The Research on the Failure Mode Classification Based on Reconstructed Phase Space and GMM
by Han Xiao, (Mechanical Engineering School, Wuhan University of Science and Technology, P.O. Box 222, Wuhan, Hubei Province E-mail: coolxiaohan@163.com), Yourong Li, (Mechanical Engineering School, Wuhan University of Science and Technology, P.O. Box 222, Wuhan, Hubei Province E-mail: liyourong@wust.edu.cn), and Lv Yong, (Mechanical Engineering School, Wuhan University of Science and Technology, P.O. Box 222, Wuhan, Hubei Province E-mail: lvyong@wust.edu.cn)
Section: Symposium 9: Innovative Techniques in Structural Health Monitoring, pp. 2553-2559, (doi: http://dx.doi.org/10.1061/41096(366)238)
Access full text
Purchase Subscription
Permissions for Reuse 
| Document type: |
Conference Proceeding Paper |
| Part of: |
Earth and Space 2010: Engineering, Science, Construction, and Operations in Challenging Environments |
| Abstract: |
The fault signals of equipments are nonlinear and non-stationary and these characteristics are more significant with the fault level increasing. In response to this feature of condition signals, the means of phase space reconstruction is put forward to display signals’ dynamics characteristic of different failure modes. And each mode’s Gaussian mixture model of reconstructed phase space is built, and then the equipments’ failures are classified by Bayes classifier. An actual measurement of gearbox vibration signal is used for validation. The experiments results show that the gear’s fault modes can be identified exactly when the mixture model number is appropriate and the noise disturbance can be overcome by proposed algorithm. Compared with BP and Elman neural network classifiers, the proposed method is proved higher reorganization rate. |
|
