Improving the efficiency of MELCOR modelling for accident analysis of a tokamak
Fusion power is a critical part of the future energy landscape yet is all too often depicted with the focus on the fusion reaction itself with little attention on supporting systems. In practice of course a real power plant will have the tokamak surrounded by a plethora of ancillaries: the fuel cycle, cooling systems, diagnostics, radiation management, controls, and so on. Understanding the consequences and mitigation for possible accidents is crucial for public acceptance but requires navigating this complex web of support systems. MELCOR is a software tool for accident analysis with a long history in the fission sector. A model has been developed for use in MELCOR, using proposed layouts of DEMO and assumptions based on current design choices, to facilitate safety design and development. We have streamlined the running of MELCOR tasks to allow simultaneous execution on a computing cluster. The input decks are version controlled within UKAEA’s internal GitLab repository, and shell scripts wrap the separate MELCOR invocations, embed specific job parameters, and keep separate the inputs and outputs of multiple separate instances. We illustrate the improved workflow with two examples studying tritium release following a Postulated Initiating Event of a guillotine break in the pumping line of a torus pumping system: a study of the effect of various locations for safety relevant valves around the system, and a 20-point parameter sweep looking at the effect of varying delays in valve closure times on the amount of tritium released from primary pipework. Results of the sensitivity analysis for valve closure delay indicate counter-intuitively that as the delay initially increases up to 120 s the total amount of tritium released decreases; with response delays of 180 s and longer the tritium release increases as expected. This parameter sweep outcome demonstrates the advantage of readily running MELCOR tasks in bulk.