ARS | Publication request: GRANULATION SENSING OF FIRST-BREAK GROUND WHEAT USING A NEAR-INFRARED REFLECTANCE SPECTROMETER: STUDIES WITH SOFT RED WINTER WHEATS

Submitted to: Journal of the Science of Food and Agriculture
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: September 30, 2002
Publication Date: January 8, 2003
Citation: Pasikatan, M. C., E. Haque, C. K. Spillman, J. L. Steele, and G. A. Milliken. 2003. Granulation sensing of first-break ground wheat using a near-infrared reflectance spectrometer: studies with soft red winter wheats. J. Sci. Food Agric. 83:151-157 (online 2003).

Interpretive Summary: A sensor for granulation could change roll gap settings automatically to follow changes in granulation of ground wheat, unlike the operator-fixed setting of present roller mills. A fully automated roller mill could help optimize flour extraction in flour milling systems. Previously, we studied the feasibility of developing a granulation sensor out of a near-infrared (NIR) reflectance spectrometer, using ground wheat from six wheat classes and hard red winter wheats. This time we studied ground wheat from soft red winter (SRW) wheat. Two sets of 35 wheat samples, representing seven SRW wheat cultivars, were ground independently using five roller mill gaps (0.38, 0.51, 0.63, 0.75, and 0.88 mm). NIR reflectance of one set was used to develop calibration to estimate granulation from spectral data of the other set. Granulation models based on partial least squares regression were developed with cumulative mass of size fractions as reference value. Different ways of treating the spectral data (log (1/R), baseline correction, unit area normalization, and derivatives) and subregions of the 400-1700 nm wavelength range were evaluated. Models that corrected for pathlength effects (those that used unit area normalization) predicted the bigger size fractions well. The model based on unit area normalization-first derivative predicted 34 out of 35 validation spectra with square root of the sum of squared differences between reference and predicted data of 3.53, 1.83, 1.43, and 1.30 for the >1041, >375, >240, and >136 mm size fractions, respectively. SRW wheat granulation models performed better than the previously reported models for six wheat classes owing to less variation in mass of each size fraction. However, SRW wheat flour has tendency to stick to the underside of sieves that affected the reference values. Thus the finest size fraction of these models did not perform as well as the HRW wheat models.

Technical Abstract: A near-infrared reflectance spectrometer, previously evaluated as a granulation sensor for first-break ground wheat from six-wheat classes and hard red winter (HRW) wheats, was further evaluated for soft red winter (SRW) wheats. Two sets of 35 wheat samples, representing seven cultivars of SRW wheat ground by an experimental roller mill at five roll gap settings (0.38, 0.51, 0.63, 0.75, and 0.88 mm), were used for calibration and validation. Partial least squares regression was used to develop the granulation models using combinations of four data pretreatments (log (1/R), baseline correction, unit area normalization, and derivatives) and subregions of the 400-1700 nm wavelength range. Cumulative mass of size fraction was used as reference value. Models that corrected for pathlength effects (those that used unit area normalization) predicted the bigger size fractions well. The model based on unit area normalization-first derivative predicted 34 out of 35 validation spectra with standard errors of prediction of 3.53, 1.83, 1.43, and 1.30 for the >1041, >375, >240, and >136 mm size fractions, respectively. Because of less variation in mass of each size fraction, SRW wheat granulation models performed better than the previously reported models for six wheat classes. However, because of SRW wheat flour's tendency to stick to the underside of sieves, the finest size fraction of these models did not perform as well as the HRW wheat models.