VISIBLE LIGHT SCATTER AS A QUANTITATIVE INFORMATION SOURCE ON MILK CONSTITUENTS A. Melenteva 1, S. Kucheryavski 2, A. Bogomolov 1,31Samara State Technical University, Molodogvardeyskaya Street 244, 443100 Samara, Russia. 2Aalborg University, campus Esbjerg, Niels Bohrs vej 8, 6700 Esbjerg, Denmark. 3J&M Analytik AG, Willy-Messerschmitt-Strasse 8, 73457 Essingen, Germany. email@example.com Fat and protein are two major milk nutrients that are routinely analyzed in the dairy industry. Growing food quality requirements promote the dissemination of spectroscopic analysis, enabling real-time monitoring of processes and products. Optical analysis is generally performed in near and middle infrared (NIR and MIR) regions and relies on the component absorbance and Beer’s Law. The light scatter effect is therefore considered as a disturbance to be avoided during the measurement or eliminated at the data analysis stage. The region of visible (Vis) light (400-800 nm) is economically attractive, because it offers a range of inexpensive light sources, optics and detectors. At present, however, it is commonly considered useless, because of the light scatter by fat globules (1-10 μm) and protein micelles (80-200 μm) that strongly dominates; therefore, making the classical absorbance analysis hardly applicable. At the same time, diffused light by itself delivers information on the milk composition, specifically, fat content as illustrated in Fig. 1, and can potentially be used for the quantitative analysis. The main task here is to extract individual quantitative information on milk fat and total protein content from spectral data. This is particularly challenging problem in the case of raw natural milk, where the fat globule sizes may essentially differ depending on source. Fig. 1. Spots of light transmitted through homogenized milk samples with different fat content. The preceding research  has shown that individual scatter patterns of fat and protein in non-homogenized milk can be distinguished, thus, enabling their quantitative multivariate analysis. In the present study, a representative designed set of raw milk samples with simultaneously varying fat, total protein and particle size distribution has been analyzed in the Vis spectral region. The feasibility of raw milk analysis by PLS regression on spectral data has been proved. The root mean-square errors below 0.10% and 0.04% for fat and protein, respectively, have been obtained. PLS components were interpreted in terms of captured information. The results obtained provide a basis for the replacement of traditional spectroscopy by custom optical analyzers, optimized for the purpose of milk analysis. Preliminary achievements in this new research area are presented and discussed. References:  A. Bogomolov, S. Dietrich, B. Boldrini, R.W. Kessler, Food Chemistry (2012), doi:10.1016/j.foodchem.2012.02.077.
Xiii International Conference on Chemometrics in Analytical Chemistry, 2012