Analysis of the Injection of Urea-water-solution for automotive SCR DeNOx-Systems: Modeling of Two-phase Flow and Spray/Wall-Interaction

Transcription

1 Analysis of the Injection of Urea-water-solution for automotive SCR DeNOx-Systems: Moeling of Two-phase Flow an Spray/Wall-Interaction Felix Birkhol, Ulrich Meingast, Peter Wassermann Robert Bosch GmbH, Germany Olaf Deutschmann Institute for Chemical Technology an Polymer Chemistry, University of Karlsruhe, Germany Copyright c 2006 SAE International ABSTRACT The selective catalytic reuction (SCR) base on ureawater-solution is an effective technique to reuce nitrogen oxies (NO x ) emitte from iesel engines. A 3D numerical computer moel of the injection of urea-water-solution an their interaction with the exhaust gas flow an exhaust tubing is evelope to evaluate ifferent configurations uring the evelopment process of such a DeNOxsystem. The moel accounts for all relevant processes appearing from the injection point to the entrance of the SCR-catalyst: momentum interaction between gas phase an roplets evaporation an thermolysis of roplets hyrolysis of isocyanic aci in gas phase heat transfer between wall an roplets spray/wall-interaction two-component wall film incluing interaction with gas phase an exhaust tube The single moeling steps are verifie with visualizations, patternator measurements, phase-oppler-anemometer results an temperature measurements. CFD simulations of a SCR DeNOx-system are compare to experimental ata to etermine the ecomposition parameters for ureawater-solution roplets. Numerical results for an injection incluing all processes aresse above are iscusse. INTRODUCTION In automotive applications, the urea-water-solution base selective catalytic reuction (SCR) is a promising metho to control NO x emissions. Urea-water-solution (UWS, containing 32.5 wt% urea; bran name: ABlue) is spraye into the hot exhaust stream an subsequently the reucing agent (ammonia, NH 3 ) is generate by evaporation of water, thermolysis of urea an hyrolysis of isocyanic aci (HNCO) [1]. The resulting spatial istribution of the reucing agent upstream to the catalyst is a crucial factor for the conversion of NO x. In actual exhaust configurations impingement of roplets on the catalyst an the walls cannot be avoie ue to the slow evaporation an thermolysis an ue to the inertia of the roplets [2]. Especially in passenger car applications [3], where osing systems without air-assiste atomizer have to be applie in combination with small tube iameters, a noticable amount of spray impacts on the exhaust tube surfaces. Spray impingement causes local cooling of the wall. Deposition of roplets an wall film formation can occur if the surface temperature ecreases below a critical temperature. Evaporation from the wall film leas to further cooling an an increasing risk of formation of melamine complexes [4]. Analyzing the literature, several stuies on the evaporation an thermolysis of UWS from spraye roplets can be foun, e.g. [2], [5], [6], an [7]. To the best of our knowlege there are no stuies publishe on the effects of spray impact on surfaces an wall film formation at the injection of UWS. Therefore the interaction of the spray with both the hot gas stream an walls is consiere in this work. To preict the generation an istribution of the reucing agent a etaile three-imensional numerical moel of the behavior of the urea-water-solution spray in the exhaust system is evelope an implemente in the commercial CFD coe Fire v8.3 from AVL [8].

2 Evaporation of the two-component roplets in the gas phase is escribe using the Rapi Mixing (RM) moel, which consiers the influence of urea concentration an variable flui properties. The vapor pressure of urea is erive from experimental ata [9] to calculate the thermal ecomposition of the urea particles. Thus, the physical conitions of the roplets are etermine as an important bounary conition in case of impingement on the walls an the catalyst. The use spray/wall-interaction moel of Kuhnke [10] accounts for ry an wet as well as for col an hot walls by using imensionless numbers which are influence by the thermo-physical properties of the roplets. Heat transfer between spray an wall is escribe accoring to Wruck [11]. The film on the wall is moele as a two-component flui of urea an water couple by momentum, species, an energy balances to the gas phase an the walls. A moel of the evaporation an thermal ecomposition of UWS from the wall film is evelope in this stuy. The calculate spray behavior is verifie with results from visualizations gaine in a flow channel at varying flow conitions. Patternator measurements, phase-oppleranemometer (PDA) results an temperature measurements are presente an compare with numerical results. EVAPORATION AND THERMOLYSIS OF UREA- WATER-SOLUTION DROPLETS When UWS is spraye into the hot exhaust gas, the roplets are heate up an water evaporates first [12] (NH 2 ) 2 CO(aq) (NH 2 ) 2 CO(l) H 2 0(g), (1) followe by the thermolysis of urea into ammonia an isocyanic aci (NH 2 ) 2 CO(l) NH 3 (g) + HNCO(g). (2) Solve urea at the roplet surface causes the vapor pressure of water to ecrease. This results in a lower water evaporation rate. The effect is suitably escribe using a Rapi Mixing (RM) moel [2]. Within the RM moel infinite fast transport is assume in the liqui phase, resulting in spatialy uniform but time-epenent temperature, concentration an flui properties of the roplet [13, 14]. The variation of the urea concentration in the roplet can be evaluate by Y u, = t ṁvap Y u,, (3) m where mass flow from the liqui to the gaseous phase is efine to be negative. For the gas phase the quasi-steay moel [13] is use. This approach is suitable to escribe the evaporation process in the entire range of present conitions, incluing the free convection case [15], using the 1/3-rule [16] for the reference values for flui properties. Integration of the transport equations for mass an enthalpy outsie the roplet yiels analytical expressions for the iffusive transport fluxes. The ifferential equations for roplet mass an temperature can be erive from mass an energy balance [13, 17] an writes for water evaporation T t m t = πd ρ g,ref Γ g,ref Sh ln (1 + B M ) (4) ( ) = ṁvap cp,vap,ref (T g T ) h vap. m c p, B T The Spaling heat an mass transfer numbers B M an B T are calculate as an B T = (1 + B M ) x 1, (5) B M = Y vap,s Y vap,g 1 Y vap,s (6) x = c p,vap,ref c p,g,ref Sh Nu 1 Le. (7) If saturation temperature is reache uring the evaporation of water, it is assume that the roplet remains at this temperature. Thus the evaporating mass can be etermine from with m t = πd λ g,ref c p,vap,ref Nu ln (1 + B T ) (8) B T = c p,vap,ref (T g T s ) h vap. (9) Equations (4) to (7) are also applie for urea ecomposition after the evaporation of water is finishe. It is assume that urea evaporates from its melt an ecomposes instantly into ammonia an isocyanic aci on the particle surface. Therefore the properties of an equimolar ammonia/isocyanic aci mixture are use in the bounary layer. The vapor pressure of urea, unknown so far, is etermine from experimental results as shown below. The approach accounts for non-unit Lewis number an the effect of Stefan flow on heat an mass transfer. Convective transport is consiere by a moifie Sherwoo an Nusselt number using the well-establishe Frössling correlations [18]. The moel for evaporation an thermal ecomposition is implemente into the CFD coe Fire v8.3. In this coe, the UWS roplets are treate with Lagrangian particle tracking, which solves the equation of motion for parcels of roplets with ientical properties using the Discrete Droplet Metho of Dukowicz [19]. Turbulence ispersion is efine by the Ey-Lifetime moel [20]. Between roplets an gas phase two-way coupling is consiere for momentum, mass an heat. For turbulence kinetic energy an issipation one-way coupling is applie. Hyrolysis of isocyanic aci HNCO(g) + H 2 0(g) NH 3 (g) + CO 2 (g) (10)

3 is consiere as a homogeneous gas phase reaction using the interface of Fire to the CHEMKIN chemistry solver [21]. The rate of hyrolysis is given by Yim et al. [22] by r hy = c HNCO e R T. (11) SPRAY/WALL-INTERACTION SPRAY/WALL-INTERACTION MODEL The physical mechanisms occurring uring the interaction of spray with a wall are very complex, because the behavior of the impinging roplet is influence by a variety of parameters such as roplet properties like velocity, iameter, flui properties an surface properties like wall temperature, surface roughness or wall film height [23]. Thus semi-empirical moels base on imensionless variables are evelope for numerical simulations (e.g. [24]). The use moel of Kuhnke [25] consiers all relevant impingement phenomena by a classification into the four regimes eposition, splash, reboun an thermal breakup base on the two parameters: K = (ρ D)3/4 u 5/4 σ 1/2 µ 1/4 = Ca 5/4 La 3/4 (12) T = T w T sat (13) As sketche in Figure 1 the critical transition temperature is in the range of Tcrit 1.1 for a variety of fluis [25]. For UWS the saturation temperature increases with increasing urea concentration [26], leaing to a higher tenency to wall film formation compare to water. In our experiments we ientify a high critical transition temperature Tcrit for UWS, which will be iscusse below. The two critical K numbers between reboun an seconary atomisation at high temperatures an between eposition an splash for low temperatures range within an , respectively. The regimes eposition an splash lea to the formation of a wall film. For more etaile information we refer to Kuhnke [25]. SPRAY IMPINGEMENT HEAT TRANSFER The cooling of the wall ue to spray impingement plays an important role, because only if the imensionless wall temperature T ecreases below the critical value eposition of roplets will occur an a wall film will evelop. The implemente heat transfer moel is base on the approach of Wruck [11]. When a roplet impinges on a hot wall, a short perio of irect contact between flui an soli exists until a vapor cushion is forme. For small roplets the heat transferre uring this irect contact is etermining the total heat transfer, because of the insulation effect of the vapor cushion [11]. The heat transfer is calculate assuming the contact by two semi-infinite boies, roplet an wall, which are in contact for a certain time with an effective contact area. It is base on Figure 1: Regime map for spray/wall-interaction accoring to Kuhnke [25]. one-imensional transient heat conuction in both contact partners. The heat transferre by the roplets in a parcel accoring to Meingast [27] is given as Q w = A cont 2 t c π b w b b w + b (T w T ) (14) with the heat penetration coefficients of the contact partners b i = (λρc p ) i. (15) The contact area A cont = π 4 D2,eff is calculate using the time average iameter D,eff. The correlation for the maximum spreaing iameter of the roplet uring impact foun by Akao et al. [28] D,max D = 0.61 W e 0.38 (16) is use to etermine the average iameter. The contact time is calculate using the K number epenence: K < 40 K > 40 t cont = π 4 t cont = π 2 (ρ D 3 ) σ [ ρ D 5 σ u 2 ] 0.25 (17) (18) The transferre heat is limite by the heat for complete evaporation of the roplet an the roplet temperature is ajuste accoring to the change in its internal energy. The transferre heat is consiere as a source term in the wall energy equation for the corresponing wall cell. WALL FILM MODELING Deposition of roplets leas to a wall film which is moele with a 2D finite volume metho in the Fire wall film moule [8]. Gas an wall film flow are treate as separate

4 single phases, couple by semi-empirical bounary conitions. The film is transporte ue to shear forces, gravity an pressure graients. CONTINUITY EQUATION The continuity equation transforme into conservation of film thickness h f writes h f t + (h f u 1 ) x 1 + (h f u 2 ) x 2 = S m ρ f A cell, (19) where u i an x i are the velocity components an the wallparallel cartesian coorinates, respectively, an S m is the mass source, e.g. eposition of roplets. Assuming a spatially constant wall film thickness in a cell, equation (19) can be solve explicitly, if velocity components an the source term are known. Sub-cycling is applie to fulfill the CFL conition ū f t/ x < 1. Film velocity profiles are solve analytically for laminar an turbulent flow, where the transition from laminar to turbulent film flow is characterize by a film Reynols number Re f = ūf h f ν f an transition is assume at Re f = 2.7. τ w h f σ f (20) Waviness of the film is moele as a surface roughness, leaing to moifie logarithmic wall law for turbulent flows u + = 1 κ ln y+ + C +. (21) The value C + is constant for technically smooth surfaces but varies accoring to Sattelmayer [29] with the roughness Reynols number Re ks = k su τ ν. (22) The equivalent san grain roughness is etermine k s = ψ(τ w ) 2h f (23) with the correction function ψ propose by Rosskamp [30]. Thus the logarithmic wall law is couple with the wall shear equations, which can only be solve iteratively. This approach works well for wall films of water or fuels [8],[31]. Elsässer [32] ientifies the relevant flui properties for the evelopment of a wall film as the kinematic viscosity ν an the surface tension σ. Our own measurements of the ynamic surface tension of UWS show only a slight increase compare to water surface tension (0.075 N/m vs N/m at 30 C), so its influence can be neglecte in a first approximation, at least for low urea concentrations. The viscosity increases with increasing urea concentration [26] an leas to an increase of the film thickness an shear stress an a tenency to a laminar film profile [32]. For high urea concentrations the above approach may only be a rough estimation of the film ynamics. UREA TRANSPORT EQUATION To account for the urea-water-solution a species transport equation for the urea fraction in the wall film is introuce Y u t + (Y uu 1 ) + (Y uu 2 ) = (24) x 1 x 2 1 (Y u, ṁ ep + ṁ th ), ρ f V cell where the right-han sie accounts for sources from eposite roplets an urea thermolysis. Diffusive effects are neglecte in this stuy, because they contribute typically less than 0.1 % to the mass flux between ajacent cells. Equation (24) is solve in the same way as the film continuity equation (19) with an explicit Euler scheme with up-wining of flux terms. ENERGY EQUATION AND MASS TRANSFER The enthalpy balance equation of the wall film writes h t + (h u 1) + (h u 2) = (25) x 1 x 2 1 ( Q w f + ρ f V Q g f + ṁ vap h vap + ṁ th h th + Q ) imp cell an is solve using a semi-implicit proceure for improve numerical stability. The terms on the right han sie account for heat transfer from the wall an the gas phase, cooling ue to evaporation an thermolysis, an sources from impinge roplets. Evaporation an thermolysis from wall film is base on Fick s law of iffusion [ ] ṁ ρi Γ Y = 1 Y i y, (26) where Y i is the interfacial vapor concentration an Γ the iffusion coefficient. For laminar or rather low Reynols number flows the iscretise concentration graient is approximate with a secon orer interpolation. For moerate or high Reynols number flows equation (26) unerestimates the evaporation rates. For such conitions the evaporation moel introuce by Sill [33], which assumes an analogy between momentum an mass transfer, is more accurate. The evaporation rate writes ṁ = ρ i u g St m Y g Y i 1 Y i, (27) where u g is the gas velocity parallel to the wall. St m is the imensionless Stanton number for mass transfer an correlates with the interfacial shear force. To account for Stefan flow a correction accoring to Ackermann [34] is applie. During the calculation the maximum of the two compute evaporation rates is taken. The saturation temperature of UWS can be reache uring the evaporation of water ue to the ecrease of vapor pressure resulting from increasing urea concentration.

5 Then the film enthalpy equation (25) is transforme to recalculate the evaporating mass flux ṁ vap to get a wall film temperature just below the saturation point. Wall film energy equation an the evaporation routines are couple in this stuy through a time step aaptation to avoi numerical instabilities an to solve the steep graients of heating an evaporation of the wall film. WALL MODELLING Knowlege about the thermal behavior of the wall is important for an accurate calculation of the spray/wallinteraction an the wall film formation. Therefore the 2D transient energy equation for the wall enthalpy H w = c p,w ρ w h w A cell T w (28) is introuce an solve semi-implicitly for every wall cell: with H w t a w 2 H w x 2 1 a w 2 H w x 2 2 = (29) + Q con + Q amb w + Q f w + Q g w Q w t Q i w = α i w A cell (T i T w ). (30) The inex i enotes amb, f an g. The temperature rop in the wall is neglecte, because the wall heat resistance is much lower than the resistance of convection (λ w /h w << α amb w an α f w ). The maximal temperature ifference between inner an outer surface of the wall is typically less than 2 C in exhaust systems. Either Q f w or Q g w are use in equation (29), epening of whether there exists a wall film in the ajacent flui cell. The heat transfer coefficients between film an wall, α f w, are calculate assuming conuction with linear temperature profiles in film an wall α f w = 2 λ f λ w h f hw λ f h f + λw h w. (31) If the wall temperature is above the saturation temperature of UWS, vapor bubbles are generate at the liqui-soli interface influencing the surface heat flux. The heat flux epening on the excess temperature T e = T w T sat is calculate following the Nukiyama boiling curve accoring to Incropera an De Witt [35]. In the free convection boiling region (0 C < T e 5 C) the heat flux is linear interpolate from the heat flux etermine with equation (31) an the nucleate boiling heat flux [ ] 1/2 ( ) 3 q w f g(ρl ρ vap ) cp,l T e = µ l h ev (32) σ l h ev P r l with T e = 5 C. In the nucleate boiling region (5 C < T e 30 C) equation (32) is applie. The maximum heat flux is reache at T e = 30 C an is given by [ ] 1/4 q w f,max σl g(ρ l ρ vap ) = h ev ρ vap ρ 2. (33) vap The heat flux in the transition region (30 C < T e T L ) results from linear interpolation between equation (33) an the minimum heat flux at the Leienfrost temperature T L following q w f,min = 0.09 h ev ρ vap [ σl g(ρ l ρ vap ) (ρ l + ρ vap ) 2 ] 1/4. (34) The heat transfer coefficients are calculate from the etermine heat fluxes using equation (30) an are applie in the enthalpy equation of wall an film. The α g w value is provie by the gas phase solver an the ambient heat transfer coefficient α amb w has to be efine by the user, epening on the ambient conition. In this stuy we use a Nu number correlation given in reference [36] for a flat plate. RESULTS In this section, a valiation of the spray behavior of a swirl nozzle uner quiescent an exhaust flow conition with col an hot ambient temperature, respectively, is iscusse. Subsequently the calculate conversion of ureawater-solution to gaseous ammonia is compare to experimental ata. Finally simulations of an injection with spray impingement on a wall are presente an compare to transient wall temperature measurements. SPRAY VALIDATION To evaluate the influence of varying nozzles, injector mounting positions an exhaust tube configurations, wellknown spray parameters as a bounary conition are essential for a CFD simulation. As moeling of primary breakup is still uner investigation, the spray is initialize with a efine roplet size istribution. Since the aeroynamic Weber numbers are below the critical value of W e crit = 12, seconary breakup oes not occur. The swirl atomizer use throughout this investigation prouces a hollow cone spray. From PDA measurements a mean iameter of 28 µm an a Sauter mean iameter (SMD) of 85 µm is etermine. The static flow rate is about 9 kg/h. The injection velocity an spray angles are etermine from backgroun visualizations in a spray box using a N-YAG laser. Image processing is one with the software DaVis from LaVision [37]. Figure 2 shows a comparison of the measure an simulate penetration for the spray an the pre-spray. As expecte, there is a goo agreement, because the start velocity of the roplets is estimate from this penetration. A patternator is use to measure the spatial spray istribution, which irectly influences the flui istribution in the exhaust tube cross section. The istribution is measure at a given istance to the nozzle an agrees well with the numerical results. Both are isplaye in Figure 3. Droplet size/roplet velocity correlations at two istances (30 mm an 50 mm) from the nozzle are extracte from PDA measurements to verify the preicte momentum exchange between roplet an gas phase. Figure 4 epicts

6 tic for big roplets. Small roplets are accelerate ue to a gas flow inuce by momentum transfer from spray to roplet. This effect is overestimate in the simulation, whereas for roplet sizes between 30 an 80 µm the simulate velocities agree well with the measure ata. The comparison of the simulate spray pattern an shaowgraphic visualization in Figure 5 shows that the main characteristics of the spray can be preicte well by the simulation. Figure 2: Penetration of pre-spray an spray. Figure 5: Simulate spray shape (left) compare to visualization gaine in a spray box at quiescent flow conition using shaowgraphy (right), 10 ms after start of injection. Figure 3: Measure an preicte spray mass istribution. the measure characteristics an the simulate correlations. Especially at 30 mm big roplets are still oscillat- After verifying the spray calculations uner col an quiescent conitions, visualizations of the spray pattern are carrie out in a flow channel at varying flow conitions up to Tg = 250 C an ug = 50 m/s. A light sheet illumination technique is applie, with the camera orthogonal to the light sheet. A triggere camera is use to provie time resolve ata. For aequate comparison the simulations are evaluate using only a thin sheet of the whole spray. Exemplary the calculate spray propagation an the visualizations are epicte in Figure 6 for Tg = 150 C an ug = 25 m/s, showing qualitatively goo agreement. EVAPORATION AND THERMOLYSIS OF DROPLETS Figure 4: Droplet size/roplet velocity correlation in a istance from nozzle of 30 mm an 50 mm. ing an eforme from the lamella breakup. Therefore they are omitte in the PDA measurements, since the analyze roplets have to be spherical. Non-spherical roplets are not etecte an lea to an uncertain statis- To get a vapor pressure correlation for urea, CFD simulations are compare to an experimental investigation of Kim et al. [9]. They stuie the conversion from injecte UWS to ammonia. The UWS is irectly injecte axial to the center of a tube (Figure 7) at varying flow conitions. The average conversion to ammonia ue to thermolysis of UWS roplets an hyrolysis of isocyanic aci is measure ownstream the injection at ifferent sampling points, which result in varying resience times. Figure 8 epicts the comparison between calculate an measure urea conversion to ammonia showing goo agreement. The vapor pressure is etermine as pu = e /T. (35) This formula is also use to calculate the ecomposition rate of urea from the wall film later on.

7 Figure 6: Simulate spray propagation (above) an visualization using light sheet illumination for T g = 150 C an u g = 25 m/s, 10 ms after start of injection. Figure 7: Sketch of the experimental setup of Kim et al. [9] (not to scale). SPRAY/WALL-INTERACTION AND WALL FILM FORMATION The spray/wall-interaction an wall film formation is investigate in a flow channel whose sie walls consist of glas for optical access. A metal plate with a thickness of 2 mm is installe in the center of the flow channel to get a efine spray impact on the wall an well-known bounary Figure 8: Calculate conversion to NH 3 for ifferent gas velocities u g an gas temperatures T g compare to experimental ata of Kim et al. [9]. conition for the lower surface of the plate (Figure 9). The gas velocity is kept constant at 30 m/s an the gas temperature is varie from 105 C up to 390 C. The transient temperature evolution of the plate at ifferent locations is measure with 0.5 mm thermocouples, which are installe in flutes just below the upper surface. Figure 9: Sketch of the experimental setup for investigation of spray/wall-interaction (not to scale). The time scale for spray inuce cooling of a typical exhaust tube wall is in the orer of secons ue to the high heat reservoir of the soli wall. As this is not a suitable time range for multi-phase 3D-CFD calculations,

8 a spee-up factor f is introuce. For our calculation a moifie wall thickness h f /f is use to reuce the cooling time. Figure 10 epicts the evolution of calculate an meassure wall temperature for ifferent combinations of gas temperatures an nozzle mass fluxes, for which no wall film occur. The simulation time is scale with the speeup factor f = 150 use in the calculations. The simulate wall temperatures agree well with the measure ata, hence the use of a reuce wall thickness seems to be an aequate approach to reuce the simulation time. Figure 11: Measure an simulate evolution of wall temperature with wall film formation for positions # 1 an # 2, ṁ inj = 3.72 kg/h, T g = 340 C. heat transfer, which coul not be taken into account in the simulation. Figure 10: Measure an simulate evolution of wall temperature without wall film at positions # 1 for ifferent gas temperatures an nozzle mass fluxes. Figure 12 shows a sie view of the preicte spray penetration at early times after start of injection. The temperature of the wall is still equal to the gas temperature of T g = 340 C. Therefore the moel preicts small roplets ue to thermal breakup. As the wall cooles own an film If the mass flux of the nozzle is increase at a constant gas temperature, the wall temperature can ecrease below the critical transition temperature an film formation can begin. The transition point is characterize by a suen change of the slope of the wall temperature curve. The transition temperature for UWS was etermine from various cooling curves (e.g. see Figure 11) to be in the range of T w = C, which correspons to a critical transition temperature of Tcrit 1.4. This high temperature cannot solely be explaine with the higher saturation temperature of UWS. Further investigations are necessary to clarify the reason for this behavior. In Figure 11 the calculate wall temperatures are compare to the experimental ata for a nozzle mass flow of 3.72 kg/h. The wall is coole ue to spray impact. Film formation starts at t 35 s, when the wall temperature ecrease below the transition temperature. The cooling of the wall is increase ue to evaporation from the wall film. The simulate shape of the cooling curves agree well with the experimental ata, however the film cooling is overestimate in the simulation. Also there is a small ifference in the calculate an measure aiabatic film temperature at times > 90 s. A possible reason for this eviation is ue to eposition of soli urea or melamine complexes, which alreay appears at the borer area of the film uring the measurements. This leas to a moifie flow situation with aitional roughness effects an therefore enhance Figure 12: Spray pattern at 4 ms, 10 ms an 20 ms after start of injection, T g = 340 C. formation begins, the critical K number is shifte to higher values (see Figure 1), leaing to eposition of most of the roplets. Only roplets with high kinetic energy are splashe. The preicte wall film thickness in Figure 13 reflects the hollow cone character of the spray with the highest film thickness in the main impact area. The preicte wall film shape agrees with our visual observations uring the experiment. Figure 14 epicts the urea concentration in the wall film. In the main impact area of the spray, the urea concentration is ominate by the urea fraction of the impinge roplets. Due to evaporation of water from the roplets this is slightly above the initial concentration of 32.5 %. As the film is transporte ownstream in regions with less spray impingement, the evaporation of water from the film leas to higher urea concentrations. After evaporation of water is complete, thermolysis of urea begins in these areas.

9 The spray propagation an reucing agent istribution in the exhaust stream can be escribe well with the moel. The urea vapor pressure moeling has been valiate by comparing CFD calculations of an evaporating spray with experimental ata from Kim et al. [9]. Figure 13: Preicte wall film thickness, m inj = 0.83 g, T g = 340 C. As spray wall interaction is expecte to occur in real exhaust gas systems, the erive approach inclues application of a sophisticate roplet wall interaction moel. This is extene by a semi-empirical approach to account for heat transfer between liqui roplets an wall. Thus the formation of a wall film can be calculate precisely as well as the seconary breakup of roplets at the wall. Spray impingement on hot surfaces, e.g. mixing elements or the exhaust tube walls, can lea to a better spatial istribution of the reucing agent ue to thermal breakup. This also yiels better conversion to ammonia, because small roplets with an enhance heat an mass transfer are generate. Aitionally heat is transferre from the hot surface to the roplets uring contact an accelerates evaporation. Figure 14: Calculate mass fraction of urea in the wall film, m inj = 0.83 g, T g = 340 C. Figure 15 shows sie views of the ammonia fraction in gas phase for the center plane an a plane laterally move by 14 mm. Due to the moerate temperature of gas an wall, the concentrations are low an therefore each scale with the maximum. In the center position ammonia is Spray impact also results in cooling of surfaces an can lea to wall film formation if the surface temperature ecreases below a critical value. The critical transition temperature for urea-water-solution is etermine within this stuy. The influence of flow conitions, exhaust tube properties an spray paramters on the film formation can be evaluate with the evelope moel. The erive moels implemente in the CFD-coe Fire v8.3 help to preict real processes uring the layout of exhaust tube configurations an injector mounting positions with respect to the spatial istribution of the reucing agent upstream the catalyst. Figure 15: NH 3 fraction in gas phase in mile plane (0 mm) an a plane move laterally by 14 mm, m inj = 0.83 g, T g = 340 C. prouce in the upper part ue to thermolysis of small roplets which ecelerate fast after injection an therefore remain at the top of the flow channel. In the 14 mm-plane, one also can see ammonia generate by urea ecomposition from the boarer area of the wall film on the plate. CONCLUSION An analysis of the injection of urea-water-solution for DeNOx-systems is presente. The evelope 3D numerical moel accounts for all relevant physical effects. The single moeling steps are verifie with experimental ata.

10 NOMENCLATURE LATIN LETTERS a thermal conuctivity, m 2 /s A area, m 2 b heat penetration coefficient, W s 0.5 /m 2 K B M,T Spaling numbers c molar concentration, mol/m 3 c p heat capacity, J/kg K Ca Capillary number D iameter, m f spee up factor h thickness, m h specific enthalpy, J/kg H enthalpy, J k s san grain roughness, m K K number, [23] La Laplace number Le Lewis number m mass, kg ṁ mass flux, kg/s ṁ specific mass flux, Kg/m 2 s Nu Nusselt number p pressure, P a Pr Prantl number Q heat, J Q heat flux, W q specific heat flux, W/m 2 r rate of reaction, mol/m 3 s R unversial gas constant, J/mol K S source, kg/s Sh Sherwoo number St Stanton number t time, s T temperature, K u velocity, m/s u τ friction velocity, m/s V volume, m 3 We Weber number, ρ u2 D σ x,y coorinates, m Y mass fraction GREEK LETTERS α heat transfer coefficient, W/m 2 K κ Karman constant λ heat conuctivity, W/m K µ ynamic viscosity, kg/m s ν kinematic viscosity, m 2 /s ρ ensity, kg/m 3 σ surface tension, N/m τ shear stress, N/m 2 Γ iffusion coefficient, m 2 /s SUBSCRIPTS amb con cont crit,g c ep e eff f hy i inj imp l,vap L m u ref s sat th w REFERENCES characteristic ambient conuction contact critical roplet, gas irect contact eposition excess effective film hyrolysis interface injecte impingement liqui, vapor Leienfrost mass urea reference roplet surface saturation thermolysis wall [1] M. Koebel, M. Elsener, an T Marti. NO x -reuction in iesel exhaust gas with urea an Selective Catalitic Reuction. Comb. Sci. an Techn., 121:85 102, [2] F. Birkhol, U. Meingast, P. Wassermann, an O. Deutschmann. Moeling an simulation of the injection of urea-water-solution for automotive SCR DeNOx-systems. Applie Catalysis B: Environmental, accepte, [3] C. Enerle, H. Breitbach, M. Paule, an B. Keppeler. Selective catalytic reuction with urea - the most effective nitrous oxie aftertreatment for lightuty iesel engines. In 26th International Vienna Motor symposium, April 28-29, Vienna, Austria, [4] H.L. Fang an H.F.M. DaCosta. Urea thermolysis an NO x reuction with an without SCR catalysts. Applie Catalysis B: Environmental, 46:17 34, [5] R. van Helen, R. Verbeek, an F. Willems. Optimization of urea SCR enox Systems for HD Diesel Engines. SAE, , [6] J.C. Wurzenberger an R. Wanker. Multi-Scale SCR Moeling, 1D Kinetic Analysis an 3D System Simulation. SAE, , [7] M. Chen an S. Williams. Moelling an Optimization of SCR-Exhaust Aftertreatment Systems. SAE, , 2005.

11 [8] FIRE 8.3. AVL LIST GmbH, A-8020 Graz, Austria, [9] J.Y. Kim, S.H. Ryu, an J.S. Ha. Numerical preiction on the characteristics of spray-inuce mixing an thermal ecomposition of urea solution in SCR system. In Proc Fall Technical Conference of the ASME Internal Combustion Engine Division, Long Beach, California USA, [10] D. Kuhnke. Spray/Wall-Interaction Moelling by Dimensionless Data Analysis. Shaker Verlag, ISBN [11] N. W. Wruck. Transientes Sieen von Tropfen beim Wanaufprall. PhD thesis, RWTH Aachen, [12] M. Koebel, M. Elsener, an M. Kleemann. Urea- SCR: a promising technique to reuce NO x emissions from automotive iesel engines. Catalysis Toay., 59: , [13] G.M. Faeth. Evaporation an combustion of sprays. Prog. Energy Combust. Sci., 9:1 76, [14] W.A. Sirignano. Fuel roplet vaporization an spray combustion theory. Prog. Energy Combust. Sci., 9: , [15] M. Burger, R. Schmehl, P. Prommersberger, O. Schaefer, R. Koch, an S. Wittig. Droplet evaporation moeling by the istillation curve moel: accounting for kerosine fuel an elevate pressures. Int. J. Heat Mass Transfer, 46: , [16] E.M. Sparrow an J.L. Gregg. The variable fluiproperty problem in free convection. Trans. ASME 80, [17] B. Abramzon an W.A. Sirignano. Droplet vaporization moel for spray combustion calculations. Int. J. Heat Mass Transfer, 32: , [18] N. Frössling. Über ie Verunstung fallener Tropfen. Gerlans Beitrag zur Geophysik, 52: , [19] J.K. Dukowicz. A particle-flui numerical moel for liqui sprays. Journal of Computational Physics, 35: , [20] A.D. Gosman an E. Ioannies. Aspects of computer simulation of liqui-fuele combusters. In AIAA AIAA, [21] R.J. Kee, F.M. Rupley, E. Meeks, an J.A. Miller. Chemkin III: A fortran chemical kinetics package for the analysis of gas-phase chemical an plasma kinetics. Sania National Laboratories, Livermore, CA , [22] D.S. Yim, S.J. Kim, J.H. Baik, I. Nam, Y.S. Mok, J.W. Lee, B.K. Cho, an S.H. Oh. Decomposition of Urea into NH 3 for the SCR Process. In.Eng.Chem.Res., 43(1): , [23] Chr. Muno, M. Sommerfel, an C. Tropea. Dropletwall collisions: Experimental stuies of the eformation an breakup process. Int. J. of Multiphase Flow, 21(2): , [24] C. Bai an A.D. Gosman. Development of methoology for spray impingement simulation. SAE, , [25] D. Kuhnke. Spray/Wall-Interaction Moelling by Dimensionless Data Analysis. PhD thesis, Universität Darmstat, [26] L. Jaeger, J. Nývlt, S. Horáĉek, an J. Gottfrie. Viskositäten von Harnstoffwasserlösungen. Collection Czech. Chem. Commun., 30: , [27] U. Meingast. Spray/Wan-Wechselwirkung bei er ieselmotorischen Direkteinspritzung. PhD thesis, RWTH Aachen, [28] J. Akao, K. Araki, S. Mori, an A. Moriyama. Deformation behaviors of a liqui roplet impinging onto hot metal surface. Trans. Iron an Steel Institue of Japan, 20: , [29] T. Sattelmayer. Zum Einfluss er ausgebileten turbulenten Luft-Flüssigkeitsfilm-Strömung auf en Filmzerfall un ie Tropfenbilung am Austritt von Spalten geringer Höhe. PhD thesis, Universität Karlsruhe (TH), [30] H. Rosskamp. Simulation von rallbehafteten Zweiphasenstrmungen mit schubspannugsgetriebenen Wanfilmen. PhD thesis, Universität Karlsruhe (TH), [31] J. Himmelsbach. Zweiphasenströmung mit schubspannungsgetriebenen, welligen Flüssigkeitsfilmen in turbulenter Heissgasströmung-Messtechnische Erfassung un numerische Beschreibung. PhD thesis, Universität Karlsruhe (TH), [32] A. Elsässer. Kraftstoffaufbereitung in Verbrennungskraftmaschinen: Grunlagen er Strömung schubspannungsgetriebener Wanfilmen. PhD thesis, Universität Karlsruhe (TH), [33] K. H. Sill. Wärme- un Stoffübertragung in turbulenten Strömungsgrenzschichten längs verunstener welliger Flüssigkeitsfilme. PhD thesis, Universität Karlsruhe (TH), [34] G. Ackermann. Wärmeübergang un molekulare Stoffübertragung im gleichen Fel bei grossen Temperatur- un Partialruckifferenzen. VDI- Forschungshefte, (Nr.382):1 16, [35] F. P. Incropera an D. P. De Witt. Funamentals of Heat an Mass Transfer. Wiley an Sons, 5th eition, ISBN [36] VDI-Wärmeatlas. VDI-Gesellschaft, eighth eition, Springer, [37] Davis LaVision GmbH, D Göttingen, Germany,