Improving Aircraft Turn Around Reliability Specific Aircraft Body Design Parts hamper Ground Handling and Airport Performance
Hartmut Fricke
Michael Schultz
Institute of Logistics and Aviation Technische Universität Dresden Dresden, Germany
[email protected]
Institute of Logistics and Aviation Technische Universität Dresden Dresden, Germany
[email protected]
The airport, and specifically the turn around time (TAT) of aircraft at the gate or a remote position from the terminal have been recognized as crucial element to ATM system performance. Currently, the TAT ranges around 30 min for short/medium range aircraft. For the 2020 Single European Sky, SESAR claims as performance target for a cut down to 15 min while also increasing process reliability. There are several reasons, why the turn around is still remarkably uncertain, mainly caused by shared responsibility for the individual ground handling processes, a frequent distortion of gate occupancy schemes at the airport and still deficient interfaces with the aircraft body. All this leads to only a limited predictability of the “Earliest Off Block Time”, this being an important time constant to trigger the departure and consequently the arrival sequence. This paper reveals the current data quality as found during a large field study at German Airports, derives the reasons for largely varying process times both on the technical and procedural level and shows the potential for improved TAT reliability through aircraft interface optimization. Aircraft Turn Around, Aircraft Body, Ground Handling, Boarding, Critical Path, Monte Carlo Simulation
I.
INTRODUCTION
The turn around time (TAT) of aircraft has been defined by IATA’s Aircraft Handling Manual (AHM) 810 [1] as the time period an aircraft occupies a stand or a gate at the airport. More specifically this period is framed by two activities: The positioning and removal of the aircraft wheel chocks, respectively named as On and Off Block Time. As this time is directly impacting the airport / terminal capacity, there is a vital interest in predicting exactly the Gate Occupancy Time (GOT) by means of stable ground handling processes from the airport operator’s point of view and a similar one in minimizing that time from an airline’s perspective since block time is commercially lost time. Therefore, the GOT has become a central planning parameter for an airport / apron terminal design processes [4]. The largest component during on block is the boarding and de-boarding of passengers, as field measurements clearly show [2]. Hence, it could also be learned, that ground handling events are characterized by a remarkable diversity in processes, occurrence of such processes and their service depth, making every aircraft turn around
somehow a unique procedure in terms of required times, interfacing and services. Our referred field study was performed in summer 2006, to learn about these constraints and to gather relevant data with the aim to exploring ways how to improve the reliability of and also shorten the time needed for a turn around. This with SESAR’s performance target in view, to cut down the TAT for domestic flights to 15 min, and 30 min for international flights [5]. Section II presents the results of the field study; section III discusses the management of the gathered data in a relational database to allow systematic analysis and dissemination of the results to the various stake holders. Section IV reveals the statistical data modeling to determine the level of reliability in today’s turn around operations by means of Monte Carlo (MC) Simulations. Section V creates an improved aircraft interface scenario and derives the expected increase in turn around reliability, again by applying MC modeling. II.
CURRENT TURN AROUND PRACTICE
A. Overview The turn around has been described in several studies such as [6], many of them putting emphasis on the boarding and deboarding processes, [7], [8], [9] and is also part of many standard documents such as the Aircraft Characteristics For Airport Planning Manual of any modern aircraft [3]. It is generally represented as a bunch of processes, from which a subgroup may run in parallel, and others are required to run sequentially, e.g. due to legal or logistical requirements such as limited space around the aircraft, tool availability, or legal constraints such as to prohibit fueling with passengers onboard. The collection of sequential processes consuming the maximum time during turn around is called critical path. As stated, typical process members of the critical path are boarding, de-boarding, fueling, loading, unloading and service processes such as cleaning or catering [2], [3]. The following picture depicts the typical process and dependencies:
Demin. Water Air Starter De-icing Nonroutine Maintenance / On-demand Check Routine Maintenance / Flight Check Cooling and Heating Fueling Wheel and tire check Portable Fresh Water Waste Water Handling (Sanitary) Exterior Aircraft Cleaning
Open PAX door Position wheel chocks
Disembark PAX
Disembark Crew
Position PAX stairs
Cabin & Cockpit
Final clean
Board Crew
Crew Check
Board PAX
Close PAX door Remove PAX stairs
catering Duty-Free stocks
Start & Remove wheel chocks
In-flight entertainment Position Cargo 1
Open Cargo
Unload Bags
Unload cargo
Load cargo
Load bags
Close cargo door
Remove cargo Loader 1
Figure 1. Parallel and sequential turn around processes
B. Set up of the Field Study* To identify possibilities to reduce individual process times and their expected remarkable variance in time especially on the critical path, the field investigation aimed at exploring potential changes to the aircraft-design and arguments for innovative technologies to be installed in the aircraft or applied on ground in order to smoothen and accelerate the ground services resulting in shorter turnaround times with increased reliability. This activity shall be seen in concert with e.g. the physical implementation of supervision technologies such as process scanning and tracing devices or e.g. data networking at the apron, as e.g. [6] is calling for, and modification of current legal regulations (see section III). The datasets collected had to cover nearly all models and types of Airbus-manufactured aircraft and at least some representative Boeing aircraft. It was further anticipated to consider different airport types, such as the Hub-Airport Munich (MUC) and mid size airports such as Leipzig/Halle (LEJ) and Dresden (DRS) in Saxony. The data gathered also covered interviews with ground handling staff as well as detailed time sequence measurements to allow determination of specific process indicators (fuel quantity, passenger rates, etc., see TABLE I. ). Further, process interruptions due to limited or distorted logistics were noted. These data did form the platform for various process analyses looking at optimization potential. Based on the Aircraft Handling Manual [1] providing a detailed framework on how to cluster and categorize ground handling services, then * The field studies needed for this research were kindly supported by Airbus Deutschland, Dresden, Leipzig/Halle and Munich Airport, and the Ground Handling Agents MUCGround GmbH and PortGround GmbH.
extended for the boarding related activities, the following collections of services were gathered (see also figure 1): •
Aircraft Servicing (covering exterior services, such as water and lavatory service, e.g. attachment of stairs and passenger bridges);
•
Loading and Unloading (on- and offloading of belly payload, container and bulk);
•
De-boarding (differentiating in remote and gate positions as well as number of doors used);
•
Catering (Handling at Ramp);
•
Interior Cleaning (including cabin and crew rest compartment cleaning);
•
Crew Change and Cabin Preparation (including duties to be performed by cabin crew);
•
Boarding (hence without distinguishing into different boarding procedures, this later recognized as important data and being subject to further studies at Dresden University of Technology (TUD)); and
•
Fueling (number and location of fuel valves used).
Based on that structure, the measurements performed in the field did include timestamps assigned at the lowest observable level, so linked to a process, task or element as respective subprocesses. A distinct process start and end was required to be observed. As such, those timestamps were tested on their traceability and practicability before they were implemented into the process templates (see figure 2).
At some processes, intermediate timestamps were defined in order to allow gathering information about process interruption times, waiting times or other intermediate points (milestones). Besides collecting time sequences, the sampling of the following additional information affecting the turnaround was performed: • • • • • •
web application to create easy access to the data, and providing a clear structure and documentation platform:
Manpower (amount of personnel performing a specific process, task or element) The equipment types used; The equipment quantities; Load figures (e.g. passenger figures, seating, baggage-, cargo-, mail-figures etc.); Aircraft layout information (if possible: amount of lavatories, galleys, etc.); Transfer volumes (e. g. fuel quantity figures).
The various data collection templates designed for the field activity is depicted in the following figure e.g. for the loading case1: Figure 3. Web front end to access data of turn around database
D. Data Preparation for Turn Around Comparison Means For each of the analyzed processes, and sub-processes, dedicated values were derived from the field measurements in order to allow comparing the individual turnarounds. The following parameters were defined for the individual processes [2]: TABLE I.
SPECIFIC PROCESS PARAMETER
Specific Process Data generated (extract) Process (De-)Boarding
Catering
Cleaning
Fueling
Figure 2. Data collection template, the loading case
C. The Collected Data: A Web Application To allow efficient handling of the remarkable amount of data collected, a relational MYSQL Database was implemented. Further a PHP application was designed so forming a powerful 1
ULD = Unit Load Device, Standard cargo container. Typical dimension: 317cm length, 243 cm width, 299 cm height. Various subclasses do exist according to IATA coding definitions.
(Un-)Loading
Servicing
Value
Unit
A/C (De-)Boarding Rate
[PAX/min]
A/C (De-)Boarding Time
[min]
Avg. Boarding Flow Interruption Time
[min]
Catering Time Split up AFT Galley
[%]
Catering Time Split up FWD Galley
[%]
Total Catering Vehicle Retention Time
[min]
Cleaning Rate
[per seat]
Cleaning Time Cabin, Lavatory, Galley
[min]
Average Fuel Flow Rate
[l/min]
Starboard/Portside Fueling Split Up
[%]
Tanker-Dispenser Split Up
[%]
(Un-)Loading AFT/FWD
[ULD/min]
(Un-)Loading Bulk
[kg/min]
A/C Service Vehicle Retention Time
[min]
Pushback Waiting and Standby Time
[min]
A/C Servicing Equipment Split up
[%]
These specific data could be sampled for over 120 complete turn around events, measured at the three sites MUC, DRS and LEJ. III.
STATISTICAL PROCESS DATA ANALYSIS
A. Legal and Procedural Constraints In accordance with ICAO Doc 9626, the Manual on the Regulation of International Air Transport [10], ground handling consists of those processes as listed previously and implemented in our database. Further, the ICAO Doc 9562 Airports Economics Manual [1] separates ground-handling services into terminal handling (passenger check-in, baggage and freight handling, flight plan processing) and ramp handling (aircraft handling, cleaning and servicing). For all of these processes, operational interdependencies do exist, limiting the parallelizing of these processes, as the area around the aircraft is very limited in light of the various service equipment used during ground handling. Further, legal requirement call for granted maneuvering capability for the fire brigade on one side (typically located right hand to the aircraft) and passenger related activities (stairs typically installed left hand). Obviously, airlines and ground handling companies are permanently investigating possibilities to reduce process constraints, since further parallelizing of processes besides mainly fueling, catering, cleaning and servicing nowadays as found during the field study (shown in figure 4) is seen as mandatory to substantially reduce the TAT. This aspect will be focused in subsequent papers.
Candidates for (increased) parallelized processing so remain • loading/unloading and • fueling. Further, • shortening and • increasing the reliability of the boarding and deboarding time do represent a second field for improvement, being subject of this paper. Investigation was consequently directed to the technical aspects of these processes in the field for getting a clear picture of all operational requirements and systematic deficiencies linked to them, presented in the next section. B. Technical Constraints Based on the interviews done with ground handling personnel in the field and observations gathered during the monitoring phase, the following representative process distortion cases were found [2]: TABLE II.
TECHNICAL DEFICIENCIES AT TURNAROUND
Technical Deficiencies at Turnaround (extract) Process
Deficiency
(De-)Boarding
Installation of second passenger bridge on gate position does not result in the time saving of the two passenger stairs on remote position due to the marginal distance between the two doors.
Boarding
Boarding is influenced by the mutual obstructions of the passengers.
Catering
The catering door is opened by the purser after all passengers left the aircraft and the catering vehicle arrived on position. Usually, the employee signalized his arrival by knocking at the door which is occasionally not realized by the crew and result to a delay of the beginning of the process. Leaving of the vehicle affected by obstructions because of loading equipment especially on smaller aircraft with bulk load.
Cleaning
High quantity of waste produced because of in-flight services. The lack of disposal units leads to defilement of the aircraft and time consuming removal by the cleaning staff.
Fueling
Obstructions between the service vehicles result to delayed start of the fueling process and a non-compliance of the safety requirements Simultaneous fueling on both wings enhances the flow rate not to the double of a single side fueling The fuel computer has problems calculating the exact filling quantity due to temperature variations. So, the centre tank is opened too late and fuelled with a flow rate lower than for the other tanks.
(Un-)Loading
Cargo Door: Panel for opening hatches of the A300, A330 and A340 attached to high to be reached manually by all personnel Conveyer belts used for (un)loading bulk. Height of aircraft influenced by the current weight of Payload and Fuel. Permanent Altitude adaptation needed inducing repositioning of the equipment
Servicing
Figure 4. Sample parallel processing during turn around – field study
Cockpit misses the information about the removal of the GPU by the loading crew. However, clear commitment necessary before disconnection. Reasoned by weight of the adapter, negative effects can occur for the employee by attaching the unit as well as for the holding of the plug connection itself.
For the sample case presented below, this offset is at 10 PAX/min:
Class frequency
25
Figure 5. Physical deficiencies: displaced control panel for cargo door (left) / manual positioning of load compartment (right)
22
20
22
17 13
15 9
10
5 5
4
4
9
10
1
0 0
The following section looks into the turnaround database and determines the statistical behavior of the processes. C. Determination of Process Stability To determine the nature of the generated process descriptors as listed in TABLE I. typical density distributions were compared to the data. Since the data has a non normal distribution character, a WEIBULL density function f(x) with α and β as shape parameter was finally selected for probing the data fitting, since it is most commonly used in life data analysis and due to its flexibility. It can mimic the behavior of other statistical distributions such as the normal and the exponential. A Weibull distribution may generally written in the form
f ( x, α , β ) =
α β −1 −( x / β )α x e βα
(1)
1
2
3 4 5 6 7 8 Class (4 PAX/min width)
Figure 6. Class building for the de-boarding case
The comparison between the classified real data set and the functional approximation is shown in figure 7, followed by the summary of the statistical fit data set for the chi-square test (TABLE III). 25 real Data
20 [PAX/min]
To conclude, quite a bunch of technical deficiencies were found during the field study that effectively hamper the manual and also partly automated activities during turn around in terms of delays and consequently reduced reliability for process times.
approx. Data
15 10 5 0 0
2
4
6
8
10
Class
Where x >= 0 and f(x) = 0 for x < 0 x to substituted by (x + x_offset) α = scale parameter β = shape parameter
For all processes found to be timely critical during today’s turn around (see figure 4), data distributions were analyzed, statistically tested on its significance against chi-square. The probing procedure is shown exemplarily for the de-boarding process, referring to the process parameter passenger rate per door [PAX/min] to allow normalizing of the varying seat load factors. It may be noticed that for some processes, an offset is required to adopt the data range to the Weibull distribution by x_offset. De-boarding: • x offset: • Number of counts: • Minimum flow rate: • Maximum flow rate: • Number of classes: • Class with:
10 97 4.1 38.9 10 4
PAX / min PAX / min PAX / min (see TABLE III) PAX/min
Figure 7. Approximated versus real data distribution – de-boarding case
TABLE III.
STATISTICAL FIT AND PROOF DATA – DE-BOARDING CASE Statistical Fit and Proof Data
(De-)Boarding [PAX / min] Class number logic
5 log(number of counts)
α
2.23619
β
19.2339
Value
x offset
10
Mean
17.0354
Standard Variation
7.92729
Variance
0.465343
Density (Weibull)
0.0030067 exp(-0.00134457 x2.23619) x1.23619
Distribution
1 – exp(-0.00134457 x2.23619)
Chi²
10.3228
Level of Significance
0.05
Proof Pass Status (Chi2 < Proof)
12.5916 TRUE
board, de-boarding, sequence;
For some other timely critical processes, the following distributions are depicted below, showing the remarkable magnitude in the distribution variations: TABLE IV.
2.23619
Value
β
19.2339
x offset Boarding [PAX / min] α
10
2.29263
β
19.5224
x offset loading AFT&FWD [bags / min] α
0
3.13022
β
165.289
x offset Unloading AFT&FWD [bags/ min] α
0
4.05691
β
215.208
x offset
0
These parameter sets do allow sorting the processes against their stability, as with α (scale parameter) tending to smaller values and β (shape parameter) tending to larger values do express increasing variance of the distribution. It comes to the following ranking: 1. Unloading 2. Loading 3. Boarding 4. De-boarding As observations did further show, the reason for this interprocess uncertainty trend refers to the varying standardization and automation level of the individual processes as well as the quality of technical support means, as described above. IV.
in
loading and unloading, obviously running in sequence, too, and
•
either fueling, catering or cleaning, all running in parallel
have been identified as critical processes according to figure 4. Concerning the fueling process, process interruptions were recorded on a regular basis, resulting from the fact, that fueling is generally initiated by the responsible personnel without knowing the exact quantity beforehand, to be ordered by the cockpit crew. So a typical volume is being fueled in a first step, than for most of the tracked cases, a break occurs within which communication takes place and the exact quantity will be told. This effect was also included in the process modeling. The following density and distribution shapes do result from functional fitting as presented in section III, reflecting also the x offset applied to the de-boarding case (see figure 9, page 7). The MC method is then applied to the collection of these processes, linked according the prescribed constraints, and executed for 104 simulation runs. It comes to the following TAT distribution: 3000 2509
2500
Class frequency
(De-)Boarding [PAX / min] α
running
•
REAL DISTRIBUTIONS – CRITICAL TA PROCESSES
Real Distributions – Critical TA Processes
obviously
2000
1833
1812
1500
1307
1000
852 506
500
300 240
222
122
92
53
33
20
25
10
15
12
5
5
10
11
12
13
14
15
16
17
18
19
20
0 1
2
3
4
5
6
7
8
9
Class (190 s witdh)
Figure 8. Turn around time distribution
UNCERTAINTY EFFECT ON THE TURN AROUND
When accumulating the critical processes along the line, the agglomeration of all partial process uncertainties will obviously induce a total uncertainty for the turn around. To determine the magnitude of that total variation, analytically modeling was replaced in favor of statistical random probing. This was achieved by applying the Monte Carlo method, relying on repeated random sampling of all critical processes, using the calibrated distributions as found above. The results are presented in the next section.
The shape of the TAT distribution so complies again with a WEIBULL behavior with the following over all process characteristics:
A. MC Simulation for overall turnaround uncertainty determination The processes
Obviously, the central limit theorem according to which under certain conditions such as referring to independent and identically-distributed data with finite variance, the sum of a large number of random variables is approximately normally distributed, cannot not applied without inacceptable errors for this data case.
•
boarding, including the emphasis on the door closing procedure after the last Passenger did
•
Minimum Time: 1135 s
(19 min)
•
Maximum Time: 12486 s
(208.1 min)
•
Mean: 1872 s
(31.2 min)
•
Standard Deviation: 23 s
0.008
0.8
0.008
0.8
0.006
0.6
0.006
0.6
0.004
0.4
0.004
0.4
0.002
0.2
0.002
0.2
0.0 400
0.000
0.000 200
300
0
100
200
Bags [kg / min]
1.0
0.08
0.8
0.08
0.8
0.06
0.6
0.06
0.6
0.04
0.4
0.04
0.4
0.02
0.2
0.02
0.2
0.0
0.00
0.00 0
10
20
30
40
50
Density Function
0.10
60
0.0 0
10
20
PAX per min and door
50
60
Catering
0.10
0.08
0.8
0.08
0.8
0.06
0.6
0.06
0.6
0.04
0.4
0.04
0.4
0.02
0.2
0.02
0.2
0.0
0.00
0.00 5
10
15
20
25
Density Function
1.0
Distribution
Density Function
40
PAX per min and door
Fueling
30
5
10
15
20
25
30
Catering Time [min]
Last Pax until doors closed
Cleaning
0.50
1.0
0.08
0.8
0.40
0.8
0.06
0.6
0.30
0.6
0.04
0.4
0.20
0.4
0.02
0.2
0.10
0.2
0.0
0.00
0.00 0
5
10
15
20
25
30
Density Function
1.0
Distribution
0.10
1.0
0.0 0
Fueling Time [min]
Density Function
30
0.10
0
Distribution
De-boarding
1.0
Distribution
Density Function
Bags [kg / min]
Boarding
0.10
0.0 400
300
Distribution
100
1.0
0.0 0
2
Cleaning Time [min]
Figure 9. Modeled distributions for all time critical processes
4
6
Delay [min]
8
10
Distribution
0
Density Function
0.010
Distribution
Density Function
Unloading
1.0
Distribution
Loading
0.010
Instead, it is shown, that all type of optimization of single (and critical) processes does lead to a turn around time distribution with Weibull character. This to be proven by the different spot cases analyzed in section V. The 3 “critical” processes (running in parallel) in between deboarding/boarding and unloading/loading were the limiting cases with the following distribution, based on another 104 simulation runs:
TABLE V.
ADDRESSED TIME EFFECTS TO TECHNICAL DEFICIENCIES Fighting Technical Deficiencies at Turnaround Deficiency counter measure Second passenger bridge on gate position Innovative Boarding Concept / New seat arrangement. Catering door signaling technically supported Enhanced vehicle logistics lowering obstruction effects Reduced waste produced Enhanced vehicle logistics lowering obstruction effects Optimized Simultaneous fueling on both wings fuel computer upgrade to calculate FOB precisely Optimized Cargo Door: Panel Auto adjustment for Conveyer belts & equipment Cockpit Info granted about GPU removal Light-weight adapter for GPU
Process (De-)Boarding Boarding Catering
Number of occurences
6000
5657
5000
Cleaning
4000
3544
Fueling
3000 2000 782
1000 0 Fueling:
Catering:
Cleaning:
(Un-)Loading
Figure 10. Distribution of limiting process cases linking de-boarding and boarding Servicing
So fueling was the limiting process out of the parallel processes for most of the cases (56%), followed by catering (36%), cleaning (8%), and other processes (0.17%). V.
POTENTIALS TO INCREASE TA RELIABILITY
The field study has shown a set of technical deficiencies which lead to uncertain, non standard processing in many cases. Based on representative interviews with ground handling personnel, individual impact effects were affiliated to the specific aspects. These have to be transferred into adopted process shape distributions, those in turn being subject of further MC simulations to derive the overall potential gain on the turnaround. A. Addressing time burden to technical deficiencies With reference to TABLE II, the following potential for reducing the impact effect to uncertainty of the individual deficiencies was found to be applicable according to expert judgment in the field (bold values mean consideration for the following MC simulations as time critical processes, see TABLE V). It shall be noticed that the expected gains shown in the above table may be seen as (realistic) indicators only, and will be subject of alteration when performing interviews at other airports. Nonetheless, the following section will clearly show how optimization potential and TA reliability do correlate for that special case. Section D will further systematically investigate the sensitivity between a given potential and the resulting TAT reliability.
Judged Variance reduction potential 10%
20% 10% 15% 25% 15% 20% 5% 10% 30% 10% 10%
B. Conversion into Adopted Distributions The assumed potential has to be transformed into process distribution behavior, respectively the modeling of reduced process uncertainty achieved with •
α (scale parameter) tending to larger values, and
•
β (shape parameter) tending to smaller values.
The approximation for the new shape and scale parameter fulfills with the Gamma Function Γ the following relationship between the parameters α, β, and the variance: ∞
∫
with Γ = t
x −1 −t
e dt it comes to
0
(2)
⎛ 2 ⎞ 1 Var ( x) = α −2 / β ⎜⎜ Γ( + 1) − Γ( + 1) 2 ⎟⎟ β β ⎝ ⎠
(3)
This leads us to the following adopted shape parameters for the time critical processes (TABLE VI).
TABLE VI.
ADOPTED DISTRIBUTIONS – CRITICAL TA PROCESSES
TABLE VII.
Adopted Distributions – Critical TA Processes (De-)Boarding [PAX / min] α
2.534656
min
1135 s
1222 s
7,67%
β
19.193186
max
12486 s
7790 s
-37,61%
x offset Boarding [PAX / min] α
10
mean
1872 s
1784 s
-4,70%
Variance
535 s²
395 s²
-26,17%
3.4680421
23 s
19 s
-17,39%
β
19.23082
x offset loading AFT&FWD [bags / min] α
0
3.252597
Value
Figure
β
154.5100 0
4.047813
β
206.81878
x offset
0
1434
Class frequency
5631
5000 4000
3564
3000 2000 788
1000 0 Fueling
Catering
Cleaning
So equal to the baseline scenario, fueling did constrain the parallel processes for 56%, followed again by catering (36%) and cleaning (8%).
1400 1200
1104 923 756
800
530
600
459 321
400
235
177
178
131
58
64
47
21
18
14
10
14
15
16
17
18
19
20
0 1
2
3
4
5
6
7
8
9
10
11
12
13
Class (190 s witdh)
Figure 11. Improved “stabalized” turn around time distribution
The shape of the improved TAT distribution now follows the over all process characteristics: • • • •
Gain / Loss
Figure 12. Distribution of the limiting process cases– improved scenario with parallel processes
1718 1755
1000
Number of occurences
6000
2000
1600
Optimized
The three critical processes now follow the distribution as depicted below:
x offset Unloading AFT&FWD [bags/ min] α
1800
Baseline
SD
These adopted distributions are now being re-applied to the MC simulation set up to explore the effect of every single counter-measure on the turn around time. With another 104 simulation runs, it comes to the following distribution:
200
POTENTIAL FOR INCREASED PROCESS RELIABILITY
Expected Benefits of technical counter measures for increased process reliability
Minimum Time: 1222 s Maximum Time: 7790 s Mean: 1784 s Standard Deviation: 19s
(20.4 min) (129.8 min) (29.7 min)
When comparing both scenarios, the following table depicts the calculated stability gain for the turn around time (TABLE VII).
C. Reduced Process Uncertainty Affecting TA Reliability Two effects can be noticed when comparing the reference scenario with the improved: Although the mean of the TA time is only decreasing by roughly 4% its stability, expressed through the variance is increasing by roughly 25%. So obviously, installing efficient technical upgrades at the aircraft for the sample case only would significantly improve the turn around reliability achieved through stable (deterministic) processes. TABLE VII indicates a potential for increasing the turn reliability by roughly 25% when retrofitting the aircraft body accordingly, and optimizing the manual procedures for the processing. This is seen as an important contribution which obviously is worse to be investigated further. Hence, it also came clear that this type of counter measures will in no way allow complying with the SESAR performance targets for the turn around. This one can so only be achieved by parallelizing fueling and hereafter catering with the critical processes (de-)boarding and (un-)loading. For the boarding processes, an in-depth study has recently been completed at TUD, exploring the effects of different boarding strategies (random, inside-out...) onto the process time.
Analogous to subsection A of this section, again the four processes boarding, de-boarding, loading and unloading were taken into consideration and compared with the reference scenario. Referring to the findings as shown in TABLE V, the potential was altered between 10% and 50% each. The already discussed transformation into process distributions, again modeled through increasing α values and decreasing β values of the Weibull distribution functions are shown in the following TABLE VIII for the boarding respectively de-boarding process and in TABLE IX for the loading and unloading process according to (2): ADOPTED DISTRIBUTIONS - BOARDING/ DE-BOARDING
Adopted Weibull Distribution Parameters Variance De-boarding Boarding reduction potential β β α α 0%
2.31
19.23
2.29
19.52
10%
2.53
19.19
2.60
19.47
20%
2.91
19.10
2.98
19.37
30%
3.39
18.97
3.47
19.23
40%
4.03
18.79
4.12
19.05
50%
4.92
18.57
5.04
18.83
TABLE IX.
0%
2.88
155.36
3.60
208.18
10%
3.25
154.51
4.05
206.82
20%
3.72
153.43
4.61
205.27
30%
4.31
152.15
5.33
203.55
40%
5.10
150.66
6.30
201.67
50%
6.21
149.00
7.65
199.64
These adopted distributions are now being applied to the MC simulation set up to explore the effect of every single process on the mean TAT and on the standard deviation of the turn around time. Figure 13 depicts the impact of improved process stability on the mean TAT, the red line highlighting the mean TAT for the reference scenario.
Bo arding
Unloading
1850 1800 1750 1700 10%
20%
30%
40%
50%
process stability improvement
Figure 13. Reduced mean TAT through increased process stability
As seen at figure 13, increased boarding or de-boarding reliability will consequently lead to a significantly decreasing mean TAT, while the impact of the boarding reliability is slightly higher than de-boarding. Note that only the stability of the processes was improved, whereas mean values for every process were kept unchanged. Increasing the boarding reliability by e.g. 50% alone is responsible for a decreased mean TAT of roughly 90 s or 4.7%. Improving de-boarding stability by 50% will lead to a shortened mean TAT, leading to 66 s or 3.5% when compared with the reference scenario. Otherwise, an improved loading and unloading reliability obviously have no significant impact onto the total TAT, due to the fact that loading/ unloading processes are in most cases not part of the critical path, although being “critical” processes. The influence of changing process distribution parameters onto the standard deviation of the turn-around time is shown in following figure 14:
ADOPTED DISTRIBUTIONS – LOADING/ UNLOADING
Adopted Weibull Distribution Parameters Variance Loading Unloading reduction potential β β α α
Debording
Loading
TAT Reference
Debo rding
25
B o arding
standard deviation [s]
TABLE VIII.
TAT Reference
1900
mean TA-time [s]
D. Sensitivity Analysis of Process Reliabilties Effect on the Turn Around Time It could be shown, that improving process reliability of critical processes will lead to significant benefits for the entire turn-around. Finally, the influence of every process, explored by changing distribution parameters for only one process while keeping all other constant, did deliver the specific reliability gain of every process to the TAT.
Lo ading
23
Unlo ading
21 19 17 15 10%
20%
30%
40%
50%
process stability improvement
Figure 14. Increased TA-standard deviation hough increased process stability
As seen in figure 14, an increase of 50% in the de-boarding process stability will result in a reduced standard deviation for the TAT of about 6.4%. An increase of the boarding reliability by 50% could lead to a reduced TAT standard deviation of 13.9%, while improving loading or unloading reliability again has no impact onto the turn-around time reliability. Both measurements (mean and standard deviation of the TAT) show the same ranking and trend for improvement potentials - the strong potential of increasing boarding reliability first and de-boarding reliability second and the de
facto non-existent influence of loading or improvement are the core findings of this study.
unloading
ACKNOWLEDGMENT The authors thank Susann Lehmann and Denise Hentschel, who conducted the time consuming field evaluations even at really bad weather conditions. Many thanks also to Christoph Thiel who supported the set up of the various statistical analyses and Daniel Fiedler for setting up the Web portal.
REFERENCES [1] [2]
IATA, Airport Handling Manual AHM 810, January 2003 H. Fricke, S. Lehmann, Cabin Operation – Service Handling within Final Status Report within Phase II, “Cabin Operation – Service Handling” within the project “Cabin Architecture and Design Concept” (CAaDC), December 2006 [3] Airbus Industrie, A320, Airplane Characteristics For Airport Planning, AC, Chapter 5 Terminal Operation, July 1995 [4] ICAO, Airport Planning Manual, Doc 9184-AN7902 Part 1, 1987 [5] SESAR Consortium, SESAR Definition Phase, D2 Report “The Perfomance Target”, Page 10, Brussels, November 2006 [6] Wu, C. L. Monitoring Aircraft Turnaround Operations — Framework Development, Application and Implications for Airline Operations. Transportation Planning and Technology, 2007 (under review) [7] Tom Caswell, Kyle Story, and Rafael Frongillo. 2007. The Airplane Seating Problem. Mathematical Contest in Modeling, Consortium for Mathematics and Its Applications (COMAP), 2007 [8] Michael Bauer, Kshipra Bhawalkar, Matthew Edwards. 2007. Boarding at the Speed of Flight. Mathematical Contest in Modeling, Consortium for Mathematics and Its Applications (COMAP), 2007 [9] Bachmat, Elkin, Bounds on the Performance of back-to-front airplane boarding policies, July 2006 [10] ICAO, Doc 9626 - Manual On The Regulation Of International Air Transport (2nd Edition), August 12, 2005, Montreal, Canada [11] ICAO, Doc 9562 . Airport Economics Manual, (2nd Edition), Montreal, Canada
