Recurrence CFD (rCFD), aims at an efficient representation of long-term processes which slowly evolve on highly-dynamic pseudo-periodic flow fields. In the framework of rCFD, short-term full CFD simulations deliver recurrence databases of the governing flow. Based on statistical reasoning, rCFD then exploits these databases in order to evolve generic flow fields, which subsequently serve as basis for the long-term process under consideration.

In transport-based rCFD, transport processes are modelled by a time-series of cell-to-cell shift patterns and face swap patterns, representing convection and diffusion. In applying rCFD to a set of single-phase and multiphase flows, we experienced a computational speed-up of up to four orders of magnitude.

Despite these successes, rCFD still faces open challenges, especially in cells of slow velocity where convection can hardly be represented by cell-to-cell shifts. In such cells, propagation of information is only enabled by isotropic face swaps, which might deteriorate rCFD predictions.

In this talk, we therefore present a novel numerical calibration method, aiming at levitating this existing limitation of rCFD. Based on a continuous displacement fields, which are extracted from full CFD simulations, we calibrate our cell-to-cell shift patterns such that information propagation is guaranteed in all cells, regardless of their flow velocity (see Fig.1).

After presenting this novel calibration methodology and thoroughly discussion first results of a lid-driven cavity test case, we apply our calibrated rCFD to the industrial process of grade change in a steel tundish.

Fig.1: Numerical simulation result of species propagation in a cubical lid-driven cavity – initially the left side was filled with red and the right side was filled with blue: (left) full CFD simulation, (middle) standard rCFD simulation with unfavorably small time-step width and (right) calibrated rCFD simulation with the same time-step width.