Fault Detection Mechanism Based On Neighborhood-Collaboration For Wireless Sensor Networks

Wireless sensor network with the characteristics of node device heterogeneity, dynamic topology, networking flexibility, is composed of a large number of sensor nodes, and it’s widely used in data acquisition. Wireless sensor nodes transmit data to the monitoring center in real-time sensing, collecting data, making the data monitoring center can be informed timely and effective information, in order to achieve remote monitoring, control and deployment capabilities. For the sensor nodes deployed in harsh and complex environments, they are more susceptible to the factors of humidity, temperature and other aspects of the environment and equipment, resulting in node failures frequently. Once a node failure occurs, it will have an impact on the overall system performance of the network, and therefore the research of how fast and effectively make the fault detection has very important practical significance.Currently, wireless sensor fault detection methods include centralized and distributed detection methods. In a centralized detection, the central node aggregates information from other nodes and make the decision, thus the central node can easily become the bottleneck of the network, and the centralized approach has obvious disadvantages in large-scale networks. With respect to the centralized approach, the distributed ways overcome the bottlenecks of the central node, so it has a relatively wide range of application. However, in existing distributed fault detection method, there still are some problems, such as fault detection accuracy to be improved, excessive communication consumption, excessive energy consumption and other issues.In this paper, about the wireless sensor network communication and data collaboration features, I proposed the fault detection algorithm based on neighborhood-collaboration and the fault detection algorithm based on improved principal component analysis method. In the detection method based on neighbor-cooperation, we use historical data to establish the residual model and forecast the normal data to detect whether new data is faulty in order to overcome existing shortcomings of the neighbor-collaborative detection algorithm, then judge the state of other neighbor nodes. The simulation results show that the algorithm not only improves the accuracy of fault detection, but also reduces the interaction of the message. In order to overcome the shortcoming of traditional PCA include a large number of noise and error data when extracting data features, this paper supposes the fault detection algorithm based on improved principal component analysis. The error sample is removed from the raw data, then get the model threshold through the adaptive model and determine whether the new data is failure.