Data-Driven Fault Detection Based on Multi-Neighbors Ensemble Preserving Embedding

In this study, a fault detection method called multi-neighbors ensemble preserving embedding (MNEPE) for analyzing data local and global structure information is proposed. By extending the analysis of different neighborhood information for neighborhood preserving embedding (NPE), the proposed MNEPE integrates the results of reconstruction samples with different number of nearest neighbors to preserve spatial structure, and improves the traditional NPE to capture both local neighborhood information and global spatial structure information. In addition, due to the integration of multiple nearest neighbor reconstruction modes for sample reconstruction, the impact of monitoring results relative dependence on the optimal number of nearest neighbors can be avoided. Finally, case study of a benchmark dataset illustrates the effectiveness of MNEPE.