Production of Aluminum AA6061 Hybrid Nanocomposite from Waste Metal Using Hot Extrusion Process: Strength Performance and Prediction by RSM and Random Forest

To date, various studies have analysed the effects of reinforced ceramic on the properties of AA6061 recycled aluminum alloy chips, such as the tensile strength and fractography. However, a comprehensive analysis of the properties of hybrid composite with the addition of nano-silica oxide and nano-copper oxide reinforcements is still very limited. Therefore, this study aimed to optimise the factors comprising the preheating temperature (PHT), preheating time (PHti), and volume fraction (VF) of reinforcements then determine their impacts on the physical and mechanical properties of the recycled solid-state extruded composite aluminum chips. A total of 45 specimens were fabricated through the hot extrusion technique. The response surface methodology (RSM) was employed to study the optimisation at a PHT range of 450–550 °C with PHti of 1–3 h and VF of 1–3 vol% for both reinforcements (SiO2 and CuO). Moreover, a random forest (RF) model was developed to optimize the model based on a metaheuristic method to improve the model performance. Based on the experimental results the RF model achieve better results than response surface methodology (RSM). The functional quadratic regression is curvature and the tested variable shows stable close data of the mean 0 and α2. Based on the Pareto analysis, the PHT and VF were key variables that significantly affected the UTS, microhardness, and density of the product. The maximum properties were achieved at an optimum PHT, PHti, and VF of 541 °C, 2.25 h, 1 vol% SiO2 and 2.13 vol% CuO, respectively. Furthermore, the morphological results of the tensile fractured surface revealed the homogenous distribution of nano-reinforced CuO and SiO2 particles in the specimens’ structure.


Introduction
Composite materials have useful properties due to their constituent matrix and reinforced materials [1,2]. One of the latest generations of composites to date is the heterogeneous hybrid composite that comprises at least three distinct components or phases with various shapes and compositions.
The conventional approach for aluminum production from ores requires approximately 113 GJ per tonne of aluminum, while the secondary fabrication and production from conventional aluminum recycling methods from scrap requires around 13.6 GJ per tonne of energy [3,4]. In contrast, the production of recycled aluminum can save up to 88% The billet was passed through the extrusion process with pre-designated parameters and conditions, as shown in Table 1. The parameters for PHT, PHti, and VF were varied between 450-550 °C, 1-3 h, and 1-3%, respectively. The temperature limit was set to 550 °C to avoid the formation of hot cracks at the surface of the final extruded products [28]. A ceramic heater (SOV140B), was employed to achieve the required temperature. In addition, a graphite-based lubricant was applied to the inner surface of the extrusion die and cylindrical container during the extrusion process to avoid the increase of the extrusion load due to frictional forces. The final extruded products were machined to prepare tensile test samples according to the ASTM E8/E8M-15a standard.  The billet was passed through the extrusion process with pre-designated parameters and conditions, as shown in Table 1. The parameters for PHT, PHti, and VF were varied between 450-550 • C, 1-3 h, and 1-3%, respectively. The temperature limit was set to 550 • C to avoid the formation of hot cracks at the surface of the final extruded products [28]. A ceramic heater (SOV140B), was employed to achieve the required temperature. In addition, a graphite-based lubricant was applied to the inner surface of the extrusion die and cylindrical container during the extrusion process to avoid the increase of the extrusion load due to frictional forces. The final extruded products were machined to prepare tensile test samples according to the ASTM E8/E8M-15a standard.

Response Surface Methodology (RSM)
The Response Surface Methodology (RSM) through the Design of Experiments (DOE) was employed to analyse the key parameters affecting the tensile strength of the extruded products. Given that RSM provides an appropriate direction required to optimise the UTS, the number of experimental runs was obtained to optimise the most suitable parameters expected to influence the response. Two replicates were included to analyse the effects of PHT, PHti, and VF of CuO and SiO 2 on the tensile strength. Three centre points were involved in the full factorial design to examine the curvature effect. Then the RSM model (Equation (1)) suggested the adequacy and capability of the linear model in defining the relationship between the process factors over the response or otherwise.
In the regression equation, the letter y is the response variable, b 0 is the constant, b 1 , b 2 , . . . , b k are the coefficients and X 1 , X 2 , . . . , X k are the values of the terms. The interactions between different factors were also investigated. The factors considered and associated ranking levels during the full factorial design are presented in Table 2.

Random Forest Model
The Random Forest (RF) model consists of multiple decision trees (h) that are used and developed with the help of a random subset of different variables that categorise and replace the independent original data sets [32]. With the use of binary partitions known as splits on different variables, the DT is isolated in the predictor space. The tree "root" node constitutes and forms the comprehensive predictor space. The non-split nodes are referred to as the "terminal nodes" and form the final partition of the predictor space. According to the value of one of the predictor variables, each non-terminal node splits into two left and right descendant nodes. The data is separated into the two descendant nodes once a split had been chosen, and each of these nodes is classified in the same way as the original node. RF is an assimilated classifier, which is made up of a group of DT classifiers and is expressed as Equation (2): R = h (x., θ k ); k = 1,2, . . . (2) where θ k represents the random vector that observes identical and independent distribution and K represents the number of DT in an RF [33]. Each DT classifier defines the optimum classification result by voting in the case of a given independent variable x.
The RF model applied in this study consisted of 100 separate DT estimators. The Mean Square Error (MSE) was used in each DT to measure the quality of the split. The minimum number of samples needed to split an internal node and to be at a leaf was set to 2 and 1, respectively [34]. The maximum depth of all trees was selected to none (nodes keep expanding until the leaves contains less than min sample split). The mentioned hyperparameters are the default RF parameters used in Scikit-learn. The hyper-parameter optimisation was performed to model and reach the ultimate performance. The RF model was also implemented with the use of the Scikit-learn framework in Python. Particle Swarm Optimisation (PSO) is an important set of evolutionary algorithms commonly employed in optimisation issues and problems. PSO algorithms have been frequently used in biological population studies to analyse individual and social behaviours [35]. PSO has been used and successfully employed to enable the groups of particles, known as a swarm to traverse the search space in a semi-random manner. PSO algorithms can also determine and identify the optimum solutions by sharing information with individual particles groups. In PSO analysis, a group of n particles within a swarm S (Equation (3)) is used, while each particle of S i is represented by a vector [36]: where x i represents the current position, v i represents the current velocity, and p i shows the well-known best position within the swarm. After the position and velocity for each particle were identified, the current position and recorded position were evaluated and analysed with the performance score. In the next and upcoming iteration, the velocity v i for each particle was altered based on the previous position p i and the current global optimal position p, as defined in the following Equations (4) and (5): where ω represents the inertia factor, c 1 and c 2 are the acceleration constants, r 1 and r 2 represent the random numbers in the interval of [0, 1], and k represents the iteration numbers. The above processes were used until convergence or termination constraints was achieved.

RF Hyper-Parameter Optimisation and Evaluation
An important part of the process for the development of the RF model was measuring the model accuracy in order to analyse the performance and predictions. The Mean Squared Error (MSE), Mean Absolute Error (MAE), and correlation coefficient (R 2 ) were primarily used during the regression analysis to determine the rate of prediction error and efficiency. The MSE describes the difference between both original and predicted values that were calculated and derived from the square average difference in the data set. In addition, the MAE provides the discrepancy and difference within the original and predicted values due to the average absolute difference in the data set. The value of R 2 explains the accuracy of the matched values with the original values. The higher value of R 2 from a scale between 0 and 1 indicates a good RF model. The fitting error during the prediction and training analysis in the developed model was validated and tested through MSE, MAE, and R 2 , which were calculated through Equations (6)- (8).
The performance metric, evaluation, and analysis methods were carefully configured. For various experiment through the data sets, the 10-fold cross-validation was used to evaluate the performance of the model. For each experiment, the optimum RF model architecture having the lowest MSE and the optimum hyper-parameter configuration would be returned. The hyper-parameter configurations are presented in Table 3.  [1,4] Parameter tuning in machine learning models is very important in order to obtain desired and required results. Through PSO optimisation, the hyper-parameters were tuned in such possible ways to obtain the best results during the use of the RF model during experimentation. The values for tuned hyper-parameters in the RF model, the random state guaranteed the generation of splits that were reproducible. The random state was applied in the train test split to provide a similar set of train and test data points every time and would not assist during debugging problems and issues.
Although the generation of more trees yielded better results, certain issues and limitations were still present. The time complexity of RF improved through the selection of various large numbers of trees. An adequate estimator's value was required for better results. The max depth of a tree is known as the largest and longest path between the leaf and the root node. The parameter that formed the DT in RF to split the minimum number of observations required for any given model is known as the min sample split. The hyper-parameter result is shown in Table 4.

Comparison between Single and Hybrid Nano-Reinforcement
According to Figure 2, the SiO 2 reinforcement in the aluminum alloy chips improved the UTS by 22%, while the CuO reinforcement improved the UTS by 16%. However, the addition of 2 vol% SiO 2 and 2 vol% CuO improved the UTS by 11% only. It was important to note that the UTS of a composite depends on various factors, such as reinforcement particle size, dispersion, morphology, reinforcement vol%, process parameter, and heat treatment. The RSM process was conducted using the full factorial design, which produced 35 runs followed by 10 runs. The suggested sequence was used in conducting each experiment. The maximum UTS of 239.86 MPa was obtained at PHT, PHti, and VF of 550 • C, 1 h, and 1 vol% of CuO and 3 vol% of SiO 2 , respectively. The UTS of the nanocomposite was significantly improved with the addition of reinforcements compared to pure aluminum alloy chips. As a result of the maximum PHT and the minimum VF of CuO and SiO 2 , the recycled nature of the aluminum chips and the addition of reinforcements improved the UTS of the fabricated composite from the AA6061 alloy.
The least UTS of 127.91 MPa was obtained at PHT, PHti, and VF of 450 • C, 1 h, and 1 vol% of CuO and 1 vol% of SiO 2 , respectively. This was a clear indication that lower PHT and lower VF of the reinforcements were significant towards the tensile strength of the recycled solid-state extruded AA6061 aluminum alloy chips. Although the addition of CuO was supposed to increase the density, instead, the density decreased due to the rise of SiO 2 in the composite powder [14], as shown in Figure 3a. The reduction was compensated with the high sintering temperature since the activation energy required to drive the sintering mechanism leads to the neck growth as a result of the surface and volume diffusion. The microhardness of the material, which refers to the hardness of the composite, increased with the temperature increase due to the reduction in the nano-reinforcement [37] and the fine grain size of the constituents. In addition, the SiO2 and CuO also improved the microhardness [17,38,39], as shown in Figure 3b. Pure aluminum alloy chips have a low hardness due to the non-uniform size of the chips, which was found to be the main reason behind the fluctuating values in some of the experimental results. These results indicated after produced hybrid composite, the uniform and homogenous dispersion of reinforcement particles, development of strong interfacial properties between matrix and reinforcement material, which led to the uniform distribution of loads in the composite samples to improve these mechanical properties.

Response Surface Methodology (RSM) for UTS
Following the interpretations of the RSM, it was obvious that the list of key parameters on the investigated factors affecting the UTS of extruded samples was PHT and VF. SEM showed the sources of variance from Table 5, displaying the p-value for the liner model, temperature, CuO, and SiO2 was significant, but the lack of fit was not significant. On the list of factors, these factors were indicated by p < 0.05, as shown in the Pareto chart in Figure 4. Other parameters, such as PHti (B), the combination of the PHT and PHti (AB), were insignificant. In addition, the mixture of PHT, PHti, and VF (ABC) was also insignificant towards the UTS [40].   The microhardness of the material, which refers to the hardness of the composite, increased with the temperature increase due to the reduction in the nano-reinforcement [37] and the fine grain size of the constituents. In addition, the SiO2 and CuO also improved the microhardness [17,38,39], as shown in Figure 3b. Pure aluminum alloy chips have a low hardness due to the non-uniform size of the chips, which was found to be the main reason behind the fluctuating values in some of the experimental results. These results indicated after produced hybrid composite, the uniform and homogenous dispersion of reinforcement particles, development of strong interfacial properties between matrix and reinforcement material, which led to the uniform distribution of loads in the composite samples to improve these mechanical properties.

Response Surface Methodology (RSM) for UTS
Following the interpretations of the RSM, it was obvious that the list of key parameters on the investigated factors affecting the UTS of extruded samples was PHT and VF. SEM showed the sources of variance from Table 5, displaying the p-value for the liner model, temperature, CuO, and SiO2 was significant, but the lack of fit was not significant. On the list of factors, these factors were indicated by p < 0.05, as shown in the Pareto chart in Figure 4. Other parameters, such as PHti (B), the combination of the PHT and PHti (AB), were insignificant. In addition, the mixture of PHT, PHti, and VF (ABC) was also insignificant towards the UTS [  The microhardness of the material, which refers to the hardness of the composite, increased with the temperature increase due to the reduction in the nano-reinforcement [37] and the fine grain size of the constituents. In addition, the SiO 2 and CuO also improved the microhardness [17,38,39], as shown in Figure 3b. Pure aluminum alloy chips have a low hardness due to the non-uniform size of the chips, which was found to be the main reason behind the fluctuating values in some of the experimental results. These results indicated after produced hybrid composite, the uniform and homogenous dispersion of reinforcement particles, development of strong interfacial properties between matrix and reinforcement material, which led to the uniform distribution of loads in the composite samples to improve these mechanical properties.

Response Surface Methodology (RSM) for UTS
Following the interpretations of the RSM, it was obvious that the list of key parameters on the investigated factors affecting the UTS of extruded samples was PHT and VF. SEM showed the sources of variance from Table 5, displaying the p-value for the liner model, temperature, CuO, and SiO 2 was significant, but the lack of fit was not significant. On the list of factors, these factors were indicated by p < 0.05, as shown in the Pareto chart in Figure 4. Other parameters, such as PHti (B), the combination of the PHT and PHti (AB), were insignificant. In addition, the mixture of PHT, PHti, and VF (ABC) was also insignificant towards the UTS [40].   After eliminating the non-significant terms, the final regression model to predict the ultimate tensile stress of the developed reinforced nanocomposite was expressed using Equation (9)   After eliminating the non-significant terms, the final regression model to predict the ultimate tensile stress of the developed reinforced nanocomposite was expressed using Equation (9) below: According to the regression coefficients and analysis of variances Figure 5, the quadratic model suggested the explanation of the effects of PHT and the mixing between SiO 2 and CuO volume fraction to analyse the UTS for the extruded composite samples. Moreover, the RSM indicated various processing parameters (PHT (A), PHT*, PHti (AB), VF (SiO 2 )*, VF (CuO) (CD), VF (CuO) (C), VF (CuO)*, VF (CuO) (CC), VF (SiO 2 ), and PHti (BB)) were important. The overall quality of the model was analysed and evaluated through the results of R 2 , adjusted R 2 (adj. R 2 ), predicted R 2 (pred. R 2 ), and proper, precise values. The quadratic model was obtained along with mathematical modelling with a strong and accurate determination of R 2 = 92.24%, which was a better regression model fit to the research studies. The adj. R 2 and pred. R 2 values were 88.23% and 80.84%, respectively. A great agreement between the adj. R 2 and pred. R 2 prevented the over-fitting of the mathematical model [41]. Table 6 shows the R results for UTS by RSM. According to the regression coefficients and analysis of variances Figure 5, the quadratic model suggested the explanation of the effects of PHT and the mixing between SiO2 and CuO volume fraction to analyse the UTS for the extruded composite samples. Moreover, the RSM indicated various processing parameters (PHT (A), PHT*, PHti (AB), VF (SiO2)*, VF (CuO) (CD), VF (CuO) (C), VF (CuO)*, VF (CuO) (CC), VF (SiO2), and PHti (BB)) were important. The overall quality of the model was analysed and evaluated through the results of R 2 , adjusted R 2 (adj. R 2 ), predicted R 2 (pred. R 2 ), and proper, precise values. The quadratic model was obtained along with mathematical modelling with a strong and accurate determination of R 2 = 92.24%, which was a better regression model fit to the research studies. The adj. R 2 and pred. R 2 values were 88.23% and 80.84%, respectively. A great agreement between the adj. R 2 and pred. R 2 prevented the over-fitting of the mathematical model [41]. Table 6 shows the R results for UTS by RSM.  From the microstructural point of view, a higher temperature was a requisite in obtaining a refined microstructure. The specimen extruded using the highest temperature may result in higher tensile strength. The explanation for this observed relationship was due to the fact that an increase in temperature strengthened the aluminum matrix composite AMC better by supporting the bonding of the matrix, SiO2, and CuO. Therefore, increasing the VF of SiO2 to 3 vol% and decreasing the VF of CuO to 1 vol% supported the UTS. On the other hand, the decrease of reinforcements showed that the extruded billets had the least UTS. The results presented were in agreement with the findings reported by Shazarel et al. [42]. It was obvious from the main effect plot shown in Figure 6, that the relationship between the parameters was close. All the centre points were quite close to the lines connecting the average tensile strength from low to high setting for PHT and from high to low setting for VF.  From the microstructural point of view, a higher temperature was a requisite in obtaining a refined microstructure. The specimen extruded using the highest temperature may result in higher tensile strength. The explanation for this observed relationship was due to the fact that an increase in temperature strengthened the aluminum matrix composite AMC better by supporting the bonding of the matrix, SiO 2 , and CuO. Therefore, increasing the VF of SiO 2 to 3 vol% and decreasing the VF of CuO to 1 vol% supported the UTS. On the other hand, the decrease of reinforcements showed that the extruded billets had the least UTS. The results presented were in agreement with the findings reported by Shazarel et al. [42]. It was obvious from the main effect plot shown in Figure 6, that the relationship between the parameters was close. All the centre points were quite close to the lines connecting the average tensile strength from low to high setting for PHT and from high to low setting for VF. Materials 2021, 14, x FOR PEER REVIEW 10 of 26 The plot was relevant to justify the trends observed in the experimental results and the relationship described in Figure 5 earlier. An interaction plot was developed during the analysis of the fitness of the proposed model to represent the relationship between the parameters of interest and the UTS, as shown in Figure 7.  A very similar trend was observed in the interaction plot. The model selected was sufficient in describing the effect of all the parameters on UTS. Generally, the UTS was higher when the PHT and PHti increased. On the other hand, the curvature effect was not significant, indicating that the level of interference of unwanted factors on the results reported was negligible compared to the factors captured by the presented model. The plot was relevant to justify the trends observed in the experimental results and the relationship described in Figure 5 earlier. An interaction plot was developed during the analysis of the fitness of the proposed model to represent the relationship between the parameters of interest and the UTS, as shown in Figure 7. The plot was relevant to justify the trends observed in the experimental results and the relationship described in Figure 5 earlier. An interaction plot was developed during the analysis of the fitness of the proposed model to represent the relationship between the parameters of interest and the UTS, as shown in Figure 7.  A very similar trend was observed in the interaction plot. The model selected was sufficient in describing the effect of all the parameters on UTS. Generally, the UTS was higher when the PHT and PHti increased. On the other hand, the curvature effect was not significant, indicating that the level of interference of unwanted factors on the results reported was negligible compared to the factors captured by the presented model. A very similar trend was observed in the interaction plot. The model selected was sufficient in describing the effect of all the parameters on UTS. Generally, the UTS was higher when the PHT and PHti increased. On the other hand, the curvature effect was not significant, indicating that the level of interference of unwanted factors on the results reported was negligible compared to the factors captured by the presented model.

Optimisation of UTS
The specific function of the response optimiser was to identify the effect of individual parameters on the UTS. The parameters from this analysis were a set of specific values in terms of all the parameters that were useful to obtain the optimum UTS. The optimum value for UTS from the response optimiser approach is presented in Figure 8. The resultant composite specimen, according to the optimised PHT, PHti, and VF of 550 • C, 1.46 h, and 1.52 vol% CuO and 3 vol% SiO 2 , respectively, was used to validate the developed quadratic regression model, as shown in Figure 9. The value of the experimental UTS for the specimen was 245.51 Mpa. Optimisation of UTS The specific function of the response optimiser was to identify the effect of individual parameters on the UTS. The parameters from this analysis were a set of specific values in terms of all the parameters that were useful to obtain the optimum UTS. The optimum value for UTS from the response optimiser approach is presented in Figure 8. The resultant composite specimen, according to the optimised PHT, PHti, and VF of 550 °C, 1.46 h, and 1.52 vol% CuO and 3 vol% SiO2, respectively, was used to validate the developed quadratic regression model, as shown in Figure 9. The value of the experimental UTS for the specimen was 245.51 Mpa.  The established relationship and the effect of the reinforcement in addition to the VF of CuO increased the UTS, which then decreased after the addition of 1.58 vol% SiO2. Figure 10a shows the response surfaces plot, which was further used to visualise the significance of these parameters, while Figure 10b shows the relationship and effect of the PHT and increase in PHti during the analysis of UTS. Based on the results, the UTS improved significantly due to the increase in PHT and PHti. Furthermore, Figure 10c shows the contour plot of UTS with the interaction between the CuO and SiO2. According to the contour plot, it can be concluded that the UTS of the composite increased after the addition of 2.75 vol% SiO2 and 2.25 vol% CuO. Figure 10d shows the contour plot of UTS that describes Optimisation of UTS The specific function of the response optimiser was to identify the effect of individual parameters on the UTS. The parameters from this analysis were a set of specific values in terms of all the parameters that were useful to obtain the optimum UTS. The optimum value for UTS from the response optimiser approach is presented in Figure 8. The resultant composite specimen, according to the optimised PHT, PHti, and VF of 550 °C, 1.46 h, and 1.52 vol% CuO and 3 vol% SiO2, respectively, was used to validate the developed quadratic regression model, as shown in Figure 9. The value of the experimental UTS for the specimen was 245.51 Mpa.  The established relationship and the effect of the reinforcement in addition to the VF of CuO increased the UTS, which then decreased after the addition of 1.58 vol% SiO2. Figure 10a shows the response surfaces plot, which was further used to visualise the significance of these parameters, while Figure 10b shows the relationship and effect of the PHT and increase in PHti during the analysis of UTS. Based on the results, the UTS improved significantly due to the increase in PHT and PHti. Furthermore, Figure 10c shows the contour plot of UTS with the interaction between the CuO and SiO2. According to the contour plot, it can be concluded that the UTS of the composite increased after the addition of 2.75 vol% SiO2 and 2.25 vol% CuO. Figure 10d shows the contour plot of UTS that describes The established relationship and the effect of the reinforcement in addition to the VF of CuO increased the UTS, which then decreased after the addition of 1.58 vol% SiO 2 . Figure 10a shows the response surfaces plot, which was further used to visualise the significance of these parameters, while Figure 10b shows the relationship and effect of the PHT and increase in PHti during the analysis of UTS. Based on the results, the UTS improved significantly due to the increase in PHT and PHti. Furthermore, Figure 10c shows the contour plot of UTS with the interaction between the CuO and SiO 2 . According to the contour plot, it can be concluded that the UTS of the composite increased after the addition of 2.75 vol% SiO 2 and 2.25 vol% CuO. Figure 10d shows the contour plot of UTS that describes the interaction between the PHT and VF of the aluminum. It was clear that the increment of PHT increased the UTS, while the optimum UTS was obtained with 1.58 vol% CuO and 3 vol% SiO 2 . A comparable relationship was reported in the study conducted by Madeva et al. [43]. the interaction between the PHT and VF of the aluminum. It was clear that the increment of PHT increased the UTS, while the optimum UTS was obtained with 1.58 vol% CuO and 3 vol% SiO2. A comparable relationship was reported in the study conducted by Madeva et al. [43].

RSM for Microhardness
The DOE results of the microhardness are presented the model analysis proposed the determination of the levels of related experimental variables are shown in Table 7. Since the calculated p-value of the developed model was less than the standard (p < 0.05), the model was statistically significant, as described in previous literature [44,45]. The effect of PHT on the microhardness is shown in the Pareto chart in Figure 11. The model showed significant results according to the RSM of microhardness results versus PHT, PHti, and VF.

RSM for Microhardness
The DOE results of the microhardness are presented the model analysis proposed the determination of the levels of related experimental variables are shown in Table 7. Since the calculated p-value of the developed model was less than the standard (p < 0.05), the model was statistically significant, as described in previous literature [44,45]. The effect of PHT on the microhardness is shown in the Pareto chart in Figure 11. The model showed significant results according to the RSM of microhardness results versus PHT, PHti, and VF.
For microhardness, the values of R 2 , adj. R 2 , and pred. R 2 are shown in Table 8. The R 2 value of 0.9054 was favourable for this analysis and provided an excellent explanation of the relationship between the independent variables and the response. The pred. R 2 of 0.8565 was within a reasonable agreement with the adj. R 2 of 0.7679 [41]. For microhardness, the values of R 2 , adj. R 2 , and pred. R 2 are shown in Table 8. The R 2 value of 0.9054 was favourable for this analysis and provided an excellent explanation of the relationship between the independent variables and the response. The pred. R 2 of 0.8565 was within a reasonable agreement with the adj. R 2 of 0.7679 [41]. Table 8. R value for microhardness by RSM.

R-Sq R-Sq (Adj) R-Sq (Pred) 90.54%
85.65% 76.79% Figure 11. Pareto chart of the standardized effects for microhardness. Figure 11. Pareto chart of the standardized effects for microhardness. The factors affecting the microhardness of composite samples were evaluated from the factorial model. The PHT was observed to have direct effects on the hardness of composite samples. The lower values of microhardness were observed at low PHT and low VF. In comparison, the high PHT and VF of SiO 2 and CuO led to the strengthening of the composite, especially SiO 2 that had a great effect on the microhardness. A low VF resulted in a low microhardness [46,47]. Figure 11 shows the Pareto chart of the standardised effects. The microhardness was affected mainly by PHT and the VF, while PHti only showed minor effects.
Furthermore, the contour plot for the microhardness shown in Figure 12 indicates that any difference in VF of SiO 2 and CuO produced the highest microhardness when the temperature increased to the peak value of 521 • C. After eliminating the non-significant terms, the final regression model to predict the microhardness of the developed reinforced composite was expressed using Equation (10) below: The factors affecting the microhardness of composite samples were evaluated from the factorial model. The PHT was observed to have direct effects on the hardness of composite samples. The lower values of microhardness were observed at low PHT and low VF. In comparison, the high PHT and VF of SiO2 and CuO led to the strengthening of the composite, especially SiO2 that had a great effect on the microhardness. A low VF resulted in a low microhardness [46,47]. Figure 11 shows the Pareto chart of the standardised effects. The microhardness was affected mainly by PHT and the VF, while PHti only showed minor effects.
Furthermore, the contour plot for the microhardness shown in Figure 12 indicates that any difference in VF of SiO2 and CuO produced the highest microhardness when the temperature increased to the peak value of 521 °C. After eliminating the non-significant terms, the final regression model to predict the microhardness of the developed reinforced composite was expressed using Equation (10)

Optimisation of Microhardness
The optimised result was relatively close to the previous optimisation results. The maximum microhardness obtained at 58.69 HV coincided with the PHT, PHti, and VF of 521 °C, 3 h, and 2.75 vol% CuO, respectively, as shown in Figure 13. The optimised solution was consistent with the experimental results of 61.32 HV. These experimental results supported the discussion earlier related to the influence of process parameters on the response [48]. The equation indicates that the PHT has a significant effect on the microhardness.

Optimisation of Microhardness
The optimised result was relatively close to the previous optimisation results. The maximum microhardness obtained at 58.69 HV coincided with the PHT, PHti, and VF of 521 • C, 3 h, and 2.75 vol% CuO, respectively, as shown in Figure 13. The optimised solution was consistent with the experimental results of 61.32 HV. These experimental results supported the discussion earlier related to the influence of process parameters on the response [48].

RSM for Density
The DOE results for the density are shown in Table 9, while the effect of the reinforcement on the density is shown in the Pareto chart in Figure 14. The model was in accordance with the RSM response results versus PHT, PHti, and VF. Since the calculated p-value of the developed model was less than the standard (p < 0.05), the model was statistically according to the desired model [44,45,49]. The model was statistically correct.

RSM for Density
The DOE results for the density are shown in Table 9, while the effect of the reinforcement on the density is shown in the Pareto chart in Figure 14. The model was in accordance with the RSM response results versus PHT, PHti, and VF. Since the calculated p-value of the developed model was less than the standard (p < 0.05), the model was statistically according to the desired model [44,45,49]. The model was statistically correct.   The value of R 2 , adj. R 2 , and pred. R 2 for density are shown in Table 10. The R 2 value of 0.9759 was favourable, which was closer to 1, and provided an excellent explanation of the relationship between the independent variables and the response. The pred. R 2 of 0.9634 was within a reasonable agreement with the adj. The value of R 2 , adj. R 2 , and pred. R 2 for density are shown in Table 10. The R 2 value of 0. 9759 was favourable, which was closer to 1, and provided an excellent explanation of the relationship between the independent variables and the response. The pred. R 2 of 0.9634 was within a reasonable agreement with the adj. R 2 of 0.9458. The factors affecting the density of composite samples were evaluated from the RSM model. The PHT was observed to have little to no effect on the microhardness of the hybrid nanocomposite material. The lower density was observed at very low PHT and VF. In contrast, the high PHT and VF of SiO2 and CuO led to the strengthening of the composite, especially SiO2 that had a great effect on the density. A low VF and PHT resulted in low density [46,47].
The density was affected mainly by the VF of SiO2 and CuO. As shown in Figure 15, the contour plot for density indicates that the increase in VF of SiO and CuO increased the density at PHT and PHti of 500 °C and 2 h, respectively.  The factors affecting the density of composite samples were evaluated from the RSM model. The PHT was observed to have little to no effect on the microhardness of the hybrid nanocomposite material. The lower density was observed at very low PHT and VF. In contrast, the high PHT and VF of SiO 2 and CuO led to the strengthening of the composite, especially SiO 2 that had a great effect on the density. A low VF and PHT resulted in low density [46,47].
The density was affected mainly by the VF of SiO 2 and CuO. As shown in Figure 15, the contour plot for density indicates that the increase in VF of SiO and CuO increased the density at PHT and PHti of 500 • C and 2 h, respectively.
After eliminating the non-significant terms, the final regression model to predict the density of the developed reinforced nanocomposite was expressed using Equation (11) below: The equation indicates that the PHT had a significant effect on density.

Optimisation of Density
Based on the full factorial optimisation, the optimised results were in accordance with previous optimisation results. The maximum microhardness obtained coincided with PHT, PHti, and VF of 265 • C, 3 h, 1 vol% SiO 2 and 1.88 vol% CuO, respectively. As shown in Figure 16, the microhardness was 58.693 HV, which was consistent with the experimental results of 61.3266 HV. The experimental results supported the early discussion related to the influence of process parameters on the response [50,51].  After eliminating the non-significant terms, the final regression model to predict the density of the developed reinforced nanocomposite was expressed using Equation (11) below: The equation indicates that the PHT had a significant effect on density.

Optimisation of Density
Based on the full factorial optimisation, the optimised results were in accordance with previous optimisation results. The maximum microhardness obtained coincided with PHT, PHti, and VF of 265 °C, 3 h, 1 vol% SiO2 and 1.88 vol% CuO, respectively. As shown in Figure 16, the microhardness was 58.693 HV, which was consistent with the experimental results of 61.3266 HV. The experimental results supported the early discussion related to the influence of process parameters on the response [50,51].

Optimisation of Density
Based on the full factorial optimisation, the optimised results were in accordance with previous optimisation results. The maximum microhardness obtained coincided with PHT, PHti, and VF of 265 °C, 3 h, 1 vol% SiO2 and 1.88 vol% CuO, respectively. As shown in Figure 16, the microhardness was 58.693 HV, which was consistent with the experimental results of 61.3266 HV. The experimental results supported the early discussion related to the influence of process parameters on the response [50,51].

Multi-Objective Optimisation
The response service methodology (RSM) was employed to optimise the parameters and obtain the maximum value of UTS, microhardness, and density. The optimised results

Multi-Objective Optimisation
The response service methodology (RSM) was employed to optimise the parameters and obtain the maximum value of UTS, microhardness, and density. The optimised results were in accordance with the results of the previous optimisation. The maximum response obtained coincided with PHT, PHti, and VF of 541 • C, 2.25 h, and 1 vol% SiO 2 and 2.13 vol% CuO, respectively. As shown in Figure 17, the UTS, microhardness, and density were 228.82 MPa, 47.82 HV, and 2.63 g/cm 3 , respectively [52].

Confirmation Test
Three confirmation tests (CT) were performed to validate the empirical results. Based on the factorial analysis, the optimum parameters for PHT, PHti, and VF. Three composite specimens were produced based on these optimum parameters to validate the developed quadratic regression RF model, in addition to the RSM model which was developed and expressed in Equations (9)-(11) respectively. Based on the confirmation test presented on Table 11, then comparison the experimental values with the predicted values from the two developed models RF and RSM of the three specimens are given in Tables 12 and 13. The calculated errors were negligible in comparison with the obtained UTS, microhardness, and density. The calculated errors between the experimental and the optimum results were in the range of ±3%. The results successfully confirmed the reproducibility of the experimental data. The RF model outperforms the RSM model as shown in Table 13, where the RF model provides a flexible way to predict the response surface function. Moreover, RF constructs a sequence of DT, with each previous DT aiming to remove the errors of the current sequential DT model, and produces the final prediction using a weighted total of the predictions provided by the sequentially constructed DT. Also, RF provides the cross-validation technique which helps to avoid overfitting. were in accordance with the results of the previous optimisation. The maximum response obtained coincided with PHT, PHti, and VF of 541 °C, 2.25 h, and 1 vol% SiO2 and 2.13 vol% CuO, respectively. As shown in Figure 17, the UTS, microhardness, and density were 228.82 MPa, 47.82 HV, and 2.63 g/cm 3 , respectively [52].

Confirmation Test
Three confirmation tests (CT) were performed to validate the empirical results. Based on the factorial analysis, the optimum parameters for PHT, PHti, and VF. Three composite specimens were produced based on these optimum parameters to validate the developed quadratic regression RF model, in addition to the RSM model which was developed and expressed in Equations (9)-(11) respectively. Based on the confirmation test presented on Table 11, then comparison the experimental values with the predicted values from the two developed models RF and RSM of the three specimens are given in Tables 12 and 13. The calculated errors were negligible in comparison with the obtained UTS, microhardness, and density. The calculated errors between the experimental and the optimum results were in the range of ±3%. The results successfully confirmed the reproducibility of the experimental data. The RF model outperforms the RSM model as shown in Table 13, where the RF model provides a flexible way to predict the response surface function. Moreover, RF constructs a sequence of DT, with each previous DT aiming to remove the errors of the current sequential DT model, and produces the final prediction using a weighted total of the predictions provided by the sequentially constructed DT. Also, RF provides the cross-validation technique which helps to avoid overfitting.

Overall RF Results for Validation and Prediction
The microhardness, UTS, and density prediction analysis and framework were modelled with higher accuracy and precision using the RF technique to avoid complex experiments and mathematical calculations. A very comprehensive and detailed comparative analysis was carried out between the model and the experiment to verify the performance. The training performance of the models was dependent on the data set under the supervision of the learning algorithms. According to the No Free Lunch (NFL) theorem, the performance and working of the RF model will not always be highly accurate and precise for each data set [53]. The four parameters comprising PHT, PHti, VF of CuO, and VF of SiO 2 were used as the input parameters, while microhardness, UTS, and density were the output parameters to evaluate the training performance of the RF model.
Before dealing with the data, the data were normalised to fit a range of [0, 1]. Data normalising boosts the evaluation and prediction while in use. The microhardness, UTS, and density of the data samples were used in the training data, and k-fold cross-validation was set to 10. The confirmation tests were used as the test data for the verification of the accuracy of the RF model. For this model, a complete data set of 48 data samples were used. A total of 45 samples were obtained from the full factorial, while the output (microhardness, UTS, and density) was obtained from the experimental results. In addition, 45 data samples were chosen and selected at random for the training and samples validation, while the remaining 3 data samples were employed as the testing samples for output prediction assessment and the confirmation test to validate the results.
Low MSE and MAE values demonstrated better results in the RF model, as shown in Table 14. Overall, the data set, the confirmation test errors, and MSE and MAE values of the RF model were low, indicating that the model performed well in predicting the microhardness, UTS, and density. The highest R 2 value for the RF model was 0.9647. Moreover, the MSE and MAE values of the RF model were lower, indicating the highly accurate results with no significant over-estimation or under-estimation of the targeted values. This study concluded the significance and efficiency of the RF model in predicting the microhardness, UTS, and density of the nano-reinforced AA6061 aluminum chips based on the MAE and MSE values with low error rate between the experimental and predicted results.

Fractographic Analysis of Tensile Sample
The Scanning Electron Microscope (SEM) was used in analysing the morphology of fractured surface for tensile profiles of hot extruded composite samples. It was noticed that the fractured surface among tensile samples differed in contours. As shown in Figure 18a, samples extruded at PHT, PHti, and VF of 550 • C, 3 h, and 2 vol% CuO, respectively, exhibited the lowest UTS of 208 MPa, which had equiaxed dimples with microvoids causing poor material consolidation [17]. The fractographic analysis revealed improper metallurgical interface bonding at the time of extrusion. The uniformly void and equiaxed dimples cracks could easily be observed with added SiO 2 to the alloy in PHT to 550 • C, 3 h and SiO 2 2 vol% with few effects on UTS of 217 MPa as shown in Figure 18b the recycled aluminum samples could have mixed through various fracture mechanisms that formed the microvoids and cleavage planes, demonstrating the effect of reinforced VF on the fractured surface.
The effect of reduction in preheating temperatures from 550 °C to 450 °C was clear effected on the extruded sample microvoids and other defects and impurities served as the stress concentration sites and were responsible for the decrement in strength and other mechanical properties. The fracture surface exhibited ridges and large crack ridge instead of equiaxed dimples, as seen in Figure 18d. Such sample showed rough and large cracks, while the fracture occurred near the grip area of the sample due to the brittle failure with little plastic deformation before the fracture.
The Field Emission Scanning Electron Microscope (FESEM) was used in analysing the morphology of fractured surface for AA6061 chips extruded sample at PHT, PHti, and VF of 450 °C and 3 h the presence of cleavage fractures, microvoids, and unequal dimples could easily be seen in some regions, as shown in Figure 19a In addition, the uniformly distributed dimples and small cracks could easily be observed with the increase in PHT to 541 °C and VF of CuO 2.13 vol%, SiO2 1 vol%, as shown in Figure 19c,d. The non-uniform distribution of reinforcement particles led to the stress concentration site, which caused the crack initiation and subsequent failure. The high PHT ensures the uniform distribution of these reinforcement particles. The high PHT also improved the interfacial properties for a better and homogenous stress distribution. When the reinforced particles were well distributed in the aluminum matrix, the interfacial properties were significantly improved, where these improvements enhanced the mechanical properties, such as UTS and compressive strength, due to the proper distribution of stress in the composite samples. At optimal parameters, the sample exhibited the highest UTS of 239 Mpa. The positive effect of high PHT was proven in the fracture analysis. The presence of microvoids and fine dimples indicated the ductile nature of the fracture. Furthermore, clear shear and tears were observable on the fractured surface, while cleavage planes decreased significantly. The presence of a large number of small and fine dimples on the fractured surface was evidence of the influence of high PHT and low VF of reinforcements. These microvoids were formed due to the elongation of voids that filled up the available spaces to form dimples, as shown in the fractured surface. The fractography analysis further revealed that the maximum PHT and minimum VF resulted in adequate consolidation and high strength recovery of the composite material [18]. The fracture surface analysis of the optimal sample exhibited conical equiaxed dimples and the fine to- The fractographic analysis also revealed that the weakest point served as the crack initiation site for the failure. This was an indication of the sudden change in the force equilibrium in the samples, as seen in Figure 18c. The failure was also due to the improper bonding in the extruded aluminum chips, which directly influenced the mechanical strength (127 MPa) of the extruded aluminum samples at PHT, PHti, and VF of 450 • C, 1 h, 1 vol% CuO and 1 vol% SiO 2 .
The effect of reduction in preheating temperatures from 550 • C to 450 • C was clear effected on the extruded sample microvoids and other defects and impurities served as the stress concentration sites and were responsible for the decrement in strength and other mechanical properties. The fracture surface exhibited ridges and large crack ridge instead of equiaxed dimples, as seen in Figure 18d. Such sample showed rough and large cracks, while the fracture occurred near the grip area of the sample due to the brittle failure with little plastic deformation before the fracture.
The Field Emission Scanning Electron Microscope (FESEM) was used in analysing the morphology of fractured surface for AA6061 chips extruded sample at PHT, PHti, and VF of 450 • C and 3 h the presence of cleavage fractures, microvoids, and unequal dimples could easily be seen in some regions, as shown in Figure 19a,

X-ray Diffraction (XRD) Analysis
Figure 20a, shown the XRD of the angle measured at 2θ in the 10-90 range. The XRD patterns of the aluminum matrix with different reinforcement concentrations. The figure shown that the laser peened sample exhibited peak broadening and a considerable shift in XRD peaks to larger Bragg's angle that indicated the presence of microstrain and decrement of the lattice constant, which was the (hkl) spacing after the laser peening. Apart from that, the (222) and (200) planes were much less pronounced in the base material, whereas the laser peened surface showed more intense and higher diffractions from those planes [54].
The crystallinity analysis was conducted on the three most important samples utilising their highest, optimum, and lowest parameters, as presented in Figure 20b, and Table  15. In addition, the uniformly distributed dimples and small cracks could easily be observed with the increase in PHT to 541 • C and VF of CuO 2.13 vol%, SiO 2 1 vol%, as shown in Figure 19c,d. The non-uniform distribution of reinforcement particles led to the stress concentration site, which caused the crack initiation and subsequent failure. The high PHT ensures the uniform distribution of these reinforcement particles. The high PHT also improved the interfacial properties for a better and homogenous stress distribution. When the reinforced particles were well distributed in the aluminum matrix, the interfacial properties were significantly improved, where these improvements enhanced the mechanical properties, such as UTS and compressive strength, due to the proper distribution of stress in the composite samples. At optimal parameters, the sample exhibited the highest UTS of 239 Mpa. The positive effect of high PHT was proven in the fracture analysis. The presence of microvoids and fine dimples indicated the ductile nature of the fracture. Furthermore, clear shear and tears were observable on the fractured surface, while cleavage planes decreased significantly. The presence of a large number of small and fine dimples on the fractured surface was evidence of the influence of high PHT and low VF of reinforcements. These microvoids were formed due to the elongation of voids that filled up the available spaces to form dimples, as shown in the fractured surface. The fractography analysis further revealed that the maximum PHT and minimum VF resulted in adequate consolidation and high strength recovery of the composite material [18]. The fracture surface analysis of the optimal sample exhibited conical equiaxed dimples and the fine topography of the fracture surface in the absence of microcracks. The failure was observed along the middle part of the sample due to the ductile failure mode, which showed plastic deformation just before the final fracture.
3.9. X-ray Diffraction (XRD) Analysis Figure 20a, shown the XRD of the angle measured at 2θ in the 10-90 range. The XRD patterns of the aluminum matrix with different reinforcement concentrations. The figure shown that the laser peened sample exhibited peak broadening and a considerable shift in XRD peaks to larger Bragg's angle that indicated the presence of microstrain and decrement of the lattice constant, which was the (hkl) spacing after the laser peening. Apart from that, the (222) and (200) planes were much less pronounced in the base material, whereas the laser peened surface showed more intense and higher diffractions from those planes [54].  The crystalline structure was identified by matching the data and peaks with the standard JCPDS data. Based on the results, the extruded sample using the optimum parameters (541 °C, 2.25 h, 2.13 vol% CuO and 1 vol% SiO2, respectively) recorded the lowest intensity of almost 37,892 a.u. compared to other samples. In contrast, for the extruded sample using the optimum parameters (450 °C, 1 h, 1 vol% CuO and 3 vol% SiO2, respectively) and (550 °C, 3 h, 1 vol% CuO and 1 vol% SiO2, respectively), the intensity at the plane (111) was 130,112 and 182,419 a.u., respectively.

Conclusions
A thorough investigation on the effects of PHT, PHti, and VF of CuO and SiO2 on UTS, microhardness, and density of the extruded nano-reinforced aluminum composite was reported in this study. The recycled hybrid aluminum alloy chips reinforced with nano-CuO and nano-SiO2 were fabricated using the direct recycling method. Based on the RSM results, the PHT, PHti, and VF of hybrid reinforcements demonstrated significant influence to obtain an optimum UTS of the nano-reinforced aluminum composite. In ad- The crystallinity analysis was conducted on the three most important samples utilising their highest, optimum, and lowest parameters, as presented in Figure 20b, and Table 15. The crystalline structure was identified by matching the data and peaks with the standard JCPDS data. Based on the results, the extruded sample using the optimum parameters (541 • C, 2.25 h, 2.13 vol% CuO and 1 vol% SiO 2 , respectively) recorded the lowest intensity of almost 37,892 a.u. compared to other samples. In contrast, for the extruded sample using the optimum parameters (450 • C, 1 h, 1 vol% CuO and 3 vol% SiO 2 , respectively) and (550 • C, 3 h, 1 vol% CuO and 1 vol% SiO 2 , respectively), the intensity at the plane (111) was 130,112 and 182,419 a.u., respectively.

Conclusions
A thorough investigation on the effects of PHT, PHti, and VF of CuO and SiO 2 on UTS, microhardness, and density of the extruded nano-reinforced aluminum composite was reported in this study. The recycled hybrid aluminum alloy chips reinforced with nano-CuO and nano-SiO 2 were fabricated using the direct recycling method. Based on the RSM results, the PHT, PHti, and VF of hybrid reinforcements demonstrated significant influence to obtain an optimum UTS of the nano-reinforced aluminum composite. In addition, the high PHT and high VF of CuO and SiO 2 improved and strengthen the bonding of aluminum chips through the inter-particle diffusion transport of matter and produced the highest microhardness. Furthermore, the density was affected by the hybrid reinforcements and preheating temperate with less effect for preheating time. Using the CuO as the reinforcement particle, the overall energy reduced according to the change in PHT and PHti of 550 • C and 3 h, respectively, to PHT and PHti of 541 • C and 2.25 h, respectively, to achieve the optimal properties. Based on the fractured surface analysis, the hot extrusion parameters affected the tensile strength of the specimen, while the fine topography and dimples were apparent under SEM and FESEM imaging with the increase in PHT and VF of CuO and decrease in VF of SiO 2 , based on the experimental results the RF model achieve better results than response surface methodology (RSM) and more accurate Such models could assist in predicting the outcome of the processing materials, which will be helpful to minimise and limit the number of experiments, therefore, minimising the time and energy consumption due to the high prediction rate. Moreover, this study demonstrated that the use of machine learning was significant to obtain high efficiency in research studies and manufacturing processes.