A Study On Fault Diagnosis In Wireless Sensor Networks

Because of the wretched environment where wireless sensor networks deployed and low-cost of sensor nodes, the network will inevitably fail during running, and the efficiency and quality of the network will be reduced. Therefore, the fault diagnosis technology have emerged and developed rapidly with the development of wireless sensor networks and requirements of applications, improve.We analysised the cause of the failure and the possible consequences of the failure in wireless sensor networks, and we got the classification of component level failure based on different basises. We give the model of the failure which we concerned in this paper and the evaluation criteria of the performace of fault diagnosis. Currently, the most fault diagnosis technology used in wireless sensor networks achieved with distributed method because of the flexibility and self-organized of distributed method which is very fit for wireless sensor networks. But there are still serious shortcomings in existing distributed fault diagnosis technology, for example the bad energy efficient and fault diagnosis accuracy is worse in netwoks where the node fault possibility is higher.In order to solve the problem of too large energy consumption during the fault diagnosis process, we propose a tree topology based fault diagnosis algorithm (TDF algorithm). The cause of the energy consumption problem is the large scale of data communication during fault diagnosis process using voting mechanism. We introduce the tree topology network to distributed method to supply a reliable reference.The algorithm maintains high node fault detection rate and low fault alarm rate in wireless sensor networks under low node distribution density. First, the algorithm finds a good node with a multi-layer detection method, then, detects the status of other nodes by means of the status relation with the good node which is inferred by the parent-child relation in the tree topology, and up to achieves fault detection of the whole network. In ZigBee tree network, the simulation results show that the algorithm performs better in energy efficiency, and identify the faulty sensors with high accuracy and robustness. The most application is applied with ZigBee standard tree topology, so this algorithm can be applied to these applications easily without extra overhead.In multi-sensor networks, the Distributed Fault Detection algorithm (DFD) compares the sensing data among its neighboring nodes to achieve fault detection. However, in areas with high density of faulty nodes, the increase of faulty nodes will lead to a significant decline in the accuracy of fault detection. The cause of this problem is the voteing mechanism can't supply enough effective nodes for judgement of node. In this paper, we exploit sensor data correlation between multi sensors to improve performance of the DFD fault detection algorithm in the multi-sensor networks. The MDFD algorithm is proposed for multi-sensor network fault detection to enhance the accuracy of fault detection in the network area with high fault distribution. The performance analysis and simulation results demonstrate that the fault detection accuracy of MDFD algorithm is about 20%~30% higher than the DFD algorithm, MDFD algorithm is more suitable for fault detection in sensor networks with low node density and higher failure ratio.