Correcting NIR spectra of dimethyl fumarate in milk measured for different brands and in different dates

The calibration transfer was used to correct near-infrared spectra (NIR) of dimethyl fumarate (DMF) for their variations resulting from differences of sample brand and measurement date. Because the concentration of DMF in milk is very small, an enrichment method called fluidized bed enrichment was used to concentrate the analyte before NIR measurement. For correcting spectra between different brands and measurement dates, the spectra of Bright dairy measured on 2/5/2011 were assigned as primary spectra. And the spectra of Yili dairy measured on 2/5/2011 and Bright dairy measured on 17/5/2011 were set as secondary spectra for changes in brand and measurement date, respectively. Two spectral pretreatment methods were applied to preprocess spectra: multiple scattering correction (MSC) and the combination of MSC and differentiation (Savitzky–Golay differentiation). After spectral pretreatment, two calibration transfer methods including canonical correlation analysis (CCA) and piecewise direct standardization were used to correct the secondary spectra. The results showed that although the variations in both brand and measurement date can cause secondary spectra to deviate from primary spectra, the CCA under the combination of MSC and differentiation can correct the secondary spectra to use the model of primary spectra with high accuracy.