Spectrometric classification models are usually constructed by multivariate analysis of measurements from multiple samples from authenticated reference database. In this study (open access) the authors used a simplified approach. They tood a single measurement: the integrated IR spectrum between 3000-2800 cm2. They used factorial mixture design, on an Excel spreadsheet, to construct a calibration curve based only upon 3 reference samples: 100% Arabica, 100% Rustica and a 50/50 mix. They then validated the curve using a range of other mix proportions, and concluded that it was suitable for detecting Rustica adulteration in “pure Arabica” down to 2.5%. The authors propose this as a useful and cheap strategy for building specific classification models for the routine checking of adulteration in individual coffees that purportedly come from a consistent and well-characterised source.
Comments