Multi-fault feature extraction method based on Manifold Learning

Aiming at the difficulty of signal processing and feature extraction in multi-fault mode, a multi-fault feature extraction method based on manifold learning is proposed. On the basis of maintaining the linear relationship between neighborhood sample points in low dimensional space, the nearest neighborhood sample points are found based on Euclidean distance through locally linear embedding (LLE) algorithm and the optimal local geometric embedding is obtained by linear combination. In the problem of regularization, by using the method of multiple weight vectors, the optimal solution is ensured to identify when it converges to the desired embedding. Finally, the multi-fault feature extraction is visually displayed in two-dimensional and three-dimensional images respectively to verify the effectiveness of the method based on the CMAPSS data set provided by NASA.