Research On Multi-Source Data Fusion In Wireless Sensor Networks

ABSTRACT:With the rapid development and extensive application of Wireless Sensor Networks (WSNs), the data generated in WSNs have increased greatly. How to implement data fusion in WSNs has become one of the most important research topics in WSNs. In the complex Wireless Sensor Networks, because of the limited sensor energy, the vast and variety data in the communication of WSNs, it is necessary to undertake the fusion to the multi-source data, which can reduce the amount of communication and improve data accuracy. Additionally, data fusion can prolong the lifetime of WSNs.After analyzing the theoretical basis about the multi-source technology, this paper studies several important issues applying Dempster-Shafer theory of evidence (DS theory) in the process of multi-source data fusion in Wireless Sensors Networks. The main work of this thesis is stated as follows:(1)Based on the DS theory, the author proposed a new belief assignment strategy, which can be applied in the Wireless Sensor Networks. Considering the reasonable impact brought by the distribution characteristics of the known sample data to the belief assignment result, this paper put forward the belief assignment strategy applying Mahalanobis distance as the distance calculation method between the measured target and the sample date set. Moreover, taking into account closed world assumption and open world assumption, this thesis constructed the basic belief assignment function based on the exponential function and designed the corresponding belief assignment algorithm.(2)The paper proposed a new representing method of conflict evidence based on the law of cosines, after importing the concept of vector space. The author introduces several existing classical identification method of conflict among evidence and analyzes the characteristic and deficiency of them. Considering the computation complexity, this thesis proceeds dimensionality reduction to the belief assignment vector applying Pignistic probability function. Additionally, the author uses angle in vector space to express the conflict between the evidences.(3)In order to undertake the data fusion in conflict evidence, using the evidence weighted average thought for reference, the author improves a data fusion algorithm based on the supporting degree between the evidences. After calculating the supporting degree of evidences applying the law of cosines, this paper constructed the correlation matrix of evidence and got the weights of credibility about evidence. Finally, the thesis applied the Murphy’s method to complete the fusion of conflict evidence.At the end, according to the requirements of rail traffic safety operation, the author designs the rail traffic transformer fault diagnosis system based on the evidence theory multi-source data fusion strategy. The paper completes the functional architecture design, the database design and the core function module design. Finally, this fault diagnosis system was verified by empirical simulation and achieved good effects.