Elsevier

Computer Communications

Volume 31, Issue 14, 5 September 2008, Pages 3469-3475
Computer Communications

Fault detection of wireless sensor networks

https://doi.org/10.1016/j.comcom.2008.06.014Get rights and content

Abstract

This paper presents a distributed fault detection algorithm for wireless sensor networks. Faulty sensor nodes are identified based on comparisons between neighboring nodes and dissemination of the decision made at each node. Time redundancy is used to tolerate transient faults in sensing and communication. To eliminate delay involved in time redundancy scheme a sliding window is employed with some storage for previous comparison results. Simulation results show that sensor nodes with permanent faults are identified with high accuracy for a wide range of fault rates, while most of the transient faults are tolerated with negligible performance degradation.

Introduction

Wireless sensor networks are emerging as computing platforms for monitoring various environments including remote geographical regions, office buildings, and industrial plants [1]. They are composed of a large number of tiny sensor nodes equipped with limited computing and communication capabilities. Since low-cost sensor nodes are often deployed in an uncontrolled or even harsh environment, they are prone to have faults. It is thus desirable to detect, locate the faulty sensor nodes, and exclude them from the network during normal operation unless they can be used as communication nodes.

Fault detection and fault tolerance in wireless sensor networks have been investigated in [2], [3], [4], [5], [6], [7], [8], [9], [10], [11]. Jaikaeo et al. [2] presented techniques for sensor network diagnosis, where response implosion problem was addressed and overcome. Koushanfar et al. [4] have proposed a cross-validation-based technique for on-line detection of sensor faults, where statistical techniques are used to identify the sensors that have the highest probability to be faulty. System-level diagnosis based on the PMC model was presented in [6]. In [5] a fault detection scheme for an event-driven wireless sensor network using an external manager was presented. Although the external manager can perform more complex functions compared to the sensor nodes, communication between the sensor nodes and the manager may be problematic.

Ding et al. [8] have proposed a localized fault identification algorithm, where each sensor node compares its own sensed data with the median of neighbors’ data to determine its own status. The performance of the localized diagnosis, however, is limited due to the non-uniform nature of node degrees in sensor networks with random deployment. Fault-tolerant event boundary detection algorithm has also been proposed. In [3] Bayesian fault recognition algorithm was presented to solve the fault-event disambiguation problem in sensor networks. Luo et al. [11] have proposed a fault-tolerant energy-efficient event detection paradigm for wireless sensor networks. For a given detection error bound, minimum neighbors are selected to minimize the communication volume. Both Bayesian and Neyman-Pearson detection methods are presented.

A distributed fault detection scheme for sensor networks has been proposed in [9]. It uses local comparisons with a modified majority voting, where each sensor node makes a decision based on comparisons between its own sensing data and neighbors’ data, while considering the confidence level of its neighbors. The scheme, however, is a little complex in the sense that information exchange between neighboring nodes has to occur twice to reach a local decision based on a threshold. In addition, it does not allow transient faults in sensor reading and internode communication, which could occur for most normal sensor nodes. Transient faults in sensing and communication have been investigated in [10]. A simple distributed algorithm has been proposed to tolerate transient faults in the fault detection process. Some other fault management schemes can be found in the survey written by Yu et al. [12].

In this paper, we propose a distributed algorithm for detecting and isolating faulty sensor nodes in wireless sensor networks. Nodes with malfunctioning sensors are allowed to act as a communication node for routing, but they are logically isolated from the network as far as fault detection is concerned. It employs local comparisons of sensed data between neighbors and dissemination of the test results to enhance the accuracy of diagnosis. Transient faults in communication and sensor reading are tolerated by using time redundancy. Faulty nodes are isolated by correctly identifying fault-free nodes. Both the network connectivity and accuracy of diagnosis are taken into account since fault-free nodes isolated might be of little or no use even if they are determined to be fault-free, unless they can participate in the network via intermediate communication nodes with faulty sensors.

Section snippets

Fault model

In the fault detection of wireless sensor networks, we assume that all the sensor nodes have the same transmission range. Sensor nodes can be randomly deployed or placed in predetermined locations. Nodes with faulty sensors and permanent communication faults are to be identified. Sensor nodes which generate incorrect sensing data or fail in communication intermittently are treated as usable nodes, and thus are diagnosed as fault-free. Sensor nodes with malfunctioning sensors could participate

Fault detection

We first define communication and test graphs. A communication graph of a wireless sensor network can be represented as a digraph G(V,E), where V represents the set of sensor nodes in the network and E represents the set of edges connecting sensor nodes. Two nodes vi and vj are said to have an edge in the graph if the distance d(vi,vj) between them is less than l (transmission range). The communication graph can be a test graph in our fault detection if two nodes with an edge connecting them

Performance evaluation

The performance of the proposed fault detection algorithm is evaluated by computer simulation. It depends on various parameters: the number of sensor nodes in a target area (or average node degree), the probability that a sensor node is faulty p, and the thresholds θ1 and θ2. In the simulation, we assume that faults are independent of each other. The following two metrics, detection accuracy (DA) and false alarm rate (FAR) are used to evaluate the performance, where DA is defined to as the

Conclusion

In this paper, we have proposed a fault detection algorithm for wireless sensor networks. Each sensor node identifies its own status based on local comparisons of sensed data with some thresholds and dissemination of the test results. The algorithm is simple and detects faulty sensor nodes with high accuracy for a wide range of fault probabilities, while maintaining low false alarm rate. Moreover, it can tolerate transient faults in sensor reading and communication with negligible performance

Acknowledgements

This work was supported by Grant No. R01-2006-000-10073-0 from the Basic Research Program of the Korea Science and Engineering Foundation.

References (12)

  • I.F. Akyildiz et al.

    Wireless sensor networks: a survey

    Computer Networks

    (2002)
  • C. Jaikaeo et al.

    Diagnosis of sensor networks

    International Conference on Communications

    (2001)
  • B. Krishnamachari et al.

    Distributed Bayesian algorithms for fault-tolerant event region detection in wireless sensor networks

    IEEE Transactions on Computers

    (2004)
  • F. Koushanfar et al.

    On-line fault detection of sensor measurements

    IEEE Sensors

    (2003)
  • L.B. Ruiz, I.G. Siqueira, L.B. Oliveira, H.C. Wong, J.M.S. Nogueira, A.A.F. Liureiro, Fault management in event-driven...
  • S. Chessa, P. Santi, Comparison-based system-level fault diagnosis in ad hoc networks, in: 20th Symp. Reliable Dist....
There are more references available in the full text version of this article.

Cited by (0)

View full text