Fault Diagnosis System for Auxiliary Converters in HXD3C Series Electric Locomotives

Abstract

This article proposes a fault diagnosis system for auxiliary converter (ACU) in HXD3C-based electric locomotives operating in the railways of Uzbekistan. Auxiliary converter defects are important because they affect traction motors, cooling systems, compressors, and other auxiliary machinery necessary for the safe operation of locomotives. Currently, in Uzbekistan's locomotive depots, fault investigation mainly depends on visual inspection and manual review of major exports of train control and Monitoring System (TCMS) malfunctions. This process takes a long time, depending on the experience of the staff, which may lead to the absence of important error patterns in historical records downloaded from TCMS. The entire proposed system consists of three main parts: an ARM-based minicomputer, diagnostic software with a machine learning model integrated for future predictive maintenance, and a screen for displaying and sending results to staff. The proposed method uses uploaded TCMS error records and converter auxiliary unit records as input. Special software then analyzes data and output reports, visual statistics, and recommendations for possible ACU failure maintenance guidance. First, inappropriate and non-specific records are cleared. ACU-related errors are then selected, grouped by time and error type, and used to identify patterns when errors occur. The results of the initial fault diagnosis show that with the help of simple visual output graphs such as line and bar graphs, maintenance personnel can roughly analyze the occurrence of faults and possible causes of malfunctions. These schedules save a lot of time for staff during maintenance error analysis. And this is very important for fast and high-quality maintenance of electric locomotives. This initial result is itself an important part of this study, as it partially automates the troubleshooting process in Uzbek Railways locomotive depots. The results show that the proposed diagnostic system can support and accelerate maintenance processes before full real-time remote monitoring is available in Uzbekistan's railways. The study promotes a practical approach to using existing TCMS historical data to improve error diagnosis, reduce manual analysis, and prepare O'Z-EL-based electric locomotives for potential predictive maintenance integration in the future.