Train Communication Ethernet Fault Diagnosis

The train communication network is the central nervous system of the train and is of great significance for the normal operation of the train.With the continuous improvement of the intelligent level of trains,the amount of data that needs to be transmitted in the train communication network continues to increase,which makes the bandwidth demand in the train network increase.The traditional fieldbus communication method has been far from satisfactory.Ethernet has become the research direction of the future train communication network with its transmission rate advantage.TRDP(Train Real-time Data Protocol),which is the international standard for train communication Ethernet,has become a new research hotspot.TRDP is the guarantee for the efficient and reliable operation of the future train communication network,but at the same time the complexity of the network greatly increases the difficulty of maintenance.Based on this,This thesis takes TRDP as the research object,carries out research on its fault diagnosis method,and builds a fault diagnosis platform.The main tasks are as follows:(1)The architecture and communication mechanism of train real-time data protocol TRDP are analyzed In-depth,the communication steps of the process data is proposed.TRDP process data communication is realized,the transmission messages and packets captured from wireshark software verified the TRDP communication test.(2)Combined with the standard Ethernet Simple Network Management Protocol(SNMP)and its Management Information Base(MIB),the SNMP of the train real-time data protocol is analyzed and its inclusion characteristics are given.A feature extraction method is designed to collect TRDP features and switch features respectively.A TRDP fault diagnosis platform was built using PC and ARM equipment.(3)In order to establish a fault diagnosis model,11 kinds of working conditions such as program jam,device disconnection,network disconnection,memory leak,program crash,port unsubscription and normal communication,are simulated.Data preprocessing(unitization,normalization and dimensionality reduction)isdesigned,and result is compared.(4)Combining the principle of support vector machine,the one-to-one method is proposed to realize multi-classification and cross-validation and grid search are proposed for parameter optimization,accuracy,precision,recall and F value are designed for evaluation.The evaluation results show that the accuracy of the model is 92.9%,and the accuracy,recall and F value of the overall test set are both above 90%,which proves that the diagnostic model has a good diagnostic effect.Finally,the research content of the full text is summarized,and the research direction of the next step is prospected.