Cane sugar crystallization using submerged vacuum membrane distillation crystallization (SVMDC)

Abstract

The performance of Submerged Vacuum Membrane Distillation and Crystallization (SVMDC) for cane sugar concentration and crystallization was investigated in this study. Using hollow fiber membrane, the effect of operation parameters, such as feed concentration, feed temperature, and feed agitation were evaluated against the permeate flux. Following the operation parameters optimization, long-term SVMDC tests were performed using cane sugar model solution and raw sugarcane juice as the feed. Porous fouling layer was formed in test using cane sugar model solution which led to membrane fouling and wetting. However, sugar crystals were successfully formed in this test, despite under-saturated final feed concentration of 73.3°Brix. This indicated the occurrence of heterogeneous crystallization in the feed solution, that was induced by the sugar crystals detached from the membrane surface. In test using raw sugarcane juice as the feed, extremely low flux was observed due to the presence of impurities. However, membrane wetting did not occur as the implication of weak drag force occurred due to the low permeate flux. In this test, there was no observable crystal formed as the final feed concentration was much lower than the saturation concentration. In addition, the impurities hindered the interaction of sucrose molecules and disrupted crystal growth.

Introduction

Sugar is one of the biggest commodities in the food industry with annual sales worldwide reaches more than 170 metric tons (USDA 2019). Sugar is mainly produced from sugarcane and the main processes of sugar production are juice extraction, clarification, concentration, and crystallization (de Souza Dias et al. 2015). Clarified juice concentration is conducted in the multiple-effect evaporator at a temperature of 117 °C and pressure of 180 kPa to achieve concentrated juice with concentration of 65°Brix (Rein 2007). Further water removal in the concentrated juice is conducted in a vacuum pan, which is accompanied by simultaneous sugar crystallization (Rein 2007). This establish process has proven to be able to economically produce high-quality sugar. However, continuous improvement of the production process is a must, particularly as energy efficiency has become a stressing issue (Gul and Harasek 2012). The highest energy consumer in the sugar production process is the evaporator as it requires great amount of steam to heat the clarified juice and vaporize the water (Madaeni and Zereshki 2010).

Lately, membrane distillation (MD) has gained much interest in solution concentration sucrose concentration (Izquierdo-Gil et al. 1999a) and water removal in various applications (Edwie and Chung 2013). Membrane distillation is a separation process with vapor pressure gradient as the driving force. As the hydrophobic membrane is used to separate the feed solution and permeate carrier, water in the feed solution should travel through the membrane as vapor (Chen et al. 2018), promoting almost 100% solute rejection (Alkhudhiri et al. 2013). Even though most studies in MD focus on desalination application, MD has a great potential to replace evaporator in sugar production due to the compactness and low energy consumption of membrane-based processes (Ruiz Salmón and Luis 2018). Several studies investigated the effect of feed temperature, concentration, and flow rate on the permeate flux in Vacuum Membrane Distillation (VMD) configuration (Hirota et al. 2016). Using polypropylene membrane with pore size of 0.2 µm and sucrose solution with concentration ranged from 0 to 17 wt.% as the feed, the permeate flux increased with decreasing concentration and increasing feed temperature and flow rate (Al-asheh et al. 2006). In another study, VMD was conducted using Al₂O₃ membrane and the results were in agreement with the research by Al-Asheh et al. (2006) (Hirota et al. 2016). Highest flux of 7 Lm⁻² h⁻¹ was achieved at feed temperature of 60 °C and feed concentration of 10%-wt. Further research on other MD configurations, such as Direct Contact Membrane Distillation (DCMD) and Air Gap Membrane Distillation (AGMD) have also been conducted (Izquierdo-Gil et al. 1999b). However, the focus of most investigations was limited to the evaluation of operation parameters on the permeate flux and the utilization of sucrose solution as the feed, which could not justify the robustness of MD for industrial application of sugarcane juice concentration.

Long-term performance of AGMD was explored in a study using sucrose solution with concentration of 150 g/L as the feed solution. The results suggested insignificant flux change over a month experiment period, which indicated the potential of MD as an evaporator replacement in sugar industry (Izquierdo-Gil et al. 1999a). In another study, the DCMD flux profiles in operations with raw and clarified sugarcane juice as the feed solutions were investigated. The raw sugarcane juice was clarified by microfiltration and the concentration of the clarified juice was 20°Brix. However, the permeate flux of test conducted using clarified juice as the feed was significantly lower than that of the test with raw juice. Further investigation indicated higher polysaccharides concentration in the clarified juice, which resulted in more severe concentration polarization. At condition of higher feed concentration on the membrane-feed interface, the vapor pressure decreased and subsequently, the driving force for water vapor transport was reduced at test with clarified sugarcane juice. (Nene et al. 2002).

Recently, process development by combining MD with crystallizer has gain popularity as the proposed process is claimed to have higher energy efficiency (Chen et al. 2014) with the potential of zero waste discharge (Guan et al. 2012). In the application of desalination with brine water as the feed solution, 89% of high-quality water could be removed, and simultaneously, salt crystal could be harvested (Mericq et al. 2010). Further process intensification in the MD-crystallizer study involved the submersion of the membrane module in a tank that serves as feed tank and crystallization tank (Julian et al. 2016). The submerged MD-crystallizer configuration is claimed to be more energy-efficient as the need for solution recirculation is eliminated and more uniform feed temperature can be achieved to promote better heat transfer (Julian et al. 2018a). As the sugar production process involves successive juice concentration and sugar crystallization process, the application of submerged MD-crystallizer becomes an attractive alternative to increase the energy efficiency of the process.

Even though the performance of simultaneous MD-crystallizer has been thoroughly studied, crystallization is a complex process and many efforts are needed to further understand its mechanism in MD-crystallizer operation. Previous study on submerged vacuum membrane distillation crystallization (SVMDC) for inland brine water treatment highlighted the occurrence of both homogeneous and heterogeneous crystallization. While homogeneous crystallization of the feed solution was highly desirable, the heterogeneous crystallization on the membrane surface resulted in severe fouling and poor SVMDC performance (Julian et al. 2018b). Effort to alleviate the fouling deposition on membrane surface in SVMDC operation focus on the engineering of feed hydrodynamic to induce shear force on the fouling layer (Phattaranawik et al. 2001). The shear force reduced the concentration polarization and temperature polarization, which indicated lower feed temperature on the feed-membrane interface compared to that on the bulk feed solution (Yun et al. 2011). In addition, the shear force could also scrape out the foulant from the membrane surface (Julian et al. 2016). This strategy was not only prolonging the MD operation and increase the permeate flux but also might induce the heterogeneous crystallization in the bulk feed solution as the detached crystal could act as the nucleation site in the bulk feed solution (Sluys et al. 1996). However, several studies resulted in opposite findings and suggested that the increase of feed hydrodynamic to a particular condition inhibits the crystallization in the feed solution due to the disruption on nuclei formation and crystal growth (Shin and Sohn 2016). This indicated that the evaluation of the effect of feed hydrodynamic on the SVMDC performance should be specifically conducted for particular application.

To date, there is no study on simultaneous MD-crystallizer for cane sugar production, particularly in submerged configuration. In this research, the application of SVMDC for cane sugar crystallization was investigated. Operation parameters including feed temperature, feed concentration, and feed hydrodynamic were examined to determine their influence on the permeate flux. At optimum operation parameters, crystallization of cane sugar was conducted in a long-term experiment using commercial cane sugar model solution and raw sugarcane juice as the feed. The SVMDC performance, in terms of permeate flux and sugar crystal production, were evaluated. Furthermore, the effect of feed composition on the fouling formation and membrane wetting in SVMDC operation was discussed in detail.

Material and methods

Materials

Membranes used for SVMDC in this study were polypropylene hollow fiber membranes with inside diameter, outside diameter, pore size, and contact angle of 0.3 mm, 0.4 mm, 50 nm, and 105° respectively. The membranes were made into modules by potting 50 fibers into a nylon tube with total area of 0.012 m². The cane sugar model solution at various concentrations was made from commercial sugar crystals dissolved in demineralized water. Sugarcane juice was freshly pressed and used as the feed solution without further purification. The concentration of the dissolved solids in the fresh raw sugarcane juice was 15°Brix.

Evaporation of cane sugar model solution

Evaporation of cane sugar model solutions was conducted to investigate the critical concentration for crystallization of cane sugar solution in water. A series of 100 mL cane sugar model solution with uniform initial concentration of 35°Brix was prepared by dissolving 350 g of commercial sugar into demineralized water to make 1 L of sugar solution at room temperature. The solutions were then immersed in a water bath with temperature of 80 °C. Each solution was removed from water bath in a predetermined period then underwent natural cooling until it reached room temperature of 25 °C. The concentration of sugar in the final solutions was measured by refractometer and the formation of sugar crystal in the solution was observed visually.

Submerged vacuum membrane distillation crystallization (SVMDC) experimental set-up

The experimental set-up of SVMDC is illustrated in Fig. 1. The feed tank with a volume of 50 mL was immersed in a water bath and the temperature of the feed was continuously monitored to maintain consistent temperature throughout the test. The membrane module was submerged on the feed solution, and the open end of the membrane module was connected to a peristaltic pump to create vacuum condition of − 80 kPa in the lumen side of the membrane. A static stirrer was inserted into the feed tank to create shear force on the membrane surface in particular tests. To avoid water evaporation during the tests, the feed tank was carefully sealed with minimum opening to insert the membrane module. The water accumulation in the permeate tank was measured in an hourly period to determine the permeate flux.

Fig. 1.

Submerged VMDC experimental set-up

To determine the optimum operating condition of SVMDC, the tests were conducted using cane sugar model solution for one hour at varied feed temperature (50–70 °C), feed concentration (0–50°Brix), and feed stirring speed (0–400 rpm). All experiments were performed in duplicates. The Reynolds number of the feed solution was proportional to the stirring speed and can be determined as follow:

$$N_{\mathit{Re}}=\frac{\rho N{D}^2}{\mu }$$ 1

with N, D, ρ, and µ were the stirrer revolution per second, baffle diameter, fluid density, and viscosity at particular conditions, respectively. At the optimum operation condition, long-term SVMDC tests were conducted using cane sugar model solution and raw sugarcane juice as the feed solution. The SVMDC was operated until the permeate flux reached zero or wetting occurred, then the feed tank underwent natural cooling to the room temperature of 25 °C. In all SVMDC tests, the concentration of feed and permeate were hourly evaluated to detect membrane wetting. Sugar concentration in the cane sugar model solution and raw sugarcane juice, as well as in the permeate tank were measured by refractometer and reported as degrees Brix, which corresponds to the amount of dissolved solid of an aqueous solution.

Statistical analysis

The data obtained from tests to determine the optimum operating condition of SVMDC were analyzed using the Analysis of Variance (ANOVA) to determine the significance of various operation conditions on the permeate flux at significance level of 0.05.

Results and discussion

Operating conditions optimization on SVMDC operation for cane sugar crystallization

In this set of tests, cane sugar model solution was used as the feed solution. The tests were conducted at varied feed concentration, feed temperature and feed stirring ranged from 0–50°Brix, 50–70 °C and 0–400 rpm, respectively. The flux of VMD at various operating conditions are shown in Fig. 2a. Exponential flux decline was noticed as the feed temperature increased at tests with similar feed concentration due to the increase of vapor pressure, in accordance with the Antoine’s equation. At equal feed temperature, permeate flux reduction was observed as the feed concentration increased. This was due to the reduction of water activity in the solution of higher solute concentration and consequently, lower vapor pressure as the driving force for mass transport. In addition, at higher solute concentration, the effect of concentration polarization becomes more severe (Schofield et al. 1990). This resulted in extremely higher solute concentration and even lower vapor pressure in the feed-membrane interface. Two-way ANOVA at a confidence level of 95% was performed to investigate the significance of feed temperature and feed concentration on the permeate flux. The P-value of feed temperature and feed concentration on the permeate flux was 1.98 × 10⁻⁶ and 2.9 × 10⁻⁴, respectively. This indicates that the permeate flux was only significantly affected by the feed temperature and feed concentration.

Fig. 2.

VMD permeate flux at varied a feed temperature and concentration (feed stirring rate of 0 rpm) and b feed Reynolds number (feed temperature of 70 °C and feed concentration of 50°Brix

The impact of feed agitation on the VMD permeate flux was examined by stirring the feed solution at 0, 250, and 400 rpm. The tests were conducted at feed temperature of 70 °C and feed concentration of 50°Brix. The results are presented in Fig. 2b as a function of Reynolds number associated with the aforementioned stirring speeds. The enhancement of permeate flux was observed with the increase of feed Reynolds number. At increased Reynolds number, the feed was more turbulent and the shear rate on the membrane surface reduced the boundary layer thickness in the feed-membrane interphase. Hence, the severity of temperature polarization and concentration polarization on the membrane surface reduced, which resulteinon higher feed temperature and lower sugar concentration at the feed-membrane interface, specifically when compared to the test without feed stirring. At this particular condition of VMD with feed stirring, the driving force for vapor transport increased. However, it is worth to note that a maximum of 18% flux enhancement was observed at test with 400 rpm (N_Re = 1200) feed stirring. One-way ANOVA of feed stirring on the permeate flux was conducted with confidence level of 95% and resulted on P-value of 0.72. This indicates the insignificance influence of feed stirring on the permeate flux, which might occur as the Reynolds number of the feed solution indicated laminar flow.

At test with feed stirring, the crystallization of sugar, which has a positive temperature-solubility coefficient, on the membrane surface was inhibited due to the reduction of temperature polarization and concentration polarization. While sugar crystallization on the membrane surface increases the fouling propensity, at test with feed stirring, the formed sugar crystals could be detached from the membrane surface and served as the seed for heterogeneous crystallization in the bulk feed solution. Apart from hindering the seed crystal formation, the feed stirring did not promote sugar homogeneous crystallization as well. Rapid feed concentration to up to supersaturation concentration is essential for homogeneous crystallization to occur; however, this could not be achieved by feed stirring due to insignificant flux enhancement.

SVMDC application for sugar crystallization from cane sugar model solution

To investigate the applicability of SVMDC to concentrate and crystallize sugar solution, long-term test using cane sugar model solution was conducted. The feed temperature was 70 °C as it was resulted in highest permeate flux (Fig. 2a) and no stirring was conducted as no significant flux enhancement could be achieved by feed stirring, even at the high speed of 400 rpm. To determine the feed concentration used in this test, a preliminary experiment on sugar crystal formation at various concentrations of cane sugar model solution was conducted and the results are presented in Fig. 3. At the concentration below 58°Brix, no sugar crystal was observed. As the solution concentration increased to up to 68°Brix, smooth sugar crystal was formed, however, the crystal could be easily dissolved back to the solution with mild feed agitation. Firm sugar crystals were observed at feed concentration of more than 68°Brix and this is in accordance with the literature stated sucrose solubility of 67.47 g/100 g of water at 25 °C (Mathlouthi et al. 1995). Based on the results in Fig. 3, feed solutions with initial concentration of approximately 50°Brix were used to accelerate the sugar crystal formation.

Fig. 3.

Sugar crystal formation at various concentration of cane sugar model solution

As indicated in Fig. 4a, initial flux of 0.2 Lm⁻² h⁻¹ was achieved prior to gradual flux decline. Rapid flux decline occurred at t = 6 h and the flux dropped to approximately 0.03 Lm⁻² h⁻¹ at t = 9 h, followed by sudden increase of permeate flux at t = 11 h. The flux decline occurred due to the synergistic effect of feed concentration enhancement and fouling formation. The gradual flux decline, which observed until t = 6 h, was occurred due to the increase of the feed concentration. As more vapor was transported to the permeate side, the feed concentration increased over time, which resulted in vapor pressure reduction on the feed solution. Furthermore, the increase of feed concentration would also lead to an increase of sugar concentration at feed-membrane interphase and induce the sugar crystallization on the membrane surface due to the concentration polarization phenomena. The formation of sugar crystals in the membrane surface resulted in rapid flux decline, as observed from t = 6 h to t = 9 h. In test with cane sugar model solution, fouling layer consisted of tiny sugar crystals which created porous fouling structure. This particular structure still allowed vapor transport with added resistance.

Fig. 4.

a Permeate flux profile, b feed and permeate concentration profile of SVMDC tests using cane sugar model solution (Feed temperature of 70 °C, feed stirring of 0 rpm, initial cane sugar model concentration of 49°Brix)

In addition, the deposited sugar crystals on the membrane surface created hydrophilic layer which could promote water intrusion through the membrane pores. Wetting is not an instantaneous process, and the occurrence of bulk feed intrusion to the permeate side was likely to be induced by partial wetting, in which some pores of the membrane were partially lost its hydrophobicity and bulk feed solution could penetrate to some extent. Sudden increase of permeate flux at t = 11 h indicated membrane wetting, evident by the evaluation of sugar concentration in the permeate tank. While no sugar was detected in the first 11 h of test, instantaneous sugar concentration was detected at t = 12 h (Fig. 4b). This indicated the transport of bulk feed solution in aqueous phase, in addition to the expected vapor, through the wetted membrane pores. The presence of aqueous phase on the membrane pores contributed to the flux decline, as it enhanced the mass transport resistance of water vapor in the membrane pores. Once the pores lost their hydrophobicity, total wetting occurred, and the aqueous feed solution could transport through the wetted pores with less resistance, resulted in increase flux and reduced separation performance. Observation of the permeate conductivity showed continuous increase of concentration in the permeate tank and the test was terminated at t = 15 h.

In test with cane sugar model solution as the feed, even though fouling became a drawback in sugar crystallization, significant amount of sugar was successfully crystallized and harvested from the feed tank. As the feed concentration increased from 49 to 73°Brix, clearly visible sugar crystals were formed in the feed tank (Fig. 4a) after 15 h of SVMDC operation and feed concentration reached 73.3°Brix. This highlighted the potential applicability of SVMDC for simultaneous sugar concentration and crystallization. However, it is worth noting that the crystallization of sugar in the feed solution occurred below its supersaturation concentration at 70 °C, which is 76.5°Brix (Mathlouthi et al. 1995). This suggested that the bulk crystallization mechanism during SVMDC in this study was heterogeneous, instead of homogeneous. As no seeds were deliberately added to the bulk feed solution, heterogeneous crystallization could only be induced by sugar crystals detached from the membrane surface.

SVMDC application for sugar crystallization from industrial raw sugarcane juice

Further test using industrial raw sugarcane juice with initial concentration of 15°Brix was conducted to study the effect of impurities on SVMDC permeate flux. The concentration of the juice was increased to 50.5°Brix by evaporation at 70 °C, prior to SVMDC operation. The feed temperature for SVMDC was 70 °C and no feed stirring was conducted. The flux profile of the test is presented in Fig. 5a indicating an initial flux of 0.14 Lm⁻² h⁻¹. Compared to the test with cane sugar model solution, the initial flux in test with sugarcane juice was lower due to the presence of impurities, such as suspended solids, pigment, and polysaccharide. The impurities further reduced the vapor pressure and driving force for vapor transport.

Fig. 5.

a Permeate flux profile and b feed and permeate concentration profile of SVMDC using industrial raw sugarcane juice (Feed temperature of 70 °C, feed stirring of 0 rpm, initial raw sugarcane concentration of 50.5°Brix)

Gradual flux decline was also observed in test with the industrial sugarcane juice, however, neither rapid flux decline nor instantaneous flux enhancement occurred. This was in opposite to the test with cane sugar model solution and indicated the absence of membrane wetting, evident by extremely low permeate concentration throughout the test (Fig. 5b). While severe fouling occurred in both tests, the difference in permeate flux significantly affected the wetting propensity. As presented in Fig. 5a, the permeate flux of test with cane sugar model solution was twice as high as the test with raw sugarcane juice. The relatively higher flux resulted in higher drag force that induced solute intrusion into the membrane pores. In addition, as the permeate flux in test with the industrial raw sugarcane juice was lower than that with cane sugar model solution, the concentration polarization was not as severe. Therefore, the concentration of solute, which had hydrophobic characteristics at the feed-membrane interface was lower and the membrane hydrophobicity was further maintained.

However, at equal operation time, the final feed concentration in test with sugarcane juice was lower than cane sugar model solution due to the lower permeate flux throughout the test with sugarcane juice (Fig. 5a). As a comparison, the final feed concentration of test with raw sugarcane juice as the feed was 64.8°Brix, which was similar to feed concentration of test using cane sugar model solution as the feed at t = 10 h. There was no crystal formed at test with the industrial raw sugarcane juice at the end of the test as the feed concentration was much lower than the supersaturation condition. In addition, viscous solution was observed as the membrane surface, instead of sugar crystal (Fig. 6a). This was due to the presence of impurities in sugarcane juice, particularly pigment and salts, which hindered the interaction of sugar molecules to form a nucleus and start the crystal growth.

Fig. 6.

Visual observation of sugar crystal formed on the membrane surface at the end of the SVMDC test using a industrial raw sugarcane juice and b cane sugar model solution

As a comparison, the deposition of sugar crystal in the membrane surface at the end of the test with cane sugar model solution could be visually observed in Fig. 6b. As concentration polarization occurred during SVMDC operation, sugar concentration at the feed-membrane interface was higher than that on the feed solution and the propensity of crystal formation increased. In addition, membrane surface could act as the nucleation site to promote heterogeneous nucleation, which requires lower energy activation than that of homogeneous nucleation. Thus, the supersaturation concentration of sugar was not necessarily needed to be achieved for the crystal formation in the membrane surface to occur. The crystal formation in membrane surface was inevitable and undesirable as it hindered the vapor transport, which resulted in flux reduction and higher propensity of membrane wetting. However, it was beneficial in terms of inducing heterogeneous sugar crystallization in the bulk feed solution, as discussed in “SVMDC application for sugar crystallization from cane sugar model solution” section.

Conclusion

This study provides insight into the simultaneous concentration and crystallization of sugar solution using SVMDC operation. SVMDC tests at various operation parameters indicated the significant influence of feed temperature and feed concentration on the permeate flux. In opposite, feed agitation to up to 400 rpm was unfavorable as it showed insignificant influence on the permeate flux and prohibit sugar crystallization. At long-term SVMDC operation using cane sugar model solution as the feed at optimum operation parameters (70 °C feed temperature, 50°Brix initial feed concentration, and 0 rpm feed stirring), crystallization of sugar on the membrane surface occurred, constructing porous fouling layer that resulted on flux decline and membrane wetting. However, sugar crystallization on the feed solution, which usually occurs through the homogeneous nucleation mechanism, was observed at feed concentration below saturation condition. This highlighted the advantages of crystal formation on the membrane surface as the detached crystals induced the heterogeneous crystallization of sugar on the feed tank.

At test with SVMDC with industrial raw sugarcane juice as the feed solution, lower flux was observed compared to the test with cane sugar model solution. This was due to the presence of impurities and resulted in delayed feed concentration enhancement. During 15 h of SVMDC operation, no wetting occurred as a result of the low permeate flux which was directly implicated to low drag force for solute intrusion into the membrane pores. However, there was no crystal observed in the bulk feed solution and at the membrane surface as the feed concentration was much lower than its supersaturation condition and the impurities in the raw sugarcane juice hindered the interaction of sugar molecules to form a nucleus and start the crystal growth. Therefore, further investigation of the crystallization of sugar from clarified raw sugarcane juice is essential.