A novel fault analysis and diagnosis method based on combining computational intelligence methods. (August 2013)

• Record Type: Journal Article
• Authors: Deng, Wu; Yang, Xinhua; Liu, Jingjing; Zhao, Huimin; Li, Zhengguang; Yan, Xiaolin
• Is Part Of: Proceedings of the Institution of Mechanical Engineers. Volume 227:Number 3(2013)
• Page Start: 198
• Page End: 210
• Publication Date: 2013-08
• Subjects: Computational intelligence -- hybrid intelligence method -- fault analysis and diagnosis -- completeness -- complicated CNC system; Mechanical engineering -- Periodicals; Production engineering -- Periodicals; Manufacturing processes -- Periodicals; 621.05
• DOI: 10.1177/0954408912459161
• Languages: English
• ISSNs: 0954-4089

Abstract

In order to improve the correctness and efficiency of fault diagnosis, a novel hybrid intelligence method based on integrating rough set, genetic algorithms, and radial basic function neural network (RGRN) was proposed for motor fault diagnosis in the complicated CNC system in this article. In the proposed RGRN method, combination and condition supplement algorithm was used to deal with the incomplete fault data and the original data were discretized using genetic algorithms to construct a decision table. Rough set theory as a new mathematical tool was used to eliminate the redundant and irrelevant attributes in order to obtain the minimum rule set for reducing the number of input nodes of the radial basic function neural network. Genetic algorithms were directly used to optimize the structure and weights of radial basic function neural network to establish an optimized radial basic function neural network (GRN) model; then, the minimum rule set was inputted into the GRN model in order to obtain the optimized RGRN model. Finally, the completed fault symptom information was inputted into the RGRN model to obtain the fault diagnosis results. The robustness of the RGRN method was tested. Simulating experiments on motor fault diagnosis in the complicated CNC system show the RGRN method not only improves the global optimization performance and quickens the convergence speed, but also obtains the robust solution with a better quality.