Bubbly flow is quite common in various natural and engineering phenomena. In particular, nuclear engineers are interested in fundamental understanding of the bubbly flow behavior due to its importance in cooling light water reactor cores. Given the extreme conditions and complex support structures in nuclear reactor cores, it is very challenging to study the flow behavior using high-fidelity experiments. Typically validated computational codes are chosen as practical tools for the thermal-hydraulic and safety analyses. As the new generations of nuclear reactors are being developed, more advanced modeling techniques are required to design safe and efficient systems.
Different from most simulation approaches, direct numerical simulation (DNS) employs no turbulence closure assumptions, which makes it a promising tool for model development. The major bottleneck of DNS was and remains to be the high computational cost, increasing exponentially with the Reynolds number. However, thanks to the on-going improvements in computer power, these computationally expensive simulations are becoming more and more affordable. Coupled with level-set interface tracking method (ITM), DNS can be used for the high-fidelity studies of two-phase bubbly flows with unprecedented details.
Meanwhile, another concern that arises is how one can best take advantage of the ‘big data’ generated from large-scale DNS and translate it into new knowledge. The traditional level-set method utilizes a signed distance field to distinguish different phases while the interface is modeled by the zero level-set. Although level-set method can distinguish gas bubbles from the liquid phase, it cannot recognize and track individual bubbles which hinders the collection of useful bubble information. As a result, the bubble tracking capability has to be developed to improve the data extraction efficiency.
In the present work, a marker field is created and advected for bubble distinction and extraction of detailed bubble parameters from the simulations. Each bubble in the flow gets assigned a unique ID, based on which the code will collect the corresponding bubble information. It has been demonstrated that bubble tracking capability can significantly improve the data extraction efficiency for level-set based two-phase flow simulations. Statistical analysis tools are also developed to post-process the recorded information about the bubbles to study the dependencies/correlations of bubble behavior with bubble local conditions. For example, in the pressurized water reactor (PWR) subchannel geometry investigated in this paper, bubbles are observed to experience different relative velocity when presenting at different distance from fuel rod surfaces. With proper grouping criterion, statistical analysis would allow introducing variable drag coefficient for bubbles based on their positions. These new insights are contributing to more accurate modeling of the multiphase computational fluid dynamic (M-CFD) simulations, and better prediction of two-phase flow behavior in engineering systems. Together with the analysis tools, bubble tracking capability will open a new door to study and understand two-phase flows.