Evaluation of SIMCA and PLS algorithms to detect adulterants in canola oil by FT-IR

Adulteration of canola oil with four potential edible oils was analysed using FT-IR and chemometric methods. The adulterants (corn, peanut, soya bean and sunflower oils) were studied in four different proportions (canola oil + adulterant oils: 90 + 10, 95 + 5, 98 + 2 and 99 + 1 in volume). Excellent classification results were obtained when multi-class approaches were performed with a maximum error of 3%, using 1630 or 16 wavenumbers as variables. In the case of one-class approaches, the selection of variables (16 wavenumbers) was necessary, improving the classification error to 5%. The differences observed using the different methods were related to the nature of each model depending on how the boundaries are set in each of them, responding either to a PCA-based or PLS-based algorithm.