Transformer breather thermal image decomposition for fault diagnosis

Image processing techniques replace human vision in most of the applications that incur danger to human lives for maintenance purposes. As a step towards automation in condition monitoring, the proposed work involves, representing an acquired thermal image of a transformer breather showing temperature variations in the discrete domain, by transforming the image using wavelet transform. Symlet wavelet is used for transformation that captures the temporal information of the image pixels. The representative parameters of the transformed image are studied at various decomposition levels to get better insights for further analysis thereby enabling accurate condition monitoring of the transformer under study. The statistical findings of the image details at various levels of decomposition expose the variation in the distribution of temperature in and around the breather vent pipes, depending upon the transformer operating conditions. Significant differences are observed in the parameters due to transformer winding faults when compared with the parameters under normal operation. Extending the work to include various transformer faults will enable the fault diagnosis process more exhaustive.