Assessment of NIR and Raman Spectroscopy as Analytical Tools to Predict Viscosity of Ice Cream Mixes

Ice cream is a complex product containing four different phases that affect its microstructure. Viscosity is a critical ice cream quality parameter which is typically measured using off-line methodologies, such as rheometry. In-line viscosity measurements allow continuous and instant analysis compared to off-line methodologies, yet they still constitute a challenge. This work focused on the preliminary study of the potential application of Near-Infrared (NIR) and Raman spectroscopy as analytical tools to assess the viscosity of ice cream mixes. Historically, partial least squares regression (PLS) is a standard algorithm used for analysis of spectral data and in the development of predictive models. This methodology was implemented over a range of viscosity values, obtained by varying the ice cream fat content and homogenization conditions. Individual PLS models showed some predictive ability and a better performance compared to the integrated model obtained by data fusion. Lower prediction errors and higher coefficients of determination were obtained for NIR, making this technique more suitable based on model performance. However, other considerations should be accounted during the selection of the best method, such as implementation limitations. This study offers a preliminary comparison of the spectroscopic methods for quantitative analysis of viscosity of aged ice cream mixes and a starting point for an in situ application study.