Mathematical Methods for Detecting and Localizing Failures in Complex Hardware/Software Systems

Abstract

This article addresses the problem of creating and applying methods of automatic failure self-diagnostics in complex hardware/software systems (SCADA systems included). We conducted a review of modern methodologies used for solving problems of this class. A structural-logical model was proposed to describe the system’s architecture. We developed a mathematical method for diagnosing failures based on analysing changes in the state of system information flows. This included failure diagnostic criteria for the method we developed. As a result, it allowed us to obtain the analytical functional dependencies for the detection and localization of failures in the system’s structural modules. Graphs are presented to illustrate the probability of the system’s non-failure versus time-of-failure detection and localization. We obtained logical conclusions about improving the quality of system functions using the methods we developed for automatic failure self-diagnostics.