Abstract
This thesis applies artificial intelligence, primarily through machine learning and deep learning algorithms, alongside advanced analytical tools and innovative data fusion strategies for food fraud testing. Initially, a systematic review of reported tea fraud cases identifies prevalent types and highlights the limitations of current detection methods, with geographical origin substitution emerging as a critical concern. To address this, the research develops a novel two-tier system that transcends traditional chemometric analysis methodologies and improves food fraud detection capabilities. In tier one, non-targeted spectroscopic fingerprinting (using Fourier Transform Infrared spectroscopy and Near-Infrared spectroscopy), targeted elemental profiling (via X-Ray Fluorescence spectroscopy), and artificial intelligence algorithms classify black tea samples based on their geographical origins. The high-level data fusion methods developed integrate results from three spectroscopic tools, enhancing discrimination accuracy and demonstrating potential for real-time, on-site rapid testing throughout the global tea supply chain. In tier two, the developed non targeted liquid chromatography Time-of-Flight mass spectrometry based testing method discriminates black tea samples with 100% accuracy in both internal and external validations, paving the way for laboratory-based confirmation testing.Thesis is embargoed until 31 December 2026.
| Date of Award | Dec 2024 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Sponsors | Queen's University & China Scholarship Council |
| Supervisor | Christopher Elliott (Supervisor), Di Wu (Supervisor) & Hui Wang (Supervisor) |
Keywords
- Artificial intelligence
- food fraud
- geographical origin
- detection
- Tea
- Machine learning