Fault diagnosis method based on improved discriminative dictionary learning

In recent years, the sparse representation-based classification method is successfully employed in pattern recognition. The learning of dictionary and the training of classifier in this method are usually two independent modules to reduce the methods identification accuracy. Here, a novel fault diagnosis method based on improved dictionary learning model was proposed, to integrate sparse coding discriminative error, classification error and reconstruction error. And this model was solved with the K-singular value decomposition (K-SVD) algorithm to realize the synchronization learning of dictionary and classifier. With this method, the original signal was decomposed firstly using the empirical mode decomposition, the features of time domain and frequency domain were extracted from the decomposed intrinsic mode functions to form faulty samples. Then, the training samples were input into the improved model optimized with K-SVD. Finally, the testing samples were identified by using the learned dictionary and classification weights. Experimental results showed that the proposed algorithm can not only be applied in small sample fault diagnosis problems, but also its robustness and classification performance are significantly higher than those of other algorithms.