Generation of effective functional control of water chemistry at Nuclear Power Plant (NPP) is one of the guidelines of NPP safeguarding. This problem can be solved by means of water chemistry automatic monitoring system. This system provides the on-line monitoring of quality indexes. Analyzing and estimating the water chemistry parameters is the prerogative of the operator as an expert of the system.

Formalization and automation of these functions can be realized only by comprehensive approach. This approach supposes the use of the statistical information and expert experience for diagnosing and forecasting.

Large multidimensional data arrays of quality indexes referring to different operating conditions of NPP allow to analyze complex relations between different factors. For analyzing these relations it's necessary to carry out the decomposition of all data volume to the subsets with similarity of system factors and characteristics.

Classification and modeling of water chemistry was realized by Kohonen neuron network. In general, this network is an algorithm, based on estimating of proximity of classification objects.

The facilities of identification of water chemistry conditions by neuron network were verified by algorithms described in normative documents and by the data of the diagnostic system based on expert knowledge.