K-means and k-centers clustering algorithms need to pre-configure the number of clusters and its clustering granularity are rough. To solve this problem, the AP clustering algorithm was applied to mechanical fault diagnosis field. The EEMD and approximate entropy theory were used to extract fault features from fault data set. Then the AP algorithm was used to discover the fault pattern from extracted fault features. Finally the new sample's fault type was diagnosed according to the clustering result. Experimental results showed that AP clustering algorithm could effectively improve the accuracy of clustering and improve the accuracy of fault diagnosis without pre-configuring the number of cluster centers.
展开▼
