This thesis participates to the effort launched after the Fukushima-Daiichi disaster to improve the robustness of national institutions involved in nuclear safety because of the role that the failing nuclear regulator had in the accident. The driving idea is to investigate how engineering techniques used in high-risk industries can be applied to institutions involved in nuclear safety to improve their robustness. The thesis focuses specifically on the Office for Nuclear Regulation (ONR), the British nuclear regulator, and its process for structured inspections. The first part of the thesis demonstrates that the hazard and operability (HAZOP) technique, used in the nuclear industry to identify hazards associated with an activity, can be adapted to qualitatively assess the robustness of organisational processes. The HAZOP method was applied to the ONR inspection process and led to the identification of five significant failures or errors. These are: failure to focus on an area/topic deserving regulatory attention; failure to evaluate an area/topic of interest; failure to identify a non-compliance; failure to identify the underlying issue, its full extent and/or safety significance and failure to adequately share inspection findings. In addition, the study identified the main causal chains leading to each failure. The safeguards of the process, i.e. the mechanisms in place to prevent, detect, resolve and mitigate possible failures, were then analysed to assess the robustness of the inspection process. The principal safeguard found is the superintending inspector who performs reviews of inspection reports and debriefs inspectors after inspections. It was concluded that the inspection process is robust provided recruitment and training excellence. However, given the predominant role of the superintending inspector, the robustness of the process could be improved by increasing the diversity of safeguards. Finally, suggestions for improvement were made such as establishing a formal handover procedure between former and new site inspectors, formalising and generalising the shadowing scheme between inspectors and setting minimum standards for inspection debriefs. These results were shared with ONR, which had reached the same conclusions independently, thus validating the new application for the HAZOP method. The second part of the thesis demonstrates that computational modelling techniques can be used to build digital twins of institutions involved in safety which can then be used to assess their effectiveness. The knowledge learned thanks to the HAZOP study was used in association with computational modelling techniques to build a digital twin of the ONR and its structural inspection process along with a simple model of a nuclear plant. The model was validated using the face-validity and predictive validation processes. They respectively involved an experienced ONR inspector checking the validity of the model’s procedures and decision-making processes and comparing the model’s output for oversight work done to data provided by the ONR. The effectiveness of the ONR was then evaluated using a scenario where a hypothetical, newly discovered phenomenon threatens the integrity of the plant, with ONR inspectors gradually learning and sharing new information about it. Monte-Carlo simulation was used to estimate the cost of regulatory oversight and the probability that the ONR model detects and resolves the issue introduced before it causes an accident. Different arrangements were tested and in particular with a superintending inspector reviewing inspection reports and a formal information sharing process. For this scenario, these two improvements were found to have a similar impact on the success probability. However, the former achieves it for only half the cost.
Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:744944 |
Date | January 2018 |
Creators | Lavarenne, Jean |
Contributors | Shwageraus, Eugene ; Weightman, Mike |
Publisher | University of Cambridge |
Source Sets | Ethos UK |
Detected Language | English |
Type | Electronic Thesis or Dissertation |
Source | https://www.repository.cam.ac.uk/handle/1810/277145 |
Page generated in 0.0015 seconds