INFORMATION TECHNOLOGY FOR AVAILABILITY ASSESSMENT OF INTELLIGENT DIAGNOSTIC SYSTEMS WITH RELEARNING BASED ON DIGITAL TWINS

Abstract

Assessing the reliability of Intelligent Diagnostic Systems (IDS) in Industry 4.0/5.0 environments poses a challenge due to the dynamic nature of AI-based tools that evolve through relearning via Digital Twins (DT). Traditional static reliability models fail to capture the variation in diagnostic trustworthiness and the impact of latent failures. The paper aims to develop an information technology for assessing the availability of IDS with relearning capabilities, enabling quantitative evaluation of how dynamic updates affect system reliability and automating the model selection process. The study employs systems analysis and set theory to construct a model classifier; Multi-Fragment Markov Models (MFMM) to describe the stochastic behavior of systems with variable parameters; and the IDEF0 functional modeling methodology to formalize the assessment technology. A classifier of availability models has been developed, distinguishing distinct features of relearning processes (triggers, fragmentation types, recovery strategies). Based on this classifier, a complex of basic models (ranging from linear BM1 to matrix-structured BM5federated) was synthesized, covering scenarios from simple IoT devices to critical Small Modular Reactor (SMR) equipment. An information technology was developed and formalized as a hierarchical set of IDEF0 diagrams, defining the workflow for model selection, transition rate matrix synthesis, and simulation result verification. Conclusions. The proposed technology forms the architectural basis for the Reliability Simulation Integration Module (RSIM) software. It enables engineers to justify the implementation of adaptive diagnostic algorithms by balancing the reduction of hidden failure risks against the downtime required for system relearning.