Physical properties and modeling for sunflower seeds
Keywords:sunflower seed, normal modeling, two-parameter log-normal modeling, two-parameter Weibull distribution modeling, physical properties
For designing the dehulling, separating, threshing, sizing and planting machines for sunflower, physical and mechanical properties of sunflower seeds should be known. In this work some physical properties of three varieties of sunflower seeds, distance between the adjacent seeds on the sunflower head (SH), length, width, thickness, mass of the individual seeds, 1000- seeds mass, and changing these parameters with location of seeds on SH were measured. Then shape properties, including geometric mean diameter, sphericity, surface area, projected area and volume of the seeds were calculated. Variations of the shape properties of the seeds on the SH were studied. Statistical indices for dimensional and shape parameters were calculated. For Mikhi, Sirena, and Songhori varieties, true and bulk densities, porosity, angle of repose on wood and galvanized surfaces were calculated by using standard methods in the moisture of 9.15, 5.26 and 5.62% (w.b.), respectively. The distribution of distance between adjacent seeds on SH was modeled by using three continuous statistical distributions namely Normal, two-parameter Log-normal and two-parameter Weibull distribution. Size and mass of seeds were modeled with two-parameter Weibull distribution. The parameters of the probability density functions (PDF) were estimated, then evaluated, and the predictive performances of the models were compared. Log likelihood goodness of fit test was used to test how well different PDFs work for prediction of the distance between seeds on sunflower head, size and mass of seeds.
The results for three varieties showed that when the distance between locations of the seed from center of the sunflower head increased, size, shape properties and mass of seed, increased, too. The values of true and bulk density, porosity and angle of repose on wood and galvanized surfaces for Mikhi variety were 497.500 kg/m3, 331.027 kg/m3, 33.46% , 25.08° and 22.23°, for Sirena were 580.368 kg/m3, 422.015 kg/m3, 27.28%, 26.80° and 23.86°, and for Songhori were 471.746 kg/m3, 319.346 kg/m3, 32.30%, 24.39° and 21.70° respectively. Modeling result for the distance between adjacent seeds on SH showed that, Log-normal distribution model fits the empirical probability density well and two-parameter Weibull distribution had worst performance for prediction. Also modeling result for the distance between adjacent seeds on showed that whenever Skewness and Kurtosis had negative value, Weibull distribution was best fit. Statistical analyses for dimensional properties and mass showed that in most cases, both Skewness and Kurtosis had negative values. Therefor for modeling dimensional properties and mass, Weibull distribution was used.
Keywords: sunflower seed, normal modeling, two-parameter log-normal modeling, two-parameter Weibull distribution modeling, physical properties