Gaseous pollutant dispersion measurements on an wind tunnel model of a waste incinerator Daniele Contini\ Andrea Corti* & Alan Robins^

Transcription

1 Gaseous pollutant dispersion measurements on an wind tunnel model of a waste incinerator Daniele Contini\ Andrea Corti* & Alan Robins^ Mar/a contini@fias.de.imifi.it; corti@pineung.unifi.it 2 EnFlo, School ofmechanical & Materials Engineering, a.robins@surrey.ac.uk Abstract A wind tunnel study of plume rise and dispersion at a twin-stack waste incinerator has been carried out in the large stratified wind tunnel facility at the EnFlo Research Centre (University of Surrey). The model scale was 1 to 300 and measurements of mean scalar gas concentration were performed with a flame ionisation detector (FID) based system in a neutral boundary layer simulating a rural environment. Standard buoyancy scaling relationships were used to determine the discharge conditions at model scale. Mean plume heights, lateral and vertical spreads were obtained from the analysis of vertical and crosswind concentration profiles measured at different downstream distances from the sources. Effects of the wind direction with respect to the alignment of the two chimneys were investigated. Vertical concentration profiles showed a slight increase of the average plume height, as a consequence of the plumes interaction, when the two stacks were aligned along the wind direction. Experimental results have been compared with the standard ISCST2 gaussian code, with the ADMS2 integral model and with a modified version of the ISCST2 code. 1 Introduction The problem of efficiently removing from the environment the large amounts of waste from human activities (both civil and industrial) has

2 710 Air Pollution become crucial for the development of modern cities. Depending on the site and on the prevailing industrial activities and level of development, most communities produce a daily amount of waste of the order of 1.5 to 4 Kg/day per person. Even when an integrated waste management program operates, a notable fraction of this waste (typically 10 to 30%, depending on waste composition and recycling capabilities) has necessarily to be disposed for incineration. Very often installations of this sort are located in or near populated areas and it is necessary to perform quite accurate evaluations of the impact of their emissions on local air quality prior to operation. These environmental impact studies are usually carried out by using simple and fast gaussian code [1, 3], like the EPA standard ISCST2 code. These kind of codes rely on empirical parameters and formulae and are not based on a full description of the three dimensional turbulent flow-field in which dispersion take place. They quite often give good results further downstream but prediction near the sources or when the site presents a complicated orography with buildings around the sources itself are beyond their scope. Lagrangian models or computation fluid dynamics codes may give accurate results in a larger variety of conditions but they are usually time consuming, difficult to implement in non neutral conditions and, as a consequence, they are not usually employed for a complete assessment of environmental impact. Full scale experiment are very expensive and difficult to do because the researcher has no control over the atmospheric conditions so it is not possible to change a parameter at a time and this limits the use of full scale experiments for developing and validating models of atmospheric diffusion. What is more environmental impact studies have to be carried out before the construction of the plant, so only full scale experiment on similar conditions may be available during the project phase. Wind tunnel simulations are thus a valuable alternative for both environmental impact analysis and model developing. By using appropriate scaling relationships is possible to study dispersion behaviour over simple and complicated orography and buildings and, depending on wind tunnel capability, atmosphere stability conditions. In this paper we discuss a wind tunnel dispersion study of the emission discharged by a model of a waste incinerator which is to be built in Italy near Florence. All the measurements of flow and dispersion were carried out in the large stratified wind tunnel of the EnFlo research centre at the University of Surrey. This is an open circuit tunnel with a working section 3.5 m wide, 1.5 m tall and 20 m long. Stable and unstable boundary layers may be simulated in the tunnel, those these capabilities were not used here and attention was focused on neutral

3 Air Pollution 711 conditions. Results have been compared with standard gaussian dispersion codes and with the UK integral code ADMS2 [4]. 2 The flow-field The approaching flow was a 1m deep neutral boundary layer artificially generated by the method of Couniham [5, 6]. The system consists of a castellated barrier wall used to create a momentum deficit and seven quart-elliptic vorticity generator placed near the beginning of the working section. Roughness elements, 20 mm tall, were placed downstream of the vorticity generators covering the entire tunnel length. Beyond an initial development fetch an ordinary two dimensional boundary layer which does not change significantly with downstream distance is generated. This kind of arrangement is known, from experience [7, 8], to produce a boundary layer with a height similar to the height of the vorticity generators and with turbulence levels similar to the ones usually measured in the planetary boundary layer. Measurements of vertical velocity and turbulence profiles have been taken by using a dual-beam laser-doppler anemometer at a number of positions downstream from the sources locations. Reference flow conditions were measured by an ultrasonic anemometer at a height of 100 mm and a vortex shedding flow meter at the boundary layer edge. The output of these instruments provided a quick and reliable check on the adequacy of flow conditions in the tunnel, which in some circumstances may be adversely affected by external conditions. Measured mean velocity, turbulence and shear stress profiles, for two different position X downstream from the sources, are presented in figure 1. This includes the velocity data plotted in log-linear coordinates, which allows ready evaluation of the log-law profile parameters, equation (1), where k=0.41 is von Karman's constant. The shear stress, u*, and roughness length, Zo, evaluated from this procedure were u*=0.060uref and Zo=0.7mm, where Uref is the wind speed at the boundary layer edge. A power law profile, equation (2), was also fitted to the data, with exponent Op = (1) -V =(?...}* (2, U(z = 100mm) V100mm/

4 712 Air Pollution In the log-law profile fit the value of u* was not actually a fitting parameter as it was obtained directly from the shear stress measured at 1000 M velocity pronies / jf r o X^4000mm # X=2000mm I r\n fif Power fit ) U(z)/Uref Longitudinal turbulence O # #X^2000mml, _ QX=4000mm / s 2 inn ^200 i nn Shear stress O#, # X=2000 mm O # o X=4000mm 0 ) # J*OQ %Rm Vertical turbulence O# 0 # 0# : # i < s 9--- ^ - ' ^ 1 inn 0# : # 8' ^200 : 100 [ X=2000 mml ifc ox=4000mm ^##% i... A ' *^* Fig. 1)Flow-field characteristics measured at two different values of X. about 60 mm above the ground; u* = - J< u'w' >. This height z=60mm corresponds to three times the roughness element height, where flow conditions are not influenced by individual elements. Results show that the velocity and turbulence profiles taken at the two measurement stations are very similar and therefore the flow is effectively completely developed with no significant changes in the flow direction. Measured turbulence characteristics indicate that the vertical turbulence intensity is smaller than longitudinal throughout the boundary layer and especially near the ground, as is typical of behaviour in the neutral planetary boundary layer [6].

5 Air Pollution The model under study Thefiillscale model has two nominally identical stacks 90 m in height, 1.44 m in diameter and 10 m apart releasing gas with a vertical speed Wgas of 15 m/s and a temperature of 419 K. The total flow rate Q, for the two stacks will therefore be NmVs. These values have been calculated for incinerator operation based on Selective Catalytic Reaction (SCR) for the reduction of the NO% emissions and a maximum allowable emission of 80 mg/nmu Discharge conditions for the small scale model have been obtained considering a set of standard buoyancy scaling relationships [8-9] because complete scaling [9] implies tunnel operating speeds which are too low to assure an acceptable boundary layer simulation. At theses speeds the flow would be excessively sensitive to changes in external conditions; what is more the roughness Reynolds number (U*ZQ/V) would be too low to assure that the floor of the tunnel is aerodynamical ly rough. When it is not possible to use complete scaling relationships some of the requirements have to be relaxed in order to simulate the real scenario in the wind tunnel. Plume rise is determined by the emitted momentum and buoyancy fluxes and the wind speed and this provides the basis for the scaling relationships. Suitable non-dimensional buoyancy and momentum flux parameters can be conserved leaving the density ratio a as a free parameter; a is the ratio between the gas mixture used to generate the plume and the air in the small scale model with respect to the full scale conditions. Once the complete scaling has been relaxed there are many approximate scaling relationships that are justifiable from a mathematical point of view (only some of which and may give accurate results [8-10]) as the different parameters that influence the plume trajectory can be combined in many different ways [11] giving different scaling rules that converge to a unique solutions only when a complete scaling is achieved. We chose to use a standard scaling rule, based on conservation of a Richardson number and momentum flux ratio; it is described by equation (3): = ^p ^ (<s P air U stack P air U stack where Qg is the total contaminant flow rate from the sources evaluated at NTP conditions and C is the concentration by volume. The left hand side of equation (3) is the non-dimensional concentration that has the same value in the model and in the full scale scenario; in the right hand side there are the parameters that have to be conserved in order to have a

6 714 Air Pollution physical similarity among the model and the full scale scenario. The scaling conditions following from (3) are presented in equation (4) u stack 1-a Wgas ToT Ustack IK (4) Table 1) Parameters of the real scenario and of the small scale model evaluated by using the buoyancy scaling relationships. Characteristic Boundary layer depth, H Stack height, h Stack internal diameter, D* Distance between stacks Speed above the BL, Uref Speed at the stack height, Usiack U(z= 100mm) Speed at 10m above the ground Gas releases flow rate, Q, Vertical speed of the gas, W^ Density ratio, a Flow Reynolds number (U^r H/v) Gas Reynolds numb. ( W, Di/v^,) Roughness Reynolds number Ambient temperature Emitted gas temperature Model 1 m 300mm 4.8 mm 33.3 mm m/s m/s m/s /min 2.88 m/s * 10" K K Real scenario 300m 90m 1.44m 10m 9.54 m/s 7.59 m/s 5 m/s m/s * * 10' K 419 K where the symbol ( ) refers to values in the model, a is the density ratio pgas/pair and y is the scale ratio (H)m/H. The density ratio (a)^ can be chosen quite freely in order, for example, to increase the required tunnel speed and/or to use a different mixture of gas. For example, referring to the measurements reported here with a mixture of 5.07% by volume of ethylene in helium we have (a)m=0.18 compared with 0.71 at full scale. In this mixture, the ethylene provides a trace gas for detection by a flame ionisation detector (FID). The equivalent full scale and the model parameters are listed in table 1. The emission Reynolds number, evaluated by using the vertical release speed, is much lower in the model than at full scale and therefore some differences in the exit velocity profile are likely which may lead to differences in the behaviour of the jet near the stack exit. This is a usual and, often unavoidable, problem when small scale models of tall and slender stacks are used. The flow

7 Air Pollution 715 Reynolds number referred to the stack diameter is also much smaller in the model creating differences in the wake structure behind the stacks. This will not create any evident effect because the ratio (W^/U^cOm is 3.9 and there is therefore no stack-tip down-wash because the exit vertical speed is too high for the gas to be trapped in the wake of the stacks. 4 Dispersion measurements results The gas sampling system used enables 16 samples of gas to be taken from various positions in the tunnel, and their hydrocarbon content analysed by a Signal Model 300 hydrocarbon analyser based on flame ionisation detection. The samples are first stored and then sequentially analysed after the experimental run is completed. The signal from the hydrocarbon analyser, once the background upstream of the sources has been subtracted, is compared with the signal from a calibration sample of known hydrocarbon content and hence the concentration of hydrocarbon in the sample is determined. A complete description of the experimental set-up is given in Ref. [9]. Two different compositions of calibration gas were used: mixtures of air and ethylene at volume concentrations of 212 and 1000 ppm. Repeatability checks were carried out on measured profiles by using both calibration gases and by repeating experiments - repeatability was high (within +/-5 to 10%) by virtue of careful experimental procedure. Vertical concentration profiles were measured at different distances X from the source along the wind direction (the centreline of the two stacks). The two stacks were arranges so that the line joining them was perpendicular to the tunnel centreline - a 0 degrees wind direction. Crosswind profiles were measured at a number of downstream positions at the stack height and at different heights 2 m downwind of the sources. Measured profiles shape were very close to gaussian. A comparison with the standard EPA gaussian code ISCST2 which use Brigg's formulae [1,2] for the plume rise, the vertical spread a, and the lateral spread Cy were carried out. A few examples of comparisons are reported in figure 2 for the vertical concentration profiles and in figure 3 for the crosswind profiles. The data are presented scaled for the full scale scenario. The results show that the standard ISCST2 code, used for rural environment, underestimate both the plume rise and the vertical spread. This confirms the results of comparisons with full scale experiment presented by Hanna and Chang [3]. The gaussian code with the BNL formulae gives a somewhat better agreement with the measured values of concentration,

8 716 Air Pollution CAUstack/Q CAUstack/Q % 2) Measured and calculated vertical profiles for different X. tough performance is less satisfactory for X greater than say 1000 m. The crosswind concentration profiles agreed quite well with one another, tough even in this case the BNL formulae give better results. Comparisons of the plume rise are shown in figure 4 and of vertical and lateral spread in figure 5. The ADMS2 plume rise predictions are close to the observation presumably because the integral model takes into account the structure of the external boundary layer. Comparisons were also made with the results of the UK integral code ADMS2 and with a modified version of the ISCST2 code. This modified gaussian code uses the formulae proposed by the National Brookhaven Laboratory (BNL) for the lateral and vertical spread that are claimed to be more suitable for releases from elevated sources as in the present case. The Brigg's plume rise formula is again used but a different method is used to evaluate the final plume rise. The distance downstream from the sources when the plume is levelled is taken to be the location where the slope of the plume centreline becomes less than 5%. A few experiments were carried out with the two stacks aligned with the wind direction (a wind direction of 90 degrees) in order to measure the effect of plume interactions and the increased rise which may occur. An example of the results is shown in

9 Air Pollution 717 figure 6 where the vertical profiles at two different values of X are plotted for both wind directions. Results show that the plume with the stacks aligned with the wind direction (hollow marks) rises above that the other ones. This effects is present for all the downstream distances analysed, though it is stronger near the sources. This effect can be explained because in the case of 90 degrees the two plumes mix with each other very soon after release, creating one single plume in which the lift effect due to the buoyancy force is increased with respect to the single emission. Actually even in the case of zero degrees the plumes mix but they do it far from the sources when they are already very diluted and the effect is almost negligible. Measurements carried out for different flow directions between 0 and 90 degrees show that this extralift effect is present only in a limited range of direction around 90 degrees. Results for all wind directions lie between the two profiles obtained with a single stack with the same total flow rate (i.e. maximum interaction) and a single stack with a half of the total flow rate (i.e. no interaction) X=300m d4 2 0 o Y(ra) T Trw^nr"^ I * Y(ra) Y(m) Fig. 3) Measured and calculated lateral profiles for different X, Similar behaviour has been found in measurements carried out by changing the gas flow rate to one half (medium flow rate) and to one

10 718 Air Pollution quarter (low flow rate) of the original. Plume rise effects are still present but analysis is complicated by the fact that the velocity ratio, Wgas/Ustack, decreases and stack-tip down-wash in the wake of the chimneys may become important, decreasing the average height of the plume. Measurements carried out for different flow directions between 0 and 90 degrees show that this extra-lift effect is present only in a limited range of direction around 90 degrees. 1OU * 0 _^ <y O 120 \x V V 110 X ISCST2 100 D 0 BNL D ADMS ^ Exp. an,,i,. i, i, i 4.. j *... i J,... -J ' 1 1 ' ' < 4 L Downwind distances X(m) Fig. 4) Measured and calculated values of the mean plume height > X(m) X(m) Fig. 5) Measured and calculated values of the plume spread Discussion and conclusions Results show that the gaussian codes may give good results in an average sense because looking at all the 60 profiles measured the predictions of ISCST2 within a factor two from the measured concentrations are 47%. The same analysis performed for ADMS2 and BNL shows that 82% and 79%, respectively, of the predictions are within a factor 2 of measured results.

11 N ^-^:^_ X=180m in...i.jt,,..,...1, III!.Wl 1 ' < i..lw CAUstack/Q Air Pollution 719 Transactions on Ecology and the Environment vol 21, 1998 WIT Press, ISSN zzu ^140 Nl Af\ 4U H 0 c^ <x X=300m C CAUstack/Q Fig. 6) Example of effects of flow direction on vertical profiles. This is of course an average behaviour on many different detector positions, however looking just to one detector the predictions could be considerably worse. The underestimation of the vertical spread of the ISCST2 code may leads to an underestimation of the ground level concentration that is only partially compensated by the smaller average height of the plume with respect to the measurements results. For the BNL case the estimation of the vertical spread is quite correct but the underestimation of the mean plume height will leads to an increased value of the ground level concentration along the centreline due to a contribution of the reflection term larger than in the measured results. The ground level concentration values are often the only indicator used in the environmental impact analysis and, as the results presented show, the estimations predicted by the gaussian codes are quite sensitive to the empirical formulae and parameters used to describe ambient air entrainment on the plume, and consequently plume rise, vertical and lateral spreads. With the experimental set-up used the detectability threshold was about 2.5 ppm and the plume at ground level was too dilute to allow measurements of concentration. Basically for ground level measurements a gas mixture with a higher concentration of ethylene has to be used at the stacks to generate the plume. Therefore the effect of the floor of the tunnel on plume dispersion were not evident in the measurements. The effect of the ceiling is negligible because the plume is always inside the boundary layer which terminate half meter below the ceiling. Acknowledgements Authors wish to thank Mr. Paul Hayden, EnFlo laboratory, for his help in setting up the measurements and Mr. Nathan Steggel for useful help and discussions.

12 720 Air Pollution References [1] Finzi G., Brusasca G., La qualita dellaria. Modelli previsionali e gestionali, Masson S.p.A., Milano [2] Zannetti P., Air pollution modeling, Computational Mechanics Publication, [3] Hanna S.R., Chang J.C., "Hybrid plume dispersion model (HPDM) improvements and testing at three sites", Atm, Env. 27AN.9,pp , [4] Carruthers, D J, Holroyd, R J, Hunt, J C R, Weng, W S,Robins, A G, Apsley, D D Thomson, D J, Smith, F B," UK-ADMS - a new approach to modelling dispersion in the earth's atmospheric boundary layer", Journ. Wind Eng. Ind. Aerodyn. 52, pp , [5] Couniham J., "An improved method of simulating atmospheric boundary layer in a wind tunnel", Atm. Env. 3, pp , [6] Couniham J., "Simulation of an adiabatic urban boundary layer in a wind tunnel", Atm. Env., pp , [7] Irwin H.P.A.H., "The design of spires for wind simulation", Journ. Wind. Eng. Ind. Aerodyn. 7, pp , [8] Cermak J. E., "Laboratory simulation of the atmospheric boundary layer", AIAA Journal N. 9, pp , [9] Robins A. G., "Wind tunnel modelling of plume dispersal", CEGB report, [10] Robins A. G., "Wind tunnel modelling of buoyant emissions", Atmospheric pollution 1980, Proceeding of the 14* International Colloquium, Studies in Environmental Science n. 8, France [11] Melbourne W.H., "Wind tunnel modelling of buoyant chimney plumes", proceeding of the Third Australian Conference on Hydraulics and Fluid Dynamics, pp , 1968.