研究目的
To develop a method for diagnosing alcoholic beverages by determining the concentrations of ethanol, methanol, fusel oil, and ethyl acetate in water-ethanol solutions using artificial neural networks and Raman spectroscopy.
研究成果
Raman spectroscopy with MLP processing of spectra can be successfully used to detect hazardous impurities in aqueous ethanol solutions. Autonomous determination gives better results for ethanol and slightly better for other components. Feature selection provides a small improvement in solution quality, with cross-entropy being the most effective method.
研究不足
The method's accuracy for methanol and ethyl acetate is several times greater than the maximum permissible concentration but more than an order of magnitude less than the lethal concentration for 100 ml of drink. The complexity of the problem and the representativity of the dataset may affect the results.
