THE IDENTIFICATION OF THE GASES ATOMS IN THE EMISSION SPECTRA USING THE PRINCIPLES
OF ARTIFICIAL NEURAL NETWORKS

Shcherbakova Anna Alekseevna, programmer, sub-department of instrument engineering, Penza State University (40 Krasnaya street, Penza, Russia), anutka7790@mail.ru
Solov'ev Vladimir Aleksandrovich, doctor of technical sciences, professor, sub-department of general professional disciplines Penza branch of Military academy of material-support logistics named after army general A. V. Khrulev (Military town, Penza-5, Russia), v.soloviev@bk.ru

004.942

10.21685/2307-5538-2018-2-5

Background. The object of the study is an information-measuring system for the identification of gases. The subject of the study is a method of reliable identification of gases by emission spectra based on the principles of artificial neural networks. The aim of the work is to develop a method for reliable identification of gas atoms based on the analysis of radiation spectra using the principles of artificial neural networks.
Materials and methods. To solve the set tasks we used methods of solving ill-posed problems in spectral analysis, mathematical modeling, statistical data processing, the theory of artificial neural networks, back-propagation, simulation of the mathematical model were performed in the software environment of Visual Basic for Applications (VBA) in Microsoft Excel.
Results. On the basis of spectrometric laboratory setup the method of gas identification using the principles of artificial neural networks and spectral characteristics in the visible wavelength range is developed, the structure and mathematical model of the artificial neural network of gas identification by its spectral characteristics are developed; the parameters of the activation function allowing to train the artificial neural network, where the reference values of the intense wavelengths of different gases are used as a teacher.
Conclusions. The use of the proposed system and the method for gas atoms identification will allow: to make a reliable identification of gases using artificial neural networks with the training of the measuring system by the teacher; the teacher is the reference gases with known wavelengths; the training minimizes the target functions that establish a relationship between the intensities and wavelengths in the visible wavelength range. Recommendations on modernization of the structural scheme and equipment of the information and measuring system of identification of substances depending on the solved problem are offered.

identification, wavelength, intensity of radiation, spectrometer, gas atoms, objective function, the activation function