The Study Of Data Fault Detection In Wireless Sensor Networks

Wireless sensor networks (WSNs) have attracted lots of researchers’ effort because of its potential wide applications, such as military, environmental monitoring, health care, etc.One of typical tasks of WSNs is to collect the interested data in the sensor field for further processing and analyzing. Due to the vulnerability of WSNs, it is necessary to apply specific methods for assuring the accuracy of data. In this thesis, two algorithms for detecting the data fault in WSN will be proposed.The first one is the data fault detection algorithm based on Bayesian (DBA). In this algorithm, the Bayesian network is introduced to compute the nodes’ failure probability. A node calculates its posterior failure probability based on its neighbor nodes’ failure probability and own failure prior probability. In order to improve the accuracy of the fault detection, border nodes are introduced to reduce effectively the effect of a large number of fault nodes. DBA is simulated and evaluated in NS-2 by comparing with the classic fault detection algorithm (DFD). The result of simulations show that DBA improves effectively the fault detection accuracy especially in the situation with a large number of faulty nodes.The other one is the data fault detection algorithm based on sampling which can infer effectively the type of data failure with Causality Diagram. Considering complexity of the application environment of WSN, the threshold of detecting need to be adjusted according to the changes of environmental condition. However, in the traditional methods of fault detection, the threshold of detecting has been set as a constant in advance, which will affect the applicability of the fault detection algorithms. In data fault detection algorithm based on sampling, the threshold of detecting varies with the changes in the environmental condition by sampling. This algorithm is also simulated and evaluated in NS-2. In the simulations, the impact of the sampling frequency on the fault detection accuracy and the impact of the noise size on the accuracy of fault detection accuracy are evaluated roundly.