RTDs
In the first part of a two-part study by Yu et al. (2014), the cold flow operation of ICFAR’s downer gas-solids separator design was simulated numerically using the commercial computational fluid dynamics (CFD) code ANSYS Fluent. The work by Yu et al. (2014) is provided for reference in Appendix E, which gives all relevant details of the theoretical framework on which the CFD model was based. The goal of the first part of the study was to improve upon the preliminary CFD simulations by Huard (2009) and to validate the simulated multiphase hydrodynamics and solids collection efficiency of the gas-solids separator by comparing with experimental results. The solids collection efficiency predicted by CFD by Yu et al. (2014) was in good agreement with the experimental results. Yu et al. (2015) then implemented heat transfer and reaction kinetics equations in the second part of the same CFD study. The gas RTD was simulated for the entire downer and gas-solids separator, as shown in Yu et al. (2015), while the gas RTD for the gas-solids separator only was performed by X. Yu at Aston University in the United Kingdom. The CFD separator RTD was shared for comparison with experimental results in this thesis.
In short, the multiphase gas-solids hydrodynamics in the downer and separator were simulated using an Eulerian-Eulerian method, wherein both phases were treated as continuous interpenetrating media. This method is contrasted with the Eulerian- Lagrangian method, wherein the solids phase is treated as a discrete dispersed phase, which is more physically meaningful than in the Eulerian-Eulerian method. However, the Eulerian-Lagrangian approach is limited to very small solids volume fractions and neglects particle-particle interactions (Cortés & Gil, 2007). For this reason, and for the ability to simulate the multiphase flow at high solids concentrations, the Eulerian- Eulerian approach was deemed more appropriate and was thus selected for the CFD simulations.
To simulate the gas RTD in the separator, a post-processing massless tracer particle method was used. The massless tracer method, described in detail by Yu et al. (2015), Mellin et al. (2014), and Aubin et al. (2009), involved the release of a large number (~ 500) of massless particles distributed uniformly over the downer cross section at the same height where the experimental tracer sparger was located (14 cm above the gas outlet pipe). The simulated tracer particles released at the same height as the actual tracer sparger followed gas streamlines that were calculated from the previously solved gas velocity field in the actual simulation.
The simulated gas velocity field and trajectories of a sample of all tracer particles in the downer and gas-solids separator are shown in Figure 3.22, which were performed by Yu et al. (2015). In the cold model scenario, ~ 90 % of the tracer particles successfully reached the gas outlet, while the remaining ~ 10 % of tracer was retained and recirculated in the downer for the entire period of simulated process time (~ 8 s). It is interesting to note that the reverse gas flow pattern observed below the gas outlet (Figure 3.22(a)) was very similar to the original simulations performed by Huard (2009). Furthermore, the predicted tracer particle path lines showed a very sharp reversal just below the cone rim, which indicated that the majority of the gas did not penetrate far below the gas outlet, as demonstrated by the experimental gas RTD and local gas velocity experiments described in Section 3.4.8.
Figure 3.22 – Simulated (a) gas velocity field and (b) massless tracer particle path lines and residence time predicted by CFD (modified from Yu et al., 2015)
Figure 3.23 compares the experimental and CFD predicted gas RTDs at Ug = 0.67 m/s and for solids loading values in the range of 1.0 wt/wt to 20 wt/wt. The most striking feature observed in the experimental RTDs at all solids loading values was the sharp initial peak immediately after the downstep, which was not observed in the simulated RTDs. The sharp initial peak in the experimental results indicated very strong gas bypassing, with a significant fraction of tracer exiting the separator almost instantaneously after the downstep. The main reason explaining the presence of the sharp initial peak in the experiments but not in the CFD results was the tracer injection method. In the experiments, a strong rush of air from the downer was observed in the first instants after the downstep due to the sudden decrease in pressure in the separator. This pulse of air may have caused the sharp initial tracer bypass immediately after the downstep.
However, in the CFD simulation, the injection of tracer was “perfect” since the hydrodynamics were not disturbed by massless tracer particles. Therefore, it may be useful to modify the CFD simulations in future to approximate the tracer downstep injection used in the experiments, since a massless tracer method cannot be implemented experimentally.
In general, the CFD-predicted RTDs were much more uniform and narrower than the experimental RTDs. One main peak centred roughly at t ~ 0.2 s was observed in the CFD RTDs, as well as small secondary and tertiary peaks between 0.5 s to 1 s after the downstep. Similar, though much smaller, peaks occurring between zero to 1 s after the downstep were also observed in the experimental RTDs. Additional peaks occurring more than 1 s after the downstep were observed in the experimental RTDs, but not in the simulated RTDs. The main reason explaining the wider distribution of peaks in the experimental RTDs was that roughly 10 – 20 % of the tracer particles CFD RTDs were manually removed from the RTD calculation since they were observed to recirculate for a very long time in the separation zone or were backmixed into the downer. Therefore, if these particles were included in the CFD RTDs, additional peaks occurring later than 1 s after the downstep would also have been observed.
Figure 3.23 – Comparison of experimental and CFD predicted gas RTDs at Ug = 0.67 m/s and at various solids loading values: (a) m&s m&g =1.0 (b) m&s m&g =10 (c) m&s m&g =20