1 Introduction

The in-service performances of materials most especially steels, are solely reliant on processing parameter sets. It is note-worthy to say that various methods have been developed to control response parameter measures that affect performances of components produced via machining, one of which is Minimum Quantity Lubrication (MQL). The MQL technique is one method that has proven to be a sustainable machining method as disadvantages associated with conventional wet, dry, high pressure cooling and cryogenic machining are eliminated. Kumar et al. [1] expressed that conventional flood cooling system does not meet expected outcome as much as MQL because coolant is restricted from reaching the tool/chip interface. MQL provides both cooling and lubrication to the cutting tool/workpiece and cutting tool/chip interfaces by penetrating these interfaces in mist condition. Cooling via MQL occurs in two modes which are convective and evaporative [2]. A study by Kedare et al. [3] showed that the cooling effect of MQL oil resulted in better surface finish of End-Milled bright mild steel over the conventional wet system at 225 rpm and it was attributed to lubricating capacity of the oil. Besides high cutting speed, high depth of cut and low feed rate that produced highest degree of chip formation in the work of Darsin et al. [4], it was also recorded that cutting fluid application mode had greater influence. Consequently, the authors showed that MQL oil application mode produced lower chip serration degree and this means lower plastic deformation or shearing stress occurred at the workpiece/tool interface. However, Bhosetty et al. [5] made it known that the quantity of fluid dispensed through MQL nozzle orifice has little effect on milling response parameters whereas the type of fluid was most influential. Lisowicz et al. [6] corroborates Bhosetty et al. [5] statement that oil/fluid type in MQL system has much influence on response parameters and stated further that MQL assisted turning using oil made by mixing vegetable oil and diester performed better than two other pure vegetable oil types in terms of surface roughness. Similarly, Yuan et al. [7], Sen et al. [8] and Liu et al. [9] reported that castor oil-ethanol blend (31–69%), castor-palm oil blend and castor oil-ethanol blend (1:2) respectively, performed better than respective individual base fluid in MQL assisted turning. In another study by Sadeghifar et al. [10], it was established that MQL flow rate during turning operations has influence on response parameters such as surface roughness where it was noted that lower flow rates of 3.5 and 10.0 ml/min gave the least surface roughness values at low feed rate. They also noted that flow rate had no effect on chip forms generated. Increase in flow rate from 70 to 100 ml/h was rather recorded by Pervaiz et al. [11] as a contributory factor to decreased surface roughness and thermal softening for a MQL variant which incorporated both cooling and lubrication, at low turning operation feed rate of 0.1 mm/rev. Gurbuz et al. [12] report that MQL flow rate had no significant effect on surface roughness has been refuted by earlier stated studies by Sadeghifar et al. [10] and Pervaiz et al. [11], though the authors [12] claimed it had effect on main cutting forces tool flank wear. According to the work of Ghuge et al. [13], MQL has the capability to reduce surface roughness by average of 47% and 3% compared to dry and flood cooling respectively. The authors as well gave value for cutting temperature reduction by average of 16% and 5% in the respective order of cutting condition stated previously. Same decline in surface roughness values was recorded in Hossain and Abedin [14] over the entire depth of cut range and cutting time considered for MQL assisted turning operation, against that obtained for dry and flood cooling. Yildirim et al. [15] in their turning experiment conducted revealed the preference of MQL over Cryogenic assisted MQL cooling-lubrication system such that MQL reduced surface roughness and tool wear by 13.8 and 50.67% respectively, contrary to what was stated in Urmi et al. [16]. Kumar et al. [17] comparatively studied the performances of coconut and castor oils applied via MQL and emulsified mineral oil in flood cooling for turning of Aluminium AA 150 and found that both MQL oils which are vegetable based, out-performed mineral counterpart by generating better surface finish. Nguyen et al. [18] opined that increasing feed rate or cutting speed leads to increase in metal removal rate while the surface quality is compromised for milling process aided by MQL. Feed rate was found by Mumtaz et al. [19] and Nguyen et al. [20] to be the most significant process parameter for surface roughness in milling operation. It was further explained by Mumtaz et al. [19] that increase in feed rate led to corresponding increase in surface roughness while increase in spindle speed and depth of cut resulted in decrease in surface roughness. Increasing machining speed has direct influence on decrease in surface roughness as a result of prevention in built-up-edge formation [21]. According to the study by Mia et al. [22], minimum surface roughness can be attained when machining is done at medium cutting speed, low feed rate and high depth of cut under dry cutting condition relative to other conditions considered. Low surface roughness is usually characterized by low feed rate and high cutting speed [2, 23] but the work of Mia et al. [24] showed that minimum surface roughness and maximum metal removal rate can be achieved with MQL technique at maximum cutting speed, feed rate and depth of cut set; tool flank wear became minima at the minimum cutting parameters settings with cutting speed having most influence on surface roughness. The work of Martowibowo [25] showed that surface roughness can be reduced by a factor of 5.3 while metal removal rate can be increased by multiple of 6.6 when MQL technique is employed than dry machining. More efficient heat transfer is attainable using MQL as demonstrated by Mia et al. [22], where oil droplets from MQL spray reduced cutting zone temperature than other techniques of dry, flood and solid lubricant with compressed air respectively. Singh et al. [26] noted that percentage temperature reduction by MQL is dependent on nozzle distance from the cutting zone. Nozzle position (rake and flank faces), workpiece hardness and tool type has also been reported by Masoudi et al. [27] to be of significant influence on output turning parameters. Benkai et al. [28] made it known that the higher viscosity vegetable oils for MQL possess the capability to lower cutting force but becomes non beneficial in terms of heat exchange capacity as it reduces heat exchange potentials of vegetable oils. Also, the authors propounded that saturated fatty acids in oils are more efficient in their lubricating influence than unsaturated fatty acids. Strain hardening or phase transformation of titanium alloy material machined under MQL condition is eliminated due to cooling effect offered by the technique [29]. Revankar et al. [2] reported that machining results in increased surface hardness higher than the as-received specimen due to strain hardening effect but below the top surface, reduced hardness was obtained as a result of aging. However, specimen subjected to MQL had lower hardening effect than dry machining. Results from Revankar et al. [2] corroborates that obtained in Kaynak et al. [29] who observed that MQL had the least hardening effect on specimen among the three methods of cutting fluids application systems considered. Commonly used MQL fluid types by researchers are of plant seeds sources making this technique environment friendly, cost effective and poses no health hazard. These oils offer promising alternative to petroleum based oil as they are less toxic, renewable and eco-friendly [30]. It has been recorded that vegetable oils possess polar atoms and groups such as S, O, N, –COOR, –COOH and OH which have high affinity for metals, bonding on surface of workpiece to form a molecular film layer that enhances lubrication at the cutting point [31]. Lubricating properties of vegetable oils are determined by their molecular structure and chemical compositions [31]. Best lubricating performance from vegetable oil can be obtained from oil with higher straightness and longer carbon chain length [33]. Joshy et al. [34] confirmed that the oleic acid in vegetable oils produced lowest coefficient of friction at lower load as the unsaturation caused adsorption that allowed formation of thin lubricating film. Contrarily, Xu et al. [35] established a correlation index of 0.83 between coefficient of friction and degree of fatty acid unsaturation with the assertion that the greater the degree of fatty acid unsaturation, the greater the coefficient of friction. As noted in Wang et al. [36] and corroborated by the results of Yin et al. [37], palm oil contains high level of saturated fatty acid (80%) which makes it better than other oils but as pointed out in Joshy et al. [34], this assertion is only true when it is operating at high load condition. Jatropha curcas oil as metal working fluid, has better friction coefficient than palm oil and palm fatty acid fraction under extreme pressure condition as reported in Zhang et al. [31], in respective of fatty acid saturation of palm oil. In this work, the influences of three pure vegetable oils i.e., Rubber Seed Oil (RSO), Neem Seed Oil (NSO) and Jatropha Seed Oil (JSO) with varying degree of fatty acid unsaturation, are evaluated in application as cutting fluids using MQL technique for turning operation. The fatty acid compositions of the oils are as defined in Table 1 where RSO and JSO can be seen to posse more unsaturated fatty acid than NSO. However, RSO has more of Linoleic than Oleic, compared to other fluid types and this means RSO is most fluidic as increase in Linoleic content of vegetable oil leads to decrease in viscosity[38, 39]. Other machining parameters also considered include spindle speed, cutting feed, depth of cut and workpiece type; the workpiece type is defined by the composition and hardness value. It is noteworthy that to the best of our understanding, this is probably the first time such work will be conducted as previous works have only carried out study using pure vegetable oil, synthetic oil and mixture of pure vegetable oil and ethers. Also, the fatty acid compositions of vegetable oil types employed in MQL machining are rarely studied for their effects on machining processes.

Table 1 Fatty acid compositions of fluid type
Full size table

2 Materials and methods

2.1 Characterization of specimens

Specific gravity of each fluid type was determined using a density bottle and measurement made against distilled water as a standard liquid in conformity with ASTM D1217 [53]. The thermal resistivity measurement was conducted using thermal property analyzer by Decagon Devices Inc and followed procedure contained in KD2 Pro [54]. Oil was poured in a tube, thermal needle sensor KS-1 inserted in the oil, knob was rotated to on position while the needle supplied heat to the liquid for 10 min before the process was stopped for each reading to be taken. Dynamic viscosity of each fluid type was measured using a Brookfield viscometer (Model No.: RVDVE230) with a rotary disc marked spindle 2. Measurement of dynamic viscosity was carried out in accordance with ASTM D2983-09 [55] by filling the beaker up to the marked point on the disc spindle and the spindle rotated at speed of 60 rpm. The pH measurement was performed using a digital pH meter after re-calibration using distilled water as reference. The elemental compositions of the low-alloyed carbon steel samples, categorized as Low-Alloyed Low Carbon Steel (LALCS), Low-Alloyed High Carbon Steel (LAHCS) and Low-Alloyed Medium Carbon Steel (LAMCS) were determined using Arc Spark Optical Emission Spectroscope (OES) in accordance with ASTM E415 [56] while the hardness of the samples was determined according to ASTM E18 [57] via Rockwell Hardness tester.

2.2 Experimental setup and turning operation

The turning operation was conducted on the three different grades of low-alloyed carbon steel materials designated as LALCS, LAMCS and LAHCS, with chemical compositions presented in Table 2. Each of the turned pieces was of diameter 25 mm. Controllable turning parameters applied are spindle speed, cutting feed, depth of cut, cutting fluid type and workpiece type at three different levels as coded in Table 2. The experimental design for the turning operation was based on Taguchi L27 Orthogonal Array, determined using Minitab 7, a statistical analysis software. At different level of parameters settings as defined by in the L27 orthogonal array, individual fluid type was applied using MQL technique. The controllable machining parameters employed are spindle speed, cutting feed, depth of cut, fluid type and workpiece type while the response parameters measured are surface roughness, tool temperature, chip thickness/thickness ratio and metal removal rate. The machining operation was plain turning on center lathe “Graziano Tortona” (Model No.: SAG 508) using carbide tool of 6° rake angle, fed to the workpiece at an approach angle 90°. Oil was delivered to cutting zone at flow rate 50 mL/hr through a nozzle of orifice diameter 0.2 mm positioned at an angle 45° and distance 70 mm from the tool cutting point, driven by air pressure 2 bar from a pneumatic compressor, nozzle position angle and air pressure are adopted from the work of Krishman et al. [58]. Basis for selection of the spindle speed, cutting feed and depth of cut ranges is that the present MQL turning experiment conducted considering rough turning operation.

Table 2 Controllable parameter value level and level codes
Full size table

2.3 Response parameter measurement

Cutting temperature at the tool cutting point for each experimental run was measured using a handheld infrared thermometer. The laser target of the thermometer enabled accurate targeting of laser beam to the cutting zone. Maximum temperatures attained during each turning experiment was displayed on the LED screen and recorded accordingly. Average surface roughness (Ra) of each machined piece was measured using surface profilometer (Model: SRT-6200) at a cut-off point of 0.8 µm and taken at three distinct points on each sample. Chips collected during cutting process for each experimental run was measured for their thickness using micrometer screw gauge. Subsequently, chip thickness ratios were determined based on Eq. 1.

$$ {\text{r}} = \frac{{{\text{t}}_{{\text{o}}} }}{{\text{t}}} = \frac{{{\text{F}}\sin {\uptheta }}}{{\text{t}}} $$
(1)

where to is uncut chip thickness; t is cut chip thickness; F is cutting feed; θ is approach angle.

Metal removal rates, generally defined mathematically as in Eq. 2 [59], were determined by measuring the initial diameter of each workpiece before commencement of turning operation and the final diameter measured as well after each cutting operation and Eq. 3 [60] was applied.

$$ {\text{MRR}} = \frac{{{\text{Volume}}\;{\text{of}}\;{\text{metal}}\;{\text{removed}}}}{{{\text{Cutting}}\;{\text{time}}}} $$
(2)
$$ {\text{MRR}} = \frac{{{\pi L}({\text{D}}^{2} - {\text{d}}^{2} )}}{{\frac{{{\text{4L}}}}{{{\text{FN}}}}}} = \frac{{\uppi }}{4}({\text{D}}^{2} - {\text{d}}^{2} ){\text{FN}} = \frac{{\uppi }}{2}{\text{D}}_{m} ({\text{D}} - {\text{d}}){\text{FN}}\;\;\;{\text{mm}}^{{3}} /{\text{min}} $$
(3)

where L is workpiece length; F is Cutting Feed; N is Spindle Speed; D is Initial workpiece diameter; d is Final workpiece diameter; Dm is Mean workpiece diameter.

2.4 Response data analysis

Response data obtained based on factor level settings in the Taguchi orthogonal array experimental design were analyzed using Analysis of Variance (ANOVA) at significance level of 5%, Taguchi signal-to-noise ratio (S/N) and mean-of-means. The ANOVA was conducted in order to test significance and obtain contributions of input machining parameters on response parameters. Furthermore, Taguchi signal-to-noise ratio and mean-of-means analysis were conducted to rank influence of input machining parameter on response parameter and obtain optimal value of each input parameter. The parameters’ response data generated were checked for their normal distribution at confidence interval of 99% by employing normal probability plots. Since minimum surface roughness, minimum cutting temperature, high chip thickness/thickness ratio and high metal removal rate are desired in any machining processes, objectives set for the Taguchi signal-to-noise ratio (S/N) and mean-of-means analysis are ‘Smaller is Better’, ‘Smaller is Better’, ‘Lager is Better’ and ‘Lager is Better’ respectively. These objective set for S/N are as defined mathematically in Eqs. 4 and 5 [61]. A multi-criteria optimization scheme known as Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) was adopted in order to obtain single optimum machining parameter rather than the individual optimum parameter based on each response. Employing the scheme, the response parameters were made into a single response parameter. TOPSIS is based on the idea that alternatives should possess the least distance to the ideal positive solution while having the farthest distance to the ideal negative solution and it has been executed in the present work according to Parida and Routara [62]. Decision matrix formulated according to the number of experimental runs by number of response parameters, was normalized using vector normalization [63] as in Eq. 6. Weights were applied to these normalized values (Eq. 7) to obtained weighted value Vij. Positive and negative ideal solutions were evaluated by employing scheme in Eqs. 8 and 9. Consequently, the distances of the weighted values to ideal solutions were obtained in accordance with Eqs. 10 and 11. Finally, the coefficients of closeness for each distance to ideal solutions were determined using Eq. 12 and the value closest to unity was found as the optimum.

$$ S/N = - 10\log \left( {\frac{1}{{\text{n}}}\sum\limits_{{\text{i = 1}}}^{{\text{n}}} {\frac{1}{{{\text{y}}_{{\text{i}}}^{2} }}} } \right)\;\;\;({\text{Larger}}\;{\text{is}}\;{\text{Better}}) $$
(4)
$$ S/N = - 10\log \left( {\frac{1}{{\text{n}}}\sum\limits_{{\text{i = 1}}}^{{\text{n}}} {{\text{y}}_{{\text{i}}}^{2} } } \right)\;\;\;({\text{Smaller}}\;{\text{is}}\;{\text{Better}}) $$
(5)

where \({\text{y}}_{{\text{i}}}\) is performance characteristics value; n is number of \({\text{y}}_{{\text{i}}}\) values

$$ {\text{x}}_{{{\text{ij}}}} = \frac{{{\text{y}}_{{{\text{ij}}}} }}{{\sqrt {\sum\nolimits_{{\text{i = 1}}}^{{\text{m}}} {{\text{y}}_{{{\text{ij}}}}^{2} } } }} $$
(6)
$$ {\text{v}}_{{{\text{ij}}}} = {\text{w}}_{{\text{i}}} \times {\text{x}}_{{{\text{ij}}}} $$
(7)
$$ {\text{I}}^{ + } = \{ {\text{V}}_{1}^{ + } ,{\text{V}}_{2}^{ + } \ldots \ldots \ldots \ldots \ldots \ldots \ldots ,{\text{V}}_{{\text{m}}}^{ + } \} = \{ (\left. {{\text{Max V}}_{{{\text{ij}}}} } \right|{\text{j}} \in {\text{J),}}\;{(}\left. {{\text{Min V}}_{{{\text{ij}}}} } \right|{\text{j}} \in {\text{J}}^{\prime } {\text{)\} }} $$
(8)
$$ {\text{I}}^{ - } = \{ {\text{V}}_{1}^{ - } ,{\text{V}}_{2}^{ - } \ldots \ldots \ldots \ldots \ldots \ldots \ldots ,{\text{V}}_{{\text{m}}}^{ - } \} = \{ (\left. {{\text{Min V}}_{{{\text{ij}}}} } \right|{\text{j}} \in {\text{J),}}\;{(}\left. {{\text{Max V}}_{{{\text{ij}}}} } \right|{\text{j}} \in {\text{J}}^{\prime } {\text{)\} }} $$
(9)
$$ {\text{D}}_{{\text{i}}}^{ + } = \sqrt {\sum\nolimits_{{{\text{j}} = 1}}^{{\text{n}}} {({\text{v}}_{{{\text{ij}}}} } } - {\text{v}}_{{\text{j}}}^{ + } )^{2} $$
(10)
$$ {\text{D}}_{{\text{i}}}^{ - } = \sqrt {\sum\nolimits_{{{\text{j}} = 1}}^{{\text{n}}} {({\text{v}}_{{{\text{ij}}}} } } - {\text{v}}_{{\text{j}}}^{ - } )^{2} $$
(11)
$$ {\text{CC}}_{{\text{i}}} = \frac{{{\text{D}}_{{\text{i}}}^{ - } }}{{{\text{D}}_{{\text{i}}}^{ + } + {\text{D}}_{{\text{i}}}^{ - } }} $$
(12)

where \(i\;\;is\;\;1, \ldots \ldots ,m\); j is 1,…….., n; yij is actual value of ith value of jth experimental run; xij is corresponding normalized value of ith value of jth experimental run; Wi is weigth of jth criteria; Vij is weigthed normalized value of ith value of jth experimental run; I+ is positive ideal solution (PIS); I− is negative ideal solution (NIS); \({\text{D}}_{{\text{i}}}^{ + }\) is distance from PIS; \({\text{D}}_{{\text{i}}}^{ - }\) distance from NIS; CCi is closeness coefficients of alternatives.

3 Results and discussion

3.1 Some physicochemical properties of rubber, neem and jatropha seeds oils; composition and rockwell hardness of low-alloyed steel

Contained in Table 3 are the results of specific gravity, thermal resistivity, dynamic viscosity and pH for the respective raw vegetable oil. It can be seen that rubber seed oil has the least of specific gravity value while jatropha seed oil has the highest. Based on the average values of specific gravity obtained which may not be significantly different and in order RSO < NSO < JSO, JSO has highest density. This indicates presence of higher Oleic fatty acid in JSO as shown previously in Table 1 and closely followed by NSO, as higher degree of unsaturation resulting from increased Oleic acid means higher density [64]. Unsaturated fatty acid contributes to enhanced lubricity by forming adherent thin film to prevent metal-to-metal contact [65], making JSO of higher potential to improve machinability when used as MQL oil. Highest thermal resistivity possessed by rubber seed oil showed that it is least conductive, may not be able to effectively cause cooling at the tool cutting point as the other two. Dynamic viscosity measured for the oils as indicated in Table 3 shows that value obtained for RSO < JSO < NSO and this indicates best spread-ability and penetration of RSO from the MQL nozzle into tool/chip interface and cutting zone [66]. This assertion as relating to RSO is supported by value of Linoleic acid presented previously in Table 1. The pH values of these non-edible vegetable oils are below neutral, though jatropha and rubber seeds oils tend to be close to being neutral in pH which may quite favourable for use as cutting fluids. Nevertheless, it has been confirmed in the work of Adesusi et al. [67] that these oils contain free fatty acids whose constituents are of carbonyl and hydroxyl groups that contributed to enhancement in corrosion resistance of low-alloyed steel in corrosive solution by acting as active centres. Hence, the respective pH value obtained may not be a thing of concern, even as rubber seed and jatropha seeds oils are near being neutral. Tables 4 and 5 show elemental compositions and Rockwell hardness values respectively, of low-alloyed steels as obtained.

Table 3 Physiochemical properties of rubber, neem and jatropha seeds oils
Full size table
Table 4 Elemental compositions of low-alloyed carbon steel samples
Full size table
Table 5 Rockwell hardness of low-alloyed carbon steels
Full size table

3.2 Response data and normal probability plots of response data

Contained in Table 6 are the results of Taguchi L27 experimental design showing input parameters variables combinations with their corresponding response parameters. These input parameters are spindle speed, cutting feed, depth of cut, fluid type and workpiece type differentiated by their compositions and Rockwell hardness value as given in Tables 4 and 5 respectively, as presented above. The abbreviations to these parameters are as defined in footnote of Table 6. Probability plots of response parameters SR, CT, CTR and MRR are presented in Fig. 1a–d, for checking distribution of the data generated from response parameters at confidence level of 99%. Data shown on the plotted probability distribution show means of each response parameter, standard deviations (StDev), Anderson–Darling values (AD) and p values. The StDev indicates departure of response data from their mean values and these data points are approximately aligned with the individual mean value lines of normal distribution plot (middle lines). The p values of 0.021, 0.163, 0.154 and 0.346 for SR, CT, CTR and MRR respectively are greater than 0.01 significance while AD values are small, all these are indications that response data obtained are reliable and useful for optimization of the turning process assisted by three vegetable oil based MQL system.

Table 6 Experimental design based on Taguchi L27 (35) orthogonal array
Full size table
Fig. 1
figure 1

Normal probability plots of a surface roughness, b cutting temperature, c chip thickness ratio, d metal removal rate

Full size image

3.3 Analysis of variance for response parameter data

The analysis of variance conducted at significance level of 5% (0.05) on data obtained from machining operation under MQL condition are as given in Tables 7, 8, 9 and 10 for surface roughness, cutting temperature, chip thickness ratio and metal removal rate respectively. Resulting p values of machining parameters and their interactions showed that workpiece type characterized by hardness and composition is significant (p < 0.05) in effect on the surface roughness generated under MQL condition (Table 7). The other machining input parameters have their p-values higher than 5% significance level, which render these parameters of no statistical significant effect on surface roughness. That is, the change in value of surface roughness obtained as a result of change in these parameters value whose effects are non-significant statistically, only happened by chances. Most especially, the effect of cutting fluid type was not felt in the result of surface roughness value because p-value associated is greater than 5% significance level. Hence, the type of vegetable oil adopted in this work does not have a significant effect on surface roughness of machined pieces which is in accordance with the assertion made in the work of Tuan et al. [69] who compared pure soybean oil and emulsion oil as against three pure vegetable oil considered herein. Similarly, the percentage contribution and F-value of workpiece type is the highest with values 48.77% and 19.94 respectively, followed by interaction effect of depth-of-cut and workpiece type having 23.84% contribution at R2 = 95.11%. This shows that workpiece type has the dominant effect on surface roughness values obtained and it is statistically significant. Results obtained in the work of Jeevan and Jayaram [70] which did not consider workpiece type, affirms the order of factors’ contribution in this current work based on spindle speed, cutting feed, depth of cut and fluid type. However, is unlikely that level of fatty acid unsaturation in the respective oils employed affected change in surface roughness of the machined pieces as evident from p value (p > 0.05) obtained; the lubricating effect caused by thin film formation [34] is not a function of oils’ unsaturation level. Similar to this is the work of Elewa et al. [71] who engaged in rough turning with application of five vegetable oils types as cutting and was found that four among the five produced surface roughness with no significant difference in value (3.0–3.1 µm) and this supports assertion in Kazeem et al. [72] that fluid type has lesser influence than feed rate in lowering surface roughness. Having determined that workpiece type (classified based on hardness and composition) to be the most contributing/significant factor in MQL turning for this case, it imperative to know that this was not the case for dry hard turning by Khan et al. [73]. Khan et al. [73] in their case observed that workpiece hardness has lesser significance and percentage contribution to surface roughness than insert type and feed rate, the assertion by the authors is also in contrast with work of Tuan et al. [69].

Table 7 ANOVA for surface roughness
Full size table
Table 8 ANOVA for cutting temperature
Full size table
Table 9 ANOVA for chip thickness ratio
Full size table
Table 10 ANOVA for metal removal rate
Full size table

The p values obtained for analysis of variance conducted for machining parameters to determine statistical significance of these parameters on cutting temperature measured at tool/chip interface indicate that they are of no statistical significance as all p values obtained are higher than 5% confidence as seen in Table 8. Nevertheless, spindle speed has the highest percentage contribution and F-value of 36.32% and 10.65 respectively, which is conformity with results from the work of Kazeem et al. [74]. This is followed by interaction effect of cutting feed and workpiece while the fluid-type has second least percentage contribution of 1.72% at R2 = 93.18%. The highest percentage contribution obtained from spindle is as expected as increase in spindle speed is established to result into increase in deformation at the primary shear zone which translates to increase in cutting temperature measured at tool cutting point. Since all varied machining parameters considered are of no statistical significance to the recorded tool/chip interface temperature, the variation in cutting temperature only happened by chances. Moreso, it has been stated elsewhere in Salur et al. [75] and Kazeem et al. [76] that feed rate has no significant effect on cutting temperature but on tool life and power consumption. However, the non-significant contributions made by all parameters considered for tool temperature measured is due to the low spindle speed employed for rough turning operation which is unlikely to generate wide tool temperature differential.

This case of machining response parameter “chip-thickness-ratio” seem quite different from the two previously presented results for surface roughness and cutting temperature. All the machining factors except spindle speed have their p values less than 5% significance level (p < 0.05), most especially “workpiece type” with least p value of 0.000 as shown in Table 6. This agrees with the assertion made by Khan et al. [77] that almost all parameter involved in machining have direct or indirect influence on chip thickening during deformation. It can be seen that for “chip-thickness-ratio”, the effects of these factors that have p < 0.05 are statistically significant, including “fluid-type” with p value of 0.017. Factor “workpiece-type” has the highest percentage contribution and F-value of 50.91% and 97.13 respectively, which means “workpiece-type” is most influential on “chip-thickness-ratio” and this influence is also most statistically significant. The influence of “fluid-type” on “chip-thickness-ratio” is not as pronounced as that of “cutting feed” as obtained from F-value but more influential than all interaction effects of parameters considered, including “depth of cut” and “spindle speed”. On the other hand, “fluid-type” percentage contribution is only higher than that obtained for “spindle speed” and “depth-of-cut”. Decrease in chip-thickness-ratio ultimately leads to increase in chip thickness thereby increasing chip load on tool rake face. This increase in chip load leads to corresponding increase in the temperature at the primary and secondary shear zones [72]. Hence, change in cutting feed, depth of cut, fluid type and workpiece type either cause decrease or increase in chip load at the cutting interface, resulting to increase in cutting temperature which may be effectively dissipated by applied cutting fluid depending on fluidity. It should be noted that the influence of fluid type on chip load variation is dependent on thin film formability of fluid type employed, dictated by molecular structure [32].

Presented in Table 10 is the ANOVA results generated for Metal Removal Rate (MRR) on turning the low alloy carbon steel under MQL condition involving three varieties of vegetable oils, with other turning parameters varied accordingly. It is evident from this table that “spindle speed” and “cutting feed” are the only machining parameters which have statistically significant effects on MRR with their respective p-value less than 5% significance level (p < 0.05). These two machining factors have their percentage contributions as 23.11% and 26.53% respectively, which shows that cutting feed has the highest percentage contribution to MRR and most influential of all factors based on its highest F-value of 17.62. The “workpiece” and “fluid-type” in this case are of no statistical significance in differences in value of MRR. Since the relationship between machining Power Input (P), Specific Power (Ps) and MRR is defined according to Eq. 13 [78], it is affirmed that workpiece type, fluid type and depth of cut are not determinants of power inputs in this work.

$$ {\text{P }} = {\text{ P}}_{{\text{s}}} \, \times {\text{ MRR}} $$
(13)

3.4 Taguchi signal-to-noise ratio analysis

The Taguchi signal-to-noise ratio employed here checks the relative importance and optimal parameter set of machining factors. In order to do this, the mean signal-to-noise ratios and mean-of-means are obtained for each of the factor’s levels and differences between highest and lowest mean signal-to-noise ratios and mean-of-means respectively are obtained for each factor. These differences are denoted as Delta. The Delta value determined has been used to rank factors from highest to lowest, that is “workpiece type” with Delta values 6.548 (mean signal-to-noise ratios) and 2.027 (mean-of-means) is ranked 1st in relative importance for determination of surface roughness under “smaller-is-better” objective set as set-out in Tables 11 and 12, this follows in that order. It should be noted that “fluid-type” ranked 4th, below “spindle-speed” and “cutting-feed” in signal-to-noise ratio analysis. This indicates that vegetable oil does not importantly effect surface roughness value variation as much as workpiece type, spindle speed and cutting feed for the MQL assisted turning operation. The minimization of surface roughness is based on “smaller is better” objective and in that wise, parameter level with highest “signal-to-noise ratio” and lowest “mean” values are selected. Hence, optimal parameter set for minimization of surface roughness are highest spindle speed (level 3—870 rpm), lowest cutting feed (level 1—0.08 mm/rev.), mid depth-of-cut (level 2—0.5 mm), JSO (level 3 according to L27 orthogonal array table but rearranged alphabetically by the Minitab software to become level 1 while conducting Signal-to-Noise Ratio analysis and Mean-of-Means, which is also evident in the interaction and main effect plots as shown in Figs. 2 and 3), workpiece-LAMCS (level 3 in both orthogonal array table and Minitab alphabetical rearrangement. It should be noted that LAMCS has the highest Rockwell hardness of 109 HRB). Workpiece type being the only statistically significant factor in this case can be said to have caused improvement in surface quality at highest hardness value as explained in the work of Struzikiewicz and Sioma [79] that at higher material hardness, workpiece material adhesion to the cutting edge is limited. Consequently, the formation of built-up edges that contributes to poor surface finish is controlled by the hardness level of the workpiece material. Hardest material generated lesser built-up edges and resulted in lesser surface roughness. The highest hardness value of workpiece materials producing optimal surface roughness means that machinable hardness increased [69]. There exists a difference between optimal value in terms of “spindle-speed” in signal-to-noise ratio and mean-of-means where signal-to-noise ratio detected highest spindle speed (level 3—870 rpm) as optimal while mean-of-means detected mid spindle speed (level 3—620 rpm) as the optimal. The difference in optimal values detected under signal-to-noise ratio and mean has no difference in effect judging by result from main effect plot for mean in Fig. 2b where the line joining both values seemed parallel to the mean-of-means line. However, these optimal spindle speed conform with the general knowledge that increase in spindle speed lowers surface roughness in turning operation [80] as both spindle speeds are higher than the minimum considered in this work. According to Fig. 3a, b, there exist interactions between pairs of factors, effect of a particular factor depended on the factor level of the other because the plotted lines are no-parallel to one another and optimal values obtainable from the interaction plots of signal-to-noise ratio and mean correspond with that obtained in effect plots in Fig. 2a, b, most especially with effect plot of signal-to-noise ratio for spindle speed. The phenomenon of obtaining minimum surface roughness can be explained from the perspective of turning operation mechanics as indicated in Eq. 14 [73] where surface roughness is directly proportional to square of feed rate and inversely proportional to tool nose radius. As stated in Kazeem et al. [76] and Kazeem et al. [72], increase in feed rate results in greater feed load on tool of which built-up edge is promoted with consequent emergence of greater feed marks on machined piece and vice-versa. Though not significant in effect, the cutting fluid type that optimized surface roughness is that with minimum thermal resistivity (JSO—664.40 °C−cm/W) and viscosity value (JSO—140.0 cP) that is higher than RSO but lower than NSO. The position of JSO viscosity value corresponds with its linoleic acid content, making have better flowability than NSO and its oleic acid content which is highest among the fluid types contributes to its lubricating capability. This means JSO has better capability to conduct heat away from the tool cutting point and lubricated the tool/chip interface best, thereby reducing risk of tool softening and eventual wear that generates poor surface finish. This preferred feature of raw JSO in lowering surface roughness as also been demonstrated in the work of Shashidhara and Jayaram [81] for lowering cutting power during turning operation.

$$ {\text{R}}_{{\text{a}}} = \frac{{{\text{f}}^{2} }}{{32{\text{r}}}} $$
(14)
Table 11 Surface roughness response for mean signal-to-noise ratio–smaller is better
Full size table
Table 12 Surface roughness response for mean-of-means–smaller is better
Full size table
Fig. 2
figure 2

Main Effect Plots for a Signal-to-Noise Ratio of Surface roughness, b Means of Surface roughness

Full size image
Fig. 3
figure 3

Interaction Plots for a Signal-to-Noise Ratio of Surface roughness, b Means of Surface roughness

Full size image

Presented in Tables 13 and 14 are Taguchi signal-to-noise ratio and means values for cutting temperature considering smaller-is-better objective set. The Delta value ranking in both tables show that spindle speed ranks 1st in effect on cutting temperature, followed by workpiece. It is evident from both tables that fluid type ranks 4th behind spindle speed, workpiece and cutting feed. Since the objective is smaller-is-better, determining optimal parameters values that minimize cutting temperature will be based on highest signal-to-noise ratio and lowest means. Hence, optimal parameters values as obtainable from Tables 13 and 14 are lowest spindle speed (level 1—435 rpm), mid cutting feed (level 2—0.12 mm/rev), lowest depth of cut (level 1—0.3 mm), fluid type—RSO (level 1 according to L27 orthogonal array table but rearranged alphabetically by the Minitab software to become level 3 while conducting Signal-to-Noise Ratio analysis and Mean-of-Means, which is also evident in the interaction and main effect plots as shown in Figs. 4a, b and 5a, b) and workpiece—LAHCS (level 2 according to L27 orthogonal array table but rearranged alphabetically by the Minitab software to become level 1 while conducting Signal-to-Noise Ratio analysis and Mean-of-Means, which is also evident in the interaction and main effect plots as shown in Figs. 4a, b and 5a, b). It is evident from Fig. 5a, b that effect of a particular factor depended on the factor level of the other. As expected, low spindle speed and cutting feed leading to minimized cutting temperature is a result of lowered shear deformation taking place at the primary shear zone. Also, RSO having lowest viscosity value and highest thermal resistivity resulted in minimization of cutting temperature due to its ability to flow best into the interface between cutting tool and cut chip, initiating cooling by convective and evaporative heat transfers [82]. The flowability of RSO is made possible by the higher presence of unsaturated linoleic acid which is responsible for decrease in viscosity of the fluid as previously stated.

Table 13 Cutting temperature response table for signal to noise ratio—smaller is better
Full size table
Table 14 Cutting temperature response table for means—smaller is better
Full size table
Fig. 4
figure 4

Effect plots for c signal-to-noise ratio of cutting temperature, d means of cutting temperature

Full size image
Fig. 5
figure 5

Interaction plots for a signal-to-noise ratio of cutting temperature, b means of cutting temperature

Full size image

Contained in Tables 15 and 16 are Taguchi signal-to-noise ratio and means values for Chip-Thickness-Ratio considering Larger-is-better objective set. The Delta value ranking in Tables 15 and 13 show that workpiece type ranks 1st in effect on Chip-Thickness-Ratio, followed by Cutting-Feed. Least ranked of the input parameters is spindle speed in 5th position while fluid-type ranks 4th behind depth-of-cut. The objective set being larger-is-better, optimal parameters values that maximize Chip-Thickness-Ratio is based on highest signal-to-noise ratio and highest means. Therefore, optimal parameters values as obtainable from Tables 15 and 16 are highest spindle speed (level 3—870 rpm), mid cutting feed (level 2—0.12 mm/rev), mid depth of cut (level 2—0.5 mm), fluid type—RSO (level 1 according to L27 orthogonal array table but rearranged alphabetically by the Minitab software to become level 3 while conducting Signal-to-Noise Ratio analysis and Mean-of-Means, which is also evident in the interaction and main effect plots as shown in Figs. 6a, b and 7a, b) and workpiece—LAHCS (level 2 according to L27 orthogonal array table but rearranged alphabetically by the Minitab software to become level 1 while conducting Signal-to-Noise Ratio analysis and Mean-of-Means, which is also evident in the interaction and main effect plots as shown in Figs. 6a, b and 7a, b). It is revealing in Fig. 7a, b that effect of a particular factor depended on the factor level of the other. Optimal parameter values obtained in Figs. 7a, b differ from that obtained in Fig. 6a, b and Tables 15 and 16 in terms of cutting feed and depth of cut which gave highest cutting feed (level 3—0.18 mm/rev) and lowest depth of cut (level 2—0.3 mm) as optimal values respectively. The ability of RSO to spread more effectively than other types of oil as MQL oil chiefly gives it edge to be ranked as the optimal, owing to its lowest viscosity value obtained [66]. It is this lowest viscosity value that enhances easier spread-ability through the MQL nozzle and penetration into interface between cutting tool and chip formed leading to better lubrication performance [83]. It has been stated previously that all machining parameters considered for this work did not have statistical significant contribution towards cutting temperature variation. Hence, it can be affirmed that the lubricating action of RSO occasioned by its ability to easily spread and penetrate cutting zone must have significantly resulted in reducing shear at the cutting zone. This that could have led to decrease in temperature through which chip thickness produced became comparatively lesser.

Table 15 Chip thickness ratio response for signal to noise ratio—lager is better
Full size table
Table 16 Chip thickness ratio response for means—lager is better
Full size table
Fig. 6
figure 6

Main effect plots of a signal-to-noise ratio of chip thickness ratio, b means of chip thickness ratio

Full size image
Fig. 7
figure 7

Interaction plots of a signal-to-noise ratio of chip thickness ratio, b means of chip thickness ratio

Full size image

Presented in Tables 17 and 18 are Taguchi signal-to-noise ratio and means values for metal-removal-rate considering Larger-is-better objective set. The Delta value ranking as contained in Table 17 show that spindle speed ranks 1st in effect on metal-removal-rate, followed by cutting-feed and depth-of-cut respectively. Least ranked of the input parameters is fluid-type in 5th position. Optimal values are highest spindle speed (level 3—870 rpm), highest cutting feed (level 3—0.18 mm/rev), highest depth of cut (level 3—0.7 mm), fluid type—JSO (level 3 according to L27 orthogonal array but level 1 based on Minitab software alphabetical arrangement) and workpiece—LALCS (level 1 according to L27 orthogonal array but level 2 according to Minitab software alphabetical arrangement). Contrarily, Delta values obtained in Table 18 shows that cutting feed ranks 1st followed by spindle speed which ranks 2nd, then workpiece, fluid type and depth of cut rank 3rd, 4th and 5th respectively. Optimal values obtained in this case of mean in Table 18 has highest spindle speed (level 3—870 rpm), highest cutting feed (level 3—0.18 mm/rev), mid depth of cut (level 2—0.5 mm), fluid type—RSO (level 1 according to L27 orthogonal array table but rearranged alphabetically by the Minitab software to become level 3 while conducting Signal-to-Noise Ratio analysis and Mean-of-Means, which is also evident in the interaction and main effect plots as shown in Figs. 8a, b and 9a, b) and workpiece—LALCS (level 1 according to L27 orthogonal array table but rearranged alphabetically by the Minitab software to become level 2). Optimal values obtained in Fig. 8a, b correspond with that obtained in Tables 17 and 18 respectively. According to Fig. 9a, b, effect of a particular factor depended on the factor level of the other and optimal parameter values obtained in these respective figures tally with that obtained in Tables 17 and 18, same as Fig. 8a, b.

Table 17 Metal removal rate response for signal to noise ratio—larger is better
Full size table
Table 18 Metal removal rate response table for means—larger is better
Full size table
Fig. 8
figure 8

Means and main effect plots of c signal-to-noise ratio of metal removal rate, d means of metal removal rate

Full size image
Fig. 9
figure 9

Interaction plots of a signal-to-noise ratio of metal removal rate, b means of metal removal rate

Full size image

3.5 Multiple response optimization using technique for order preference by similarity to ideal solution (TOPSIS)

The multiple response parameter values obtained previously in Table 6 are converted to a single equivalent response at each experimental level using TOPSIS as contained in Table 19 which show previously obtained raw response parameters, normalized and weighted normalized values. Equal priorities have been applied to all responses because it is assumed that all these parameters contribute to features possessed by components after machining equally and as such the weight criterion of 0.25 was evaluated using Analytic Hierarchy Process (AHP) with a consistency ratio 0.00 < 0.10. Desirability for surface roughness and cutting temperature are their minimum values while that of chip thickness ratio and metal removal rate are their maximum values. Hence, Positive Ideal Solution (PIS) for surface roughness and cutting temperature are their Minimum Weighted Normalized Values (MiWNV) as denoted symbolically with superscript ǂ while chip thickness ratio and metal removal rate had their Maximum Weighted Normalized Values (MaWNV) as denoted symbolically with superscript ˥. The Negative Ideal Solution (NIS) on the other hand, for surface roughness and cutting temperature are their MaWNV denoted symbolically with superscript ˥ and chip thickness ratio with metal removal rate having their NIS obtained as MiWNV values as denoted symbolically with superscript ǂ. Separation measures (D+ and D−) of the alternatives are given in Table 20 which is a measure of closeness to the PIS and NIS respectively. Closeness Coefficients (CCi) of the alternatives are graphed in Fig. 10 where the value 0.719233 immerged as the closest to unity, corresponding to parameter sets in experiment No. 27 having spindle speed-870 rpm, cutting feed-0.18 mm/rev, depth of cut-0.5 mm, RSO and LAMCS. These are optimum conditions for machining as the global solution for MQL assisted turning process conducted.

Table 19 Raw and normalized values of machining response parameters
Full size table
Table 20 Separation measures of alternatives
Full size table
Fig. 10
figure 10

Closeness coefficients of alternatives

Full size image

4 Conclusion

The optimization of low-alloyed steel in turning operation while applying vegetable oil-based cutting fluids via minimum quantity lubrication technique has been conducted and the following are conclusions drawn based on values obtained from TAGUCHI, ANOVA and TOPSIS optimization techniques:

  • Workpiece type, classified based on hardness has dominant effect on surface roughness value while the fluid type employed as cutting fluid in minimum quantity lubrication assisted turning has no significant effect.

  • Workpiece type which happened to be the only statistically significant factor influencing surface roughness showed that hardest material of low-alloy medium carbon steel gave the optimal surface roughness value.

  • Jatropha seed oil having highest value of oleic acid which contributed to lubricity of oil, gave optimum value of surface roughness despite the none significant effect cutting fluid type has on surface roughness.

  • Most statistically influential factor on chip thickness ratio is the workpiece type while being followed by cutting feed and cutting fluid type respectively.

  • Flowability of rubber seed oil occasioned by its highest value linoleic acid with resulting lowest viscosity value gave optimum value of chip thickness ratio.

  • Three parameters that are statistically significant on chip thickness ratio and optimized its value are low alloyed high carbon steel, 0.12 mm/rev cutting feed and rubber seed oil.

  • Spindle speed and cutting feed are the only influencing factors on metal removal rate while fluid type and workpiece type are of no statistical significance.

  • Highest values 870 rpm and 0.18 mm/rev spindle speed and cutting feed respectively, gave optimal value of metal removal rate.

  • Global optimization conducted using TOPSIS approach showed that irrespective of statistical significance, 870 rpm, 0.18 mm/rev, 0.5 mm, rubber seed oil and low-alloy low carbon steel gave optimum values of response parameters.

  • Rubber seed oil which is an inclusive parameter in the global optimization TOPSIS scheme is basically due to its low viscosity value as a result of its relatively higher linoleic acid content, making flow easily into cutting zone.