutting forces. R-Sq and R-sq (predicted) values were very close to each other and also closely inclined to unity which suggested that the developed model was well fit. The desirability function optimization method was implemented to choose the optimal data set for multiple responses. Panda et al. [15] utilized the regression as well as multi-variant response optimization to elaborate the entire machining processes. ANOVA also play a key figure to evaluate the role of cutting variables on turning responses. The cutting feed was the highly dominated control term affecting multi-cutting responses at 95% of certainty levels. Ferreira et al. [16] worked on hardening a metallic piece of steel using regular and wiper ceramic cutting tool. The observation confirmed that the wear at the flank surface was greatly affected by turning speed whereas the tool experienced rapid wear when cutting was done with a speed of 240 m/min. In the majority of runs, the wiper tool presented a better response compared to the regular tool. Experimental work carried by Sarnobat and Raval [17] reveals that light honed tool edge geometry provides optimum machining performance with superior values of surface finish and compressive residual stress. Xavior and Jeyapandiarajan [18] performed finish turning operation on AISI D2 steel and optimal values of process parameter were determined using orthogonal array design and Grey relational analysis. The investigation revealed that feed is the most influencing input parameter affecting finish turning of hardened steel. Inadequate amount of turning works on hardened D2 steel (55 HRC) was noticed to date. Regression concept illustration in the past published articles is not enough. However, this work definitely will provide the benefits to the researcher to accomplish their work under dry surrounding and provide a good illustration of the concept of regression modeling. 2. Procedure and details of Experiment AISI D2 (55±1 HRC) tool steel cylindrical shape of dimensions (ϕ 48 mm x L 200 mm) is chosen as trial specimen due to its broad uses in industries like punches and dies, blanking and, mill rolls, spinning tools, shear blades, and automobile components. Taguchi design L16 has been selected for experimentation with three turning variables i.e. rotational speed, tool feed rate and depth of cutting with ranges varies from (63-182 m/min), (0.04 - 0.16 mm) and (0.1-0.4 mm) respectively. Moreover, accessible commercially WIDIA mixed ceramic (Al2O3+TiCN) inserts of

2488

R. Kumar et al. / Materials Today: Proceedings 18 (2019) 2486–2495

grade CW2015 issued. The geometry of cutting insert having CNMG 120408 (ISO designation) tightly fixed into a (PCLNR 2525 M12) tool holder according to ISO designation with (95o) approach angle, (-60) rake angle and (0.8 mm) of nose radius. The experimentations have performed on a semi-automatic NH 22 lathe (HMT) of highest spindle speed 2040 rpm in dry surrounding as shown in Fig. 1. Surface roughness (Ra) measurement is accomplished by surface roughness tester (Taylor Hobson, Surtronic 25). FLUKE Ti-32 infra-red (IR) camera has been employed to measure and confine the picture of T, VBc of worn out inserts are determined by an optical microscope (STM6 Olympus, Japan). MINITAB-16 software has been utilized for regression modeling, normal probability plot and ANOVA analysis. OriginPro 2016 (64Bit) is used to draw the graph between experimental and predicted data.

AISI D2 steel

Ceramic insert

Fig. 1. Hard turning process using ceramic insert. 3. Experimental results overview In current work, analysis of surface roughness (Ra), chip-cutting tool interface temperature (T) and flank wear (VBc) of D2 steel hard turning using the mixed ceramic tool are focused. The experimental results revealed the excellent performance of ceramic inserts during the tuning process. The surface roughness lies within the limit of 1.3 microns. Chip-tool interface temperature varies in between 1800C to 3500C whereas flank wear width varies from 0.03 to 0.09 mm. The optical image of tool-tip at run 7 is displayed in Fig. 2a and its corresponding temperature is displayed in Fig. 2b. a

b

Wear at flank surface

Run-7 d = 0.2 mm v = 182 m/min f = 0.12 mm/rev

Fig. 2. (a) Diagram of wear (flank) image (b) Chip-tool interface temperature.

R. Kumar et al. / Materials Today: Proceedings 18 (2019) 2486–2495

2489

Quality of surface gets deteriorate with higher feed rated whereas growth of wear width rises with rotational speed. Chip-tool interface temperature improves with rising speed and cutting feed. The depth of cut doesn’t produce any significant role for considered output characteristics. Abrasion is only seemed to be regular wear mechanism involved in the process. Similar observations have been reported in various works [7, 10]. 4. Regression modeling Regression modeling is a statistical approach in the form of empirical equations for analysis of numeric problems. The regression models co-relate the output variables with input variants. The Regression is applied to the observed data collected, while the examined variables are closely related to the functional relationships between the variables with one or more inputs. In addition, responses signifying the (ϒ) output can be stated as the function of the turning inputs, for example, v, f and d and as presented in Eq. 1. ϒ = λ0 + λ1 (d)+ λ2 (f) + λ3(v) + λ4 (d2) + λ5 (f 2) + λ6 (v2)+ λ7 (df)+ λ8 (fv) + λ9 (vd)

(1)

Where, ϒ is output variant, λ0is intercept of the plane and λ1, λ2…..λ9 is regression coefficient, the coefficient of each λ displayed in Eq. 1 is evaluated by least square methods. The expressions d, f and v are the entering variants; d 2 , f 2 and v2 are the square variants; and df, fv and vd are interaction variants correspondingly [17, 18]. A 2nd order regression model is exploited usually when the output function is unspecified. Furthermore, the model fitness (2nd order regression equation) can be verified by its estimated determination coefficient (R-Square) and its validity approval was dependent on more to very more R-Square value. The R-sq term specifies the variant of data in a percentage while the R-Square (adjusted) data is helpful when the model is compared to a mixed set of data. When R-square values reach hundred, the response output model will be efficiently fitted among actual results. Equations for investigational responses were developed using regression techniques considering uncoaded units with a 95% confidence level. The implications for the analysis and the capability of the models were identified using ANOVA derived from the computed values of f and p. For ANOVA, the squares sum is calculated to calculate the square of the divergence as of the decent mean. In addition, the average squares are calculated by the sum of the squares and the degrees of freedom [17, 18]. When the p-value is less than 0.05 (confidence level 95%), the significance of the analogous variable is recognized, and this model has said to be noteworthy [18]. The consequence of the regression model is derived from the normal probability plot based on the central limit theory. In a probability plot, the residuals must be on a straight line, signifying that the error is usually scattered and that the sample is found to be important with good correlations. 4.1 Model based on surface roughness The surface roughness (Ra) is a key factor to understand the machinability performance in any machining process. Therefore, the current work focused on the modeling of surface roughness in hard part turning condition. The performance of ceramic tool is comparable with the grinding operation as more or less the Ra value lies about 0.8 µm except at elevated feed rate (0.12 and 0.16 mm/rev). From the regression model for Ra (Eq. 2), the numerical value of R-Square and R-Square (adj) are 0.9913 and 0.9782 respectively. Both the values are very nearer to unity and seem close to each other which confirm a good fineness developed model with better mathematical relevant and the irrelevant factors are not linked with the model. Comparative statics of results (experimental vs. fitted) for surface roughness is displayed graphically in Fig. 3a and traced to be very close (average error = 2.816 % as displayed in Table 1) which confirms good fit of the established model. The graph (Fig. 3b) representing the normal probability distribution and it confirms that the residuals moderately scattered close to a linear line representing that these are normally scattered and ascertained that regression model is well adequate. As of 3D graph (Fig. 4a-f), the surface roughness increases with rising in d-f, f-d, f-v and v-f (Fig. 4a-d) whereas the terms (v-d and d-v) are not significant. However, the rate of dominance for feed is more than speed and depth of cutting. In accordance with ANOVA (Table 2), the regression model is important with 95 % confidence level i.e. p-value is lower than 0.05 but linear terms are insignificant as their probability is larger than 0.05 whereas other terms (square and interaction) are significant.

2490

R. Kumar et al. / Materials Today: Proceedings 18 (2019) 2486–2495

Ra = 0.5686 + 0.3282d ˗ 0.1654f ˗ 0.0024v + 3.1250d2 + 22.7344f2 + 0.0000v2 + 1.3254df ˗ 0.0173dv ˗ 0.0089fv R2 = 99.13% R2 (adj) = 97.82%

(2)

b

a Experimental data Predicted data

Normal Probability Plot (response is Ra)

1.2

99

1.0

95 90 80 70 60 50 40 30 20 10 5

0.8

Percent

Surface roughness, Ra (µm)

1.4

0.6 0.4 0.2 0.0

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 Test run

1

-0.050

-0.025

0.000 Residual

0.025

0.050

Fig. 3. (a) Comparative results in between experiment and model for Ra (b) Normal probability graph for Ra.

a

b 3-D Surface Graph of Ra vs f, d

3-D Surface graph of Ra vs d, f 1.25 Ra 1.00 0.75 0.50 0.05 0.10 0.15 f

1.25 Ra 1.00 0.75 0.50 0.1 0.2 0.3 0.4 d

0.4 0.3 0.2 d 0.1

d

c

1.25 Ra 1.00 0.75 0.50 0.05 0.10 0.15 f

150 100 v 50

1.25 Ra 1.00 0.75 0.50 50

0.15 0.10 f 0.05

100 150 v

0.15 0.10 f 0.05

f

e 3-D Surface Graph of Ra vs v, f

3-D Surface Graph of Ra vs f, v

3-D Surface Graph of Ra vs v, d 1.25 Ra 1.00 0.75 0.50 0.1 0.2 0.3 0.4 d

150 100 v 50

Fig. 4. (a-f) 3-D surface graph for Ra.

3-D Surface Graph of Ra vs d, v 1.25 Ra 1.00 0.75 0.50 50

100 150 v

0.4 0.3 0.2 d 0.1

R. Kumar et al. / Materials Today: Proceedings 18 (2019) 2486–2495

2491

Table 1. Percentage mean error between trial and predicted data.

% Mean error

Table 2. Ra model (ANOVA). Source DF Regression 9 Linear 3 Square 3 Interaction 3 Residual Error 6 Total 15

Seq SS 0.960862 0.836239 0.079373 0.045250 0.008471 0.969333

Ra

T

VBc

2.816

1.744

7.061

Adj SS 0.960862 0.003149 0.079533 0.045250 0.008471

Adj MS 0.106762 0.001050 0.026511 0.015083 0.001412

F 75.62 0.74 18.78 10.68

P 0.000 0.564 0.002 0.008

Remarks Significant Insignificant Significant Significant

4.2 Cutting temperature model In machining problem, the temperature is a novel factor which affects tool life and quality of finished product. Therefore, the present work emphasized on modeling of chip-tool interface temperature (T) in hard turning process. Regression model for T is represented in Eq.3. From the model, R-Square and R-Square (adj) are noticed to be 0.9895 and 0.9737 respectively. In addition, the R-Square value noticed to be extremely close to the hundred which signifies that the regression model precisely predicts the response outputs and the regression model is significant statically. Also, R-square (adj) is very nearer to R-Square however, it can be understood that the inappropriate expressions are not correlated with the model. T = 271.75˗ 301.73d + 1720.21f ˗ 2.53v + 30.63d2 ˗ 3605.47f2 + 0.01v2 ˗ 238.18df + 2.65dv ˗ 2.45fv R2 = 98.95%; R2(adj) = 97.37%

(3)

b

a

340

99

320

95 90 80 70 60 50 40 30 20 10 5

300 280

Percent

Temperature, T (0C)

Normal Probability Plot (response is T)

Experimental data Predicted data

360

260 240 220 200 180 160

1 0

2

4

6

8 10 Test run

12

14

16

18

Fig. 5. (a) Comparative results in between experiment and model for T

-15

-10

-5

0 Residual

5

10

(b) Normal probability graph for T.

Graphical representation of results (Experimental and predicted) displayed in Fig. 5a. It clearly revealed that the predicted results are very nearer (average error = 1.744 % as displayed in Table 1) to experimental outcomes.

2492

R. Kumar et al. / Materials Today: Proceedings 18 (2019) 2486–2495

Nevertheless, the regression model is thoroughly important and fitted. The normal probability graph for T is displayed in Fig. 5b, which revealed that the residual is scattered fairly near to a straight linear line demonstrating that the error is normally distributed and it confirmed that the regression model is well adequate. From 3D graph (Fig. 6a-f), the temperature at chip-tool interface improves with an increase in f-v, v-f, v-d and d-v (Fig. 6c-f) whereas the terms d-f and f-d (Fig. 6a-b) are not significant. However, the elevation in temperature due cutting speed is highly dominants compare to feed and depth of cut. From ANOVA (Table 3), the regression model for temperature was noted with a confidence level of 95%. All the source terms are significant as their p term lies below to 0.05. a

b

c

3-D Surface Graph of T vs d, f 350 300 T 250 200 0.05 0.10 0.15 f

3-D Surface Graph of T vs f, d 350 300 T 250 200

0.4 0.3 0.2 d 0.1

d

0.1 0.2 0.3 0.4 d

3-D Surface Graph of T vs f, v 350 300 T 250 200

0.15 0.10 f 0.05

50

e

f

3-D Surface Graph of T vs v, f 350 300 T 250 200 0.05 0.10 0.15 f

100 150 v

0.15 0.10 f 0.05

3-D Surface Graph of T vs d, v

3-D Surface Graph of T vs v, d 350 300 T 250 200

150 100 v 50

0.1 0.2 0.3 0.4 d

T 150 100 v 50

350 300 250 200 50

100 150 v

0.4 0.3 0.2 d 0.1

Fig. 6. (a-f) 3-D surface graph for T. Table 3. ANOVA for cutting temperature model. Source Regression Linear Square Interaction Residual Error Total

DF 9 3 3 3 6 15

Seq SS 38781.1 30693.9 6991.9 1095.3 411.9 39193

Adj SS 38781.1 4705.4 6972.3 1095.3 411.9

Adj MS 4309.02 1568.47 2324.09 365.11 68.65

F 62.77 22.85 33.85 5.32

P 0.000 0.001 0.000 0.040

Remarks Significant Significant Significant Significant

4.3 Model based on Flank wear In the hard turned process, the growth of flank wear significantly influences the performances such as surface quality, cutting forces, vibrations and dimensional accuracy. Prediction of a favorable level of flank wear is of great relevance for hard turning concern. Therefore prediction regression model is being focussed as mentioned in Eq. 4. From the model (Eq.4), the R-Square is especially nearer to 100 % (R-Square = 0.9896), that confirms the perfection

R. Kumar et al. / Materials Today: Proceedings 18 (2019) 2486–2495

2493

of fit for the established model with higher mathematical importance. In addition, it revealed the solid interrelationship between the experimental data and the expected data. VBc = 0.048942+ 0.069856d + 0.182737 f ˗ 0.000704v ˗ 0.175000d2 ˗ 0.312500 f2 + 0.000003v2 ˗ 0.751357df + 0.000946dv + 0.000857fv R2 = 98.96% R2 (adj) = 97.40% (4) a

b 0.09 0.08

99

0.07

95 90 80 70 60 50 40 30 20 10 5

0.06 0.05

Percent

Flank wear, VBc (mm)

Normal Probability Plot (response is VBc)

Experimental data Predicted data

0.04 0.03 0.02 0.01 0.00

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 Test run

1 -0.0050

-0.0025

0.0000 Residual

0.0025

0.0050

Fig. 7. (a) Comparative results in between experiment and model for VBc (b) Normal probability graph for VBc. a

b 3-D Surface Graph of VBc vs d, f 0.08 VBc 0.06 0.04 0.02 0.05 0.10 0.15 f

c

3-D Surface Graph of VBc vs f, d

0.08 VBc 0.06 0.04 0.02 0.1 0.2 0.3 0.4 d

0.4 0.3 0.2 d 0.1

3-D Surface Graph of VBc vs v, f 0.08 VBc 0.06 0.04 0.02 0.05 0.10 0.15 f

150 100 v 50

0.08 VBc 0.06 0.04 0.02 50

0.15 0.10 f 0.05

e

d

3-D Surface Graph of VBc vs f, v

100 150 v

0.15 0.10 f 0.05

f 3-D Surface Graph of VBc vs v, d 0.08 VBc 0.06 0.04 0.02 0.1 0.2 0.3 0.4 d

150 100 v 50

Fig. 8. (a-f) 3-D surface Graph for VBc.

3-D Surface Graph of VBc vs d, v 0.08 VBc 0.06 0.04 0.02 50

100 150 v

0.4 0.3 0.2 d 0.1

2494

R. Kumar et al. / Materials Today: Proceedings 18 (2019) 2486–2495

Similarly, R-square adjusted (0.9740) is near R-square (0.9896), which clearly shows that inappropriate factors do not exist in the model established. The experimental and predicted graph of flank wear is illustrated in Fig. 7a and it is noticed to be approximate to each other (average error = 7.061 % as displayed in Table 1) entailing the importance of the regression model established. The normal probability graph is displayed in Fig. 7b, that suggested that residuals are spread near the straight line signifying that the errors are usually scattered to ensure that the model-related factors are imperative. It shows good correlations between trial and predicted results. Flank wear is of a rising trend with improving d-f, f-d, f-v and v-f (Fig. 8a-d) whereas the terms v-d and d-v (Fig. 8e-f) are not significant. The maximum impact due to speed on the flank wear is evident from the 3D area (Fig. 8). ANOVA shows that the regression models have been developed are important because p (probability) is less than 0.05 represented in Table 4. Table 4. ANOVA for VBc model. Source Regression Linear Square Interaction Residual Error Total

DF 9 3 3 3 6 15

Seq SS 0.005412 0.004759 0.000422 0.000231 0.000057 0.005469

Adj SS 0.005412 0.000275 0.000412 0.000231 0.000057

Adj MS 0.000601 0.000092 0.000137 0.000077 0.000009

F 63.53 9.70 14.49 8.15

P 0.000 0.010 0.004 0.015

Remarks Significant Significant Significant Significant

Conclusion Current work deals on the application of the regression approach to predict surface roughness, chip-tool interface temperature and flank wear using the mixed ceramic tool in hardened AISI D2 steel in dry condition. The subsequent concluding remarks are as follows:

The turning performance using ceramic insert was superior as the surface roughness lies within 1.3 micron and chip-tool interface temperature varies in between 1800C to 3500C whereas flank wear width varies from 0.03 to 0.09 mm among all tests. Regression models for Ra, T, and VBc clearly denotes that the numerical value of R-Square and R-Square (adj) are very near to unity and very closely to each other which confirms a good fit for the developed model with better mathematical relevant and the irrelevant factors are not linked with the model. The normal probability graphs for all three responses are combined and the residuals are distributed as close to the straight line, signifying that errors created are usually scattered. Normal probability plot for all three responses and residuals that are about the straight line, representing that errors created are usually scattered which in turn the model is significant. 3D surface graph indicates that the feed is a most dominant term for Ra whereas cutting speed was more prevailing for T and VBc. ANOVA for all responses clearly highlighted that developed regression models are significant as p (probability) lies within 0.05 at 95 % of the level of confidence.

Acknowledgement This research is supported by AICTE, New Delhi, India under Research Promotion Scheme (RPS) project vide Ref No: 8-154/RIFD/RPS/POLICY-4/2013-14. The authors express their thanks and gratitude AICTE for granting financial support to perform the research work. References [1] J.P. Davim, Machining of hard materials, Springer- Verlag London Limited (2011) 1-46 [2] A.K. Sahoo, B. Sahoo, A comparative study on performance of multilayer coated and uncoated carbide inserts when turning AISI D2 steel under dry environment, Measurement 46 (2013) 2695–2704. [3] A.K. Sahoo, B. Sahoo, Experimental investigations on machinability aspects in finish hard turning of AISI 4340 steel using uncoated and multilayer coated carbide inserts, Measurement 45 (2012) 2153–2165.

R. Kumar et al. / Materials Today: Proceedings 18 (2019) 2486–2495

2495

[4] S. Chinchanikar, S.K. Choudhury, Machining of hardened steel—Experimental investigations, performance modelling and cooling techniques: A review, International Journal of Machine Tools and Manufacture 89 (2015) 95–109. [5] H. Bensouilah, H. Aouici, I. Meddour, M.A. Yallese, T. Mabrouki, F. Girardin, Performance of coated and uncoated mixed ceramic tools in hard turning process, Measurement 82 (2016) 1–18. [6] S.R. Das, D. Dhupal, A. Kumar, Experimental investigation into machinability of hardened AISI 4140 steel using TiN coated ceramic tool. Measurement 62 (2015) 108–126 [7]R. Das, D. Dhupal, A. Kumar, Study of surface roughness and flank wear in hard turning of AISI 4140 steel with coated ceramic inserts, Journal of Mechanical Science and Technology 29 (10) (2015) 4329–4340 [8] V. Bhemuni, C.S. Rao, P.V. Vinay, Effect of machining parameters on tool wear in hard turning of AISI D3 steel, Procedia Engineering 97 (2014) 338–345. [9] A. Pal, S.K. Choudhury, S. Chinchanikar, Machinability assessment through experimental investigation during hard and soft turning of hardened steel, Procedia Materials Science 6 (2014) 80 – 91 [10] M. Elbah, M.A. Yallese, H. Aouici, T. Mabrouki, J-F Rigal, Comparative assessment of wiper and conventional ceramic tools on surface roughness in hard turning AISI 4140 steel, Measurement 46 (2013) 3041–3056. [11] N. Mandal, B. Doloi, B. Mondal, Predictive modelling of surface roughness in high speed machining of AISI 4340 steel using yttria stabilized zirconia toughened alumina turning insert, International Journal of Refractory Metals and Hard Materials 38 (2013) 40–46 [12] C. J. Rao , D.N. Rao, P. Srihari, Influence of cutting parameters on cutting force and surface finish in turning operation, Procedia Engineering 64 ( 2013 ) 1405 – 1415. [13] A.J. Makadia, J.I. Nanavati, Optimisation of machining parameters for turning operations based on response surface methodology, Measurement 46 (2013) 1521 – 1529. [14] M.W. Azizi, S. Belhadi, M.A. Yallese, T. Mabrouki, J.F. Rigal, Surface roughness and cutting forces modelling for optimization of machining condition in finish hard turning of AISI 52100 steel, Journal of Mechanical Science and Technology 26 (12) (2012) 4105 – 4114 [15] A. Panda, A.K. Sahoo, A.K. Rout, Multi-attribute decision making parametric optimization and modelling in hard turning using ceramic insert through grey relational analysis: A case study, Decision Science Letters 5 (2016) 581–592. [16] R. Ferreira, Jan Řehoř, C.H. Lauro, D.Carou, J.P. Davim, Analysis of the hard turning of AISI H13 steel with ceramic tools based on tool geometry: surface roughness, tool wear and their relation, Journal of Brazilian Society of Mechanical Science and Engineering 38 (2016) 2413– 2420. [17]Sarnobat, S. S., & Raval, H. K. (2019). Experimental investigation and analysis of the influence of tool edge geometry and work piece hardness on surface residual stresses, surface roughness and work-hardening in hard turning of AISI D2 steel. Measurement, 131, 235-260. [18]Xavior, M. A., & Jeyapandiarajan, P. (2018). Multi-Objective Optimization during Hard Turning of AISI D2 Steel Using Grey Relational Analysis. Materials Today: Proceedings, 5(5), 13620-13627. [19] D.C. Montgomery, Design and Analysis of Experiments, fourth ed., Wiley, New York, 1997. [20] A. K. Sahoo, B. Sahoo, Performance studies of multilayer hard surface coating (TIN/TICN/AL2O3/TIN) of indexable carbide inserts in hard machining: Part-II (RSM, Grey relation and techno economical approach), Measurement 46 (2013) 2868-2884.