The UK Advanced Gas-Cooled reactors (AGRs) have cores made of graphite bricks with dual functions: as structural elements of the core, providing space for and separating fuel and control rods; and as moderator of the nuclear reaction. Nuclear graphite is a quasi-brittle material, where the dominant mechanism for failure is cracking. While cracking of isolated bricks is expected due to operation-induced changes in graphite microstructure and the gradual build-up of stresses due to irradiation, these could be tolerated as far as the overall structural function of the core is maintained. Assessment of the whole core behaviour has been previously done with full-scale models where bricks have been considered as rigid body elements connected by springs. This approach does not allow for the realistic assessment of the stresses in the bricks and associated brick cracking. This thesis is dedicated to the development of the first physically realistic full-scale model of AGR reactor cores. The work explores the existing capabilities of the commercial software ABAQUS to represent the complexity of the core in terms of the geometries of various components, their interactions, and different formulations of graphite mechanical behaviour. The resulting model is shown to resolve the evolution of strains, stresses, and damage during reactor operation with higher fidelity than previous models. The simulation results presented in the thesis illustrate how the developed model can be used in a fitness-for-service assessment. One set of results show the gradual build-up of damage in the graphite core over 30 years of reactor operation, as well as how damage evolves in different regions of the core. This information is useful for planning targeted inspections, which can then be used for model validation. A second set of results show the changes in the channels diameters and alignments, which is essential information for deciding on continuing safe operation of the reactor. Once the proposed model is validated or recalibrated with operational data on local damage, the proposed methodology will become a trusted predictive tool.
|Date of Award||31 Dec 2021|
- The University of Manchester
|Supervisor||Andrey Jivkov (Supervisor), Graham Hall (Supervisor) & Paul Mummery (Supervisor)|
- Large-scale Modelling
- Finite Element Modelling