Monitoring the texture of high-moisture extrudates in the cooling die of an extruder using near-infrared spectroscopy

Abstract

High-moisture extrusion processing (HMEP) is used to produce high-moisture extrudates (HME) with fibrous textures that mimic animal meat. However, many factors affect the final product, and industries require in-line control tools to monitor and optimise the process. This study aims to evaluate the feasibility of using near-infrared spectroscopy (NIRS) to monitor HMEP in the cooling die of an extruder through the prediction of the textural properties of the final product. Different strategies to minimise the temperature effects over the spectra and different modelling approaches were evaluated. To do so, NIR spectra were acquired in the cooling die of the extruder during HMEP at different cooling die temperatures (10–30 °C) and flow rates (10.0, 12.5, 16, 19.5, and 22.0 g/min). Then, the moisture content and textural properties of the final HME were determined physicochemically. Various correction techniques were used to minimise the effects of temperature on the spectra and improve the in-line prediction accuracy of the extrudates' textural properties. Results showed that, with adequate preprocessing, the textural properties could be estimated using both partial least squares regression (PLSR) and principal component regression. Using PLSR models, the lowest predictive errors obtained were 0.87 N for transversal cut force, 9.15 N for hardness, and 7.90·10−3 for springiness. However, the data proved to be insufficient to train a convolutional neural network properly. Although more experimental work is needed, NIRS and chemometric techniques demonstrate potential for monitoring HMEP in the cooling die, enabling in-line optimisation of this process.