Moisture Determination of Static and In-Motion Powdered Infant Formula Utilising Multiprobe near Infrared Spectroscopy

Abstract

Moisture content of powdered infant formula is a critical factor governing product quality, especially in terms of its physicochemical stability. High levels of moisture accelerate the conversion of amorphous lactose to α-lactose monohydrate, which is the main cause of sticking and caking problems. Conversely, milk powders may become more susceptible to lipid oxidation at relatively low moisture levels. Traditionally, moisture content has commonly been determined by methods based on the loss of weight when drying under controlled conditions in an oven. However, these methods are time consuming and not suitable for in-line measurement. Near infrared (NIR) spectroscopy is a rapid method and perhaps the most used and accepted alternative technique for the measurement of moisture content in the dairy industry. Significant challenges in using NIR spectroscopy in the manufacturing process are in-line measurement and real-time monitoring. In this work, a novel multiprobe NIR system based on a Fabry–Perot interferometer combined with four fibre probes was assessed for predicting the moisture content of samples with varying moisture levels (ranging from ca 2% to 13%) and recorded under static conditions and various levels of motion. Partial least squares (PLS) regression was used to correlate the spectral response to the reference moisture values. The coefficient of determination of calibration ( R2) and root mean square error of cross validation for the best model were 0.99% and 0.57%, respectively. The PLS calibration model was then applied to an independent set of samples recorded under both conditions: static [root mean square error of prediction ( RMSEP) = 0.62%] and motion at 0.01 m s−1 ( RMSEP = 0.66%), 0.07 m s−1 (1.07%) and 0.16 m s−1 (0.97%). Therefore, moisture predictions for both measuring modes agreed well with the reference values, even when samples were travelling at speeds of 0.16 m s−1. These results demonstrated that the approach is suitable for in-line moisture analysis of powdered infant formula.