Reactor pressure vessel embrittlement: Insights from neural network modelling

• Published in: Journal of nuclear materials 2018-04, Vol.502, p.311-322
• Main Authors: Mathew, J., Parfitt, D., Wilford, K., Riddle, N., Alamaniotis, M., Chroneos, A., Fitzpatrick, M.E.
• Format: Article
• Language: English
• Subjects: Artificial neural networks; Embrittlement; Irradiation; Irradiation damage; Light water; Mathematical models; Neural networks; Nuclear reactor components; Nuclear reactors; Predictions; Pressure; Pressure vessels; Radiation; Reactor pressure vessel embrittlement; Reactors; Steel
• ISSN: 0022-3115; 1873-4820
• DOI: 10.1016/j.jnucmat.2018.02.027

Irradiation embrittlement of steel pressure vessels is an important consideration for the operation of current and future light water nuclear reactors. In this study we employ an ensemble of artificial neural networks in order to provide predictions of the embrittlement using two literature datasets, one based on US surveillance data and the second from the IVAR experiment. We use these networks to examine trends with input variables and to assess various literature models including compositional effects and the role of flux and temperature. Overall, the networks agree with the existing literature models and we comment on their more general use in predicting irradiation embrittlement. •ANN models are developed to predict irradiation embrittlement based on different surveillance datasets.•Dependence of embrittlement on a range of variables have been examined.•Performance of the model are assessed against various literature models.•Neural network predictions perform with a similar level of uncertainty compared to literature-based models.