Research Directions on AI and Nuclear

Abstract

The development of applications and systems for the nuclear domain involves the interplay of many different disciplines and is, therefore, particularly complex. Additionally, these systems and their innovations have to be compliant with strict international regulations and recommendations. The scientific and industrial communities have been studying, developing and applying advanced Artificial Intelligence (AI) techniques and tools in several (non-nuclear) application domains. Their encouraging results have pushed the nuclear community to pay increasing attention to the field of AI. Among the expected benefits of AI is the simplification of complex procedures, the reduction in the execution of time-consuming operations, the increase of safety levels, and the reduction in the overall cost. At the French Atomic Energy Commission (CEA), we have identified and have started to address several open questions, such as: where in the nuclear domain can AI-based techniques be implemented in the most productive way? What do the nuclear standards and recommendations say about its use? Can we identify some core challenges and issues common to multiple areas of the nuclear domain? In this paper we provide a first analysis and answers to the above questions and we conclude by emphasizing some cross-domain high priority challenges.

Ähnliche Arbeiten