Detection and Recovery of Faulty Nodes Using Clustering Optimization and Bayesian Analysis in Wireless Sensor Networks

An WSN may often consist of hundreds of distributed sensors to monitor, collect and transmit information from various environments. Signals are relayed from one sensor to another sensor until it reaches the sink node which could either be mobile or a central control unit. The breaking down of sensors would inevitably affect the transmission of signals or even cause signal lost. Hence it is normal that inactive nodes miss their communication in network hence split the network. The clustering optimization technique efficiently groups the nodes into clusters and selects a cluster head based on the maximum energy. The number of members is limited based on the communication distance and the number of cluster per network is also constrained to number of nodes in the network. The fault nodes are detected based on the signals collected from them and classified into scene fault and node fault. The data from the fault nodes are recovered by using Bayesian analysis of compressive sensing to the cluster head. The secondary cluster head is chosen based on energy level and it takes the position of cluster head in the case of failure of the head. This approach allows faulty nodes recover their data by selecting neighbourhood cluster properly and increases detection accuracy of fault diagnosis and reduces time complexity.