Artigo Acesso aberto Revisado por pares

Estimating Aircraft Landing Weights from Mode S Data

2023; American Institute of Aeronautics and Astronautics; Volume: 60; Issue: 2 Linguagem: Inglês

10.2514/1.c036689

ISSN

1533-3868

Autores

Frank Holzäpfel, Grigory Rotshteyn,

Tópico(s)

Aerospace and Aviation Technology

Resumo

Open AccessEstimating Aircraft Landing Weights from Mode S DataFrank Holzäpfel and Grigory RotshteynFrank Holzäpfel https://orcid.org/0000-0003-3182-1832DLR, German Aerospace Center, 82234 Oberpfaffenhofen, Germany and Grigory RotshteynDLR, German Aerospace Center, 82234 Oberpfaffenhofen, GermanyPublished Online:1 Feb 2023https://doi.org/10.2514/1.C036689SectionsRead Now ToolsAdd to favoritesDownload citationTrack citations ShareShare onFacebookTwitterLinked InRedditEmail AboutI. IntroductionAircraft mass is a key parameter controlling aircraft performance, and thus the prediction of aircraft path routings, and climb and descent profiles; hence, it also constitutes relevant information for aircraft traffic management. However, these proprietary data are usually not easily accessible because airlines consider them confidential information.During the definition process of the concepts of operations (Operational Services and Environmental Definition) for transmitting aircraft-derived meteorological data to enable a wide range of Next Generation Air Transportation System and Single European Sky Air Traffic Management Research applications in the areas of wake turbulence, air traffic management, and meteorology, it was generally agreed that aircraft mass should be incorporated into the downlink and crosslink of aircraft-derived data [1,2]. However, it was also stated that aircraft mass is traditionally considered sensitive information due to operator concerns about potentially releasing data to competitors [2]. This classification may continue to hamper the provision of mass data in future data protocols.Because wake turbulence constitutes a potential risk for following aircraft, separation standards between consecutive aircraft have been established that limit the capacity of congested airports [3]. Because the strength of aircraft wake vortices is directly proportional to the aircraft weight [4], mass information is required for the prediction of pairwise dynamic aircraft separations for approach and landing with a wake vortex advisory system [5]. Here, the consideration of landing weight extends from the operational empty weight to the maximum landing weight, leading to wide ranges of initial wake vortex strength, which in turn are limiting the potential capacity gains of such advisory systems [6,7]. For efficient onboard prediction of wake turbulence, the wake vortex volumes that need to be avoided by other aircraft should be as compact as possible. To accomplish this objective, the knowledge of aircraft mass is of paramount importance [8]. Also, the wake turbulence re-categorization (RECAT) process for the optimization of wake turbulence separations initiated by the International Civil Aviation Organization (ICAO) requires adequate landing weights for the classification into different categories [9,10]. The estimation of runway occupancy times, which is another important element for optimizing the arrival capacity of busy airports, also depends on reliable estimates of aircraft landing weights. Moreover, the planning of continuous descent operations or adaptive increased glide slopes aiming at reduced fuel consumption and noise mitigation depend on the aircraft gross mass [11,12].A number of partly quite extensive methods for the estimation of aircraft weights for different phases of flight have been reported in the literature. Chati and Balakrishnan applied statistical machine learning techniques (Gaussian process regression) to model the aircraft takeoff weight using flight data recorder data from the takeoff ground roll, achieving a mean absolute error of 3.6% [13]. Alligier et al. focused on the estimation of aircraft mass during the climb phase using point-mass model parameters from observations fitting the modeled power to the observed energy rate [14,15]. Sun et al. estimated takeoff mass from total energy model estimates at different flight phases where validation experiments yield a mean absolute error of 4.3% of the actual aircraft mass [16]. Fricke et al. presented a technique to estimate the aircraft gross mass for calculating the optimum rate of descent by employing a flight mechanical approach using the measured approach ground speed [11]. The validation based on 12 flight trajectories yields an average deviation between modeled and recorded gross masses of 12.3%. Measurements of the actual landing weights at the airports of Memphis and Dallas/Fort Worth resulted in an overall average of 85% of the maximum landing weight (MLW) [17]. This average value was also adopted as a typical landing weight in the "European Proposal for Revised Wake Turbulence Categorisation and Separation Minima on Approach and Departure" (known as RECAT-EU) [18].This Note employs equations suggested for estimates of descent speeds depending on aircraft gross mass, following the Base of Aircraft Data (BADA) [19] to estimate aircraft weights before touchdown. The required aircraft type, calibrated airspeed, and air density data are derived from mode S data protocols. The landing weights of 3328 aircraft approaching Vienna International Airport, provided by Austrian Airlines for the analysis of the wake turbulence mitigation field measurement campaign [20,21] conducted from May to November 2019, serve as reference masses. A simple correction for the observed inclination of pilots to fly somewhat faster than prescribed in reference handbooks eliminates the bias of the mass estimates in the current database.II. MethodA. Mode-S-Based Mass EstimatesAt the Vienna International Airport, a measurement campaign has been accomplished from 6 May 2019 until 28 November 2019 in order to assess the potential of so-called plate lines to mitigate wake turbulence risks during final approach [20,21]. The wake vortex measurements employing three lidars (light detection and ranging) were conducted in five measurement planes in which the flight altitudes above ground were determined from mode S data to 40.8, 45.8, 54.3, 64.8, and 74.5 m, respectively, with a standard deviation of 4.9 m. The mass estimates, derived from mode S data interpolated to these five measurement planes, are all adopted in the statistics in order to smoothen the scatter caused by intended airspeed variations and those introduced by turbulence as well as random measurement error.Mode S is a secondary surveillance radar process that allows selective interrogation of aircraft employing ground-based interrogators and airborne transponders [22]. The provided mode S data protocols contain the date, time, aircraft position, attitude, and airspeed data as well as the atmospheric pressure, wind speed, and temperature data at an update rate of 4 s. Metadata in the file for each landing provide the aircraft type, call sign, origin, destination, time stamps for the start and end of aircraft tracking, as well as the runway intended for landing. From these available parameters, the determination of the aircraft gross mass employs the aircraft type and the values of the true airspeed, air pressure, and temperature.According to BADA [19], calibrated airspeeds VCAS for jet and turboprop aircraft in the descent phase amount to VCAS=CVmin⋅Vstall+VdDES(1)where CVmin=1.3, and Vstall is the stall speed for the landing configuration of the particular aircraft type. The descent speed increment VdDES amounts to 5 kt for flight altitudes below 1000 ft, to 10 kt between 1000 and 1500 ft, to 20 kt between 1500 and 2000 ft, and to 50 kt between 2000 and 3000 ft. Aircraft operating speeds vary with the aircraft mass. This variation is considered according to Vstall=Vstall,ref⋅mmref(2)where Vstall,ref is the reference stall speed for the landing configuration of the particular aircraft type, and mref is its respective reference mass. From Eqs. (1) and (2), the landing mass can be estimated according to m=(VCAS−VdDESCVmin⋅Vstall,ref)2mref(3)The true airspeed VTAS from the mode S protocol is translated into calibrated airspeed VCAS with the relation VCAS=VTAS ρ/ρ0(4)simplified for low flight altitudes, where ρ0 is the density of air at sea level of the International Standard Atmosphere. Air density ρ is retrieved from the pressure p and temperature T, employing the equation of state for perfect gases of ρ=p/RT, where the gas constant amounts to R=287 J/kg⋅K for dry air. More elaborate computations of VCAS from VTAS (see BADA user manual [19]) on average increased the landing mass estimates by only 0.027%, and are therefore not considered here.B. Reference Mass DataAustrian Airlines kindly provided the landing weight m of all recorded landings of their fleet during the Vienna field trial. The weight data were provided for 2958 landings of the A320 family representative of the ICAO weight category medium and category D (upper medium) of the RECAT-EU scheme. The landings of the A320 family comprise 560 A319 aircraft, 1932 A320 aircraft, and 466 A321 aircraft. Landings of the B767-300ER falling into category C (lower heavy) of the RECAT-EU scheme amount to 225. The B777-200ER with 145 landings is the most frequent representative of the RECAT-EU scheme B (upper heavy) within the Vienna database (1.2% traffic share).These weights serve as a reference to assess the weight estimates derived from mode S data. The reference landing weights are quoted from the estimated initial aircraft weight and its fuel consumption. Before every flight, a load and balance sheet are established, where the load sheet contains the planned takeoff mass. After the flight, the fuel consumption corrected by the consumption for the taxiing to and from the runway is subtracted from the takeoff mass. The fuel consumption is captured by flow rate measuring devices located directly at the jet engines and is displayed in the cockpit where the display is set at zero with every engine start.III. Statistical AnalysesTable 1 presents landing weight (LW) statistics of the considered aircraft types. Although the average mass of the B767-300ER almost perfectly coincides with the 85% of the MLW found in the statistics collected at Memphis and Dallas/Fort Worth airports [17], the other aircraft types are heavier than 85% of the MLW by 6.9 (B777) and 7.8 (A320) percentage points. The respective standard deviations, σ[(LW)/MLW], varying between 2.7 and 5.4% appear relatively small.Figure 1 depicts the landing weights estimated according to Eqs. (3) and (4) from mode S data plotted against the reference masses provided by Austrian Airlines. To avoid unrealistically high landing masses, the estimated masses were alternatively clipped off at the MLW (black dots) or at 95% of the MLW (blue dots), where the 95% fraction constitutes an arbitrarily chosen empirical value. The clusters representing the A320 family (bottom left), the B767-300ER (center), and the B777-200ER aircraft (top right) are clearly delimited. Figure 2 shows the corresponding plots for the individual aircraft types to exhibit the performance of our estimation approach in more detail. Whereas in Fig. 1 the good alignment of the linear regression curves is mainly related to the distinct weight clusters of the three considered aircraft weight categories, the slopes of the regression curves for the individual aircraft types plotted in Fig. 2 tend to tilt the more the estimated masses are clipped off.The data from the five measurement planes for one landing aircraft frequently appear as five vertically aligned data points because any speed variation of the aircraft while passing the five measurement planes corresponds to mass estimate variations, whereas the reference data only provide one mass value for each landing aircraft. This effect is most apparent with the few outliers of the B777-200ER weight estimates. Obviously, the aircraft weights are overestimated for all aircraft types while the scatter of the mass data is reasonably low.Fig. 1 Estimated landing weights partly clipped off at different fractions of the MLW against actual weights of several aircraft types.Fig. 2 Estimated landing weights as in Fig. 1 for individual aircraft types.Table 2 lists the corresponding biases and standard deviations between estimated and reference landing weights of the data plotted in Fig. 1 without clipping. The A320 family features the largest relative weight overestimates and corresponding standard deviations, whereas the B767-300ER data show the smallest deviations.The exclusion of estimated weights above the MLW reduces the bias and the standard deviation mostly for the A320 family (Table 3). Because the A320 family exhibits both the highest relative landing weights (see Table 1) and the highest mass overestimates (Table 2), the relative frequency of landings with clipped off masses is also the highest. For the B767-300ER with an average landing weight of 0.85 MLW, the MLWs are almost not exceeded by the estimates such that the figures are almost unchanged as compared to Table 2.Clipping estimated masses above 0.95 MLW (Table 4) reduces the bias for all aircraft types with little impact on the standard deviations. As can be seen in Figs. 1 and 2, the clipping partly compensates for the overestimation of the mode S data approach such that the biases are reduced.Typical landing speeds are slightly higher than quoted in reference handbooks to have at command a certain buffer for unforeseen events, such as rapid changes in wind speed or direction. A comparison of the measured approach speeds with those prescribed for landing in the A320 Quick Reference Handbook [23] indicates that A320 pilots fly on average 1.2 m/s faster than scheduled. This comparison considers the aircraft gross mass and headwind strength, and it assumes that the pilots employed the landing configuration "full" and activated the autothrust system.Figures 3 and 4 exhibit that a doubling of the descent speed increment VdDES to 10 kt reduces the mass estimates substantially. As a consequence, the bias for all aircraft types almost completely vanishes as detailed in Table 5, whereas the standard deviation remains almost the same. With the descent speed adjustment, the slopes of the regression curves deviate from the ideal slope of one by not more than 6%. Figure 4 indicates that the doubling of VdDES appears slightly overdone for the Boeing aircraft but seems appropriate for the Airbus aircraft.Figure 5 delineates the distribution of the relative differences between the estimated landing masses with VdDES=10 kt and the reference masses. Unsurprisingly, the distribution is skewed with a steeper decline toward the MLW.Fig. 3 Estimated and actual landing weights for different values of VdDES and several aircraft types.Fig. 4 Estimated landing weights as in Fig. 3 for individual aircraft types.Finally, all mode S data available during the descent from 3000 ft to touchdown are used to assess the suitability of the mass estimates at higher altitudes by employing the values for VdDES introduced in Sec. II.A. For this purpose, the deceleration of the aircraft within a height range was considered by linear interpolation of VdDES, leading to a somewhat reduced bias without much impact on the standard deviation. It was found that the standard deviation in the height range from 0 to 1000 ft amounts to 6.7%. The standard deviations are roughly doubled in the height range 1000 to 1500 ft to 12.7%, tripled in the range 1500 to 2000 ft to 18.9%, and increased fourfold between 2000 and 3000 ft to 27.4%. Obviously, in the higher-altitude ranges above 1000 ft, the standard deviations increase substantially, presumably because air traffic control prescribes the descent speeds depending on the wind and traffic situation. Already, a doubling of the standard deviation limits the application of the weight estimate substantially for many purposes.Fig. 5 Distribution of relative differences between estimated and measured masses for all considered aircraft types with a doubling of VdDES.IV. ConclusionsAircraft landing weights are estimated from mode S protocol data comprising the aircraft type and the values of the true airspeed, air pressure, and temperature. These parameters are translated into air density and calibrated airspeed from which the aircraft mass can be deduced by employing the reference stall speed and reference mass of the particular aircraft type obtained from the BADA database. The weight estimates are compared to reference data provided by Austrian Airlines for 3328 landings of the A320 family, the B767-300ER, and the B777-200ER. The agreement between the estimated and actual landing weights achieved with the unmodified approach is fair, and the scatter of the mass data is reasonably low. The bias of the weight estimates can be reduced by clipping off overestimated weights exceeding the maximum landing weight or 95% of the MLW. The gross mass bias can be almost completely eliminated by consideration of the tendency of pilots to slightly exceed the prescribed approach speeds. At flight heights above 1000 ft, the scatter of the mass estimates increases substantially because actually flown airspeeds deviate increasingly with height from the average values assumed in the BADA database.AcknowledgmentsThis project has received funding within the framework of the Single European Sky Air Traffic Management Research Joint Undertaking "Increased Runway and Airport Throughput" project (PJ.02 EARTH) and the "Safely Optimized Runway Throughput" project (VLD3-W2 SORT) within the European Union's Horizon 2020 Research and Innovation Programme under grant agreement nos. 731781 and 874520 as well as from the DLR, German Aerospace Center (DLR) project "Wetter und Disruptive Ereignisse." The provision of landing masses by W. Wurzinger from Austrian Airlines is greatly acknowledged, and the provision of the mode S data by L. Strauss and C. Weiß from MeteoServe Wetterdienst GmbH is highly appreciated. Finally, we give thanks to D. Vechtel from the DLR's Institut für Flugsystemtechnik for the instructive discussions. References [1] "Aircraft Derived Meteorological Data via Data Link for Wake Vortex, Air Traffic Management and Weather Applications, Operational Services and Environmental Definition (OSED)," RTCA STD DO-339, Washington, D.C., June 2012. Google Scholar[2] "Standards Development Activities for using Near-Real-Time Aircraft-Derived Data in Future Applications," RTCA STD DO-360, Washington, D.C., Sept. 2015. Google Scholar[3] Hallock J. N. and Holzäpfel F., "A Review of Recent Wake Vortex Research for Increasing Airport Capacity," Progress in Aerospace Sciences, Vol. 98, April 2018, pp. 27–36. https://doi.org/10.1016/j.paerosci.2018.03.003 CrossrefGoogle Scholar[4] Gerz T., Holzäpfel F. and Darracq D., "Commercial Aircraft Wake Vortices," Progress in Aerospace Sciences, Vol. 38, No. 3, April 2002, pp. 181–208. https://doi.org/10.1016/S0376-0421(02)00004-0 CrossrefGoogle Scholar[5] Holzäpfel F., Strauss L. and Schwarz C., "Assessment of Dynamic Pairwise Wake Vortex Separations for Approach and Landing at Vienna Airport," Aerospace Science and Technology, Vol. 112, May 2021, Paper 106618. https://doi.org/10.1016/j.ast.2021.106618 CrossrefGoogle Scholar[6] Holzäpfel F., Gerz T., Frech M., Tafferner A., Köpp F., Smalikho I., Rahm S., Hahn K.-U. and Schwarz C., "The Wake Vortex Prediction and Monitoring System WSVBS, Part I: Design," Air Traffic Control Quarterly, Vol. 17, No. 4, 2009, pp. 301–322. https://doi.org/10.2514/atcq.17.4.301 LinkGoogle Scholar[7] Gerz T., Holzäpfel F., Gerling W., Scharnweber A., Frech M., Kober K., Dengler K. and Rahm S., "The Wake Vortex Prediction and Monitoring System WSVBS Part II: Performance and ATC Integration at Frankfurt Airport," Air Traffic Control Quarterly, Vol. 17, No. 4, 2009, pp. 323–346. https://doi.org/10.2514/atcq.17.4.323 LinkGoogle Scholar[8] Sölch I., Holzäpfel F., Abdelmoula F. and Vechtel D., "Performance of On-Board Wake Vortex Prediction Systems Employing Various Meteorological Data Sources," Journal of Aircraft, Vol. 53, No. 5, 2016, pp. 1505–1516. https://doi.org/10.2514/1.C033732 LinkGoogle Scholar[9] Cheng J., Tittsworth J., Gallo W. and Awwad A., "The Development of Wake Turbulence Recategorization in the United States," AIAA Paper 2016-3434, June 2016. https://doi.org/10.2514/6.2016-3434 Google Scholar[10] "RECAT-EU: European Wake Turbulence Categorisation and Separation Minima on Approach and Departure," EUROCONTROL, Brussels, ed. 1.2, 2018, p. 26, https://www.eurocontrol.int/publication/european-waketurbulence-categorisation-and-separation-minima-approach-and-departure [retrieved 17 Jan. 2023]. Google Scholar[11] Fricke H., Seiß C. and Herrmann R., "Fuel and Energy Benchmark Analysis of Continuous Descent Operations," Air Traffic Control Quarterly, Vol. 23, No. 1, 2015, pp. 83–108. https://doi.org/10.2514/atcq.23.1.83 LinkGoogle Scholar[12] Clarke J.-P. B., Ho N. T., Ren L., Brown J. A., Elmer K. R., Tong K.-O. and Wat J. K., "Continuous Descent Approach: Design and Flight Test for Louisville International Airport," Journal of Aircraft, Vol. 41, No. 5, 2004, pp. 1054–1066. https://doi.org/10.2514/1.5572 LinkGoogle Scholar[13] Chati Y. S. and Balakrishnan H., "Modeling of Aircraft Takeoff Weight Using Gaussian Processes." Journal of Air Transportation, Vol. 26, No. 2, 2018, pp. 70–79. https://doi.org/10.2514/1.D0099 LinkGoogle Scholar[14] Alligier R., Gianazza D. and Durand N., "Learning the Aircraft Mass and Thrust to Improve the Ground-Based Trajectory Prediction of Climbing Flights," Transportation Research, Part C: Emerging Technologies, Vol. 36, Nov. 2013, pp. 45–60. https://doi.org/10.1016/j.trc.2013.08.006 CrossrefGoogle Scholar[15] Alligier R., Gianazza D., Ghasemi Hamed M. and Durand N., "Comparison of Two Ground-based Mass Estimation Methods on Real Data," International Conference on Research in Air Transportation (ICRAT), Istanbul, Turkey, May 2014, p. 8, https://drive.google.com/file/d/1UH4gqx9jF3onvyHouPxOza_PlsSAr2jq/view [retrieved 17 Jan. 2023]. Google Scholar[16] Sun J., Ellerbroek J. and Hoekstra J., "Aircraft Initial Mass Estimation Using Bayesian Inference Method," Transportation Research, Part C: Emerging Technologies, Vol. 90, May 2018, pp. 59–73. https://doi.org/10.1016/j.trc.2018.02.022 CrossrefGoogle Scholar[17] Delisi D. P., Pruis M. J., Wang F. Y. and Lai D. Y., "Estimates of the Initial Vortex Separation Distance, bo, of Commercial Aircraft from Pulsed Lidar Data," 51st AIAA Aerospace Sciences Meeting including the New Horizons Forum and Aerospace Exposition, AIAA Paper 2013-0365, 2013. https://doi.org/10.2514/6.2013-365 LinkGoogle Scholar[18] "European Proposal for Revised Wake Turbulence Categorisation and Separation Minima on Approach and Departure—'RECAT—EU' - Safety Case Report," EUROCONTROL, Brussels, 2017. Google Scholar[19] "User Manual for the Base of Aircraft Data (BADA) Revision 3.11," EUROCONTROL Experimental Center Technical/Scientific Rept. 13/04/16-01, Brétigny-sur-Orge, May 2013. http://upcommons.upc.edu/bitstream/handle/2099.1/24342/AnnexI.pdf?sequence=2 [retrieved 12 March 2021]. Google Scholar[20] Holzäpfel F., Stephan A., Rotshteyn G., Körner S., Wildmann N., Oswald L., Gerz T., Borek G., Floh A., Kern C., Kerschbaum M., Nossal R., Schwarzenbacher M., Strobel M., Strauss L., Weiß C., Kauczok S., Schiefer C., Czekala H., Maschwitz G. and Smalikho I., "Mitigating Wake Turbulence Risk During Final Approach via Plate Lines," AIAA Journal, Vol. 59, No. 11, 2021, pp. 4626–4641. https://doi.org/10.2514/1.J060025 LinkGoogle Scholar[21] Holzäpfel F., Vechtel D., Rotshteyn G. and Stephan A., "Plate Lines to Enhance Wake Vortex Decay for Reduced Separations Between Landing Aircraft," Flow: Applications of Fluid Mechanics, Vol. 2, March 2022, pp. E6-1–E6-26. https://doi.org/10.1017/flo.2021.16 Google Scholar[22] "Manual on Mode S Specific Services," International Civil Aviation Org. Doc. 9688, AN/952, 2nd ed., Montreal, 2004, http://www.anteni.net/adsb/Doc/9688_cons_en.pdf [retrieved 12 March 2021]. Google Scholar[23] N.N., "A320 Quick Reference Hand Book," Airbus Industries, June 2010. Google ScholarTablesTable 1 Landing weight statistics LandingsLW/MLW, %σ(LW/MLW), %A320 family295892.85.4B767-300ER22584.92.7B777-200ER14591.93.6Table 2 Deviations between estimated and actual landing weights (not clipped) LandingsBias, %σ, %A320 family29588.65.1B767-300ER2253.83.7B777-200ER1454.63.6Table 3 Deviations between estimated and actual landing weights (clipped at 100% MLW) LandingsClippedBias, %σ, %A320 family295817785.33.9B767-300ER22523.83.6B777-200ER145304.13.2Table 4 Deviations between estimated and actual landing weights (clipped at 95% MLW) LandingsClippedBias, %σ, %A320 family295824141.64.7B767-300ER225203.63.4B777-200ER145992.03.2Table 5 Deviations between estimated and actual landing weights varying VdDES VdDES, ktLandingsBias, %σ, %All aircraft533288.15.21033280.35.3 Next article FiguresReferencesRelatedDetails What's Popular Articles in Advance CrossmarkInformationCopyright © 2023 by Deutsches Zentrum für Luft- und Raumfahrt (DLR). Published by the American Institute of Aeronautics and Astronautics, Inc., with permission. All requests for copying and permission to reprint should be submitted to CCC at www.copyright.com; employ the eISSN 1533-3868 to initiate your request. See also AIAA Rights and Permissions www.aiaa.org/randp. AcknowledgmentsThis project has received funding within the framework of the Single European Sky Air Traffic Management Research Joint Undertaking "Increased Runway and Airport Throughput" project (PJ.02 EARTH) and the "Safely Optimized Runway Throughput" project (VLD3-W2 SORT) within the European Union's Horizon 2020 Research and Innovation Programme under grant agreement nos. 731781 and 874520 as well as from the DLR, German Aerospace Center (DLR) project "Wetter und Disruptive Ereignisse." The provision of landing masses by W. Wurzinger from Austrian Airlines is greatly acknowledged, and the provision of the mode S data by L. Strauss and C. Weiß from MeteoServe Wetterdienst GmbH is highly appreciated. Finally, we give thanks to D. Vechtel from the DLR's Institut für Flugsystemtechnik for the instructive discussions.PDF Received24 September 2021Accepted1 January 2023Published online1 February 2023

Referência(s)