On the role of numerical calibration in real-time recurrence CFD (rCFD) simulations of multiphase flows
Sprache des Vortragstitels:
Englisch
Original Tagungtitel:
11th International Conference on Conveying and Handling of Particulate Solids
Sprache des Tagungstitel:
Englisch
Original Kurzfassung:
rCFD has been shown to speed up pseudo-periodic flows like fluidized beds by three to four orders of magnitude. While early success stories just focused on the bright side of rCFD by emphasizing on its fascinating computational efficiency, this methodology still suffers from a series of potential pitfalls.
To start with, the existing version of rCFD exhibits significant propagation errors in nearly stagnant flow regions. In order to overcome this problem, an automated numerical calibration procedure for cell-to-cell shifts is introduced, which is based on a novel discretization scheme for the Lagrangian propagator of fluid flow. In a first lid-driven cavity test case, we prove that this new version of rCFD reduces propagation errors significantly. Finally, this new method is applied to secondary gas injection in a fluidized bed.
Next, we discuss the case of non-passive propagation of temperature. Obviously, resulting gradients in the temperature field give raise to buoyancy driven natural flow, thus changing the original flow field. Consequently, any isothermal flow database will be violated. Based on a lab-scale demonstrator experiment, we show that in this case multiple databases are needed and we further discuss ways of how to establish and execute them.
Since multiple databases are prone to excess data requirements, we finally propose a grid coarsening concept for rCFD for the purpose of data reduction. This coarsened rCFD can be regarded as a spatially filtered representation of a high-resolution Lagrangian propagator. Based on a metallurgical process, we show the functionality of this data-reduction methodology.
While all of these potential pitfalls of rCFD still exist, we introduced conceptual counter-measures. We believe that these under-the-hood improvements will lead to an extended applicability of rCFD in the realm of process modelling.