Data-driven computational modelling of fracture of nuclear graphite bricks

英国先进气冷反应堆的石墨核芯出现裂缝是由于长期暴露在反应堆恶劣的环境中，以及材料性能的退化。格拉斯哥大学开发了一种独特的基于物理的建模能力来预测这种不稳定的裂纹扩展。这将被法国电力公司用于未来的安全案例。使用机器学习（ML）技术，统计分析和图论，学生将从实际操作环境中获得的数据进一步改进这些基于物理的模型。这种技术也将使砖的客观选择进行进一步的分析，这是不确定性的关键来源。还可以根据裂缝形态对砖进行分类。同样，作业数据将补充基于物理模型的数据，以改进延长寿命的决策。ML还将使模型参数的不确定性得以传播，并探索其影响。

Cracks in the nuclear graphite core of the UK's advanced gas-cooled reactors are the result of long-term exposure to the reactor's aggressive environment, and degradation of material properties. The University of Glasgow has developed a unique physics-based modelling capability for predicting this unstable crack propagation. This will be used by EDF Energy for future safety cases. Using machine learning (ML) techniques, statistical analysis and graph theory the student will further improve these physics-based models from data obtained from the real operating environment. Such techniques will also enable the objective selection of bricks to undertake further analysis, which a key source of uncertainties. It will also be possible to classify bricks in terms of cracks morphologies. Similarly, operational data will be supplemented with data from the physics-based models for improved decision making in life extension. ML will also enable uncertainty in model parameters to be propagated, and their influence explored.