2.1 Raw materials

American soybeans were provided by “Carthage Grains” corporation. They were pretreated at an industrial scale before achieving the solvent extraction operation. The sequential pretreatments were the cracking (C), cracking-flaking (CF), and cracking-flaking-expanding (CFEx) (Fig. 2). Afterward, these differently pretreated C, CF, and CFEx soybeans were textured by Instant Controlled Pressure-Drop DIC using the laboratory-scale equipment at different conditions of saturated dry steam pressure P and treatment time t, as defined by the Design of Experiments (DoE). The differently pretreated soybeans were all adjusted in terms of moisture content using an airflow drying at 40 °C and 3 m/s to reach 11% wet basis.

Close

Fig. 2

Diagram and experimental protocol of soybean oil extraction inserting a DIC texturing stage.

Export to PPT

2.2 Chemical product

For the extraction and analysis of soybean oils (fatty acids and tocopherols), the solvents used (methanol, ethanol, toluene, dichloromethane, acetonitrile, chloroform, and hexane) were HPLC and GC -grade solvents were supplied from Thermo Fisher Scientific (Courtaboeuf, France) by UniLasalle Polytechnic Institute (Beauvais, France) and La Rochelle University (France).

2.3 Experimental methodology

N-hexane solvent extraction of soybean oil was accomplished with differently pretreated samples C; CF; and CFEx. To recognize the best manner of Instant Controlled Pressure-Drop DIC texturing, we defined a specific Design of Experiment DoE integrating the main DIC operating parameters. The responses in terms of yields and kinetics were measured as the most effective way to assess the modification of the technological aptitude of the materials. Randall extraction method was adapted to compare various yields while dynamic maceration allowed defining the extraction kinetics. Additionally, chromatography analyses were used for measuring the concentration of different fatty acids and specifying the extracted oil quality.

The protocol based on the integration of DIC texturing in the preparation stage of the soybean seeds before solvent extraction.

Fig. 2 presents different routes for vegetal oil extraction including the possible ways to improve yields and kinetics of soybean oil extraction by adding DIC treatments using a Design of Experiment DoE on each type of pretreated seed at the industrial scale of the solvent extraction line.

2.4 Main intensification ways and technology

DIC texturing process (Fig. 3) had different impacts depending on the operating parameters, the characteristics of soybeans. It changed the porosity and the tortuosity of the treated samples to, consequently, increase the solvent diffusivity within the bean. It also allowed breaking the cell walls and increasing the oil availability in the grain (Jablaoui, 2018). The modifications of oil diffusivity and availability allowed n-hexane extraction to revise Randall yields and dynamic maceration kinetics, respectively.

Close

Fig. 3

Schematic diagram of DIC laboratory scale unit: 1/ Treatment vessel; 2/ Vacuum tank; 3/ Large section-instant opening valve; 4/ water-ring vacuum pump; 5/ Saturated steam generator with condensed water trap.

Export to PPT

Fig. 4 shows that the difference in temperature just before and after DIC flash decompression mainly alters its impacts. By using the saturated dry steam as the heating source, the treatment temperature ranged from 152 to 180 °C depending on the pressure range between 0.5 and 1 MPa. DIC processing time t usually ranged between 10 and 100 s, was defined to ensure adequate thermal and moisture homogeneities within the seed. The pressure-drop is performed towards a vacuum of 4 kPa, with an instantaneous mode in less than 20/40 10⁻³ s. Usually, the decompression rate ΔP/Δt should be higher than 500 kPa s⁻¹.

Close

Fig. 4

Stages of DIC and expander treatments, respectively: A/ Grains to set in the treatment vessel to immediately connect with the vacuum tank (only for DIC); B/ Saturated dry steam to inject till reaching the desired pressure and temperature treatment levels; C/ High pressure/high temperature to maintain for dozens of seconds (for DIC) and some minutes (for expander); D/ Pressure to abruptly drop towards a vacuum (3–5 kPa for DIC), and towards the atmospheric pressure (for expander); E/release from the vacuum towards the atmosphere pressure (for DIC) and a cooling stage from 100 °C towards an almost ambient temperature (for expander).

Export to PPT

Fig. 4 shows the DIC stage. It is worth noticing that a first vacuum phase is achieved to set a close superficial interaction between the material and the saturated dry steam in the treatment vessel. The high-temperature/high-pressure stage is usually defined to allow the grain to reach suitable internal thermal and moisture homogeneities. By abruptly dropping the pressure towards a vacuum, the autovaporization of some amount of the grain water instantaneously happens, provoking an inevitable cooling and a possible texturing.

The structure modification of the grains in terms of expansion ratio and tortuosity inevitably depends on the quantity of the generated vapor m_v and more specifically its volume. For similar treatment parameters of saturated steam pressure and temperature T_t, the autovaporized quantity of m_v strictly depends on the final pressure, and, obviously the final temperature T_f. Since the vacuum level of DIC was about 4–5 kPa, the temperature abruptly reached just after the instant decompression an equilibrium level of about 30–33 °C. This means that m_v was about twice more in DIC with $\left({\mathrm{T}}_{\mathrm{t}}-{\mathrm{T}}_{\mathrm{f}}\right)={\left(152-33\right)}^{°}\mathrm{C}$ than in expander $\left({\mathrm{T}}_{\mathrm{t}}-{\mathrm{T}}_{\mathrm{f}}\right)={\left(152-100\right)}^{°}\mathrm{C}$ . It is well-known that the higher the amount of vapor generated by autovaporization, the higher the texturing effect (Ben Amor and Allaf, 2009).

Subsequently, by instantaneously dropping the saturated dry steam towards a vacuum, DIC incites a much higher possible expansion leading to much better porosity and greater tortuosity than expander can do. The optimized DIC treatment should inexorably imply higher texturing and tortuosity, hence possibly increasing the technological aptitudes of the product in terms of effective diffusivity and starting accessibility with a higher interaction exchange surface. Likewise, the autovaporization/texturing process may also lead to breaking cell walls, which increases the availability of the oil and active molecules in the cells (Allaf and Allaf, 2014) and (Jablaoui et al., 2019).

During the outstandingly short-time pressure-drop (dozens of milliseconds), only the thermodynamics of instantaneity can adequately interpret this far-from the quasi-static process. Since the temperature is fundamentally defined as the density per mole of the kinetic energy of anisotropic translational random micro-agitation, the one-dimension organization of velocity during such so short time decompression results in a precipitous cooling, easily crossing the glass transition and greatly improving the vapor separation (Chemat et al., 2017).

2.5 Study and modeling of the extraction kinetics

After the starting stage of superficial interaction of the solvent, and since n-hexane is capable of greatly dissolving the vegetable oils, the two phenomena of solvent–solute diffusion within the grain, and mass convection at the exchange surface, simultaneously occur. Since the extraction process was made under high-speed stirring conditions, the external superficial transport was so intensified that the extraction process briefly became Negligible External Resistance (NER). The internal effective diffusion of the solvent–solute was the limiting process.

Thus, the solvent extraction kinetics was studied through the phenomenological model of Coupled Washing/Diffusion (CWD). In these conditions, the internal mass transfer revealed the whole extraction process. Subsequently, it was modeled based on similar Fick's law with the effective diffusivity D_eff (m² s⁻¹) to expose the global internal liquid transfer (Fernández et al., 2012). The mathematical solutions change depending on the specific conditions of the shape and average size of the granules. In the present case, we assumed D_eff as a constant, independently on the position within the grain and operation time. Crank solutions can then be adopted as series or error functions for short diffusion times, or trigonometric series or Bessel functions for larger time (Lebovka et al., 2016).

By limiting Crank series to its first term, the logarithm transformation allows linearizing Eq. (3), to identify the slope K. The experimental results to use must excluding the data ranged between the initial time (t = 0) and t₀ assumed as a time that let the extraction be controlled by the diffusion. Thus, the effective diffusivity D_eff is estimated from Eq. (4):

It is noteworthy that only the experimental results for t > to were used in pursuing D_eff. By extrapolating this theoretical model towards the origin (t = 0), an initial yield $Y_0$ usually higher than Y_i = 0 should reveal the starting accessibility ${\delta Y}_s$ yield (g oil/g db):

The difference between the initial value of yield (Y_i = 0) and the theoretical value calculated from the diffusion model divulges the starting accessibility to link with the initial superficial interaction (Bouallegue et al., 2019). The starting accessibility δY_s unfolds the oil quantity issued from the interaction between the solvent and the interaction surface of the product, normally carried out over a very short time.

2.6 Assessments and characteristics

3.1 Yield extraction versus DIC processing parameters

The samples of different mechanically pretreated cracked (C), flaked (CF), and expanded (CFEx) soybeans, as well as the respective DIC textured forms (C-DIC) and (CF-DIC) were all submitted to the solvent extraction at the same conditions (Randall velp-148 using n-hexane at 55 °C, for 6-h). The values of oil yields expressed in (g oil/g db) were shown in Table 1.

In the same solvent extraction conditions, following the two mechanical pretreatments of cracking C and cracking-flaking CF, the soybeans had increased oil yields. However, as we expected, DIC texturing systematically allows the oil availability and yields to be noticeably greater than cracked C and flaked CF soybeans, by 15.55% and 7.34%, respectively.

3.2 Statistical analysis of oil yields

To express and evaluate the kinetic parameters of solvent oil extraction (yields, D_eff, and δY_s), versus the DIC texturing parameters such as the saturated dry steam pressure and processing time, we used Statgraphics software in analyzing the effect plots, estimating the surface response, defining the variance, and fitting the empirical equation of the experimental results.

We also systematically introduced the normalized Pareto diagrams. It is a convenient way to possibly recognize the significant effects of the linear, quadratic, and interaction DIC parameters P and t. The statistical results are represented in Fig. 5. In the case of oil yields, it confirmed that P and t had the highest significant effects with a p-value of 0.05.

Close

Fig. 5

Impact of DIC Parameters on oil yields in cracked C and flaked CF soybean seeds.

Export to PPT

Similarly, Eq.6 gives the second-order empirical regression model of the extraction yields Y_oil (expressed in g oil/g db) for cracked soybeans, versus the DIC processing parameters; P in MPa and t in s with R² = 81.72%.

The DIC parameters for cracked and flaked soybeans were optimized at 0.48 MPa for 105 s and 0.49 MPa and 96 s, with oil yields of 0.2452 and 0.24215 g oil/g db, which presents an increase of 0.033 and 0.00664 g oil/g db or about 14.56% and 3% higher than initial C and CF beans, respectively.

Fig. 5 also shows a higher significant effect of the DIC processing parameters on the extraction yield of flaked soybeans CF with a regression coefficient R² = 89.13 and 0.05 as p-value. Oil availability Y_oil,CF was given as an empirical regression model in Eq. (7):

By DIC texturing the flaked soybeans at 0.49 MPa for 96 s, oil yield reached 0.2452 g oil/g db revealing an increase of 0.017 g oil/g db higher than the simply flaked soybeans.

In both cases C and CF, the optimal saturated steam pressure was identical (0.49 MPa). However, the optimal treatment times were different (105 and 96 s, respectively). This is obviously because the DIC treatment time depends on the thickness and the moisture and heating diffusivities.

The study confirms an improvement in extraction yields and kinetic extraction essentially generated by a change in the technological aptitude of the grains treated with DIC.

Based on the results of the statistical study (separate effects Fig. 5), variation in extraction efficiencies as a function of the calculated amount of self-sprayed steam (Eq (1)), and for a pressure variation interval of [0.27, 0.63] confirms a positive correlation represented in Fig. 6 for both the crushed and crushed - flaked soybean study.

Close

Fig. 6

Correlation of the amount of the autovaporized vapor and oil extraction yield.

Export to PPT

3.3 Statistical analyses of the impact of DIC on extraction kinetics

To study the impact of various modifications of seed structure through the pretreatment operations (cracking, flaking, and expansion) and DIC texturing technology on the kinetics of oil extraction by n-hexane at each pretreated level of soybean, a study of the extraction yields as a function of time was conducted. The purpose of the dynamic maceration extraction was to compare the TRY (time required to achieve the same yields as cracked C beans) for CF beans those from expander and suitable DIC-textured seeds (under optimal DIC conditions). The experimental results (Fig. 7) show that regardless of the level of pretreatment; Cracked, flaked, or expanded, DIC texturing greatly accelerated the extraction process. Thus, while for cracked soybeans, the oil yield of 0.207 g of oil/g db was reached in 160 min, a DIC-texturing made it possible to obtain this level in only 35 min (TRY = 35 min), meaning three times faster extraction capacity. For flaked soybean seeds, the value of TRY was almost similar to the cracked beans, thus 160 min. The effect of expander was to reduce this TRY time to almost 120 min, which is about 1.33 times faster solvent extraction than for cracked or cracked/flaked beans. It is worth highlighting that the optimized DIC texturing is about 4.7 times faster for solvent extraction (TRY = 35 min) (Table 2).

Close

Fig. 7

Impact of Instant Controlled Pressure-Drop DIC texturing on oil extraction kinetics of different pretreated soybeans.

Export to PPT

Based on preliminary results and data issued from literature, the operations of the kinetics were considered for 160 min the differently treated seeds. In the cases of solvent extraction process of soybean, the highly stirred dynamic maceration was defined as a NER Negligible External Resistance process, which means that the resistance of the solute oil-in-solvent transport compared to the internal diffusion transfer was so weak that the properly said unit operation was methodically controlled by the internal mass transfer. We thus reviewed these different cases of kinetics through the Coupled Washing Diffusion CWD phenomenological model. With beans assumed to be spherical form, the slope K (s⁻¹) issued from Eq (4) was used to identify the effective diffusivity D_eff (m²/s). Fig. 7 shows its evolution versus the different processing parameters of DIC.

From the diffusion theoretical curve, the starting accessibility was estimated as the yield of extracted oil before the diffusion process became the limiting phenomenon. It is projected by subtracting the inferred theoretical diffusion value Y₀ and the experimental value Y_i at t = 0 (Eq. (5)). The starting accessibility δY_s divulges the oil extracted through a direct interaction between the grain surface and the solvent.

Cracked-flaked CF soybeans normally had lower effective diffusivity (D_eff = 1.7 against 3.2 10⁻¹² m²/s), and higher starting accessibility δY_s (0.1487 against 0.1242 g oil/g db) than only cracked C seeds. Certainly because of the decrease in porosity and an increase in exchange surface that flaking normally fetches.

Oppositely, by creating much higher porosity than expansion did, DIC-texturing resulted in twice more effective diffusivity while the starting accessibility was 60% higher (Fig. 7 and Table 2). These experimental data demonstrate that DIC texturing is a crucial pretreatment, which marks much better than expander, the higher diffusivity and apparent exchange surface.

Specifically, the experiments confirmed what was predicted from the fundamental instantaneous thermodynamic analyses. Thus, although similar in terms of saturated steam pressure and process temperature, the DIC texturing resulted in a greater intensification of the solvent extraction. The greater the structure modification, the higher the porosity and tortuosity, and thus, the better the technological capabilities of DIC textured soybeans. This was confirmed by scanning electron microscopic (SEM) observations in (Fig. 8) to better recognize the biological microstructure and textural modifications in each pretreated group of seeds.

Close

Fig. 8

Impact of expander and Instant Controlled Pressure-Drop DIC texturing on structure of Crushed - flaked soybeans by scanning electron microscopic SEM observation: (a) non thermal textured seeds; (b) Expander treated seeds; (c) DIC textured seeds; (d) DIC-textured of Expander treated seeds.

Export to PPT

The microstructure of Crushed - flaked soybeans presents effectively organized cells, while expander thermal texturing implies slight modifications of these cells. The impact of DIC was much more important in opening the cells and, thus, increasing the oil kinetic extraction. The most expansion impact became from coupling both expander and DIC treatments.

Moreover, depending on the severity of each operation, possible broken-wall cells may occur. However, the DIC stands out by reaching a more expanded structure, getting, just after decompression, lesser equilibrium temperature (usually 33 °C instead of 100 °C) that is to say no further cooling stage is needed and requiring much lower energy consumption. DIC allows increasing oil yields, quality preservation, and effective kinetics.

3.4 Gas chromatography analyses and soybean oil quality versus DIC treatment conditions

Gas chromatography was used to identify and quantify the fatty acid compounds of the different oil samples. Table 3 presents the two groups of saturated and unsaturated lipids. The amounts of the palmitic acid (C16:0), stearic acid (C18:0), and arachidic acid (C20:0), which are the saturated fatty acids, were preserved after DIC treatment at their initial percentages of 11%, 4.5%, and 0.4%, respectively. The same behavior was observed for the unsaturated fatty acids with oleic acid (C18:1), linoleic acid (C18:2), and linoleic (C18:3) whose percentages were 24.1%, 53%, and 7%, respectively. In optimized DIC texturing conditions, the yield enhancement was accompanied by substantial preservation of the oil quality.

Such a similarity in the lipid profiles of oil extracted from raw material (RM) and DIC-textured soybeans allowed increasing yields, enhancing the kinetics, and preserving the oil and cake quality.

Moreover, the data (Table 3) based on a comparative determination of the tocopherols content (δ, γ, and α), show a slightly-significant variation according to the differently DIC treated soybeans. These modifications inside the values of tocopherol compounds; δ, γ, and α occurred but without any negative impact on total tocopherol content. Moreover, as usual, optimized DIC treatment allows preserving or even increasing the antioxidant availability in various types of plant-based materials, because of the new porous texture performed at high temperature for a very short time (High-Temperature/Short-Time HTST DIC process). The fact that Expander treatment implies much larger thermal treatment time and resulting in much lower porosity than DIC perfectly explains this behavior.

As confirmed in the work of (Mannaï et al., 2019) and (Jablaoui et al., 2018), a high-temperature/short-time treatment, a suitably optimized DIC involves significant conservation of the biochemical composition, even for thermal sensitive molecules. Indeed, since the high-temperature treatment is achieved in a very short time and the final low temperature is immediately reached thanks to the vacuum level, DIC treatment didn’t imply any notable biochemical degradation.

3.5 Industrial application: Energy balance and environmental impact

The integration of DIC in the industrial process of solvent extraction of soybean oil results in a great reduction of the operation time and an overall increase in oil yield with preservation of the quality. Thus, for a total quantity of one ton (1000 kg) of raw material, the oil yields become 177 kg for DIC-soybeans in about 35 min and 163 kg for expander-soybeans in about 120 min against 154 kg in 160 min for nonthermal textured cracked soybeans (Table 4).

The “Expander” and DIC operations need energy consumptions, which can be issued from Eq. (1). They both depend on the treatment temperatures and oil and water contents:

This table allows justifying the interest of using DIC texturing as an intensification way of solvent extraction of soybean oil. The environmental impact is directly correlated with energy consumption.