Classification of geographical origin of wheat flour using NIR spectroscopy and femtosecond LA-ICP-MS coupled with chemometric models

Abstract

Wheat flour is a staple commodity widely consumed globally as a wheat-derived product with various characteristics. These characteristics are essential as they influence product security, quality, and pricing. Geographical origin (GO) is a valuable way to classify products based on their physical and chemical attributes. This study combined near-infrared spectroscopy (NIRS) and femtosecond laser ablation-inductively coupled plasma-mass spectrometry (fsLA-ICP-MS) results with a chemometric approach to classify the geographic origin of domestic and imported wheat flour. Partial least-squares discriminant analysis (PLS-DA) and radial basis function support vector machine (RBF-SVM) were employed to classify and analyze the correlation structure and variance within the data with significant differentiating elements including 24Mg, 55Mn, 31P, 56Fe, 66Zn, 121Sb, 107Ag, 208Pb, 140Ce, and 133Cs. Six preprocessing analyses were performed to enhance data quality and accuracy. Combining NIRS with fsLA-ICP-MS data in the RBF-SVM model, the standard normal variate (SNV) achieved an accuracy of 95 %. Meanwhile, the PLS-DA method employed mean normalization, smoothing, Savitzky-Golay first derivative (SG1), and Savitzky-Golay second derivative (SG2), identified as the highest-performing preprocessing techniques, resulting in 100 % accuracy.