Understanding variability and calibration challenges in NIR miniaturized sensors: the impact of particle size and analytical session in almond powder analysis

The increasing application of miniaturized Near-Infrared (NIR) sensors highlights their potential for rapid, non-destructive, and cost-effective analysis, particularly in food industry. These portable instruments are often marketed as easy-to-use solutions, intended for use by non-specialists rather than analytical chemistry experts, which has contributed to their widespread adoption. This study investigates the contamination of bitter almond in almond powder using various low-cost miniaturized NIR sensors, including the SCiO sensor, two NeoSpectra Micro Development Kits, and the NeoSpectra Scanner, with and without the Rotator accessory. Almond powders with different levels of contamination of bitter almond (0–100 wt%) were analysed, and Principal Component Analysis (PCA) was used as an initial data screening step, showing the importance of particle size, thus providing a valuable quality control in this type of measurements. Partial Least Squares (PLS) regression models were developed to predict the percentage of contamination of bitter almonds and to evaluate the performance of each NIR sensor. The best regression models were obtained using the NeoSpectra Scanner spectrometrer being to predict concentration values with an error around 2.5% and a limit of detection around 4.5% of bitter almond in almond powder. Performance discrepancies were observed between sensors of the same type and model, as well as across different experimental sessions. These results emphasize the importance of understanding the limitations of miniaturized NIR sensors, while also highlighting their effectiveness, affordability, and portability, which make them a valuable and reliable tool for on-site food safety applications .