Distributed Traversal Based FaultDiagnosis in Wireless Sensor Network

Abstract

Wireless Sensor Networks (WSNs) have become a new data collection and monitoring solution for a variety of applications. Wireless sensor networks (WSNs) are widely used in various real-life applications where the sensor nodes are randomly deployed in hostile, human inaccessible and harsh environments. Faults occurring to sensor nodes are common due to the complexity of sensor device and the harsh environment where the sensor nodes are deployed. One of the major research focus in wireless sensor networks in the past decades has been to diagnose the sensor network of faulty nodes. This helps to provide service of the network without interruption despite the occurrence of failure due to harsh conditions. Some of the issues concerned to fault diagnosis in wireless sensor networks have been addressed in this work mainly focusing on improvement of detection accuracy, reduction of message overhead, reduction in consumption of energy, and robustness to erroneous data by using statistical methods. The fault diagnosis techniques are classified based on the methods they employ to determine the faults. Here we have proposed a traversal-based diagnosis algorithm that seeks to diagnose both permanent as well as intermittent fault. The proposed algorithm employs an anchor node to traverse the field. The traversal of the field is decided by the traversal algorithm taking into consideration the length and breadth of the sensor field and transmission range of the nodes.

The anchor node stops at defined positions in the field where it executes the fault diagnosis algorithm in its neighbourhood. The diagnosis algorithm uses timeout method to identify hard faults and adjusted boxplot method to identify permanent and intermittent faults. The adjusted boxplot method takes into consideration the skewness of the data generated by the nodes in the sensor field.Real life data are not perfectly symmetric,so we have taken care of this fact and given independence to the algorithm in terms of data assumption. The proposed algorithm is simulated in Omnet++ environment which shows very promising results. Detection accuracy, false positive rate, false alarm rate and energy consumption shows an improvement over the existing algorithms.