Science-based nuclear design and safety in the emerging age of data based analytics

Highlights: • This paper provides a review of the science- vs. data-based shifts in institutional investment in the U.S. • As the knowledge of nuclear power systems has increased, science-based solutions of these phenomena have emerged but are difficult to close. • Independent of analysis fidelity, data-based methods typically serve as surrogates to science when knowledge is incomplete. • Today's nuclear industry is primarily interested in the potential of modeling and simulation to extend plant lifetimes. • Multi-physics/multi-scale modeling and simulation can and is playing a valuable role related to aging and plant life extension. • The USDOE R&D investment needs to be balanced with short- and long-term industry needs. - Abstract: Over the many decades of nuclear power development, pursuits of reason and observation, appearing as science- and data-based evaluations, have alternated in priority by technologists as they mutually serve to refine the understanding and utility of nuclear science. The rapid advancement of both of these methods principles for nuclear applications, among others, has been synergistic with advances in computing. Stimulated by recent data analytics achievements in the realm of computer visualization, machine learning and the internet-of-the-things, this paper reviews the historical trends of science- vs. data-based analytics with a focus on nuclear-thermal-hydraulics modeling and simulation. The benefit of this progress is then presented in terms of its value to the nuclear industry. An underlying assumption is maintained that sustainable progress relative to present-day investment in research and development is only possible where benefits for the commercial nuclear power industry can be realized.