The article presents a method for analyzing causal relationships aimed at ensuring the reliability and safety of complex human-machine systems (HMS). The developed approach combines system decomposition with modeling of dynamic interactions, considering the influence of human, technical, and organizational factors. A model based on Markov processes with variable transition intensities dependent on subsystem states is proposed, which allows for identifying cascade failures and reducing the dimensionality of tasks without loss of informativeness. In the literature review, we performed an analysis of modern methods, including dynamic Bayesian networks, fuzzy cognitive maps with genetic tuning, and system-theoretic process analysis. The effectiveness of hybrid algorithms in identifying critical risks is demonstrated. As a result, we have conducted numerical experiments on aviation systems, such as Airbus A-320, Bombardier CRJ-200, Boeing 747, and Sukhoi Superjet 100 (RRJ-95). These experiments illustrate the calculation of stationary state probabilities and risk dynamics. Conclusions are formulated regarding the methodological flexibility of the approach, its advantages in overcoming the “curse of dimensionality,” and prospects for application in other industries.
