Skip to Main Content
Existing vibration fault diagnosis methods' application was limited because the lack of randomness, fuzziness and the relevance between the two. So a new algorithm-cloud neural network(CNN) based on cloud transformation is presented in this paper. And the vibration fault diagnosis steps of CNN based on cloud transformation are: First, extract the spectral feature vectors in the frequency domains of the generation sets as training samples, and the digital characteristic of clouds of training samples are obtained by cloud transformation; Then the feature vectors are used as training samples and the digital characteristic of clouds as initial weight to train the CNN to realize the mapping relationship between spectral feature vectors and fault types, thus achieving the purpose of diagnosing faults. The result shows that the application of CNN based on transformation on vibration fault diagnosis is feasible.
Date of Conference: 27-31 March 2009