This project is concerned with using machine learning to identify potential operational issues among National Grid's 700+ 400/132 kV autotransformers. One tool for condition assessment is dissolved gas analysis of oil samples from operational transformers. Oil samples are typically taken at 6 monthly intervals and the aim of this project is to establish from analysis of historical data whether it is possible to predict ultimate modes of failure, whether there are similarities in development of degradation mechanisms because of manufacturer, operational age, location on the network etc. and ultimately if it is possible to use current measurement data and previous data trends to predict future issues within the fleet.