Research On Management Model And Fault Diagnosis Algorithm For Railway Wireless Mesh Networks

With its flexible structure and easy deployment, wireless mesh networks (WMN) are widely deployed along the railway to provide network access. However, several factors in the practical environment, such as the openness of the wireless channel and instability of the power supply, lead to node failures frequently. And the mobility of mesh nodes makes them difficult for management, which has greatly limited the development of railway mesh networks. Hence, an accurate and comprehensive network management system could monitor the network status effectively, while improve the network performance. To address this issue, a three-tier network management model and the fault diagnosis algorithm based on the decision tree for wireless mesh network (FDDT-WMN) are proposed in this work.This paper summarizes the research status of management systems and fault diagnosis algorithms for WMN firstly, through giving insight into typical management protocols and models. Then considering the characteristics and management difficulties of WMN, a three-tier network management model which is applicable for the banyan-tree topology networks is proposed. Simulation results have shown that it could reduce the administrative overhead and improve the network stability significantly.Based on the analysis of decision tree algorithms, the attribute data of WMN are classified into several groups, each of which is assigned with a weight. Then a fault diagnosis algorithm on the basis of improved C4.5-I algorithm, i.e., FDDT-WMN is proposed to establish a knowledge base of the data. And the decision tree and rule set are generated by training sample data. Experimental results show that it could diagnose and classify the faults accurately and efficient.After the faults get classified by using FDDT-WMN, a fault recovery strategy for different types of faults is proposed, which is based on the calculation rule of TQ in the BATMAN-adv routing protocol. And the simulation results demonstrate that it leads to much faster recovery.A management system with the fault diagnosis unit integrated is implemented using Django framework. And a testbed is set up to collect related experimental data, which have shown that proposed scheme could achieve better network management and be able to locate and recover the failure quickly.