Вы находитесь на странице: 1из 18

Simulation of Aircraft Refueling System Activities at Airports and their Optimization

Shantanu Maskeri and M K Hada Defence Institute of Advanced Technology Girinagar, Pune 411025 Maharashtra, India The activities involved in refueling of aircraft at airfields are simulated and optimized using discrete event simulation using AutoMod Simulation software. Animation is done to visualize, understand and optimize the whole system. The refueling of aircraft at airfields is a simple process yet one which requires a high level of coordination, efficiency, cost effectiveness and streamlined procedures. It consists of refueling of aircraft by tankers, topping up of tankers at a filling station/bunk, documentation under supervision to prevent any mistake to avoid any incident or accidents since it is related to aircraft. All this has to be achieved within stringent time restrictions to ensure no delays to aircraft and flight goes as per schedule. This work will help those working in the area of discrete event simulation for resource optimization and can also help people involved in aviation field. The creation of airfield infrastructure is highly time and capital intensive. Thus measures that increase the efficiency of existing facilities are often the only available solution to the need for increased capacities at airfields. Increase in use of airfields by aircraft sometimes necessitates addition of capacities in terms of space, manpower, passenger utilities, aircraft servicing with utmost accuracy. Airfields operations are an ideal application area for simulation. The processes are in a continuous state of change, are complex and stochastic, involve many moving objects, and require a good performance that can be measured in several different performance indicators. To best of our knowledge we have not come across similar work, however many problems of discrete event simulation are report in many journals and books the of which is given in reference. This study can be used to design the refueling facility for higher frequency of aircraft arrival and departure then is an increase in the number of parking bays necessary or will increasing the number of tankers be sufficient? If this be so, then how many tankers and have what capacity should we have keeping the

costs to a minimum? In deciding the quantity of each parameter to achieve the desired result keeping in mind the costs and feasibility. The optimization problem is to achieve a desired performance, say over 99.90% timely refueling of aircraft making minimum changes to the present infrastructure and minimizing the costs.

Simulation of Aircraft Refueling System Activities at Airports and their Optimization


Shantanu Maskeri Defence Institute of Advanced Technology Girinagar, Pune 411025 Maharashtra, India M K Hada Defence Institute of Advanced Technology Girinagar, Pune 411025 Maharashtra, India mkhada@hotmail.com Keywords: - Air Craft, Refueling, AutoMod, Discrete Event Simulation, Optimization. Abstract:- Increase in use of airfields by aircraft necessitates addition of capacities in terms of space, manpower, passenger utilities, aircraft servicing, aircraft availability. The creation of airfield infrastructure is highly time and capital intensive. Thus measures that increase the efficiency of existing facilities are often the only available solution to the need for increased capacities at airfields within limited resources. In this paper work and results of a M Tech dissertation are presented. A simulation model is presented for optimization of resources for aircraft refueling activities involving the use of resources like tankers of different capacities, their crew, supervisors, tanker filling stations, parking areas for tankers, supervisors vehicles. The simulation and modeling is done using AutoMod application software. All the activities mentioned above have different initial and operating costs, which add up to the total cost of the refueling. The problem involves deciding the quantity of each parameter to achieve the desired result keeping in mind the costs and feasibility. Optimization is to achieve a desired performance of 99.90% timely refueling of aircraft making minimum changes to the present infrastructure and minimizing the costs. The simulation is carried to visualize the system which includes realistic features such as routings, dependencies, updating of attributes of aircrafts and resources, capture of relevant statistics, color schemes for easy identification of resources and their animation. The simulation run can be viewed in a 3D

graphical environment to enhance easy understanding and visualization of each activity. Optimization has been carried out using AutoStat, a statistical analysis tool integral to AutoMod. The optimization process is a culmination of a series of several analyses including sensitivity analysis, vary one factor analysis, and vary multiple factor analysis and use of the optimization tool of AutoStat. 1. Introduction

Discrete event simulation is often used to model systems where complex processes are combined with a limited infrastructure capacity. The systems are complex, the numbers of entities involved are high, and the systems are in a continuous state of change. The boundary conditions for solutions are interesting, because solutions have to comply with strict regulations, financial limits, and environmental conditions. The airfield refueling process is in a continuous state of change are complex and stochastic, involve many moving objects and require a good performance that can be measured in several different performance indicators. The refueling of aircraft at airfields is a simple process yet one which requires a high level of coordination, efficiency, cost effectiveness and streamlined procedures. It consists of refueling of aircraft by tankers (trucks), topping up of tankers at ATF (Aviation Turbine Fuel) bunk or a filling station, documentation and supervision of these activities for quality control and error prevention. All this has to be achieved within stringent time restrictions to ensure delays to aircraft. An increase in the number of aircraft operating from an airfield inevitably brings about the need for higher number of resources used in the aircraft refueling activities. These resources, which include parking space, maintenance area, fuel bunks, tanker (trucks), trained crew (manpower) are expensive to obtain and may not be always available. Thus the optimum utilization of existing resources is the best solution. Large and sophisticated models of airport operations already exist such as SIMMOD (www.atac.com) and Preston Airport Solutions (www.preston.net) but they encompass the whole gamut of airport activities like staffing, baggage, check in equipment and are not focused solely on the aircraft refueling activities. This motivated us to take the problem for simulation study and optimization using AutoStat. 4

2.

Model design

To design a simulation model that will realistically depict the aircraft refueling activities on an airfield. The simulation software AutoMod is used. The model incorporates features such as size of tanker, speed of tanker, routing of tanker, numbers of tankers, frequency of arrival of aircraft, size of aircraft, speed of travel of aircraft, capacities of aircraft, refueling rate (pumping), routing of aircraft. To analyze the output data of simulation runs AutoStat software is used that is an add-on statistical analysis tool compatible with AutoMod. 2.1 Refueling Activities The activities are modeled as:

An aircraft lands and taxies to its allotted parking bay. A tanker looking for work is dispatched to aircraft after checking that the balance fuel in the tanker is more than the fuel requirement of the aircraft.

Tanker crew begins coupling the tanker to the aircraft while Shift Officer (SO) is waiting. Refueling of the aircraft is carried out after arrival of the SO. The SO carries out the inspections, refueling is completed and after necessary documentation the SO leaves. The tanker crew carries out decoupling and looks for work at other parking bays. If work is found then they proceed to that bay else the tanker proceeds for topping up.

At the topping bay the tanker crew begins coupling the tanker to the top up bunk while waiting for the SO. The SO arrives, inspects the process and fuel quality. He waits till there are no more tankers waiting for topping up at the bay. Tanker is topped up, the tanker is decoupled and it proceeds to park at the tanker park where it begins to look for work again. Whenever an SO becomes free, he looks for work at both, the aircraft parking bays and the top up bunk and proceeds to the location where he finds work. If he does not find work at either of the areas he parks in the SOs park.

Travel area for tankers and SOs

Tanker Top up Area

Aircrafts parking

SOs/Tanker Park

Taxiing Area

Aircrafts land
Figure 1: Layout of simulated airport area

Aircrafts Take off

Table 1: Operating Parameters Parameter Number of Tankers Number of Filling Station Filling Rate Refueling Rate Number of shifts Shift Timings Description 7 each type
11KLtr, 16 KLtr, and 27 KLtr

Two 1,500-2,000 Ltr/min


Filled up to 95% of capacity

1,500 Ltr/min or 1.5 KLtr/min Three


1st: 7:00 a.m. to 3:00 p.m. 2nd: 3:00 p.m. to 11:00 p.m. 3rd: 11:00 p.m. to 7:00 a.m.

Number of drivers and helpers

7 each in 1st and 2nd shift 6 each in 3rd shift

Each crew consists of a driver, helper and a shift officer for supervision Manpower allocation 1st and 2nd shift 3rd shift Distance:- bunk to aircraft Average speed Travel time

5 shift officers per shift

7 drivers + 7 helpers + 5 shift officers = 19 6 drivers + 6 helpers + 5 shift officers = 17 0.5 km 25 kmph 0.02 hr

Figure 2: View of simulation run 3. Analysis of Simulation Runs

Analysis of the system was carried out using AutoStat, a statistical analysis tool of the AutoMod suite. Simulation runs were carried out for carefully designed experiments and the output of runs were studied and analyzed. The system response or KPI (Key Performance Indicator) considered is the average refueling time i.e. mean of the times taken to refuel each aircraft. The objective of the analysis is to optimize the resources for an average refuel time of 20 to 30 minutes. 3.1 Sensitivity analysis A sensitivity analysis was carried out to test the sensitivity of the system to variations in the parameter values. The parameters varied were: a. b. c. Number of tankers 11 Kltr, 16 Kltr, and 27 Kltr Number of Shift Officers Capacity of filling station/bunk

The bar chart below is plotted for five parameters as shown in Figure 3. The results of simulation runs and conclusions drawn are as follows: Negative effect indicates that the parameter has a reducing effect on the response. The factor quantity of 16 KLtr tankers has the highest effect (-18.05) on reducing the average refuel time followed by quantity of 27 KLtr tankers (-14.4),

quantity of 11 KLtr tankers (-11.3) and quantity of shift officers (-2.87).

The parameter Capacity of filling station/bunk has a negligible effect on the average refuel time.

Figure 3: Sensitivity analysis 3.2 Vary One Factor (VOF) Analysis

This analysis consisted of four sub analyses carried out by varying following parameters each time: a. b. c. Number of 11 KLtr tanker 1 to 5 Number of tanker 16 KLtr and 27 KLtr 1 to 7 Number of Shift Officers 1 to 5 The chart below is plotted for three type of tankers as shown in Figure 4. The results and conclusions drawn are as follows: Average refuel time is very high when number of tanker is 1 and reduces greatly when number of tankers is increased. The flattening of the curves after number of tanker is 3 indicates that the effect of number of tanker above 3 has a lesser effect than that seen when it is increased from 1 to 3.

Average refuel time is least for the parameter number of 16 KLtr tankers followed by that for 27 KLtr tankers and highest for 11 KLtr tankers. This confirms with the results of the sensitivity analysis.

Number of Shift Officer versus refueling time is shown in Figure 5 as number of Shift Officer increases time to refuel decreases.

Summary - VOF Analysis


140 120
Avg Refuel Time

100 80 60 40 20 0 1 2 3 4 5 No of Tankers 6 7

tkr_11_1_no
tkr_16_1_no
tkr_27_1_no

Figure 4: VOF Analysis Tankers

Figure 5: VOF Analysis Shift Officers

10

3.3

Vary Multiple Factors (VMF) Analysis Four parameters were varied between different ranges as follows: Number of 11 KLtr tanker 0 to 4 Number of 16 KLtr tanker 0 to 7 Number of 27 KLtr tanker 1 to 3 Number of Shift Officers 3 to 5

A total of 252 combinations were generated with five runs carried out per combination using Common Random Numbers (CRNs). The use of Common Random Numbers (CRNs) enables comparison of the response between different combinations. The runs produced results in the form of tables of confidence intervals, and summary statistics. It also generates bar graphs and multiple factor graphs. The multiple factor graphs and summary statistics were used to carry out further study and analysis to find out the optimum combination of parameters which give a response (average refuel time) in the desired time bandwidth of 20 to 30 minutes. Two methods were employed for this analysis Intuitive method and Below Average method. 3.4 Intuitive Method This was done using the summary statistics on an Excel sheet. Following filters were applied to reject combinations of the factors: a. Filter 1: - Reject combinations with average refuel time greater than 30 and less than 20 b. c. Filter 2: - Reject combinations with number of shift officers greater than 4 Filter 3: - Reject combinations with total no of tankers greater than 6 and less than 4 These filters yielded the following combinations as shown in Table 2

11

Table 2: Intuitive Method - Stage 1


f_sm_no tkr_11_1_no tkr_16_1_no tkr_27_1_no Total NoofTkrs r_tab_refueltime Std. Dev. Minimum Maximum Median # of Runs Scenario no. 3 0 3 1 4 4 0 3 1 4 3 1 3 1 5 4 1 3 1 5 3 2 3 1 6 3 0 4 1 5 4 0 4 1 5 3 1 4 1 6 4 3 1 2 6 4 2 2 2 6 3 0 3 2 5 4 3 0 3 6

28.93 27.84 23.19 21.53 20.91 22.63 20.73 20.31 28.22 25.76 21.10 25.27 0.40 0.59 1.07 0.19 0.43 0.85 0.72 0.51 2.01 1.01 0.90 0.67 28.36 27.24 21.99 21.29 20.52 21.73 19.69 19.67 25.01 24.51 20.29 24.23 29.48 28.72 24.55 21.78 21.64 23.96 21.47 20.94 30.10 26.97 22.50 25.95 28.95 27.68 22.84 21.49 20.80 22.68 20.58 20.25 29.00 25.65 20.86 25.45 5 5 5 5 5 5 5 5 5 5 5 5 1 2 3 4 5 6 7 8 9 10 11 12

d.

Filter 4: - Reject combinations with same configuration of tankers but with higher number of shift officers. Thus, scenario numbers 2, 4, 7 were rejected. Table 3: Intuitive Method - Stage 2
f_s m _ n o tkr_ 11 _ 1_ n o tkr_ 16 _ 1_ n o tkr_ 27 _ 1_ n o T o ta l N o o fT krs r_ ta b _ re fu e ltim e S td. D ev. M in im u m M a xim u m M e dia n # o f R un s S c e n a rio n o . 3 0 3 1 4 28 .9 3 0 .4 0 28 .3 6 29 .4 8 28 .9 5 5 1 3 1 3 1 5 23 .1 9 1 .07 21 .9 9 24 .5 5 22 .8 4 5 3 3 2 3 1 6 20 .9 1 0.43 20 .5 2 21 .6 4 20 .8 0 5 5 3 0 4 1 5 2 2 .6 3 0.85 2 1 .7 3 2 3 .9 6 2 2 .6 8 5 6 3 1 4 1 6 2 0 .31 0.5 1 1 9 .67 2 0 .94 2 0 .25 5 8 4 3 1 2 6 2 8.22 2 .0 1 2 5.01 3 0.10 2 9.00 5 9 4 2 2 2 6 2 5.7 6 1 .0 1 2 4.5 1 2 6.9 7 2 5.6 5 5 10 3 0 3 2 5 21 .1 0 0 .9 0 20 .2 9 22 .5 0 20 .8 6 5 11 4 3 0 3 6 25 .2 7 0 .67 24 .2 3 25 .9 5 25 .4 5 5 12

e.

Filter 5: - Reject combinations with total number of tankers greater than 5. This resulted in scenario numbers 5, 8, 9, 10 and 12 being rejected, leaving us with the following combinations.

12

Table 4: Intuitive Method - Final Stage


f_sm_no tkr_11_1_no tkr_16_1_no tkr_27_1_no Total NoofTkrs r_tab_refueltime 3 0 3 1 4 28.93 0.40 28.36 29.48 28.95 5 1 3 1 3 1 5 23.19 1.07 21.99 24.55 22.84 5 3 3 0 4 1 5 22.63 0.85 21.73 23.96 22.68 5 6 3 0 3 2 5 21.10 0.90 20.29 22.50 20.86 5 11

Average Std. Dev. Minimum M aximum Median # of Runs

Scenario No.

3.5

Below Average Method

This was also done using the summary statistics on an Excel sheet. A single filter was iteratively applied to select below average values of a fitness function. Since our aim is to yield a combination which gives the least average refuel time for the least number of shift officers and tankers, the fitness function was defined as below Number of shift officers * Total no of tankers * average refuel time The average of the fitness function for all 252 combinations was calculated and those having values below this average were selected. The average was then calculated for the balance combinations and again those with below average values were selected, and so on till the following combinations remained after 5 iterations.

13

Table 5: Below Average Method - 5 Iterations


f_sm_no tkr_11_1_no tkr_16_1_no tkr_27_1_no r_TotalNoofTkrs Average Noof (sm + tkrs) r_tab_refueltime Average Std. Dev. Minimum Maximum Median # of Runs Scenario No No of Scenarios Fitness Function 3 0 3 1 4 7 28.93 0.40 28.36 29.48 28.95 5 46 1 3 1 3 1 5 8 23.19 1.07 21.99 24.55 22.84 5 49 2 3 2 3 1 6 9 20.91 0.43 20.52 21.64 20.80 5 52 3 3 0 4 1 5 8 22.63 0.85 21.73 23.96 22.68 5 61 4 3 1 4 1 6 9 20.31 0.51 19.67 20.94 20.25 5 64 5 3 0 5 1 6 9 19.98 0.57 19.20 20.59 19.89 5 76 6 3 0 3 2 5 8 21.10 0.90 20.29 22.50 20.86 5 166 7 3 1 3 2 6 9 19.43 0.28 19.02 19.72 19.57 5 169 8 3 0 4 2 6 9 19.29 0.42 18.61 19.75 19.38 5 181 9

347.22 347.87 376.34 339.47 365.63 359.68 316.49 349.81 347.21

A filter of rejecting combinations with average refuel time (f_tab_refueltime) below 20 minutes was applied to ensure compliance of results within the desired time bandwidth. Thus scenario numbers 76, 169 and 189 were rejected and the following six combinations were obtained. Table 6: Below Average Method - Final
f_ s m _ n o tk r_ 1 1 _ 1 _ n o tk r_ 1 6 _ 1 _ n o tk r_ 2 7 _ 1 _ n o r_ T o ta lN o o fT k rs N o o f (s m + tk rs ) r_ ta b _ re fu e ltim e 3 0 3 1 4 7 2 8 .9 3 0 .4 0 2 8 .3 6 2 9 .4 8 2 8 .9 5 5 46 1 3 1 3 1 5 8 2 3 .1 9 1 .0 7 2 1 .9 9 2 4 .5 5 2 2 .8 4 5 49 2 3 2 3 1 6 9 2 0 .9 1 0 .4 3 2 0 .5 2 2 1 .6 4 2 0 .8 0 5 52 3 3 0 4 1 5 8 2 2 .6 3 0 .8 5 2 1 .7 3 2 3 .9 6 2 2 .6 8 5 61 4 3 1 4 1 6 9 2 0 .3 1 0 .5 1 1 9 .6 7 2 0 .9 4 2 0 .2 5 5 64 5 3 0 3 2 5 8 2 1 .1 0 0 .9 0 2 0 .2 9 2 2 .5 0 2 0 .8 6 5 166 6

A ve ra g e

A ve ra g e S td . D e v. M in im u m M a xim u m M e d ia n # of R uns

S c e n a rio N o N o o f S c e n a rio s F itn e ss F u n ctio n

3 4 7 .2 2 3 4 7 .8 7 3 7 6 .3 4 3 3 9 .4 7 3 6 5 .6 3 3 1 6 .4 9

Both methods - Iterative method and Below Average method have yielded the same combinations for number of shift office, number of 11 KLtr tankers, number of 16 KLtr tankers and number of 27 KLtr tankers viz. 3-0-3-1, 3-1-3-1,

14

3-0-4-1 and 3-0-3-2 respectively. Below average method yielded two additional combinations that also fit the required criteria. These are 3-2-3-1 and 3-1-4-1. 3.6 Optimization Analysis

The AutoStat software contains an inbuilt optimization tool that was used to obtain an optimum combination of resources to minimize the average refuel time. A total of six analyses were carried out, each with a different fitness function. The ranges of variations of parameter values were kept same as that for the Vary Multiple Factor analysis to enable comparison between the two. Table 6 showing the six analyses carried out and their respective responses. Table 6: Fitness Function Analyses Analysis No Fitness Function Response (Combinations) 3-0-4-1 1 * avg refuel time 1 + 5 * (No of SOs+ No of Tkrs) 3-1-4-1 4-0-4-1 3-0-3-2 3-0-3-3 3-0-4-1 1 * avg refuel time 2 + 6 * (No of SOs+ No of Tkrs) 3-1-4-1 4-0-4-1 3-0-3-2 3-0-3-3 3-0-4-1 1 * avg refuel time 3 + 7 * (No of SOs+ No of Tkrs) 3-1-4-1 4-0-4-1 3-0-3-2 3-0-3-3 3-0-4-1 1 * avg refuel time 4 + 8 * (No of SOs+ No of Tkrs) 3-1-4-1 4-0-4-1 3-0-3-2 3-0-3-3

15

1 * avg refuel time + 5 * No of SOs + 5 1 * No of 11KLtr Tkrs + 2 * No of 16KLtr Tkrs + 3 * No of 27KLtr Tkrs 1 * avg refuel time + 10* No of SOs + 6 2 * No of 11KLtr Tkrs + 4 * No of 16KLtr Tkrs + 6 * No of 27KLtr Tkrs

3-0-4-1 3-1-4-1 3-2-4-1 3-3-4-1 3-4-3-1 3-0-4-1 3-1-4-1 3-2-4-1 3-3-4-1 3-3-3-1

It is concluded from above results that An analysis using a coefficient of 1 for all parameters yielded combinations that had low average refuel times but used large number of resources. Hence a range of coefficients was applied to the resources in two ways o Coefficient was applied to sum of all parameters o Different coefficient was applied to each parameter All the above experiments yielded efficient combinations of parameters. All the analyses using coefficients ranging from 5 to 8 gave the same five combinations. The other two analyses that were carried out using different coefficients for each parameter gave the same set of six combinations. Two combinations 3-0-4-1 and 3-1-4-1 were found in all analyses and hence can be said to be robust combinations.

4.

Results and Conclusion

A sensitivity analysis was carried out to discern the level of sensitivity of the system to different parameters. The Vary One Factor (VOF) analysis helped to know the range of variation of parameters for further analyses. These ranges were: o o o Shift Officers 3 to 5 Number of 11 KLtr tankers 0 to 4 Number of 16 KLtr tankers 0 to 7 16

Number of 27 KLtr tankers 1 to 3

The Vary Multiple Factor (VMF) analysis gave statistics for 252 combinations of parameters. These were further analysed by two methods to obtain combinations with response (average refuel time) within a defined time bandwidth of 20 to 30 minutes and using less number of resources. The optimization analysis was aimed at using the optimization tool of AutoStatTM to obtain optimum combinations of parameters within the defined time bandwidth. It was found that the restriction in defining the fitness function for being unable to define boundary conditions for responses reduces the effectiveness of the optimization module. Therefore the analysis was carried out by changing the fitness function and varying the coefficients of the parameters in the fitness function. It was found that for the given data and parameter specifications, he good combinations of parameters were as under: Table 7: Analysis of Optimization Results S No 1. 2. 3. 4 5 6 7 8 9 10 11 12 No of Shift Officers 3 3 3 3 3 3 3 3 3 3 3 4 No of 11 KLtr Tankers 0 0 1 0 2 1 0 2 3 3 4 0 No of 16 KLtr Tankers 3 4 3 3 3 4 3 4 4 3 3 4 No of 27 KLtr Tankers 1 1 1 2 1 1 3 1 1 1 1 1 Avg Refuel Time 28.93 22.63 23.19 21.10 20.91 20.31 24.33 20.13 20.12 22.31 20.17 20.73

It is concluded that combinations 3-0-3-1, 3-0-4-1, and 3-1-3-1 are the optimum combinations. The balance combinations employ larger numbers of resources without significantly reducing the average refuel time and hence are not optimum.

17

This paper has sought to highlight the use of discrete event simulation to optimize resources. The domain of aircraft refueling system on airports provide a good challenge because of its dynamic nature, varied parameter set and stochastic nature of activities. If accurate data on breakdowns and repair are used the model can also be used for studying the effects of breakdowns and other contingencies. Cost data can used in the model to carry out a cost benefit analysis. Hence the system model can also be used to optimize to cater for contingencies and costs.

References [1] Law Averill M. and Kelton W David 2000: Simulation Modeling and Taha Hamdy A: Operations Research An Introduction, 4th Edition, Analysis, 3rd Edition, McGraw Hill. [2]

Macmillan Publishing. [4] [5] AutoMod Users Guide, Volumes 1 and 2, June 2003. Banks Jerry 2004: Getting Started with AutoMod, 2nd Edition, Brooks

Automation Inc. [6] Edward J. Williams and Ramu Narayanaswamy, Application of Simulation

to Scheduling, Sequencing and Material Handling, Proceedings of the 1997 Winter Simulation Conference. [7] [8] [9] www.airnewzealand.com, Aircraft speeds. www.boeing.com, Aircraft sizes and fuel capacity www.airbus.com, Aircraft sizes and fuel ca pacity

18

Вам также может понравиться