Page Banner

United States Department of Agriculture

Agricultural Research Service

Research Project: THE ADVANCEMENT OF SPECTROSCOPIC SENSORS/CHEMOMETRIC ANALYSIS/BIOBASED PRODUCTS FOR QUALITY ASSESSMENT OF FIBER, GRAIN, AND FOOD COMMODITIES

Location: Quality and Safety Assessment Research Unit

Title: Rapid characterization of poultry feed composition by infrared spectral analysis

Author
item Holser, Ronald

Submitted to: European Poultry Conference Proceedings
Publication Type: Abstract Only
Publication Acceptance Date: May 27, 2010
Publication Date: August 23, 2010
Citation: Holser, R.A. 2010. Rapid characterization of poultry feed composition by infrared spectral analysis; Abstract #161-163. Thirteenth European Poultry Conference, 23-27 August, 2010, Tours, France.

Interpretive Summary: Measurements of the major components in poultry feed formulations were performed using a combination of near infrared and mid-infrared spectral analysis. This provides a rapid and non-destructive technique for quality control purposes at the feed mill and for quality assurance after storage. Commercial samples of starter, grower, and finisher feeds were obtained and analyzed using a PerkinElmer Spectrum 400 system. The instrument was equipped with dual sources and configured for attenuated total reflectance (ATR) measurements over the mid-infrared range and diffuse reflectance measurements over the near-infrared range. Spectra were analyzed by a 2-dimensional correlation technique to identify the spectral regions associated with compositional changes that could be used to develop predictive quantitative models. Spectra were processed with Grams/AI 8.0 (ThermoElectron Corp.) and Unscrambler (CAMO Software, Inc.). Each spectrum was the sum of 128 co-added scans. The array basic program 2Dgen was used with Grams/AI to produce synchronous and asynchronous correlation maps. These maps indicated the spectral regions of strongest correlation. These results were used to generate models based on partial least squares methods. Models were developed from principal component analysis (PCA) and multivariate curve resolution (MCR). The PCA model used three principal components while the MCR model was implemented with the non-negativity, closure, and unimodality constraints. The optimized models were tested with meal samples prepared in the laboratory and shown to predict fat, moisture, and protein content of poultry meals with less than 0.5% error. The advantages of this approach include minimal time required for sample preparation and analysis, acceptable accuracy and reproducibility, simple interpretation of model results, and potential for on-line or automated analysis. This technique is also applicable to portable or handheld infrared systems that are now available.

Technical Abstract: NA

Last Modified: 8/22/2014
Footer Content Back to Top of Page