WSN Nodes And Network Fault Diagnosis Base On United Neural Network

In recent years, the Internet of Things has been widely concerned, regarded as revolutionary technology of the information industry once again after computer, the Internet and mobile communications technology. As one of realization forms of Internet of Things, the WSN(wireless sensor networks) shows a good prospect. In the actual application process, WSN often works in a so complex, hostile environment as to fail by interference and corruption, which affects seriously the efficiency and quality of working. Meanwhile, for its limitations of the work environment and its own technical characteristics, these faults are difficult to overcome by people personally, and its diagnostic techniques are different from the traditional networks. Therefore, fault diagnosis and fault tolerance of node and network are an important part of the wireless sensor network research.According to the structure, features and the existing fault diagnosis method of wireless sensor networks, for diagnosing node and network faults in wireless sensor networks that may occur, the paper reasonably extracts fault symptoms signals. On this basis, proposes a kind of nodes and network fault diagnosis solutions, which is realized by united neural network comprised by two different functions neural network.The first stage neural network is to predicate output of the node sensor as a predictor, in order to detect WSN’s node sensor failure and generate a fault symptoms signal. To realize the first stage neural network, a new radial basis network structure of Elman neural network are proposed, and its training algorithm is derived. Then the sensor unit fault detection method based on radial basis Elman neural network is simulated through the establishment of a sensor fault simulation model.The second stage neural network using RBF neural network is used to classificate faults characteristic signal pattern, in order to discriminate the possible fault type of WSN. In order to enhance the efficiency of the neural network training, the paper propose a double-parameters real-coding of quantum genetic algorithm by doing thorough research to quantum genetic algorithm, and combine with the existing hierarchical genetic algorithm, to devise a hybrid hierarchical quantum genetic algorithm for learning of RBF classifier.To verify the fault diagnosis scheme, the manner by combining with computer simulation and physical experiments is used. Firstly simulates fault diagnosis method involving various techniques on the computer, then tests the whole in wireless sensor networks. Simulation experiments show that the wireless sensor network fault diagnosis method can detect simultaneously node-level and network-level faults in wireless sensor networks, and has a higher fault detection rate, fulfill the basic requirements of the practical application. This exploration, in theory, will play a certain reference role to further the development of neural network and evolutionary algorithms.