Machine learning methods in recognition of harmful impurities in the atmosphere from spectral data
Ph.A. Kozhevnikov1, M.R. Konnikova1, A.S. Sinko1, A.A. Angeluts2,3
1Lomonosov Moscow State University, Faculty of Physics, Moscow, Russia
2Lomonosov Moscow State University, Faculty of Physics, Irkutsk, Russia
3Matrosov Institute for System Dynamics and Control Theory of Siberian Branch of Russian Academy of Sciences
Keywords: terahertz spectroscopy, neural network, deep convolutional neural networks, 1D convolutional networks, transformation of neural network architecture, gas analysis
Abstract
Expanding the instrumental and analytical methods for identifying harmful impurities in the atmosphere is an important task for solving environmental problems. In this regard, the work focuses on developing a comprehensive approach to detection of harmful impurities in atmospheric air. This approach is based on measurements of the absorption spectra of air containing harmful impurities along a path by pulse terahertz spectroscopy methods. To analyze the obtained spectral data, a neural network is created and applied, and arrays of model absorption spectra of gas mixtures with different qualitative and quantitative compositions are generated for its training. It is shown that the neural network is capable of identifying six gas components in concentrations of up to 0.01 ppm with accuracy of 90-95%. A series of experiments with real gases confirms the sensitivity of the THz spectroscopy method to low gas concentrations in the mixture. The results show that the combined method is sufficiently sensitive for identifying both single gases and gas mixtures, which can be used for environmental monitoring.
|
