Fault Diagnosis of Rolling Bearing Based on Fisher Discrimination Sparse Coding

In response to mechanical fault in feature extraction problem, this paper presents a Fisher discrimination sparse coding method. This method is achieved by optimizing an objective function that includes two steps. First, this objective function works well in denoising where signals need to be reconstructed. Second, another objective function is added to the sparse coding framework, the discrimination power of the Fisher discriminative methods with the reconstruction property, and the sparsity of the sparse representation that can deal with the fault signal which is corrupted. Finally, the feature is extracted. In rolling bearing fault classification experiments, the new method improves the accuracy of classification.