Comparison of three kinds of models for detecting tartrazine in tea powder by synchronous fluorescence spectroscopy combined with chemometrics.

This study developed a rapid and green method for detecting tartrazine in tea powder using front-face synchronous fluorescence spectroscopy coupled with chemometrics (PLS algorithm). A total of 38 samples (18 unadulterated and 20 adulterated at 2 ‰-40 ‰) were analyzed. Three model types were constructed for qualitative and quantitative analysis: a local model (single spectrum), a full-spectrum model, and an ensemble model combining multiple single-spectrum members. For qualitative classification, the ensemble model achieved 100% accuracy, outperforming the optimal local model (40 nm interval) and the full-spectrum model. For quantitative prediction, the ensemble model also performed best, followed by the full-spectrum model. This method provides a simple, simultaneous solution for determining the presence and concentration of tartrazine adulteration in tea powder.