Functional Diagnostic System for Multichannel Mine Lifting Machine Working in Factor Cluster Analysis Mode

Abstract

The primary direction of the increase of reliability of the automated control systems of complex electromechanical machines is the application of intelligent information technologies of the analysis of diagnostic information directly in the operating mode. Therefore, the creation of the basics of information synthesis of a functional diagnosis system (FDS) based on machine learning and pattern recognition is a topical task. In this case, the synthesized FDS must be adaptive to arbitrary initial conditions of the technological process and practically invariant to the multidimensionality of the space of diagnostic features, an alphabet of recognition classes, which characterize the possible technical states of the units and devices of the machine. Besides, an essential feature of FDS is the ability to retrain by increasing the power of the alphabet recognition classes. In the article, information synthesis of FDS is performed within the framework of information-extreme intellectual data analysis technology, which is based on maximizing the information capacity of the system in the process of machine learning. The idea of factor cluster analysis was realized by forming an additional training matrix of unclassified vectors of features of a new recognition class obtained during the operation of the FDS directly in the operating mode. The proposed algorithm allows performing factor cluster analysis in the case of structured feature vectors of several recognition classes. In this case, additional training matrices of the corresponding recognition classes are formed by the agglomerative method of cluster analysis using the k-means procedure. The proposed method of factor cluster analysis is implemented on the example of information synthesis of the FDS of a multi-core mine lifting machine.