Applied Thermal Engineering 123 (2017) 635–645

Contents lists available at ScienceDirect

Applied Thermal Engineering

journal homepage: www.elsevier .com/locate /apthermeng

Research Paper

Study of the performance, simplification and characteristics of SNCRde-NOx in large-scale cyclone separator

http://dx.doi.org/10.1016/j.applthermaleng.2017.04.1221359-4311/� 2017 Published by Elsevier Ltd.

⇑ Corresponding author.E-mail address: yuanqx1991@163.com (Q. Yuan).

Zhizhong Kang, Qixin Yuan ⇑, Lizheng Zhao, YuKun Dai, Baomin Sun, Tao WangKey Laboratory of Condition Monitoring and Control for Power Plant Equipment, North China Electric Power University, Beijing 102206, China

h i g h l i g h t s

� The optimized mechanism of the SNCR reaction was presented verified.� The optimized mechanism and CFD software were combined to simulate the process of SNCR de-NOx.� CFD software was used to study gas–solid characteristics.

a r t i c l e i n f o

Article history:Received 16 October 2016Accepted 24 April 2017Available online 10 May 2017

Keywords:SNCRMechanism simplificationCHEMKINCFDSeparator characteristics

a b s t r a c t

CHMKIN software was used to optimize the mechanism of SNCR reaction. Then the optimized 18-elementmechanism and CFD software were combined to simulate the process of SNCR de-NOx and meanwhilethe gas–solid characteristics in a cyclone separator of a supercritical CFB boiler were studied. The gas–solid characteristics and SNCR characteristics of the large-scale cyclone separator were obtained. Thetotal pressure difference of the cyclone separator increases with the inlet velocity, and the rate of increaseincreases gradually. When the inlet velocity is 20 m/s, the total pressure difference is 1987.6 Pa. The clas-sification efficiency of the cyclone separator increases with the inlet velocity, but the rate of increasedecreases gradually. When the inlet velocity is 20 m/s, the classification efficiency is 66.74%. The windowof the reaction temperature for the SNCR process is in the range of 1123–1323 K. The de-NOx efficiencyfirst increases and then decreases with the increase of temperature, reaching a peak at 1223 K. At lowtemperature, improving the NSR has little effect on the de-NOx efficiency. When the reaction tempera-ture is in the temperature window, improving the NSR can effectively enhance the de-NOx efficiency.But after increasing the NSR to a certain extent, the de-NOx efficiency increase will slow down. SettingNSR = 1.5 is suitable.

� 2017 Published by Elsevier Ltd.

1. Introduction

Ox nitride (NOx) is one of the main pollutants from powerplants, causing photochemical smog pollution by chemical reactionwith light and combining with the water in the air to produce acidrain. There is great international concern about NOx emission con-trol. The circulating fluidized bed (CFB) boiler has the characteris-tics of low temperature in the furnace. So when the externalconditions are the same, the amount of NOx generated in the fur-nace is low relative to the pulverized coal furnace [1]. The CFB boi-ler is now widely used in various countries. In order to furtherstrengthen environmental protection, China put forward the goalof ‘‘super-low emission”—under the condition of 6% of the standard

oxygen content, NOx emission concentration is less than 50 mg/N m3; SO2 emission concentration is less than 35 mg/N m3; sootemission concentration is less than 5 mg/N m3. A power plant willbe built with a 660 MW CFB boiler, burning low calorific value coal(coal gangue + middling). To achieve ‘‘super low emission” it isnecessary to further strengthen the de-NOx unit. SNCR has theadvantages of low investment, simple equipment and, most impor-tantly, the temperature in the CFB boiler cyclone separator is justwithin the SNCR reaction temperature window, so a CFB boilerequipped with an SNCR de-NOx device at the cyclone separatorcan achieve more than 50% de-NOx efficiency [2]. A 660 MWsupercritical CFB boiler is designed with SNCR de-NOx equipmentinstalled at the cyclone separator to achieve the goal of super-lowemission.

The 660 MW supercritical CFB cyclone separator is character-ized by its large scale. The preliminary design is to scale up thestructure of the common cyclone separator, but the large scale of

636 Z. Kang et al. / Applied Thermal Engineering 123 (2017) 635–645

the structure will lead to a more complex internal flow field in thecyclone separator. In practical operation, SNCR de-NOx is largelydependent on the mixing of flue gas and reducing agent. The com-plexity of the flow field will have a major influence on the SNCR de-NOx. At the same time, the complexity of the flow field will makethe boundary of the inner and outer swirl zone of the cyclone sep-arator irregular, which will affect the gas-solid separation andenergy loss of the cyclone separator.

In this paper, we use CHEMKIN software to optimize the SNCRreaction mechanism. The optimized mechanism and computa-tional fluid dynamics (CFD) numerical simulation are combinedto study the SNCR de-NOx in the cyclone separator of a 660 MWsupercritical CFB. Meanwhile, the CFD software is used to studythe performance of the cyclone separator—the pressure differenceand separation efficiency. It provides a valuable reference for theSNCR de-NOx and the practical operation of the large-scale cycloneseparator in the supercritical CFB.

2. Simplification of SNCR reaction mechanism

Although the mechanism of SNCR de-NOx has been very thor-oughly studied, in practical engineering application, with the SNCRde-NOx reducing agent input from a limited number of nozzlesinto the reaction zone, the flow and mixing of flue gas in thecyclone separator are the key factors affecting the SNCR de-NOxefficiency [3]. Therefore, it is necessary to use the combination ofCFD and SNCR chemical reaction process simulation to simulatethe process of SNCR de-NOx in the cyclone separator of the CFBboiler.

FLUENT itself contains a SNCR de-NOx reaction module. Thereaction mechanism of this module involves 14 components,including 9 basic chemical reactions [4]. Table 1 is the reactionmechanism of the reaction module. Reactions (1)–(7) are the SNCRde-NOx mechanism. Reactions (8) and (9) are the decomposition ofurea.

Because the module is in a black box, the chemical reactionsand their parameters cannot be modified.

When the simulation of the chemical reaction process is com-bined with the CFD simulation of the flow process, the detailedchemical reaction mechanism will occupy a large memory, andthus seriously reduce the efficiency of the simulation and affectits accuracy. In order to simulate the process of SNCR de-NOx inthe cyclone separator under the coupled action of flow mixingand chemical reaction kinetics, according to the reaction condi-tions of the CFB boiler cyclone separator, the optimized SNCR reac-tion mechanism is essential.

2.1. Analysis of SNCR mechanism

The chemical reaction mechanism is based on the micro level,from the point of view of the basic elements, to explore thedynamic behavior of the total reaction. The corresponding reaction

Table 1SNCR reaction mechanism (Fluent reaction module).

Serial number Chemical reaction

1 NH3 + NO? N2 + H2O + H2 NH3 + O2 ? NO + H2O + H3 HNCO + M? H + NCO + M4 NCO + NO? NO + CO + H5 NCO + OH? NO + CO + H6 N2O + OH? N2 + O2 + H7 N2O + M? N2+O + M8 CO(NH2)2 ? NH3 + HNCO9 CO(NH2)2 ? 2NH3 + CO2

rate constants are needed in the calculation of each basic elementreaction: pre-exponential factor A, reaction order n and the activa-tion energy E.

Numerous studies [5–7] have found that the reducing agentitself, through rapid pyrolysis, produces NH3 through the reactionof NH3 + OHM NH2 + H2O to reduce NO. NO selectively and rapidlyreacts with NH2 in the SNCR reaction temperature window. Thegenerated products have two paths, as shown in reactions (1)and (2):

NH2 þ NO $ NNHþ OH ð1Þ

NH2 þ NO $ N2 þH2O ð2ÞReaction (1) is the reaction when producing chain factors and

reaction (2) is the reaction when not producing chain factors [8].If it is short of chain factors, the self-sustaining reaction of NOand NH3 cannot be continued. So to a certain extent, the relativereaction rate of reactions (1) and (2) determines the extent of de-NOx. As a response to reactions (1) and (2), Miller and Bowmanput forward the branch coefficient a showing a reaction that pro-duces active routes. That is, reaction (1) accounts for the propor-tion of the total reaction rate. In order to maintain the NH3 andNO self-sustaining reactions, Miller [9] proposes that the branchcoefficient a should be greater than 0.25. Reaction (1) generatingOH can react with NH3 to regenerate NH2. The competitive reactionof NH2 includes a reaction between NH2 and free radicals, as shownin reactions (3)–(6):

NH2 þH $ NHþH2 ð3Þ

NH2 þ O $ HNOþH ð4Þ

NH2 þ O $ NHþ OH ð5Þ

NH2 þ OH $ NHþH2O ð6ÞThe generated NH2 can be reduced and oxidized. The reduction

reaction takes the dominant position at lower temperature, and theoxidation reaction at high temperature produces a greater impact.From experimental and theoretical analysis, the survival time ofthe NNH active route generated in reaction (1) is between 8 and11 s. If the survival time of the NNH active route is too short, andthe coefficient a is too large, it will cause an explosive reaction [10].

By superimposing reaction (1) and reaction (7), the followingreaction can be obtained: NH2 + NOM N2 + H + OH. This reactionnot only reduces the NO but also produces chain factors. Whenthe reaction temperature increases, the reaction (1) rate increases,producing more chain factors and yielding a more in-depth reac-tion and improving the de-NOx efficiency. Reaction (7) is asfollows:

NNH $ N2 þH ð7ÞNH free radicals are generated by the NH2 extraction of hydro-

gen atoms, such as reaction (3), reaction (5) and reaction (6). At thesame time, the reaction of NH radicals with NO generates N2O orN2, such as in reactions (8) and (9):

NHþ NO $ N2OþH ð8Þ

NHþ NO $ N2 þ OH ð9ÞThe reaction of NO with an H or CO direct amines is shown in

(10) and (11). The reaction between NO and H atoms at high tem-perature is dominant; however, due to the large amount of heatabsorption, this almost does not occur under 1400 K. The reactionof NO and CO first generates N atoms. The rate at which the

Table 2Optimized mechanisms.

Optimized equation Pre-exponentialfactor

Activationenergy

Temperatureexponent

Oþ OH () Hþ O2 2.0E14 0 �0.402OH () H2Oþ O 4.3E03 �2486 2.70Hþ O2ðþMÞ () HO2ðþMÞ 2.1E18 0 �1.0HO2 þ OH () H2Oþ O2 2.9E13 �497 0NH3 þ OH () NH2 þ H2O 2.0E06 566 2.04NH2 þ OH () NHþ H2O 4.0E06 1000 2.0NH2 þ NO () NNHþ OH 8.9E12 0 �0.35NH2 þ NO () N2 þ H2O �8.9E12 0 �0.35NHþ OH () NþH2O 5.0E11 2000 0.5NHþ NO () N2Oþ H 2.9E14 0 �0.4NNH () N2 þ H 1.0E07 0 0NNHþ O2 () N2 þ HO2 2.0E14 0 0COþ OH () CO2 þH 1.5E07 �758 1.3NH2 þ O () HNOþH 6.6E14 0 �0.5NHþ O2 () HNOþ O 4.6E05 6500 2.0HNOþ O2 () NOþ HO2 1.0E13 25000 0NOþHO2 () NO2 þ OH 2.1E12 �480 0NO2 þ O () NOþ O2 3.9E12 �238 0

Z. Kang et al. / Applied Thermal Engineering 123 (2017) 635–645 637

reaction occurs at low temperature is relatively slow. However,usually a higher reaction rate constant is used in the study of themechanism of the early stage [11].

NOþH $ Nþ OH ð10Þ

NOþ CO $ Nþ CO2 ð11ÞThe formation of HNO is mainly through the reaction of NO

with H2 or the recombination of NO and H. The series of reactionsof NH generated by HNO are as follows. Usually reactions (12)–(14)under the condition of the reaction are slow. Only reaction (13) hasa strong competitive ability at high temperature [12].

HNOþH $ NHþ OH ð12Þ

HNOþ CO $ NHþ CO2 ð13Þ

HNOþH2 $ NHþH2O ð14ÞExisting studies have found that when the temperature is below

1400 K, CO or H2 in the SNCR reaction conditions can lead to a sig-nificant reduction in NO, but the mechanism remains to be furtherstudied.

2.2. Simplification of SNCR mechanism in cyclone separator

The SNCR de-NOx process is very complicated. The detailedmechanism of the elementary reaction model usually contains sev-eral hundred elementary reactions, leading to a large number ofnumerical solutions of the reaction kinetics equations.

In 1996, Brouwer et al. [13] optimized the detailed mechanismof Miller and Bowman by means of sensitivity analysis and curvefitting. They developed a set of 6 substances, and a 7-step reactionmechanism in which NH3 was the reducing agent in the reactionand the effect of CO on the reaction process of SNCR was also con-sidered. In 2002, Rota et al. [14] used the detailed mechanism ofurea decomposition and the SNCR reaction processed with ammo-nia as the reducing agent to describe the process of NOx OUT reac-tion. In 2003, Xiaohai et al. [15], through an automaticsimplification of the program code, optimized the detailed basisof the reaction mechanism as the optimized mechanism of 10 stepsand 14 components for the coal-fired boiler. In 2006, Lu [8] used asensitivity analysis to analyze and optimize the 1989 Miller modeland the Rota model. The optimized model, including a 14-stepreaction from the Rota model could predict the change of NO andNH3 concentration with the change of temperature in high andlow oxygen levels.

CHEMKIN 4.1 software has a strong sensitivity analysis tool.Considering the flue gas composition, inlet velocity and residencetime of the CFB boiler cyclone separator in the process of SNCRreaction, we make a sensitivity analysis of the 18 components:NH3, N2O, NO, NO2, NH2, HNO, H, O, OH, O2, HO2, H2O, NH, NNH,N, N2, CO, CO2. If the absolute value of the sensitivity coefficientof any one of these substances in a basic element reaction is larger,this shows that the effect of the base element reaction on the SNCRreaction system in the cyclone separator is greater. By ignoring ele-ment reactions with less absolute values of the sensitivity coeffi-cient, we remove those intermediate components that have littleeffect on specific issues to achieve the purpose of optimizing thedetailed reaction mechanism.

In this paper, using the CHEMKIN software to push the flowreactor and adopting the consolidation mechanism of combiningØyvind Skreiberg et al.’s mechanism of ammonia oxidation withthe SNCR de-NOx mechanism, with ammonia as the reducingagent, we calculate the chemical kinetics of NH3 reduction byNOx. After using the sensitivity analysis, the optimized mechanismis as Table 2:

2.3. Optimized mechanism verification

In order to verify whether the optimized mechanismcan accurately simulate the process of de-NOx in setting theCFB boiler cyclone separator SNCR de-NOx reaction conditions,in this paper, an SNCR de-NOx experiment is carried outunder laboratory conditions. The experimental results andsimulation results are compared to verify the optimizedmechanism.

The experiment was carried out on a quartz tube reaction sys-tem. The experiment equipment comprised a reaction gas cylinder,mass flowmeter, electric resistance furnace, quartz reactor and gasanalyzer (Testo 350). The schematic diagram of the experimentalsystem is shown in Fig. 1.

N2, O2, NO and NH3 were provided from reaction gas cylinders(O2 bottles, NO bottles and NH3 bottles were 1% of the content ofthe mixture with N2). The ratio of the various reactive gases wascontrolled by the mass flow meter. A thermal resistance furnacewas used to heat the quartz reactor, increasing the temperatureby 10 K per minute at the preheating stage and setting the pre-heating time to 1.5 h. With a uniform temperature increase, thereaction gas with a good proportion of the prior mixture wasreacted in a high-temperature quartz reactor. The gas analyserwas used to measure the volume concentration of the reactiongas before and after the reaction. The experiments were carriedout under atmospheric pressure. The total mass flow rate of thereaction gas was constant, at about 1250 cm3/s. The NO flow inthe experiment remained unchanged for 125 cm3/s. N2 in the reac-tion gas was the protective gas.

Setting normalized stoichiometric ratio (NSR) = 1.5, O2 = 6%conditions, the experimental results and simulation results arecompared, as shown in Fig. 2.

From Fig. 2, when the reaction temperature is between 1073 Kand 1323 K, the trend of the de-NOx efficiency curve with thechange of temperature is the same. The de-NOx efficiency curveof the optimized mechanism and the de-NOx efficiency curve ofØyvind Skreiberg’s detailed mechanism are similar. But it is foundthat the experimental values are higher than the simulated values.The analysis shows that the reaction gas in the experiment beforethe reaction is fully mixed to improve the SNCR de-NOx efficiency.Under the conditions of setting the CFB boiler cyclone separatorSNCR de-NOx reaction, the optimized mechanism can simulatethe process of de-NOx.

Fig. 1. Schematic diagram of the experimental system.

1100 1150 1200 1250 1300 135045

50

55

60

65

70

75

80

85

90

SNC

R d

e-N

Ox e

ffic

ienc

y/%

Temperature(K)

SNCR mechanism(Oyvind Skreiberg)Experimental resultSimplified mechanism(18)

Fig. 2. The change curve of the denitrification efficiency with temperature.

Fig. 3. Schematic diagram of cyclone separator.

638 Z. Kang et al. / Applied Thermal Engineering 123 (2017) 635–645

3. Establishment of the calculation model

3.1. Physical model construction

The size of each part of the cyclone separator and the specificstructure of the cyclone separator are shown in Fig. 3. The initialsize of the structure is shown in Table 3.

3.2. Grid division

We adopt a method for partitioning the grid in the partitiondomain and divide the separator into 3 sub-regions: the inlet sec-tion, the exhaust pipe section, and the other section. Respectively,mesh the sub-regions. Finally, we choose a model divided intoabout 500,000 as the number of the grid which can get a goodresult. The specific grid of the cyclone separator is shown in Fig. 4.

3.3. Selection of turbulence model

Commonly used turbulence models are the k-emodels and RSMmodel. Fig. 5 shows velocity nephogram of x = 0 cross-sectionusing RSM, RNG k-e, Realizable k-e model. Fig. 6 shows velocitynephogram of z = 12 cross-section using RSM, RNG k-e, Realizablek-e model.

It can be seen from Fig. 5 that the simulation results using RSMmodel accord with the flow field distribution in cyclone separator.

The deviation of simulation results using k-e model with thedecrease of axial position will become increasingly large. This isbecause turbulent pulsation is strong in the conical part of cycloneseparator. At the same time, the k-e model based on the samedirection eddy viscosity hypothesis will show its own limitationand insufficiency.

It can be seen from Fig. 6 that the simulation results using RSMmodel accord with the flow field distribution in cyclone separator.Velocity nephogram exhibits an axisymmetric form. The simula-tion results using RNG k-emodel are in disorder, so the simulationresults do not conform to the internal flow field distribution. Thesimulation results using Realizable k-emodel are much better thanthe simulation results using RNG k-emodel. However, there is stilla gap.

Fig. 7 shows the tangential velocity of x = 0 cross-section usingRSM, RNG k-e, Realizable k-e model.

When the air enters from the inlet of the separator, the air flowbegins to accelerate. After the airflow enters the cyclone separator,the air flow continues to accelerate, reaching the maximum at theturn. Then the air flows along the cylinder wall, and the tangentialvelocity decreases slowly. In the separation space below the exitsection, the axial symmetry of the tangential velocity distribution

Table 3Size table of structure of cyclone separator.

Body diameter Riser diameter Diameter of dust discharge port Depth of insertion Inlet section height Inlet section width Total height Cone height

D Dx Dd S a b H Hc

8.5 m 4.25 m 1.5 m 3.6 m 8.5 m 3.6 m 21.8 m 10.8 m

Fig. 4. Schematic diagram of the grid of cyclone separator.

RSM RNG k- Realizable k-

Fig. 5. velocity nephogram of x = 0 cross-section using RSM, RNG k-e, Realizable k-emodel.

Z. Kang et al. / Applied Thermal Engineering 123 (2017) 635–645 639

is quite good, which shows the characteristics of strong swirlingflow in the separator.

Fig. 8 shows the tangential velocity of the x = 0 cross-sectionusing RSM, RNG k-e, Realizable k-e model.

The axial velocity of air flow in the cyclone separator in thecylinder and the upper part of the cone is basically quasi symmet-ric distribution. But the deviation is relatively large at the lowerpart of the cone. The axial velocity distribution is not symmetricalalong the geometric center of the cyclone separator. It has a certaineccentricity distance. This shows that in the cyclone separator atthe end of the trachea, there is the phenomenon of airflow shortcircuit.

It can be seen from the figures that the simulation results usingRSM model accord with the flow field distribution in cycloneseparator.

RNG k-e model was proposed by Yakhot et al. [16] in 1986. Themodel was optimized based on the standard k-emodel. The turbu-lent viscosity was modified and meanwhile the swirling of theaveraged flows and the swirling flow were considered. Realizablek-e model was a newer turbulence model proposed by Shih et al.[17] in 1995. Turbulent eddy viscosity coefficient was related tostrain rate in Realizable k-e model.

The k-emodels are based on the isotropic property assumption.This is not in conformity with the strong spin property of thecyclone separator. The RSMmodel is no longer based on the isotro-pic property assumption and has significant advantages in simulat-ing anisotropic turbulence [18]. The biggest difference betweenRSM model and the k-e models is that RSM model completelyabandons the isotropic eddy viscosity Boussinesq hypothesis andmore rigorously considers streamline bending, vortex, rotationand tension rapid change. For complex flows, RSM model hashigher accuracy in predicting the potential and in many casescan give results better than various k-e models. But this predictionis limited to the Reynolds stress related equation. When need toconsider the anisotropy of Reynolds stress, must use RSM model.Many scholars [19–21] use the RSM model to study the cycloneseparator, and get results that are close to the actual situation. Sowe use the RSM turbulence model for the simulation. The govern-ing equations of the RSM model are shown below.

Transport equation:

@ðqu0iu

0jÞ

@tþ Cij ¼ Pij þ Dij þ£ij þ eij ð15Þ

Cij - convection term; Pij - Reynolds stress; Dij - diffusion term; uij -stress and strain term; eij - dissipation term.

Turbulent diffusion model:

DT;ij ¼ @

@xk

ui

rk

@u0iu

0j

@xk

!

rk - Prandtl number of turbulent kinetic energy: 0.82; ul - tur-bulent viscosity.Stress strain model:

£ij ¼ £ij;1 þ£ij;2 þ£ij;w

uij;1 - pressure strain term; uij;2 - fast stress and strain term;uij;w - wall radiation term.Buoyancy model:

Gij ¼ bui

Prigi

@T@xj

þ gj@T@xi

� �

Pri - turbulent Prandtl number: 0.85.Dissipative term model:

eij ¼ 23dijqe

e - scalar dissipation.3.4. Difference schemes and algorithms

The commonly used difference schemes are the first-order dif-ference scheme, two-order difference scheme and QUICK differ-

RSM RNG k- Realizable k-

Fig. 6. velocity nephogram of z = 12 cross-section using RSM, RNG k-e, Realizable k-e model.

RSM RNG k- Realizable k-

Fig. 7. The tangential velocity of the x = 0 cross –section using RSM, RNG k-e,Realizable k-e model.

640 Z. Kang et al. / Applied Thermal Engineering 123 (2017) 635–645

ence scheme. The first-order difference scheme is easy to convergebut the error is relatively large. The two-order difference schemeoptimizes the accuracy error and the simulation error is reduced,

RSM RNG k- Realizable k-

Fig. 8. The axial velocity of the x = 0 cross –section using RSM, RNG k-e, Realizablek-e model.

but it still has some error. The QUICK difference scheme not onlyhas stability but also carries on the further optimization to the pre-cision cutting and further reduces the false diffusion term. Themost accurate results can be obtained by using the QUICK differ-ence scheme [22]. According to the characteristics of high swirlflow in the cyclone separator, the PRESTO pressure interpolationscheme should be adopted. The SIMPLEC coupled method is theoptimization of the SIMPLE coupled method. The convergence ofthe high spin fluid flow problem is limited by the pressure-velocity coupling. We use the SIMPLEC pressure-velocity couplingmethod to accelerate the convergence of the solution and set thedefault values for the SNCR component equation, momentumand energy equations.

3.5. Boundary condition setting

According to the 660 MWCFB cyclone separator SNCR designnozzle, we select 3 reducing agents nozzle. Two spray points arepositioned at the air inlet and the tube body tangent, and the thirdspray point is located on the outer wall of the tube body at thesame height. The reducing agent and flue gas can be well mixed.The concrete structure schematic diagram is shown in Fig. 9. Thespecific locations are shown in Table 4.

Based on the full analysis of the flow field in the cyclone sepa-rator, the model is set as follows:

(1) Input: speed of input (inlet velocity), 20 m/s.(2) Exit: free flow (outflow), flow rate of 1.(3) Particle capture port: the wall, set to trap type when adding

particles.(4) Wall: standard wall function, no slip wall.(5) The effect of solid particles on the SNCR reaction is not con-

sidered in the cyclone separator.(6) The main components of the flue gas are N2, CO2, H2O, O2

and NO, and the components are calculated by the thermo-dynamic calculation of the boiler.

(7) We select ammonia solution with a concentration of 15% asthe reducing agent. NSR is set to 1.5.

The liquid droplet evaporation process is set up by the DPMmodel. Evaporation generates gaseous NH3.

The main components of flue gas are N2, CO2, H2O, O2 and NO,and each component is obtained by boiler thermodynamic calcula-tion. Imported gas phase statistics and urea injection statistics areshown in Tables 5 and 6.

The mixing degree of reducing agent and NOx in flue gas is eval-uated by investigating the concentration distribution of reducingagent. The concentration distribution of reducing agent is simu-lated. The calculated results are shown in Fig. 10. Coordinate valueindicates the mass fraction of ammonia reductant.

Fig. 9. Schematic diagram of injection point positions.

Table 4Injection point positions of reducing agent.

x y z

Injection point 1 �2.42 4.26 19.8Injection point 2 �2.42 4.26 15.3Injection point 4 �4.25 0 15.3

Table 6Urea injection statistics.

Urea mass flow rate(kg/s)

Dilution water mass flow rate(kg/s)

Urea concentration(%)

0.024 0.136 15

Fig. 10. Mass fraction of ammonia reducing agent.

Z. Kang et al. / Applied Thermal Engineering 123 (2017) 635–645 641

As can be seen from Fig. 10, the ammonia nitrogen molar ratio isbetween 1 and 2. It can not only take into account the mixing in theflue gas enrichment area, but also take into account the mixing offlue gas entering cyclone separator dead zone. Due to offset theeffect of upward rotating airflow, the distribution of the ammoniareducing agent in the core cylinder is more uniform, which ensuresthe good mixing of the reducing agent and the flue gas.

The SNCR de-NOx reaction is slow. The various components ofthe equation and the momentum and energy equations shouldbe solved first. On the basis of the flow field, temperature fieldand concentration of the cyclone separator, we solve the chemicalreaction of SNCR denitrification [23]. The metal composition of thecirculating ash particles entering the cyclone will have an effect onthe SNCR reaction. SiO2 and Al2O3 have little effect on SNCR reac-tion, but Fe2O3, Fe3O4 and CaO have an effect on SNCR [24]. Butit is difficult to calculate the metal composition of the circulatingash (relating to combustion coal, operating parameters, boiler size,

Table 5Imported gas phase statistics.

Gas flow rate(kg/s)

Temperature(K)

Composition (mass fraction%)

789.48 1173 N2:62.83%; CO2:22.79%; H2O:9.95%;O2:4.42; NO:0.01%.

etc.), and the content of metal components in circulating ash isless. Circulating ash has little effect on SNCR reaction, which canbe ignored in Engineering Research. In the study of SNCR denitrifi-cation, only the gas phase steady flow field is considered forcyclone separator. Without considering the influence of solid par-ticles on SNCR reaction, many scholars [4,25] use this method tostudy the characteristics of SNCR reaction and get accurate results.The same method is used to simplify the simulation of SNCR reac-tion. The SNCR de-NOx efficiency calculation is shown in formula(16):

gNO ¼ £NO;in �£NO;out

£NO;inð16Þ

4. The SNCR characteristics of large-scale cyclone separator

In the supercritical CFB boiler the SNCR equipment is installedto achieve the low-emission standard in the cyclone separator.Although a large number of scholars have done extensive researchon the SNCR de-NOx system, this mainly focused on the SNCR ofregular CFB boilers. When it comes to the supercritical CFB boiler,whose inner flow field is more complicated, affecting the charac-teristics of the SNCR definitely, few people have done this kind ofresearch. With the method of the optimized mechanism and CFDsimulation, we study the characteristics of the SNCR of the cycloneseparator.

4.1. Effect of temperature

SNCR reactions are affected most by temperature, where thetemperature window is usually between 1123 K and 1323 K [26],and the main reaction follows the pathway shown below [27,10]:

4NH3 þ 4NOþ O2 ! 4N2 þ 6H2O ð17ÞUnder high temperature, the NH3 oxidation reactions are as

follows:

642 Z. Kang et al. / Applied Thermal Engineering 123 (2017) 635–645

4NH3 þ 3O2 ! 2N2 þ 6H2O ð18Þ

4NH3 þ 5O2 ! 4NOþ 6H2O ð19ÞWhen the temperature is low, the SNCR reaction is too slow to

achieve the goal of de-NOx. On the other hand, under a relativelyhigh reaction temperature, the NH3 will take pathways (18) and(19) above, which will decrease the reducer and produce NO, caus-ing a drop in the SNCR de-NOx efficiency. As a result, for actualoperation, choosing an appropriate reaction temperature is critical.The SNCR de-NOx efficiency affected by temperature is shown inFig. 11.

As Fig. 11 shows, the SNCR de-NOx efficiency first increasesthen decreases, and it reaches a peak at the temperature of1223 K. Choosing NO transformation efficiency higher than 40%,the SNCR de-NOx reaction temperature window is 1123–1323 K.

In the system based on NH3 for deoxidizing flue gas NOx, NH3

reacts with OH and H groups to generate NH2 under the optimizedmechanism reaction (5), then NH2 group reduces the NO in the fluegas under the optimized mechanism reactions (6) and (7).

Optimized mechanism reaction (5) is affected most by the tem-perature, and when the temperature is below 1123 K, some of theNH3 turns into NH2, and furthermore the optimized mechanismreactions (7) and (8) are inhibited, which decreases the SNCR de-NOx efficiency. However, at temperatures higher than 1223 K,the OH group will increase greatly, and be oxidized into NH2 groupwith simplified mechanism reactions (6) and (14), which alsodecreases the SNCR de-NOx efficiency.

4.2. Effect of molar ratio of ammonia and nitrogen

With NH3 playing a role in the reducing agent to NOx without acatalyst, the quantity of NH3 is critical to de-NOx. According to theprinciple of chemical reaction equilibrium, reaction (17) will movetowards the positive direction as the quantity of NH3 increases,which will raise the SNCR de-NOx efficiency and make reactions(18) and (19) move towards the positive direction. As reactions(18) and (19) move towards the positive direction, the reducingagent will be consumed and the newly generated NO will reducethe transformation rate [28]. As a result, increasing the quantityof reducing agent will increase the cost and vice versa. So it isimportant to choose the appropriate quantity of reducing agentto balance the efficiency and cost in actual operation. The SNCRde-NOx efficiency affected by the NSR under different tempera-tures is shown in Fig. 12.

Fig. 12 shows that, when the temperature is lower than thereaction temperature window, increasing the NSR has little effect

1050 1100 1150 1200 1250 1300 135010

20

30

40

50

60

70

80

90

SNC

R d

e-N

Ox e

ffic

ienc

y/%

Temperature (K)

NSR=1 NSR=1.2 NSR=1.4 NSR=1.6 NSR=1.8 NSR=2

Fig. 11. Effect of temperature on SNCR de-NOx efficiency under different NSR.

on the SNCR de-NOx efficiency. With the NSR at 1073 K from 1to 2, the SNCR de-NOx efficiency increases from 16.3% to 21.1%;with the NSR at 1123 K from 1 to 2, the SNCR de-NOx efficiencyincreases from 26.4% to 45.1%. This is because when the tempera-ture is low, inhibiting the formation of OH and O groups, the NH2

content is less, resulting in a reduction in SNCR de-NOx efficiency.At low temperature, the temperature is the controlling factor of theSNCR reaction. So when the reaction temperature is low, and add-ing the reducing agent NH3, the SNCR de-NOx efficiency cannot beeffectively improved. When the temperature exceeds 1123 K, thetemperature is no longer the controlling factor of SNCR and theincrease of NSR can significantly increase the SNCR de-NOx effi-ciency. At different temperatures, the effects of the NSR on theSNCR de-NOx efficiency show the same change trend. But whenthe NSR increases to a certain extent, the rate of increase in SNCRde-NOx efficiency gradually slows down. Analysis shows that NH2

can generate four OH free radicals, so the NH3 has sufficient OHfree radicals to produce NH2. NH2 undergoes the chain reactionof NH2 + NO generation of N2 and NH2 + NO generation of NNH.At around NSR = 1.5, the two chain reactions compete to reachequilibrium [29]. The NSR then continues to increase, only makingthe SNCR de-NOx efficiency increase slowly, and the increase in theamplitude tends to be gentle. After comprehensive considerationof the SNCR de-NOx efficiency and operational economic factors[30–32], the NSR is set to 1.5.

5. Performance of large-scale cyclone separator

A large-scale cyclone will lead to the flow field becoming moredisordered in the cyclone separator and will have a major effect onthe energy loss and the separation of solid particles [33]. Researchon the performance of large-scale cyclone separators can provide ameaningful reference for practical operation.

5.1. Study of gas flow field

The tangential velocity, axial velocity and radial velocity of thex = 0 cross-section are shown in Fig. 13.

The axial symmetry of tangential velocity distribution is better.Show the characteristics of strong swirl in separator. All the max-imum tangential velocity point form an interface which divides theinternal flow field of the separator into two parts. The velocity ofthe central area is large, so the centrifugal force is large. It is favor-able for separation. While velocity of the external area is small, andthe effect of the airflow carrying particles weakens, which is favor-able for the particles to be captured near the wall. Partial area

1.0 1.2 1.4 1.6 1.8 2.0

20

30

40

50

60

70

80

90

SN

CR

de-

NO

x effi

cien

cy /%

NSR

1073K 1123K 1173k 1223k 1273k 1323k

Fig. 12. Effect of NSR on the SNCR de-NOx efficiency at different temperatures.

Fig. 14. velocity vector at the z = 12 m section.

5 10 15 20 25 30

0

1000

2000

3000

4000

5000

Tota

l pre

ssur

e/Pa

Inlet velocity/m/s

Fig. 15. Variation curve of total pressure difference at different inlet velocities.

Z. Kang et al. / Applied Thermal Engineering 123 (2017) 635–645 643

velocity is negative indicating secondary swirls existing. The axialsymmetry of axial velocity distribution is also better. The boundarypoint is related to the shape of the cyclone. Radial velocity is muchsmaller than tangential velocity and axial velocity. Most are cen-tripetal, and only the central vortex core has the radial flow. Schol-ars [34] have found that the radial velocity is mostly centripetal,between 0 and 3 m/s. Lower part of the outlet section has the lar-gest radial velocity, up to 7 m/s.

Fig. 14 is velocity vector diagram at z = 12 m section.As can be seen from Figure 146 after the airflow enters the

cyclone separator, due to the restriction of the wall surface of thecyclone separator, the air flow turns downward to form the outerswirl. When the air reaches the bottom, the air has to flow upwardto form an internal swirl. External airflow flows downward andinternal airflow flows upward. Direction of rotation is the same.

5.2. Pressure distribution of large-scale cyclone separator

Fig. 15 is the variation curve of the total pressure differencebetween the inlet and outlet of the cyclone separator at differentinlet velocities.

It can be seen from Fig. 15 that with the increase of the inletvelocity, the pressure drop of the cyclone separator increases,and the rate of increase is also more and more powerful. The pres-sure drop from 295.84 Pa soon increases to 4231.2 Pa when theinlet velocity rises from 5 m/s to 30 m/s. From an energy point ofview, increasing the inlet velocity of the cyclone separator willincrease the loss of energy. This is because too high inlet speed willaccelerate the wear of the cyclone separator. It should be based onthe separation performance of the cyclone separator as far as pos-sible to use a slightly lower inlet velocity, saving energy. In normaloperation at the entrance speed of 20 m/s or so, the diagram showsthat with the large-scale cyclone’s normal operation, the total pres-sure difference is 1987.6 Pa.

5.3. Classification efficiency of large-scale cyclone separator

Gas-phase turbulent flow will affect the particle phase, simi-larly, and particles will also affect the gas phase. But the interactionbetween particles can be neglected. So the phase coupling stochas-tic trajectory model is used to simulate the particle motion [35].Particles do high-speed rotating flow in the separator. The largeparticles are thrown out under the centrifugal force and fall intothe dust exhaust port after several collision walls [36,37]. Set these

tangential velocity axi

Fig. 13. The tangential velocity, axial velocity an

particles to be captured; Gas fluid carries small particles which aredischarged from the exhaust pipe. Set these particles to escape. Setthe number of particles into the cyclone n, the number of particlescaptured ni. Classification efficiency is calculated by formula (20):

al velocity radial velocity

d radial velocity of the x = 0 cross-section.

Fig. 16. The particle velocity trajectories.

5 10 15 20 25 3045

50

55

60

65

70

75C

lass

ifica

tion

effic

ienc

y/%

Inlet velocity/m/s

Fig. 17. Curve of different inlet velocity classification efficiency.

644 Z. Kang et al. / Applied Thermal Engineering 123 (2017) 635–645

g ¼ ni

n� 100% ð20Þ

In this paper, set solid particles as spherical ash particles. Parti-cle density is the average of ash density. According to the m = qv,sphere volume v = 4/3pr3, the ratio of mass is proportional to thethird power of particle size. The analysis of particle size includesthe consideration of particle quality. Use DPM model to simulatethe particle phase in Euler-Lagrangian coordinate system. The par-ticle phase motion equation is as follows:

mdV

!

dt¼ FD

!þ FP

!þ FA

!þ FB

!þ FS

!þ FM

!þ FG

!þ FC

!ð21Þ

FD—Drag force;FP—Pressure gradient force:

Fp

!¼ �4

3pR3rPf

FA—Additional mass force:

FA

!¼ �1

243pR3qf

� �dup

!

dt� duf

!

dt

!

FB—Basset force:

FB

!¼ 6R2 ffiffiffiffiffiffiffiffiffiffiffiffi

plqf

p Z t0

ts0

dðu0f � u0

pÞ=dtffiffiffiffiffiffiffiffiffiffiffiffiffitp � s

p dt

FS—Saffman force:

FS

!¼ 81:2R2ðqflÞ

12k

12

FM—Maguns force:

FM

!¼ pR3qf x

! �ðuf! �up

! Þ½1þ OðRepÞ�FC—Volume forceIn the gas-solid two-phase flow, the particle size is small and

the concentration is very thin. So in this paper, fluid drag force isthe main force, and the other compared with very small, can beneglected [38,39].The particle velocity trajectories are shown inFig. 16.

Classification efficiency is one of the important indexes to eval-uate the separation performance of the cyclone separator. The effi-ciency of classification is the separation efficiency of a givenparticle size, which has nothing to do with the particle size atthe inlet [40]. So it can be seen that evaluating the performanceof the cyclone separator with classification efficiency is more sig-nificant. Without consideration of the operating conditions, thetotal efficiency of the cyclone separator is meaningless [34].Because the cyclone can be completely separated from the largeparticles, the key lies in the separation of small particles so as tomainly simulate the small particles. The particle diameter obeysthe Rosin–Rammler distribution, with particle sizes 0.05–0.1 mm,average particle size 0.075 mm, and 2000 particles are selectedrandomly from the particle mass flow to calculate the classificationefficiency. Fig. 17 is the classification efficiency change curve at dif-ferent inlet velocities.

As can be seen from Fig. 17, when the inlet velocity increases,the classification efficiency of the cyclone separator will increase;when the inlet velocity decreases, the separation efficiency of thecyclone separator will decrease. At the same time, it can be seenthat the influence of the change of inlet velocity on the classifica-tion efficiency curve is relatively large. When the inlet velocityincreases, the centrifugal force of the particles is increased. But tur-bulence is enhanced, and the ash bucket back mixing becomesmore serious, causing the small size particle classification effi-ciency not to obviously improve and become even lower [41].

To sum up, the entry speed is not a matter of the bigger the bet-ter. It should be based on the cleanliness of the outlet air flow

requirements to select the appropriate inlet velocity which canachieve the purpose of separation, save energy and extend the lifeof the cyclone separator. When the inlet velocity of the normaloperation is 20 m/s, the classification efficiency of the large-scalecyclone separator is 66.74%.

6. Conclusion

Based on Oyvind Skreiberg’s SNCR de-NOx reaction mechanismand under the reaction conditions of the cyclone separator of theCFB, the sensitivity analysis method of CHEMKIN software is usedto optimize the SNCR reaction mechanism. The optimized SNCRreaction mechanism and CFD software are combined to simulatethe performance of SNCR de-NOx and the characteristics in a660 MW supercritical circulating fluidized bed cyclone separator.The following conclusions are obtained:

(1) CHEMKIN software is used to determine the degree of corre-lation between each reaction in the model and the compo-nents of interest and we obtain the optimized mechanismof the 18 basic elements and compare these with the exper-imental and detailed mechanism. It is proved that the opti-mized mechanism can simulate the SNCR de-NOx processin the cyclone separator of the supercritical CFB boiler.

Z. Kang et al. / Applied Thermal Engineering 123 (2017) 635–645 645

(2) At reaction temperatures of 1073–1323 K, the SNCR de-NOxefficiency first increases and then decreases with theincrease of temperature, reaching a peak at 1223 K. The con-version rate of NO is more than 40% and the correspondingtemperature is set to the temperature window. The SNCRde-NOx reaction temperature window is 1073–1323 K.

(3) When the temperature is lower than the reaction tempera-ture window, temperature is the controlling factor andimproving the NSR cannot effectively improve the SNCRde-NOx efficiency. When the temperature is increased tothe reaction temperature window, the NSR can effectivelyimprove the SNCR de-NOx efficiency. But by enhancing theNSR to a certain extent, the SNCR de-NOx efficiency increasewill gradually slow down. Considering the SNCR de-NOxefficiency and economic factors, the NSR is set to 1.5 asappropriate.

(4) The total pressure difference of the cyclone separatorincreases with the inlet velocity and the rate of increasegradually increases. When the inlet velocity of the cycloneseparator is 20 m/s, the total pressure difference is1987.6 Pa. The classification efficiency of the cyclone separa-tor increases with the inlet velocity and the rate of increasegradually decreases. When the inlet velocity of the cycloneseparator is 20 m/s, the classification efficiency is 66.74%.

Acknowledgment

This study was supported by the Fundamental Research Fundsfor the Central Universities (JB2015RCY06).

References

[1] W. Qu, J.Q. Zhou, J.H. Yang, G.Y. Qin, SNCR denitrification system in CFB boilers:optimization and application, Therm. Power Gener. 43 (1) (2014) 133–136.

[2] C.H. Luo, E.L. Wang, Selective non-catalytic reduction technology and its de-NOx system used in CFB boilers, J. Power Eng. 28 (3) (2008) 442–446.

[3] M. Oliva, M.U. Alzueta, A. Millera, R. Bibao, Theoretical study of the influence ofmixing in the SNCR process comparison with pilot scale data, Chem. Eng. Sci.55 (22) (2000) 5321–5332.

[4] D.B.N. Thanh, L. Young-II, J.K. Seong, H.E. Won, S.Y. Kyung, Experiment andcomputational fluid dynamics (CFD) simulation of urea-based selectivenoncatalytic reduction (SNCR) in a pilot scale flow reactor, Energy Fuels 22(6) (2008) 3864–3876.

[5] Z. Zhong, J. Lan, Desulfurization in the reducing state and denitrification withammonia injection in an FBC test facility, Fuel 77 (1998) 757–762.

[6] P. Glarborg, K. Dam-Johansen, J.A. Miller, The reaction of ammonia withnitrogen dioxide in a flow reactor: implication for the NH2 + NO2 reaction, Int.J. Chem. Kinet. 27 (12) (1995) 1207–1220.

[7] P.D.J. Glabrorg, M. Kim, A. Jmaes, et al., Modeling the thermal DENOx process inflow reactors. Surface effects and nitrous oxide formation, Int. J. Chem. Kinet.26 (4) (1994) 421–436.

[8] Z.M. Lu, Study on the Mechanism of SNCR Reaction and Its MixingCharacteristics, Zhejiang University, 2006.

[9] J.A. Miller, C.T. Bowman, Mechanism and modeling of nitrogen chemistry incombustion, Prog. Energy Combust. Sci. 15 (4) (1989) 287–338.

[10] M. Shiva, B. Jan, P.K. Jonathan, NOx formation and selective non-catalyticreduction (SNCR) in a fluidized bed combustor of biomass, Sci. Direct 5 (2010)1393–1409.

[11] P. Kilpinen, Department of Chemical Engineering, Combustion Chemistry (Ph.D. thesis), Akademi University, Research Group, 1992.

[12] M. Röhrig, H.G. Wagner, Combust. Inst. 25 (1994) 975–981.

[13] J. Brouwer, M.P. Heap, D.W. Pershing, et al., A model for prediction of selectivenon-catalytic reduction of nitrogen oxides by ammonia, urea, and cyanuricacid with mixing limitations in the presence of CO, in: The Twenty SixthInternational Symposium on Combustion, 1996.

[14] R. Rota, D. Antos, E.F. Zanoelo, et al., Experimental and modeling analysis of theNOxOUT process, Chem. Eng. Sci. 57 (2002) 27–38.

[15] X. Han et al., Detailed modeling of hybrid reburn/SNCR processes for NOxreduction in coal-fired furnaces, Combust. Flame 132 (2003) 374–386.

[16] V. Yakhot, S.A. Orszag, Renormalization group analysis of turbulence, J. Sci.Comput. 1 (1986) 3–11.

[17] T.H. Shih, W.W. Liou, A. Shabbir, et al., A new k-e eddy viscosity model for highReynolds number turbulent flows, Comput. Fluids 24 (1995) 227–238.

[18] R. Huang, A partitioning parallel procedure for numerical simulation of gasflow in a cyclone separator, J. Hvdrodyn. 1 (2002) 52–58.

[19] J.G. Ran, L. Zhang, et al., Numerical simulation of gas-solid flow motioncharacteristics and deposition efficiency of particles in water-film cycloneseparator, J. Chem. Indust. Eng. 54 (10) (2003), 13,1–1396.

[20] H.G. Wang, S. Liu, Application and comparison of different turbulence modelsin the three-dimensional numerical simulation of cyclone separators, J. Eng.Therm. Energy Power 18 (4) (2003) 337–342.

[21] G.Y. Zhou, Numerical simulation and experimental study on the performanceof gas-liquid spiral separator, Chin. J. Chem. Eng. 13 (5) (2005) 596–603.

[22] T.D.B. Nguyen, Y. Lim, S. Kim, et al., Experimental and computational fluiddynamics (CFD) simulation of urea-based selective noncatalytic reduction(SNCR) in a pilot-scale flow reactor, Energy Fuels 22 (6) (2008) 3864–3876.

[23] Z. Huang, B. Sun, et al., Numerical simulation of cyclonic separator for one CFBboiler and retrofit thereof, Therm. Power Gener. 37 (6) (2007) 38–41.

[24] X.S. Hou, D.W. Chang, H. Zhang, Experimental study on the effect of circulatingashes on ammonia oxidation, Proc. CSEE 28 (5) (2008) 57–62.

[25] H.S. Kim, M.S. Shin, D.S. Jang, et al., Numerical study of SNCR application to afull-scale stoker incinerator at Daejon 4th industrial complex, Appl. Therm.Eng. 24 (14–15) (2004) 2117–2129.

[26] Q. Li, Y.X. Wu, H.R. Yang, J.F. Lv, Simulation and optimization of SNCR process,CIESC J. 64 (5) (2013) 1789–1796.

[27] J.A. Miller, M.C. Branch, R.J. Kee, A chemical kinetic model for the selective ofnitric oxide by ammonia, Combust. Flame 43 (1981) 81–98.

[28] V.M. Zamansky, V.V. Lissianski, P.M. Maly, et al., Reactions of sodium species inthe promoted SNCR process, Combust. Flame 117 (4) (1999) 821–831.

[29] Z.X. Tang, H.M. Zeng, C.B. Cen, et al., Experimental study of selective non-catalytic reduction of NOx by NH3 in flue gas containing SO2, Proc. CSEE 33(20) (2013) 34–39.

[30] J.T. Hao, W. Yu, P. Ru, et al., Influences of flue gas components and additives onNO reduction in the selective non- catalytic reduction process, Proc. CSEE 35(12) (2015) 3054–3060.

[31] T.D.B. Nguyen, Y. Lim, S. Kim, et al., Experimental and modeling analysis of theNOxOUT process, Chem. Eng. Sci. 57 (2002) 27–38.

[32] R.K. Lyon, J.E. Hardy, Discovery and development of the thermal DeNOxprocess, Ind. Eng. Chem. Fund. 25 (1) (1986) 19–24.

[33] T.M. Shih, A.L. Ren, Primitive variable formulations using nonstaggered grids,Numer. Heat Transfer (1984) 413–428.

[34] F.P. Qian, M.H. Zhang, Numerical study of the influence of temperature onseparation performance of cyclone separators, J. Power Eng. 26 (2) (2006) 253–257.

[35] O.Y. Jie, J.H. Li, Particle-motion-resolved discrete model for simulation gas-solid fluidization, Chem. Eng. Sci. 54 (1999) 2077–2083.

[36] H. Shalaby, K. Wozniak, G. Wozniak, Numerical calculation of particle- ladencyclone separator ow in LES, Eng. Appl. Comput. Fluid Mech. 2 (2008) 382–392.

[37] P. Kozolub, A. Klimanek, R.A. Bialecki, W.P. Adamczyk, Numerical simulation ofa dense solid particle flow inside a cyclone separator using the hybrid Euler-Lagrange approach, Particuology (2016), http://dx.doi.org/10.1016/j.partic.2016.09.003.

[38] M. Renwei, J. Ronald, J.H. Thomas, Particle dispersion in isotropic turbulenceunder stokes drag and Basset force with gravitational settling, Fluid Mech. 225(1991) 95–481.

[39] J.F. Lu, L. Yu, Y.J. Zhang, et al., A 600MW supercritical circulating fluidized bedboiler, J. Power Eng. 27 (4) (2007) 497–501.

[40] B. Wang, M. Xu, K. Chu, A. Yu, Numerical study of gas-solid ow in a cycloneseparator, Appl. Math. Model. 30 (2006) 1326–1342.

[41] J. Derksen, S. Sundaresan, H. Van den Akker, Simulation of mass-loading effectsin gas-solid cyclone separators, Powder Technol. 163 (2006) 59–68.

本文献由“学霸图书馆-文献云下载”收集自网络，仅供学习交流使用。

学霸图书馆（www.xuebalib.com）是一个“整合众多图书馆数据库资源，

提供一站式文献检索和下载服务”的24 小时在线不限IP

图书馆。

图书馆致力于便利、促进学习与科研，提供最强文献下载服务。

图书馆导航：

图书馆首页 文献云下载 图书馆入口 外文数据库大全 疑难文献辅助工具