AACC International AACC International



doi:10.1094/CFW-55-6-0293 |  VIEW ARTICLE

AACC International Report

A New AACC International Approved Method for Assessment of a Calibration from a Near Infrared Spectrometer

S. R. Delwiche (1) and S. Millar (2). (1) USDA ARS, Beltsville, MD, U.S.A.; (2) Campden BRI, Chipping Campden, Gloucestershire, United Kingdom. Cereal Foods World 55(6):293-296.

Near-infrared (NIR) spectroscopy is used throughout the cereals and food industries, as well as other agricultural, pharmaceutical, and petrochemical industries, primarily because of its simplicity of sample preparation, speed, and accuracy. Arguably, NIR methodologies have become the preferred means to measure quality at all handling and processing stages of grain, from instream monitors on combine harvesters, through elevator storage operations, and on to milling, extraction, and baking and cooking stages. Successful applications include the measurement of moisture, protein, oil, hardness, ash, and starch components in wheat, corn, barley, and rice. Essential to the operation of any NIR technique, aside from the equipment itself, is the underlying calibration, which relates the optical response from the instrument to a concentration of a constituent (e.g., protein content, moisture content). In almost all cases, the NIR procedure is developed as a secondary method in which a calibration is developed through the relationship of the optical instrument response to concentration values as measured by a conventional wet chemical procedure, such as Kjeldahl or combustion for protein content. When such a calibration is initially developed, the NIR practitioner will have documented the performance characteristics in terms of overall error (SEC = standard error of calibration, SECV = standard error of cross-validation, RMSD = root mean square error of differences) and goodness of fit (R(^2) of the calibration equation developed from using various statistical methodologies, such as multiple linear regression, partial least squares regression, artificial neural networks, or support vector machines). Regardless of the statistical method employed during calibration development, periodically, it is necessary to ensure that a calibration is still in compliance with the expected level of performance. Changes to the instrument (e.g., lamp change) or the samples themselves may also necessitate an assessment. One of the traditional procedures for performing this task involves the application of the calibration equation to a set of well-characterized samples (3), as described herein. Elaboration of the statistical testing procedure, including a worked example, are provided in a new Approved Method (39-01.01) of AACC International (2010).

 

 

© Copyright AACC International  | Contact Us - Report a Bad Link