A Hybrid Data-Fusion Estimate Method for Health Status of Train Braking System

The high-speed solenoid valve is a crucial module in train braking system, which is an essential factor to ensure the safe operation of trains. How to estimate the health status of the high-speed solenoid valve accurately to improve the reliability of train braking system is a challenging issue. Most related work relies on accurate physical models or large amounts of historical data. To address this challenge, this paper proposes a hybrid data-fusion estimate method for the health status of train braking system. Firstly, the physical model of the high-speed solenoid valve is established, and physical indicators which represent the working performance are extracted. Then, the dynamic driving current is processed by ensemble empirical mode decomposition (EEMD) to calculate the information entropy. Physical indicators and information entropy indicators are combined into a feature vector, which can be reduced the dimension by the t-distributed stochastic neighbor embedding (T-SNE) algorithm. Finally, the feature vector is input into the probabilistic neural network (PNN) to estimate the health status of train braking system. The proposed method is implemented in the high-speed solenoid valve degradation dataset, which collected by the train brake system experiment platform. The result shows that it is better than other methods in the accuracy and calculation efficiency.