Dynamic Eccentricity Fault Detection in Synchronous Machines Using Principal Component Analysis

The paper proposes a new method for detecting dynamic eccentricity faults in a Salient Pole Synchronous Machine (SPSM). Several methods have been documented in literature for detecting dynamic eccentricity, which includes using search coils, motor current signature analysis (MCSA), and data-based techniques. The former approach is invasive, thereby increasing installation cost, while MCSA is affected by load variations. A major hurdle in using data-driven methods is the selection of features. To overcome these limitations, the authors propose performing principal component analysis (PCA) on the fundamental sideband frequencies of motor current signals to detect and isolate dynamic eccentricity from static eccentricity faults. Principal Component Analysis (PCA) is a data-driven technique that can extract meaningful features in a dataset by transforming correlated variables into a reduced set of uncorrelated variables through a linear transformation. Experimental data of stator currents from a 2-kW, 208 V, 3-phase SPSM was used for the analysis. Results showed that the method isolated dynamic eccentric faults irrespective of the loading condition of the SPSM.