energies

Article A Study on the Strategy for Departure Pushback Control from the Perspective of Reducing Carbon Emissions

Xinhua Zhu 1,2, Nan Li 3 , Yu Sun 3, Hongfei Zhang 4, Kai Wang 5,* and Sang-Bing Tsai 5,6,*

1 College of Economics and Management, Civil Aviation University of China, Tianjin 300300, China; [email protected] 2 College of Civil Aviation, Nanjing University of Aeronautics and Astronautics, Nanjing 211106, China 3 College of Air Traffic Management, Civil Aviation University of China, Tianjin 300300, China, [email protected] (N.L.); [email protected] (Y.S.) 4 Operation and Control Center, Shandong Airlines, Jinan 250014, China; [email protected] 5 College of Business Administration, Capital University of Economics and Business, Beijing 100070, China 6 Zhongshan Institute, University of Electronic Science and Technology of China, Zhongshan 528400, China * Correspondence: [email protected] (K.W.); [email protected] (S.-B.T.)

 Received: 2 August 2018; Accepted: 13 September 2018; Published: 17 September 2018 

Abstract: In order to reduce the time of departing aircraft and reduce the fuel consumption and exhaust emissions of the aircraft, Shanghai Hongqiao was taken as an example to study the control strategy for aircraft departure. In this paper, the influence of the number of departure aircraft on the runway utilization rate, the takeoff rate, and the departure rate of flight departures under the conditions of airport runway capacity constraints are studied. The influence of factors, such as the number of departure aircraft, the position of the aircraft, and the configuration of airport arrival and departure runways, on the aircraft taxiing time for departure is analyzed. Based on a multivariate linear regression equation, a time prediction model of aircraft departure taxiing time is established. The fuel consumption and pollutant emissions of aircraft are calculated. The experimental results show that, without reducing the utilization rate of the runway and the departure rate of flights, implementing a reasonable pushback number for control of departing aircraft during busy hours can reduce the departure taxiing time of aircraft by nearly 32%, effectively reducing the fuel consumption and pollutant emissions during taxiing on the airport surface.

Keywords: aircraft; taxi time; takeoff rate; pushback control; green transportation; carbon emissions; reducing carbon emissions

1. Introduction With the rapid development of air transportation, the number of flights at major in China has been increasing, making airport surface runways congested. This, in particular, causes the aircraft on airport surfaces to take a long time to taxi, and an excessive number of flights to wait in line at the entrance to the runway. Due to premature pushback of aircraft and waiting on crowded taxiways, an additional 10–20 kg of fuel consumption is added for each additional minute of taxi time [1,2], resulting in an increase in aircraft exhaust emissions affecting the air quality around the airport. A series of studies have been conducted by scholars at domestic and foreign terminals to control the number of aircraft departing from airport surfaces and reduce the congestion time in aircraft taxiing. In 2007, Balakrishnan H and Jung Y [3] studied the airport surface operation of Dallas–Fort Worth airport by establishing an integer programming model. This study shows that using the method of delaying the pushback time of departing aircraft can reduce the number of airport surface

Energies 2018, 11, 2473; doi:10.3390/en11092473 www.mdpi.com/journal/energies Energies 2018, 11, 2473 2 of 15 taxiing aircraft and reduce congestion, thereby reducing the average taxiing time of departing aircraft, and using airport surface aircraft taxiway optimization methods can significantly reduce the waiting time for aircraft crossing the runway, thereby reducing the average taxiing time of arriving aircraft. In 2009, Simaiakis I et al. [4] analyzed the key factors affecting the taxi time of departing aircraft. Taking the Boston International Airport (BOS) as an example, a forecasting analysis of the departure taxiing time of aircraft was made, and a queuing theory model based on the departure process of aircraft was proposed. Using this queue-pushback strategy for departing aircraft can reduce the departure taxiing time and, thus, reduce aircraft pollutant emissions. In 2010, Jung YC et al. proposed the Spot Release Planner (SRP) and Runway Scheduler (RS) [5]. The SRP aims to reduce an aircraft’s taxiing time by keeping the runway productivity at the maximum level by sorting the order of departure of the departing aircraft on the apron and controlling the pushback time of each aircraft to control the time that the aircraft enters the maneuvering area. The RS is designed to sort and time-allocate departing aircraft and arriving flights across the takeoff runway to achieve maximum runway utilization. Jung YC et al. combined the proposed two strategies to optimize the operation of busy airport surfaces. In 2010, Lee H [6] proposed two ways to optimize the operation of airport surfaces: delayed pushback and path optimization. Delayed pushback refers to the control of off-board aircraft that is applied during an airport congested period to control the airport’s congestion. The path optimization optimizes the taxi path of all aircraft at the airport based on the delayed pushback. With the application of airport surface monitoring equipment, it makes it possible to analyze the airport surface trajectory of the aircraft in detail using the airport surface monitoring data. In 2011, I. Simaiakis et al. [7] applied Airport Surface Detection Equipment Model-X (ASDE-X) data from the monitoring equipment at Boston International Airport. They considered the impact of the airport runway configuration, the different type series in the fleet, meteorological conditions, and other factors on the airport runway capacity. A study on aircraft pushback rate control at the airport was conducted. In 2013, S Ravizza [8] and others calculated the required taxiing distance, the total steering angle, the type of departing and arriving aircraft, the number of aircraft in operation on the airport surface, the usage configuration of departure and arrival runways, and the position of the gate. The establishment of a multiple linear regression model helped to provide a more accurate prediction of the aircraft into and out of the required taxi time. In 2015, Tang Y [9] elaborated on the concept of the Advanced Surface Movement Guidance and Control System (A-SMGCS) proposed by International Civil Aviation Organization (ICAO) in his doctoral dissertation and conducted a comprehensive study on an aircraft’s initial taxi route planning, real-time optimization of aircraft taxiing routes, and A-SMGCS three-dimensional (3D) simulation. In addition, Xiangling Z et al. [10] studied the issue of virtual pushback queues for departing flights at the gate position and the issue of decision-making for collaborative pushback of departing flights. In 2016, Nan L and Hongzhe L [11] analyzed the surveillance data of Hongqiao Airport, used support vector machines to classify and determine the trajectory of taxi aircraft, and applied data mining technologies to the prediction of airport surface aircraft taxi time, the determination of airport surface taxi hotspots, and conflict zone determination. Under the same runway configuration conditions, as the number of aircraft pushback into the apron and taxi systems increases, more flights are added to the takeoff queue, resulting in a gradual increase in runway utilization and departures from flights. However, due to the effect of aircraft wake spacing, when a certain number of taxiing departing aircraft is reached, the runway capacity becomes the limiting factor and the number of taxiing departing aircraft will continue to increase. The runway utilization rate and takeoff and departure rate of flights will only tend to change smoothly. Therefore, in practice, the tower controllers are more concerned with a reasonable number of departing aircraft operations in a given runway usage configuration. This paper uses the airport surface monitoring data from Shanghai Hongqiao Airport to study the influence of the number of different departing aircraft within the apron and taxi systems on the takeoff and departure rate of the flights, the departure taxiing time, and the runway utilization rate under runway capacity constraints. A departure time prediction model for departing aircraft is established. A reasonable control strategy is implemented for departing Energies 2018, 11, 2473 3 of 15 aircraft within the busy airport departure period without reducing the operating efficiency of the runway, thereby reducing the aircraft departure taxiing time and reducing aircraft fuel consumption and pollutant emissions.

2. Airport Surface Operation Data Analysis and Definition

2.1. Airport Surface Monitoring Data Analysis While the airport surface monitoring system assists controllers in performing control services more safely and efficiently, the equipment also records real-time aircraft trajectory data. The aircraft movement trajectory data recorded by the airport surface monitoring system includes a time stamp and the aircraft’s position, altitude, speed, and so forth, and by combining these with the airport topology data, we can identify the operational status of the aircraft, such as its pushback, taxiing, takeoff, and landing; calculate parameters, such as taxi distance and taxi time; identify the taxi path; and provide a data foundation for studying aircraft airport surface operation and optimization [12,13]. Through the data analysis of the Shanghai Hongqiao Airport’s March 2015 airport surface monitoring system, we have sorted out the full arrival and departure trajectories for flights when using the 18 L/18 R configuration on the departure and arrival runway (18 L runway for approach and 18 R runway for departure calculated as two complete tracks during the stop-and-depart process).

2.2. Airport Surface Operation Data Definition Aircraft departure taxiing refers to the entire process of the pushback of the aircraft from the gate position, the taxiing to the departure runway, and the wait for takeoff. In order to analyze the operation of airport surfaces, this paper gives the following definitions of the quantitative indicators for measuring airport surface operations: 1. Departure taxiing time: the total time of the aircraft’s pushback from the parking position to the time of taxiing to the departure runway, including the aircraft’s pushback, apron and taxiway taxiing, and the holding time the entrance of the runway (unit: minute). 2. Number of airport surface aircraft: the total number of departing and arriving aircraft taxiing (including taxi wait) or undergoing pushback in the apron and taxiway systems (unit: flight). 3. Number of departure aircraft: the number of departing aircraft that are taxiing (including the taxi wait) or undergoing pushback in the apron and taxiway systems (unit: flight). 4. Runway utilization rate: the ratio of the length of time the runway is accumulatively occupied over a period of time to the length of the time period of the calculation. 5. Takeoff and departure rate of flights: the number of departing aircraft per unit of time (unit: flights/minute). When calculating the runway utilization rate over a period of time, first of all, it is necessary to find out the total time taken for the runway to be occupied during this period. In general, the time spent on the following operations of the aircraft is accumulated into the occupied time of the runway.

6. Takeoff running: the duration from the point where a departing aircraft accelerates for takeoff on the runway until the aircraft’s tires are off the ground; 7. Departure waiting: the departing aircraft waits for takeoff clearance on the runway; 8. Final Approach: the duration from an arriving aircraft’s being at its final approach phase of 2.5 nm (nautical miles) from the runway’s end to the landing of the aircraft on the runway; 9. Landing: the duration of an arriving aircraft’s starting to land to when it is off the runway; 10. Cross-taxi: the aircraft crosses over the runway.

For an airport with only one runway, when calculating the runway utilization rate of the airport, the cumulative time of the five operations (6)–(10) of the aircraft needs to be taken into account as the occupied time of the runway. Energies 2018, 11, 2473 4 of 15

Energies 2018, 11, x FOR PEER REVIEW 4 of 15 ForEnergies airports 2018, 11, with x FOR multiplePEER REVIEW runways, if aircraft arrivals and departures run on different4 runways,of 15 the time takenForFor airports airports on the with with departure multiple multiple runways, runway runways, willif aircraft only arrivalsarrivals take into and and departures accountdepartures the run run aircrafton on different different operations runways, runways, (6), (7), and (10),thethe time and time taken only taken theon on the aircraftthe departure departure operations runwayrunway (8)–(10) willwill only need tatakeke tointointo be account account considered the the aircraft aircraft when operations theoperations approach (6), (6), (7), runway(7), is occupied.andand (10), (10), and and only only the the aircraft aircraft operations operations (8)–(10) needneed toto be be considered considered when when the the approach approach runway runway is occupied.is occupied. 3. Runway Capacity Analysis 3. Runway Capacity Analysis 3. Runway Capacity Analysis This sectionThis section applies applies the the airport airport surveillance surveillance data to to analyze analyze thethe actual actual airport airport operational operational data. data. This section applies the airport surveillance data to analyze the actual airport operational data. HongqiaoHongqiao Airport Airport is a narrow-distance,is a narrow-distance, dual-runway dual-runway airport; airport; 18 18L is L mainly is mainly used usedfor arrival, for arrival, 18 R is 18 R is mainlyHongqiaomainly used for used Airport departure, for departure, is a narrow-distance, and and its its taxiway taxiway dual-runway system system network airport; is isshown 18 shown L is in mainly Figure in Figure used1. 1 .for arrival, 18 R is mainly used for departure, and its taxiway system network is shown in Figure 1. 8 18R 18L 36R 8 18R 18L 36R 36L 36L

No. 2 apron No. 2 apron Figure 1. Network diagram of the Shanghai Hongqiao Airport taxiway system. Figure 1. Network diagram of the Shanghai Hongqiao Airport taxiway system. 3.1. Runway3.1. Runway Utilization UtilizationFigure Rate 1. Network Rate Analysis Analysis diagram of the Shanghai Hongqiao Airport taxiway system. This paper analyzes the main runway configuration of Hongqiao Airport, namely the 18 L This3.1. Runway paper Utilization analyzes Rate the Analysis main runway configuration of Hongqiao Airport, namely the 18 L approach and the 18 R departure configuration. Figure 2 shows the relationship between the approachutilizationThis and paper the of the 18analyzes departure R departure the 18 main R runway configuration. runway and theconfigur numb Figureationer of departingof2 showsHongqiao aircraft the Airport, relationshipon the airportnamely surface. between the 18 L the utilizationapproachThe ofcalculation theand departure the of the18 runwayR 18departure R runwayutilization configuration. and rate the is based number Figure on a statistical of 2 departingshows period the aircraft ofrelationship every on 15 themin, between airportbecause the surface. The calculationutilizationthe number of of the of the departingdeparture runway aircraft 18 utilization R runway on the rate andairport isthe based sunumbrface oner changes of a statisticaldeparting slightly aircraft period every 15on of minthe every airportand 15 this min, surface. can because the numberTheavoid calculation ofstatistical departing of errors the runway aircraft caused utilizationby on excessive the airport rate changes is based surface in the on numbera changes statistical of departing slightly period of everyaircraft. every 15 15Additionally, minmin, andbecause this can avoidthe statisticala number15 min statistical errorsof departing caused period aircraft is by not excessive tooon shortthe airport changesand the su averagerface in the changes runway number slightly utilization of departing every rate 15 canaircraft. min be calculatedandAdditionally, this can during this period so as to avoid statistical errors due to excessive statistical fluctuations caused by a 15 minavoid statistical statistical perioderrors caused is not by too excessive short and changes the average in the number runway of departing utilization aircraft. rate canAdditionally, be calculated a 15the min statistical statistical time period being toois not short. too short and the average runway utilization rate can be calculated during this period so as to avoid statistical errors due to excessive statistical fluctuations caused by the during this period so as to avoid statistical errors due to excessive statistical fluctuations caused by statistical time being too short. the statistical time being too short.

Figure 2. The relationship between runway utilization and the number of departing aircraft.

FigureFigure 2. The 2. The relationship relationship between between runway runway utilizationutilization and and the the number number of departing of departing aircraft. aircraft. Energies 2018, 11, 2473 5 of 15 Energies 2018, 11, x FOR PEER REVIEW 5 of 15

The horizontal axisaxis inin FigureFigure2 shows2 shows the the number number of departureof departure aircraft aircraft on theon airportthe airport surface surface and andthe verticalthe vertical axis axis shows shows the runwaythe runway utilization utilization rate rate of18 of R.18 ItR. can It can be seenbe seen that that when when the the number number of ofdeparture departure aircraft aircraft on theon airportthe airport surface surface is N ≤is 6,N the≤ 6, number the number of aircraft of aircraft departing departing the airport the surfaceairport surfaceat this time at this is small,time is and small, the and runway the runway utilization utilization rate of the rate departure of the departure runway 18runway R increases 18 R increases with the withnumber the ofnumber airport of surface airport departure surface departure aircraft. When aircra theft. When number the of number airport of surface airport departure surface departure aircraft is aircraftN = 7, the is N runway = 7, the utilization runway utilization rate is already rate closeis already to 1, indicatingclose to 1, thatindicating under thethat operating under the scale operating of the scalenumber of ofthe departing number aircraft,of departing the 18 Raircraft, departure the runway18 R departure has been basicallyrunway usedhas been efficiently. basically When used the efficiently.number of When departure the number aircraft inof operationdeparture onairc theraft airportin operation surface on is the N ≥airport8, the surface runway is utilizationN ≥ 8, the runwayrate fluctuates utilization at a rate value fluctuates close to at 1, a indicating value close that to the1, indicating departure that runway the departure 18 R is continuing runway 18 to R beis continuingused efficiently. to be used efficiently.

3.2. Analysis Analysis of Flight Takeoff Rate Figure3 3 showsshows thethe takeofftakeoff raterate ofof flightsflights fromfrom departuredeparture runwayrunway 1818 RR asas aa functionfunction ofof thethe numbernumber of departure aircraft on the airport surface. The takeoff rate statistics for the runway are also based on a 15 15 min min statistical statistical period, period, and and then then the the takeoff takeoff rate rate average average of of each each departing departing aircraft aircraft quantity quantity is calculated.is calculated.

FigureFigure 3. 3. TheThe relationship relationship between between the the takeoff rate and the number of departure aircraft.

When the number of departing aircraft is N ≤ 6,6, the the number number of of departure departure aircraft aircraft on the airport surface atat thisthis time time is is small, small, and and the the number number of arriving of arriving and departing and departing flights perflights unit per of timeunit increases of time increaseswith the increasewith the in increase the number in the of number departure of aircraftdeparture on aircraft the airport on the surface. airport When surface. the numberWhen the of numberdeparture of aircraft departure is N aircraft = 7, the is takeoff N = rate7, the of flightstakeoff hasrate reached of flights its maximum.has reached Correspondingly, its maximum. Correspondingly,from Figure2, it can from also Figure be seen 2, it that can the also runway be seen utilization that the raterunway is close utilization to 1, indicating rate is close that to the 1, indicatingarrival and that departure the arrival of theand flights departure began of tothe be flights restricted began by to thebe restricted runway capacity; by the runway the arrival capacity; and thetakeoff arrival rates and of flightstakeoff reach rates 0.52 of flights flights/min reach (310.52 flights/h). flights/min When (31 theflights/h). number When of departure the number aircraft of departureon the airport aircraft surface on the is Nairport≥ 9, thesurface takeoff is N rate ≥ 9, ofthe flights takeoff calculated rate of flights by the calculated statistics by shows the statistics a slight showsdownward a slight trend. downward However, trend. this does However, not simply this indicatedoes not that simply the takeoffindicate rate that will the decrease takeoff whenrate will the decreasenumber ofwhen departure the number aircraft of on departure the airport aircraft surface on isth largee airport in correspondence surface is large within correspondence Figure2. When with the Figureperformance 2. When is Nthe≥ performance9 compared to is NN =≥ 7–8,9 compared the runway to N utilization = 7–8, the is runway also close utilization to 1, indicating is also close that the to 1,runway indicating is still that fully the utilized. runway However, is still fully the utilized. percentage However, of waiting the timepercentage for the aircraftof waiting at the time runway’s for the aircraftend increases at the runway’s when N ≥end9 inincreases Figure 2when compared N ≥ 9 in with Figure the case2 compared when N with = 7–8, the indicatingcase when thatN = 7–8, the indicatingdecrease in that the the takeoff decrease rate isin due the totakeoff the increase rate is du ine the to the holding increase time in for the takeoff holding at time the runway’s for takeoff end. at theThrough runway’s further end. analysis Through of thefurther arrival analysis and departure of the arrival flight and data, departure it was found flight that data, because it was found the double that becauserunways the of Hongqiaodouble runways Airport of are Hongqiao narrow parallelAirport runways,are narrow the parallel arriving runways, and departing the arriving aircraft and of departingthe two runways aircraft of cannot the two take runways off and cannot land attake the off same and land time at because the same of time the aircraftbecause wake of the spacing.aircraft wake spacing. This means that the two runways cannot operate independently. Therefore, the continuous arrival of multiple aircraft will affect the aircraft taking off, resulting in a decrease in the Energies 2018, 11, 2473 6 of 15

This means that the two runways cannot operate independently. Therefore, the continuous arrival of multiple aircraft will affect the aircraft taking off, resulting in a decrease in the takeoff rate of the runway 18 R and a cumulative increase in the number of departure aircraft on the airport surface, such as the number of departure aircraft on the airport surface reaching 12. This situation is reflected in the performance of N ≥ 9 in Figures2 and3. Although the takeoff rate of flights has decreased, the percentage of time that an aircraft is waiting to take off at the runway entrance has increased in the runway utilization histogram. Figures2 and3 can be summarized as follows: in the fixed runway configuration, the number of departure aircraft that undergo pushback into the apron and taxiway systems increases as more flights are added to the takeoff queue. The runway utilization rate and takeoff rate have gradually increased. However, aircraft takeoff will be limited by the time interval. When the number of departure aircraft on the airport surface reaches a critical value or if the number of departure aircraft on the airport surface continues to increase, the runway capacity will become a limiting factor. The runway utilization rate and the takeoff rate will only tend to have stable fluctuations and will no longer increase significantly.

4. Aircraft Departure Time Prediction

4.1. Factors Affecting Departure Taxiing Time Analysis of the departure taxiing process shows that the aircraft departure taxiing time is related to the airport current runway configuration, the gate of the aircraft apron, and the congestion status of the departure taxiway through the apron and the taxiway systems. Under certain runway usage configurations, the influence of the gate of the apron on the taxi time of the departing aircraft can be expressed by a taxi distance parameter. The influence of the apron and taxiway systems’ congestion conditions on the taxi time of departing aircraft can be expressed by the parameter of the number of aircraft on the airport surface. Through the data analysis of the Shanghai Hongqiao Airport March 2015 airport surface monitoring system, we select the flights in UTC time 04:00–06:00 (Local Time 12:00–18:00), which is the busy time. The full sample contains 1469 departure flights and 1399 arrival flights. The data in Table1 summarizes the statistics for the departure flight sample.

Table 1. Summary of Statistics for the departure flight Sample.

Standard Mean Median Max. Min. Deviation the Departure Taxiing Time (Minutes) 16.32 14.19 7.05 30.94 2.73 the Number of the Departure Aircraft (Flights) 5.90 6 2.55 11 1 the Taxiing Distance (Meters) 2232.37 2278 458.06 3055 1005

4.2. Impact of the Number of Aircraft on the Airport Surface Figure4 is a scatter plot of the departure taxiing time of aircraft and the number of aircraft on the airport surface when each aircraft underwent pushback on 9 March 2015, UTC time 4:00–6:00. From the scatter plot, it can be clearly seen that the departure taxiing time of aircraft gradually increases with the increase in the number of aircraft on the airport surface. Energies 2018, 11, 2473 7 of 15 Energies 2018, 11, x FOR PEER REVIEW 7 of 15 Energies 2018, 11, x FOR PEER REVIEW 7 of 15

35 35

) 30 ) 30 25 25 20 20 15 15 10 10 5

the departure taxiing time (min taxiing the departure 5

the departure taxiing time (min taxiing the departure 0 0 0246810121416 0246810121416 the number of the aircraft the number of the aircraft

FigureFigure 4. 4.The The scatter scatter diagram diagramof ofthe the relationshiprelationship betw betweeneen the the departure departure taxiing taxiing time time and and the the number number Figure 4. The scatter diagram of the relationship between the departure taxiing time and the number ofof aircraft. aircraft. of aircraft. ConsideringConsidering that that there there are are large large differencesdifferences in th thee taxi taxi path path between between the the arriving arriving and and departing departing Considering that there are large differences in the taxi path between the arriving and departing aircraft,aircraft, the the number number of of arriving arriving aircraft aircraft hashas little in influencefluence on on the the taxiing taxiing time time of ofthe the departing departing aircraft. aircraft. aircraft, the number of arriving aircraft has little influence on the taxiing time of the departing aircraft. Therefore,Therefore, all all the the arriving arriving aircraft aircraft in in the the statistical statistical data data of of Figure Figure4 4are are excluded excluded and and the the scatter scatter plotplot of Therefore, all the arriving aircraft in the statistical data of Figure 4 are excluded and the scatter plot theof departure the departure taxiing taxiing time time of aircraftof aircraft and and the the number number of of departing departing aircraft aircraft on on thethe aircraftaircraft surface is isof obtainedthe departure in Figure taxiing 5. time of aircraft and the number of departing aircraft on the aircraft surface obtainedis obtained in Figure in Figure5. 5.

35 35

) 30

) 30 25 25 20 20 15 15 10 10 5 5

the departure taxiing time (min taxiing the departure 0 the departure taxiing time (min taxiing the departure 0 024681012 024681012 the number of the departure aircraft the number of the departure aircraft

Figure 5. The scatter diagram of the relationship between the departure taxiing time and the number Figure 5. The scatter diagram of the relationship between the departure taxiing time and the number Figureof departing 5. The scatteraircraft. diagram of the relationship between the departure taxiing time and the number ofof departing departing aircraft. aircraft. The linear correlation coefficient between the aircraft departure taxiing time and the number of The linear correlation coefficient between the aircraft departure taxiing time and the number of aircraftThe linearon the correlation airport surface coefficient is R1 = between 0.62; the the linear aircraft correlation departure coefficient taxiing between time and the the aircraft number aircraft on the airport surface is R1 = 0.62; the linear correlation coefficient between the aircraft ofdeparture aircraft on taxiing the airport time and surface the number is R1 = of 0.62; depa therture linear aircraft correlation on the airport coefficient surface between is R2 = the 0.79. aircraft By departure taxiing time and the number of departure aircraft on the airport surface is R2 = 0.79. By departurecomparing taxiing R1 and time R2, andit is theshown number that the of departureaircraft departure aircraft taxiing on the time airport has surface a stronger is R 2linear= 0.79. comparing R1 and R2, it is shown that the aircraft departure taxiing time has a stronger linear Byrelationship comparing with R1 and the Rnumber2, it is shown of departure that the aircraft aircraft on departure the airport taxiing surface time at the has time a stronger of aircraft linear relationshippushback.relationship withTherefore, with the the number thenumber number of departure of departure of departure aircraft aircraft onairc the rafton airport theon airportthe surface airport surface at thesurface timeat the can of time aircraft be usedof pushback.aircraft as a Therefore,predictorpushback. the variable Therefore, number of the ofthe departureaircraft number departure aircraftof departure taxiing on the airctime. airportraft on surface the airport can be usedsurface as can a predictor be used variable as a predictor variable of the aircraft departure taxiing time. of the aircraft departure taxiing time. Energies 20182018,, 1111,, 2473x FOR PEER REVIEW 88 of 1515

4.3. Effect of Aircraft Departure Taxiing Distance 4.3. Effect of Aircraft Departure Taxiing Distance The effect of the airport runway configuration and an aircraft’s gate position on the aircraft's departureThe effect taxiing of thetime airport is reflected runway in the configuration distance required and an for aircraft’s the aircraft gate to position taxi from on thethe aircraftapron to's departurethe takeoff taxiingrunway time entrance. is reflected Figure in 6 theis the distance scatter requiredplot of the for aircraft the aircraft departure to taxi taxiing from time the apron and the to therequired takeoff taxi runway distance entrance. for departure Figure6 onis the9 March scatter 2 plot015, ofUTC the time aircraft 4:00–6:00. departure From taxiing the scatter time and plot, the it requiredcan be seen taxi distancethat the fordeparture departure taxiing on 9 Marchtime incr 2015,eases UTC gradually time 4:00–6:00. with Froman increase the scatter in the plot, taxiing it can bedistance. seen that the departure taxiing time increases gradually with an increase in the taxiing distance.

35

30

25

20

15

10

5 the departure taxiing time(min) taxiing the departure

0 0 500 1000 1500 2000 2500 3000 3500 the taxiing distance(m)

Figure 6. The scatter diagram of the relationship between the flightflight taxi time and the taxi distance.

In apron and taxiway systems, the aircraft generallygenerally taxis at a low and and uniform uniform speed. speed. Therefore, Therefore, under anan idealideal no-taxiing-collision no-taxiing-collision condition, condition, when when the the aircraft aircraft is taxiing is taxiing at a constant at a constant speed, thespeed, taxiing the timetaxiing is positivelytime is positively related torelated the taxiing to the distance.taxiing di Thestance. linear The correlation linear correlation coefficient coefficient between between the aircraft the departureaircraft departure taxiing timetaxiing and time departure and departure taxiing distancetaxiing distance is R3 = 0.77.is R3 This= 0.77. indicates This indicates that the that aircraft the departureaircraft departure taxiing timetaxiing and time the departureand the departure taxiing distance taxiing havedistance a strong have linear a strong relationship. linear relationship. Therefore, theTherefore, aircraft the departure aircraft taxiing departure distance taxiing can distance be used ascan a predictorbe used as of aircrafta predictor departure of aircraft taxiing departure time. taxiing time. 4.4. Departure Taxiing Time Prediction Model 4.4. DepartureThe aircraft Taxiing departure Time Prediction taxiing time Model T can be divided into two parts: The aircraft departure taxiing time T can be divided into two parts: T = ttw + tc, (1) = + T ttw tc , (1) where ttw represents the time taken by the aircraft to taxi from the apron to the departure runway where 𝑡 represents the time taken by the aircraft to taxi from the apron to the departure runway without conflict, and the magnitude of the value is related to the taxiing distance d; and tc indicates the amountwithout ofconflict, time spent and the escaping magnitude and waiting of the value for each is related aircraft to during the taxiing the taxiing distance process d; and due 𝑡 toindicates mutual influence.the amount The of magnitudetime spent escaping of the value and reflects waiting the for degree each aircraft of airport during congestion the taxiing and isprocess related due to the to numbermutual influence. (N) of aircraft The departuresmagnitude onof the the value airport reflects surface. the degree of airport congestion and is related to theFrom number the analysis(N) of aircraft in the previousdepartures section, on the the airport departure surface. taxiing time of aircraft is linearly related to theFrom number the analysis of aircraft in the departing previous the section, airport the surface departure and taxiing the departure time of taxiingaircraft distance.is linearly Tablerelated2 showsto the number the correlation of aircraft data. departing According the toairport the correlation surface and coefficient the departure r of the taxiing independent distance. variable, Table 2 whenshows rthe is closecorrelation to 1, there data. isAccording a strong linearto the relationshipcorrelation coefficient between the r of two the independentindependent variables.variable, Itwhen represents r is close only to a1, judgment there is a on strong collinearity linear betweenrelationship two between independent the two variables. independent Therefore, variables. multiple It linearrepresents regression only a models judgment could on becollinearity used to predict between the two aircraft independent departure variables. taxiing time. Therefore, multiple linear regression models could be used to predict the aircraft departure taxiing time. Energies 2018, 11, 2473 9 of 15

Table 2. Correlation Data.

the Number of the Taxiing Distance Departure Aircraft the Number of Departure Aircraft 1 0.66551 the Taxiing Distance 0.66551 1

The multiple linear regression equation can be expressed as

y = m(x1, x2,..., xp) + ε, (2)

Since the linear regression assumes that m(x1, x2, ... , xp) is a linear function of the random variables (x1, x2, ... , xp), in this paper, the aircraft departure taxiing time T is linearly related to the number of departure aircraft on the airport surface N and the departure taxiing distance d. Therefore, the multivariate linear regression equation expression of the aircraft departure taxiing time prediction model can be expressed as T = β0 + β1N + β2d + ε, (3)

In this formula, β0, β1, and β2 are the linear regression coefficients to be solved. For convenience of description, Equation (3) is represented by the matrix expression below (Equation (4)):

Y = Xβ + ε, (4)

To ensure correct statistics, it is usually necessary to make multiple observations on the independent variable and the dependent variable corresponding to the independent variable. Assume that the observation statistics are performed n times, where Y and ε are n-dimensional column vectors, β is a (p + 1)-dimensional column vector, and the independent variable X is an n × (p + 1)-dimensional matrix whose first column is all 1. Additionally, take p = 2 corresponding to Equation (3). In order to obtain the best-estimated vector parameter β, we make the sample Xβ estimation as close as possible to the observed value Y, making the error term ε as small as possible. Using least −1 squares estimation, we can see that when β = XT X XTY, the square of ε-mode

kεk2 = (Y − Xβ)T(Y − Xβ) !2 n p (5) = ∑ yi − β0 − ∑ xijβj i=1 j=1 reaches the minimum, so the best linear unbiased estimate is

−1 βˆ = (XT X) XTY, (6)

2 The adjusted coefficient of determination RAdj can be used to measure how well the model fits the data. The expression is as follows:

n 2 ∑ (yˆi − y) /(n − p − 1) 2 i=1 RAdj = 1 − n , (7) 2 ∑ (yi − y) /(n − 1) i=1

In this formula, n is the number of observations, that is, the number of departure aircraft counted; yi is each observation value of the dependent variable, that is, the ith aircraft departure taxiing time; y is the average of the dependent variable observations, that is, the average departure taxiing time of the aircraft calculated; and yˆi is the estimated value of the dependent variable for each observation, that is, Energies 2018, 11, 2473 10 of 15

2 the multiple linear regression model prediction of the departure time of the ith aircraft. RAdj values between 0 and 1, with values closer to 1 indicating a better fit [14]. To sum up, we assume that the number of airport surface departure aircraft is N and the required departure taxiing distance is d. Equations (3) and (6) can then be used to obtain the fitting prediction formula for the aircraft departure taxiing time T as

T = −8 + 1.35N + 6.9d, (8)

Table3 shows the parameters of the multiple linear regression. By Equation (7), the goodness of fit is described by an Adjusted R2 = 0.835, which shows that the aircraft departure taxiing time prediction model is reasonable.

Table 3. The tables of the parameters for regression.

Variable Coefficient t-Statistic Sig C −8.078409 −2.81201 ** N 1.353452 4.748534 *** d 6.958 4.259463 *** Sig. indicates if the p-value is 0.05 (*), 0.01 (**), or 0.001 (***).

5. Pushback Strategy for Departure on the Airport Surface

5.1. Implementation of Control Strategies Taking the departing flight of the No. 2 apron of Hongqiao Airport shown in Figure1 as an example, data analysis is conducted to compare the flight departure taxiing time, the runway utilization rate, and the takeoff rate under different departure aircraft numbers on the airport surface. When the airport runway configuration is 18 L/18 R, the average departure taxiing distance of the No. 2 apron departing flight is 2.25 km, and we substitute the taxiing distance into Equation (8). Thus, the correspondence between the departure taxiing time of the flight and the number of departure aircraft on the airport surface can be obtained. Then, the corresponding relationship between the runway utilization rate, the flight takeoff and departure rate, and the number of departure aircraft on the airport surface can be calculated according to the statistics of Figures2 and3. Thus, Table4 can be obtained.

Table 4. The operating parameters for different numbers of departure aircraft.

Airport Surface Takeoff and Departure Aircraft Taxi Time (min) Departure Rate Runway Utilization (%) Number (N) (flight/min) 1 8.87 0.11 32.0 2 10.2 0.21 47.5 3 11.6 0.29 61.5 4 12.9 0.37 72.1 5 14.3 0.43 83.3 6 15.6 0.49 94.0 7 16.9 0.53 98.9 8 18.3 0.53 97.3 9 19.7 0.50 97.0 10 21.0 0.44 98.5 11 22.3 0.42 100

According to Table4, when the number of aircraft departures on the airport surface is N ≤ 6, the number of departure aircraft is small and the flight departure taxiing time, runway utilization rate, and takeoff and departure rate of flights increase with an increase in the number of aircraft Energies 2018, 11, 2473 11 of 15 departures on the airport surface. When the number of aircraft departures on the airport surface is N = 7, the average departure taxiing time of No. 2 apron flights is 16.9 min, the runway utilization rate begins to be close to 1, the takeoff and departure rate is 0.53 flight/min, and the flight departure process begins to be controlled by the runway capacity limits. When the number of airport surface departure aircraft is N ≥ 8, the runway utilization rate and the flight takeoff and departure rate no longer increase, while the average taxi time of departure flights on the No. 2 apron no longer increases. Therefore, when the number of airport surface departure aircraft is N ≤ 7, the tower aircraft controller can act in accordance with the first-come first-service (FCFS) principle based on the pushback request clearance issued by the flight. When the number of airport surface departure aircraft is N ≥ 8, the controller can first control the aircraft pushback at the gate position, and then set up a virtual pushback sequence for these departure flights. When N ≤ 7, the departure flights will be queued according to the virtual pushback sequence. The implementation of the departure control strategy for the aircraft did not reduce the runway utilization rate and flight takeoff and departure rate, but it slowed the airport surface congestion so that the departure taxiing time can be effectively reduced without increasing the total delay in departure flights.

5.2. SIMMOD Simulation The SIMMOD software was used to simulate the use of the departure control strategy. SIMMOD is a discrete-time simulation software released by the U.S. Federal Aviation Administration. SIMMOD provides dynamic decisions based on user-defined rules, and each process of a flight is controlled based on user rules. Its performance indicators are: the flight time of the aircraft, the capacity per unit of time, delays, etc. [15–19]. The SIMMOD simulation model relies mainly on a detailed description of the airport and airspace network, and the traffic flow moves on the nodes and connections of the network. The operating path of the aircraft can be specified either by the user or automatically by the Dijkstra Algorithm [20–23]. First, the Computer Aided Design (CAD) map of Hongqiao Airport is imported into SIMMOD to establish the airport topology map and the waypoints are inputted to establish the arrival and departure procedures and routes. Then, according to the flight plan of on 9 March 2015, UTC time 4:00–6:00 (local time 12:00–14:00), the arrival and departure flight information is established. The departure control strategy simulation is implemented. The simulation results show that the average wait time of the 52 departure flights was 2.24 min, the total taxi time was decreased by 119 min, and the maximum waiting time for the gate was 14 min. As shown in Figure7, the blue line represents the case where the departure control is not used, and the red line represents the case where the departure control is used. It can be clearly seen from the figure that by the use of the departure control strategy, the total departure taxi time is reduced. The taxi time predicted in Section 5.1 by the Section4 multiple linear regression prediction method is basically consistent with the taxi time obtained by the Section 5.2 SIMMOD simulation Energies 2018, 11, 2473 12 of 15 Energies 2018, 11, x FOR PEER REVIEW 12 of 15

the average departure taxi time Comparison

) 25 (per 15-min interal)

20

15

10 the actual departure taxiing time

5 the taxiing time of the departure control strategies 0 the departure taxiing time (min 12:00 12:15 12:30 12:45 13:00 13:15 13:30 13:45 14:00 the local time at start of taix

FigureFigure 7. TheThe scatter scatter diagram diagram of of the the relationship between the flight flight taxi time and the taxi distance.

5.3. Analysis Analysis of Fuel Saving and Emission Reduction Data During the busy hours of airport operation, th thee number of departure ai aircraftrcraft on the airport surface was controlled at N = 7. Compared Compared with N = 9–11, 9–11, the the runway runway utilization utilization rate rate and and the the takeoff takeoff and departure rate did not change substantially; substantially; ho however,wever, the average departure taxiing time of each aircraft onon the the No. No. 2 apron2 apron decreased decreased by 2.8–5.4 by 2.8–5.4 min, accountingmin, accounting for 16.7–31.9% for 16.7–31.9% of the aircraft of the departure aircraft departuretaxiing time taxiing at this time time. at this time. Taking the CFM56-5B4/P CFM56-5B4/P engine engine of of the the Airbus Airbus A320 A320 as asan an example, example, the the total total amount amount of fuel of fuel oil andoil and pollutant pollutant gas gasemissions emissions consumed consumed for foreach each departing departing aircraft’s aircraft’s taxiing taxiing on onthe the No. No. 2 apron 2 apron is calculatedis calculated when when the the aircraft aircraft impl implementsements different different departure departure control control strategies strategies [24,25] [24,25 as] asshown shown in Tablein Table 5. 5.

Table 5. The total fuel consumption and total pollutant emissions of departing aircraft under different Table 5. The total fuel consumption and total pollutant emissions of departing aircraft under different departure control strategies. departure control strategies. Airport Surface Departure Fuel Consumption (kg) Pollutants Total Emissions (kg) AirportAircraft Surface Number (N) Departure Aircraft6 Fuel Consumption 194.6 (kg) Pollutants Total 6.28 Emissions (kg) Number (N)7 210.9 6.81 6 8194.6 228.4 7.37 6.28 9 245.9 7.94 7 10210.9 262.1 8.46 6.81 8 11228.4 278.3 8.98 7.37 9 245.9 7.94 As can be seen10 from Table2, when the number262.1 of airport surface departure 8.46 aircraft is controlled to be N = 7, the11 fuel consumption per A320 departure278.3 flight is reduced by 35 kg 8.98 to 67 kg compared to N = 9–11. Gas emissions decreased by 1.13–2.17 kg. AsFor can other be commonseen from aircraft Table 2, types when on the the number apron, whenof airport the airportsurface implements departure aircraft a departure is controlled control tostrategy be N = with 7, the N fuel = 7 duringconsumption busy hours, per A320 the reduction departure in flight fuel consumption is reduced by and 35 kg total to pollutant67 kg compared emissions to Nper = flight9–11. Gas is shown emissions in Figure decreased8. by 1.13–2.17 kg. For other common aircraft types on the apron, when the airport implements a departure control strategy with N = 7 during busy hours, the reduction in fuel consumption and total pollutant emissions per flight is shown in Figure 8. Energies 2018, 11, 2473 13 of 15 EnergiesEnergies 2018 2018, 11, 11, x, x FOR FOR PEER PEER REVIEW REVIEW 1313 of of 15 15

FigureFigure 8. 8. The The fuel fuel consumption consumption reduction reduction graph. graph.

FromFrom Figures Figures8 8 8and and and9 ,9, 9, it it canit can can be be be seen seen seen that that that the the the fuel fuel fuel consumption consumption consumption and and and pollutant pollutant pollutant emissions emissions emissions of of of each each each typetype areare are greatlygreatly greatly reduced. reduced. reduced. Therefore, Therefore, Therefore, the the the departure departure departure control control control strategy strategy strategy was was adoptedwas adopted adopted during during during the busy the the hoursbusy busy ofhourshours airport of of airport airport operation, operation, operation, which which effectivelywhich effectively effectively reduced redu redu thecedced fuel the the consumption fuel fuel consumption consumption during during during the taxiing the the taxiing taxiing stage ofstage stage the aircraftofof the the aircraft aircraft and reduced and and reduced reduced their pollutant their their pollutant pollutant emissions. emissions. emissions.

FigureFigure 9. 9. The The reduction reduction in in the the pollutant pollutant emissions. emissions. 6. Conclusions 6.6. Conclusions Conclusions In this paper, the Shanghai Hongqiao Airport is taken as an example to study the control strategy InIn this this paper, paper, the the Shanghai Shanghai Hongqiao Hongqiao Airport Airport is is taken taken as as an an example example to to study study the the control control strategy strategy for departure aircraft pushback on the airport surface. The influence of the different numbers of forfor departure departure aircraft aircraft pushback pushback on on the the airport airport su surface.rface. The The influence influence of of the the different different numbers numbers of of departure aircraft within the apron and taxiway systems on the runway utilization rate and the takeoff departuredeparture aircraft aircraft withinwithin thethe apronapron andand taxiwaytaxiway systemssystems onon thethe runwayrunway utilization utilization raterate andand thethe rate was studied under airport runway capacity constraints. Additionally, the influence of factors, such takeofftakeoff rate rate was was studied studied under under airport airport runway runway capa capacitycity constraints. constraints. Additionally, Additionally, the the influence influence of of as the number of departure aircraft in the apron and taxiway systems, the position of the apron, and factors,factors, such such as as the the number number of of dep departurearture aircraft aircraft in in the the apron apron and and taxiway taxiway systems, systems, the the position position of of the configuration of airport arrival and departure runways, on the departure taxiing time of aircraft thethe apron, apron, and and the the configuration configuration of of airport airport arriva arrival land and departure departure runways, runways, on on the the departure departure taxiing taxiing was analyzed. Multiple linear regression equations were used to establish an aircraft taxi departure timetime of of aircraft aircraft was was analyzed. analyzed. Multiple Multiple linear linear regr regressionession equations equations were were used used to to establish establish an an aircraft aircraft time prediction model and the reductions in fuel consumption and pollutant emissions were calculated. taxitaxi departure departure time time prediction prediction model model and and the the reductions reductions in in fuel fuel consumption consumption and and pollutant pollutant emissions emissions The results show that reasonable control of the pushback of departing aircraft during the airport’s werewere calculated. calculated. The The results results show show that that reasonable reasonable co controlntrol of of the the pushback pushback of of departing departing aircraft aircraft during during busy hours can reduce the aircraft departure taxiing time without reducing the runway utilization thethe airport’s airport’s busy busy hours hours can can reduce reduce the the aircraft aircraft dep departurearture taxiing taxiing time time without without reducing reducing the the runway runway rate and takeoff and departure rate of the aircraft, thereby reducing aircraft fuel consumption and utilizationutilization raterate andand takeofftakeoff andand departuredeparture raterate ofof thethe aircraft,aircraft, therebythereby reducingreducing aircraftaircraft fuelfuel pollutant emissions during the taxiing phase. consumptionconsumption and and pollutant pollutant emissions emissions during during the the taxiing taxiing phase. phase. Author Contributions: Writing: X.Z. and N.L.; Providing the case study and the idea: X.Z., N.L., Y.S., and H.Z.; Revising and editing: S.-B.T. and K.W. Energies 2018, 11, 2473 14 of 15

Funding: This research was funded by the National Natural Science Foundation (No. U1533112) and the National Social Science Foundation (No.13CGL005). Acknowledgments: The authors would like to thank the National Natural Science Foundation (No. U1533112) and the National Social Science Foundation (No.13CGL005) for the support. Conflicts of Interest: The authors declare no conflict of interest.

References

1. María, V.M.; Hall, C.A. Modelling Performance and Emissions from Aircraft for the Aviation Integrated Modelling project. J. Aircr. 2010, 47, 812–819. 2. Jung, Y. Fuel Consumption and Emissions from Airport Taxi Operations. Available online: https://core.ac. uk/display/10559073 (accessed on 17 August 2018). 3. Balakrishnan, H.; Jung, Y. A Framework for Coordinated Surface Operations Planning at Dallas-Fort Worth International Airport. Available online: https://arc.aiaa.org/doi/abs/10.2514/6.2007-6553 (accessed on 13 September 2108). 4. Simaiakis, I.; Balakrishnan, H. Queuing Models of Airport Departure Processes for Emissions Reduction. Available online: https://arc.aiaa.org/doi/10.2514/6.2009-5650 (accessed on 13 September). 5. Jung, Y.C.; Hoang, T.; Montoya, J. A Concept and Implementation of Optimized Operations of Airport Surface Traffic. Available online: https://arc.aiaa.org/doi/abs/10.2514/6.2010-9213 (accessed on 13 September). 6. Lee, H.; Simaiakis, I.; Balakrishnan, H. A comparison of aircraft trajectory-based and aggregate queue-based control of airport taxi processes. In Proceedings of the 29th Digital Avionics Systems Conference, Salt Lake City, UT, USA, 3–7 October 2010; pp. 1–15. 7. Simaiakis, I.; Khadilkar, H.; Balakrishnan, H.; Hansman, R.G. Demonstration of reduced airport congestion through pushback rate control. Transp. Res. Part A 2014, 66, 251–267. [CrossRef] 8. Ravizza, S.; Atkin, J.A.D.; Maathuis, M.H.; Burke, E.K. A combined statistical approach and ground movement model for improving taxi time estimations at airports. J. Oper. Res. Soc. 2013, 64, 1347–1360. [CrossRef] 9. Tang, Y. Aircraft Taxi Routes Planning and 3D Simulation for A-SMGCS. Ph.D. Thesis, Nanjing University of Aeronautics and Astronautics, Nanjing, China, December 2014. 10. Zhao, X.; Tang, J.; Lu, F.; Han, B. Strategy Analysis for Delayed Fights Pushback and Sensitivity Analysis of the Length of Virtual Queue. J. Sichuan Univ. 2016, 3, 115–123. 11. Li, N.; Lv, H. Method to Find out Path of Aircraft on Surface Based on Surface Movement Radar Data. Aeronaut. Comput. Tech. 2016, 46, 6–9. 12. Lv, H. The Analysis of Aircraft Taxiing and Prediction Study Based on Surveillance Data. Master’s Thesis, Civil Aviation University of China, Beijing, China, April 2016. Available online: http://kns.cnki.net/KCMS/detail/detail.aspx?dbcode=CMFD&dbname=CMFD201701&filename=1016 776031.nh&uid=WEEvREcwSlJHSldRa1FhdkJkVWI0UTA0a3RxcEdTZnNJVW1ZVVFkM3gvVT0=$9A4 hF_YAuvQ5obgVAqNKPCYcEjKensW4IQMovwHtwkF4VYPoHbKxJw!!&v=MDAwNDFyQ1VSTEtm YnVkb0Z5bmhWYjdMVkYyNkdMUy9HTkhQcnBFYlBJUjhlWDFMdXhZUzdEaDFUM3FUcldNMUY= (accessed on 17 August 2018). 13. Eurocontrol Standard Document for Surveillance Data Exchange Part 9: Category 062 SDPS Track Messages. Available online: https://www.google.com.tw/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&ved=2 ahUKEwiHtb3gtLndAhWLyrwKHdT3AxoQFjAAegQIARAC&url=https%3A%2F%2Fwww.eurocontr ol.int%2Fsites%2Fdefault%2Ffiles%2Fcontent%2Fdocuments%2Fnm%2Fasterix%2Farchives%2Fasterix -cat062-system-track-data-part9-v1.11-042010.pdf&usg=AOvVaw389n3cb5i1jElDR9PsFJ3s (accessed on 14 September 2018). 14. Anderson, R.C.; Reeb, D.M. Founding-family Ownership and Firm Performance: Evidence from the S&P 500. J. Financ. 2003, 58, 1301–1328. 15. Tsai, S.-B. Using the DEMATEL Model to Explore the Job Satisfaction of Research and Development Professionals in China’s Photovoltaic Cell Industry. Renew. Sustain. Energy Rev. 2018, 81, 62–68. [CrossRef] 16. Lee, Y.-C.; Hsiao, Y.-C.; Peng, C.-F.; Tsai, S.-B.; Wu, C.-H.; Chen, Q. Using Mahalanobis-Taguchi System, Logistic Regression and Neural Network Method to Evaluate Purchasing Audit Quality. Available online: http://journals.sagepub.com/doi/10.1177/0954405414539934 (accessed on 13 September 2108). Energies 2018, 11, 2473 15 of 15

17. Liu, B.; Li, T.; Tsai, S.B. Low carbon strategy analysis of competing supply chains with different power structures. Sustainability 2017, 9, 835. 18. Qu, Q.; Tsai, S.-B.; Tang, M.; Xu, C.; Dong, W. Marine ecological environment management based on ecological compensation mechanisms. Sustainability 2016, 8, 1267. [CrossRef] 19. Lee, Y.-C.; Wang, Y.-C.; Chien, C.-H.; Wu, C.-H.; Lu, S.-C.; Tsai, S.-B.; Dong, W. Applying revised gap analysis model in measuring hotel service quality. SpringerPlus 2016, 5, 1191. [CrossRef][PubMed] 20. Wang, J.; Yang, J.; Chen, Q.; Tsai, S.-B. Collaborative Production Structure of Knowledge Sharing Behavior in Internet Communities. Mob. Inf. Syst. 2016, 2016, 8269474. [CrossRef] 21. Tsai, S.-B.; Lee, Y.-C.; Guo, J.J. Using modified grey forecasting models to forecast the growth trends of green materials. Proc. Inst. Mech. Eng. Part B J. Eng. Manuf. 2014, 228, 931–940. [CrossRef] 22. Ge, B.; Jiang, D.; Gao, Y.; Tsai, S.-B. The influence of legitimacy on a proactive green orientation and green performance: A study based on transitional economy scenarios in china. Sustainability 2016, 8, 1344. [CrossRef] 23. Lee, S.-C.; Su, J.-M.; Tsai, S.-B.; Lu, T.-L.; Dong, W. A comprehensive survey of government auditors’ self-efficacy and professional Development for improving audit quality. SpringerPlus 2016, 5, 1263. [CrossRef] [PubMed] 24. ICAO Aircraft Engine Emissions Databank. Available online: https://www.easa.europa.eu/easa-and-you/ environment/icao-aircraft-engine-emissions-databank (accessed on 17 August 2018). 25. Li, N.; Zhang, H.F. Calculating aircraft pollutant emissions during taxiing at the airport. Acta Sci. Circumstantiae 2017, 37, 1872–1876.

© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).