Recent Progress in Efficient Gas–Solid Cyclone Separators with a High Solids Loading for Large-scale Fluidized Beds †

Circulating fluidized beds (CFB)s are important technical equipment to treat gas–solid systems for fluid catalytic cracking, combustion, gasification, and high-temperature heat receiving because their mass and heat transfer rates are large. Cyclones are important devices to control the performance of CFBs and ensure their stable operation; heat-carrying and/or solid catalyst particles being circulated in a CFB should be efficiently separated from gas at a reduced pressure loss during separation. In commercial CFBs, a large amount of solids (> 1 kg-solid (m 3 -gas) –1 or > 1 kg-solid (kg-gas) –1 ) is circulated and should be treated. Thus, gas–solid cyclones with a high solids loading should be developed. A large number of reports have been published on gas–solid separators, including cyclones. In addition, computational fluid dynamics (CFD) technology has rapidly developed in the past decade. Based on these observations, in this review, we summarize the recent progress in experimental and CFD studies on gas– solid cyclones. The modified pressure drop model, scale-up methodology, and criteria for a single large cyclone vs. multiple cyclones are explained. Future research perspectives are also discussed.


Introduction
Circulating fluidized beds (CFB)s are commercially used for fluid catalytic cracking (FCC), and combustion and gasification of coal, biomass, and waste material because their heat and mass transfer rates are very large (Basu P., 2015;Grace J.R. et al., 1997;Gräbner M., 2015;Lettieri P. and Macri D., 2016;Li C.-Y. et al., 2017;Pecate S. et al., 2019;Scala F., 2013;Stolten D. and Scherer V., 2011). Recently, novel CFB solar receivers have been developed for concentrated solar power generation (Ansart R. et al., 2017;García-Triñanes P. et al., 2016, 2018. In a CFB, because heat-carrying and/or catalyst particles are circulated to transfer heat between each reactor, fast and efficient gas-solid separation is extremely important for a stable operation and reducing particle loss due to entrainment. Cocco R et al. (2017) investigated particle entrainment and clustering in a fluidized bed. Cyclones are widely used as primary gas-solid separators owing to their simple configuration and ease of operation. A stable operation of cyclones is important to reduce load in secondary gas-solid separators, makeup of the heat-carrying and/or catalyst particles, and pressure loss. Thus, much attention has been paid to the research and development of cyclones over the past several decades (Hoffmann A.C. and Stein L.E., 2010).
In commercial CFB boilers where carbonaceous solids are combusted to generate heat, a high solids-loading gas (C T > 10 kg-solid (kg-gas) -1 as defined in Eq (1)) should be treated in the primary gas-solid separator in order to efficiently separate solids from the gas ( Van de Velden M. et al., 2007;Dewil R. et al., 2008).
total mass flow rate of solid to cyclones C total mass flow rate of gas to cyclones  (1) Cortés C. and Gil A. (2007) extensively reviewed the models developed for the flow behavior, velocity profiles, pressure drop and collection efficiency in inverse-flow cyclones under the conditions of C T < 0.23 kg-solid (kggas) -1 . Huard M. et al. (2010) comprehensively reviewed gas-solid separators including cyclones, impact separators, and other separators with a riser top and downer bottom configuration. They summarized the solidsseparation mechanism and separation efficiency of each separator and reported that recent research on reverseflow cyclones is directed toward the influence of high solids loadings on cyclone performance and computational fluid dynamics (CFD) simulations. However, in the past, CFD calculations were not highly reliable for large solidsloading cyclones and fast separators due to limitations on computational power.
On the basis of such reviews and reports, in the current review, we summarized progress in the past decade in terms of experimental and numerical studies on high solids-loading gas-solid separators (mainly cyclones) in fast solids-circulation systems in CFBs. In the improved C-S model, geometric and velocity variables were normalized with respect to cyclone diameter. As shown in Fig. 1, the performance predicted by the improved C-S model agrees well with experimental data when C s = 1-5 kg-solid (m 3 -gas) -1 , thus effectively amending the overprediction of the original C-S model.

Multi-cyclone and non-uniform distribution of particles
In large-scale industrial reactors with high gas-solid flow rates, small parallel cyclones are often preferred to achieve a high separation efficiency when the distribution of gas-solid flow at each cyclone inlet is uniform. However, it is difficult to place a refractory in such small cyclones (Nowak W. and Mirek P., 2013) and there is strong evidence that gas-solid flow in parallel cyclones is non-uniform, which reduces the total separation efficiency (Zhang C. et al., 2016). Masnadi M.S. et al. (2010) examined gas-solid flow distribution across two parallel and identical cyclones based on an analytical model and compared flow distribution through parallel pipes. They confirmed the consistency of their analytical model by comparing the experimental data of two identical cyclones (barrel diameter D = 101.6 mm). They reported that a non-uniform (or maldistribution) gas-solid flow is unavoidable for a high solids loading (upstream solids hold up > 0.01 %) and that fouling can significantly affect maldistribution of gassolid flow through the identical cyclones. Zhang C. et al. (2010) calculated a three-dimensional full-loop CFB boiler model with two parallel cyclones using an Eulerian granular multiphase model. In their study, they found minor differences in the average solids mass flux (G s ) in the two cyclones (5.74 and 6.05 kg m -2 s -1 ) and pointed out that the maximum G s alternates between these two cyclones. Zhou X. et al. (2012) investigated gas-solid flow through six parallel cyclones located asymmetrically on the left and right walls of the riser (i.e., three cyclones on the left and three cyclones on the right) in a CFB cost test apparatus. They observed that the distribution of gassolid flow was non-uniform across three cyclones on one side and that the middle cyclone on each side exhibited higher particle velocities while their G s was lower than that of other cyclones. Jiang Y. et al. (2014) conducted numerical calculations on gas-solid flow hydrodynamics at a CFB boiler test facility with six parallel cyclones using an Eulerian-Lagrangian model and computational particle fluid dynamics (CPFD). They validated simulation data using experimental data obtained by electrical capacitance tomography (ECT) in cold model tests (at ambient temperature and atmospheric pressure). The geometry of the six cyclones was either axis-symmetric or point-symmetric, as shown in Fig. 2.
They also observed that the solids concentrations of the four cyclones located at the corners of the chamber were greater than those of others and stated that an "axissymmetric" arrangement (case A) for cyclones is better than a "point-symmetric" arrangement (case B) from the view point of uniform distribution of solids (Fig. 3). Wang S. et al. (2017) investigated the hydrodynamics of gases and solids in six parallel cyclones with central symmetry and axial symmetry arrangements combined with a full-loop CFB by CFD and a discrete elemental method (CFD-DEM, Fig. 4). They reported that for a uniform distribution of solid mass flux in parallel cyclones, axial symmetry is better than central symmetry. The middle Front view of simulation geometry, (b) Axis-symmetric, (c) Point-symmetric (units: mm). Reprinted with permission from ref. (Jiang Y. et al., 2014). Copyright (2014) Elsevier B.V. *Numbers and characters are rewritten to improve the clarity. cyclones on both sides have higher solid velocity and solid holdup than other corner cyclones (Fig. 5). Shuai D. et al. (2017) also investigated the flow behavior of gas and solids in six cyclones in parallel in an annular furnace. They observed a non-uniform distribution of G s and cyclone pressure drop (Δp c ) in the six cyclones. However, no regularity could be observed. The relative deviation of G s in the six cyclones was 8.0 % under typical operating conditions.

Design principles of multi-cyclones
The design principles of small multi-cyclones and a large single cyclone were analyzed. Mo X. et al. (2015) investigated the influence of wall friction and solid acceleration on the non-uniform distribution of gas-solid flow in two parallel cyclones. They reported that pressure drop in cyclones has an inflection point with respect to the mass flow rate ratio between solids and the gas (C T,inf , Eq. (1)), which is around 0.5-3.33 kg-solid (kg-gas) -1 , as calculated from the reported values. They also suggested that the inflection point increases with an increase in the wall friction but decreases with an increase in solid acceleration. The inflection point has a large effect on nonuniform gas-solid distribution in the two parallel cyclones. As summarized in Table 1, the solid flow distribution is non-uniform when C T approaches C T,inf . Zhang C. et al. (2016) analyzed multi-phase interactions and investigated instabilities in uniformity in two parallel cyclones installed after a fluidized bed reactor. They provided a novel design principle to avoid nonuniform distribution of solids using C T and dimensionless vortex finder diameter (d r ) defined in Eq. (15). Fig. 6 shows a phase diagram of uniformity of the two parallel cyclones as a function of fraction of solid flow to cyclone 1 (i.e. path 1) and C T . They reported that under the condition that inlet velocity of air is 15 m s -1 , Reynolds number is 3.08 × 10 5 and glass beads are used at 20 °C and 101.3 kPa, a low solids loading (C T < 1.35), the uniform distribution is stable but at C T = 1.35, there occurs a turning point from uniform to non-uniform distribution. When C T > 1.35, the maldistribution of solids is stable.
Main contributor Wall friction Both greatly influence the pressure drop

Solid acceleration
Gas fraction in cyclone 1 (ψ) and solids fraction in cyclone 1 (γ) ψ > 0.5 and γ > 0.5 or ψ < 0.5 and γ < 0.5 ψ ~ 0.5 and highly uneven solid distribution at certain flow perturbation ψ > 0.5 and γ < 0.5 or ψ < 0.5 and γ > 0.5 They found that d r should be less than 0.32 and that d r strengthens the swirl in a cyclone for enhancing the stability of uniformity and cyclone efficiency under this condition. Fig. 7 shows a phase diagram of stability based on the data in literature (Zhang C. et al., 2016). For parallel cyclones that treat high solids loading, d r should be small enough to ensure the stable operation. These criteria are very useful in the selection of single large cyclones or multi-cyclones for uniform and steady operation.  Wu X. et al. (2011) experimentally measured and calculated the pressure drop of two scroll-type cyclones connected in series for a wide range of solids loading (C T = 0-30 kg-solids (kg-gas) -1 ) and cyclone inlet velocity (v i = 16-24 m s -1 ). The experimental data showed that the cyclone pressure drop decreased dramatically as C T increased to 7.5 kg-solids (kg-gas) -1 after which it remained almost constant.

Cyclones in series
Fushimi C. and Guan G. et al. developed a triple bedcombined CFB for coal gasifiers (Fushimi C. et al., , 2014Guan G. et al., 2010Guan G. et al., , 2011 and experimentally investigated the hydrodynamics of silica sand (average particle diameter of 126 μm) at a high flux (G s in the range of 200-546 kg m -2 s -1 ) in cold model tests. In these experiments, three cyclones were connected in series to separate solids from air. Separation efficiency at the outlet of the third cyclone was 99.998 % when G s was  Reprinted with permission from ref. (Zhang C. et al., 2016). Copyright (2016) John Wiley and Sons.

Fig. 8
Schematic diagram of the proposed iG-CLC system Reprinted with permission from ref. (Wang X. et al., 2015). Copyright (2015) The American Chemical Society.
490-510 kg m -2 s -1 . Meanwhile, Hoffmann A.C. and Stein L.E. (2010) re commended that the underflow pipes of series-connected cyclones should be independently sealed. Wang X. et al. (2015) developed a high-flux CFB reactor, comprising of a fuel reactor, an air reactor, a J-valve, a downcomer, and solid separation systems, for in situ gasification chemical looping combustion (iG-CLC). They optimized the iG-CLC system by developing an inertial separator for primary gas-solids separation and a cyclone for secondary gas-char and remaining solids separation to improve operation stability and solid separation efficiency. They reported that the selective separation efficiency of coal particles in the first-stage inertial separator was 77.7 % when the mass f low rate of the solid was 1021 kg h -1 and the overall separation efficiency after the second-stage cyclone reached 99.5 % when the mass flow rate to cyclone was 35 kg h -1 in their cold model tests.

Combination of an inertial separator and a cyclone for a high-flux CFB
In their subsequent work (Shao Y. et al., 2017), they carried out three-dimensional full loop CFD calculation using an Eulerian-Eulerian two-fluid model combined with the standard k-e turbulence model for the gas phase and the kinetic theory of granular flow for the solid phase to investigate the hydrodynamics of coal and iron ore particles on the basis of the experimental results of cold model tests. They reported continuous and stable solid circulation in the iG-CLC model when G s was around 400 kg m -2 s -1 .

Design principles of the cyclone inlet and body for high solids-loading multi-cyclones
Hoffmann A.C. and Stein L.E. (2010) recommended that 1) bends should be located at a distance of ten pipe diameters or its equivalent before cyclone inlets to avoid non-uniform distribution of gases and solids in the cyclone inlet piping and 2) a scroll or wrap-around type of inlet should be used. At high solids or liquid loadings (> ~10 vol%), care must be taken not to restrict the discharge opening or the annular space around a vortex stabilizer. Wei Z.G. et al. (2016) examined the pressure of a catalyst powder (Sauter mean particle diameter of 63.6 μm) flow in a dipleg (150 mm inner diameter and 9 m height) when G s was in the range of 50.0 to 385.0 kg m -2 s -1 .   . The bottom of the dipleg was immersed into the fluidized bed. Fig. 10 shows the observed fluidization pattern in the dipleg. At small G s (G s < 50.0 kg m -2 s -1 ) (cf. Fig. 3a, b), a dilute-dense coexisting falling flow. The swirl flow (just below the cyclone), the dilute particle falling flow (in the middle) and the dense flow (bottom) were observed. In the dense flow, some ascending gas bubbles were observed like a bubbling fluidized bed. When G s = 200.0-250.0 kg m -2 s -1 (cf. Fig. 3c), no interface between the dilute particle falling flow and the dense flow was observed. Pressure fluctuation intensity reached a maximum in the fluidization pattern being transformed. Further increase in G s (G s > 350.0 kg m -2 s -1 ) (cf. Fig. 3d), the fluidization regimes developed the dense conveying flow with a high particle concentration.

Scale-up methodology and grade efficiency
The scale-up of cyclones is important for reactor design and estimating separation efficiency. Mirek P. (2016Mirek P. ( , 2018 investigated the scaling rules of cyclones in CFB boilers and estimated the separation efficiency of cyclones by setting up the following model (Eqs. (16)- (21)). In this model, the total separation efficiency of a cyclone η tot is the sum of the efficiency at the wall (η wall ) and in the inner vortex (η vtx ) (Eq. (16)). A portion of the incoming solids not collected by the wall is collected by the inner vortex, which is based on the Muschelknautz model (Muschelknautz E., 1972 Here, the constant K lim = 0.02 for fine particles and 0.03 for coarse particles. For initial load, μ e < 2.2 × 10 −5 , the exponent y has a value of 0.81 and for μ e > 0.1, y = 0.15. The value of d e is calculated using Eq. 18 (Mirek P., 2018).
Wang J. et al. (2019) reported that when the solids loading is higher than the limit loading (μ e > μ lim ), dense particles cannot flow through the rotating/swirling stream because of their high inertia; in this case, these particles hit the outside wall of the cyclone and descend to the bottom.
Mirek P. also described the particle size distribution of the carryover, R F (d), and it can be determined using Eq. (19).
The total separation efficiency is determined by Eq. (20).

   
The same researcher summarized the scaling relationship using five dimensionless parameters sets shown in Table 2. They compared estimation curves based on the Table 2 Scaling relationships and parameters used for the analysis of the cyclone performance in cold model tests (cf. Fig. 11).
Reprinted with permission from ref. (Mirek P., 2018). Copyright (2018) Elsevier B.V.  (21)). . 11 shows the result. All the proposed scaling relationships resulted in a very high separation efficiency (> 99.4 %). However, the results strongly depended on the selected scaling relationships set (Mirek P., 2018). Wang J. et al. (2019) recently conducted experiments extensively and introduced sophisticated models to improve the prediction of grade efficiency of cyclone separators. They constructed generalized linear mixed-effects (GLME) models that are functions of the parameters (St,

Scaling relationships Parameters chosen independently Notation
, Re, Fr ,H ) to understand and control grade efficiency (η F (d)) variation. Note σ ξ is the size deviation in particle diameter at the inlet. In their GLME models, the observed grade efficiency vector (η) is considered to be a summation of the expected grade efficiency vector (μ) and the noise vector (e), which is assumed to follow a normal distribution with mean 0 (Eq. (22)).
where ϕ i is a random effect corresponding to the i th com-ponent and has a normal distribution with a variance σ i 2 . The g(μ) and the expected grade efficiency (μ) are linked in the following equations ((24), (25)). Table 3 lists the values estimated for the parameters (β 0-7 , σ 1 , σ 2 , σ 6 ) in Eq. (23), including a 95 % confidence interval for random effects. Fig. 12 shows the parity plot of experimental results and fitted values predicted by using the GLME model. It can be seen that the GLME modeling predicts the experimental grade efficiency accurately. It is considered that the GLME model can provide setting conditions for cyclone separators .

Fig. 11
Influence of a set of scaling relationships on the total grade efficiency curve based on the parameter sets shown in Table 2 and the Stokes equation. Reprinted with permission from ref. (Mirek P., 2018). Copyright (2018) Elsevier B.V.  Mirek P.    Table 4 summarizes the experimental and numerical studies on gas-solid cyclones conducted in the last decade. Recently, a number of numerical studies using CFD simulations have been conducted to investigate the velocities and volume fractions of gases and solids and pressure drop in cyclones by varying their configurations. Most of these experimental and numerical studies were conducted under dilute conditions (C s or C T~0 .001) . Huang A.N. et al. (2018a) investigated the effect of particle mass loading on the hydrodynamics of solids and separation efficiency using the Eulerian-Lagrangian approach with a two-way coupling method for CFD and compared with the experimental results of laboratory-scale cyclones at C s of 0.0016-0.153 kg-solid (m 3 -gas) -1 and v i = 15 and 18 m s -1 . They reported that the overall separation efficiency increased from 74.5 % for a conventional cyclone to 80.7 % for a cyclone with a lower cone slit. More detailed experimental and numerical studies under high solids loading conditions on gas-solid cyclones are expected in the future.

Conclusions and future prospects
In this review, recent progress in high solids-loading gas-solid cyclones has been summarized. The improved C-S model proposed by Li S. et al. (2011) is highly useful for predicting pressure drop in cyclones. For commercialscale CFBs, 2 to 6 multi-cyclones are often used in parallel. However, the non-uniform distribution of solids in the parallel section is a problem. Mo et al. investigated the effect of wall friction and solid acceleration on the nonuniform distribution of gas-solids flow and found that the inflection point has a significant influence on the maldistribution of gas-solids in two cyclones in parallel. Zhang C. et al. (2016) found that the phase graph of the stability of uniform distribution/maldistribution of solids in two parallel cyclones as functions of C T (kg-solid (kg-gas) -1 ) and d r.
To improve the scale-up methodology, Mirek P. (2018) summarized and modified the Muschelknautz model by applying five sets of scaling relationships and found that all the employed sets of scaling relationships resulted in a very high separation efficiency (> 99.7 %). The GLME model developed by Wang J. et al. (2019) can predict grade efficiency with a large prediction variability.
It is expected that CFB boilers will be used for biomass co-combustion to reduce CO 2 emissions due to power generation. In waste-treatment plants, CFB incinerators may be used in the future because of their high thermal efficiency and operability. In CFB boilers used for power generation, non-steady-state operations (rapid changes in load) are required. In these operations, a high solids separation efficiency and control over the cut-off diameter are required. We expect for future research, flow behaviours of gas and solids, and separation efficiency of solids in cyclones under non-steady state or transient operation conditions should be investigated. η tot : Total separation efficiency [-] η vtx : Separation efficiency in cyclone vortex [-] η wall : Separation efficiency at wall [-] ϕ i : Random effect of the i th component [-] σ i : Standard deviation of ϕ i [-] σ ξ : Size deviation of particle diameter at inlet [-] ψ: Gas fraction in cyclone 1 [-] γ: Solids fraction in cyclone 1 [-] Abbreviations