Detecting whey adulteration of powdered milk by analysis of volatile emissions using a MOS electronic nose

Abstract

The ongoing investigation into detecting fraudulent additions to milk powder is a collaborative effort among governmental agencies, industry, and academia. Recent advancements are emphasizing the utilization of fingerprinting methodologies, which involve characterizing the odor of samples using gas sensors and using chemometrics methods to detect “out-of-class” or adulterated samples. We developed an advanced method using an electronic nose (e-nose) equipped with 8 metal oxide semiconductor (MOS) sensors to detect whey adulteration in powdered milk by analyzing volatile emissions. We examined pure powdered milk adulterated with whey at six concentration levels (10%, 20%, 30%, 40%, and 50% v/v) in both dry and rehydrated forms. Statistical analyses, including Principal Component Analysis (PCA) and Artificial Neural Network (ANN), were employed to interpret the sensor output responses from the e-nose. The ANN analysis demonstrated a total variance of 85%, with only eight out of 180 samples (4.4%) being misclassified in detecting whey adulteration in powdered milk. The model achieved an impressive detection accuracy of 95.6%. Notably, sensors MQ9 and TGS822 exhibited the most robust responses to wet samples, while sensors MQ136 and TGS822 showed the highest reactivity to dry test samples. PCA analysis revealed that the first principal component (PC-1) accounted for 90% of the total variance, whereas PC-2 contributed only 4% to the variance. In summary, this article offers insights into the application of an e-nose portable device that enables non-invasive analysis, and it provides a promising tool for rapid quality assessment of commercial powdered milk.