Experimental Study On Fault Diagnosis Method Of Centrifugal Fans Based On Statistical Pattern Recognition

In accordance with the problems of rotating machinery condition monitoring and fault diagnosis, several kinds of common mechanical faults is simulated in a large fan test-bed and the fault mechanism and its spectrum characteristic are researched. Based on which, a serial of problems are studied in conjunction with statistical pattern recognition techniques, such as initial feature obtaining, feature extraction, condition distinguishing, fault diagnosis, etc. the feature space abnormal state recognition method and the multi-KL subspace method are proposed. The value of the method has been showed in the fault diagnosis of fans. In the end, a fan performance monitoring and fault diagnosis system was developed in this thesis.