A key element of understanding the health of nuclear power plant reactors is the analysis of condition monitoring data generated during routine operation. As reactors age, there is the desire to extend their operational lives, provided it is still safe to do so, to ensure a continued supply of low-carbon base load generation. This project is examining the use of machine learning techniques to supplement existing diagnostic knowledge and operational expertise to provide enhanced understanding of reactor core component health. Case studies are being explored from two different reactor designs to develop novel semi-supervised machine learning techniques to aid with assessment of existing health and make predictions of future condition of key plant components. The specific novel challenges for the research focuses on a) dealing with the imbalance of data where the vast majority of the data represents normal, routine operation and little faulty data exists and therefore techniques to incorporate human expertise is required and b) providing suitable explicability of results within the regulated, safety related nuclear industry.