Optimization studies of stir casting parameters and mechanical properties of TiO2 reinforced Al 7075 composite using response surface methodology

Stir casting is a common metallurgical route in the casting of aluminum composites. Series of work done in this aspect considered the development of the composites with fixed stir casting parameters without applying an optimization approach. These parameters affect the microstructure and performance of the composites. The study is focused on the optimization of the stir casting parameters in the production of Al 7075 reinforced with TiO2 microparticles for performance improvement. Three stir casting parameters of stirring temperature, speed, and time were varied and optimized using the central composite design technique of the response surface method. Properties evaluated were ultimate tensile strength, hardness, impact strength, elastic modulus, and compressive strength. ANOVA results showed that the three stir casting parameters had a significant impact on the property responses. Five quadratic models were established for the properties linking them to the factors. The models were confirmed to be statistically significant at a confidence level of 95% and variations were observed to be < 5%. The interaction profile of the parameters as per response surface was analyzed. Contour plots associated with each interaction gave different ranges of stirring parameters in which each property can be maximized. Simultaneous optimization of the properties using Minitab 19 software showcased 779.3 °C, 574.2 rpm, and 22.5 min as the optimal stir casting parameters for temperature, speed and time respectively.

, this led to enhancement of hardness and bending strength. Their study revealed that the microstructure showed a particle dispersion within the matrix, thereby enhancing the properties. It was reported that titanium boride dispersion in Al 7075 matrix resulted in the enhancement of tensile strength and hardness at particulate dosage of 4, 8, and 12 wt% 9 . Infusion of TiC in Al 7075 led to the improvement of yield and tensile strength even as optimum enhancement was attained at 9 wt% TiC inclusion. Similar trend was achieved when Kumar et al 7 added Titanium carbide at 3, 4, 5, 6, and 7 wt% of the Al matrix. Microhardness, yield, and ultimate tensile strength were enhanced as the particulates increased in the stir cast product. Optimum enhancement of strength was achieved at 7 wt% content. The choice of TiO 2 in reinforcing and enhancing properties of Al 7075 was taken by Murali et al 10 , the compressive and tensile strength were enhanced with titanium dioxide addition even as the optimum enhancement was attained at 15 wt%. Comparative study showed that titanium dioxide proportions of 1, 3, 5, 7, and 9 wt% led to a progressive and linear increase in yield and ultimate tensile strength while microstructural examination showed even dispersion of particles within the matrix 11 . Al 6061 was reinforced with 50 µm average sized titanium dioxide, as investigated by Kumar etal 12 . Evaluation of the properties showed that relative to the base metal, the hardness was improved by 20.7, 52.6, and 66.7% and the ultimate tensile strength was enhanced by 31.6, 55.8, and 89.5% as the particulate increased at 1, 2, and 3 wt%. It is concluded that the titanium dioxide particles reinforced composite exhibited improved properties compared to the base metal. Authors Alagarsamy and Ravichandran 13 employed the stir casting route in the development of TiO 2 -AA 7075 composites with varying proportions of 5, 10, and 15 wt% titanium dioxide particles. The mechanical properties; tensile impact, flexural and compressive strengths, and hardness were enhanced as the particulate increased up to 10 wt%. Authors Kumar et al 14 and Kumaret al 15 have performed the investigation of the influence of the stir casting parameters on the distribution of the reinforcement particles using the visualization technique. From their investigation, it was concluded that 40% impeller position from the base 45º blade angle shows the considerable improvement in the mechanical properties. Computational Fluid Dynamics simulations were employed to evaluate the effect of vortex pressure, which was created during the stirring action, on the processing of the MMC. The results concluded that the mechanical properties were improved by the optimal vortex height using Taguchi design of the experiment 16 . Kumar et al 17 also performed the microstructural evaluation to identify the influence of stir casting parameters on the mechanical properties of the composites. The microstructural results show a uniform dispersion of reinforcement particles which consecutively increases the mechanical properties of the composites. Even though various techniques were involved in the evaluation of the stir casting process parameters, no one has performed the RSM optimization technique to determine the best stir casting parameters for the processing of Metal Matrix Composites. The afore-discussed literature centered on the effect of composition on the properties of Al 7075 without considering the effect of varying stir casting process parameters like stirring temperature, time, and speed on properties of the alloy. Other process parameters are blade angle, stirrer height, feed rate of reinforcement, and direction of impeller rotation. These is important to ascertain how these parameters affect the microstructure and optimize for the most appropriate conditions for improvement of the alloy. This research reveals the effect of stirring parameters; temperature, time, and speed on the ultimate tensile strength, microhardness, impact strength, elastic modulus, and compressive strength of Al 7075/10 wt% microparticles of titanium dioxide (average size 13 µm) according to the study of Alagarsamy and Ravichandran 13 and El-Mahallawi et al 18 in which 10 wt% yielded the optimum performance.

Methodology
Material preparation. Aluminum alloy ingot AA 7075-T6 was employed in this study. The chemical composition as obtained via spectrometer test is highlighted in Table 1 while the properties are stated in Table 2.
Stir casting process was done in a graphite crucible (diameter 100 mm and height 175 mm). Titanium dioxide (TiO 2 ) microparticle of size 13 µm was preheated to 500 °C for 10 min and was introduced into the melt at 10 wt%. The range of stirring speed, temperature, and time considered are presented in Table 3. The specimens were prepared in accordance with the experimental runs obtained via the central composite design as stated in Table 4. To enhance the wettability of matrix, 1 wt% magnesium was added into the melt as previously reported by Kumar et al 19 .
Test procedure. Ultimate tensile strength is the optimum stress in which the given material can tolerate under applied tension without breaking. Machined specimens whose dimensions are 30 mm in length and 5 mm in diameter were tested for tensile strength applying a universal testing machine (Instron 3369 Series). In line  20 , a load of 10 kN was applied at a rate of 10 −4 /s and a cross head speed of 3.0 mm/ min' . As prescribed by ASTM E09-9 21 , the compressive strength of the samples was examined using the universal testing machine applying a load of 100 kN at a cross speed of 1 mm/min. Vickers microhardness test was done on the specimens in accordance with ASTM E 384-17 22 on the surfaces of the samples applying a load of 10 N for 10 s on each sample. Impact toughness was also probed subjecting a specimen 10 × 10 mm 2 initially notched at 45° to high strain impact with the use of pendulum of 300 N in weight while measuring the absorbed energy to failure (ASTM E-23) 23 . In accordance with ASTM E 384-17 24 , field emission scanning electron microscope (JSM-7610 E) was used in accessing the microstructure of the developed samples. From Table 1, the compositional elements of Al 7075 are displayed, Zn and Mg were observed to be present in considerable amounts while Ti and Cr occurred in trace proportions. Table 2 highlights the properties of the base material with the ultimate tensile strength of 521 MPa and relative density of 2.699.

Design of experiment.
Experimental process involves the design of an experiment via the response surface method (RSM) in which the process parameters are optimized. The variables are stirring temperature (A), stirring speed (B), and stirring time (C). RSM utilizes mathematical and statistical means in analyzing the relationships between process parameters and the response parameters. This process has been employed in several literatures for optimization and the outcome showed that experimental runs in the laboratory can be minimized via this process [25][26][27][28][29] . With the aid of Minitab 19 software, a central composite design was employed involving five-level-three-factor. Twenty (20) experimental runs were undergone for each property evaluated, entailing 6 axial runs, 8 factorial runs, and 6 replicates at the center point as carried out in previous studies 22,30 . The secondorder polynomial expression in Eq. (1) was employed in accessing the relationship between the process variables and predicted responses. (1) DXiXj + E  www.nature.com/scientificreports/ Z is the predicted response, A is the intercept, B is a linear coefficient for first-order expression, C is a quadratic coefficient for the second-order expression, D is the coefficient of the interaction effect, and E is the random error. First order polynomial model is expressed in Eq. (2).
where X 1 , X 2 … X n are the process variables, A is the constant, and Cn is linear of the nth factor constant, while E is the error.
Accuracy of the models were verified by the predicted values of the coefficient of correlation for predicted and adjustable data, relative deviation, root mean square, and mean square error. In the design, temperature was selected between 600 and 800 °C, time between 10 and 20 min, and speed between 400 and 600 rpm as observed in Tables 3 and 4 in accordance with Aynalem 31 .

Results and discussion
Analysis of variance and regression models. As highlighted in Table 5, the p values for the process variables stirring temperature (A), stirring speed (B), and stirring time (C) are less than 0.05, which reflect the significance of these variables as they determine magnitude of the response. Squared interactions A*A and C*C are also statistically significant, whereas B*B is insignificant. Two-way interaction A*B, A*C, and B*C are significant since p > 0.05. Meanwhile, the contribution of parameters for linear variables A, B, and C are 36.5, 9.03, and 27.89%, implication of which shows the order of significance of the variables is in descending order of stirring temperature, stirring speed and stirring time.
Second order polynomial model obtained for the ultimate tensile strength which incorporates the input variable is expressed in Eq. (3).
UTS is the ultimate tensile strength, A is the stirring temperature, B is the stirring speed, and C is the stirring time.
From Table 6, the p values for the linear terms A, B, and C are less than 0.05, hence, are significant with contributions of 39.58, 20.81, and 29.87% respectively. This implies that the input variables have a significant effect on the hardness response of the properties as the input parameter varies. The square terms A*A, B*B, and C*C are insignificant likewise cross interactions A*B, A*C, and B*C. From the contributions of the linear terms, the stirring temperature has the highest contribution to the response while the stirring time is the next. Stirring speed showed the least contribution amongst the three parameters. Second order polynomial model obtained for the hardness is expressed in Eq. (4).
Hd is hardness, A is the stirring temperature, B the is stirring speed, and C is the stirring time. ANOVA analysis carried out on impact strength revealed the response is significantly dependent on the input variables (Table 7). Linear terms A, B, and C are significant, squared interactions A*A and C*C are significant while B*B is insignificant. Cross interactions A*B, A*C, and C*C are insignificant on the response. Contributions of A, B, and C are 29.03, 14.87, and 26.50%, therefore, the stirring temperature has the highest contribution. Next is the stirring time while stirring speed contributes the least. Second order polynomial model obtained for hardness is expressed in Eq. (5). www.nature.com/scientificreports/ IM is the impact strength, A is the stirring temperature, B is the stirring speed, and C is the stirring time. Table 8 highlights the ANOVA results on the response of elastic modulus to process variables at 95% confidence level and 5% significance level. As reflected, the probability value (p-value) for stirring temperature (A), stirring speed (B), and stirring time (C) as process variables under the linear model is < 0.05 depicting the fact EM is the elastic modulus, A is the stirring temperature, B is the stirring speed, and C is the stirring time. From Table 9, the p values for the linear terms A, B, and C are less than 0.05, hence, they are significant with contributions of 37.80, 7.23, and 29.62% respectively, the relevance of which showed that the input variables have a significant effect on the compressive strength response of the composite. The square terms A*A, B*B, and C*C are significant while the cross interactions A*B, A*C, and B*C are insignificant. From the contributions of the linear terms, the stirring temperature has the highest contribution to the response while the stirring time is the next. Stirring speed showed the least contribution amongst the three parameters. Equation (7) represents the model for compressive strength.
CS is the compressive strength, A is the stirring temperature, B is the stirring speed and C is the stirring time. Table 10 reveals coefficients of correlation R 2 , R 2 (adj), R 2 (pred), and ð R 2 (difference between R 2 (adj) and R 2 (pred)). As for ultimate tensile strength (UTS), R 2 = 97.75% depicting a strong relationship between the model and the dependent variable that is 97.75% of the observed variation can be explained by the model. R 2 (adj) is 96.57% while R 2 (pred) which indicates the level of prediction of future data model is 93.51%. It was reported that the difference (ð R 2 ) between R 2 (adj) and R 2 (pred) should be < 20% for a reliable model and since the difference is < 20% for UTS there is a good correlation, thus the requirement is satisfied [32][33][34] . Likewise, the hardness, impact strength, elastic modulus, and compressive strength have a value of R 2 to be 94.76, 95.9, 96.36, and 97.39 showing that the model has over 90% representation of the relationships. ð R 2 (%) for hardness, impact strength, elastic modulus and compressive strength are all less than 20% hence satisfying the requirement and said to have good correlation. www.nature.com/scientificreports/ Normal probability plots and residual versus fit. Figure 1a-e illustrates the normal probability plot of residuals for ultimate tensile strength, hardness, impact strength, elastic modulus, and compressive strength. It is observed that most of the plotted points form a nearly linear pattern with few departures from the straight line showing consistent data, further confirming the validity of the model. Scatter plot of the residual against fitted data is demonstrated in Fig. 2a-e in which the data are plotted randomly round the dashed line is heteroskedastic as the points are concentrated towards the line [33][34][35] . The residual for the responses is uniformly scattered around the mean response, reflecting the adequacy of the model.

Coefficient of correlation for mechanical properties.
Analysis of the response surface and contour plot. Response surface and contour plot for ultimate tensile strength. Effect of interaction of stirring temperature (Tm) versus stirring speed (Sp) on the ultimate tensile strength of composite. Property responses as a function of interactions between experimental variables are represented by a 3D response surface plots and 2D contour mapping. The graphical illustration of the model is plotted as a function of two process parameters holding the other variable constant. As for the response surface, the response is plotted on Z axis while the input variables are plotted on X and Y axis. The effect of interaction between speed and temperature at a constant time of 15 min is highlighted in Fig. 3a for the response surface. As the stirring speed, temperature, and speed increased, ultimate tensile strength response increased. However, at the speed of 500 rpm, and a temperature of 800 °C, the ultimate tensile strength response value reduced. There is therefore a strong dependence of the ultimate tensile strength on the interactive pattern between speed and temperature. The two parameters displaced a parabolic reflection profile with points of inflection at 800 °C for temperature and 500 rpm for speed, yielding a maximum value of 644.8 MPa at the point of inflexion. www.nature.com/scientificreports/ that the microhardness of the composite depends on the interaction between the two parameters. Maximum microhardness of 119.2 HV was attained when the process was carried out at 580 rpm for 20 min. The plot of Fig. 4d shows that the optimum hardness can be attained at the portion "A" value of which is between 118 and 120 HV. Corresponding range of speed is 560-650 rpm while that of time is 19.4-22.5 min.
Effect of interaction of stirring time (T) versus stirring temperature (Tm) on Vickers microhardness of composite. From Fig. 4e, the effect of the interaction of stirring time and the temperature on Vickers microhardness of the developed aluminum composite reinforced with TiO 2 ceramic microparticles at a constant speed of 500 rpm. With the increase in temperature and time, there was a consecutive increase in hardness. This shows that the hardness of the composite depends on the interaction between the two parameters. Temperature reflected a concave profile while time depicted an upward linear profile. Maximum hardness of 119.2 was attained at the interplay of 600 rpm and 835 °C. The figure further shows that the response of microhardness depends on the interaction. The plot of Fig. 4f shows that the optimum hardness can be attained at the portion "A" value of which is 118-120 HV. Corresponding range of speed is 785-850 °C while that of time is 16.5-22.5 min.
Response surface and contour mapping for impact strength. Effect of interaction of stirring temperature (Tm) versus stirring speed (Sp) on the impact strength of composite. The effect of interaction between speed and temperature at a constant time of 15 min is presented in Fig. 5a for the response surface. As the stirring speed, temperature, and speed increased, the impact strength response increased, nonetheless, at the stirring speed of 500 rpm, and a temperature of 800 °C, the impact strength response value decreased. There is therefore a strong dependence of impact strength on the interactive pattern between speed and temperature. The two experimental parameters displaced a parabolic profile with the ends facing down. The points of inflection at 800 °C for temperature and 500 rpm for speed yields a maximum value of 5.27 J/m 2 . Figure  Response surface and contour mapping for elastic modulus. Effect of interaction of stirring temperature (Tm) versus stirring speed (Sp) on elastic modulus of composite. The influence of interaction between speed and temperature at a constant time of 15 min is highlighted in Fig. 6a for the response surface. As the stirring speed, temperature, and speed increased, the elastic modulus increased, although an interplay between the speed of 500 rpm and a temperature of 750 °C engendered decrease in the modulus. There is a strong dependence of the ultimate tensile strength on the interactive pattern between speed and temperature. The two parameters displaced a parabolic interactive profile with points of inflection at 750 °C for temperature and 500 rpm for speed with a maximum value of 95.5 GPa at the point of inflexion. Figure 6b shows the contour plot of the effect of the interaction temperature versus speed at constant 15 min on the elastic modulus. The plot reflects different segments where varying elastic moduli are attained at varying temperature and speed interaction. Segment "A" is the optimal zone in which the interaction of temperature and speed yield an optimum strength range of 94-96 GPa attainable at a temperature range of 720-824 °C and speed of 475-580 rpm. www.nature.com/scientificreports/  Fig. 6c. As the stirring time and speed simultaneously increased, there was improvement in the modulus. Notwithstanding, there was a significant negative interaction between speed and temperature at speed beyond 500 rpm, the effect of which led to decrease in the modulus. Stirring speed showed a parabolic profile with the two-end pointing downwards while time demonstrated a linear uptrend profile. The contour plot of the two parameters is presented in Fig. 6d where   Table 11. It is evident that the deviation for each property is less than 5%, therefore, it is concluded that there is a good agreement between the experimental values and the predicted value, thus, validating the model.    Figure 8b reflects the dispersion of particles and coagulation, which is attributed to the samples prepared at low temperature of 350 °C, speed of 331.821 rpm and time 10 min. At lower temperature, the viscosity is higher which has the tendency for premature solidification leading to coagulation 36 . Figure 8c and d features intermetallic phases which occurred at high temperature of 868.179 °C leading to the formation of intermetallic phases culminating in the lowering of strength. Meanwhile, the moderate temperature allows the mobility of atoms allowing the TiO 2 to be dispersed within the matrix. Figure 8 e and f are images of samples produced at temperature and speed of 800 °C and 600 rpm, respectively. At high temperature and speed, there is turbulence during stirring leading to gas entrapment causing voids and blow holes as observed in the figures 37 . Figure 9a presents the microstructural image of the pure aluminum base alloy in which inherent pores are detected. Figure 9b showed the morphology of the developed composites at optimum stirring parameters of 779.3 °C, 574.2 rpm, 22.5 min for temperature, speed, and time, respectively. The image showed uniform dispersion of the particulate within the matrix giving rise to enhanced properties as compared to the pure Al 7075 alloy. Figure 10a and b indicates the phases present in the pure Al 7075 alloy and the developed composite at optimal stirring parameters. In Fig. 10a, the XRD patterns of the pure alloy showed the presence of crystalline aluminum and traces of other metal components. The patterns in Fig. 10b identified other phases alongside the crystalline aluminum. Titanium dioxide is present confirming the presence of particulate which suppresses the existence www.nature.com/scientificreports/ of other trace elements identified in the base metal. Traces of intermetallic phases are also identified, TiAl 3 and Al 5 T 12 which occurred as a result of high temperature reaction. Since in trace, they are suppressed, therefore not playing a major role on the properties of the composite.

Conclusion
Influence of individual and combined interaction of three processing factors of stir casting processing route in the development of Al 7075/TiO 2 composite was evaluated by means of the central composite design using the response surface method. Outcome of the findings showed that the response of ultimate tensile strength, hardness, impact strength, elastic modulus, and compressive strength depends on the interactions between these parameters. The ANOVA results showed that each property is influenced by the stirring parameters while selected square interactions had considerable effect at confidence level of 95%. Five predictive models were developed for the properties were observed to be statistically significant and from the coefficient of correlation, it was deduced that over 94% of the data for each response was well represented by the model. Response surface method revealed that each evaluated has a strong dependence on the interactions between the process variables. Parameter values not greater than 750 °C and 500 rpm for temperature and speed, respectively, were observed to have a positive influence on the responses, while values beyond that could have negative contributions. In the same manner, a stirring time of up to 30 min may provide enough time for even dispersion engendering a positive contribution to the responses. Simultaneous optimization of the properties showed that the optimum of mechanical properties is achievable at experimental conditions of 779.3 °C, 574.2 rpm, and 22.5 min.

Data availability
All data generated or analyzed during this study are included in this published article.