Assessment of 'Golden Delicious' Apples Using an Electronic Nose and Machine Learning to Determine Ripening Stages

Area of Science:

• Food Science
• Analytical Chemistry
• Sensory Science

Background:

• Consumer purchasing decisions are heavily influenced by apple appearance, often lacking objective quality indicators.
• Electronic noses offer a non-destructive method to assess fruit quality by detecting Volatile Organic Compounds (VOCs).

Purpose of the Study:

• To evaluate the effectiveness of an electronic nose system for classifying the ripening stages of 'Golden Delicious' apples.
• To determine the optimal sensor combination and data analysis strategy for accurate apple ripeness assessment.

Main Methods:

• Utilized four Metal Oxide Semiconductor (MOS) sensors (MQ3, MQ135, MQ136, MQ138) to collect VOC data from apples over two years.
• Applied Principal Component Analysis (PCA) and K-means clustering for initial ripening stage identification.
• Employed the K-Nearest Neighbors (KNN) model for classification, testing on various datasets including independent experiments.

Main Results:

• Successful classification of apple ripening stages (less ripe, ripe, overripe) with accuracies exceeding 75% on specific datasets.
• Achieved 100% accuracy in classifying apple ripeness using data from all experiments and independent test sets.
• Correlation and PCA analyses indicated that using two or three sensors could yield comparable results to using all four.

Conclusions:

• Electronic nose technology provides a highly accurate and reliable method for non-destructively assessing apple ripeness.
• Analyzing data from multiple experiments over extended periods and considering seasonal VOC variations enhances predictive accuracy.
• Optimizing sensor selection and data processing is key to maximizing the performance of e-nose systems in fruit quality assessment.