Multi-Fault Diagnosis of Rotating Machinery Based on Spatially Constrained ICA

Most conventional methods cannot get effective performance during multi-fault diagnosis of rotating machines. Blind source separate techniques have applied to this problem, but the results show some limitations. Some prior knowledge considering the spatial or temporal characteristics of signals can be incorporated into these BSS approaches. Here spatial topography information will be chosen as spatial constraints and combine them with FastICA algorithm to get spatially constrained ICA (SCICA) method. This new technique will deal with multi-fault signals of rotating machinery through simulation process. SCICA is effective to multi-fault diagnosis and can separate all the source signals.