An energy-efficient clustering algorithm with self-diagnosis data fault detection and prediction for wireless sensor networks

Wireless sensor networks are used to track and regulate physical conditions like temperature and the environment's humidity. Wireless sensor networks with their advanced features are adopted for many real-time applications. The limited capacity batteries usually power the sensor nodes. But the big challenge of limited battery capacity obstructs remote and inaccessible areas where their use is the most favorable. For extending the lifetime of the network, the battery should optimally utilize it for different operations. The requirement toward low complexity and low energy consumption motivate the wireless sensor networks' efficient clustering algorithm. The sensor nodes group into clusters; one sensor node is chosen as a cluster head, and communication to the sink node from the sensor nodes occurs through the cluster head (CH) node. In the proposed method, cluster heads are determined based on the sensor node's weighted metric. The sensor nodes are then self-adaptive by making correct decisions in real-time based on the sensed data, but detected information is often inaccurate due to some mechanical, wireless loss, and battery problems. The erroneous or irrelevant data should be overlooked to avoid unnecessary data transmission, which contributes to reducing the network's lifetime. In the neighborhood-dependent self-diagnosis fault detection technique, the faulty sensed data are filtered at the sensor node itself. In the data prediction algorithm, the filtered data are predicted at the cluster head. All the factors collectively contribute to enhance the network's lifetime. The lifetime improvement of the proposed approach is almost doubled compared with LEACH one time better than QLEACH and ECH, and 51% better than temporal approach.