Fault Diagnosis in Wireless Sensor Networks Using a Neural Network Constructed by Deep Learning Technique

Abstract

Sensor nodes in wireless sensor networks (WSNs) are randomly deployed in hostile environments. Real-time experience shows that sensor nodes are prone to faulty. Different faults of sensor nodes are inevitable due to internal and external influences such as adverse environmental conditions, low battery, calibration and sensor ageing effect. Since WSNs applications rely on the fidelity of data reported by the sensor nodes, it is important to detect a faulty sensor and isolate them. Most of the existing fault detection techniques in literature are statistical based which demands sensor domain knowledge and the data from the neighbouring sensors. There may be a problem of detecting a sensor fault by analyzing the sensor data in distributed approach is non-trivial since a faulty sensor reading could mimic non-faulty sensor data. Currently, machine learning algorithms have been successfully used to identify and classify various types of faults in WSNs to avoid such kind of problems. However, the application of deep learning (DL) methods has sparked great interest in both the industry and academia in the last few years. In this chapter, neural network methods will be used in fault diagnosis in WSN with DL algorithms. The focus on diagnosis of fault includes hard, soft, intermittent and transient types.