Optimisation of the link volume for weakest link failure prediction in NBG-18 nuclear graphite

This paper describes the process for approximating the optimal size of a link volume required for weakest link failure calculation in nuclear graphite, with NBG-18 used as an example. As part of the failure methodology, the link volume is defined in terms of two grouping criteria. The first criterion is a factor of the maximum grain size and the second criterion is a function of an equivalent stress limit. A methodology for approximating these grouping criteria is presented. The failure methodology employs finite element analysis (FEA) in order to predict the failure load, at 50% probability of failure. The average experimental failure load, as determined for 26 test geometries, is used to evaluate the accuracy of the weakest link failure calculations. The influence of the two grouping criteria on the failure load prediction is evaluated by defining an error in prediction across all test cases. Mathematical optimisation is used to find the minimum error across a range of test case failure predictions. This minimum error is shown to deliver the most accurate failure prediction across a whole range of components, although some test cases in the range predict conservative failure load. The mathematical optimisation objective function is penalised to account for non-conservative prediction of the failure load for any test case. The optimisation is repeated and a link volume found for conservative failure prediction. The failure prediction for each test case is evaluated, in detail, for the proposed link volumes. Based on the analysis, link design volumes for NBG-18 are recommended for either accurate or conservative failure prediction.