Browsing by Author "Rust, Alexandra Elizabeth"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    Honey authentication: effect of irradiation and ageing on near-infrared (NIR) spectroscopy classification models
    (Stellenbosch : Stellenbosch University, 2020-03) Rust, Alexandra Elizabeth; Manley, Marena; Williams, Paul James; Allsopp, Mike; Stellenbosch University. Faculty of AgriSciences. Dept. of Food Science.
    ENGLISH ABSTRACT: Near-infrared (NIR) spectroscopy was used to investigate the effects of syrup dilution, ageing, storage temperature and irradiation treatment on the NIR spectra of honey. Additionally, NIR spectroscopy and partial least squares discriminant analysis (PLS-DA) were employed to develop a classification model for the rapid screening of high-fructose corn syrup (HFCS) and invert cane sugar syrup (ICSS) diluted honey. Detection of irradiation treatment was also investigated, to assess NIR spectroscopy-based models as a potential screening tool for detecting mislabelled honey. Unfiltered and unheated honey samples (n = 17) obtained from South African beekeepers were uniformly strained and subjected to treatment combinations of 10 kGy gamma irradiation and dilution with 0, 10 or 20% (w/w) ICSS or HFCS to create sub-samples (n = 174) which were stored at 25°C. Another three undiluted subsets were stored at 4°C, 40°C and in uncontrolled ambient conditions (n = 51). A benchtop BÜCHI Fourier transform-near infrared (FT-NIR) spectrometer and a portable MicroNIR NIR spectrometer, with ranges of 1000–2500 nm and 908-1676 nm respectively, were used to acquire triplicate spectral measurements, over a period of 9 months. ANOVA-simultaneous component analysis (ASCA) indicated that honey type, diluent type, storage temperature and age had significant (p > 0.05) effects on the spectral dataset, while diluent level and irradiation treatment did not. Despite this, irradiation treatment was found to reduce the validation accuracy and efficacy of authentication models, by 5.82% and 7.19% respectively, when PLS-DA models based on only irradiated and only non-irradiated spectral data were compared, suggesting that authentication may be impeded by irradiation treatment to some degree. However, a PLS-DA model discriminating on the basis of irradiation treatment obtained an unsuccessful validation classification of 59.7%, suggesting that there is little or no utilisable effect of irradiation on the spectral data. The best-performing authentication solutions were individual two-class PLS-DA models for detecting ICSS (75.95% accuracy, 86.31% sensitivity) and HFCS (73.95% accuracy, 82.14% sensitivity) dilution, which demonstrated predictive power adequate for screening purposes. PLS-DA models based on spectral data acquired with the benchtop BÜCHI instrument performed best when compared with the portable MicroNIR instrument and its two sample presentation formats. Despite this, the MicroNIR with Teflon cup sample presentation was shown to be a feasible and cost-effective alternative, demonstrating similar accuracies (70.0-75.47%) and efficiencies (68.22-74.51%). In addition, quantification of the level of diluent with partial least squares regression (PLSR) was poor for both ICSS (R2 Pred = 0.118, RMSEP = 6.795%) and HFCS (R2 Pred = 0.147, RMSEP = 6.596%) dilutions. This was attributed to an inadequate range of dilution levels in the reference data, as well as the insignificant effect (p < 0.05) of diluent level on the overall variation in the spectral data. The findings of this study highlighted the potential shortcomings of NIR spectroscopy models in providing definite authentication, while demonstrating the capabilities of this technique for authenticity screening.