Customized e-nose sensor array configuration and VOCs pattern analysis for Cabernet Sauvignon grape post-harvest quality monitoring

Abstract

Cabernet Sauvignon grape, a globally renowned red wine grape variety, is valued for its distinctive aromatic profile. And this makes quality maintenance during post-harvest storage, transportation, and winemaking processes critical. Traditional quality assessment methods, which rely on expert sensory evaluation, are inherently subjective and impractical for large-scale operations. This study innovatively applies electronic nose (e-nose) technology, which uses clustering analysis of volatile organic compounds (VOCs) emitted by grapes, to monitor the quality states of Cabernet Sauvignon grapes. A customized sensor array made up of eight specialized sensors was developed to capture unique VOCs signatures across different quality stages. Principal Component Analysis (PCA) reduced the data volume by 50% while preserving 97% variance, streamlining downstream analyses. Clustering analysis was performed on the processed e-nose data. Its validity was confirmed by Rand Index (RI = 0.8336) and Fowlkes–Mallows Index (FMI = 0.7831), both exceeding empirically established reliability thresholds (RI $>$ 0.8; FMI $>$ 0.75). The patterns identified by clustering were then used to construct a VOC prediction model, and achieved an overall accuracy of 0.9583 and weighted performance metrics of precision = 0.9632, recall = 0.9583, and F1-score = 0.9599. This demonstrates the method’s effectiveness in identifying and predicting VOC patterns. From the cluster analysis results, it was found that the e-nose showed superior sensitivity to grape condition changes compared to human senses. The early warning rules defined by clustering outcomes triggered 100% successful alerts prior to observable quality state transitions. This early warning capability provides actionable intervention windows before irreversible deterioration. This work advances Cabernet Sauvignon grapes post-harvest quality monitoring and offers broader implications for quality control in perishable agricultural supply chains.