The Application of Response Surface Methodology in the Investigation of the Tribological Behavior of Palm Cooking Oil Blended in Engine Oil

The purpose of this study was to determine the optimal design parameters and to indicate which of the design parameters are statistically significant for obtaining a low coefficient of friction (COF) and low wear rate with waste palm oil blended with SAE 40. The tribology performance was evaluated using a piston-ring-liner contact tester.The design of experiment (DOE) was constructed by using response surface methodology (RSM) to minimize the number of experimental conditions and to develop a mathematical model between the key process parameters such as rotational speeds (200 rpm to 300 rpm), volume concentration (0% to 10% waste oil), and applied loads (2 kg to 9 kg). Analysis of variance (ANOVA) test was also carried out to check the adequacy of the empirical models developed. Scanning electron microscopy (SEM) was used to examine the damage features at the worn surface under lubricant contact conditions.


Introduction
Tribological Studies of Waste Oil Biolubricant.In the late 1800s, petroleum had been discovered and that led to the replacement of animal fats, vegetable oils, and mineral oils with synthetic oils.Petroleum oil gradually started to be the main lubricant base stocks and that was because of their low cost and superior performance.Lubricants are being used widely in all fields of manufacturing and industrial applications.Studies showed that more than thirty-eight million metric tons of oils was used for lubrication techniques in 2005 for different industrial applications in the United States (USA).Lubricants are commonly used to reduce overheating and friction in various engines, machinery, turbines, and gear.
The excessive usage of petroleum-based oils has significantly contributed to the environmental pollution and triggered awareness from the environmental sectors [1].Besides that, the demands for fossil fuel and oil products are increasing in numerous areas.Based on the reported works, alternative oil should increase to cover about 36 billion gallons in 2022 [2].In other words, there is a great demand for oil in the coming few years and high attention should be paid to find alternative resources.To overcome such issue, researchers start developing an alternative fuel and/or oil products from natural resources aiming to replace the fossil products which become the main goal of many researchers, environmental and government bodies especially in the developed countries such as Australia, US, and Europe.
In the current decade, there are a few attempts aiming to study the potential of using bio-oil such as sunflower oil, castor oil, soybean oil, and pollock oil as biofuel for diesel engines.Most of the works showed good and promising results.However, there is a tribological issue raised by most of the researchers in which biofuels deteriorate the engine components.On the other hand, currently there is an effort to try to use pure bio-oil as a lubricant.From 2010 until recently, several biolubricants have been investigated in different countries [3], for example, soybean oils (the USA and South America), rapeseed oil (Europe), and palm oil (Asia) [4][5][6].Those studies are still at the initial stage and there are many 2 Advances in Tribology issues and limitations that need to be addressed before using such oil [7].Moreover, the literature highly recommends deep investigation on the performance and the potential of using biolubricants.Developing a friendly low cost biolubricant attracts the attention of researchers to use waste cooking oil as the main resources of lubricant.
Waste cooking oil can be considered the most promising bio-oil feedstock despite its drawbacks, that is, high free fatty acid (FFA) and water contents [8].As reported by many researchers, biofuels produced from waste cooking oils have numerous advantages such as low pollution (CO 2 , CO, and NO  ), low cost, and acceptable brake specific fuel consumption.An interest can be drawn to use the waste cooking oil as a lubricant.Kalam et al. [9] experimentally investigated the friction and wear characteristics of normal lubricants, which is an additive-added lubricant and waste vegetable oil-(WVO-) contaminated lubricants.The WVO-contaminated lubricants with amine phosphate as antiwear additive reduced the wear and friction coefficient and increased the viscosity; thus palm oil waste with a normal lubricant and amine phosphate additive could be used as a substitution for lubricant (maximum 4%).Based on the four-ball tribo testing result, the WVO-contaminated lubricant with the presence of antiwear additives showed promising results due to better thermal and oxidative properties of waste vegetable oils which consist of long chain saturated fatty acids [10].Masjuki and Maleque [11] have experimented the effect of palm oil diesel (POD) fuel contaminated lubricant on sliding wear of cast iron against mild steel and investigated the sliding contact using the pin-on-disc type of friction and wear apparatus.Based on the results, the use of pure commercial (0% POD contamination) lubricant resulted in a moderate wear rate while pure POD 100% lubricant produced the highest wear rate compared to other contaminated lubricants.

Lubrication and Material Preparation.
Base oil used in this experiment was SAE 40.Palm oil was chosen because it is commonly used in Malaysia.Volume concentrations of 0% and 10% waste oil were blended with base oil using magnetic stirrer and ultrasonic bath.For the preparation of waste cooking oil as biolubricant, the waste cooking oil underwent three types of processes: coarse filtering, dewatering, and fine filtering.Wear and friction performance for biolubricant were evaluated using a piston ring-liner contact tester and the material use was aluminium 6061 which is the common material for a piston ring.After the lubrication preparation had been completed, the details of lubricant compositions which are viscosity, density, and moisture content for all lubrication were determined.

Evaluation of Tribological Properties.
Wear test involves making linear movements similar to the pair of cylinderpiston ring operating under real conditions.Figures 1-3 show the picture of the wear tester and setup.The type of material for specimen used in this experiment was aluminium 6061 which are the material commonly used for a piston ring.Normal loads were applied to the device by hanging weights on the bearing lever where the piston ring sample is attached in order to produce the desired loads.The load chosen was between 2.0 kg and 9 kg.Low engine-speed intervals (200 rpm and 300 rpm) were selected during testing because such conditions generate the greatest friction in engines, particularly during the first movement and at the top dead centre (TDC) [12].The temperature used was the same as room temperature and the operating time was 10 minutes per specimen.The coefficient of friction (COF) was measured using ARDUINO Software and wear rate was determined via weight difference using weight scale with sensitivity 0.1 mg.Calculation of soefficient is shown in Figure 4.The test conditions are presented in Table 1.

Calculation of Coefficient of Friction and Specific Wear
Rate.Consider where   is coefficient of kinetic friction,   is applied force, and  is the load.For specific wear rate evaluation, Wear region where Δ is weight loss of the specimen,  1 is weight of the specimen before test, and  2 is weight of the specimen after test.
Volume loss (Δ) of the specimen is computed as per below: where  is experimental density of the specimen.The specific wear rate (  ) of the specimen was calculated using the following equation: where   is sliding distance and (m),   is normal load (N).

Design of Experiment (DOE)
2.4.1.Response Surface Methodology.The design of experiment (DOE) for this study was constructed using response surface methodology (RSM) to obtain the optimization for different parameters in the tribological behavior using Minitab software.RSM is the procedure to determine various relationships between process parameters and tribological criteria and explore the effect of these process parameters on the coupled responses Montgomery [13].RSM techniques are based on the use of factorial design in which the main effect of the factor is defined as the variation in response caused by a change in the level of the factor considered, while the other ones are kept constant [14].In order to study the effects of the tribological parameters, the two most important tribological criteria which are wear rate (WR) and coefficient of friction (COF) act as the response.Table 2 shows the suitable levels of the factors used to design the parameters for a tribological experiment while Table 3 shows the design values obtained from the Minitab.
where () is the response which is wear rate (WR) and coefficient of friction (COF).It is created by various process variables of tribological parameters. 0 ,   ,   , and   are the regression coefficients for intercept, linear, quadratic, and interaction terms, respectively.  and   are the independent variables.Contour plots were obtained using the fitted model by keeping the least effective independent variable at a constant value while changing the other two variables [15].A Box-Behnken design with three levels of variables was used for the current study.15 tribology tests were conducted according to the design as shown in Table 3.To ensure that the quadratic mathematical models for the wear rate and average COF of the analysis were adhered to, all of the experimental data were checked through the residual plot to verify that the mathematical models displayed standard normal distribution.

Physicochemical Properties of Waste Cooking Oil in Base
Lubricant.The data were used to evaluate the differences between base lubricant stock (SAE 40) and blended lubricant of palm oil and waste oil.Table 4 shows the properties of base oil (SAE 40), palm oil, and waste oil.
A good lubricant should have a high boiling point, adequate viscosity, low freezing point, high oxidation resistant, noncorrosive properties, and good thermal stability.The most important property of oil is viscosity.It indicates the resistance to flow and is directly related to temperature, pressure, and film formation.High viscosity indicates low resistance of flow [16].Lubricants are generally less dense than water.If the density of an object is less than that of water, then that object will float.This is why if there is a moisture problem in the lube system, the water settles at the bottom of the sump and is drained out first whenever the plug is pulled or the valve is opened.The density of a lubricant fluid can provide indication of its composition and nature [17].The presence of water does not only have a direct harmful effect on the machine components but it can also trigger the progress of oxidation up to tenfold increase and thus resulted in premature aging of oil [18].Less moisture content in lubricating oil indicates rust and corrosion prevention.Table 5 shows the results of blended lubricant composition for 5% and 10% waste palm oil.The analysis of variance (ANOVA) and the -ratio test were performed to justify the goodness of fit of the empirical models.The calculated values of -ratio for lack of fit were compared to the standard values of -ratio corresponding to their degrees of freedom to find the adequacy of different empirical models.The -ratio was calculated as a ratio of mean sum of the experimental error [19].

Analysis of the Developed Empirical Models and Regression Analysis.
Tables 7 and 8 represent the estimated regression coefficient and analysis of variance (ANOVA) for average COF for blended waste cooking oil with SAE oil.The fit summary recommends that the empirical model is statistically significant for the analysis of COF.The value of  2 was more than 99.10% which means that the empirical model provides an excellent explanation of the relationship between the independent variables (factors) and the response (COF).Based on Table 7, the associated  value for the model was lower than 0.05 (95% confidence interval).This indicates that the model was considered statistically significant.Meanwhile, the lack of fit of  values for the average COF models was also significant as they were less than 0.05. Figure 5 shows the residual plot for coefficient of friction.
Tables 9 and 10 represent the estimated regression coefficient and analysis of variance (ANOVA) for the specific wear rate for blended waste cooking oil with SAE oil.The fit summary recommends that the empirical model was statistically significant for the analysis of COF.The value of  2 was over 95.99% which means the empirical model provides an excellent explanation of the relationship between the independent variables (factors) and the response (WR).Based on Table 6, the associated  value for the model was lower than 0.05 (95% confidence interval).This indicates that the model was considered to be statistically significant.Meanwhile, the lack of fit of  values for the average COF models were not significant as they were more than 0.05. Figure 6 shows the residual plot for coefficient of friction.
The  values and  values in the estimated regression coefficient of wear rate in Table 10 denote the significant influence of each input variable in the models.The smaller numerical values of "" and larger values of "" signify that the related regression coefficient is highly significant [13].Equations ( 6) are the empirical equation for the average COF and the wear rate for the lubricant as the functions of independent variables of speed (), load (), and volume concentration (VC) in coded units: According to the COF model, the highest significant level was quadratic load, followed by linear load and lastly the interaction of speed and applied load, while, for specific average wear rate, quadratic volume concentration showed  the highest significance level, followed by interaction of speed and applied load and finally the linear load.For Figure 8(c), concerning the volume composition and load, the specific wear rate gradually increases as the volume composition increases.Thus, volume composition had more significant effect than speed and load.

Multiobjective Optimization Using Response Surface
Methodology.The main advantage of using response surface methodology (RSM) is that the response can be optimized by controlling the input parameters [20].The performances of wear and coefficient of friction depend not only on the lubricant properties but also on the sliding conditions of material under lubricant contact condition.In this study, the optimization was carried out in order to determine the minimum wear and friction of the blended waste oil with SAE 40 contact with aluminium 6061.Optimization of the process parameters was carried out using RSM optimization technique.Desirability for the whole process of optimization was calculated to show the feasibility of optimization to examine whether all parameters are within the working range or not.The goal was to minimize COF and WR.Table 11 shows the target value and the upper value for both response, average COF, and wear rate.Figure 9 exhibits the optimization plot for both COF and WR responses.The optimum value shown in the plot is 0.0717  for COF and 0.7380 for WR.The relevant parameters such as speed, load, and volume composition are 200 rev/min, 6.3712 kg, and 0.2020% of volume composition respectively.The composite shown in the plot is 0.87250.

Surface Texture Analysis.
There are various types of wear in mechanical systems such as abrasive wear, adhesive wear, fatigue wear, and corrosive wear.Since the lubricant regime occurred in this experiment was boundary lubrication, thereby, abrasive wear, adhesive wear, fatigue wear, and corrosive wear were observed in the wear regions [21].All these wear mechanisms were found in this experiment but most of the wear phenomena were abrasive and adhesive wear.The major wear mechanisms that can be found in the specimen were observed to be wear grooves that resulted from abrasive wear because the asperities on the hard surface of the liner samples touched the soft surface of the ring samples and had a close relationship with the thickness of lubricant film.
The SEM images of the aluminium plate shown used various types of volume concentration of waste cooking oil blended with engine oil.Referring to Figure 10, it was found that the wear decreases at 5% concentration compared to SAE 40 and wear started to increase when 10% waste oil concentration was used.This is due to the 5% concentration of waste oil that showed the highest viscosity results compared to 10% concentration because high viscosity (thick) engine oil helps to maintain a barrier between moving part and also drag the movements between two contact surfaces.

Conclusion
As conclusions, the study examined the effects of various control parameters, namely, speed, load, and volume composition on the responses of coefficient of friction and wear rate.
The following conclusion can be derived based on the results obtained: (i) The correlations between the control parameters (speed, load, and volume composition) and responses (specific wear rate and coefficient of friction) of waste oil added with standard lubricant were successfully developed using RSM.The model showed that the speed, load, and volume composition have a significant effect on coefficient of friction (COF) and specific wear rate (WR).According to the COF model, the highest significance level was quadratic load, followed by linear load and lastly the interaction of speed and applied load, while for specific average wear rate, quadratic volume concentration showed the highest significance level, followed by interaction of speed and applied load, and finally the linear load.
(ii) The predicted optimized volume composition for the input variables to produce the lowest response of specific wear rate and average COF in the range tested for waste oil blended with SAE 40 was speed (200 rev/min), load (6.3712 kg), and volume composition (0.2020%).
(iii) According to SEM analysis on the worn surfaces, the maximum wear occurred at 10% concentration of

Figure 1 :
Figure 1: Piston ring reciprocating liner test machine, contact geometry, and test specimen.

Figure 2 :
Figure 2: Lubricant bath for specimen facilitates the linear movement.

Figure 3 :Figure 4 :
Figure 3: The wear region of a specimen.
Figures 7(a), 7(b), and 7(c) represent the three-dimensional response surface plots and the contour plots of COF regarding speed, load, and volume composition.Based on Figure7(a), as the load and speed increase, the value of COF increases.Figure7(b)shows the relationship of volume composition and load by which the COF is gradually decreasing as the volume composition decreases.For Figure7(c), concerning the volume composition, the coefficient of friction reduces to the lowest at a certain speed and then increases as the speed increases even

Figures 8 (
a), 8(b), and 8(c) represent the threedimensional response surface plots and the contour plots of wear rate regarding speed, load, and volume composition.Based on Figure 8(a), as the load and speed increase, the value of specific wear rate increases.Meanwhile Figure 8(b) shows the volume composition and load by which the specific wear rate gradually increases as the volume composition increases.

Figure 5 :Figure 6 :
Figure 5: Normal probability plot and versus fits for average COF.

Table 2 :
Process parameter and its level.

Table 3 :
Design values obtained from a Minitab.
2.4.2.Mathematical Modelling Based on RSM.Response surface regression was used to construct a complete quadratic mathematical equation for wear rate and average COF.A second-order polynomial response surface empirical model can be developed as follows to evaluate the parametric effects on the various tribological criteria:

Table 4 :
Properties of baseline oil.

Table 5 :
Properties of blended volume concentration for waste oil.
Table 6 represents the results of experiments conducted to investigate the tribological properties of waste cooking oil blended with SAE 40 engine oil for different factor variables.

Table 7 :
Analysis of variance for coefficient of friction (COF).

Table 8 :
Estimated regression coefficients for COF.

Table 10 :
Estimated regression coefficients for wear rate.

Table 11 :
Target value and upper value of average COF and wear rate.