LOCAL COMERCIAL NUMERO “2” INDICIO 1 30,851 VIDEOGRAMAS DE DIVERSOS TITULOS, 3001 FONOGRAMAS DE DIVERSOS TITULOS, 13 (trece) aparatos electrónicos: (CUATRO TELEVISORES DE

A two-fluid model has been recommended for studying large-scale flow structures occurring in pilot and industrial scale bubble columns due to its relatively lower computational cost [80]. An Eulerian-Eulerian two fluid model was therefore selected in the present research considering the large size of the pilot flotation column that was to be modelled. In the Eulerian-Eulerian approach both phases, i.e., the primary phase and the secondary phase, are treated as inter-penetrating continua. Momentum and mass conservation equations are then

-0,05 0 0,05 0,1 0,15 0,2 0,25 0,3 -0,6 -0,4 -0,2 0 0,2 0,4 0,6 M e an A xi al B u b b le Veloc ity (m /s) Radial Position (m)

51 solved for each of the phases separately. Interaction between the phases was accounted for through inclusion of the drag force between phases. The Eulerian-Eulerian model has been reviewed in Chapter 3. The continuity and momentum equations for the Eulerian-Eulerian model are therefore not repeated here and are given in Chapter 3 as equation (3.3) and equation (3.4), respectively.

In the present research, water was modelled as the primary phase while air bubbles were treated as the secondary phase. The volume fraction (or gas holdup) of the secondary phase was calculated from the mass conservation equations as:

1 𝜌_𝑟𝐺(

𝜕

𝜕𝑡(𝜀𝐺𝜌𝐺) + ∇. (𝜀𝐺𝜌𝐺𝒖⃗⃗⃗⃗⃗ ) = 𝑆𝐺 𝐺) (4.2) where ρrG is the phase reference density, or volume averaged density of the secondary phase

in the solution domain. The volume fraction of the primary phase was calculated from the secondary phase one, considering that the sum of the volume fractions is equal to 1. In order to obtain the correct local distribution of the gas phase, previous researchers implemented compressibility effects in their CFD models using the ideal gas law [22, 32]. Similarly, the axial gas holdup variation in the present study was accomplished in the CFD simulations by applying the ideal gas law to compute the density of the secondary phase (ρG) as a function of

the local pressure distribution in the column according to the following equation: 𝜌𝐺 =

𝑝_𝑜𝑝+ 𝑝 𝑅 𝑀_𝑤𝑇

(4.3)

where p is the local relative (or gauge) pressure predicted by CFD, pop is operating pressure, R is the universal gas constant, Mw is the molecular weight of the gas, and T is temperature.

4.3.3.1 Drag force formulations

Generally, the drag force per unit volume for bubbles in a swarm is given by [84]: 𝐹 𝐷 = 3𝜀_𝐺𝜀_𝐿 4 ( 𝜌_𝐿 𝑑𝐵 ) 𝐶𝐷|𝒖⃗⃗⃗⃗⃗ − 𝒖𝐺 ⃗⃗⃗⃗ |(𝒖𝐿 ⃗⃗⃗⃗⃗ − 𝒖𝐺 ⃗⃗⃗⃗ ) 𝐿 (4.4) where CD is the drag coefficient, dB is the bubble diameter, and 𝒖⃗⃗⃗⃗⃗ − 𝒖𝐺 ⃗⃗⃗⃗ is the slip velocity. 𝐿 The drag force formulation includes the correction for bubble swarms and local volume fractions of the phases [91]. There are several empirical correlations for the drag coefficient,

52 function of the bubble Reynolds number (Re). A constant value of the drag coefficient may also be used [92, 93]. The bubble Reynolds number is defined as:

𝑅𝑒 =𝜌𝐿|𝒖⃗⃗⃗⃗⃗ − 𝒖𝐺 ⃗⃗⃗⃗ |𝑑𝐿 𝐵 𝜇𝐿

(4.5)

In the present research, simulations were carried out with three different drag coefficients. The first set of simulations was performed using the Universal drag coefficient [94]. Subsequent simulations were then conducted with the Schiller-Naumann [95] and Morsi- Alexander [96] drag coefficients in order to compare the suitability of the three drag models for the average and axial gas holdup computation in the flotation column. The equations describing the three drag coefficients are outlined in Table 4.2.

The drag coefficient values calculated from the equations presented in Table 4.2 are plotted in Figure 4.6 as a function of the bubble Reynolds number for 0 ≤ Re < 1000. Typical bubble sizes used in column flotation range from 0.5 – 2 mm corresponding to 1 << Re < 500 [97]. For the 1 mm bubble size used in the simulations in this study, the Re value calculated using equation (4.5) was 115.5. Details of the drag coefficient calculations for the Re versus CD

graphs are provided in Appendix 4 (A – C). From Figure 4.6, it can be seen that increasing Re causes the drag coefficient to decrease exponentially towards a constant value. A constant CD

value of about 0.44 is therefore applicable for the Schiller-Naumann drag coefficient for Re > 1000.

On the other hand, the formulation for the Morsi-Alexander and Universal drag coefficients will change depending on the bubble Reynolds number. For the Morsi-Alexander drag model, the constants A, B, and C will be varying as Re changes from one range to another as outlined in Table 4.2. Likewise, the universal drag coefficient is defined differently for flows that are categorized to be either in the viscous regime, the distorted bubble regime, or the strongly deformed capped bubbles regime as determined by the Reynolds number.

At the moderate superficial gas velocities simulated in this study, the viscous regime (0 ≤ Re < 1000) conditions apply. The equation presented in Table 4.2 is the one which is applicable when the prevailing flow is in the viscous regime. From Figure 4.6, it can also be observed that the Universal and the Schiller-Naumann drag models will give similar values of CD in

the viscous regime. Further details about the universal drag laws are available in a recent multiphase flow dynamics book [94].

53 Table 4.2: The drag coefficients that were used in the CFD simulations in the present study. Schiller and Naumann[95] 𝐶𝐷 = { 24 𝑅𝑒[1 + 0.15𝑅𝑒 0.687_{] 𝑓𝑜𝑟 𝑅𝑒 < 1000} 0.44 𝑓𝑜𝑟 𝑅𝑒 > 1000 (4.6) Morsi and Alexander[96] 𝐶_𝐷 = 𝐴 𝑅𝑒+ 𝐵 𝑅𝑒2+ 𝐶 𝐴 𝐵 𝐶 𝑅𝑒 < 0.1 24 0 0 0.1 ≤ 𝑅𝑒 < 1 22.73 0.0903 3.69 1 ≤ 𝑅𝑒 < 10 29.1667 −3.8889 1.222 10 ≤ 𝑅𝑒 < 100 46.5 −116.67 0.6167 100 ≤ 𝑅𝑒 < 1000 98.33 −2778 0.3644 1000 ≤ 𝑅𝑒 < 5000 148.62 −4.75 × 104 0.357 5000 ≤ 𝑅𝑒 < 10000 −490.546 5.787 × 105 _0.46 10000 ≤ 𝑅𝑒 < 50000 −1662.5 5.4167 × 106 0.5191 𝑅𝑒 ≥ 50000 0 0 0.44 (4.7) Universal drag[94] 𝐶_{𝐷𝑣𝑖𝑠} = 24 𝑅𝑒(1 + 0.1𝑅𝑒 0.75₎

where Re is the relative Reynolds number for the primary phase L and the secondary phase G obtained on the basis of the relative velocity of the two phases as:-

𝑅𝑒 =𝜌𝐿|𝒖⃗⃗⃗⃗ − 𝒖𝐿 ⃗⃗⃗⃗⃗ |𝑑𝐺 𝐵 𝜇_𝑒

where μe is the effective viscosity of the primary phase considering the effects of the secondary phase.

𝜇_𝑒 = 𝜇𝐿 1 − 𝜀_𝐺

54 Figure 4.6: Drag coefficient CD as a function of bubble Reynolds number Re for 0 ≤ Re < 1000.

4.3.4 Turbulence model

Turbulence was modelled using the realizable k-ϵ turbulence model which is a RANS (Reynolds- Averaged Navier-Stokes) based model. In the RANS modelling approach, the instantaneous Navier-Stokes equations are replaced with the time averaged Navier-Stokes (RANS) equations which are solved to produce a time averaged flow field. The averaging procedure introduces additional unknowns, the Reynolds stresses. The Reynolds stresses are subsequently resolved by employing Boussinesq’s eddy viscosity concept where the Reynolds stresses (or turbulent stresses) are related to the velocity gradients according equation (3.9). The eddy (turbulent) viscosity is then modelled as outlined in equation (3.13) and equation (3.14).