A fault detection method for wireless sensor networks based on improved LTS regression algorithm

In wireless sensor network, a large number of cheap nodes are deployed in the uncontrollable environment. Therefore, the fault probability of sensor nodes in wireless sensor network is much greater than that in traditional network. According to the impact model, the impact of one event is significantly dependent on the distance between the sensor and event source. So in this paper, we propose a wireless sensor network fault detection method based on the improved LTS regression algorithm. After collecting the change of data of each node when an event occurs, improved LTS regression algorithm will select the best performance data to calculate a series of properties of event source. Then the theoretical value of each sensor would be calculated by those properties. According to the margin between theoretical value and actual value, the faulty sensor can be detected. Theoretically, the robustness of LTS algorithm ensures the stability and high accuracy of performance in this method before the failure rate reaches its break point 50%. Our experiments also demonstrate that this method performs well before the percentage of fault sensor nodes arrives to 50% mentioned above.