Temporal-Spatial Information Fusion Method of Fault Diagnosis Based on Clustering Means Centralization

Abstract:

Aiming at the uncertainty of fault signal and strong nonlinear trait of fault features,A Temporal-Spatial information fusion method of fault diagnosis is presented based on clustering means centralization. Extracting multiple template models from different types of fault,the means of which is regarded as the base point of certain clusters. Calculating the Euclidean distance between the fault features to be detected from different sensors and base point of all clusters,the reciprocal value of which is normalized to be Basic Probability Assignment (BPA) of every fault. With a period of continuous observations, multiples of BPA are combined with Dempster rule in every group of sensors in the time domain,then the results of every group combine again to get the final diagnostic results. Finally the case of motor rotor shows the effectiveness of proposed method.

• Periodical: Applied Mechanics and Materials (Volumes 635-637)
• Pages: 1639-1642
• DOI: https://doi.org/10.4028/www.scientific.net/AMM.635-637.1639
• Online since: September 2014
• Authors: Jian Jin, Guo Chu Chen*, Yu Fa Xu
• Keywords: Basic Probability Assignment (BPA), Evidence Theory, Fault Diagnosis, Temporal Spatial Information Fusion

References

[1] D.Q. Zhu and Y.A. Liu: J. Control and Decision. Vol. 22(2007), p.1321(In Chinese).

[2] J. Han and Y.G. Tao: J. Chinese Journal of Scientific Instrument. Vol. 21(2000), p.644(In Chinese).

[3] X.B. Xu, C.L. Wen and Y.C. Wang: J. Journal of Electronics &Information Technology. Vol. 31(2009), p.1635(In Chinese).

[4] Smets P: J. Information Fusion. Vol. 8( 2006), p.387.

[5] X.B. Xu, C.L. Wen, H.N. Jiang and Y.C. Wang: J. Chinese Journal of Scientific Instrument. Vol. 31(2010), p.334(In Chinese).