Quantification of macro-components in raw milk using micro NIR sensors

Near Infrared (NIR) spectroscopy is a rapid and proven method for the compositional analysis of raw milk; however, its usage is limited to laboratories due to instrumentation cost and bulkiness. The objective of this study was to investigate the potential of Fabry–Pérot Interferometer (FPI)-based micro NIR sensors to quantify macro-components in raw milk such as fat, protein, lactose, and total solids for in-situ analysis. An experimental prototype was designed to acquire the spectra of 250 raw milk samples in transmission mode using two micro NIR sensors. Calibration models were developed for macro-components of raw milk using partial least square (PLS) regression, and prediction performance was assessed using statistical metrics. It was observed that the sensor S-2.0 was able to quantify fat (RMSEP = 0.15 %), protein (RMSEP = 0.15 %) and total solids (RMSEP = 0.30 %). However, S-2.5 resulted in relatively lower prediction accuracy for fat (RMSEP = 0.35 %) and protein (RMSEP = 0.33 %), possibly due to the NIR region's lower penetration power where the sensor S-2.5 captures the response. The results showed that the micro/handheld NIR sensors could quantify certain macro-components (fat and protein) in raw milk, while they may not be suitable for other components (e.g., lactose). •The potential of micro NIR sensors to quantify raw milk macro-components.•Effect of a lower resolution and limited spectral range of micro NIR sensors.•Influence of NIR regions on penetration power and signal-to-noise ratio.•Obstacles in acquiring the reflectance spectra of raw milk using micro NIR sensors.