A Hierarchical Cluster-Based Fault Management Approach for Common Mode Failure Diagnosis in Wireless Sensor Networks

Document Type : Persian Original Article

Authors

1 Department of Computer, Islamic Azad University, Tehran, Iran.

2 Iran Telecommunication Research Center, Tehran, Iran.

Abstract

Inasmuch as sensor nodes are typically used in inaccessible environments, they are vulnerable and insecure against environmental destructive factors and against deliberate devastating attempts of enemies. Hence, fault occurrence in wireless sensor networks (WSNs) is deemed to be an unavoidable phenomenon. The main drawback of comparative fault detection methods are that in case more than half of the neighboring nodes are faulty or the nodes become faulty due to a common mode failure (CMF), they will fail to detect faulty nodes properly. Thus, in order to address this issue, the authors introduced a cluster-based hierarchical fault detection method which increases the influence of non-adjacent sensor nodes’ data in determining of sensor’s status. Therefore the proposed method not only compares the data of neighboring nodes but also compares the data of non-neighboring nodes at an upper layer in order to adopt the proper decision upon the status of the nodes. Since applying fault detection methods in determined intervals and static manner are considered as inefficient, in this paper, we put forward an intelligent and dynamic method to determine the appropriate time for the implementation of the fault detection algorithm; hence, the right time and the required number of the implementation of the algorithm are intelligently and dynamically specified and as a result, the network lifetime increases. The related simulations were carried out by means of Matlab software was conducted under different densities of the nodes and with differing probability of being faulty nodes. The simulations results indicated that the fault detection accuracy of the proposed algorithm is significantly high and its false alarm rate is noticeably low. The results obviously demonstrate that the proposed method is scalable. 

Keywords