Rapid methanol detection in ethanol solutions via non-linear Raman spectroscopy modeling.

Methanol contamination in ethanol-based products poses a significant health risk due to its toxicity at low concentrations. This study developed a rapid, non-destructive method for detecting and quantifying methanol in alcoholic beverages using Raman spectroscopy combined with chemometric models, including Partial Least Squares (PLS), Interval PLS (iPLS), Kernel PLS, and Radial Basis Function Artificial Neural Networks (RBF-ANN). The Kernel PLS model, applied to optimized spectral intervals, achieved the lowest relative prediction error (REP%) of 6.84 % on a test set. Validation on real samples spiked with 3-20 % methanol confirmed the method's ability to quantify methanol with absolute relative errors below 10 % in most cases. Compared to conventional techniques like gas chromatography (GC) and high-performance liquid chromatography (HPLC), this approach requires minimal sample preparation and shorter analysis time, making it suitable for quality control in the food, pharmaceutical, and cosmetic industries.