The Mechanisms And Agorithms Of Fault Management For Wireless Sensor Networks

Wireless sensor networks (WSNs) is a resource-constrained ad hoc network., The related technologies developed in recent years make it is applicated in widen fields. It is often deployed in extreme environment to collect data. The power, storage and computation ability of sensor nodes are limited, the communication way is multi-hop transmission, and then the sensor nodes have higher fault rates than traditional network nodes. Therefore, it is a challenge task to guarantee the data service without additional overhead. The efficient fault management is helpful to achieve the goal.Fault management of WSNs includes fault detection, fault diagnosis and fault recovery. The main task of the fault detection is probing or monitoring network to obtain the failure information. Fault diagnosis is responsible for the analysis of fault information to identify causes of failure. Fault recovery is processing the network to ensure the network operation normally.Most of the existed fault management methods for WSNs are followed the managent technique for traditional networks, without taking consider of the limited resource, multi-hop communictiaon and poor work environment of WSNs. The thesis studied the mechanisms and algothms of fault management for WSNs, focused on the fault detection and fault recovery. The main content and contributions are:1) For the additional overhead upon by fault detection in WSNs. Considered the node position distribution and fault distribution characteristic. Proposed a Simple Random Sampling based fault probing mechanism. First verify the applicability of the Pareto principle in wireless sensor network fault management. Then, according to the characteristics of the node distribution in accord with the Poisson distribution, we used simple random sampling to select the probe position. The probing tasks are evenly distributed in different nodes. Combined with the distribution characteristic of time of nodes, we adjusted the probing frequency dynamicly. Under the premise to ensure high fault detection rate and low false alarm rate, balancing the load of probing, reducing the number of probes. Thereby prolong the network lifetime.2) The neighbor data anlysis based fault detection method was proposed with the data-centric characteristics of WSNs. We first introduced the concept of credibility level of the node to filter out untrusted nodes making, which is determined by coparing its data with historical data and neighbor data. The node was compared with the trusted node in neighbor to determine whether a node failure by yoting. The algorithm have a better fault tolerant without decreasing the fault detection rate.3) Failure of certain node can cause the network is divided into several disconnected areas in the WSNs nodes can be mobile. The proposed recovery mechanism is based on the movement of nodes to restore the network connection, and particular node is choosed to move to the location of the location of the fault node. In the process of selecting nodes, all the two-hop neighbor nodes of fault node are candidated nodes. The nearest node to the fault node is choosen from the candidated nodes with minimum degree to perform the reovery task. And the target location of the nodes will moved is caluated according to the location of neighbor nodes of the fault node. The algorithm aovid the problem of excessive number of cascading moved nodes causing by moving a node to the fault point, ease the problem of decreasing to coverage ability, when a node is moved to restore the network connection.4) We proposed the data prediction based fault recovery method to ensure that data service of WSNs will not be affected. The Hausdorff distance analysis was used to select the reference nodes to avoid the problem caused by node sparse. And calculate the credibility of the reference nodes to filter out untrusted nodes. Then take the exponertial smoothing method to fill data, adjust the calculation parameters with Hausdroff distance analysis. The proposed method recover the WSNs from the point of view of that the data services are not affected.