You are on page 1of 30

Optimal Combustion to Counteract Ring Formation in Rotary Kilns

Michele Pisaroni michele.pisaroni@gmail.com


Scientific Computing Group, Delft Institute of Applied Mathematics, Faculty of Electrical Engineering, Mathematics and Computer Science Delft University of Technology, The Netherlands

Abstract Avoiding the formation of rings in rotary kilns is an issue of primary concern to the cement production industry. We developed a numerical combustion model that revealed that in our case study rings are typically formed in zones of maximal radiative heat transfer. This local overheating causes the overproduction of the liquid phase of the granular material that tends to stick to the ovens wall and to form rings. To counteract for this phenomenon, we propose to increase the amount of secondary air injected to cool the oven. Experimental validation at the plant has repeatedly shown that our solution is indeed effective. For the first time in years, the kiln has been operation without unscheduled shutdowns, resulting in a monthly five-digits cost saving. Keywords: rotary kiln, combustion model, ring formation.

I. INTRODUCTION
A. Rotary kilns A rotary kiln is a long horizontal cylinder slightly inclined tilted on its axis. The objective of a rotary kiln is to drive the specific bed reactions, which, for both kinetic and thermodynamic reasons, require high bed temperature. In direct-fired kilns, the energy necessary to raise the bed temperature to the level required for the intended reactions originates with the combustion of hydrocarbon fuels in the freeboard near the heat source of burner. This energy is subsequently transferred by heat exchange between the gas phase and the bed. The material to be processed is fed into the upper end of the cylinder. As the kiln rotates, material gradually moves down towards the lower end, and may undergo a certain amount of stirring and mixing. Heat transfer between the freeboard and the bed is rather complex and occur by all the paths established by the geometric view factors in radiation exchange. All these manifest themselves into a combined transport phenomenon with the various transport processes coming into play in one application. [2] In our kiln in Rotterdam we produce Calcium Aluminate Cement (CAC), a very white, high purity hydraulic bonding agents providing controlled setting times and strength development for todays high performance refractory products. The cement is made by fusing together a mixture of a calcium bearing material (limestone) and an aluminium-bearing material. The calcined material drops into the "cool end" of the kiln. The melt overflows the hot end of the furnace into a cooler in which it cools and solidifies (Fig.1). The cooled material is then crushed and grounded. One of the primary issues in our production is the ring formation.

B. Ring Formation As material slides and tumbles through a kiln, a thin layer of dust invariably forms on the surface of the refractory lining. Some zones of the kiln may be particularly prone to particle accumulation and the combined effect of particular thermal and flow conditions results in the formation of cylindrical deposits, or rings. As the ring grows thicker, the available opening of the kiln is decreased, hindering the flow of lime product and flue gasses through the kiln. In our CAC (Calcium Aluminate Cement) kiln in Rotterdam we observed in particular front-end / midkiln rings [9]. They are located close to the burner and are presumably caused by the high temperature in this area, particularly when the refractory surface is overheated by direct impingement of the burner flame. These are the most common and also the most troublesome type of rings. They cannot be reached from outside the kiln and is therefore impossible to remove while the kiln is in operation. In severe cases, rings grow rapidly and cause unscheduled shutdowns of the kiln. Depending on the severity of the problem, maintenance labour, make-up lime purchase, and lime mud disposal can bring the cost of a ring outage to 150,000 per shout down. In the last 3 years an average of 1 ring formation per month was registered, 70% of which caused shut down of the kiln.

II. COUNTERACTING RING FORMATION


Even though the temperature inside the kiln is obviously an important factor in ring formation, it cannot be measured easily. Since the gas temperature profile along the kiln varies significantly with burner configurations, fuel, moisture content and the insulating lining it is difficult to obtain the exact temperature at which rings are formed. To solve this problem we started our analysis from the material phase diagram [3] (Fig. 2). Rings are

caused by too much liquid phase in the region of the kiln where we expect the maximum temperature and radiative heat transfer on the interface between the gas and insulating lining. We decided to run our simulation with an empty kiln and analyse the temperature of the gas-lining interface as a parameter to understand the amount of liquid phase in the kiln. This simplification allow to reduce the complexity of the model and in particular can give us a valid estimation of the amount of heat produced by the flame and exchanged with the granular material flowing in the kiln. We decided to recreate the thermal and fluid conditions in an empty kiln model to reduce the computational efforts needed to simulate the granular moving bed, that occupies no more than 5% of the volume of the kiln.

III. COMPUTATIONAL MODEL


In the kiln we examine a flame projected from a burner-pipe inside the kiln generates the hot gases (Fig.3). In the hot zone of the kiln we inject hot air from an air inlet placed on the top of the burner, natural gas from the injectors of the burner and a small amount of cold air form the cooling slot placed around the inlets to cool down the surface of the burner. Assuming that it is not possible to modify radically the geometry of the kiln, we were interested in finding out if it was possible to achieve an effective reduction of ring formations, a setup of the kiln able to counteract the growing of rings and keep the temperature and the conditions as much as possible close to the optimum for the production. Hence we decide to investigate in the effects of the secondary air injection effects on the flame and lining temperature. We decided to run the simulation on a full 3D geometry of the kiln because the flow conditions inside the kiln are not symmetric as a result of the presence of the rectangular air intake. The geometry of the complete kiln is shown in Fig. 4 and Fig.5.

A. Grid Generation We used a polyhedral mesh with refinement in the critical zone (Fig.6, Fig.7 and Fig.8). At the end of the refinement we have 2.8 million of cells. Polyhedral mesh provide a balanced solution for complex mesh generation problems. They offer the same automatic meshing benefits as tetrahedral while overcoming some disadvantages. A major advantage of polyhedral cells is that they have many neighbours (typically of order 10); gradients can be much better approximated than with tetrahedral cells. More neighbours obviously means more storage and computing operations per cell, but this is more than compensated by a higher accuracy. Polyhedral cells are also less sensitive to stretching than tetrahedra. For a polyhedron with 12 faces, there are six optimal directions, which, together with the larger number of neighbours, leads to a more accurate solution with a lower cell count. A more detailed analysis of polyhedral mesh can be found in Peric (2004) [5]. B. Governing Equations Turbulent combustion arises from the two-way interaction between chemistry and turbulence. When a flame interacts with a turbulent flow, turbulence is modified by combustion because of the strong flow accelerations through the flame front induced by heat release, and because of the large changes in kinematic viscosity associated with temperature changes. Compared to premixed flames, turbulent non-premixed flames exhibit some specific features that have to be taken into account. Non-premixed flames do not propagate: they are located where fuel and oxidizer meet. This property has consequences on the chemistry/turbulence interaction: without propagation speed, a non-premixed flame is unable to impose its own dynamics on the flow field and is more sensitive to turbulence. Molecular diffusion may also be strongly affected (differential diffusivity

effects). We use the following list of models in our computations: C.1 Turbulence Model: Realizable K-epsilon, C.2 Chemical Reaction Model, C.3 Combustion Model: Eddy Break Up (EBU), C.4 Radiation

Model: Participating Media (DOF). These models are described in detail in Appendix 2.

IV. COMPUTATIONAL RESULTS


In this section we present the computational results obtained with the models described in the previous section. The computation was done with STAR-CCM+ [1]. We run this model on a 10-node cluster for 3500-4000 iterations. The time needed for a single computation was 3-3.5 days. We present here 2 configuration of the kiln: STD_Config is the standard configuration used in the past years in our plant with a volume air-fuel ratio of 9 and H_Air is the new configuration with a volume air-fuel ratio of 12.

A. Standard Configuration (STD_Config) With standard configuration we mean that we run our model with all the parameters set as in the standard production situation. The first key result is the temperature profile in the kiln shown in Fig.10. In the analysis of ring

formation one of the most interesting parameter that must be taken into account is the temperature at the interface between the hot gas and the solid wall (Fig.11). It is well shown that the hottest zone is placed in a particular position exactly in the interval at 4.5-7 m from the burner. This is also the interval where in reality we can find the most severe ring formations in our kiln. In Fig. 11 it is possible to see that the flame (represented as the iso-surface with concentration of CH4 higher than 0.01) peak temperature is placed exactly near the zone where we measure the maximum wall temperature. The second key result is the incident radiation on the same hot gas-solid interface. A particular zone (Fig.12 and Fig.13) is exactly placed in the highest temperature zone. The material that is flowing in this section absorbs the maximum heat due to this peak of radiation produced by the flame propagating above it. It is reasonable to hypnotize that the liquid phase of material in this zone is so high that the material becomes sticky and attack the solid wall. B. Higher Air-Fuel Ratio (H_Air) With higher air-fuel configuration we mean that we run our model with the volume air-fuel ratio set to 12. We keep the injectors of fuel and the cooling air of the burner set as in the standard configuration. We observe that in this configuration the gas temperature is lower than in the standard case (Fig. 14). The blue zone on the top, due to the secondary air injection, propagates further because more air is injected in the kiln. The secondary air injection acts as a cooler for the wall. Also the incident radiation in the gas-solid interface is lower and the region with maximum radiation is smaller compared with the previous case (Fig.16 and Fig.17). The material that is flowing in this section absorbs less heat due to this lower radiation. It is reasonable to hypnotize that the liquid phase of material in this zone is lower and that the material cannot attack the wall.

In an X-Y plot we can easily see the temperature and the radiation peak in the gas-solid interface are lower then the previous case (c.a -5% Temperature and -11% Incident Radiation) (Fig.18 and Fig.19).

V. EXPERIMENTAL VALIDATION AT PLANT


The reported simulation model, shows that the flame peak temperatures and the interface between gas and solid can be reduced significantly. The hypothesis is that reducing the flame temperature will result in reduction of heat-transfer via radiation, the clinker bed is below its melting regime. That will stop the growing of the ring dam. On August the 28th 2011, a severe ring formation was reported from the plant (Fig.20). After assessing the situation, it was decided to increase the A/F ratio substantially (from 9 to 12). The effects of A/F ratio 12 that we were expecting due to previous numerical calculation: flame temperature peaks down interface hot gas-solid wall reduced no significant change in flow pattern radiative heat transfer reduced melting and growing of ring stopped

We observed that: 4 hours later it can be seen that ring stops growing (Fig. 21) 24 hours later the ring size stabilized and started to decrease due to breaking of lumps (Fig.22) 40 hours later the kiln remain stable in operation (Fig. 23) Several days later, the kiln ring diminishes slowly until it was almost destroyed. When the growing of

the ring is stopped and we reach a temperature at which the liquid phase is very low the vibration due to the drive gears of the kiln and the rotation gradually breaks lumps from the ring and after 40 hours the ring is almost destroyed.

ACKNOWLEDGEMENT
We thank Marco Talice for his advice worth more than platina and the CD-ADAPCO London office for the support in using their software.

VI. CONCLUSION
We developed a numerical model allowing to access the effectiveness of measures implemented to counteract the formation of rings in a rotary cement kiln used in use by Almatis B.V. in Rotterdam. In this three-dimensional combustion model, the gas flow, the temperature profile, radiative heat distribution and the concentration of hydro-carbon species in the kiln is taken into account. Simulations show that increasing the volume air-fuel ratio reduces peaks in radiative heat transfer in zones critical to ring formation. This reduction results in turn in less heat being absorbed by the granular material bed, effectively reducing the amount of material liquid phase prone to sticking to the kiln's surface and to forming rings. The validity of our model has been experimental observed at the Almatis plant in Rotterdam. Since August 28th, 2010, the kiln has been operation without unscheduled shut-downs, resulting in a monthly five-digits cost saving.

Appendix 1

Fig.1. General layout of a direct fired, countercurrent feed rotary kiln.

Fig.2. Phase diagram

Fig.3. Burner setup

Fig.4. Outside view of the kiln

Fig.5. Burner zone

Fig.6. Outside view of the grid

Fig.7. Interior refinment

Fig.8. Details of the grid in the burner zone

Fig.10. Temperature profile in the axial section (A/F=9)

Fig.11. Temperature of Hot Gas Solid Wall interface (A/F=9)

Fig.12. Incident radiation Hot Gas Solid Wall interface (A/F=9)

Fig.13. Incident radiation Hot Gas Solid Wall interface (A/F=9)

Fig.14. Temperature profile in the axial section (A/F=12)

Fig.15. Temperature of Hot Gas Solid Wall interface (A/F=12)

Fig.16. Incident radiation Hot Gas Solid Wall interface (A/F=12)

Fig.17. Incident radiation Hot Gas Solid Wall interface (A/F=12)

Fig.18. Temperature vs position (blue A/F=9, red A/F=12)

Fig.19. Incident radiation vs position (cyan A/F=9,greenA/F=12)

Fig.20. Ring structure at A/F=9

Fig.21. Ring structure after 4 hours the imposition of A/F=12

Fig.22. Ring structure after 24 hours the imposition of A/F=12

Fig.23. Kiln returned in a stable condition after 40 hours

Appendix 2
C.1 Turbulence Model The description of the turbulent flows in non-premixed combustion processes using Computational Fluid Dynamics (CFD) may be achieved using three levels of computations: Reynolds Averaged Navier Stokes (RANS), Large Eddy Simulations (LES) or Direct Numerical Simulations (DNS). In current engineering practice, RANS is extensively used because it is less demanding in terms of resources but the closure models describing turbulence and combustion limit its validity. The advantage of RANS is its applicability to any configuration and operating conditions. Considering the complexities and the dimensions of our kiln, the only feasible choice was RANS. To obtain the Reynolds-Averaged Navier-Stokes (RANS) equations, the Navier-Stokes equations for the instantaneous velocity and pressure fields are decomposed into a mean value and a fluctuating component [6]. A K-Epsilon turbulence model is a two-equation model in which transport equations are solved for the turbulent kinetic energy k and its dissipation rate . Various forms of the K-Epsilon model have been in use for several decades, and it has become the most widely used model for industrial applications. In our case we used a Realizable K-Epsilon model. A critical coefficient of the model is expressed as a function of mean flow and turbulence properties, rather than assumed to be constant as in the standard model. This allows the model to satisfy certain mathematical constraints on the normal stresses consistent with the physics of turbulence (realizability). The two equations for the model are:

m the physics of turbulence (realizability). The two equations for the model are: or the model are: er ce @ @ @ t @ k y @ @ + t @ k (@ k ) + (kui ) = + k @ xj @ t ( k ) + @ xi (kui ) = @ xj (1) @t @ xi @ xj @ xj (1) k

d o for turbulent kilnetic energy k for for turbulent kilnetic energy kk turbulent kilnetic energy y is @ @ @ @ @@ t t @" @" ( " ) + ( " u ) = + i he ("ui ) = + @ t @ t (") + @ x@ @x xx i xi " " @@ jj @jxj a 22 " " " " n +C C3 P )b ) CC + P+ C P +S S 1C " 1"(P( k "3 2" "" k + "b 2 " + k k k k ar y dissipation for for dissipation " for dissipation . " Chemical Reactions Model C.2C.2 Chemical Reactions Model
C.2 Chemical Reactions Model

+P + P P " YY S M + k +k Pk + b " + S b M k

(2) (2)

The ame projected from the burner-pipeinside insidethe the kiln kiln is is The ame projected from the burner-pipe gas (more than 90 % of methane (CH4)) that reacts with hot (500-600 C) air (23% of O2) injected from continuously the injection natural gas(more (morethan than 90 90 continuously fedfed by by the injection ofof natural gas the top window visible in Fig.3. that reacts with hot (500-600 C) air (23% %methane of methane (CH4)) % of (CH4)) that reacts with hot (500-600 C) air (23% of O2) injected from the top window visible in Fig.3. through M on a nu Consider a system chemical system of N species reacting Consider a chemical of N species reacting through M reactions: of O2) injected from the top window visible in Fig.3. reactions: The N N X X the Da 0 00 vkj Mk () vkj Mk f or j = 1, M (3) is the k=1 k=1 When 0 00 where Mk is a symbol fot species k, vkj and vkj are molecular kiln, th where M is a symbol for species k, v and v are molecular stoichiometric coefficients of species k brings stoichiometric coefcients of species k in reaction j . Consider mass reaction rates, for species k , ! k is the sum of the Sta becaus rates ! kj produced by all M reactions:
k kj kj

The flame projected from the burner-pipe inside the kiln is continuously fed by the injection of natural

0 00 is a symbol k, vkj and vkj are molecular stoichiometric coefcients of species k reaction 00 j . 00in and 00 whereM M is a symbol fot species k, v v aremolecular molecular where is a symbol fot species k, v and v k kj kj are k kj kj Consider mass reaction rates, for species k , ! is the sum of k j. stoichiometric coefcients of species k in reaction

k =1 =1 wherekM k

k =1 k =1 fot species

stoichiometric coefcients reaction j in reaction j. Consider mass reaction rates,of for species species k, kk is in the sum of rates kj.produced by all M rates ! produced by all M reactions: kj mass reaction rates, for species k , ! Consider is the the sum sum of of Consider mass reaction rates, for species k , ! kk is reactions: rates! kj producedby byall all M reactions: reactions: rates ! produced kj M MM X X ! kj 0 ! k = M ! kj = Wk M vkj Qj with Qj = (4) M M W v X X k kj X X ! kj ! j =1 j =1 0 0 kj kk = = kj =W W v Q with Q Q = (4) ! ! ! ! v Q with (4) kj = k kj k jj jj = kj W v W v k kj k kj j =1 j =1 j =1 j =1 where Qj is the rate of progress of reaction j . The progress rate Qj of reaction j is written: whereQ Q isthe therate rateof ofprogress progressof ofreaction reactionjj.. The Theprogress progressrate rate where jj is where Qj is the rate of progress of reaction j. The progress rate Qj of reaction j is written: 0 00 Q ofreaction reactionjj is iswritten: written: vkj vkj Q N N jj of Y Y Y k Y k Q j = Kf j K (5) rj 00 00 00 N N v v N N W W kj kj Y Y k vkj k vkj Y Y Y Y k =1 k =1 Y Y k k k k Q =K K K (5) Q K (5) rj jj = ff jj rj W W W Wkk k =1 k =1 rates where Kf j and Krj are thekk forward and reverse of reaction k =1 k =1 j. whereK K andK K arethe theforward forwardand andreverse reverserates ratesof ofreaction reaction where rj are ff jj and rj Kfj and Krj are the forward and reverse rates of reaction j. j jwhere .. These rate constants are the central problem of combustion
modeling. They are usually modeled using the empirical ArrheTheseArrhenius rate constants are the central problem of combustion the These empirical law: nius law: rate constants are the central problem of combustion modeling. They Theyare areusually usuallymodeled modeledusing usingthe theempirical empiricalArrheArrhemodeling. niuslaw: law: Ej Taj nius j Kf j = Af j T j exp = A T exp fj T (6) RT E T E T aj jj aj jj exp jj exp K = A T = A T (6) Kffjj = AffjjT exp = AffjjT exp (6) RT T RT T Expressing the individual rates means Qj providing for each Expressing the individual progress rates progress Qj for each reaction data reaction for the premeans providing data for the pre-exponential constant Af j , the exponential constant Afj, individual the temperature progress exponent j and the activation temperature Taj (or the Expressing the rates Q for each reaction Expressing individualand progress rates Qtemperature jj for each reaction temperaturethe exponent the activation Taj (or j activation energy EJ = RT aj ). In for numerical approach for reacting flows, the chemical scheme (or means providing data for the pre-exponential constant A j,, the the means providing data the pre-exponential constant A fj f the activation energy EJ = RTaj ). In numerical approach temperature exponent and the activation temperature T (or temperature the activation (or ajone mechanism) is one exponent database of data, which must be available for the temperature In our T calculations jj and aj for reacting ows, the chemical scheme (orcomputation. mechanism) is the activation activation energy E = RT ).areIn In numerical approach the energy E = ). numerical approach J aj J aj we used and of tested different reaction mechanisms; some very simple and contains only the database data which must beRT available for the computation. forreacting reacting ows, the the chemical scheme (or mechanism) mechanism) is one one for ows, chemical (or is In our calculations we used scheme and tested different reaction databaseof ofdata data which must beavailable available forthe thecomputation. computation. database which must be for mechaisms; some are very simple and contains only the reacIn our calculations we used and tested different reaction In our and other tested different reaction tions of calculations methane and we fewused species, more detailed mechThese rate constants are the central problem of combustion modeling. They are usually modeled using

is the th is kiln, Whe When bring kiln, kiln, the S bring bring beca the S S the micr beca becau micr micro EB centr EB EB tion centr centr the m tion r tion depe them m the entha depe depen equa enth entha calcu equa equat tions calcu calcu actio tions tions to th actio actio insta to the th to insta instan

in wh diffu in wh w in The diffu diffu neou The s The neou neou

wher den wher wher den den

wher

reactions of methane and few species, other more detailed mechanisms contain also other alkanes reactions and more species. Since it is well known that EBU degrades its global accuracy as the number of species increases, we decided to restrict our calculations to a 5-step mechanism with 14 species. C.3 Combustion Model We have now to link the reactions and their proprieties with the turbulent flow produced by the flame and vice versa. To do so a combustion model is needed. The purpose of combustion models is to calculate the reaction state space and the quantities they influence: density, viscosity, and temperature. A large chemical reaction set can span a wide range of time and length scales. Resolving all the length and time scales affecting the grid-mean properties in a reacting flow system demands computational resources beyond those currently available. We therefore need combustion models to account for the processes that occur at length and time scales below what we can resolve on a numerical simulation grid. The choice of combustion model is decided by knowing the Damkohler number. When the Damkohler number is very large, as in the case of the kiln, the turbulent mixing that brings reactants together at the molecular scale controls the reaction rate. In this limit, the Standard Eddy Break Up (EBU) model is fairly accurate because it assumes that the reaction occurs instantaneously upon micromixing. EBU combustion model track individual mean species concentrations on the grid through transport equations. The reac- tion rates used in these equations are calculated as functions of the mean species concentrations, turbulence characteristics and, depending on the specific model used, temperature. A mean enthalpy equation is solved in addition to the species transport equations. The mean temperature, density and viscosity are then calculated knowing the mean enthalpy and species concentra- tions. In the standard EBU, individual species in the global reaction are assumed to be transported at different

equations. The mean density are then enthalpy equation is temperature, solved in addition toand theviscosity species transport (5) (5) equations. The mean temperature, density and viscosity are then the mean and species concentraequations.knowing The mean temperature, density and viscosity are then (5) calculated calculated knowing the mean enthalpy enthalpy and species concentrations. In the standard EBU, individual species in the global recalculated knowing the mean enthalpy and species concentrations. In the standard EBU, individual species in the global retion assumed to be transported at different rates tions. are In the standard EBU, individual species in the according global rection action action are assumed to be transported different rates according rates according to their own governing equations. These areat derived from the instantaneous governing ction to their own governing equations. These are from action are assumed to be transported at different rates according to their governing equations. These are derived derived from the the equations for own species i. instantaneous equations for species .. to their own governing equations. These are i derived from the instantaneous governing equations for species i tion instantaneous governing equations for species i. stion @ rhestion @ ( Y i ) + r ( U Y i F i ) = S i rrhe(7) Fi (7) i ) + r (U Yi i i ) = Si i rrhe@@ t ( Yi (7) @ t ( Y i ) + r ( U Y i F i ) = S i @t in which Fi = Di rYi is the diffusion ux, Di is the molecular i = Di irYi i is the diffusion ux, Di i is the molecular in which Fi (6) diffusivity of species i , and S is the reaction rate species i. in which F = D r Y is the diffusion ux, D is S the molecular (6) diffusivity i diffusivity of species i, i Dii iis i in which Fi is the i diffusion flux, the molecular and iof is the reaction rate. of species , and S is the reaction rate of species i. i (6) The i standard model uses time-average of the instantadiffusivity ofEBU species i, and S isinstantaneous the reaction rate of species i. ithe The standard EBU model uses the time-average of the governing of equations: The standard EBU model uses the time-average the instantaneous governing equations: The standard EBU model uses the time-average of the instantation neous governing equations: ction neous governing equations: the ction @ , the @ ( Y i ) + r ( U Y i F i ) = S i (8) ,(or the @ ( Y ) + r ( U Y F ) = S (8) @t i i i i i i i i j (or j ( Y ) + r ( U Y F ) = S (8) @ t i i i i ach (or j @ t oach oach one where F i now includes the effect of turbulent diffusion and is one where F includes includes thediffusion effect of turbulent diffusion and is i now i where Fi now theincludes effect of turbulent and is defined as: sn. one dened where F now the effect of turbulent diffusion and is as: i on. dened as: t on. tion dened as: F i = Di + t (9) t rY i ction F = D + r Y (9) t Sc i i i ction eaci i i i F = D + r Y (9) Sc i i i reaci eaci Sc reac- where t is the turbulent diffusivity echi Sci is the turbulent ech- where t is the turbulent diffusivity and and Scii is the turbulent mechcies. Schmidt t where is the turbulent diffusivity and Sci is the turbulent number. The reaction rate is modeled through an excies. Schmidt tnumber. The reaction rate is modeled through anrate excies. pression ind) where t is the turbulent diffusivity and Sci is therate turbulent Schmidt number. The reaction is Schmidt number. The reaction is modeled through an exthat takes the turbulent micromixing process into kind) pression that takes the turbulent micromixing process into acackind) ses, count. modeled through an expression that takes thedimensional turbulent micromixing micromixing process into account. This is done pression that takes the turbulent process into acThis is done through arguments. Thus, for a ases, count. This is done through dimensional arguments. Thus, for a ases, reaction own count. This is done through dimensional arguments. Thus, for a of the form: through dimensional arguments. Thus, for a reaction of the form: own reaction of the form: hown reaction of the form: vF F + vO O ! vP 1 P1 + vP 2 P2 + ....... + vP j Pj (10) v F F + v O O ! v P 1 P1 + v P 2 P + ....... + v P (10) 2 Pj j j j 2 P vF F + vO O ! vP 1 P1 + vP2 P + ....... + v P (10) 2 2 Pj j rate of fuel depletion is assumed to be: with the the rate of fuel depletion is assumed to be: with the rate of fuel depletion is assumed to be: owith do the rate of fuel depletion is assumed to be: o do To do tion YO stion 1 Aebu min stion ties RF = Y , 1 F Y O, O 1 Y O, M s tities ities R = A Y F R ebu min F ,, O ebu F tities RF A min Y , arge F = ebu F MF R sO O

vF F + vO O ! vP 1 P1 + vP 2 P2 + ....... + vP j Pj the rate of fuel depletion is assumed to be:

(10)

with o do stion tities arge ngth cting ands We esses olve

1 YO RF = Aebu min Y F , , R sO Y P1 Y Pj Bebu + ... + (11) sP 1 sP j limiting reactant divided by the time scale of the large eddies. |v P i | M P i vO MO k The reaction source term for each species can then be obtained where sO = , s = and = = P i R mix v M v M ". F each F species can then be F obtained F The reaction source term for from: from:(12) essentially states that the integrated micromixing rate Eqn. nr X is proportional to the mean (macroscopic) the S =M v R concentration of(12) MF
i i ij j j =1

a disc

Ad [4] an of the si

C.4 Radiation Model


C.4 Radiation Model

A Participating Media Radiation Model (DOF) is needed A Participating Media Radiation Model (DOF) is needed to simulate In the some heat transfer due to to simlate the heat transfer due to radiation. cases, thermal can occur as a surface phenomenon. The radiation. In radiation some cases, thermal radiationonly can occur only as a surface phenomenon. The media media separating surfaces (such as dry air)However, is transparent separating the surfaces (suchthe as dry air) is transparent to thermal radiation. other as in our to thermal radiation. However, other as in our kiln we require kiln we require consideration of participating media. This term refers to media lying between the consideration of participating media. This term refers to media radiating surfaces that can absorb, emit, and/or scatter thermal radiation. lying between the radiating surfaces that can absorb, emit, In the m to and/or scatter thermal radiation. As radiation travels through a medium, its radiant intensity I in the direction is absorbed and As radiation travels through a medium, its radiant intensity I (RTE) which, written in terms of absorbed radiant intensity for a specific wavelength ,by is given in the direction is and incremented theby: intervening material. This process is governed by the radiative transfer equation (RTE) which, written in terms of radiant intensity for a specic wavelength , is given by: dI = ds I + ka Ib ks + 4 Z I () d() (13)
incremented by the intervening material. This process is governed by the radiative transfer equation

The e given

k pa

Z 4

where

equation (RTE) which, written in terms Z of radiant intensity for a dIwavelength , is given by: ks specic = I + ka Ib + I () d() ds 4 4 Z (13) k ps dI k ps s + k I + I () d() pa pb = I + ka pb Ib + 4 pa ds 4 4 Z 4 (13) k C 2ps C1 +I kpa = Ipb + 2 I () d() 1 (14) Ib = (14) 4 b 5 C / T 4 2 5 e C / T 1 2 e 1 16 2 2C where = 0.595522 10 /s = 0.01439 mK .. where C C1 = 10 16 W Wm m21 /s and and C C2 1 = 0.595522 2 = 0.01439 mK Ib (14) 5 e C2 / T 1 When coefcients of When the the absorption absorption and and scattering scattering coefcients of the the medium medium are of medium is called gray. In are independent independent of wavelength, wavelength, the medium is= called gray. In where C1 = 0.595522 10 16 Wthe m2 /s and C2 0.01439 mK . When the absorption and scattering coefficients of the medium are independent of wavelength, the that the be integrated over to that case, case, the RTE RTE can can be integrated over wavelength wavelength to proproWhen the absorption and scattering coefcients of the medium duce equation. boundary conmedium a is wavelength-independent called grey. In that case, the RTE can be integrated The over wavelength to produce duce a wavelength-independent equation. The boundary conare independent of wavelength, the medium is called gray. In a dition applied to the RTE for diffusely emitting (with emissivwavelength independent equation. Thebe boundary condition applied to the RTE for diffusely dition applied to the RTE for diffusely emitting (with emissivthat case, the RTE can integrated over wavelength to emitting proity " ) and reecting (with reectivity ) boundaries is, for w ity and (with reectivity w )for boundaries is, for (with" emissivity w)reecting and reflecting (with reflectivity w) boundaries each wavelength: duce a) wavelength-independent equation. The boundary conw w is, each each wavelength: wavelength: dition applied to the RTE for diffusely emitting (with emissivZ Z ity "w ) and reecting (with reectivity w0 ) boundaries is, for w 0 w 0) |n s 0| d I ( Iw + I ( s (15) Iwavelength: (s s) )= =" "w I ( s ) | n s | d (15) each w Iw + 0 ms 0<0) Z( (ms <0) w 0given by the The radiant heat ux in a particular direction r is I ( s ) = " I + I (s0 ) |q n s | d by(15) w ux w in a particular direction The radiant heat q is the The radiant heat flux in a particular direction qr is given by the integration r of the given radiant intensity over 0 ( m s < 0) integration of the radiant intensity over all solid angles and over integration of the radiant intensity over all solid angles and over all solid angles and over spectrum: the wavelength spectrum: the wavelength The radiant heatspectrum: ux in a particular direction qr is given by the the wavelength Z 1Z integration of the radiantZ intensity over all solid angles and over 1Z (r ) = I ( s ) s d d (16) the wavelength q spectrum: qr I ( s ) s d d (16) r (r ) = 0 4 Z 01 Z 4 The radiation solution coupled to the solution qr (r) =is )s uid d ddynamic (16) The radiation solution is coupledIto(s the uid dynamic solution through the divergence of the radiative heat ux. This term ex0 dynamic 4 solution through the divergence of the radiative The radiationthe solution is coupled to the through divergence offluid the radiative heat ux. This term exchanges energy between the uid to and the radiant energy eld. The radiation is coupled the uid dynamic solution heat flux. This term solution exchanges energy between the and fluid and theradiant radiant energy field. Given the changes energy between the uid the energy eld. Given the intensity eld, the divergence of the heat ux is comthrough the divergence ofthe the radiative ux. This exGiven the intensity eld, divergence of the heat uxterm is comintensity field, the divergence of the heat flux is computed as: heat puted as:energy between the uid and the radiant energy eld. changes puted as: Z of the heat 1 the Given the intensityZ eld, divergence ux is comZ 1 Z qr = ka 4 Ib I d d (17) puted as: r r qr = 0 ka 4 Ib I d d (17) 4

The e k pa given

k pa where tion f pende where the the fo tionfo f pende Ii i the I fo

Ii The The r r dent) dent) term term a The ra dent) term a

The The n n a a give give

The n a give

To To conve conve an To he an he in our conve in ou simul an he simul outsid in ou outsid obvio simul obvio outsid obvio

es. ined es.

The discrete ordinates method solves eld equations for radia- 6 The discrete ordinates method solves with for direction radiation intensity associated with a by fixed tion intensity associated a xed s, represented a discrete solid angle. field equations

changes energy between the uid and the radiant energy eld. Given the intensity eld, the divergence of the heat ux is computed as: Z 1 Z r qr = ka 4 Ib I d d (17)
0 4

in ou simul outsid obvio


6

ined (12)

(12)

a discrete solid angle. A detailed description of the method can be found in Modest A detailed description of the method can be found in Modest [4] and Siegel and Howell [8]. As a result [4] and Siegeldescription and Howell [8]. As a result of be Eqn. (20), the form A detailed of the method can found in Modest of Eqn. (20), the form of these ordinate equations (for each wavelength band) is: of ordinate equations (for each wavelength is: form [4]these and Siegel and Howell [8]. As a result of Eqn. band) (20), the

direction s, represented by a discrete solid angle.

eded ases, eded The ases, arent The quire arent media quire emit, edia emit,

ity I venity I nsfer venfor a nsfer for a

(13) (13)

of these ordinate equations (for each wavelength band) is: n X ks s i rI i = Ii + ka Ib + n w j Ij X 4 ks j =1 w I s i rI i = Ii + ka Ib + j j 4 X n k ps j =1 +k pa Ipb + n w j Ij X 4 k ps j =1 w I +k pa Ipb + j j 4 j =1 (18) (18) In the above equation, represents a wavelength band from In the above equation, represents a wavelength band from m to n. The black body emission in this to above black body emission in a this band is given by: m n . The In the equation, represents wavelength band from band is given by: band is given by: Z nemission in Z this m to n . The black body m Ib = Z n Ib d (19) Z m Ib d 0 0 Ib = Ib d Ib d (19) 0 0 The equivalent particle absorption and scattering coefcients are given by [100] particle as: The equivalent absorption and scattering coefcients are The equivalent particle absorption coefficients are given as: given by [100] as: and scattering 2 X X di d2 i k pa = X Qa,i Ni 2 , k ps = X Qs,i Ni 2 (20) 4 d 4 d k pa = i Qa,i Ni i , k ps = i Qs,i Ni i (20) 4 4 i i where i indicates the parcel index number. The transport equation for each ordinate direction is discretized and solved inde-

venvennsfer nsfer nsfer for a for for a a

0 0

0 0

(13) (13) (13)

The equivalent particle particle absorption scattering coefcients are The absorption and and scattering scattering coefcients coefcients are are The equivalent equivalent particle absorption and scattering coefcients are given by [100] [100] as: given given by by [100] as: as: 2 2 X X 2 2 X d d 2 2 X X i i d d 2 2 X X d i i di di kpa = Qa,i , k = Q N (20) pa a,i N i ps s,i i i i k = Q N , k = Q N (20) Q N (20) i ps s,i i a,i i ps s,i i kpa = Q N , k = Q N (20) 4 4 pa a,i i 4 ps s,i i 4 4 i i 4 4 i i i i i i

(14) (14) (14)

mK.. mK mK . dium dium dium y. In y. In y. In proproproconconconssivssivssivs, for , for s, for

where i i indicates indicates the the parcel The transport equawhere parcel index index number. number. The The transport transport equaequawhere i indicates the parcel index number. The transport equawhere i indicates theordinate parcel indexdirection number. The is transport equation for each solved ordinate direction is tion for each discretized and indetion for each ordinate direction is discretized and solved indediscretized and solved indetion for each ordinate direction is discretized and solved independently. The angularly boundary conditions take discretized and solved independently. The discretized angularly discretized boundary conditions take the form: pendently. boundary conditions take conditions take pendently. The The angularly angularly discretized discretized boundary boundary conditions take the form: the the form: form: X w X 0 X 0 w I = " I + I n s (21) 0 i ,w w b j w j 0 I = " I + I n s (21) w n s (21) j j i ,w w b j Ii ,w = "w Ib + Ij n sj (21) w w j 0 w (ns00<0) ( (n n s s0< <0) 0) The reection reection term term represents represents summation over incoming (inciThe summation over incoming (inciover incoming (inciThe reflection term represents summation over incoming (incident) ordinate directions. The discretized The reection term represents summation over incoming (incident) ordinate directions. The discretized form of this source dent) form of of this this source source ordinate directions. The discretized form dent) ordinate directions. The discretized form of this source form of this source term at any cell is: term at any cell is: term term at at any any cell cell is: is: 0 1 0 1 1 0X 1 n X n X X @ A X X r qr = ka @ n 4wj Ij 4 I (22) b A A 4 I (22) r qr = k 4 w I 4 I (22) j j b a j j b r a j j b j =1
=1 j j=1 =1
(ns <0)

(15) (15) (15)

y the y the y the over over over

(16) (16) (16)

The net absorption rate of radiative energy by all particles within The by all all particles particles within within net absorption rate of radiative energy by The a given cell is given by: The net absorption rate of radiative energy by all particles within a given cell is given by: a a given cell is given by: 0 1 0 1 1 n 0X X n n X A X k a @X r qpr = X 4wj Ij 4 Ipb A (23) A @ 4 I (23) r qpr = k 4 w I 4 I (23) j j j pb a j pb pr a j j pb j =1
=1 j j=1 =1

ution ution ution m exm exm exeld. eld. eld. comcomcom-

To resemble the real setup of the kiln we impose a forced To the real setup of of the the kiln kiln we we impose impose a a forced forced To resemble resemble theexternal real setup convection on the layer of the solid wall by imposing convection solid wall wall by by imposing imposing on the external layer of the solid convection an heat transfer coefcient evaluated from measurements done an measurements done done heat transfer coefcient evaluated from measurements an in our plant. With this long list of models we are essentially in our With this long list list of of models models we we are are essentially essentially in our plant. plant. this long simulating all With the thermal exchanges that happen inside and simulating all the thermal exchanges that that happen happen inside inside and and simulating theand thermal exchanges outside the all kiln through the solid wall, the ame and outside the kiln and through the solid wall, wall, the the ame ame and and

REFERENCES
[1] Star-CCM+Vers6.06.011UserGuide,2011. [2] Akwasi A Boateng. Rotary Kilns: Transport Phenomena and Transport Processes. ButterworthHeinemann, 1st edition, 2006. [3] Taylor H.F.W. Cement Chemistry. Academic Press, 1990. [4] M.F.Modest.RadiativeHeatTransfer.McGraw-Hill,NY,1993. [5] M. Peric. Flow simulation using control volumes of arbitrary polyhedral shape. ERCOFTAC Bulletin, No. 62, September 2004. [6] Veynante D. Poinsot, T. Theoretical and Numerical Combustion. R.T. Ed- wards, Inc., 2nd edition, 2005. [7] W. Rodi. Experience with two-layer models combining the k-e model with a one-equation model near the wall. 29th Aerospace Sciences Meeting, January 7-10, Reno, NV, AIAA 91-0216, 1991. [8] R. Siegel and J.R. Howell. Thermal Radiation Heat Transfer. Hemisphere Publishing Co., Washington, 3rd edition, 1992. [9] Barham D. Tran H.N. An overview of ring formation in lime kilns. Tappi Journal, (74 [1]), 1991.

You might also like