Power Auto-transformer Mechanical Faults Diagnosis based on Frequency Response Analysis

Frequency response analysis (FRA) is a sensitive method established for testing the mechanical integrity of transformers. However, interpretation of FRA signature still requires specialist's advice, while FRA interpretation codes have not been approved officially. In order to improve interpretation of FRA test, different types of mechanical faults with various extents are required to be tested on power transformers. To address this challenge, in this paper a 3D-FEM model of a power auto-transformer was constructed and verified by experimental measurements. The extracted FRA signatures of FEM based various radial deformations and axial displacements were compared using two existing statistical indicators and two new condition indictors viz. the crest factor and frequency center. The studies demonstrated that frequency range between 1–100 kHz is mainly affected by the mechanical faults and the 1st resonance of the double-peak feature of the auto-transformer FRA remains unchanged during winding mechanical faults. Also the results showed that the crest factor is likely to be capable of diagnosing of fault type.