The literature review on CFD modelling of cyclonic separators (Sec. 3.5) revealed that the modelling approach for separators operating at low particles loading is generally
based on an Euler-Lagrange approach, and it will be used in the present study. Also as proposed by Newton (2007), if one phase dominates the volumetric flow rate in the vessel, then it should be possible to model the section of the vessel in which the highest volume fraction for the primary phase can be found. For such cases, the upper section of the Pipe-SEP will be modelled. Two assumptions are made in the present simulation study. The liquid droplets are simulated with one-way coupling method within the flow field of the model, and moreover the droplets do not interact with each other, and do not coalesce or break up. Due to the fact that the major objective of current study regards conditions with high efficiency Pipe-Hi-SEP systems operating with high GVF (>90 Vol%) incoming flow, the assumptions can be deemed to be valid. More to the point, the second stage separator Hi-SEP is usually subject to more than 95 Vol% gas flow. Another thing need to mention is that the Hi-SEP has similar geometry with the Pipe- SEP. To get realisable results with limited computer effort, only the Pipe-SEP will be simulated in detail. The Hi-SEP simulation will be used to validate the scale rule. The detailed components within the CFD simulations and the reason for selecting them are explained in the following sub-sections.
5.2.1 Computational Domain and Grid
The computation domains for the Pipe-SEP were developed based on the prototype used in previous experimental work (see Sec. 4.2). The Pipe-SEP section above the liquid level is simulated. The liquid level should always be kept below the inlet (200mm in this case). To explore the effects of the outlet section on flow profile, two cases without the presence of the “L” outlet in Pipe-SEP called Pipe-SEP I, and one with the “L” outlet called Pipe-SEP II were developed, as shown in Figure 5-1 and Figure 5-2. The diameter of the Pipe-SEPs was 150mm. The diameter and the length of the “L” outlet in Pipe-SEP II were 50mm and 500mm, respectively. The dimensions of the computation domain and observation planes employed in this study are also given in Figure 5-1 and Figure 5-2. Four plotting sections are used to reveal the flow filed inside Pipe-SEP as given by Table 5-1.
Figure 5-1 Schematic of the Pipe-SEP I geometry and coordinate definition
Figure 5-2 Schematic of the Pipe-SEP II geometry and coordinate definition
9 0 0 2 0 0 1 0 0 5 0
Table 5-1 The locations of each section within the plots
Section X1 X2 X3 X4
X(mm) 150 300 450 600
X/D 1 2 3 4
5.2.2 Selection of Turbulence Model
The numerical simulation quality depends largely on the type of turbulence modelling applied for approximation of turbulence. There are various turbulence models available in ANSYS FLUENT to simulate the rotational turbulent flow within the Pipe-SEP. The literature review on CFD modelling of cyclonic separators (Sec. 3.5) revealed that the k- ߳models are not appropriate for strongly swirling flows. Both of the RSM model and LES model have been widely applied and provide accurate predictions of the swirl flow pattern and velocity profile. Since the RSM model can yield reasonable results with limited computer resources, the RSM will be used in the present study to reveal the flow field.
5.2.3 Selection of the Discretization Schemes
Kaya and Karagoz (2008) and ANSYS FLUENT recommended the QUICK scheme for the momentum discretization. Kaya and Karagoz (2008) also recommended the Presto scheme for pressure and SIMPLE algorithm for pressure-velocity coupling. In terms of the discretising the Reynolds’ stress equation, the first order upwind method is suggested, but Shukla et al., (2011a) suggested the second order one in terms of the kinetic energy equation including the study of the dissipation rate. Table 5-2 lists the numerical schemes used in the present simulations.
Table 5-2 The numerical settings for the present simulation
Pressure discretization PRESTO Pressure velocity coupling SIMPLE Momentum discretization QUICK
Turbulent kinetic energy Second order upwind Turbulent dissipation rate Second order upwind
5.2.4 Boundary Conditions
A velocity inlet is applied at the inlet boundary, with an air inlet velocity ofܸ௦ି௧ having an initially value of 10 m/s, corresponding to an air inlet volumetric flow rate ܳ=145 Sm3/h. The air properties used were: air density 2.5 kg/m3, viscosity 1.73E-5 Pa·s, with a resulting Reynolds number 72254 based on the inlet diameter and the velocity of inlet nozzle. The turbulence intensity was determined as recommended by FLUENT 12 guidelines, using the following equation:
ܫ= 0.16ܴ݁ି.ଵଶହ (5-1)
The turbulence intensityܫis 3.95% and the hydraulic diameter equals the diameter of the inlet nozzle. The near-wall treatment and the effect of the wall roughness were modelled using non-Equilibrium wall functions. The out flow is gas outlet and its operational pressure is set as 0 Pa.
5.2.5 Grid Independency Study
The grid was generated in ANSYS Workbench. The Pipe-SEP I consists of purely trihedral grids. The Pipe-SEP II consists of trihedral grids at the separator body and tetrahedral grids at the gas outlet section. Figure 5-3 shows the grid configuration that was used for the simulation study. The Pipe-SEP I grid domain contains 169955 elements. The Pipe-SEP II grid domain contains 414369 elements.
The grid independence investigation was carried out within the Pipe-SEP I and Pipe- SEP II simulations. In order to make sure the final simulation results are insensitive to the grid condition, three levels of grids were applied as shown in Figure 5-4 and Figure 5-5.
(a) Pipe-SEP I (b) Pipe -SEP II
(a) Coarse grid (b) Medium grid (c) Fine grid
(a) Coarse grid (b) Medium grid (c) Fine grid
The simulated results of tangential velocity profile at X2 for Pipe-SEP I and Pipe-SEP II are shown in Figure 5-6 and Figure 5-7, respectively. It can be seen that the results for the three grids are very similar to each other in the centre region. In the near wall region, as the grid density increase, the result of medium grid and fine grid are very close.
Furthermore, Table 5-3 lists the grid details, which indicate that the outcomes for the fine and medium grids have very close values, particularly for the Eu values and cut size D50. Due to the better performance, even though the medium grid is able to offer a good prediction, to obtain the best results and avoid unnecessary uncertainties, the fine grid was adopted for all the simulations.
Figure 5-6 The tangential velocity profile at X2 for Pipe-SEP I with different grids -0.08 -0.06 -0.04 -0.02 0 0.02 0.04 0.06 0.08 -10 -8 -6 -4 -2 0 2 4 6 8 10 Y distance (mm) T an g e nt ia l v e lo ci ty at X 1 (m /s ) Coarse grid Medium grid Fine grid
Figure 5-7 The tangential velocity profile at X2 for Pipe-SEP II with different grids
Table 5-3 Grid details of the Pipe-SEP in the simulation
Pipe-SEP I Pipe-SEP II Elements Eu D50 Elements Eu D50 Coarse mesh 33903 0.8616 8.7 96354 0.9188 6.2 Medium mesh 93658 0.8536 9.2 298357 0.9113 6.9 Fine mesh 169955 0.8584 8.9 414369 0.9146 6.3 5.2.6 Convergence Criteria
As for the convergence criteria, two aspects are adopted. Firstly, the solution is assumed to be converged with pre-set scaled residuals of 1E-07. Secondly, the gas velocity at gas outlet should become constant and stable.