3.1. Parameters Optimization
The results of the Taguchi experiments according to L9 orthogonal array were analyzed and the response values of mean particle size distribution obtained from particle size analyzer are shown in
Table 2. All the responses with three repetitions from each experiment were evaluated individually. The results showed that each experiment exhibited a various range of particle sizes. The finest size of POA particles was obtained in experiment number nine with 0.09 (± 0.07) µm of mean particle size. Meanwhile, experiment number one obtained a greater size with 0.48 (± 0.06) µm of mean particle size.
The TEM images of the POA nanoparticles obtained in nine experiments (
Table 2) are shown from
Figure 3,
Figure 4 and
Figure 5. The particle size distribution histogram
Figure 3A–C and TEM image of POA nanoparticles from experiment one to three (E1–E3) are shown in
Figure 3. The TEM images only reveal the lateral dimension of the nanoparticle but no information on the nanoparticle height. Though, with the dimension of the width particles from the TEM images, the lateral particle size distribution can be obtained [
22]. From the histogram, the size distribution is narrow, and the major peak is centered at 480, 410, and 230 nm. The TEM image of experiment one (E1) illustrated the identical of the POA shape which is spherical. However, The TEM image significantly reveals that the smaller the particle size distribution, POA tends to agglomerate, and identification of particle shape and size becomes more complicated.
Meanwhile, a distribution curve for particle size histograms in
Figure 4D,E and TEM images from experiment four to six (E4–E6) are shown in
Figure 4. It can be observed that the TEM images illustrated the irregular shape and size of POA nanoparticles with a wide range of distribution. The morphology showed the irregular shape of POA nanoparticles probably due to rigorous impact and forces during the ball milling process. For the particle size distribution according to the experiments, a major peak is centered at 280, 190, and 150 nm, respectively. The last three experiments of L9 orthogonal array with the TEM image and distribution of particle size histograms are shown in
Figure 5. The TEM images from experiment seven to experiment nine (E7–E9) showed that the POA nanoparticles still retain the spherical morphology although through the different ball milling process parameters. This behaviour could increase the interest of POA utilization in a wide range of applications. In the meantime, the lower peak of distribution in experiment 7
Figure 5G may be due to the agglomeration of POA nanoparticles compared to experiment eight
Figure 5H and nine
Figure 5I. Furthermore, the smaller POA nanoparticles tend to adhere to each other due to growing of inter-particle bonding. It is significant to mention that there is a great possibility the aggregation of the nanoparticles occurred and leads to the declining volume of POA nanoparticles. Nevertheless, the distribution particle analysis still obtained the major peak at 110, 130, and 90 nm of each experiment, respectively, as presented in
Figure 5.
Although experiment number nine resulted in a higher distribution of POA nanoparticles, to determine the optimum parameters and most influential parameter, the signal to noise (S/N) needs to be calculated. In this experiment, the signal to noise ratio was used to identify the control factors that reduce the variability of the results by minimizing the effects of uncontrollable factors. The S/N value was used to determine the most significant parameters and which level is highly contributing to the output [
23]. The S/N output was calculated according to the “smaller is better” equation for static design. The Taguchi method provided by commercial software Minitab 16 was used to calculate the signal to noise ratio of each experiment. The computed signal to noise ratio value was presented in
Table 3. As shown in Equation (1), from the objective of the-lower-the-better quality characteristic which is to minimize the response, a greater S/N ratio corresponds to a smaller variance of the output characteristic around the desired value.
The S/N response table by factor level for the milling time, milling speed, and size of balls was created in an integrated way and the results are tabulated in
Table 4. The delta (∆) value was calculated using (highest S/N–lowest S/N) for each parameter and compared. From the point of the
“smaller is better” quality characteristic, the size of the ball’s performance showed the larger delta (∆) value compared to other parameters. This is evidence that the size of balls showed the greatest influential parameter in the ball milling process followed by milling time and milling speed. The greater delta (∆) values for a parameter, the effect of the parameter on the process will correspond to a smaller variance of the output and generate better performance of the experiment [
24,
25]. Therefore, the size of the ball’s performance was the most significant parameter for producing POA nanoparticles.
3.2. The Influence of the Studied Parameters on the Responses
The unpredictable effects of the processing parameters on POA particle size are shown in
Figure 6. According to the main effect of the S/N ratio presented in
Figure 6, the effects on the production of POA nanoparticles increase with higher milling time. Despite the milling time considered as an influential parameter in this process, nevertheless with longer milling time, the POA particles tend to agglomerate due to the rapid reduction of POA size and would be challenging to analyze by particle size analyzer. This is also supported by Goya [
26], where nanoparticles tend to agglomerate and are challenging to disperse even though using the ultrasonic treatment. Moreover, there is a possibility of powder contamination due to chipping or breakage from the stainless steel ball and the inner wall of the milling jar. As tested by Kumar and Kumar [
27], an increase of milling time will affect mineralogy and physicochemical of POA due to the creation of new surface and particle breakage, which undergoes structural changes and mechanical activation.
Lower milling speed, as illustrated in
Figure 6, showed a higher impact on the production of POA nanoparticles. In the ball milling process, rotational at higher speed could affect the particle size distribution and particle characteristics. Findings from [
16] revealed that heat would be produced due to impact and abrasion between balls and milling jar walls. Hence, the energy input into the powders will increase and could deform the properties of particles. Additionally, the high temperatures generated may also contaminate the transformations of powders. As reported by Suryanarayana [
28], higher milling intensities could increase the average crystal size due to the enhanced dynamical recrystallization. However, there are certain limitations to the maximum speed that could be engaged depending on the design of the mill and milling jar. In fact, almost all researchers applied more than 100 rpm even higher up to 2000 rpm to produce nanoparticles [
29]. Though, the ball milling practice beyond the maximum speed limit will cause the balls to pin to the inner wall of the milling jar and could not produce any impact force.
In the meantime, an increase in the size of balls in the ball milling process, as illustrated in
Figure 6, showed a significant effect in producing POA nanoparticles. The ball size is a critical parameter influencing the performance of a ball milling process. It is well known that to produce effective breakage of large particles, larger balls are needed, whereas smaller balls are needed for the effective breakage of fine particles. The optimization of balls size was very limited due to variables data obtained from the previous experiment. Nevertheless, research by Nkwanyana and Loveday [
30] found out that using a larger size of balls would contribute minimal impact on the rate production of particles. Based on the effect of product fineness study by Fuerstenau, Lutch [
31], to produce the finer particle distribution, the higher quantity of smaller balls are needed. This was due to the total expended energy between balls and particles that assist the deagglomeration of the fine particles. According to Cho and Kwon [
32], to ensure the milling efficiency of materials, a comprehensive investigation should be carried out to delineate the potential optimum parameters influencing the distribution of particles such as mill rotational speed, mill speed, mill diameter, and powder filling.
3.3. Analysis of Variance (ANOVA) Approach
ANOVA is one of the statistical models approach by using computational techniques to detect the performance of factors and their interactions by comparing the percentage of contribution of each parameter to the response [
33]. The primary purpose of applying ANOVA is to investigate which design parameters significantly affect the quality characteristic. Thus, it can be decided which independent factor dominates over the other and the percentage contribution of that particular independent variable. This analysis is measured by the sum of the squared deviations from the total mean signal to noise ratios (S/N) to obtain which parameters significantly affect the preparation of POA nanoparticles. Statistically, there is a tool called the F-test, which was named in honour of Sir Ronald Fisher. The F-test was conducted to see which design parameters have a significant effect on the quality characteristic. In this analysis, the F-ratio is a ratio of the mean square error to the residual error and is traditionally applied to determine the significance of a factor [
34].
From the results of ANOVA, the last column in
Table 5 illustrated the percentage contribution of each parameter. Percentage (%) contribution is defined as the significant rate of the process parameters and larger values represent a more significant effect on the preparation of POA nanoparticles. It can be observed that the size of balls has the most significant effect with 43.42% contribution, followed by milling time (34.08%) contribution, and milling speed (15.39%) contribution. However, the significant effect of the relationship between parameters could be seen in contributing factors (%). It is observed that contributing factors (%) reveal there was an interaction effect of process parameters. Moreover, this theory supported by Sharma and Chattopadhyaya [
35] found out that the more significant the contributions of a particular factor to the total sum of squares, the larger the ability is of that factor to influence process parameters.
The optimization results from the combination of parameters, which were predicted as 20 mm of ball’s size, 24 h milling time, and 100 rpm on milling speed, must be supported through the confirmation test. The confirmation experiments were conducted to validate the optimal parameters obtained by the Taguchi method. This test aims to validate the optimal process parameters and to show how close the respective prediction with the real experiment is. Moreover, the confirmation test is highly recommended by Taguchi to verify the accuracy of the optimal process parameters that have been selected. Under these conditions, to validate the palm oil ash (POA) properties, the particle size distribution and morphology characteristics were carried out by particle size analyzer and TEM micrograph, respectively.
3.4. Characteristic of Palm Oil Ash Morphology and Particle Size Distribution
The particle size distribution of the POA nanoparticles under predicted optimal condition from the Taguchi method was illustrated in the histogram graph in
Figure 7. A study by Akbari and Tavandashti [
36] proposed that to obtain reliable and valid frequency versus particle size curve, hundreds of particles should be measured for an optimum sample size. The histogram graph for particle size distribution shows the intensity or frequency versus the size range typically. It was observed that the majority of the POA particle size distributed between 70–110 nm of particle sizes. From the histogram profile, the average diameter of the POA nanoparticles was estimated to be 80 nm. The smallest particles were obtained with the lowest intensity and were estimated to be eight (8) nm. Meanwhile, the largest particles exhibited at 0.2% intensity were estimated to be 200 nm. Although the histogram reveals the POA nanoparticle’s distribution, the information is still inadequate to characterize the behaviour of the POA nanoparticles. Moreover, the distribution of the POA nanoparticles probably influences by the agglomeration resulting in variation of POA nanoparticles sizes. This behaviour occurred because of the robust and attractive interaction between nanoparticles derived from strong Van Der Waals force. A recent study by Aqeel Ashraf and Peng [
37] revealed that a high concentration of nanoparticles with extraordinary surface area commonly aggregated more rapidly, hence formed larger nanoparticles. Therefore, the preparation of POA nanoparticles mainly depends on the optimal parameters that could increase the number of nanoparticles and significantly decrease the agglomeration of nanoparticles. Thus, transmission electron microscope will be used to view the image of POA nanoparticles individually.
Transmission electron microscope (TEM) micrograph visualized the morphology and distribution of POA nanoparticles. Moreover, the TEM image supplies quantitative information about the shape, size, and distribution of particles on a very local scale.
Figure 7 showed the TEM image of POA nanoparticles, which is in accordance with the particle size distribution, which is evaluated by laser diffraction particle size analyzer. From the TEM image, the POA nanoparticles appear in black colour. The black colour was probably due to that less transparent towards electron beam and electronic density of POA nanoparticles element as reported by Pishvaei, Farshchi [
38], Yazdimamaghani, and Pourvala [
39]. The TEM micrograph reveals that the POA nanoparticles possess an irregular size and uniformly distributed within 70 to 200 nm range without having agglomeration. This similar finding also reported by Abdul Khalil and Rus Mahayuni [
11] that ground POA particles had an irregular shape and the size was measured less than 50 nm. In addition, within that nanometric range, there was slightly an agglomeration that occurred between POA nanoparticles. It was assumed that the agglomeration occurred between POA nanoparticles due to the increase in surface area and energy of POA nanoparticles. In this study, the ball milling process involved complex mechanical forces such as particles that are repeatedly fractured and flattened. In the case of the spherical shape of POA nanoparticles that can be observed in the TEM image, it is probably due to the balanced formation of milling force between balls, milling jar, and POA particles. Hence, the POA particles resist being fractured and flattened, causing particles to obtain a size more than 100 nm. Therefore, the confirmation test indicated that the results obtained from the selection of optimal parameters could produce a significant size of palm oil ash (POA) nanoparticles. The results confirming the success of Taguchi statistical analysis in obtaining optimal parameters to produce POA nanoparticles were presented in
Table 6.