The Research On Node Fault Diagnosis Algorithm Of Wireless Sensor Network

Wireless sensor network is composed of a large number of nodes deployed in the target area. The network system is highly integrated. It plays an increasing important role in modern society and has a good application prospect. With the expansion of WSN, nodes are easily broken down because of their manufacturing process, physical environment, limited energy and other factors in their life cycle, thus influencing the system construction. Wireless sensor network nodes are often deployed in some special areas where human couldn't directly conduct equipment maintenance directly. So it's very essential to get a practical fault diagnosis technology applied to wireless sensor network.Instead of the traditional diagnosis technology, in wireless sensor network node fault diagnosis technology, the particularity of network structure determines the requirement of energy efficiency. We must propose a fault diagnosis algorithm relied on WSN, and it not only can give full play the advantage of information exchange, bus also can avoid the disadvantage of large redundant information and complex link. This paper researched the level structure of WSN based on efficient clustering algorithm, and achieved the cluster node's accurate fault detection in WSN. Based on the information differences between the cluster nodes detected by the cluster head node, it can achieve better fault information classification.This paper makes a further study in two aspects: WSN node fault diagnosis and fault-tolerant algorithm. The nodes fault diagnosis algorithm can achieve real-time analysis of the operational status of the network nodes. When some nodes got failure or ran out of energy, they will be away from the network and will no longer send and receive information between each other. The fault-tolerant algorithm for WSN is to enhance the stability of the network, even if parts of the nodes get failure, it will still be able to improve network performance to maximize, including improve the coverage rate that may dropped by nodes failure and establish the stability of the routing between the nodes. The main research contents and innovation points in this paper are as follows:(1) Proposed a quantum wavelet neural network binding with WSN model, using the inherent fuzziness of quantum wavelet neural network and realizd the strong coupling sensor terminal nodes fault information identification. Quantum wavelet neural network introduced the quantum states superposing thought into wavelet neural network. It makes the hidden layer activation function be a linear superposition by several Morlet wavelet. Neurons can express more status and order of magnitude to improve convergence rate and diagnosis accuracy. Through the adjustment of quantum interval, the network can reduce the uncertainty in fault mode classification.(2) The data collected by sensor nodes in the complicated target area are often contain certain noise and interference. In order to eliminate the singular and trend data, and take digital filter, data preprocessing is often needed. This paper uses the smoothing Kalman filter and makes sample data keep original features well and tend to expect goal.(3) Aimed at the disadvantage of previous distributed node fault detection(DFD) algorithm's high computational complexity and greatly influenced by nodes failure rate, this paper proposed an improved clustering DFD wireless sensor network node fault diagnosis algorithm. By using the LEACH-DFD clustering algorithm proposed in this paper and detecting the abnormal nodes by cluster heads, each cluster using optimum threshold, the diagnostic accuracy can be improved.(4) After the completion of self-organization in WSN, some nodes failure will cause the coverage rate drop. This paper uses the global search ability and particle swarm optimization algorithm in the fault area to realize the re-cover parts of the fault nodes. So it can improve the coverage rate and ensure the connectivity between the nodes.(5) This paper establishes stable routing in the nodes based on Hamilton circle from graph-thesis. When Sink node access to Hamilton circle, the communication distance that all the nodes traverse to the Sink node will be minimal. So it can effectively save node energy. Even when more nodes fails caused by fault or energy exhausted and leave off the network, it still can establish stable routing structure to prolong the lifecycle of the WSN.