Online recurrence-CFD: enhancing multiphase flow simulation performance with online recurrence analysis

This study proposes a novel approach to accelerate computational fluid dynamics (CFD) simulations by employing simultaneous recurrence analysis. The methodology involves creating recurrent trajectories by recording the average values of samples through a probability sampling approach. These recurrent trajectories are then used for both qualitative and quantitative recurrence analysis in the simulations. A solver with a user-defined condition for detecting recurrent patterns is employed to activate recurrence-CFD (rCFD). By utilizing the previous hydrodynamic results of the recurrent trajectory, rCFD accelerates the simulation by a factor of 3 for 20 s of simulation time of a lab-scale air-water bubble column. The results show excellent agreement between the hydrodynamic fields obtained from rCFD and regular CFD simulations. Furthermore, the performance of rCFD was evaluated by solving a scalar transport equation using hydrodynamic fields for both CFD and rCFD, and an average correlation coefficient of 0.98 was observed between the results of CFD and rCFD. The proposed approach has the potential to significantly reduce the computational time required for CFD simulations in various applications with a general recurrent.