A novel portable analyzer for raw milk quality control during the material purchase at dairy plants was developed, by which the percentages（mass fraction） of main components including total protein, fat, and lactose...A novel portable analyzer for raw milk quality control during the material purchase at dairy plants was developed, by which the percentages（mass fraction） of main components including total protein, fat, and lactose of an unhomogenized milk sample could be determinated in 1 min with the help of non-dispersive short-wave near-infrared （NDSWNIR） spectrometry in a wavelength range from 600 nm to 1100 nm and multivariate calibration. The analyzer was designed with a single-beam optical system, which comprised a temperature control module, a multi-channel narrow-band light source（16 wavelengths）, a glass absorption cell with 15 mm sample thickness, a silicon photodiode detector, several compound lenses and a recorder module. A total of 80 raw milk samples were collected at a dairy farm twice a month for 4 months. The samples were scanned with a common UV-Vis-NIR spectrometer and analyzed according to China GB standard methods. The uninformative variables elimination（UVE） method was carried out on the spectrum data and the percentages of main components of all the samples to choose the peak emitting wavelength of each channel of the light source. Another 90 raw milk samples were collected from the same dairy farm thrice a month for 3 months. The samples were analyzed according to China GB standard methods and with the proposed analyzer. The percentages of the main components and the NDSWNIR absorption data of the samples were used for the construction and validation of the multivariate calibration model with partial least squares（PLS） method. The root-mean-square errors of prediction（RMSEP） of total protein, fat and lactose were 0.201, 0.172 and 0.247 and the coefficients of correlation（R） were 0.932, 0.981 and 0.933, respectively.展开更多
文摘A novel portable analyzer for raw milk quality control during the material purchase at dairy plants was developed, by which the percentages（mass fraction） of main components including total protein, fat, and lactose of an unhomogenized milk sample could be determinated in 1 min with the help of non-dispersive short-wave near-infrared （NDSWNIR） spectrometry in a wavelength range from 600 nm to 1100 nm and multivariate calibration. The analyzer was designed with a single-beam optical system, which comprised a temperature control module, a multi-channel narrow-band light source（16 wavelengths）, a glass absorption cell with 15 mm sample thickness, a silicon photodiode detector, several compound lenses and a recorder module. A total of 80 raw milk samples were collected at a dairy farm twice a month for 4 months. The samples were scanned with a common UV-Vis-NIR spectrometer and analyzed according to China GB standard methods. The uninformative variables elimination（UVE） method was carried out on the spectrum data and the percentages of main components of all the samples to choose the peak emitting wavelength of each channel of the light source. Another 90 raw milk samples were collected from the same dairy farm thrice a month for 3 months. The samples were analyzed according to China GB standard methods and with the proposed analyzer. The percentages of the main components and the NDSWNIR absorption data of the samples were used for the construction and validation of the multivariate calibration model with partial least squares（PLS） method. The root-mean-square errors of prediction（RMSEP） of total protein, fat and lactose were 0.201, 0.172 and 0.247 and the coefficients of correlation（R） were 0.932, 0.981 and 0.933, respectively.
