CC BY
2The 30th International Conference on Advanced Laser Technologies LD-O-3
ALT'23
Integration of data of various types of spectroscopy for determination of concentrations of ions in multi-component aqueous solutions
A. Guskov1'2, I. Isaev2, S. Burikov1'2, T. Dolenko1'2, K. Laptinskiy2, O. Sarmanova12, S. Dolenko2
1-Department of Physics, Lomonosov Moscow State University, Leninskiye Gory 1/2, 119991 Moscow, Russia 2- Skobeltsyn Institute of Nuclear Physics, Lomonosov Moscow State University, Leninskiye Gory 1/2,
119991 Moscow, Russia
dolenko@srd. sinp. msu. ru
The problem of simultaneous determination of concentrations of several types of ions in aqueous solutions is very topical in various subject areas, ranging from industry wastewater monitoring to environmental monitoring and quality control of beverages, e.g. mineral waters. Heavy metals are among the dangerous pollutants of water, therefore in many areas of industry and ecology there is a need to determine the concentration of heavy metal ions dissolved in water. Ions of various inorganic salts are simultaneously present in natural waters including those consumed by people, e.g. drinking or mineral waters.
The most accurate methods for determining the chemical composition of solutions are methods of conventional chemical analysis. However, this approach is time-consuming, it requires good sample preparation and consumption of expensive reagents. At the same time, most practical problems require easy-to-use, express and contactless methods. Optical spectroscopy (OS) techniques have the listed advantages; therefore, they make a promising alternative to chemical analysis. It should be also noted that different types of OS possess different properties, making them provide different information about the studied object. Therefore, integration of data of different OS types may result in more precise determination of the composition of the solution.
The most widely used types of OS are laser Raman spectroscopy, absorption spectroscopy and infrared spectroscopy. However, at present there is no analytical solution to the problem of determining the concentrations of each component in multicomponent solutions by their spectra for any of these methods. This problem is an inverse problem having properties that make solving it a hard task: it is non-linear, ill-conditioned or even ill-posed. Moreover, even the direct problem of modeling optical spectra cannot be adequately solved for spectra of liquids.
One of the few ways to solve such problems is the use of machine learning (ML) methods based on training ML models on experimental data. Here we present data integration of OS methods - simultaneous use of Raman, optical absorption and/or IR spectra to determine the types and concentrations of ions present in aqueous solutions.
In this paper, we report the results of solving two problems of the described type: determining the concentrations of 10 ions of inorganic salts by Raman and optical absorption spectra [1], and determining the concentrations of heavy metal ions in solutions by Raman absorption and infrared spectra [2]. The problems are solved using various ML methods. The results are compared with those obtained using data of each OS type separately.
The study was carried out at the expense of the grant from the Russian Science Foundation, project No. 19-11-00333, https://rscf.ru/en/project/19-11-00333 /.
[1] I.Isaev, I.Gadzhiev, O.Sarmanova, S.Burikov, T.Dolenko, K.Laptinskiy, S.Dolenko. Using Method Integration Transfer Learning for Neural Network Solution of an Inverse Problem in Optical Specroscopy. Proc.SPIE, 2022, V.12193, Laser Physics, Photonic Technologies, and Molecular Modeling, art.121930Y. DOI: 10.1117/12.2626358.
[2] A.Guskov, K.Laptinskiy, S.Burikov, I.Isaev. Integration of Data and Algorithms in Solving Inverse Problems of Spectroscopy of Solutions by Machine Learning Methods. In: Kryzhanovsky, B., Dunin-Barkowski, W., Redko, V., Tiumentsev, Y. (eds). Advances in Neural Computation, Machine Learning, and Cognitive Research VI. NEUROINFORMATICS 2022. Studies in Computational Intelligence, 2023, V.1064, pp. 395-405. Springer, Cham. DOI: 10.1007/978-3-031-19032-2 41.
