Design and Evaluation of Composite Fault Diagnosis Protocols for Wireless Sensor Networks.

Abstract

Wireless sensor networks (WSNs) are spatially distributed battery-operated devices interconnected wirelessly to support various applications. However, due to well-known deployment issues such as human-inaccessible environment, environmental hazards, the devices are susceptible to faults leading to failures. These faults can be either hardware fault or computational fault or sometimes both. The undesirable behaviors of the sensor node caused by these faults affect the computational efficiency and quality of service (QoS).Detection, identification, and isolation of faults in WSNs could improve assurance of quality, reliability, and safety. Automated fault diagnosis is a well-studied problem in the research community. Fault diagnosis is considered as a very challenging problem in WSNs research. The works in this thesis attempt to solve these problems. Mostly, the works concentrate on the design of fault diagnosis methodology to build highly accuracy detection method by considering the constraints and resources of WSNs. The simulations (using NS–2.35) and testbed experiments are conducted for the performance measurement of the proposed methods. The set of works that have been conducted in this thesis are summarized as follows.
A neural network based fault diagnosis algorithm is proposed for WSNs to handle the composite fault environment. Composite fault includes four different kinds of faults such as hard, soft, intermittent, and transient faults. The fault diagnosis protocol designed are based on (1) gradient descent and (2) evolutionary algorithm (gradient-free) approach. It detects, diagnose, and isolate the faulty nodes in the network. The proposed protocol works in four phases such as clustering phase, communication phase, fault detection and classification phase, and isolation phase. Furthermore, a feed forward neural network based on gradient descent is modeled for automatic detection of link quality in a sensor network. Simulation results show that the proposed protocol using gradient-free optimization performs better than the existing protocols in terms of detection accuracy, false alarm rate, false positive rate, and detection latency.
A composite fault diagnosis protocol is proposed for wireless sensor networks using statistical and neural network approach. The proposed protocol consists of three phases, such as clustering phase, fault detection phase, and fault classification phase to diagnose the composite faulty nodes in the WSNs. The protocol strategy is based on a timeout mechanism to detect the hard faulty nodes, and analysis of variance method (ANOVA test) to detect the soft, intermittent, and transient faulty nodes in the network. To test a method of probabilistic classification, a feed forward probabilistic neural network (PNN) technique is implemented to classify the different types of faulty nodes in the network. The performance of the proposed composite fault diagnosis protocol is evaluated. The evaluation of the proposed model is also carried out by the testbed experiment in an indoor laboratory and outdoor environment.
A lightweight and less-overhead approach is proposed to automatically diagnose hard and soft faults in wireless sensor networks. Precisely, a lightweight checksum method is implemented for hard or crash fault detection. Such a method is capable of detecting multiple hard faults within a single path with the help of a timeout mechanism. For diagnosis of soft faults such as permanent, intermittent, and transient faults, we implement the Anderson-Darling statistical method. The Anderson-Darling test analyzes how the sensor readings are fitted in a specific distribution for a tested significance level. To validate the hypotheses and implementation, many testbed experiments are conducted. These experiments essentially report performance of proposed methods. Some of the performance parameters include fault detection accuracy, false alarm rate, and false positive rate and these parameters have also been studied with varying fault probabilities in a sensor network. The most important and interesting observation is that the proposed lightweight schemes can diagnose both hard and soft faults in O(1) message complexity over the network, which makes the schemes adoptable in practice.
A graph-theoretic distributed protocol is proposed to detect simultaneously the faults and cuts in the WSN. The proposed approach is accomplished mainly in four phases, such as initialization phase, hard fault and cut detection phase, soft fault detection phase, and fault tolerance phase. The protocol is an iterative method where at every time iteration, the node updates its state to calculate the potential factor. We introduced two terminologies such as a safe zone or cut zone of the network. The proposed method diagnoses the different types of faulty nodes such as hard and soft permanent, intermittent, and transient faults with better detection accuracy. The proposed method follows a fault tolerance phase where faulty sensor node values would be predicted by using the data sensed by the fault free neighbors. The proposed method is evaluated with regard to various performance evaluation measures by implementing the same in the network simulator. The obtained results show that the proposed graph-theoretic approach is simple yet very powerful for the intended tasks. The experimental evaluation of the fault tolerance module shows promising results with R-squared of 0.99. For the periodic fault such as intermittent fault, the proposed method also predicts the possible occurrence time and its duration of the faulty node so that fault tolerance can be achieved at that particular time period for better performance of the network.