Reference-free detection of spike faults in wireless sensor networks

In recent years, wireless sensor networks have been applied to many applications. These sensors are usually simple, low-cost devices, deployed in large quantities, and prone to failure. This paper presents a model-free and reference-free spike fault identification method based on pair-wise verification. When the input of a system comes from a common source, there is a linear relationship between the output of any pair of sensors. This linear relationship between sensor pairs can be obtained through training. We present a method which is able to find faulty sensors suffering from sparse spikes in their outputs by pairwise comparisons even though there is no knowledge of which sensor is normal or abnormal, and no knowledge of the common input. The performance and limitations of the algorithm are discussed. Simulation results show that our algorithm has good performance, even when the spike fault power is comparable to the output signal power or observation noise power.