In this paper, a Rough Set based approach aided with S-Transform features have been proposed for classification of short circuit faults during impulse test of a transformer. The required winding current waveforms are acquired by emulating short circuit faults across different disc positions in the analog model of a 33 kV winding of a 3 MVA transformer using developed analog fault simulator. Significant features are extracted for identification of various fault characteristics. It was found that the features extracted from S-transform data were sufficient for efficient classification of different fault cases and also application of Rough sets with a minor modification helps in improving the classification accuracy percentage.