Prediction of titratable acidity of sour cream by Fourier-transform infrared spectroscopy

Abstract

Fourier-transform infrared spectroscopy has become an increasingly valuable analytical tool in the dairy industry because it is fast, chemical-free, and can capture complex chemical information. However, predictive models for many cultured and formulated dairy products remain underdeveloped. This work evaluated the feasibility of using Fourier-transform infrared spectroscopy combined with chemometric modeling to predict an important quality attribute: titratable acidity of sour cream. This attribute is critical for product quality, regulatory compliance, and process efficiency, yet current reference or chemistry methods are labor-intensive, destructive, and limited in throughput. Commercial sour cream products were sampled weekly over 9 wk to capture batch-to-batch variability. Titratable acidity was determined by standard NaOH titration, and spectra were collected with the FOSS MilkoScan FT3. Partial least squares regression model achieved training and internal validation R2 values of 0.601 and 0.581, respectively, demonstrating that the spectra contain meaningful information about acid concentration. External validation yielded an R2 of 0.463, indicating that the model captured the general relationship between spectra and measured values, but was limited by formulation differences and sample diversity. As proof of concept, these findings suggest that expanded datasets and formulation-specific modeling strategies are needed for robust industrial applications.

Abstract as supplied by the indexing source. © the authors / publisher.