Research Of Wireless Sensor Network Fault Diagnosis Algorithm

The wireless sensor network is a kind of network distributed system which is composed of a large number of intelligent sensor nodes deployed in the area of surveillance. However, due to the large number of low cost and resource constrained nodes, randomly deployed in uncontrollable regions, plus the WSN application environment is extremely complex and harsh WSN often appear all sorts of fault and its reliability is the important challenge of sensor network applications. Fault diagnosis is one of the key technologies to improve the reliability of WSN.For the existing models and methods in the dynamic and complex changeable in fault diagnosis for sensor network self learning and adaptive ability shortage problem, this paper intends to explore the artificial immune system of danger theory applied to WSN node fault diagnosis method, and puts forward an applicable to WSN system level fault diagnosis method.Firstly, the structure of WSN network is studied, and the relationship and cooperation between them are studied. Then, the WSN fault diagnosis process is described. And then, according to the problem of WSN fault classification, fault feature extraction and feature extraction, the mechanism of WSN node fault and system fault is analyzed.Aiming at the problem of WSN node malfunction, a new algorithm based on Immune Danger Theory WSN node fault diagnosis is proposed. The algorithm first established mapping model of antibody, antigen and WSN fault feature vector, using risk trigger threshold risk source recognition and genetic algorithm to generate antibody library, based on k-nearest neighbor classification method to construct multi antibody failure detector and fault classification, update antibody library by tracing the change of the fault data online. The algorithm has the features of self-learning and dynamic updating, which has the small calculation load.For the problem of WSN system fault, a fault diagnosis method based on time weighted K-nearest neighbor is proposed. According to the system fault mechanism to establish characteristic value, the temporal correlation of WSN system failure, and the design time weighted based fault diagnosis classification rules, the method establishes a model of the fault diagnosis of the system combined with the K-nearest neighbor method. The model has the ability of correcting the fault data offset of WSN system and improving the correct rate of fault diagnosis. The results show that the time weighted K-nearest neighbor method has better anti-jamming and adaptability, and the classification accuracy is high.Finally, through experimental design determined simulation software, to obtain experimental data, verify the effectiveness of the method and other methods are compared. The results show that the above two methods have good reliability, adaptability and higher classification accuracy.