Machine Learning Methods for Classifying Gas Discharge Images of Liquid Solutions

A program has been developed that utilises machine learning methodologies for the classification of gas discharge images of liquid solutions. The glow patterns of liquid solution droplets in an electromagnetic field were recorded using the gas discharge visualization (GDV) method. The method utilises a mass-produced Bio-Well device, does not require consumables, and acquires images in approximately one minute. In the developed algorithm, classifiers were trained to demonstrate high accuracy in distinguishing various types of water, including tap-filtered water, distilled water, water from three different springs, tea with sugar, and solutions with magnesium, salt, and shungite impurities. The developed approach is employed in the research of water purification methods using electromagnetic fields.