Figure 2 above is a plot of the calibration curve for analysis of fat in cheese by FT-NIR spectroscopy. The Predicted Error Sum of Squares (PRESS) plot shows an optimum pattern with high PRESS at low factors and low PRESS at higher numbers of factors (Figure 3 at right).
The second method quantifies dry matter in the blend, 479 standards with values between 31.36 and 53.65 percent wt. were used. The optimized method included the region from 4075 to 8909 cm-1 with a first derivative spectral pretreatment and a one-point baseline. The calibration yielded an RMSEC value of 0.645 and an RMSECV of 0.776. Figure 4 at bottom is a plot of correlation and a part of the calibration results. Again, with a calibration set of over 400 samples, the dry matter calibration shows an excellent correlation with the reference data. The PRESS plot for dry matter was also very reasonable.
Conclusions
FT-NIR can offer an ideal alternative to traditional methods for determination of fat and dry matter in processed cheeses during their manufacture. The user is able to obtain the data quick enough for control of (feedback to) the process of raw materials and additives blending in melting pots, resulting in saving raw materials and sustaining a constant quality of products. In addition, FT-NIR is a non-destructive technique that allows collection of a single spectrum for the prediction of multiple chemical or physical properties in a sample. This methodology could be extended to predict other critical factors in cheese processing, such as total protein or moisture content—obviating the need for more involved testing protocols.
Hirsch is manager of research & advanced development at Thermo Fisher Scientific in Madison, Wis. For questions, contact 608-276-6234. Tenkl and Hollein both work at Nicolet CZ ltd. in Prague, Czech Republic. Reach them at [email protected] and [email protected].
Leave a Reply