Analysis of the Performance of a Dehulling System for Confectionary Sunflower Seeds ()
1. Introduction
Sunflower is mainly cultivated for commercial oilseed production by pressing and/or solvent extraction. Although the non-oilseed variety (confectionary sunflower, of larger size and lower oil content) is grown to a lesser extent, it has a wide market because it is used for human consumption and in the food industry for birds and others animals. Confectionary sunflower is generally classified into three categories: the larger seeds are roasted, salted and packaged for human consumption; medium size seeds are dehulled and packaged for use as snacks or in bakery foods; and the smaller seeds are used as poultry feed.
Commercial processing of sunflower seeds requires achieving an efficient separation of the hull from the seed in the dehulling process. In the oil industry, partial dehulling of the seed presents many advantages, such as a suitable bulk porosity for the oil extraction process, better quality of both raw oil (lower wax content) and de-oiled meal (higher protein content), as well as an increase of the life span of the machinery [1] [2] . On the other hand, the size of confectionary sunflower seeds and the easiness to remove their hull are two properties that have an essential role in the quality of the obtained product. The characteristics of use of this seed require a dehulling process that is efficient in the removal of the hull and that also allows obtaining a product consisting mainly of whole kernels.
The most efficient method for the industrial processing of sunflower seeds is based on a combination of impact and centrifugal forces. During impact dehulling, seeds are fed into the top of the equipment and a spinning rotor expels them against the impact frame of the dehuller. The force of the impact causes the hull to break away from the kernel, and then the hull is removed by aspiration as the kernels are separated by vibratory sieves.
Many investigations have shown that the dehulling ability depends on different characteristics of the seeds, such as size and density [1] -[3] , hull content [4] [5] , oil content [3] [6] and moisture content [2] [3] [7] -[9] . Subramanian et al. [2] and Gupta and Das [10] analyzed the effect of feed rate and impeller speed on the dehulling performance of oilseed sunflower varieties grown in India. Previous studies investigated the effect of moisture content of the seeds on the dehulling ability of a non-oilseed sunflower hybrid [8] , but a search of the literature does not show any studies on confectionary sunflower involving other process variables or a combined study of those variables. Therefore, the aim of the present work was to determine an optimal combination of the operating conditions (moisture content of the seeds and dehuller impact speed) in the dehulling process in order to maximize the dehulling ability maintaining a high percentage of whole kernels.
2. Materials and Methods
2.1. Sample Preparation
The confectionary variety Mycogen 9338 (Morgan) was selected for this study. The sunflower seeds used were grown in Olavarría (36.5˚S, 60.5˚W), Buenos Aires province, Argentina. The seeds were manually cleaned to remove all foreign matter, broken or immature seeds. The moisture content of the seeds (MC) was determined according to the ASAE S352.2 Method [11] . Both the hull total content (H total) and the kernel total content (WKtotal) were determined by manual dehulling from a sample of 10 g and were expressed as dry basis percentage (%, d.b.). The desired moisture content was obtained by drying the seeds in a convection air oven at 40˚C - 45˚C or by spraying with pre-calculated amounts of distilled water, and then thoroughly mixing and sealing the samples in separate polyethylene bags. The samples were kept in a refrigerator for at least 72 h to allow a homogeneous moisture distribution. Before starting a test, the required amount of seeds was taken out of the refrigerator and allowed to equilibrate to ambient temperature.
2.2. Dehulling Ability and Whole Kernel Percentage
The mechanical dehulling was carried out by impact in a dehulling pilot equipment based on a centrifugal process (Figure 1). The rotor speed was adjusted with a variable frequency drive and calibrated with a tachometer. After the mechanical dehulling of the sample, the resulting product was classified into hull, whole kernels and rest (undehulled seeds, partially dehulled seeds, broken kernels and fines, i.e. material smaller than 2 mm). The fractions were dried in a forced-air oven at 130˚C for 3 h, and the mechanically-extracted hull/kernel percentages were calculated by Equations (1) and (2), respectively, expressed as percentages in dry basis.
(1)
(2)
The dehulling ability (DA) and the whole kernels production (WK) were determined as the percentage ratio of the percentage of mechanically extracted hull/whole kernels to the total amount of hull/kernels of the seeds, Equations (3) and (4), respectively.
(3)
(4)
2.3. Experimental Design and Statistical Analysis
The response surface methodology (RSM) combined with a central composite rotatable design (CCRD) were used to point out the relationship between the response functions and the process variables, as well as to determine the conditions of these variables able to optimize the performance of the dehulling system for confectionary sunflower seeds. The design consisted of 11 experiments with three center points and four axial points added to the 22 full factorial design (Table 1). All the experiments were conducted in triplicate. The impact dehuller rotor speed (rpm) was expressed in terms of the peripheral speed (PS) of the impeller (m/s).
STATGRAPHICS Centurion XV software (Version 15.2.06, StatPoint, Inc.) was used to perform the statistical analysis of the results, develop multiple regression models from experimental data, and predict process conditions that improve the performance of the dehulling system. Experimental data were fitted to a generalized second-order model by the following equation:
Table 1. Experimental program and experimental mean values of response variables for central composite rotatable design.
X1 (moisture content, % d.b.); X2 (peripheral speed, m/s) (Standard deviation in parentheses).
(5)
where Yi is one of the two predicted responses (dehulling ability of the seeds or the percentage of whole kernels) and X1, X2 represent the coded levels of the independent variables.
2.4. Optimization
In the present work, the optimization procedure was carried out by using the desirability function (D) combined with response surface methodology. Desirability optimization methodology is based on the idea that the quality of a product or process that has multiple characteristics, when one of them is outside of some “desired” limits, is completely unacceptable. The method finds operating conditions that provide the “most desirable” response values. The optimal values of the factors are determined from the maximization of the function. A high value of D, which varies between zero and one, indicates the best combinations of factors to optimize the system studied [12] -[14] . In the present study, D was developed for the maximum DA and maximum WK criteria.
3. Results and Discussion
3.1. Fitting Models
Results of different dehulling runs are shown in Table 1. Estimated values of regression coefficients and their corresponding p-values for the DA response are presented in Table 2. They demonstrate that moisture content of the seeds (X1) and peripheral speed (X2) had the largest effect on DA. These effects were followed by the quadratic term of peripheral speed (X2). The quadratic term of moisture content (X1) and the interaction between moisture content of the seeds and peripheral speed (X1X2) did not present a significant effect on DA (p > 0.05).
The estimated values of regression coefficients and their corresponding p-values for the WK response are shown in Table 3. All the coefficients were significant (p £ 0.05), except for the moisture content of the seeds/ impact speed interaction (X1X2).
The prediction models were rearranged by removing the terms that were not significant at 5% in the second-order polynomial models, and the corresponding equations are shown in Table 4.
ANOVA tests revealed that the quadratic polynomial models adequately represent the responses of DA and WK with coefficients of determination, R2 = 0.8282 and R2 = 0.9467, respectively (Table 4). R2 and the adjusted R2 values for the studied response variables were equal or higher than 0.80, hence there was a close agreement between the experimental results and the theoretical values predicted by the proposed models. The model adequacy was tested using the lack of fit test, which was not significant for p > 0.05. Since the models were found
Table 2. Regresion coefficients of the development model for the dehulling ability.
Table 3. Regression coefficients of the development model for the whole kernel percentage.
Table 4. Adjusted models with only significant coefficients.
aWant the selected model to have non-significant lack of fit (p > 0.05).
to show insignificant lack of fit (Table 4), the responses were sufficiently explained by the regression equations.
The coefficients in Table 4 are presented in terms of coded factors. Thus, the size of the coefficients could directly be related to the observed change in the response, enabling a straight comparison between the coefficients. Accordingly, the importance of each factor in the response could be evaluated. The coefficients of moisture content (X1) were, in absolute value, in the same order as those of peripheral speed (X2), indicating that both factors had the same impact on the considered response.
The linear terms of the factors showed an opposite effect on both considered responses. On the other hand, any change in one of the studied factors in order to improve one response had the opposite effect on the other one. For example, an increase in the peripheral speed improved the DA of the seeds, but reduced the WK percentage. Likewise, an increase in moisture content of the seeds in order to achieve a higher WK production reduced the ability of the seeds to dehull (i.e., the DA of the seeds). This situation makes it difficult to optimize the dehulling system.
The regression models developed were used for each response in order to determine the specified optimum conditions. These regression models are valid only in the selected experimental domain, which was determined taking into account some economic and operational considerations of the industry, and quality characteristics of the seeds.
In order to visualize the effect of the independent variables on both responses (DA and WK), response surface plots were generated from the developed models. Figures 2 and 3 show the effect of both the moisture content of the seeds and the peripheral speed on DA and WK, respectively, for Mycogen 9338 confectionary sunflower seeds.
DA of the seeds increased with the increase in PS for all the seed moisture content. The increase in Dehulling efficiency with increasing PS at any moisture content of the seeds was also observed for an oilseed sunflower variety (Morden) [10] , safflower seeds [9] and oat genotypes [15] . The highest DA at higher PS may be attributed to a higher discharge velocity of the seed and inducing a larger impact against the casing of the dehuller.
Decreasing MC caused an increase in DA on confectionary sunflower seeds in agreement with the reported for other sunflower varieties [8] and safflower seeds [9] .
On the other hand, PS had negative linear and quadratic effects on WK, while MC had a positive linear effect on WK. Therefore, WK was found to increase as both the MC increased and PS decreased. The higher WK at low MC may be attributed to the kernel becoming brittle and thus, it broken easier.
Figure 2. Effect of moisture content of the grains and peripheral speed on dehulling ability (DA).
Figure 3. Effect of moisture content of the grains and peripheral speed on percentage of whole kernel (WK).
3.2. Optimization
Optimization of the processing operations should be performed in order to obtain the best conditions for the dehulling of seeds, resulting in a superior quality product as well as maximizing throughput capacity.
By applying the desirability function method, the optimal operating conditions, obtained with a desirability value of 0.9460, were 12.3% d.b. for the moisture content of the seeds and a peripheral speed of 32.5 m/s for the dehuller equipment. This factor-level combination would achieve the maximum response values (72.6% and 63.2% for DA and WK, respectively) for Mycogen 9338 confectionary sunflower seeds.
3.3. Confirmatory Studies
Verification experiment was performed at the optimum conditions obtained from the above study. The experimental DA and WK at the optimal operating conditions were 74.15% ± 4.31% and 61.21% ± 1.90%, respectively. It was observed that experimental optimal values were reasonably close to the predicted values confirming the validity and adequacy of the predicted models.
4. Conclusions
The models developed in the present work could be used to predict the variation of DA, as a measure of the easiness of the seeds to dehull, and WK, in terms of both the moisture content of the seeds and impact speed for Mycogen 9338 confectionary sunflower seeds. The response surface methodology and desirability function method were used to determine the optimal processing conditions within the experimental region. The results of the optimization technique showed that the optimal conditions that maximize DA and WK could be obtained if the system is operated at a peripheral speed of 32.5 m/s with a seed moisture content close to 12.3% d.b. Under these conditions, the values of DA and WK would be 72.6% and 63%, respectively.
Taking into account that confectionary sunflower seeds should not be stored with a moisture content above 10% - 11%, the optimum moisture value obtained from the experimental data (12.3%, d.b.) determined the need to moisturize the seeds prior to the dehulling process. Hence, further studies of technical and economic feasibility may be necessary.
The results obtained offer criteria for the dehulling process of confectionary sunflower seeds, especially for the oil industry. This would help to optimize the process, and obtain products of better quality and higher profits.
Acknowledgements
The authors acknowledge the financial support from Universidad Nacional del Centro de la Provincia de Buenos Aires, Argentina.
Abbreviations
CCRD: Central composite rotatable design DA: Dehulling ability D: Desirability function Db: Dry basis MC: Moisture content of the seeds PS: Peripheral speed RSM: Response surface methodology
NOTES
*Corresponding author.