Akademik Veri Yönetim Sistemi
Final SensorFINT International Conference, Cordoba, İspanya, 29 - 31 Mayıs 2024, ss.133, (Özet Bildiri)
Ensuring the quality of milk powder is crucial due to its nutritional importance. This study investigates the use of Near-Infrared (NIR) spectroscopy as a nondestructive, rapid alternative for the classification of milk powders under various relative humidity conditions. Samples, including whole, skim milk, and lactose-free milk powders, were analyzed using both a desktop (NIRFlex N-500) and a portable (NIR-S-G1) NIR spectrometer. Advanced chemometric methods, such as mean centering, baseline correction, Gaussian smoothing, and derivative processing, were applied to the raw NIR spectra using Orange data analysis software (Demšar et al., 2013). The study revealed that whole milk powder samples with different relative humidity rates showed distinct NIR spectra at 1455 nm and 1927 nm, corresponding to water absorption. High relative humidity samples exhibited higher absorbance values. Principal Component Analysis (PCA) was employed to explore sample differences, with k-Nearest Neighbors (kNN) and Support Vector Machine (SVM) models achieving 100% classification accuracy based on relative humidity. The portable NIR device demonstrated high efficiency for the project due to its fast data collection, ease of transportation, cost-effectiveness, and robust classification ability. Our findings affirm that NIR spectroscopy combined with robust chemometric techniques provides a reliable, efficient method for nondestructive moisture content determination and classification in milk powder, enhancing real-time quality control and product standardization.