Use of near-infrared spectroscopy for analysis of mixed rice bran and parboiled rice bran

Both rice bran and parboiled rice bran can be used for rice bran oil production. Parboiled rice is more desirable than rice bran due to its higher oil content and lower acid value compared to rice bran, hence higher price. The difference in price of raw materials may motivate suppliers to mix raw rice bran in parboiled rice bran. As a result, rice bran oil manufacturers frequently come across parboiled rice bran samples that have higher acid value and lower oil content. Currently, Near Infrared Spectroscopy (NIR) is widely used in rice bran oil production plant as a rapid and non-destructive method to analyze food components. At present, there are separate NIR databases for rice bran and parboiled rice bran. None of these two database can be used to accurately read or analyze the properties of mixed rice bran. Therefore, this research aimed to develop the NIR database of mixed rice bran. The research recruited 10 lots of rice bran and parboiled rice bran from the mills in Chai Nat, Phichit, Nakhon Sawan, Phitsanulok, Kamphaengphet and Nakhon Pathom. The mixed rice brans were prepared by mixing rice bran and parboiled rice bran in the 0: 10 to 10: 0 ratio. The properties of mixed rice bran were determined by using NIR spectroscopy, analyses of chemical composition and physical properties, e.g. the content of moisture, protein, fat, crude fiber, phytic acid, acid value, bulk density (tapped), bulk density (untapped), and color (whiteness index). The result showed that parboiled rice has higher protein and crude fiber content than rice bran, whereas rice bran had higher moisture content, acid value, phytic acid, whiteness index, and bulk density (p <0.05). Measurement of NIRS properties of the samples was carried out in the range of 12,500 cm-1 - 3,600 cm-1 in diffuse reflection mode.  The relationship between NIR spectrum and chemical - physical properties of the samples were done by using Partial least square (PLS) regression, with the sample data for calibration, internal validation, and external validation of 165, 55, and 60, respectively. The results showed that raw NIR spectral data could be used to correlate with fat content and bulk density (tapped) with the coefficient of determination (R2) of 0.98 and 0.91, respectively, and root mean square error (RMSE) of 0.36 and 0.01, respectively. Moisture content, fiber, and whiteness index were correlated well with the pretreated spectra using normalization method with an R2 of 0.91, 0.91, and 0.91, and RMSE of 0.13, 0.36, and 0.48, respectively. Acid value and bulk density (untapped) were correlated with the pretreated spectra using baseline method with an R2 of 0.94 and 0.92, and RMSE of 1.53 and 0.01, respectively. Protein and phytic acid were correlated with the pretreated spectra using smoothing moving average and smoothing median filter method, respectively, with an R2 of 0.92 and 0.96, and RMSE of 0.20 and 0.15, respectively. Internal validation showed that moisture content, protein, fat, fiber, phytic acid, acid value, bulk density (tapped), bulk density (untapped), and whiteness index could be predicted using NIRS readings with an R2 of 0.87 to 0.97, where the R2 for bulk density (tapped) prediction was the lowest (0.87) and the R2 for fat prediction was the highest (0.97). Further confirmation using external validation samples showed that NIRS could be used to predict the properties of mixed rice bran in the descending order for R2; fat content (0.97), phytic acid (0.89), bulk density tapped (0.87), fiber content (0.86), whiteness index (0.80), protein content (0.78), and bulk density untapped (0.75).