METHOD FOR MONITORING THE TECHNICAL CONDITION OF GTO CONVERTERS IN TRACTION ROLLING STOCK BASED ON NEURAL NETWORK REFERENCE MODELS AND OPERATIONAL PARAMETERS
This paper presents a developed fault monitoring methodology designed for gate-locked thyristor converters (GTO), which are widely used in traction locomotives. The main goal is to create an intelligent system capable of detecting failures in a timely manner, which is crucial for improving reliability and ensuring uninterrupted operation of railway transport. The methodology is based on data obtained from on-board technical diagnostic systems that collect key operating parameters in real time, such as current, voltage, overvoltage, ambient temperature, and the temperature of the converter itself. These parameters serve as the basis for the developed model, which allows you to predict and prevent possible malfunctions. The system is based on a neural network model that has been trained on a hybrid dataset. This set includes both synthetic and partially real data obtained during operation. This approach solves the problem of the lack of real data on breakdowns, allowing you to simulate various potential failures and operating modes. The aim of the study is to create a hybrid machine learning model that will have three important advantages: High diagnostic accuracy, Easily interpretable results for maintenance personnel, Practical applicability in real-world railway operating conditions. If the system is successfully implemented, it will open the door to predictive maintenance, which will significantly reduce the likelihood of sudden breakdowns and operating costs.
Сведения
- Журнал
- Вестник КазАТК
- Опубликовано
- 2026 · Т. 4, № 145, С. 62-75
- Язык
- Английский
- ISSN
- 1609-1817 · 2790-5802
- Издатель
- JSC Academy of Logistics and Transport