Submitted to: Applied Engineering in Agriculture
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 3/1/2006
Publication Date: 5/1/2006
Citation: Armstrong, P.R., Maghirang, E.B., Xie, F., Dowell, F.E. 2006. Comparison of Dispersive and Fourier-Transform NIR Instruments for Measuring Grain and Flour Attributes.. Applied Engineering in Agriculture. 22(3): 453-457.
Interpretive Summary: Near-infrared light is absorbed and reflected differently by chemical compounds or constituents found in grain such as protein, starch and water. The concentration of these grain constituents, and others, can be accurately measured by instruments that use near-infrared reflectance (NIR) spectroscopy. NIR instruments measure reflectance across a broad spectrum of near-infrared light. The reflectance of near-infrared light at specific wavelengths is used to build statistical models to predict constituents concentration. The purpose of this study was to compare two types of instruments for their ability to predict concentrations of protein, moisture, and hardness of whole grain wheat; protein, ash, and amylose of wheat flour; and corn grit fat. The study used a Fourier transform-NIR spectrometer (FT-NIR) from Bruker Optics, Billerica, MA and a Model 6500 NIR from FOSS-NIR Systems, Inc., Silver Spring, MD. The FT-NIR instrument differs from the NIR instrument in the method of light spectra measurement. Wheat flour protein and ash; whole grain wheat protein and moisture were measured with excellent accuracy by both instruments while wheat flour amylose and whole grain wheat hardness measurement were less accurate. Corn grit fat measurement was poor for both instruments. Overall the FT-NIR and NIR instruments were essentially equal in measurement accuracy and there is no apparent advantages of one over the other.
Technical Abstract: Instruments using near-infrared reflectance (NIR) and Fourier transform near-infrared (FT-NIR) spectroscopic methods were compared for their predictive performance of several wheat flour and grain constituents. Protein, moisture, and hardness of whole grain wheat; protein, ash, and amylose of wheat flour; and corn grit fat were used to develop prediction equations between reference data of these constituents and their spectra. Partial Least Squares (PLS) regression was used to develop the prediction equations. NIR and FT-NIR spectrometers collected spectra over the wavelength ranges of 1100-2498 and 1142 - 2502 nm respectively. Prediction models were selected using F-test criteria ('=0.05). Results show that FT-NIR and NIR instruments were comparable in prediction performance and there is no apparent advantages of one over the other. Wheat flour protein and ash; whole grain wheat protein and moisture models had good quantitative prediction based on RPD values, i.e. RPD values were greater than 5. Wheat flour amylose and whole grain wheat hardness predictions were qualitative with RPD values near 3. Corn grit fat predictions were poor with RPD values near 1.