Determining partial pressure and temperature of gas using artificial neural networks

The question of solving the inverse problem of gas medium optics to determine the partial pressure and gas temperature using artificial neural networks is considered. The analysis of the errors of the obtained models was carried out depending on the number of used spectral centers and the size of the training sample, which showed a tendency to decrease the magnitude of errors with the growth of these parameters. The models were obtained that provides a solution to the inverse optical problem of determining the partial pressure and temperature of carbon monoxide and water vapor with a relative error of less than 3 % and 3.5 % respectively.