Cangemi, Silvana (2025) Definizione di profili metabolici per la certificazione di qualità e di origine di prodotti agroalimentari e loro tracciabilità attraverso intelligenza artificiale e tecnologia BlockChain. [Tesi di dottorato]

Documento PDF TESI_DOTTORATO_CANGEMI_SILVANA_DR995914.pdf Download (5MB)

Abstract

The increasing demand for reliable tools for authentication, quality certification, and traceability of agri-food products has highlighted the need for advanced analytical methodologies capable of addressing the complexity of modern supply chains. In this context, the present thesis proposes an integrated approach based on the combination of NMR spectroscopy (HR-MAS and CP-MAS), artificial intelligence (AI) techniques, and blockchain technologies, with the aim of defining characteristic and validated metabolic profiles to support objective and automated certification systems. The analysis was conducted on various agri-food matrices (wheat, UHT milk, tomato concentrate, and wood chips), selected to represent different levels of industrial processing and chemical-physical complexity. Although HR-MAS and CP-MAS differ in spectral resolution, both techniques allow for direct analysis of the sample without solvent extraction, thereby preserving the metabolic integrity. From a computational standpoint, two parallel and complementary workflows were developed. The conventional workflow relied on established tools (TopSpin for spectral processing, AMIX for bucketing, and XLSTAT for statistical analysis) and enabled the identification of significant variables through PCA, ANOVA, and Tukey's test. The AI-driven workflow, fully automated and developed in R and Python, included data standardization, dimensionality reduction via PCA, and the training of supervised machine learning models (including Random Forest, SVM, XGBoost, and Logistic Regression), optimized through cross-validation and performance testing. Key spectral features were identified using Mean Decrease in Accuracy (MDA) and cross-compared with those selected by the classical approach to assess convergence and robustness. Finally, the integration between CP-MAS NMR spectroscopy and digital traceability systems based on IoT sensors and blockchain was explored in collaboration with the company Farzati S.p.A., demonstrating the practical feasibility of securely registering and verifying the origin of plant-based materials along the production chain. The results validate a scalable and reproducible approach for origin and quality certification in the agri-food sector, offering new perspectives for the digitalization and transparency of supply chains.

Downloads

Actions (login required)

Modifica documento