A Novel Fault Diagnosis Technique for Wireless Sensor Network Using Feedforward Neural Network

In recent times, research on fault diagnosis for wireless sensor networks (WSNs) has been a challenging task. The sensors (SNs) in WSN are deployed in the environment through volcanoes, forests, highways, etc., for extended periods; hence, they are subjected to frequent and unexpected faults. Existing fault diagnosis techniques such as distributed, hybrid, and centralized require transmitting the data sensed by an SN to their neighboring SNs, cluster head, or base station to identify their fault status. Hence, these techniques are energy-inefficient and suffer from high communication overhead and delay, which decreases the overall lifetime of the network. Also, these existing techniques depend on the spatial–temporal correlation between the nodes to detect the faulty SNs in the network. Therefore, to address these issues, we propose a novel fault diagnosis technique where each SN detects its fault status using its own sensed data. The performance of the proposed approach has been compared with the existing fault diagnosis techniques to demonstrate its effectiveness. The experimental results show the efficiency of the proposed fault diagnosis technique over existing techniques in terms of the various performance metrics.