## Brazilian Journal of Chemical Engineering

##
*Print version* ISSN 0104-6632*On-line version* ISSN 1678-4383

### Braz. J. Chem. Eng. vol.17 n.4-7 São Paulo Dec. 2000

#### http://dx.doi.org/10.1590/S0104-66322000000400035

**SUPERCRITICAL EXTRACTION OF OLEAGINOUS: PARAMETRIC SENSITIVITY ANALYSIS**

**M.M.Santos*, E.A.Boss and R.Maciel Filho **School of Chemical Engineering, State University of Campinas, C.P. 6066

13081-970, Campinas - SP, Brazil

*( Received: November 30, 1999 ; Accepted: April 6, 2000)*

Abstract -The economy has become universal and competitive, thus the industries of vegetable oil extraction must advance in the sense of minimising production costs and, at the same time, generating products that obey more rigorous patterns of quality, including solutions that do not damage the environment. The conventional oilseed processing uses hexane as solvent. However, this solvent is toxic and highly flammable. Thus the search of substitutes for hexane in oleaginous extraction process has increased in the last years. The supercritical carbon dioxide is a potential substitute for hexane, but it is necessary more detailed studies to understand the phenomena taking place in such process. Thus, in this work a diffusive model for semi-continuous (batch for the solids and continuous for the solvent) isothermal and isobaric extraction process using supercritical carbon dioxide is presented and submitted to a parametric sensitivity analysis by means of a factorial design in two levels. The model parameters were disturbed and their main effects analysed, so that it is possible to propose strategies for high performance operation.

Keywords:Supercritical Fluid Extraction, Natural Oil, Parametric Sensitivity, Factorial Design.

**INTRODUCTION**

**Extraction of Oleaginous**

The conventional process of oleaginous extraction uses hexane as solvent. This substance is highly toxic and flammable, thus increasing danger in the production atmosphere and damage to the environment. Besides, due to the successive crises of petroleum in the last decades, the perspective of petroleum reservations end as well as the need of using clean technologies, the studies of viability of supercritical fluids application for extraction and processing of renewed natural resources as vegetable oils have been intensified.

It is in this context that is being studied the hexane substitution by supercritical solvents. The supercritical fluids are effective solvents for the oilseeds extraction, and also present a series of peculiarities. They present advantages in relation to liquid solvents generally used. The positive points are the absolute absence of residues in the products, which is an interesting characteristic for the food industry, simpler and cheaper recovery stages. A wide variety of solvents can be used due the solvent properties at the thermodynamic conditions and the selectivity of a certain solute can be controlled manipulating the solvent density or adding a co-solvent.

The carbon dioxide, which is already used in discontinuous processes to extraction of high valuable products, has been studied as an important substitute for hexane. Besides the inherent characteristics to the supercritical means, the carbon dioxide has been proved a potential solvent for the hexane substitution because it is not toxic, not flammable, presents supercritical conditions relatively mild (critical temperature of 31ºC and critical pressure of 73.8 bar), and is available at low cost.

In this work the model for oleaginous supercritical extraction is used and the parametric sensitivity analysis of this process using factorial design in two levels is analysed.

**Mechanisms of Oleaginous Extraction**

The oilseed extraction has been made, in industrial scale, since the second decade of the century XX. However, the process mechanism is not very well established.

The mass transfer of non-adsorbed solute inside solids, whose pores are completely filled by the extractive fluid, may be described in agreement to the first Fick's law. That is, it is proportional to mass transfer coefficient and concentration gradient. However, the first works that investigated oleaginous extraction mechanisms indicate that extraction rate presents weak dependence and is not proportional to concentration gradient (Karnofsky, 1949, Othmer e Argawal, 1955).

Due to these verifications, mechanisms have been developed that propose, each one of them, a limiting stage in oilseeds extractive process.

The slow oil break-up model (Karnofsky, 1949) supposes that certain oil components, not very soluble in hexane, would inhibit the solution of the triglycerides. Later, the inhibition action was attributed (Karnofsky, 1986) to the fosfatides presence. It is admitted that the existing oil inside seeds particles can be divided in a portion dissolved in the stagnant solvent and in another non dissolved portion. The resistance to the extraction would be, above all, in the break-up of the non dissolved oil, being the diffusion of the oil dissolved to the bulk micelle (oil/solvent mixture), a process relatively fast.

According to the mechanism of capillary flowing (Othmer and Argawal, 1955), the extraction would be a fluid-dynamic problem fundamentally in which the solvent and the oil flowing in a complicated system, formed by vegetable cells broken up with the break and the lamination of the seeds. Firstly, there would be the oil break-up of the particle surface and of the closer broken cells. The solvent, then, would penetrate for the capillary spaces to dissolve more oil, establishing in these ways, concentration gradients. The capillary flowing would take the oil until the surface of the particle so that new diffusion could happen. Thus the potential for flowing decreased together with the extraction rate as if it approached the equilibrium.

Relatively recent experiments (Nieh and Snyder, 1991), accomplished with powdered soy, revealed extraction times very inferior to the necessary ones for flakes (laminate particles). If the slow oil break-up was to be the limiting stage of the flake extraction, it should be interesting to have in powdered solids.

The extraction of corn oil with ethanol presented similar behaviour for the powdered extraction and larger particles (Chien et al., 1990), when these particles were pre-treated with ethanol in a proportion of 0.4 ml_{ethanol}/g_{corn} for 18 hours. With this, the diffusion of the ethanol in "dry" areas of the particles is much smaller than in areas wet by the solvent. Thus, the diffusion of the solvent for the dry solid would be the limiting stage of the extraction process.

It is probable that the decrease in the solid particles dimension eliminates the resistance to the solvent and oil transports inside these particles, progressively. Also, the resistance to the mass transport of the solvent in the solid pores, still empty, could be elevated enough to mask the effects of solvent/oil against-diffusion.

In spite of these verifications, the most recent works treat the problem of the extraction of oleaginous in a way that do not take this into account. It is for example the case of the works that used an equation for the extractor phase based on the convection-diffusion equation (Cuperus et al. and Stastová et al., 1996, Sovová, 1994). Although the effective coefficients of mass transfer can include several dispersion effects, and due to the presence of the solid middle, that approach type has shown to be inadequate, to be used in the design of processing units. However, they have a suitable use for experimental parameter determinations as well as for optimisation and control evaluations.

**MATHEMATICAL MODELING**

The extraction model used by Sovová (1994) consists on the solvent flowing axially and continuously with superficial velocity U through a bed of milled plant material in a cylindrical extractor. The solvent is solute-free at the entrance of the extractor, and the thermal effects as well as the drop pressure along the bed are negligible (isothermal and isobaric process). The initial distribution of solute and the particle size are both constants in the extractor. It is reasonable to admit the velocity profile is plug flow and also the axial diffusion effects can be neglected and the bulk fluid phase mass transport occurs by convection. The solute is deposited in plant cells and protected by cell walls. However, the milling breaks parts of the walls. Thus some amount of solute is directly exposed to the solvent. Bearing this in mind, the following equations can be written to describe this process.

The equations consist in mass balance for solid phase:

(1) |

and mass balance for fluid phase:

(2) |

The initial and boundary conditions are:

(3a) |

(3b) |

(3c) |

where t_{lim} is the necessary lower time to the saturated solvent reaches analysed position *h*, at U velocity; r_{s} is solid density (kg/m^{3}); e is void fraction of the bed (m^{3}/ m^{3}); x is oil concentration in solid (kg oil/kg oil free solid); t is time (s); r is CO_{2} density (kg/m^{3}); y is oil content in solvent (kg oil/kg oil free solvent); h is axial position along the extractor (m) and U is solvent superficial velocity (m/s); y_{i} is solvent concentration value to the lower time t_{lim} and J is mass transfer rate.

The extractive process is divided in two stages separated by a residual concentration x_{k}. Firstly, the easily accessible solute is extracted. When the oil concentration in solid phase decreases to x_{k}, the mass transfer is limited by the diffusion in the solid phase, that is:

(4) |

Several expressions for mass transfer rate at supercritical fluid extraction are showed in Table 1. If the easily accessible solute crosses the interfacial boundary fast enough to keep the solvent at the boundary saturated, the mass transfer rate is:

(5) |

where k_{f}a_{0} is the global mass transfer coefficient based on supercritical phase. The other stage, slower than the first, has resistance to mass transport in the solid matrix, given by the expression:

(6) |

In this model, y_{r} is the oil solubility (kg oil/kg CO_{2}) in the solvent and as the solubility of various vegetable oils is similar, it can be estimated from a common correlation (del Valle and Aguilera, 1988):

(7) |

where T is temperature(K) and r is density of CO_{2}. This correlation is valid in the temperature range 20-80ºC at pressures ranging from 150 to 800 atm at 20-60ºC, 150 to 750 atm at 70ºC and 150-680 atm at 80ºC.

The data used in this work were 280 bar, the extraction temperature was 40.0ºC, and at these conditions, the density of pure carbon dioxide was 899 kg/m^{3} and the oil solubility was 0.00685 kg/kg CO_{2}.

This model was solved by 4^{th} order Runge-Kutta, in respect to time dependence, associated with a finite difference method for spatial variable discretization.

**PARAMETRIC SENSIVITY ANALYSIS**

**Main Effects**

The parametric sensitivity analysis was carried out with a mathematical description based on the Sovová´s model (1994), applying disturbances of ± 10% in the values of the normal operation conditions as showed in Table 2.

From these disturbances, the influence of these variables was analysed in the oil solid concentration, x, and in the oil solvent concentration, y, at some axial positions of extractor, in pilot scale.

The parametric sensitivity analysis is made at the extractor exit (laboratory scale). The oil residual concentration in solids and solvent are showed in Figs. 1 and 2, respectively.

In the first few minutes of extraction the saturated solvent with oil reaches the analysed position and, thus, the disturbances in U, x_{0}, y_{0} and dp do not produce effects, maintaining the oil residual concentration in solids. After some time, the non-saturated solvent reaches the analysed position and the disturbances effects are observed until the residue in the solid matrix is smaller than x_{k}, when the rates drop. The intermediate region where the rates are relatively high is denominated extraction front region.

The increase in solvent superficial velocity generates decrease in the oil residual concentration in solids (Fig. 1) and in solvent (Fig. 2). This increase in velocity generates greater turbulence in solvent flowing, increasing the mass transfer. The oil residual concentration in solvent decreases due to greater solvent amount.

Increasing the oil initial concentration in solids (x_{0}), greater concentration in solids is obtained during all first extraction period. When the extraction front reaches the analysed position, the effect in x is constant and positive. The effect in y is null. After this point is reached, the effects increase and after some time they are null. That happens because with this increase, firstly it has greater oil amount in solids and the solvent continues to extract oil until its saturation. On the other hand, the oil residual concentration in solvent is not affected. As the extraction progresses, the oil residual concentration in solids reaches its original value, without disturbances, and its effect in solids tends to be very small. Alternatively, the oil residual concentration in solvent increases because without disturbances, the available oil amount for extraction would be smaller.

With the increase in oil initial concentration in solvent, firstly the effect in solid concentration is positive because the extracted amount by the solvent is smaller. But this effect is null as the extraction progresses. The effect in solvent oil concentration, initially null, rises until it reaches the value corresponding to increase in the solvent oil concentration.

The particle diameter effect to both concentrations is null at the initial stages of extraction, that is, before the front reaches the analysed position. When the extraction front reaches the analysed position, larger particles generate smaller concentrations. In the second extraction period the behavior inverts, that is, larger solvent concentrations for larger particles (Fig. 2). The effect of the inverse response happens for the fact of having extraction in both periods in the position where the effects are analysed as well as in previous positions. Smaller solvent concentration, as the front arrives, indicates that in previous positions the extraction occurs, mainly, in the second extraction period. This generates, at the observed position, larger rates that reflect a smaller residual oil concentration in solids (Fig. 1). As the extraction progresses, however, the solvent concentrates in previous positions and the extraction starts to happen at smaller rates, leading to larger residues, up to, in the same way that in previous positions, it passes for the second period. This influence in previous positions is evidenced by the behavior presented at the extractor entrance, showed in Fig. 3. In this figure it is not observed the inverse behaviour of particle diameter effect.

**Secondary Effects**

Figure 4 shows the secondary effects of the solvent superficial velocity with the others variables in the oil residual concentration in solids (x) at the extractor exit and sampling time of 200 seconds. Figure 5 shows the effects in the oil residual concentration in solvent (y).

In agreement with figures 4 and 5, increasing the solvent superficial velocity and the particles diameter simultaneously, the secondary effect in oil residual concentration in solvent is -10% and in solid is -7% in the extraction front arrival region. Negative effects in oil residual concentration in solids and positive effects in oil residual concentration in solvent are suitable for extraction. In this case, the negative effect in solids is suitable. However, the negative effect in solvent prejudices the extraction. As the negative effect in oil residual concentration in solvent is greater, in absolute values, the simultaneous increase of U and dp worsens the extraction. In the second extraction period, the observed behavior is inverse. The increase of U and dp is suitable to extraction.

In extraction front period, the increase of U and y_{0} simultaneously presents effects of about +5% and +10% in the oil residual concentration in solids and solvent, respectively. That is, this increase worsens the extraction in 5% and helps in 10%. Thus, this interaction is suitable. In the second period of extraction, the interaction effects present inverse response, until the effects become very small.

Increasing U and x_{0} simultaneously, in the first period the interaction effects are about +13% and +18% for the oil residual concentration in solids and in solvent, respectively. This indicates that the increase is suitable for the extraction. In the second period, the effects are about -110% in solids and -75% in solvent. In this period, therefore, this increase is suitable to the extraction.

Figures 6 and 7 show the interaction effects: of the initial oil concentration in solids with the initial oil concentration in solvent; of the initial oil concentration in solids with particle diameter, and of the initial oil concentration in solvent with particle diameter. Figure 6 depicts the results for the residual oil concentration in solids whereas Figure 7 in solvent.

Increasing simultaneously the initial oil concentration in solids and in solvent, the secondary effects are about -5% and -12%, in the first period, in oil residual concentration in solids and in solvent, respectively. Thus, this increase is worsened to the extraction. In the second period, this increase aids the extraction. One has +35% in oil residual concentration in solids and +54% in solvent.

When x_{0} and dp are increased, in the first period, the extraction is aided (+8% in solids and +15% in solvent). However, the extraction is worsened in the second period because the values of effects are -47% in solids and -67% in solvent.

The simultaneous increase in y_{0} and dp aids the extraction in the first period (2.5% in solids and 8% in solvent) and worsens in the second period (-24% in solids and -53% in solvent).

**CONCLUSIONS**

Figs. 1 and 2 show that all analysed parameters have influence on the solids and solvent concentration during all the first extraction period. In the second period, the particle diameter has influence on both concentrations and on the initial oil solvent concentration. In fact, they affect the oil residual concentration in solvent. The other parameters have null effects. Due to these verifications, it is possible to manipulate the solvent superficial velocity for compensate flotation in solid and solvent phases. Disturbances in particle dimensions are not compensated by superficial velocity manipulation, since this effect in second period is null. Therefore, it is necessary to determinate the optimum particle diameter and manipulate the solvent superficial velocity to control the process.

**ACKNOWLEDGEMENT**

The authors acknowledge the financial support from Capes.

**REFERENCES**

Chien, J.T. , Hoff, J.E. , Lee, M. J. , Lin, H.M. , Chen, Y.J. 7 Chen, L.F. (1990), "Oil Extraction of Ground Corn with Ethanol", The *Chem. Eng. Journal, *vol. 49, B102-B113. [ Links ]

Cuperus, F. P., Boswinkel, G., Muuse, B. G., Derksen, T. P. (1996), "Supercritical Carbon Dioxide Extraction of Dimorphotheca pluvialis Oil Seeds", *JAOCS*, vol. 73, No. 12, 1675-1679. [ Links ]

Cygnarowicz-Provost, M., OBrien, D. J., Maxwell, R. J., Hampson, J. W. (1992), "Supercritical Fluid Extraction of fungal lipids using mixed solvents: experiment and modeling", *J. Supercrit. Fluids*, vol. 5, 24-30. [ Links ]

del Valle, J. M., Aguilera, J. M. (1988), "An improved equation for predicting the solubility of vegetable oils in supercritical CO_{2}", *Ind. Engng Chem. Res.*, vol. 27, 1551-1553. [ Links ]

Karnofsky, G. (1949), "The Theory of Solvent Extraction", *JAOCS, *vol. 26, 564-569. [ Links ]

Karnofsky, G. (1986), "Design of Oilseed Extractors. I. Oil Extraction", *JAOCS, *vol. 63, No. 8, 1011-1013. [ Links ]

Lack, E. A. (1985), "Kriterien zur Auslegung von Anlagen für die Hochdruckextraktion von Naturstoffen", Ph.D. thesis, TU Graz. [ Links ]

Lee, A. K. K., Bulley, N. R., Fattori, M., Meisen, A. (1986), "Modelling of supercritical carbon dioxide extraction of canola oilseed in fixed beds", *JAOCS*, vol. 63, 921-925. [ Links ]

Nieh, C. D. ; Snyder, H. E. (1991), "Solvent Extraction of Oil from Soybean Flour I- Extraction Rate, a Countercurent Extraction System, and oil quality", *JAOCS, *vol. 68, No. 4, 246-249. [ Links ]

Othmer, D. F. ; Agarwal, J. C. (1955), "Extraction of Soybeans: Theory and Mechanism", *Chem. Eng. Progress, *vol. 51, No. 8, 372-378. [ Links ]

Pekhov, A. V., Goncharenko, G. K. (1968), "Ekstrakcija prjanogo rastitelnogo syria szhizhennymi gazami, *Maslozhirovaja promyshlennos*, vol. 34, No. 10, 26-29. [ Links ]

Sovová, H. (1994), "Rate of Vegetable Oil Extraction with Supercritical CO_{2} - I. Modelling of Extraction Curves", *JAOCS,* vol. 49, No. 3, 409-414. [ Links ]

Stastová, J., Jez, J., Bártlová, M., Sovová, H. (1996), "Rate of the Vegetable Oil Extraction with Supercritical CO_{2} – III. Extraction from Sea Buckthorn", *Chemical Engineering Science*, vol. 51, No. 18, 4347-4352. [ Links ]

*To whom correspondence should be addressed