Research on Intelligent Diagnosis of Rail Vehicle Transmission System

Abstract

Rail vehicle transmission system is the key parts of bogies, which transmits load and power. The failures of transmission systems will result in long downtime and expensive maintenance costs. The structure of transmission system is complex and the vibration signals collected from various components are coupled. Therefore, the diagnostic accuracy of a single sensor is relatively low, and fault location is difficult. To improve the accuracy of fault diagnosis, this paper proposes a simple but practicable method based on multiple sensor fusion technology, which combined support vector machine (SVM) with fuzzy integral fusion algorithm (FI). First, the energy entropy features are extracted from multiple sensors data as the inputs of SVMs. Then, the outputs of SVMs are transformed into posteriori probabilities as the basis for calculating fuzzy memberships. Finally, the comprehensive judgment is obtained by FI operation. The data collected from running rail vehicles verify that recognition rate of this scheme is higher than the single sensor and other fusion methods. Moreover, it is also verified that the method has certain application value.