Research on Fault Diagnosis Algorithm of Subway Vehicle Door System Based on Multi-stage Feature Fusion of Vibration Signals.

The subway vehicle door system is a critical component of the railway vehicle, and its load-carrying transmission mechanism is more susceptible to failure or damage compared to other components. Conducting a quick and accurate evaluation analysis of its health status is an effective approach to detecting hidden flaws and enhancing the safety and reliability of the door system. Therefore, this paper aims to address the issue of frequent failures in the door system by analyzing vibration signals in accelerated life testing, exploring a fault diagnosis algorithm, and achieving efficient diagnosis and timely feedback on the mechanism's health status. To begin, a test platform is built to collect vibration data and identify structural failures, laying the foundation for further analysis. The wavelet packet EMD noise reduction algorithm is then applied to mitigate the phenomena of signal mode mixing, enhancing the clarity of data. Subsequently, the Person, mRMR, and PCA algorithms are utilized to effectively highlight the multi-stage features of the vibration signal, generating comprehensive characterization quantities that provide in-depth insights into the signal's behavior. Finally, a Bayesian algorithm-based SVM multi-classification model is designed, taking advantage of the multi-stage feature combination to efficiently predict the health status of the mechanism. This model proves superior, achieving an accuracy that is 6.25 percent higher than the non-segmented classification model, demonstrating the advantage of segmented analysis in understanding and predicting the mechanism's performance.