Monitoring and diagnostics of technological processes with the possibility of dynamic selection of variables

The article describes a method for monitoring and diagnosing the state of technological processes with the possibility of dynamic selection of variables. As a basis, the algorithm known from machine learning algorithm of single-class classification OneClassSVM was used, which showed a high percentage of accuracy in determining the normal state of the technological process. A model of a technological process as an infinite data stream with changing properties with a large number of parameters is proposed. A probabilistic model of the state space is used for possible dynamic selection of variables. An experiment is described in which the effectiveness of the proposed method was studied, and its results are presented. Its validity is shown, and it is noted that it can be used in modern systems for monitoring the state of technological processes to identify malfunctions and threats to information security.