Research On Link And Data Fault Diagnosis And Network Connectivity Recovery In Wireless Sensor Network

Wireless sensor network(WSN) is the core supporting technology of the Internet of things. It has been widely used in military, environmental monitoring, disaster assistance, industrial control, intelligent home and other areas. And it has become the focus research in the field of information technology. Compared with the traditional network, the WSN, which use data as its application center, has a large number of nodes, a wide range of distribution, limited communication and computing resources, and is often deployed in uncontrolled or even harsh environments. Under the influence of these internal and external factors, the network is more vulnerable to interference or failure, which would bring immeasurable loss. Therefore, it is challenging to detect and recover as quickly as possible from damaged WSN, which is the necessary means to ensure the reliability and stability of the network.With the support of the National Science & Technology Pillar Program “Research and demonstration on building energy system monitoring and optimization control technology” and others, the paper focuses on fault diagnosis and recovery mechanism for different fault types in WSN. In the context of practical applications, according to research ideas of established model, theoretical analysis, a new method, simulation and experimental testing, fault diagnosis and recovery mechanisms are in-depth discussed.Firstly, considering the transmission quality problem of communication links in WSN, we propose a loss link diagnosis method based on network tomography technology. The method combines the active and passive strategies of loss link inferences. It identifies good links using passive collection of end-to-end observations. Then it minimizes uncertain links through the logical reasoning of adjacent paths, and sends the improved unicast probe packet actively to infer the transmission characteristics of remaining uncertain links. Simulation results show that: compared with others, this method introduces combination of active and passive strategy, so it has higher fault detection rate. Meanwhile, the proposed unicast mechanism reduces the required amount of probe packets, and enhances the relevance of packets with the same strips, which is more suitable for the practical application of wireless sensor networks.Secondly, aiming at the problem of data abnormity in WSN, a data fault diagnosis method based on spatial-temporal correlation is proposed. With the temporal correlation of data, it uses a time series analysis method to determine the preliminary detection state for each node. With the spatial correlation of data, it uses adjacent nodes to obtain a comparison test result for the neighbors of each node, and adopts four rules to determine the final detection results. Simulation results show that the method can be applied to the practical application, and its four rules give different detection accuracies. Compared with others, the proposed method has high fault detection rate and low false positive rate even when the large scale nodes are faulty.Thirdly, a relay node placement strategy based on the virtual force and the triangle Steiner point is proposed for disjoint segments with damaged WSN. The segment is considered as a collection of different nodes in the network, and it is more advantageous to optimize connection by relay nodes. The approach mainly includes two stages: maximizing coverage and establishing connectivity. It uses virtual forces to adjust damaged deployment for increasing coverage degree and reducing cost of connections, then find the proper triangular Steiner points or minimum spanning tree edges to achieve the optimal connecting these disjoint segments. Simulation results show that: compared with other methods, this method requires less relay nodes for recovery communication; with the increase the number of disjoint segments in the target area, the advantage more pronounced than others.As for restoring connectivity, a multi-objective optimized interconnection of disjoint wireless sensor network segments using mobile data collectors is also proposed. Firstly, the method uses clustering to segments detection. Then considering the delay of data collection and the equilibrium of collecting task, the network connectivity and data acquisition path optimization problem are formalized into an improved multi-traveling salesman problem, called nMTSP. It use improved multi-objective optimization genetic algorithm with virtual partitions, hierarchical chromosome structure, improved population diversity strategy, encoding and coding to solve Pareto solutions for nMTSP. Simulation results show that: this method provides a new idea, which can effectively solving Pareto optimal solutions for connectivity recovery. Compared with NSGA-Ⅱ, the improved population diversity strategy of the proposed algorithm can achieve significant improvements.Finally, according to the application of building environment data acquisition, wireless sensor nodes based on CC1110 are designed; the embedded gateway based on ARM9 is developed. Using them, we establish a prototype system of wireless sensor network through the CAN bus to achieve the network data transmission. And the proposed methods are evaluated by the experiment platform.In summary, the paper establishes the fault diagnosis and recovery methods for wireless sensor network, which can improve the survivability and reliability, and have positive significance and engineering application prospects. They can provide effective theory and technical support for application of the wireless sensor network in the fields of military security, industrial control, environmental monitoring and others.