Diagnostics of inorganic substances in water by method of Raman spectroscopy using adaptive neural networks algorithms

At present, monitoring of processes taking place in natural waters, in particular, the control of qualitative and quantitative composition of mineral waters, the control of technical and residuary waters, the determination of salts composition of aquifer waters, is the actual problem. This paper is devoted to the solution of the inverse problems of laser Raman spectroscopy of water media in salts identification and determination of salts and ions concentration in solutions.
Different influence of different salts and of their concentrations on water Raman spectra provides principal preconditions for identification of salt type by the Raman band. We have used artificial neural networks (ANN) to analyze the Raman spectra with more precision and to solve the problem of recognition of the salt kind in water, with subsequent determination of the salt concentration. Because of great number of components of the solution (up to 5-6 salts) firstly in this work hierarchical structure of neural network classifiers was used in order to recognize great arrays of data [1].
New approaches with using ANN allow recognition of type of the salt with probability 70-100 % and non-contact determination of each salt (or ion) in 5-components solution with accuracy up to deciles of Mole.

1. Dolenko S.A., Orlov U.V., Persiantzev I.G., Shugay U.S. Adaptive construction of hierarchical neural network classifiers. Neurocomputers, elaboration and application, 2005, № 1-2, pp. 4-11 (in Rus.)