| Because the work environment of WSN(Wireless Sensor Networks) is always forbidden for human or bad even dangerous, all kinds of faults occur on the sensor nodes of WSN, affected by these faults, measurement on apperceived physical quantities are wrong, even some functions of the WSN are lost or the whole network is paralytic. So it is necessary for research on fault diagnosis on sensor nodes of WSN.In the paper, WNN (Wavelet Neural Network) is used to diagnose the sense module of the node.The results of diagnosis are fault detection and fault time locating. Fault detection is about detecting whether the node is fault, fault time locating is about registering the fault time based on signal singular points analysis. The main contain includes:1. Introducing two existent methods of fault diagnosis on WSN sensor node. The method of node fault diagnosis based on space relativity is expounded. The method of node fault diagnosis based on WNN is discussed.2. Deducing the training algorithm of wavelet neural network and making improvement for traditional BP algorithm.3. Making research on the model of sense module, fault model and representation.4. Obtaining source data which includes the data that sensor node works normally and the data that the sensor node goes wrong because of Additive, Calibration, Stuck-at-Value and Random Noise fault.5. Detecting fault and locating the time that the faults occur on sense module of the isolated node (have no redundancy with other nodes in WSN) based on WNN observation apparatus and wavelet decomposition.6. Detecting fault and locating the time that the faults occur on sense module of the redundancy senor nodes based on crossed WNN forecast apparatus.In the paper, WNN is used for node fault diagnosis on WSN. Though a large number of simulated examples, the satisfying fault diagnosis ability of WNN is testified. Exploring on theory of the paper will provide a reference for the development of fault diagnosis technology on WSN node. |
PDF Full Text Request