SUPERCRITICAL FLUID EXTRACTION AND MODELING OF ARGEMONE MEXICANA AND PONGAMIA PINNATA SEED OILS AND THEIR ANALYSIS

Abstract

Some natural seeds contain valuable chemical components, which are used for medicinal purposes to cure various chronic diseases. However, extracting these chemicals in pure natural form is a challenge. This problem can be somewhat resolved by supercritical fluid extraction (SFE) technique. In the present work, two seeds, namely Argemone mexicana (L.) (AM) and Pongamia pinnata (L.) (PP) are selected for extraction. This selection is purely based on the properties and applications of these two seed oils found in literature. A thorough literature review indicates that AM and PP seeds contains good percentage of oil, which contain some valuable chemical components such as fatty acids and alkaloids that can be used as medicine to cure various chronic deceases. Besides, the medicinal requirements, these oils also have a potential to produce biofuel. For commercial purposes, oils from these two seeds, are generally extracted by means of conventional extraction techniques such as soxhlet extraction, mechanical pressing, percolation methods, etc. However, due to the lower yield, loss of valuable components, refining hurdles and safety aspects make these conventional methods cost intensive and environmentally incompatible. Furthermore, the oil obtained through these methods, needs to be processed for cleaning and removal of organic solvents (e.g. n-hexane, methanol, isopropanol etc.) which are used during conventional extraction techniques (e.g. soxhlet extraction, percolation method). Even the added processing of the oils, do not remove organic solvents completely from the final product (oil). Therefore, the use of supercritical fluids (SFs), as an alternative solvent, during the extraction of seed oils has been attracting widespread attention due to their advantageous properties such as liquid-like density, gaslike viscosity and negligible surface tension. This new method (e.g. SFE) also meets environmental regulations and promotes its utilization as green method. The SFE, using supercritical carbon dioxide (SC-CO2) has also been recognized as the latest emerging eco-friendly and clean separation technology for various valuable food ingredients from natural products, pharmaceutical products, etc., with high yield and better quality products under a wide range of operating conditions. Among various SFs, SC-CO2 has achieved the unique popularity due to its non-flammability, -non-toxicity, -non-explosive behavior, -low cost, -availability in plenty amount and -easy to separate out from the extracted product. In the present experimental investigation, oils are extracted from AM and PP seeds by means of SC-CO2. The SFE experiments have been performed using SFE 1000F apparatus, supplied by ii Thar Technologies Inc., Pittsburgh. The software, ‘Design Expert 10.0’ has been used to design the number of experiments in a methodological way that actually help to develop a statistical analysis based correlation between input and output variables. The ANOVA (analysis of variance) has been performed to establish the relative significance of the individual input parameters (e.g. temperature, pressure, particle size, flow rate-CO2 and the % of co-solvent) and their interaction effects. For the present SFE process, the above five input parameters of significance have been screened out from thirteen influencing input parameters from the available published literature (as listed in section 2.3), and finally the effect of these input parameters on the cumulative extraction yield (CEY) of oil have been identified. Final five selected SFE operating parameters were optimized for maximum yield via response surface methodology (RSM) with a ‘five-factors-three-levels’ Box- Behnken design (BBD). The regression analysis of the experimental data of SFE process for AM and PP seed oils, has confirmed that the quadratic model (second-order polynomial equation) is the best for the prediction of CEY. This quadratic model offers the R2 values as 0.9737 and 0.9944 respectively for AM and PP seeds. Further, the statistical analysis has also shown that the interactions amongst all the five input parameters exist. Further, for both seeds, % of co-solvent and pressure have been observed to be the most influencing parameters followed by each one. The prediction of CEY values, through quadratic model, have been found within error range of ‘+14.4 to -11.28 %’, and ‘+18.39 to -16.32 %’ of experimental values for AM and PP seed oils respectively. Furthermore, an artificial neural network (ANN) model has been developed, using experimental data and for this, a trainable, feed-forward-back-propagation (FFBP) network has been used to predict the CEY of both seed oils with an acceptable level of accuracy. From a number of performing ANN models, an optimized model ‘FFBPN [5-6-1]’ has been selected for prediction of the CEY. The ‘FFBPN [5-6-1]’ shows average absolute relative deviation percentage (AARD %), mean square error (MSE)) and best regression coefficient (R2) for the SFE of AM and PP seed oils as ‘3.33 %, 0.0038, 0.9835’ and ‘4.39 %, 0.00051, 0.9874’ respectively. The present SFE process has been modeled by applying two mass transfer phenomenon based models (MTPBMs) (e.g. Sovova, 1994 and Reverchon, 1996). The fitting parameters (e.g. Z, W and xk) of the Sovova model (Sovova, 1994) have been optimized using the global optimization technique (genetic algorithm (GA) approach) by running a written program in MATLAB to minimize the iii AARD % between the CEY’s, obtained from the Sovova model and from the experiments during the SFE of AM and PP seed oils. On the other hand, the Reverchon model (Reverchon, 1996) is also solved and fitted with the present SFE experimental data using the COMSOL Multiphysics 5.3a. It is found that the Sovova model (SM) has produced excellent fitting with the experimental CEY data within an error band of ‘1.436 to 14.198 %’ and ‘0.7706 to 14.17 %’ for AM and PP seed oils respectively. From the results of SM, it has also been observed that the mass transfer coefficients in solvent phase (������������) are larger than the mass transfer coefficients in the solid phase (������������) for all parametric conditions. In comparison to SM, Reverchon model (RM) has produced partially good fittings with experimental data within an error band of ‘5.52 to 96.3 %’ and ‘2.64 to 19.74 %’ for AM and PP seed oils respectively. It is noted that RM is good for predicting the final value of CEY and not the intermediate CEY value with time. After the extraction of oils from AM and PP seeds, the physico-chemical properties (e.g. calorific value (MJ/kg), flash point (ºC), fire point (ºC), cloud point (ºC), pour point (ºC), acid value (mg KOH/g), peroxide value (meq/kg sample), and saponification value (mg KOH/g)) of oil samples have been determined. The chemical compositions of seed oils, obtained through both extraction methods (soxhlet and SFE methods), were analyzed and quantified using gas chromatography (GC). In addition to these, the characterization of feed (seeds) are also performed by scanning electron microscopy (SEM), Thermo-gravimetric (TG) analysis and Fourier transform infrared (FTIR) spectroscopy. The characterizations (e.g. total oil content (%), moisture (%), ash (%), SEM, TG analysis and FTIR analyses) were also performed on the seed particles of AM and PP seeds. The total oil content, moisture content and ash content are found as ‘42%, 9.6% & 3.5%’ and ‘36%, 5.4% & 1.8%’ for AM and PP seed particles respectively. The FTIR analyses of both seed particles have confirmed the presence of fatty acids, alkaloids, and protein. TG analyses has also confirmed the moisture and ash content. The morphological changes at the surfaces of AM and PP seeds have also been studied during the SFE process through SEM analysis. On the other hand, GC analysis of the product (oil) shows that AM seed oil is rich in linoleic acid (C18:2n6c), oleic acid (C18:1n9c), palmitic acid (C16:0), and stearic acid (C18:0), with the range of concentrations (weight %) as ‘22.54 - 59.07 %’, ‘25.01 - 41.46%’, ‘11.40 - 23.58 %’, and ‘2.98 – 5.97 %’ respectively. Whereas, the PP seed oil is found rich in oleic acid (C18:1n9c), arachidic acid (C20:0), cis-10-pentadecanoic acid iv (C15:1), stearic acid (C18:0), cis-8,11,14-Eicosatrienoic acid (C20:3n6), linolenic acid (C18:3n3), gamma(ϒ)-linolenic acid (C18:3n6) and cis-11-Eicosenoic acid (C20:1) and are found in the range of concentrations (weight %) as ‘45.42 - 58.62%’, ‘15.34 - 18.02%’, ‘8.64 - 11.95%’, ‘5.74 - 7.04%’, ‘2.62 - 4.45%’, ‘1.24 - 4.01%’, ‘0.34 - 1.53%’, and ‘0.0 - 3.93%’ respectively. Further, the physicochemical properties (e.g. heating value (MJ/kg), flash point, (°C), fire point (°C), cloud point (°C), pour point (°C), saponification value (mg KOH/g), peroxide value (meq/kg sample) and acid value (mg KOH/g)) of the extracted AM and PP seed oils during SFE, suggest that the oils could also be used for bio-fuel. Finally, the economic analysis of the SFE process at industrial scale (for 1000 liters capacity of extraction vessel) has confirmed the economic feasibility, based on the obtained payback periods as ‘1.63’ and ‘1.84’ years for the envisaged plant of SFE process of AM and PP seed oils respectively.