Catalytic sensor-based software-algorithmic system for the detection and quantification of combustible gases in complex mixtures

A software-algorithmic method for processing temperature characteristics of a catalytic sensor to recognize the type and determine the concentration of combustible gases in multicomponent mixtures has been developed and its feasibility has been evaluated. This method integrates principal component analysis (PCA) for processing S-characteristics of catalytic sensors with machine learning classification algorithms to enhance selectivity. A flowchart of this software-algorithmic method is presented, and experimental studies of one-, two- and multicomponent mixtures consisting of hydrocarbons and hydrogen are carried out. In this work, a sample of six flammable gases and vapors of VOCs (methane, propane, butane, ethylene, hexane, and hydrogen) was examined. The original gases were calibration gas mixtures in the range of pre-explosive concentration. The results proved that the software-algorithmic method could recognize the type of combustible gas and determine the concentration with an accuracy of up to 90 %.