Optimising Tropical Fruit Juice Quality Using Thermosonication-Assisted Extraction via Blocked Face-Centered Composite Design

: Extraction of tropical fruit juice using simple, efﬁcient, and environmentally friendly technologies is gaining importance to produce high quality juices. Juice from pink-ﬂeshed guava, pink-ﬂeshed pomelo, and soursop was extracted using direct and indirect thermosonication methods by varying intensity, time, and temperature, and compared to those extracted using water bath incubation. Improvised models of juice yield, ascorbic acid, and total soluble solids responses were generated by eliminating insigniﬁcant model terms of the factors in full quadratic model using backward eliminating procedure. Main effects, 3D, or 4D plots for each response were developed based on factors that inﬂuenced the response. Results showed that the best extraction method for guava and pomelo juices were within indirect thermosonication method of 1 kW, 55 ◦ C and 30 min, and 2.5 kW, 54 ◦ C and 23 min, respectively. Direct thermosonication method at 10% amplitude, 55 ◦ C for 2 to 10 min was more suitable for soursop juice. Thermosonicated extraction of tropical fruit juice can improve its juice yield, ascorbic acid content, and total soluble solids content.


Introduction
Preserving pink-fleshed guava, pink-fleshed pomelo, and soursop fruits in the form of juice concentrates is useful, as these three fruits are excellent in terms of fighting cancer and are claimed as cancer therapy fruits [1,2]. The pink-fleshed guava contains lycopene, which can prevent skin damage from ultraviolet rays and offers protection from prostate cancer. It is also rich in carotene, which can protect against lung and oral cavity cancers. The high vitamin C content of the pink-fleshed pomelo helps to strengthen and maintain elasticity of arteries. In addition to being good for the digestive system, pomelo can aid in weight loss process because the fat burning enzyme in pomelo can help to absorb and reduce starch and sugar in the body. The pink-fleshed pomelo is slightly sweeter and more nutritious than other pomelos because of its darker pigment. Soursop pulp is white and juicy and has a delightful sour-sweet aroma with a yogurt-like taste [3], in addition to containing annonaceous acetogenins, which are prostate cancer chemopreventive compounds [1]. The fruit is widely used in anticancer folk therapies in North, Central, and South America, and Southeast Asia [4].
In the ascending order of the amount of free-run juice from fresh fruit, the pink-fleshed guava ranks the lowest, followed by the soursop, and pink-fleshed pomelo. The pinkfleshed guava and soursop both contain high starch levels of 13% [5] and 27.3% [6], respectively. The high starch content and the thick, creamy, and fleshy, pink-fleshed guava and soursop pulps result in difficulty in cell wall disruption of the fruit tissue. Thus, only a

Fruit Pulp Preparation
Matured, pink-fleshed guava fruits at 75% ripeness with a pronounced typical aroma and yellow skin were purchased from Sime Darby Beverages Sdn. Bhd. The ripened fruits were washed under running tap water and any floral remnants at the apex were removed after the outer part dried. Then, the tip ends were cut with a sharp knife prior to dicing the fruit into small pieces.
Pink-fleshed pomelo fruits were purchased from Perniagaan Buah-buahan Ah Yew, Bidor, Perak. Pomelo with yellow rinds were chosen for peeling process. The very thick spongy rind was cut into four sections as close as possible to the flesh of the pomelo. Each section of peel was pulled away from the fruit prior to slicing the fruit and breaking it in half. The seeds and remaining pith, which are bitter, were completely removed.
Matured green soursop fruits were purchased from Federal Agricultural Marketing Authority (FAMA) of Malaysia. The fruits were allowed to ripen in an air-conditioned (18 ± 2 • C) room for 3 to 4 days prior to processing. They were considered ripened when they were soft to touch and the shiny green colour turned to lack-lustre green or yellowishgreen. Approximately 5 kg of ripened fruits were processed while waiting for remaining fruits to ripen within 1 to 2 days. The fruits were washed and cut into half, cored, their skin hand-peeled, and seeds removed from the pulp manually. For each fruit, 2.5 kg of pulp was crushed into a mash using a 1300 W commercial food blender (XB409, Ceado, Italy) with a pulse duration of 60 s at a high speed of 28,000 rpm, then 30 s off and 60 s at a low speed of 22,000 rpm for complete homogenisation of pulp mash.

Thermosonic-Assisted Extraction
The direct thermosonic-assisted extraction as shown in Figure 1 was performed using a 400 W digital ultrasonic processor (S-450D, Branson, MO, USA) with its probe tip immersed to a depth of 25 mm into a 150-mL beaker containing pulp mixture samples at ultrasonic amplitudes of 10, 55, or 100% for 2, 6, or 10 min. For indirect thermosonic-assisted extraction as illustrated in Figure 2, a glass bottle of the pulp sample mixture was partially immersed in an ultrasonic water bath [26], which contains distilled water as a medium to spread waves at a power of 1, 1.75, or 2.5 kW for 10, 20, or 30 min. The motion frequency, power, and amplitudes of ultrasonic energy dictate the intensity of treatment where high shear forces induced produce high-energy microbubbles that accelerated juice extraction process through release of energy by direct injection into fruit cell wall to obtain a high volume of extracted juice. Each extraction was conducted at three temperatures, i.e., 25, 40, and 55 • C, and the pulp mixture samples were mixtures of 50 g of blended pulp with distilled water at ratio of 1:1. The specified temperature was maintained manually by adding cold water into a water container that places the beaker for direct thermosonication or applying continuous flow of water in the bath for indirect thermosonication. The extraction by incubation in a water bath acted as a control with pulp mixture samples added in glass bottles, shaken for 30, 75, and 120 min at 20, 100, and 180 rpm in a shaking water bath (BS-21, Lab Companion, Korea) at incubation temperatures of 25, 40, and 55 • C.
The treated pulp was separated from the juice by centrifugation at 4000 rpm and 4 • C for 20 min using a refrigerated centrifuge (Mikro 22R, HettichZentrifugen, Germany). The supernatant was collected for determination of juice yield, ascorbic acid (AA) content, and total soluble solids (TSS) content.
For each fruit, 2.5 kg of pulp was crushed into a mash using a 1300 W commercial food blender (XB409, Ceado, Italy) with a pulse duration of 60 s at a high speed of 28,000 rpm, then 30 s off and 60 s at a low speed of 22,000 rpm for complete homogenisation of pulp mash.

Thermosonic-Assisted Extraction
The direct thermosonic-assisted extraction as shown in Figure 1 was performed using a 400 W digital ultrasonic processor (S-450D, Branson, MO, USA) with its probe tip immersed to a depth of 25 mm into a 150-mL beaker containing pulp mixture samples at ultrasonic amplitudes of 10, 55, or 100% for 2, 6, or 10 min. For indirect thermosonic-assisted extraction as illustrated in Figure 2, a glass bottle of the pulp sample mixture was partially immersed in an ultrasonic water bath [26], which contains distilled water as a medium to spread waves at a power of 1, 1.75, or 2.5 kW for 10, 20, or 30 min. The motion frequency, power, and amplitudes of ultrasonic energy dictate the intensity of treatment where high shear forces induced produce high-energy microbubbles that accelerated juice extraction process through release of energy by direct injection into fruit cell wall to obtain a high volume of extracted juice. Each extraction was conducted at three temperatures, i.e., 25, 40, and 55 °C, and the pulp mixture samples were mixtures of 50 g of blended pulp with distilled water at ratio of 1:1. The specified temperature was maintained manually by adding cold water into a water container that places the beaker for direct thermosonication or applying continuous flow of water in the bath for indirect thermosonication. The extraction by incubation in a water bath acted as a control with pulp mixture samples added in glass bottles, shaken for 30, 75, and 120 min at 20, 100, and 180 rpm in a shaking water bath (BS-21, Lab Companion, Korea) at incubation temperatures of 25, 40, and 55 °C.
The treated pulp was separated from the juice by centrifugation at 4000 rpm and 4 °C for 20 min using a refrigerated centrifuge (Mikro 22R, HettichZentrifugen, Germany). The supernatant was collected for determination of juice yield, ascorbic acid (AA) content, and total soluble solids (TSS) content.

Blocked Face-Centred Central Composite Design
The blocked face-centred central composite design provides high-quality predictions over the entire design space and does not require using points outside the original factor range. Blocking was required to increase precision of experiments as variations of each batch of fruits could not be controlled, and all runs could not be completed within a day. Face-centred refers to alpha value of one and was chosen because the points were at operational limits, and equipment used could set the parameters to certain values only. The combination of factors for each run was generated and arranged using commercial software (MINITAB ® Release 14, Minitab Inc., State College, PA, USA). Table 1 shows experimental runs designed for one batch of fruits. There were 20 runs for each batch of fruits with three replications for each block and batch varying three factors, intensity, time, and temperature, at three levels, −1, 0, and 1, which gave total runs of 60. Responses were measured as percentage change of juice yield, ascorbic acid (AA), and total soluble solids (TSS) content. Fitted regression models were obtained using MINITAB, whereas 4D surface plot was generated using ThreeDify Excel Grapher (v.3.3.8, ThreeDify Inc., Ottawa, ON, Canada) software. The plotted graph chosen depended on the significant factors involved in the improvised models.
The second-order model is widely used in RSM because of its flexibility in taking on a wide variety of functional forms and a good capability to estimate the true response surface [27]. As blocking was used in this study, the second-order model, Equations (1) and (2), were necessary to fit the data [28,29]. The best-fitted models were chosen, and values for the combined factors that satisfied all optimum responses simultaneously were generated.

  
where Y is the response; β0 is the constant for intercept; β01 to β09 are the constant for blocks; βj is the linear coefficient; βjj is the quadratic coefficient, and βij is the interaction coefficient. xi and xj are independent variables (i and j are in the range of 1 to k). k is the number of independent variables (k = 3).

Blocked Face-Centred Central Composite Design
The blocked face-centred central composite design provides high-quality predictions over the entire design space and does not require using points outside the original factor range. Blocking was required to increase precision of experiments as variations of each batch of fruits could not be controlled, and all runs could not be completed within a day. Face-centred refers to alpha value of one and was chosen because the points were at operational limits, and equipment used could set the parameters to certain values only. The combination of factors for each run was generated and arranged using commercial software (MINITAB ® Release 14, Minitab Inc., State College, PA, USA). Table 1 shows experimental runs designed for one batch of fruits. There were 20 runs for each batch of fruits with three replications for each block and batch varying three factors, intensity, time, and temperature, at three levels, −1, 0, and 1, which gave total runs of 60. Responses were measured as percentage change of juice yield, ascorbic acid (AA), and total soluble solids (TSS) content. Fitted regression models were obtained using MINITAB, whereas 4D surface plot was generated using ThreeDify Excel Grapher (v.3.3.8, ThreeDify Inc., Ottawa, ON, Canada) software. The plotted graph chosen depended on the significant factors involved in the improvised models.
The second-order model is widely used in RSM because of its flexibility in taking on a wide variety of functional forms and a good capability to estimate the true response surface [27]. As blocking was used in this study, the second-order model, Equations (1) and (2), were necessary to fit the data [28,29]. The best-fitted models were chosen, and values for the combined factors that satisfied all optimum responses simultaneously were generated.
where Y is the response; β 0 is the constant for intercept; β 01 to β 09 are the constant for blocks; β j is the linear coefficient; β jj is the quadratic coefficient, and β ij is the interaction coefficient. x i and x j are independent variables (i and j are in the range of 1 to k). k is the number of independent variables (k = 3).
Factors points represent by coded values: −1 = Low level; 0 = Centre point; +1 = High level. a Intensity represents motion frequency (rpm) for control, amplitude (%) for direct thermosonication and power (kW) for indirect thermosonication. b Each batch of fruits was replicated three times.

Response Analyses
The responses evaluated were the percentage change of juice yield, AA, and TSS content from the original fruit pulp-distilled water mixture before extraction process. The calculation of percentage change was performed using Equation (3) for each evaluated response.
Percentage change (%) = Treated sample − Original fruit pulp and distilled water mixture Original fruit pulp and distilled water mixture × 100 The percentage of juice yield (% w/w) was calculated following Equation (4) [23,24] based on weight of fruit pulp, distilled water and total centrifuged juice of each sample in duplicate.
Juice yield (%) = Weight of supernatant − Weight of distilled water Weight of fruit pulp × 100 (4) AA content was measured using the 2,6-dichlorophenol-indophenol visual titration method [11]. A 10 mL juice sample was brought to a 100 mL volume with 3% metaphosphoric acid prior to titration with a 10 mL metaphosphoric acid extract of the sample with 2,6-dichlorophenol-indophenol to the pink end-point. AA content was calculated following Equation (5) and averaged from quadruplicate samples.
Ascorbic acid content (mg/100mL) = Titre×Dye factor×Volume made up×100 Aliquot of extract taken for estimation×Weight of sample taken for estimation (5) TSS content was determined using a digital refractometer (PAL-Alpha, Atago, Bellevue, WA, USA) and reported as degree Brix ( • Brix). The analysis for TSS was performed in triplicate.  Table 2 shows the percentage of juice yield, AA, and TSS content of fresh pulp and fruit pulp mixture samples for guava, pomelo, and soursop. The AA and TSS content decreased by almost half from that of fresh fruit pulp after being mixed with distilled water in a ratio of one to one. The presented results are averages of all blocks.

Response Surface Regression Analysis
Both the independent and dependent variables were fitted to full quadratic models, and the response surface regressions were generated to check the goodness of model fit. The standard deviation (S) is important for checking data distribution. The R 2 is important for measuring how much variation in the response is explained by the model. The adjusted regression (R 2 adj ) is useful for comparing models with different numbers of predictors. The p-value of R 2 is useful for determining the relationship between dependent and independent variables, and the p-value of lack-of-fit indicates whether an adequate model has been chosen. A lower S value, p-value of R 2 less than 0.05 and higher R 2 , R 2 adj and p-value of lack of fit indicated a better data fit by the model. The R 2 can be small as long as the p-value of the regression coefficient is statistically significant (p < 0.05). The determination of S, R 2 , R 2 adj and p-value of lack of fit were, however, performed in the background when the model term reduction was performed. Table 3 lists the final fitted models for all responses from direct and indirect thermosonication-assisted extraction procedure and the control after considering values of S, R 2 , R 2 adj , p-value of R 2 , and lack-of-fit. The best-fitted model was necessary to obtain true information on the effect of all three factors to each response. The models were improved by removing insignificant model terms using a backwards elimination procedure to make them easier to work with while maintaining the predictive efficiency [30]. The models' terms in Table 3 interprets the relationships among responses of juice yield, AA, and TSS content and input factors of intensity, time, and temperature for guava, pomelo, and soursop.

Extraction Process Optimisation
For response optimisation, the combined goals of maximum increase in juice yield, AA, and TSS content for the studied range of all three input factors were satisfied by the conditions given in Table 4.  Figure 3 shows the responses of the optimised method of guava juice extraction. The juice yield was affected by temperature linearly, where 47 • C was a point of change of the positive and negative juice yield percentage. It is believed that 47 • C is the cut-off temperature for water reactivity due to ultrasonic waves where trapped water molecules inside guava fruit tissue moved faster until they could break the cell walls for water release. Below 47 • C, the trapped water inside guava fruit tissue was not released and remained in the solid matter during centrifugation process. The AA content displayed a minimum increase after 20 min of extraction, whereas a minimum positive change of TSS content occurred in the middle range of the studied temperatures and power levels. Figure 4 illustrates responses of optimised method of pomelo juice extraction, where temperatures above 28 • C were effective in increasing juice yield. Juice yield started to decrease drastically above 42 • C until the investigated temperature of 55 • C, where the temperature was high and water loss occurred due to evaporation. The AA content increased under a combination of high power and temperature, at 2.3 kW and a temperature of 45 • C. The change of TSS content was influenced by all three factors of power, time, and temperature. The combination of high power, time, and temperature provided minimal increase in TSS content. The highest range of TSS content increment of 2 to 2.5% occurred at high power as well as a short time and low temperature. Figure 5 illustrates responses of optimised method of soursop juice extraction. Juice yield increased when the amplitude was low and temperature was high. Amplitude affected both the AA and TSS content, and they displayed positive changes for the studied amplitude range. The minimum increase in AA content occurred at an amplitude of 60%, whereas the minimum increase in TSS content occurred at 38% amplitude. The AA content was high when low amplitude was applied, whereas high TSS content occurred at high amplitude.
The sonication methods were able to increase AA content in all three fruit juices via removal of dissolved oxygen which favoured decomposition of AA during cavitation [31], whereas the negative change of AA content in pomelo juice was due to degrading of AA in acid solutions [32]. The content of TSS of all three fruit juices extracted using optimised method was greatly affected by heat and/or intensity of direct and indirect ultrasound systems. These factors were also found to be effective in the extraction of mayhaw fruit juice [33].

Adequacy of Models and Verification
The adequacy of the models to predict optimum combined responses was validated experimentally using optimum conditions listed in Table 4. Table 5 shows that the trends and sequences were reasonably close between the predicted and experimental. The indirect thermosonication method was chosen for extracting guava juice because it achieved the highest increases in juice yield and TSS content. The indirect thermosonication method was also good for extracting pomelo juice as all response variables displayed positive trends as the other two methods presented decrease in AA content. The juice yield obtained via indirect thermosonication, however, had lower juice yield than the control. For extraction of soursop juice, the direct thermosonication method was suitable because it achieved a positive change of AA content as compared with the other two methods. A lower value of juice yield and TSS content increase was acceptable as long as AA content did not degrade, because it is the main nutrient that needs to be preserved in fruit juice production. AA also plays a role as an antioxidant, and it is an unstable compound that can degrade easily under inappropriate conditions [32].

Conclusions
The blocked face-centred central composite design was suitable for tropical fruit juice extraction optimisation because blocking helps to control the variations of the fruits' properties. Based on advantages of extraction time reduction and prevention of AA content loss, ultrasound-assisted extraction methods were found to be helpful in the studies of three difficult-to-extract fruit juices. The best method for guava juice extraction was indirect thermosonication at 1 kW using distilled water incubated at 55 • C for 30 min. The indirect thermosonication was also good for extracting pomelo juice at 2.5 kW with water incubation temperature of 54 • C for 23 min. Extraction of soursop juice was suitable via direct thermosonication at 10% amplitude with distilled water at 55 • C for 2 to 10 min.
Author Contributions: N.A. designed the experiments, collected and analyzed the data, and wrote the manuscript. N.L.C. conceptualized the study, supervised the research, edited, and revised the manuscript. All authors have read and agreed to the published version of the manuscript.