Abstract:A wide range of applications have started to use Wireless Sensor Network (WSN) as an information collection and monitoring tool. Reliable and accurate performance of sensor nodes is necessary in critical applications. For characteristics of the WSN data stream environment, the limitations of conventional principal component analysis (PCA) method which depend on the fixed model in practical application are analyzed, an online fault detection framework in WSN nodes based on recursive PCA (RPCA) model is proposed. The algorithm applies RPCA algorithm to sequentially update the model representing normal behavior of the sensed data adaptively and realize the online fault detection. Experiments with real data show that our online fault detection algorithm not only tracks the normal changes well, but also achieves good detection performance for typical node faults.

Adaptive and Online Fault Detection Using RPCA Algorithm in Wireless Sensor Network Nodes