Neural network-based prediction of technical failures in communication networks

Abstract

This article discusses the problem of automated forecasting of the technical condition of train radio communication networks within the railway sector of Uzbekistan. The technical characteristics of existing systems, the theoretical model of signal propagation, and the main causes of failures are examined in detail. Traditional forecasting approaches are shown to be limited, as they often fail to adequately reflect nonlinear processes, the influence of electromagnetic interference, and the impact of maintenance activities. To address these shortcomings, an automated forecasting approach based on artificial neural networks is proposed. This method makes it possible to identify both sudden and gradually developing faults in advance, thereby increasing overall system reliability, supporting effective planning of technical maintenance, and reducing operational costs. Practical experiments carried out on railway sections confirmed the effectiveness of the proposed methodology. Overall, the use of neural networks for forecasting is considered a scientific and practical solution for enhancing the reliability of train radio communication systems, improving safety, and accelerating the gradual transition toward digital communication technologies.