Diagnosis of transformers based on vibration data

Abstract

Transformers with loose or deformed windings can fail during an external short circuit with loss of service and heavy maintenance costs. The transformers’ tank vibration technique potentially offers a decisive solution for an on line continuous assessment of the integrity of structural elements in transformers. In this paper, the influence of sensor location on tank vibration measurements is addressed by means of Support Vector Machine (SVM) algorithms. Laboratory tests have been carried out in different points of the tank, on a typical oil filled power transformer under two extreme conditions, tight and loose windings. Preliminary results of SVM analysis of tank vibration spectra showed that it is possible to correctly identify winding looseness during repetitive sensor installation.