XAI-Supported Diagnostics of Real-Time Communication with EMS Issues

Digitalized production systems are inherently complex and characterized by numerous interdependent components, making it challenging for operators and engineers to fully understand their relationships. On-demand dependency analysis is often costly, time-consuming, and difficult to perform. Communication represents a critical digital subsystem upon which the reliable operation of digital components depends. However, communication quality may be adversely affected by the activity of physical components controlled by digital systems, primarily due to electromagnetic susceptibility. This paper presents a supervisory system designed to identify and analyze dependencies between communication failures and the activities of physical components as functions of control states and temporal conditions. The proposed approach enables improved monitoring, diagnosis, and understanding of interactions between physical and digital subsystems in digitalized production environments. The proposed system achieves this by continuously analyzing the condition of the communication network and identifying correlations with the current states of the controlled system using machine learning techniques. To determine the underlying causes of communication disturbances, Explainable Artificial Intelligence (XAI) is employed to interpret the model’s predictions and identify the system outputs that most significantly influence network performance. Experimental results demonstrate that the proposed approach can successfully identify the causes of communication degradation and indicate the specific system outputs whose activation leads to the deterioration of communication networks.