Response Modeling and Optimization of Process Parameters in Turning Medium Carbon Steel Under Minimum Quantity Lubrication (MQL) with Vegetable Oil and Oil Blends

Abstract

Machining is an important aspect of manufacturing. The application of cutting fluid in the field of engineering manufacturing has a history of hundreds of years, and it plays a vital role in the processing efficiency and surface quality of parts. The use of vegetable oil in machining operations is receiving attention currently for sustainable alternatives to mineral-based cutting oil. If the vegetable oil is applied through the minimum quantity lubrication (MQL) technique, it becomes more cost effective, eco-friendly, and sustainable. This study aims to investigate the effects (cutting force and temperature) of coconut oil, a coconut–rice bran oil blend, and a coconut–olive oil blend, and compare them with VG 68 oil using MQL. A magnetic stirrer was employed for mixing oils (coconut–rice bran oil and coconut–olive oil), performed at 40 °C and 250 rpm. The response parameter values were evaluated at different combinations of speed (78, 113.5, and 149 mm/min), feed (0.1, 0.13, and 0.16 mm/rev), and depth of cut (0.5, 0.75, and 0.1 mm). The design of the experiment (DOE) was created using the value of input parameters using response surface methodology (RSM). Percentage (%) reduction was calculated to compare the reduction in cutting force and temperature by using coconut oil, a coconut–rice bran oil blend, and a coconut–olive oil blend concerning mineral oil. Empirical models were developed for cutting force and temperature by RSM for the four cutting environments. The ANOVA result shows that the model performed satisfactorily for both temperature and force analysis. RSM-based optimization was carried out and the optimal solution was found at the cutting speed of 80.15 m/min, feed rate of 0.10 mm/min, and 0.5 mm depth of cut for the coconut–olive oil blend. Also, the model performed better in the reduction in force than temperature.

1. Introduction

Manufacturing companies have been striving to become more sustainable through green manufacturing for the past three decades, but there have been several obstacles in the way of reaching this goal [1]. Industries involved in manufacturing are important stakeholders because of the substantial impacts they have on the environment. The majority of these businesses understand the importance of environmental protection and are actively working to address these issues. The future of environmentally friendly production is one in which all negative effects are eliminated [2]. Therefore, environmental features must be guaranteed by manufacturing sectors without disrupting the economic or social components of the business [3], making green manufacturing practices an essential part of sustainable manufacturing.

A vital part of manufacturing is machining. It involves several procedures that involve the surface modification and material removal of a workpiece following the use of various techniques [4,5]. Almost all of the energy produced while cutting metal is converted into heat [6]. Figure 1 depicts the distribution of heat during an orthogonal cutting operation in the workpiece, cutting tool, and chip. The primary cutting areas, where most of the energy is converted to heat, the secondary deformation area at the tool–chip interface, where additional heat is produced by rubbing or shearing, and the strained flanks, which are caused by rubbing between the tool and the finished surfaces, are the three sources of heat that are produced at the cutting point [6]. Thermal cracking, which is caused by expansion and cracking brought on by extremely high temperatures, is another cause of cutting tool failure [7]. Additionally, the heat generated during metal cutting causes tool wear, which may hamper product quality and process productivity [8,9]. High temperatures experienced during and after machining eventually result in surface damage from oxidation, burns, and quick corrosion at the cutting area, which causes the workpiece’s dimensional deviation [10]. To minimize the above-mentioned problems, machining may require expensive high-heat and wear-resistant tools and sturdy machine tools that can withstand the machining process. Moreover, it is limited to low-strength material cutting with restrained productivity [11].

The use of cutting fluid helps overcome these obstacles. It can be used for a variety of purposes. It can lubricate the chip–tool and work–tool interfaces by their ball bearing and rolling effect, cool the workpiece and cutting zone, sweep out the metal chips, and last but not least, protect the machined surface from corrosion [13,14]. There are two main types of cutting fluids. Figure 2 shows the classification of different types of cutting fluids.

The conventional flood cooling technique involves directing a cooling jet toward the cutting area, which is most suited for operations that generate a high cutting temperature or sparks, such as turning, drilling, and grinding, because the cooling jet quickly reduces temperature and spark generation [16]. However, this technique is not economical and eco-friendly. To overcome its drawbacks, researchers are encouraged to invent innovative, low-cost, efficient, and sustainable cooling strategies shown in Figure 3. Among the most used techniques are vegetable oil-based coolant application [17,18], cryogenic cooling with LN₂ or CO₂ [19], MQL with different oil/water [20], and their hybrid applications like cryogenic–MQL, MQL with vegetable oil, etc. Based on sustainability indicators and machinability efficiency, cryogenic approach secures the highest ranking, then MQL [21].

Roughly 85% of the machining fluids used globally are mineral oil-based cutting fluids [22]. However, they have bad effects on human health and significant and pervasive occupational dangers that get airborne and form aerosols during machining [23]. Considering human health and environmental consciousness, modern machining has started to follow sustainable alternatives. Eco-friendly and biodegradable vegetable oils used as cutting fluids is the most sustainable solution. They are recommended over petroleum-based metalworking cutting fluids for performance, cost, health, safety, and the environment [24]. Most vegetable oils come from plant seeds. These oils come from sustainable sources such as coconut, sunflower, soybean, rapeseed, olive, palm, and others [25]. Figure 4 shows how vegetable oils biodegrade faster than mineral oils. It shows vegetable oils’ competitiveness. Studies demonstrate that vegetable oils can deteriorate 70–100% of their weight [26].

Thus, the use of vegetable oils as metal cutting fluids (MCFs) enables the performance of green machining [27]. Above and beyond, vegetable oil has a triglyceride structure, which makes it suitable for using as a desirable lubricant. Due to long fatty acid chains, high strength lubricant film is created, which has the ability to adhere to the metal surface. Ultimately, it works better compared with mineral oil for friction and wear reduction during machining. Vegetable oil is also used to provide a more stable and high viscosity index even at higher temperatures, making it the best alternative to mineral oil [28,29]. Xavior and Adithan [30] compared vegetable oil with mineral oil in the turning of AISI 304 austenitic stainless steel. Surface roughness and tool wear were reduced when vegetable oil was used during machining.

There are various methods for improving vegetable oil properties. One of the simplest approaches for improving vegetable oil properties is to blend or combine vegetable oils with diverse qualities [31]. Due to the intrinsic variances in the fatty acids present in oils, they have varied physical and chemical properties. Pure vegetable oil, when used without any modifications, can have poor physical, chemical, and nutritional qualities, as well as oxidative stability.

Figure 4. Vegetable oils and mineral oils comparison in terms of biodegradability. (Reproduced from [32] with copyright permission from the publisher. License number: 5899341052809).

Figure 4. Vegetable oils and mineral oils comparison in terms of biodegradability. (Reproduced from [32] with copyright permission from the publisher. License number: 5899341052809).

Cryogenic cooling is an environmentally friendly and extremely effective method for machining [33]. This technique entails directing a liquid cryogen stream directly onto the cutting interfaces, resulting in enhanced machining performance. The environmental friendliness of cryogenic cooling is notable, as it leaves no hazardous residue and does not require costly disposal systems. Thus, it is a viable alternative to conventional cooling strategies [34] because it satisfies numerous requirements for sustainable machining. In a comparative investigation conducted by Karkade and Patil [35], compressed CO₂ gas cooling was found to improve surface integrity during machining, despite possibly resulting in increased cutting forces due to work-hardening effects. However, the use of cryogenic refrigeration does have some restrictions. High equipment costs and the expense of procuring liquid nitrogen (LN₂) or liquid CO₂ pose financial challenges [36]. In addition, these cryogens cannot be reused, unlike conventional cutting fluids. Extremely low temperatures in cryogenic environments can cause work hardening, resulting in increased cutting forces and diminished dimensional accuracy [37].

The other effective, sustainable, eco-friendly, and easy-to-handle coolant application technique is MQL, which injects a small quantity of lubricant or cooling fluid into the cutting zone in two ways: either drop by drop or in aerosol form at a very high velocity (100 m/s) [37,38,39]. The most common approach uses MQL in aerosol form, which improves cooling by increasing fluid surface area and the heat transfer coefficient due to high fluid velocity and aerosol evaporation. The MQL technique has some promising outcomes as well as offers economic savings, environmental conservation, and human safety due to its low coolant usage and disposal savings [40,41]. Sharma et al. [42] studied MQL technology, which generates surfaces superior to dry machining and similar to MCFs. Dhar et al. [43] found that MQL turning improved tool wear and surface roughness over typical MCFs. Tasdelen et al. [44] observed that MQL reduced tool wear and shorter chip length than MCFs in precipitation-hardened steel drilling. Silva et al. [45] found that MQL outperformed MCFs in grinding by monitoring surface integrity and wheel wear. MQL is also more energy-efficient and environmentally friendly than dry and flood machining [46].

The MQL method can be more environmentally friendly and efficient by hybridizing with various techniques. MQL with cryogenic cooling can perform better [47]. However, this method can be expensive to set up and maintain. The MQL method can be more environment-friendly and hazard-free if mineral oil is replaced with distilled water, oil-in-water emulsion, and vegetable oil. Vegetable oil is the most favorable alternative compared with others because of its high viscosity [48]. The MQL technique with coconut oil cutting fluid machining conditions reduced tool wear, friction coefficient, chip morphology, and workpiece surface quality, according to Vardhaman et al. [49]. Khan and Dhar [50] investigated the effect of vegetable oil-based MQL cutting that affected tool wear, cutting zone temperature, surface quality, and dimensional deviation when turning AISI 1060 steel with uncoated carbide inserts. MQL cutting reduced tool wear, surface roughness, and dimensional error. Khan et al. [51] turned AISI 9310 alloy steel with vegetable oil-based MQL. The authors reported that the cutting area’s temperature was reduced significantly. If vegetable oil blends are used by MQL instead of pure vegetable oil, then the performance increases further. As stated by Susmitha et al. [52], blends of two vegetable oils in 1:1 proportions perform better compared with a single vegetable oil and commercial mineral oil during the drilling operation of mild steel under MQL. Nanoparticles can be mixed with vegetable oil to improve its performance, and recently, huge research has been conducted in this area [53,54].

The ultimate sustainability for any manufacturing process can be achieved by investigating the effect of different variables on machining responses and careful selection of each variable involved. Numerous research regarding the effect analysis and selection of optimum machining parameters was carried out previously. Currently, it is a vast area of manufacturing research. Improved product quality, lower production costs, and sustainable machining operations are a consequence of an optimized process [55]. Cutting speed, feed rate, and depth of cut are key process parameters for turning operations; yet, cooling conditions play a critical role in increasing process performance. Selection of suitable parameters can be achieved considering a single response or multiple responses, but simultaneously considering multiple important responses according to their importance is an increasing requirement for obtaining an optimized manufacturing process. There are various process responses by which the process performance can measured. Sristi et al. [56] optimized the cooling environment, cutting speed, and feed rate considering multiple responses such as temperature, force, roughness and chip reduction coefficient by using the MCDM method. Zaman and Dhar [57] selected optimum MQL parameters like nozzle diameter, nozzle angle, air pressure, and oil flow rate considering minimization of temperature, force, and surface roughness. Muthuram and Christo [58] optimized turning parameters for minimizing roughness and maximizing material removal rate.

Different prediction models are typically used as inputs for the optimization of process parameters in metal-cutting processes. Response surface methodology (RSM) is a statistical regression technique widely used to formulate prediction models. Response surface methodology (RSM) was utilized by Suresh and Basavarajappa [59] to investigate the impact of various cutting parameters on the hard turning of AISI H13 steel using a PVD-coated TiCN ceramic tool. Models for tool wear and surface roughness were created. When turning Ti-6Al-4V alloy, Mia et al. [60] employed RSM and ANN to forecast the surface roughness and cutting force. The force required to hard turn AISI 52,100 steel (60 HRC) with a CBN tool was also predicted using RSM [61]. The ideal cutting conditions were determined by Zahia et al. [62] using RSM to reduce cutting force and surface roughness. They examined AISI 4140 alloy steel that had been machined with a ceramic insert. Zaman and Dhar [63] used a composite desirability approach to find optimum process parameters considering multiple criteria. MQL application parameters were also optimized by the RSM method [57]. Various artificial intelligence-based methods, such as ANN-GA [58] and PSO-ANN [64] hybrid algorithms can also be utilized for optimization.

From the literature review, it can be concluded that lubrication is necessary for efficient and accurate machining performance. MQL is one of the promising cooling and lubrication techniques though its performance is lower than cryogenic cooling. However, there are numerous ways of improving the performance and sustainability of MQL. Selection of optimum process and MQL parameters can be a useful technique to achieve better machining performance. High-performance fluids such as vegetable oil blends and nanofluids can also be used to improve the performance of MQL. However, nanofluid preparation and maintenance is very critical. Blending vegetable oils is an easy and cost-effective method. Nevertheless, MQL machining with different vegetable oil blends is very rare in the recent literature. As per authors’ knowledge, there is no literature where a coconut oil + rice bran oil blend and coconut oil + olive oil blend were used as MQL fluids in turning/machining operations. This can be a vast scope to address in search of green and cost-effective lubrication for sustainable machining.

In this present work, different coconut oil-based blends were prepared and their applications using MQL in turning medium carbon steel were experimentally investigated compared to pure coconut oil and mineral oil in terms of cutting temperature and cutting force. Experimentation was conducted via the Box–Behnken design of experiment using RSM. Predictive models for both of the responses were formulated by utilizing MINITAB 17. Finally, the optimum value of process parameters was identified by using an RSM-based composite desirability method considering the minimum cutting temperature and minimum cutting force simultaneously. The final result has been verified by experimentation.

2. Materials and Methods

2.1. Workpiece and Cutting Tool Material

A medium carbon steel round bar with a length of 400 mm and diameter of 71 mm was used as the work material. The hardness of this material was 102–105 HRB. Medium carbon steel portrays a good balance of tensile strength, ductility, and malleability. It has high flexibility and elasticity despite its fragility and brittleness. It is a highly machinable and weldable material for many applications. Medium carbon steels are used for making shafts, axles, gears, crankshafts, couplings, railway wheels, and different equipment related to building structures and forgings, etc.

Commercially available uncoated tungsten carbide inserts (ISO designation SNMG 120408, WIDIA, Bangalore, India) with hardness 74.2 RC were used in the machining. They can be used in turning for a variety of applications. Tungsten carbide is a hard and rigid material with high impact resistance, strength, and compressive strength. It is also resistant to deformation, deflection, heat, and oxidation. The uncoated carbide insert was mounted in a PSBNR 2525 M12 tool holder (Sandvik, Maharashtra, India), which provided the required working tool geometry of −6°, −6°, 6°, 6°, 15°, 75°, and 0.8 mm.

2.2. Machine Tool and Experimental Conditions

The experimentation was conducted by turning medium carbon steel with an uncoated carbide insert in a manual lathe machine (KL-3280C/2000, Sunlike Machine and Tool Corp., Huizhou, China) under MQL conditions. The cutting speed, feed rate, and depth of cut for turning medium carbon steel by uncoated carbide insert were set based on the literature and availability in the machine. In the literature, similar differences in cutting parameters have been observed [59,65]. The experimental conditions of this work have been presented in

Table 1. Experimental conditions.

Machine Tool	: Lathe Machine (KL-3280C/2000), spindle power 7.5 KW
Work Material	: Medium Carbon Steel
Tool Material	: Uncoated Tungsten Carbide
Cooling Condition	: MQL
MQL Supply Parameters:
Air Pressure	: 23 bar
Oil Pressure	: 25 bar
Oil Quantity	: 150 mL/h
Nozzle Diameter	: 1 mm
MQL Coolants:
	i. Mineral Oil (VG 68 Cutting Fluid)
	ii. Coconut Oil
	iii. Coconut + Rice Bran Oil
	iv. Coconut + Olive Oil
Process Parameters:
Cutting Speed (V)	: 78, 113.5, 149 m/min
Feed Rate (f)	: 0.1, 0.13, 0.16 mm/rev
Depth of Cut (d)	: 0.5, 0.75, 1 mm

.

With the four input variables, a full factorial design of the experiment was created and experimentation was conducted according to it. For each experimental run, multiple responses such as cutting temperature and main cutting force values were collected. For each experimental run, three readings were taken and the average values were taken as final reading. A photographic view of the experimental setup is presented in Figure 5.

2.3. MQL Setup and Oil Preparation

As defined before, turning experimentation was performed under MQL conditions. It is an easy-to-use, economical, eco-friendly, and effective alternative to conventional cooling, also called near-dry machining [66] or micro-lubrication [67]. The primary components of the MQL delivery system include a compressor, fluid chamber, mixing chamber, and micro nozzle. In the appropriate location, a pressure regulator and flow meter were used to regulate pressure and flow rate, respectively. The compressor provided compressed air and could generate a maximum pressure of 25 bars. Oil was poured into the mixing chamber. A portion of the compressed air traveled directly to the mixing chamber, while the remainder flowed to the fluid chamber to force the fluid out of the container and into the mixing chamber. The fluid chamber’s capacity was 1 L. It was connected to the compressor by a flexible pipe through the inlet port to maintain a constant pressure within the chamber. During machining, it was necessary to maintain a constant passage of cutting fluid into the mixing chamber via a flow control valve over an extended period. The volume of the fluid chamber was maintained at one liter because it could continue for five hours at a rate of 200 mL/h. For all four environments, the MQL technique was used. The MQL nozzle was kept at approximately 35 mm distance from the tool–workpiece interface on an inclination angle of 40°~45° with horizontal surface. A similar nozzle distance and similar inclination angle was observed in the literature [65,68]. Figure 6 shows the MQL setup for this work.

2.4. Preparation of Cutting Oil Blends

Four types of cutting fluids were used in MQL. Among them, two were coconut oil-based blends that were prepared (Figure 7). The performance of these oil blends was compared with pure coconut oil and mineral oil. For both oil blends, a magnetic stirrer machine was used for blending. For the first blends, 500 mL each of coconut oil and rice bran oil were mixed at a 1:1 ratio. The oils were blended for 30 min at a constant 250 rpm at a temperature of about 40 °C to prepare 1000 mL of blended oil [22]. For the second blend, olive oil was used instead of rice bran oil.

Table 2. Oil proportions of the formulated vegetable oil blends.

Oil Sample	Vegetable Oil	Oil Quantity	Blend Ratio
Oil blend 1	Coconut oil	500 mL	1:1
	Rice bran oil	500 mL
Oil blend 2	Coconut oil	500 mL	1:1
	Olive oil	500 mL

shows the oil blending proportions.

2.5. Measurement of Responses

Two response parameters, cutting temperature (T) and cutting force (Fc), were taken into consideration and quantified as the machining performance. The temperature of cutting was measured using a tool–work thermocouple technique, in which the tool and work material function as two dissimilar metals of the thermocouple. Cold junction was created by connecting the cold part of the tool with the cold part of the workpiece, and the thermocouple circuit was completed. Here, a Cu wire and Cu rod were used for the connection. During machining, high heat was generated in the chip–tool interface, which acted as the hot junction of the thermocouple. Voltage was generated in the thermocouple. This voltage has a relationship with the hot junction temperature if the cold junction temperature is fixed. As stated by [61], a digital multimeter (Rish multi 15S, Rishabh Instruments limited, Nashik, India) was used in between the Cu wire to record the voltage generated during machining. The workpiece and the tool were mandatorily isolated from the four-jaw self-centered chuck and tool post, respectively, due to accurate functioning of the thermocouple. In the first place, leather and in the second place, a mica sheet, were used as insulation. By this method, the average cutting temperature over the entire contact area could be measured reliably for a wide range of temperatures. For a particular experimental run, the most stable cutting temperature reading was taken from 15 mm of turning operation. The thermocouple for the tool–work is depicted schematically in Figure 8.

The calibration for this tool–work pair was carried out to convert voltage (mV) to hot junction temperature (°C). For calibration, artificial hot and cold junctions were created with the help of a chip from the workpiece and a rod of tool material. A digital multimeter (Rish multi 15S, Rishabh Instruments limited, Nashik, India) was also connected to record generated reading, and a digital thermometer (Eurotherm, Watlow, Worthing, UK) was placed near (as much as possible) the hot junction. Then, the hot junction was heated artificially and the simultaneous readings of the multimeter and thermometer were recorded and plotted in a graph, shown in Figure 9.

From the scatter graph, a regression relation was formulated. In this case, an almost linear relationship between the voltage and temperature with a 99.98% coefficient of determination was found. The regression equation stated in Equation (1) was used to determine the chip–tool interface temperature.

$$\mathrm{T}\mathrm{e}\mathrm{m}\mathrm{p}\mathrm{e}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{u}\mathrm{r}\mathrm{e}\left(\mathrm{℃}\right)=37.8+57.45\mathrm{V}\mathrm{o}\mathrm{l}\mathrm{t}\mathrm{a}\mathrm{g}\mathrm{e}\left(\mathrm{m}\mathrm{V}\right)$$

(1)

The amount of cutting force generated during machining has a significant impact on the tool’s life and wear. The principal cutting force (Fc) during hard turning at different V-f-d combinations under MQL was measured using a dynamometer (Kistler, Winterthur, Switzerland). The dynamometer was placed under the tool post and connected with a PC through a data acquisition system. During machining, several values of cutting force were stored in the PC. The average value of these values was considered as the cutting temperature value for a specific condition. The setup for force measurement is shown below in Figure 10.

2.6. RSM-Based Approach for Process Optimization

RSM is a methodical approach that is used to establish a link between the independent and dependent variables. This method can be used for experimental design, evaluation of the effects of process parameters on process output responses, identification of significant parameters and interactions contributing to the response being measured, establishment of a correlation between the significant parameters and the responses, and identification of the optimal process parameter settings intended for superior process performance within the limits chosen for the experimental design [70]. Systematic data are created and summarized for analysis by a carefully planned experimental design in the first step of RSM. Experimental design can produce the most necessary knowledge with the least amount of experimentation. RSM has the advantage of saving time and money by minimizing the number of experimental runs [71]. The second-order polynomial model required for optimization can be designed by using Box–Behnken designs (BBDs) and central composite designs (CCDs) [72]. Due to its economic character, BBD is increasingly used in actual industrial research [73].

Based on regression analysis, the mathematical model used in RSM is typically a first- or second-order polynomial model. Though the true reaction surface is curved enough to allow for variable interactions, a second-order model is likely needed. The model that was built using the above method is called a quadratic model and is shown by Equation (2) [53]. This model has linear terms, terms that are quadratic, and terms that combine.

$$\mathit{Y}={\mathit{b}}_{\mathbf{0}}+{\sum }_{\mathit{i}=\mathbf{1}}^{\mathit{k}}{\mathit{b}}_{\mathit{i}}{\mathit{x}}_{\mathit{i}}+{\sum }_{\mathit{i}=\mathbf{1}}^{\mathit{k}}{{\mathit{b}}_{\mathit{i}\mathit{i}}{\mathit{x}}_{\mathit{i}\mathit{i}}}^{\mathbf{2}}+{\sum }_{\mathit{i}\mathbf{<}\mathit{j}}{\sum }_{\mathit{j}=\mathbf{2}}^{\mathit{k}}{\mathit{b}}_{\mathit{i}\mathit{i}}{\mathit{x}}_{\mathit{i}}{\mathit{x}}_{\mathit{j}}+\mathit{\epsilon }$$

(2)

Here, ${\mathit{b}}_{\mathbf{0}}$ is the constant term; ${\mathit{b}}_{\mathit{i}}$ is the coefficient of linear terms; ${\mathit{b}}_{\mathit{i}\mathit{i}}$ represents the coefficient of the quadratic term; ${\mathit{b}}_{\mathit{i}\mathit{j}}$ is the coefficient of cross-product terms; and ${\mathit{X}}_{\mathit{i}}$ are the quality variables, which in this case are V, f, and d.

The analysis of variance (ANOVA) was carried out in Minitab 17 software. The full quadratic term was used for developing the regression equation and ANOVA table. Following the method of the reduced model, backward elimination was performed with a 95% confidence level (α = 0.05). Percent contribution (Pi) of the model and model terms on different output responses can also be determined from ANOVA based on Adj SS (adjusted sum of squares) values using Equation (3) [74]:

$${\mathit{P}}_{\mathit{i}}={\displaystyle \frac{{\mathit{S}\mathit{S}}_{\mathit{i}}}{\sum {\mathit{S}\mathit{S}}_{\mathit{i}}}}\times \mathbf{100}\%$$

(3)

where SS = sum of squares and i = model terms.

The normality and randomness of the residual values of the formulated empirical models can be studied using residual plots [75]. Normal probability plots and a histogram of residual frequencies against residuals are plotted to verify the normality of residual values. Residual versus fitted values and residual versus observed order plots are plotted to verify the randomness of the residual values.

Model verification using experimental data is required before using these models for optimization of the process parameters [76]. A validity test of a model needs to be conducted by using additional experimental data that were not used during model formulation. Any created model is considered to be valid if it can predict response data with a reasonable error. The prediction accuracy of the model can be calculated by different error measures. Among them, mean absolute percentage of error (MAPE) is an effective, simple, and widely used method for model validation [77]. MAPE is the average of the absolute percentage error. The lower the value of the MAPE of a model confirms better performance. MAPE for the models was computed by utilizing Equation (4) [78].

$$\mathrm{M}\mathrm{A}\mathrm{P}\mathrm{E}={\displaystyle \frac{1}{n}}\sum _{t=0}^n{\displaystyle \frac{\left|{\mathrm{A}\mathrm{c}\mathrm{t}\mathrm{u}\mathrm{a}\mathrm{l}}_{\mathrm{t}}-{\mathrm{P}\mathrm{r}\mathrm{e}\mathrm{d}\mathrm{i}\mathrm{c}\mathrm{t}\mathrm{e}\mathrm{d}}_{\mathrm{t}}\right|}{{\mathrm{A}\mathrm{c}\mathrm{t}\mathrm{u}\mathrm{a}\mathrm{l}}_{\mathrm{t}}}}\times 100$$

(4)

Response optimization is a process of finding the optimal values of input parameters that lead to the desired or optimal values of response variables in a given system or process. The objective is to identify the combination of input settings that maximizes or minimizes the response variable(s) based on specific goals or criteria. Models developed to predict the responses can be utilized for finding the optimum process parameters. Optimization in manufacturing is very important to achieve improved productivity, reduced production cost, and hazard-free operation as a byproduct. The optimization process can be based on multiple responses or single responses. Using the response optimizer in MINITAB is very easy and this method takes very little time. This technique uses the desirability functions approach. Every response has been translated by an individual DF into a unitless desirability value (d) that varies over the range 0 ≤ d ≤ 1.

3. Results and Discussions

3.1. Design of Experiment (DOE) and Data Collection

RSM was used for the design of the experiment. In this experiment, there were four input parameters, among which three were numerical, such as cutting speed, feed rate, and depth of cut, and other was categorical, such as cooling environment.

Table 3. Process parameters with their levels.

Sl. No	Variable	Symbols	Unit	Lower Level		Upper Level
1	Cutting Speed	V	m/min	78		149
2	Feed Rate	f	mm/rev	0.10		0.16
3	Depth of Cut	d	mm	0.5		1
4	Cooling Environment	CE	Categorical	MO	CO	CO + RBO	CO + OLO

N.B. MO—mineral oil; CO—coconut oil; RBO—rice bran oil; OLO—olive oil.

shows process input parameters with their levels. By defining the levels selected for process analysis, the Box–Behnken design (BBD) was created using MINITAB 17 software. For ‘quadratic’ mathematical model formulation, BBD is considered an efficient experimental design compared to full factorial design. It can provide maximum information from minimum experimentation as well as handle continuous and categorical factors [73].

Table 4. DOE and experimental data.

Exp No.	V (m/min)	f (mm/rev)	d (mm)	CE	T	Fc
1	78	0.1	0.75	MO	828	335
2	149	0.1	0.75	MO	898	285
3	78	0.16	0.75	MO	857	432
4	149	0.16	0.75	MO	932	407
5	78	0.13	0.5	MO	811	314
6	149	0.13	0.5	MO	898	304
7	78	0.13	1	MO	851	520
8	149	0.13	1	MO	944	471
9	113.5	0.1	0.5	MO	840	265
10	113.5	0.16	0.5	MO	863	373
11	113.5	0.1	1	MO	857	382
12	113.5	0.16	1	MO	892	481
13	113.5	0.13	0.75	MO	875	353
14	113.5	0.13	0.75	MO	869	368
15	113.5	0.13	0.75	MO	875	351
16	78	0.1	0.75	CO	782	297
17	149	0.1	0.75	CO	857	270
18	78	0.16	0.75	CO	817	425
19	149	0.16	0.75	CO	869	392
20	78	0.13	0.5	CO	748	244
21	149	0.13	0.5	CO	840	236
22	78	0.13	1	CO	800	445
23	149	0.13	1	CO	915	455
24	113.5	0.1	0.5	CO	800	210
25	113.5	0.16	0.5	CO	817	230
26	113.5	0.1	1	CO	817	324
27	113.5	0.16	1	CO	863	438
28	113.5	0.13	0.75	CO	823	312
29	113.5	0.13	0.75	CO	805	307
30	113.5	0.13	0.75	CO	811	311
31	78	0.1	0.75	CO + RBO	754	265
32	149	0.1	0.75	CO + RBO	840	245
33	78	0.16	0.75	CO + RBO	777	405
34	149	0.16	0.75	CO + RBO	857	384
35	78	0.13	0.5	CO + RBO	731	235
36	149	0.13	0.5	CO + RBO	823	230
37	78	0.13	1	CO + RBO	754	427
38	149	0.13	1	CO + RBO	851	397
39	113.5	0.1	0.5	CO + RBO	800	195
40	113.5	0.16	0.5	CO + RBO	805	254
41	113.5	0.1	1	CO + RBO	811	301
42	113.5	0.16	1	CO + RBO	840	385
43	113.5	0.13	0.75	CO + RBO	782	280
44	113.5	0.13	0.75	CO + RBO	765	290
45	113.5	0.13	0.75	CO + RBO	771	285
46	78	0.1	0.75	CO + OLO	707	221
47	149	0.1	0.75	CO + OLO	811	240
48	78	0.16	0.75	CO + OLO	754	319
49	149	0.16	0.75	CO + OLO	840	361
50	78	0.13	0.5	CO + OLO	696	242
51	149	0.13	0.5	CO + OLO	811	221
52	78	0.13	1	CO + OLO	748	418
53	149	0.13	1	CO + OLO	800	379
54	113.5	0.1	0.5	CO + OLO	777	181
55	113.5	0.16	0.5	CO + OLO	794	255
56	113.5	0.1	1	CO + OLO	794	342
57	113.5	0.16	1	CO + OLO	817	355
58	113.5	0.13	0.75	CO + OLO	777	234
59	113.5	0.13	0.75	CO + OLO	771	231
60	113.5	0.13	0.75	CO + OLO	765	236

shows the design of the experiment created, with process responses (outputs). For each of the experimental runs, three readings were taken and their average values were considered as the response values (T and Fc) for any respective experimental run.

3.2. Effect of Vegetable-Based Oil Compared to Mineral Oil

Utilizing selective data from Table 4, the effects of MQL with coconut oil, the coconut–rice bran oil blend, and the coconut–olive oil blend were compared with mineral oil.

Table 5. Percentage (%) reduction for temperature and cutting force.

Sl. No.	V m/min	F mm/rev	D mm	CO % Reduction		CO + RBO % Reduction		CO + OLO % Reduction
				T	Fc	T	Fc	T	Fc
1	78	0.1	0.75	5.56	11.35	8.94	20.9	14.62	34.03
2	149	0.1	0.75	4.57	5.27	6.46	14.04	9.69	15.79
3	78	0.13	0.5	7.77	22.30	9.87	25.16	14.19	22.93
4	149	0.13	0.5	6.46	22.37	8.36	24.35	9.69	27.31
5	78	0.13	1	6.00	14.43	11.4	17.89	12.11	19.62
6	149	0.13	1	3.08	3.40	9.86	15.72	15.26	19.54
7	113.5	0.1	0.5	4.77	20.76	4.77	26.42	7.50	31.70
8	113.5	0.16	0.5	5.34	38.34	6.73	31.91	8.00	31.64
9	113.5	0.16	1	3.26	8.94	5.83	19.96	8.41	26.20
AVG				5.20	16.35	8.02	21.82	11.05	25.42

exhibited the percentage reduction in temperature and force for each vegetable-based oil compared to mineral oil.

Figure 11 reveals that for all of the vegetable cutting oils (coconut oil, coconut–rice bran oil blend, and coconut–olive oil blend), the percentage reduction in force and temperature is always positive. That means all the vegetable oils performed better than the mineral oil. This is because of the better lubrication and viscosity of vegetable oil due to its long fatty acid compared with mineral oil. A similar result is available in the literature [30]. Among three vegetable oils, the coconut–olive oil blend resulted in the greatest force and temperature reduction, followed by the coconut–rice bran oil blend and pure coconut. The lubrication properties of olive oil were better than rice bran oil. So, the oil blend with olive oil worked better. Lubrication properties were improved when more than one oil were blended together. From the literature [22,60], a similar finding was observed.

3.3. Response Surface Modeling (RSM)

RSM-based predictive models for cutting temperature and cutting force have been formulated based on the experimental data presented in Table 4 by employing the commercial statistical software package MINITAB 17. The analysis of variance (ANOVA) table was used to evaluate the effects of various variables on the dependent variable (cutting force and temperature). The sequential sum of squares, F-value, and p-value were all included in the ANOVA table along with the percentage contribution of each factor. The p-value indicated a factor’s importance at a 95% level of confidence. The p values above 0.05 (p value > 0.05) showed that the interaction terms in this model were not significant. Therefore, only significant terms (p value < 0.05) were considered for the model using backward elimination. The relevance of a factor increases with the F-value. For the RSM cutting force and cutting temperature models, the cutting speed, feed rate, and depth of cut were all statistically significant because the p-values were less than 0.05. Determined coefficients can also be used to provide insights into model significance. The correlation between experimental data and anticipated responses was quantitatively assessed by the determination coefficient ($R^2$). The reduced quadratic models for cutting temperature after applying backward elimination for four different environments are shown in Equations (5)–(8).

$$T\left(Mineraloil\right)=867.1+1.2069V–3287f+62.50d+\mathrm{14,330}f\ast f$$

(5)

$$T\left(Coconutoil\right)=818.7+1.2069V–3287f+62.50d+\mathrm{14,330}f\ast f$$

(6)

$$T(Coconut+Rice-branoil)=791.8+1.2069V–3287f+62.50d+\mathrm{14,330}f\ast f$$

(7)

$$T(Coconut+Oliveoil)=771.9+1.2069V–3287f+62.50d+\mathrm{14,330}f\ast f$$

(8)

The analysis of variance for temperature, carried out to investigate the influence of the factors, is shown in

Table 6. ANOVA table for temperature.

Source	DF	Adj SS	Adj MS	F-Value	p-Value
Model	7	150,999	21,571.3	115.55	0.000
Linear	6	148,515	24,752.6	132.59	0.000
V	1	58,739	58,738.8	314.64	0.063
f	1	5539	5538.8	29.67	0.000
d	1	7813	7812.5	41.85	0.000
Env	3	76,425	25,475.1	136.46	0.000
Square	1	2484	2484.0	13.31	0.001
V*V	1	2484	2484.0	13.31	0.001
Error	52	9709	186.7
Lack-of-Fit	44	9295	211.2	4.10	0.020
Pure Error	8	413	51.6
Total	59	160,707
Model Summary
S	R-sq		R-sq (adj)		R-sq (pred)
13.6632	93.96%		93.15%		91.89%

. The table analysis revealed the F value and the p value. The p-value for the model was <0.5, which indicates that the model was significant with a 95% confidence level, which is supported by the previous literature [53]. In addition, the model summary generated by ANOVA revealed a high value of the determination coefficient ${(R}^2=93.96\%)$, indicating that the proposed model explained $93.96\%$of the total variations in the results. Consequently, the model could not account for $6.04\%$ of the variations. In addition, a high value of the adjusted determination coefficient (${adjustedR}^2=93.15\%$) indicated that the polynomial model was highly acceptable and could be used to predict the responses.

Adjusted R² (93.15%) and predicted R² (91.89%) values are very close to each other, which means that the model was adequate to predict the temperature. The cutting temperature predicted by this model was accurately within the range of the process parameters used. This result fully complied with the previous results [57].

The quadratic model for cutting force was formulated by using RSM. The expressions of cutting force for four different environments are shown in Equations (9)–(12).

$$Fc\left(Mineraloil\right)=420.8-6.47V+1602f-47d+0.02748V\ast V+242d\ast d$$

(9)

$$Fc\left(Coconutoil\right)=371.1-6.47V+1602f-47d+0.02748V\ast V+242d\ast d$$

(10)

$$Fc(Coconut+Ricebranoil)=349.9-6.47V+1602f-47d+0.02748V\ast V+242d\ast d$$

(11)

$$Fc(Coconut+Oliveoil)=327.1-6.47V+1602f-47d+0.02748V\ast V+242d\ast d$$

(12)

The analysis of variance for the cutting force carried out to investigate the influence of the factors in the cutting force is shown in

Table 7. ANOVA table for cutting force.

Source	DF	Adj SS	Adj MS	F-Value	p-Value
Model	8	368,502	46,063	74.45	0.000
Linear	6	348,293	58,049	93.83	0.000
V	1	2228	2228	3.60	0.063
f	1	73,920	73,920	119.48	0.000
d	1	200,186	200,186	323.57	0.000
Env	3	71,959	23,986	38.77	0.000
Square	2	20,209	10,105	16.33	0.000
V*V	1	17,817	17,817	28.80	0.000
d*d	1	3401	3401	5.50	0.023
Error	51	31,553	619
Lack-of-Fit	43	31,304	728	23.36	0.000
Pure Error	8	249	31
Total	59	400,055
Model Summary
S	R-sq		R-sq (adj)		R-sq (pred)
24.8734	92.11%		90.88%		88.90%

. From the table, the F value and the p value are revealed. The p-value for the model was <0.5, which indicates that the model was significant with a 95% confidence level, which is supported by the previous literature [53]. The model summary generated by ANOVA revealed a high value of the determination coefficient ($R^2=92.11\%$), indicating that the proposed model explained $92.11\%$ of the total variations in the results. Consequently, the model could not account for $7.89\%$ of the variations. In addition, a high value of the adjusted determination coefficient (${adjustedR}^2=90.88\%$) indicated that the polynomial model was adequately fitted to the experimental data.

There were very little differences between the adjusted R² (90.88%) and predicted R² (90.88%) values, which confirms the adequacy of the model. Within the range of process parameters used, this model can predict the cutting force accurately. The result is highly aligned with the previous literature [57].

The normal probability plot of the residuals, residual versus fit, residual histogram, and residual versus order for the response surface methodology is depicted in Figure 12. The normality plot demonstrates that the majority of points lie relatively near to the straight line, indicating that errors are normally and independently distributed. The normality assumptions for this curve are considered valid. A distinct random plot of residuals over fitted values is commendable. Admittedly, there is no indication of an increasing/decreasing pattern that would disprove the constant variance assumption. The histogram of residuals displays a normal distribution curve. The residual plot over observation order demonstrates that the values are random, with abrupt ups and downs. This demonstrates that the residuals are uncorrelated. The model’s performance was considered satisfactory based on the residual plots for both temperature and force.

For the validity testing of the regression models, nine sets of experimental data were utilized, which have been presented in

Table 8. Experimental and predicted responses for a new set of experimental runs.

Run	Input Parameters				Experimental Output		Predicted Output		APE
	V (m/min)	f (mm/rev)	D (mm)	CE	T (°C)	Fc (N)	T (°C)	Fc (N)	T (%)	Fc (%)
1	113.5	0.1	0.75	MO	905	294	866	302	4.36	2.56
2	78	0.13	0.5		836	314	807	329	3.43	4.65
3	149	0.16	1		928	485	950	518	2.41	6.84
4	113.5	0.1	0.75	CO	848	258	817	252	3.64	2.39
5	78	0.13	0.5		813	273	759	279	6.65	2.16
6	149	0.16	1		882	439	902	468	2.26	6.71
7	113.5	0.1	0.75	CO + RBO	756	217	790	231	4.53	6.28
8	78	0.13	0.5		716	240	732	258	2.24	7.37
9	149	0.16	1		842	422	875	447	3.93	5.99
10	113.5	0.1	0.75	CO + OLO	727	195	770	208	5.93	6.58
11	78	0.13	0.5		664	249	712	235	7.25	5.67
12	149	0.16	1		796	398	855	424	7.41	6.65

.

In Figure 13, the experimental values of cutting temperature and force are compared with the predicted values calculated by using the regression models. The predicted values of the responses appear to be very close to the experimental values. The absolute percent error (APE) for cutting temperature under the test conditions was calculated as 2.24~7.41%, whereas for cutting force, the values of APE were 2.16~6.71%. The mean absolute percent error (MAPE) values for the temperature and force were calculated as 4.5% and 5.32%, respectively. This result indicated a good agreement with the previous literature [79]. According to Pant and Chatterjee [80], the regression models were highly accurate in predicting the responses as the MAPE values were less than 10%.

3.4. Optimum Process Parameters Selection by Response Optimizer

RSM was used for the optimization of the cutting force and temperature. The control factors were also established to a setting corresponding to the minimum cutting temperature and force. Then, using desirability analysis, multi-response optimization was carried out. The responses were optimized by using the composite desirability function of response surface methodology. In this study, the weight for each response was set equal to 1. Different weights for the responses can be applied as per requirement. Figure 14 shows that the optimization of the temperature and cutting force, a simultaneous minimization problem of both functions, divulged the 96.04% (D = 0.9604) composite desirability at a cutting speed of 80.15 m/min, feed rate of 0.10 mm/min, and 0.5 mm depth of cut. Also, the coconut–olive oil blend was found to be an optimum cutting environment. In this way, the minimum attainable cutting force and temperature were found at 182.166 N and 714.457 °C, respectively. A composite desirability value of 96.04% indicated that the optimum process parameter settings achieved satisfactory results for all the responses. A similar result has been found in the literature [52]. Also, the individual desirability of cutting force (99.65%) was much greater than that of temperature (92.558%). So, it can be said that the optimum settings are more effective for minimizing force than minimizing temperature. This statement will be validated further from the validity test in the following section. Individual desirability values were achieved with same weight for both of the responses. For any case, if minimizing temperature is more important, then more weight should be assigned for the temperature. Similar research has been found in the literature [81].

Table 9. Optimum parameter selection.

Optimum Parameters
Cutting Speed	80.15 m/min
Feed Rate	0.10 mm/min
Depth of Cut	0.50 mm
Cutting Environment	Coconut + Olive Oil Blend

shows the final optimum results.

After obtaining the optimum parameter settings, a validity test was conducted by comparing initial settings with optimal settings of the parameters. The cutting temperature and cutting force values were hugely reduced for the optimum parameter settings predicted by RSM (

Table 10. Validity test results.

Responses	Initial Settings		RSM Settings		Improvement in Responses
	Predicted Value	Experimental Value	Predicted Value	Experimental Value
Parameter settings (V, f, d, Env)	78, 0.1, 0.75, MO		80.15, 0.1, 0.5, CO + OLO		-
Cutting temperature	823	828	714	698	15.7%
Cutting force	344	335	182	194	42%

). This validity test ensured the suitability of the application of RSM in various machining processes.

From the above table (Table 10), it can be observed that the cutting temperature and force improved in the optimal parameter settings compared with the initial parameter settings. Moreover, it has also been proved that these optimum settings were more effective in reducing the force (42%) compared with temperature (15.7%), which had a high correlation with the individual desirability values of the responses (Force-0.99 and T-0.92). Similar research is available in the literature [72,82].

4. Conclusions

From the literature, it was found that vegetable oil can be efficiently used as MQL cutting fluid in machining. Vegetable oil is nonhazardous, biodegradable, and beneficial to the environment. In the present research work, turning experimentation of medium carbon steel under vegetable oil- and mineral oil-based MQL has been conducted. The RSM module of MINITAB 17 were used to formulate predictive models of cutting temperature and cutting force. Afterward, desirability based parametric optimization was conducted to find the optimum process parameter settings considering multiple responses, such as minimizing cutting temperature and minimizing cutting force. The following observations were made as a result of this study:

• Performance of three vegetable oil-based cutting fluids were compared with mineral oil under MQL conditions. Due to the long fatty acid structure and higher viscosity index of vegetable oils, they performed better compared with mineral oils concerning cutting temperature and cutting force.
• Among the three oils, both oil blends performed better than pure coconut oil because, blending can improve the properties of oil as cutting fluids. In addition, between the two oil blends, the coconut + olive oil blend outperformed the other blend (coconut + rice bran oil).
• Based on the experimental data, predictive models for cutting temperature and cutting force were developed. Both of the models preserved over 92% accuracy (R² value). This accuracy value indicated that the regression models explained more than 92% of the variability in the response data.
• For both of the models, very little differences between the adjusted R² and predicted R² values confirmed the adequacy of the model. These models can accurately predict the responses within the range of process parameters used in the experimentation.
• By using residual plots, the normality and randomness of the residuals for both regression models were also well established. Additionally, the residuals also showed constant variance and were correlated to each other. These confirmed the reliability of the regression models and data.
• The regression models were verified with new observation data where the MAPE values were less than 10% for both responses.
• By using a composite desirability-based response optimizer, input parameters for this experiment were optimized for minimum cutting temperature and force simultaneously. Optimum process parameters were found as cutting speed at 80.15 m/min, feed rate at 0.1 mm/rev, and depth of cut at 0.5 mm under the coconut + olive oil blend-based MQL. A high composite desirability value (0.96) of the optimum process parameter settings indicated that this could achieve satisfactory results for all the responses.
• The optimum result was validated and found reasonable improvement of the responses in these settings compared with the initial settings. This optimal setting was proved to be more effective for minimizing the cutting force compared with the cutting temperature. This result was aligned with the individual desirability values of the responses (Force-0.99 and T-0.92).

5. Future Recommendations

Future research could explore how MQL can be applied to other manufacturing processes, as well as the development of new lubricants specifically tailored for MQL conditions.

• Different vegetable oils can be studied for new blending. Also, the mixing ratio can be varied in further investigation to reveal the effect.
• Due to limited resources, this work only addresses two response parameters (temperature and force), but additional response parameters, such as tool wear, surface roughness, material removal rate (MRR), etc., can be experimentally analyzed.
• The addition of nanoparticles in vegetable oil blends can be considered for advanced analysis.