A new transformer-based axis temperature alarm system

The current axis temperature alarm system mainly determines the current alarm level based on the axle temperature status or characteristics detected by a single monitoring station, which has a problem: the characteristic data detected by the monitoring station is not related to the previously detected characteristic data. In this study, temperature rise, column temperature rise difference and vehicle temperature rise difference are taken as the feature data, and the attention mechanism model with mask is used to judge the current alarm level by combining the feature data measured by the previous monitoring station. The simulation results show that the optimized system has a discrimination agreement rate of 87.2% for the alarm level, which is a 13.5% increase in accuracy over the previous method.