|Barton Ii, Franklin|
Submitted to: Federation of Analytical Chemistry and Spectroscopy Societies Final Program
Publication Type: Abstract Only
Publication Acceptance Date: October 27, 1997
Publication Date: N/A
Interpretive Summary: Interpretive summary not required for literature review, book chapter (unless reporting original research), book review, or oral presentation (where only the title or abstract is published). D152.1
Technical Abstract: The study aims to develop cereal food NIR reflectance methods which are unaffected by particle size and moisture content. About 150 unground commercial cereal products were examined, including breakfast cereals, snack foods, whole grains, flours and baking mixes. A portion of the samples were treated to introduce large variance in particle size and moisture content, yielding a total of approximately 350 specimens. NIR reflectance from 400-1700nm was measured for each specimen using a hopper device for rapid loading of multiple repacks for scanning by a fast diode-array spectrometer with a large illumination area (3"). The breakfast cereals portion of the specimens (n=290) spanned large ranges of moisture content (1.4-20.1%); protein content (3.9-21.5%); and dietary fiber content (0.52-49.7%). Particle sizes ranged from Nabisco's Cream of Wheat to Kelloggs' Frosted Mini-Wheats . We developed preliminary partial lleast squares regression models using first derivative spectra in the rang 850-1700nm. Moisture response was markedly nonlinear in the first few factors, but a 7 factor model yielded a root mean-squared cross-validation error of 0.7%. Protein error was 1.2% for an 11 factor NIR model. The error of NIR determination of dietary fiber was 2.5%. For this result, the reference was a powder NIR reflectance method with an error of 1.7% relative to the A.O.A.C. enzymatic-gravimetric method, which had a precision of 0.7%. The presentation will outline further refinement of the reference data set and chemometric analyses, with discussion of the effects of sample class, optical sampling, particle size, and moisture content.