Automatic transportation process control systems with an extended data analytics circuit

Abstract

A new hybrid approach has been proposed to automate the management of complex technological processes at railway stations of industrial transport using intelligent monitoring technologies. This approach is based on the concept of predictive modeling combined with methods of statistical analysis, including a modification of the principal components analysis method for multivariate statistical analysis and the identification of violations in technological processes using a combination of well-known methods such as contribution analysis and fuzzy dynamic analysis. The principal feature of the hybrid approach is mapping the initial space of numerical parameters of the technological process onto a new space formed by fuzzy rules of an evolving system model. Applying multivariate analysis to new system variables using the principal component method allows for the formation of a few intermediate variables with different degrees of granularity and interpretability, describing the behavior of the controlled process, which makes it possible to develop mathematical models and algorithms for solving various monitoring tasks An example of using this approach for post-processing monitoring data to identify performance discrepancies in a marshalling yard and anomalies in the controlled process is considered.