Feature Extraction for Fault Diagnosis of Machine based on Kernel Nonnegative Matrix Factorization

To accurately extract the feature, a new feature extraction method based on kernel nonnegative matrix factorization is proposed to extract the features of nonlinear data in feature space. The method can extract the feature by using the nonlinear high-dimensional mapping of the kernel function, and can deal with the nonlinear data. The method of feature extraction and fault diagnosis based on kernel nonnegative matrix factorization is given. The results of nonnegative matrix factorization, principal component analysis, kernel principal component analysis and kernel nonnegative matrix factorization are compared by UCI data. The application of rolling bearing in practice shows that the method is suitable for the extraction of fault features for equipment, and effectively overcomes the shortcomings of nonnegative matrix factorization and principal component analysis, and can improve the fault diagnosis performance.