SUMMARY
The object of research is the process of functioning of the chemical-technological system of two series-connected units, designed to produce Fe-C melt. This process is evaluated on the basis of the classification rules, which makes it possible to identify that component of the system, according to which the deviations of the operating mode from the normal by chemical analysis of the content of sulfur and phosphorus are revealed. Such consideration allows determining the completeness of the desulfurization and dephosphorization processes of the Fe-C melt. One of the most problematic places is the lack of systematized data on the procedures for determining possible deviations from the normal mode of functioning of the aggregates in terms of the completeness of the desulfurization and dephosphorization processes.In the course of research, the methods of parametric classification were used, which allow one to obtain an analytical description of the discriminant function and, based on a comparison of its values at a given point in the space of factor-attributes with a threshold value, assign an object to one of the classes. In this research, each of the classes characterizes an aggregate of a chemical-technological system for the content of sulfur and phosphorus in the Fe–C melt.The obtained results allow to state that it is possible to construct a broken curve separating classes using the methods of parametric classification. This is due to the fact that the proposed five-step procedure for selecting the input data area makes it possible to remove one of the restrictions on the use of parametric classification methods, namely, the requirement of equality of covariance matrices. The proposed procedure has a number of features, in particular, the choice of the input data area for calculating the coefficients of the discriminant function and threshold values is determined through the vertices of the plans for the full factorial experiment. This ensures the possibility of obtaining one hundred percent accuracy of classification in areas corresponding to its plan. Compared to the results of the construction of classification rules for the total sample of data, that is, for all points of the space of factor-attributes, and not only for points of the chosen plan, this does not have significant advantages. But from the point of view of obtaining the form of a separating curve other than linear, this can be beneficial if the input data samples are not well separated.