Statistical approaches for milk composition determination using combined near infrared, Raman, conductivity, and refractive index measurements

Current practices for routine milk composition determination employ commercial infrared systems. The use of SW-NIR and NIR FT-Raman spectra coupled with conductivity and refractive index could lead to more frequent and less costly analysis of fat, lactose and protein in milk.; The present study examines the potential of both SW-NIR absorbance spectrophotometry and NIR FT-Raman spectrophotometry to develop a model to estimate fat, lactose, and protein in whole milk of cows. To accomplish this, 79 milk standards, spanning the range of composition seen in practice, were obtained. Acquisition of NIR spectra over the wavelength range of 700 nm to 1018 nm was conducted. Between 0 and 3700 cm-1, NIR FT-Raman spectrophotometric measurements of the milk samples were made using a 1064 nm Nd: YAG laser source. Conductivity and refractive index measurements were also obtained for the milk standards.; A partial least squares calibration with leave-N-out cross validation was made using spectra with conductivity and refractive index to estimate fat, lactose and protein contents. Calibrations were developed using 75% of the milk standards. Models were further validated using an independent test set comprised of the remaining 25% of the data that had been excluded from calibration. A second calibration was conducted using a genetic algorithm approach. (Abstract shortened by UMI.)...