An Information Fusion Based Incipient Fault Diagnosis Method for Railway Vehicle Door System

Sie sind hier: Homepage > Suche

An Information Fusion Based Incipient Fault Diagnosis Method for Railway Vehicle Door System

Zhou, Keyi / Lu, Ningyun / Jiang, Bin / Liu, Zheng / Zhang, Baoli / Chen, Jianfei

The safety and reliability of the railway vehicle door system are critical to ensure passengers’ safety and transportation efficiency. Fault diagnosis is essential for such a purpose. However, due to various uncertainties in the models and measurements, as well as the existence of incipient faults, traditional fault diagnosis methods often suffer from high false alarms. In this study, an incipient fault diagnosis method is developed using an information fusion strategy, which can greatly reduce false alarms and hence improve the reliability and accuracy of fault diagnosis results. The proposed fault diagnosis method is a data-driven approach, where current, rotational speed, rotational angle/distance signals collected from the driven motor and vibration signals collected from the supporting elements and door leaves are used for fault diagnosis. Initially, features are extracted from the signals to train two classifiers. Subsequently, these classifiers generate probabilities for different fault types. Then fault diagnosis model is developed using an information fusion strategy where evidence belief divergence and fuzzy preference relationship are employed to handle conflicts between different evidence. The principal contributions center on the elimination of information uncertainties within the railway door system, enabling precise diagnosis of incipient faults. To validate the methodology, verifications are conducted on a vehicle door test bench. Comparative experiments are also conducted to demonstrate the superiorities of the proposed method in comparison to approaches that do not incorporate information fusion or address information conflicts. The experimental results show that the proposed method significantly enhances diagnosis accuracy by at least 10%.