Cognitive Methodology as an AI Tool for Investigation of the Phenomenological Ground of Melt Electromagnetic Treatment

This chapter is devoted to using artificial intelligence (AI) methods to improve the efficiency of various technological and production processes, particularly in the foundry industry. Many experiments have ascertained that the melt treatment with electric current (ECT) in certain modes achieves an effective improvement of structural parameters and quality of castings. Still, the lack of a holistic theory of the processes occurring in the melt during ECT and a clear understanding of the physical mechanisms of the current influence on structure formation does not allow to predict confidently the results of treatment and stably reproduce ECT modes that provide favorable conditions for the structure formation of castings. Implementation of this promising technology of melt treatment requires a deep study of the elements’ interaction mechanisms of the weakly structured complex system “ECT-of-Melt” (CSEM) at all levels of its hierarchical relationships and chains of interactions; however, the development of knowledge in this poorly studied domain of metallurgy and its practical application for the creation of automated control systems for ECT modes cannot be realized without using information technologies (IT) and artificial intelligence methods. Under such conditions, one of the most important AI approaches was used to study the phenomenological ground of the CSEM – the cognitive 268 methodology of modeling, which is based on structuring knowledge in a specific subject domain (SD). Ontological analysis of the SD using taxonomic models ensured the creation of an ordered information platform for constructing a cognitive model (CM) of the CSEM. The CSEM CM has several essential peculiarities; in particular, a multi-level hierarchical structure that covers more than a hundred concepts – vertices of the corresponding graph and about 2000 edges-connections. Such a CM cannot be used in computer models in the form of a conventional cognitive map (CMp). Therefore, the connection matrix of the primary CM was subjected to decomposition into a set of partial adjacency matrices. On their base, a hierarchical network scheme of interconnections between concepts was developed. The derived partial matrices are presented as formulaic notations, which reflect the content of the CMp corresponding to certain algorithms of machine analysis of Big Data in the modeled system. CM CSEM is supplemented with an electronic catalog of codified concepts and a register of links between them with the specification of influence functions. The novelty in the presented KM consists in conforming of the tact sequence of the concepts interaction at different hierarchical levels due to the introduction of the factor of tact duration of the processes into the properties of the influence functions. The result of the presented development is the completion of the set of AI tools with the help of which it is possible to compile a wide range of algorithms for the cognitive study of the phenomenological basis of ECT, targeted at revealing the most effective modes of ECT of melts and obtaining high-quality castings.