Fault Detection in Wireless Sensor Network Based on Deep Learning Algorithms

Abstract

This study discusses fully distributed fault detection via a wireless sensor network. Initially, we suggested using the Convex hull approach to determine a range of extreme points including nearby nodes. As the number of nodes rises, the message's duration is constrained. Secondly, in order to enhance convergence performance and identify node errors, we suggested using a convolution neural network (CNN) and a Naïve Bayes classifier. Lastly, we use real-world datasets to examine CNN, convex hull, and Naïve bayes algorithms to find and classify the defects. Based on performance measures, the results of simulations and experiments demonstrate that the CNN algorithm has better-identified defects than the convex hull technique while maintaining feasibility and economy.