Process monitoring via enhanced neighborhood preserving embedding

A novel process monitoring scheme named enhanced neighborhood preserving embedding (ENPE) is proposed. Neighborhood preserving embedding (NPE) only considers the reconstruction error on the basis of each local neighborhood is linear. For the purpose of addressing both the reconstruction error and the distance, the dual weight matrix and the enhanced objective function are constructed in the ENPE method. Finally, under a numerical example and the Tennessee Eastman (TE) benchmark, the superiority of the proposed ENPE method is evaluated through comparing with principal component analysis (PCA) and NPE. •A novel process monitoring method-ENPE is developed.•The dual weight matrix is established considering the reconstruction error and the distance.•An enhanced objective function is constructed in the ENPE method.