Health State Assessment of Traction Converters under Multi-Operating Conditions Based on Feature Extraction

The traction converter, a critical component of rail transit vehicles, significantly impacts operational safety. Frequent starts/stops and high-intensity operations subject it to cyclic electrical and thermal stresses, accelerating performance degradation. Variable operating conditions further complicate assessment by introducing nonlinear and non-stationary degradation patterns. This paper proposes a feature extraction-based health assessment method for traction converters under variable conditions. Firstly, a mathematical model of the traction converter system is established, and the key components that have the greatest impact on the system are analyzed. Then, a degradation model of the key components is established, along with a degradation analysis under variable operating conditions, which is verified through simulation. Finally, key feature parameters are extracted from multiple dimensions and combined with a neural network algorithm to achieve accurate assessment of the health status.