Article pubs.acs.org/est
Abatement of an Aircraft Exhaust Plume Using Aerodynamic Baffles Michael Bennett,*,† Simon M. Christie,† Angus Graham,† Kevin P. Garry,‡ Stefan Velikov,‡ D. Ian Poll,‡ Malcolm G. Smith,§ M. Iqbal Mead,∥ Olalekan A. M. Popoola,∥ Gregor B. Stewart,∥ and Roderic L. Jones∥ †
Centre for Aviation, Transport and the Environment, School of Science and the Environment, Manchester Metropolitan University, Chester Street, Manchester, M1 5GD, U.K. ‡ School of Engineering, Cranfield University, Cranfield. Bedford MK43 OAL, U.K. § ISVR Consulting, University of Southampton, Southampton, SO17 1BJ, U.K. ∥ Department of Chemistry, University of Cambridge, Lensfield Road, Cambridge CB2 1EW, U.K. S Supporting Information *
ABSTRACT: The exhaust jet from a departing commercial aircraft will eventually rise buoyantly away from the ground; given the high thrust/ power (i.e., momentum/buoyancy) ratio of modern aero-engines, however, this is a slow process, perhaps requiring ∼1 min or more. Supported by theoretical and wind tunnel modeling, we have experimented with an array of aerodynamic baffles on the surface behind a set of turbofan engines of 124 kN thrust. Lidar and point sampler measurements show that, as long as the intervention takes place within the zone where the Coanda effect holds the jet to the surface (i.e., within about 70 m in this case), then quite modest surface-mounted baffles can rapidly lift the jet away from the ground. This is of potential benefit in abating both surface concentrations and jet blast downstream. There is also some modest acoustic benefit. By distributing the aerodynamic lift and drag across an array of baffles, each need only be a fraction of the height of a single blast fence.
■
INTRODUCTION AND OUTLINE THEORY Local air quality (AQ) is one of several issues constraining the development of airports. Studies made for the Project for the Sustainable Development of Heathrow,1 for example, showed that this area already breaches the current EU limit on annual NO2 concentrations (i.e., 40 μg m−3). The contribution of the airport to this breach would be far worse, however, were it not for the buoyant rise of the aircraft emissions. Some instructive simulations for Schiphol have been published in reference.2 For a mean emission height of 15 m, most (59%) of the local NO2 impact arises from the takeoff mode. Decreasing this height to 5 m more than doubles the impact; increasing it to 40 m reduces the impact by more than a factor of 3. We have observed such processes in our Lidar studies of aircraft emissions at Manchester and Heathrow airports,3 being able to follow exhaust plumes in light winds until their buoyant rise has carried them well clear (>50 m) of the ground. Broadly, the vertical momentum of the material emitted per unit time increases at a rate FB (the “buoyancy flux”), while its horizontal momentum is FM (i.e., the engine thrust). The two become equal, and the plume will rise strongly, after a travel time τ = FM /FB. For a modern high-bypass aero-engine, this scale time is of order 80 s.4 It would improve local AQ if τ were engineered to be as small as possible: this would then increase the wind speed necessary to observe the exhaust emissions on the ground in the neighboring community. This paper describes a field trial in © 2013 American Chemical Society
which we attempted to do this for a medium sized commercial aircraft (BAe146) by roughening the surface of the airfield sufficiently to remove a significant fraction of the momentum from the exhaust jet. We concentrated on the emissions in the initial stages of the takeoff run, since it is here that aircraft are at near maximum power and also moving most slowly. The hot exhaust plume cannot even start to rise until it has overcome the Coanda effect,3,5 which initially draws the jet toward to the surface. A free jet is conical,6 so at a distance x from the source, with an assumed spreading parameter β, it will have spread to an area A = π(βx)2. Meanwhile its excess velocity will have dropped to u, such that the momentum flux, FM = ρAu2, (where ρ is the ambient air density) remains close to the initial thrust. The volumetric flux, V = Au, through a given cross-section must therefore grow with distance, this being balanced by an external radial inflow toward the jet. Two parallel jets will therefore converge in their mutual centripetal flows, whereas a jet emitted parallel to a smooth surface converges toward its image jet behind the surface. This is the Coanda effect. A simple calculation (presented in Supporting Information (SI), Section S1) suggests that if the jet centerline starts at a height z0 above the surface then the expanding jet will Received: Revised: Accepted: Published: 2346
September 6, 2012 January 11, 2013 January 23, 2013 January 23, 2013 dx.doi.org/10.1021/es303586x | Environ. Sci. Technol. 2013, 47, 2346−2352
Environmental Science & Technology
Article
descend to touch the surface after a travel distance of xt ≈ 2z0/ β√5. Once the centerline of the jet has reached the surface, the downward momentum flux from the entrainment of air from above tends to keep it there, while any buoyancy in the jet provides a countervailing upward momentum. The two balance at a travel distance of xL ≈ (τ/π)1/2·(2FM/(πρ))1/4 ; from the definition of τ, we see that this distance is proportional to FM3/4. Beyond this point, the Coanda effect should no longer hold the jet closely to the ground. A further effect is that shear in the surface boundary layer will disrupt a plume, spreading the emission horizontally faster than it can rise. Hanna et al.7 suggest that a suitable criterion for whether a buoyant surface emission will rise or not is that the ratio U3Wπρ/FB, (where U is the ambient wind speed and W the local plume width) should be less than ∼100. This criterion in fact gives a comparable result to our formula for xL above if U is exchanged for u. The objective of our project was to enhance plume lift-off through the installation of an array of inclined baffles on the airfield behind the starting aircraft. If, overall, this imposes a drag D and a lift L on the exhaust jet, then the scale time to buoyant lift-off, τ′, is that at which FB τ′ + L = FM − D, that is, τ′ = (1 − (D+L)/FM)·τ; τ′ is thus minimized by maximizing D + L. Meanwhile, drag alone would reduce the travel distance before the jet escapes the Coanda effect by a factor of (1 − D/ FM)3/4 . It is clearly the case that a single suitably angled blast fence could successfully direct the jet away from the ground. On safety grounds, however, a tall barrier is not acceptable if aircraft are to land or take off over it. Within the safety envelope, a distributed array of low baffles should be more effective. This paper describes our experience with the implementation of such an array at near full scale and our use of a Lidar and various point samplers to monitor the influence that the baffles had on an aircraft exhaust jet. We also monitored engine noise beyond these baffles in order to exclude any adverse effect on the local noise environment. This was important, since any such effect would preclude their widespread deployment: although AQ represents the tightest legal constraint on the operation of major European airports, it is noise which provokes the greatest number of complaints. Acoustic barriers are normally made from material massive enough for their benefit to be limited by the diffraction of sound around the barrier rather than by residual transmission through it. Given the low surface density and high porosity of the material used for our aerodynamic baffles, it might be expected that any acoustic benefits would be small. Potential effects (beneficial or otherwise) include: • An acoustic barrier effect, providing a small reduction of noise at high frequencies. • Self-noise of the barriers either in the natural wind or from the jet of the departing aircraft. • A modification to the sound field due to changes in thermal refraction as the plume lifts away from the ground.
On each sortie, the aircraft would make a static burn at full thrust for a nominal 10 s, before taking off, performing a circuit, and landing. The aircraft has 4× Honeywell ALF502R-5 engines each of thrust 31 kN and a specific fuel consumption of 41.4 kg/kN.h. With a thermodynamic efficiency of ∼30%, this implies a buoyancy scale time of ∼86 s. Instruments available were (among others) a Rapid-Scanning Lidar,3,8−11 two Osiris optical particle counters (OPCs), and a set of electrochemical AQ nodes developed by the University of Cambridge12−14 (Figure 1). In addition, acoustic measurements downstream of
Figure 1. Disposition of instruments for the Cranfield trial. The starting points of the aircraft (defined as the engine exhaust) over the twelve sorties are labeled as nos. 1−12. The air quality nodes for Sortie nos. 5−12 are along the boundary fence, labeled S1−S6; they were more widely spaced for Sortie nos. 1−4. The sound level meters are numbered 1−6, with the Norsonics recorder being on the airfield adjacent to the runway threshold and the acoustic source being beside S4.
the aircraft were made by ISVR Consulting (University of Southampton). Meteorological measurements were available from a 7 m mast on the Lidar and at 2.1 m on the Osiris units. Video recordings of all operations were also taken. As noted, the objective was to install an array of baffles in the “Restricted Area” between the runway threshold and the boundary fence. Its impact on the dispersion of the exhaust jet from the aircraft would then be monitored. Two broad issues arose for the design of these baffles: they should impose sufficient drag on the jet to affect the flow significantly, and they should present no hazard to the normal operation of the airport. We will discuss the second point first. The most critical issue was that the baffles should be frangible. Formally, they should collapse cleanly if struck with a moment of >500 lb-ft, that is, 678 N-m.15 Conversely, they should also be able to withstand the jet blast. These contrasting requirements were met by having each row in the array consist of a line of independent relatively narrow baffles (“windowing”). These were also sufficiently frangible so that, if necessary, energency vehicles could drive straight through them without obstruction. It was also essential that any such array did not interfere with the navigation aids at the airport, either optical or RF. Given the layout of Cranfield airport, this was not here an issue but elsewhere it may be necessary to construct the baffles out of electrically insulating material.
■
ARRANGEMENTS FOR FIELD TRIALS Field trials were undertaken at Cranfield Airport (52° 4.33′ N, 0° 37.00′ W) over the period 14−21 September 2011, using an instrumented BAe146 aircraft (www.faam.ac.uk) as a source. 2347
dx.doi.org/10.1021/es303586x | Environ. Sci. Technol. 2013, 47, 2346−2352
Environmental Science & Technology
Article
Table 1. Parameters of Baffles As Implementeda baffle row label
arc radius/m
no. of baffles
slope height/m
inclination to horizontal
vertical height/m
individual baffle width/m
total arc length/m
A B C
36.0 53.3 74.0
23 28 33
1.4 1.7 2.0
40° 50° 60°
0.90 1.30 1.73
2.1 2.1 2.1
48.3 58.8 69.3
a The centre of curvature was at the intersection of the runway centreline with the threshold. The baffles were inclined away from the jet thrust. The ‘slope height’ is the height the baffle would have if it were deployed vertically.
Figure 2. Averaged Lidar scans for Sortie nos. 11 and 12 at low elevation. Wind speed, U, and direction is from the mast on the Lidar. Eastings and northings are relative to the Lidar. The color scale indicates the excess backscatter relative to that from clear air (cf. ref 3). Osiris 1 is close to the intersection of the boundary fence and the extended runway centerline.
surface-mounted plate inclined at 45° to the horizontal, rising to CF ≈ 1.12 at an inclination of 60°. These values are consistent with the WT simulations. With five rows of 45° baffles separated by 12× the slope height of the previous row, and with (slope height)/(travel distance) = 1/60, our scoping calculation had suggested that the array might remove 26% of the initial thrust. Note that a single 60° baffle of the height necessary to deliver the same aerodynamic drag would have to subtend an elevation angle of ∼3.0° at threshold. Given that the standard glide-path has an approach angle of 3°, this is clearly unacceptable; regulatory clear approach surfaces are typically 1.5° or less.20 In fact, the WT simulations showed that, even at this separation, there was significant interaction between rows, with 80−90% of the drag being achieved from just three rows. Given the constraints on the practicable number of baffles noted above, we thus limited ourselves to three rows. A scaled array corresponding to the parameters given in Table 1, though with α = 60° throughout (i.e., with the furthest subtending an elevation angle of 1.3° from threshold) was found to absorb >30% of the initial thrust in the presence of a very modest headwind. (An increased headwind was observed to increase the drag more than proportionately.) We see that the largest baffles (Row C) were no higher than a man’s height. In a commercial installation, more rows would be installed, increasing very gradually from initially modest baffles close to the threshold, probably to a similar ultimate height. The final design of baffle is illustrated in the image in the ToC. The baffles were constructed from aluminum, with each consisting of a 40 mm OD tube frame hinged to a front foot.
A permanent installation might be easier to design than the temporary installation for our field trial. We required a system that could be raised or lowered within 30 min, so as to be able to perform back-to-back measurements of the jet with or without the baffles. This severely constrained the feasible number of baffleswe ultimately used a total of 84and they were therefore wider individually than was desirable. Ground works were likewise excluded: permanent baffles could be set in a concrete base; ours were merely pegged down. On the other hand, for a temporary installation, there was no concern with its attracting birds, while the Airport was able to declare a displaced threshold of 150 m for approaching aircraft. This increased the safety margin of aircraft above the baffles by ∼8 m, as illustrated in the Table of Contents (ToC) image. The aerodynamic design of the baffles followed two separate paths: a simple scoping calculation based on published studies of wind loading on buildings,16,17 and wind tunnel (WT) simulations in the School of Engineering at Cranfield (Paper in preparation). The scoping calculation relied on earlier measurements at Cranfield18,19 which had suggested that the vertical spreading angle of the jet was typically β ≈ 1/9.0. Given a momentum flux within the jet, we could then translate this into the normal aerodynamic load, FN on an inclined, groundbased, flat plate.16 If the array consists of free-standing flat plates at an angle α to the horizontal, then D + L = FN (cosα + sinα). For a given vertical height of plate, τ′ is thus minimized when α ≈ 45°. The reaction of the ground (i.e., the dynamic surface pressure field) reduces L, leading to the optimum angle being probably α ∼ 60°. Extrapolating from published values, we should expect a normal drag coefficient, CF ≈ 0.97, for a 2348
dx.doi.org/10.1021/es303586x | Environ. Sci. Technol. 2013, 47, 2346−2352
Environmental Science & Technology
Article
Table 2. Point Surface Measurements within the Plume for Sortie Nos. 9−12a lidar (peak value)
Osiris 1 (peak values)
AQ node (integrated value)
aerosol/μg m−3 sortie no.
baffles
backscatter/ratio of excess to ambient
wind speed/m s−1
PM10
PM2.5
PM1
9 10 11 12
down up up down
0.47 0.47 0.27 0.66
46.4 22.4 26.0 44.8
47.92 10.11 12.35 63.15
16.61 5.24 7.27 6.18
6.70 2.08 2.70 2.74
[NO]/ppm s ± 1σ 69.5 62.8 43.2 82.7
± ± ± ±
0.6 0.5 0.4 0.6
a
The quoted Lidar backscatter is the peak observed value above the location of Osiris 1. PM values are the peak 5 s values above the preceding background. NO concentrations are summed over all sensors and integrated over the time of passage of the plume.
Cambridge. These small, robust, low power sensors have been widely used in industrial gas sensing and are now sufficiently sensitive to operate at ambient mixing ratios (∼ppb for NOx). They are intended for deployment within flexible sensor networks, investigating dispersion and individual exposure in the urban environment. Those used in this study consisted of autonomous nodes each with three electrochemical cells (CO, NO, NO2), conditioning electronics, GPS and a GPRS to enable the routine sending of data in near real time over the mobile telephone network. Their deployment thus required minimal overhead. The sensors are discussed in detail in refs 12 and 14. For this study a row of six hand-held units was deployed along the airport perimeter fence, these being attached to the fence posts ∼50 cm above ground level. (As components of a flexible network, we routinely deploy these sensors quite close to the ground; this minimizes their impact at an operational site.) For Sortie nos. 1−4, sensors were arranged at uniform 50 m intervals centered on the runway centerline. For Sortie nos. 5−12, they were evenly arranged over 50 m centered on the runway centerline, with the remaining two sensors 50 m away at either end (Figure 1). We could thus collect more detailed information on the center of the exhaust plume while still monitoring its periphery. The mixing ratio of NO was selected as a proxy for the distribution of the plume, since it was clearly recorded at all sensors for all sorties. CO and NO2 were also detected by all sensors but there were confounding local CO sources (airport vehicles), while the NO2 signal was only ∼15% of the NO value. A typical time series for NO is shown in Figure 3, which clearly shows the aircraft plume signature for Sortie no. 3. While the aircraft was stationary, the plume should be aligned with the engine thrust, that is, along the centerline toward
The top of the frame was connected through a fuse bolt to a lightweight strut hinged to a hind foot. The feet were then pegged down. By removing the fuse bolts, a baffle could be lowered in a few seconds. The baffle itself consisted of 80% solidity agricultural wind-break fabric, the porosity being intended to reduce eddy generation. The baffles were installed in the Restricted Area between the runway threshold and the airport perimeter fence (Table 1, Figure 1), and left flat prior to the trials. During trial days, a first sortie would be run with the baffles down; two would be run with the baffles up; and then a final sortie with the baffles down again. We could thus pair runs with and without the baffles. Initially, all the baffles were at 60°. In Sortie nos. 2 and 3 (14/9), however, with the aircraft engines being, respectively, 71 and 106 m from the first baffle and the static burst continuing for 15 s, the design was found to be insufficiently robust. The inclinations were therefore reduced as indicated (and the number and size of pegs increased). For Sortie nos. 6 and 7 (15/9) we were then too cautious, with the aircraft engines being, respectively, 154 and 109 m from the first baffle and the static burst continuing for only 5 s. This was insufficient to demonstrate the desired effect. On the final day (21/9), the aircraft was returned to close to its original position (93 m for Sortie no. 10 and 60 m for Sortie no. 11) and the burst duration was restored to 11 s.
■
DISPERSION MEASUREMENTS (1). Lidar. The Lidar was sited as indicated in Figure 1, with the laser source being 3 m above ground level. Alternate azimuthal scans were made at elevations of ∼0.7° (i.e., just high enough to avoid minor surface obstructions) and 4.5°, the intention being to detect any separation of the plume from the surface. The range of azimuths was as indicated in Figure 1, with successive shots being separated by 0.5° (i.e., 80 shots per scan). By scanning alternately right-to-left and left-to-right, the total cycle time was minimized to between 10 and 11 s. The Lidar data were analyzed using the algorithms described in ref 3. We could thus produce near horizontal cross sections of each scan and observe the plume propagate away from the aircraft. Useful measurements of the impact of the baffles were made on the final day (Sortie nos. 9−2). The cross sections shown in Figure 2 show the average returns from the scans at the lower elevation for the last two sorties (cf. SI Figure S1), while Table 2 lists some key near-surface observations within the jet, including the peak backscatter measured by the Lidar above the point where the jet centerline crosses the boundary fence. (2). Point Samplers. (2.1). Cambridge AQ Sensor Nodes. A range of low cost hand-held or static AQ sensor nodes, based on advanced electrochemical gas sensors (cf., e.g., www. alphasense.com), has been developed at the University of
Figure 3. Plot showing NO concentrations at the six AQ nodes (S1− S6) for Sortie no. 3. 2349
dx.doi.org/10.1021/es303586x | Environ. Sci. Technol. 2013, 47, 2346−2352
Environmental Science & Technology
Article
showed that at the smaller range, the exhaust jet was closely ground-based; at the further range, however, the peak concentration was already a few m above the ground. From the simple theory noted in the Introduction, we should have expected the exhaust jet to be at the ground from xt ≈ 18 m, while it should be starting to leave it from xL ≈ 84 m or less. This analysis thus suggests that for Sortie no. 10, the jet had already lifted far enough for the array no longer to be fully effective; conversely, it also hints at why the baffles experienced a drag greater than the design value in Sortie nos. 2 and 3. To confirm how the jet has dispersed in Sortie nos. 10 and 11, we extracted the plume centerline backscatters for the two elevations for these two sorties. These values were estimated by treating the azimuthal cross sections of mean backscatter as an array of azimuth vs range (i.e., shot number vs bin number). For each range from the Lidar, the second-largest backscatter as a function of azimuth was taken to indicate the centerline concentration. (The second-largest value is statistically more robust than the maximum value, since an anomalously large value automatically censors itself. It is a favored statistic of the EPA.21) Simple trigonometry was then used to convert range from the Lidar to downstream distance from the source. For Sortie no. 10 we found that centerline concentrations at the higher elevation only approached those at the lower elevation after a travel distance of 300 m; even at the furthest distance sampled, that is, >400 m, concentrations at the lower elevation were still somewhat greater (SI Figure S3). For Sortie no. 11 (i.e., with the first baffle at only 60 m from the source), the concentrations at the two elevations were essentially indistinguishable from a travel distance of 150 m, that is, from scarcely having cleared the baffle array. They also fell much more rapidly: by 300 m, near-surface concentrations were only a third of those observed in Sortie no. 10. It is thus clear that for Sortie no. 11, the short distance of the array from the source led to the plume leaving the ground very rapidly. The extra 33 m in Sortie no. 10 delayed this process. The array seems nevertheless still to have had some effect: for Sortie nos. 9 and 12, no coherent plume could be observed for the higher elevation (SI Figure S4), implying that the jet remained close to the ground (i.e.,
