Determination of Catechin Monomers in Tea Polyphenols Powder Using NIR and ANN

Near infrared diffuse reflectance spectra of 52 tea polyphenol powder samples were collected with FT-NIR spectrometer. The calibration models of total catechins(TC) were established with back-propagation artificial neuron network(BP-ANN) and radical based function artificial neuron network (RBF-ANN) and optimized with prediction sample set. The result showed that the RBF-ANN model is better than the BP-ANN model. Calibration models of total ester catechins(TEC), total simple catechins(TSC), catechin monomers (EGCG, GCG, ECG, D,L-C, EC and EGC) were established with RBF-ANN. The models of TC, TEC, TSC, EGCG, ECG were robust with prediction correlation coefficient(R) above 0.9 and prediction relative standard error (RSE) less than 0.10%. The models of GCG, D,L-C, EC, EGC had much higher RSE of over 0.15%. This result suggests that it is feasible to rapidly determinate the contents of TC, TEC, TSC, EGCG, ECG in tea polyphenols powder with NIR spectroscopy combined with RBF-ANN models.