Transformers are seldom loaded to their maximum capacity as per the existing industry practices. The ultimate goal of this research project is to develop a method for predicting the maximum dynamic loading capability without violating the thermal limits of the transformer's insulation. Dynamic loading must account for, at minimum, load magnitude and shape, the ambient temperature, the external cooling conditions and the thermal limits. This paper discusses methods of detecting irregularities in the cooling mode transitions for substation distribution transformers. The two HST and TOT models considered in this paper are the non-linear IEEE model and the model built using linear regression techniques.