Collaborative Analysis of Multiple Fault Characteristics Based Fault Prognostic Analytics for Large-scale Generate Unit

With the rapid improvement of the degree of large-scare unit, and the increasing complexity of the connection between various equipment, the safety of production process is threatened by the increasing risk of the equipment fault. For the purpose of insuring that the large-scale unit operates reliably and safely, it is of great significance to do fault prognostic to know the health status of the device in advance. In this work, a novel approach of fault prognostic is presented for the large-scale generate unit. Firstly, multiple features, including slow-varying features, nonstationary features and nonlinear features, are extracted to characterize fault characteristics with slow feature analysis (SFA), co-integration analysis (CA) and kernel principal component analysis (KPCA). Then, the features closely related to fault degradation process, which are monotonic, robust, and correlative are chosen and fused into a DSSIM health index to reflect fault degradation process by dissimilarity analysis. Finally, q-step-ahead fault prognostic is achieved by partial least square algorithm (PLS). An actual industry process of the large-scale generate unit is provided to validate the feasibility and effectiveness of the proposed approach.