Determination of cow's milk fat content based on segmented dielectric relaxation parameters combined with characteristic variables

Rapid and accurate online detection of cow's milk composition content is significant. This study aimed to clarify the correlation between dielectric relaxation parameters (DRPs) and fat content and explore methods to improve milk fat content prediction accuracy with segmented DRPs. Dielectric spectra were collected from 270 milk samples and divided into two relaxation segments using Cole-Cole plots. The segmented DRPs were obtained by fitting the segmented dielectric spectra using a modified Debye model. The prediction models for milk fat content based on segmented DRPs were built, and their performance was evaluated using root mean square errors of prediction set (RMSEP) and residual prediction deviation (RPD). The results showed an enhanced correlation between segmented DRPs and milk fat content. The prediction accuracy of the models with the segmented DRPs was higher than that of the unsegmented DRPs. The milk fat content prediction model based on the combination of the selected DRPs and successive projections algorithm (SPA) had the best prediction accuracy, with an RPD of 4.4 and an RMSEP of 2.05 g·kg⁻¹. This study improved the prediction accuracy of milk fat based on the segmented DRPs combing with characteristic variables. This study provides a new method for accurately detecting complex systems such as cow's milk based on dielectric spectroscopy.