This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Biography for William SwanChief Economist, Seabury-Airline Planning Group. AGIFORS Senior Fellow. ATRG Senior Fellow. Retired Chief Economist for Boeing Commercial Aircraft 1996-2005 Previous to Boeing, worked at American Airlines in Operations Research and Strategic Planning and United Airlines in Research and Development. Areas of work included Yield Management, Fleet Planning, Aircraft Routing, and Crew Scheduling. Also worked for Hull Trading, a major market maker in stock index options, and on the staff at MIT’s Flight Transportation Lab. Education: Master’s, Engineer’s Degree, and Ph. D. at MIT. Bachelor of Science in Aeronautical Engineering at Princeton. Likes dogs and dark beer. ([email protected])
1. Cost vs.. Distance is Linear Illustration Explanation Calibration Why we care
2. Cost vs.. Airplane Size is Linear Illustration Explanation Calibration Why we care
3. Cost vs.. Distance and Size is Planar Why we care
Cost vs. Distance is Linear
• Cost for a single airplane design– Example 737-700
• Cost based on Engineering cost functions– Data from 25-year Boeing OpCost “program”– Divides cost into engineering components
• Fuel, crew, maintenance, ownership• Calibrates components from airline data
– Records of fuel burn– Knowledge of crew pay and work rules– Schedule of recurring maintenance and history of failures– Market Ownership Rents allocated to trips
Engineering Approach is Different
• Not a “black box”– We made what is inside the box
• Not a statistical calibration– Although components are calibrated against data
• Less an overall average– OpCost calibrations based on detail records
• OpCost estimates costs– For standard input cost factors: fuel, labor, capital– Ongoing function recalibration
• This report from 2001 version• 2004 version now in use
We Generate “Perfect” Data Points
• Cost for exactly the same airplane– At different distances
• Each point with identical input costs– Fuel, labor, capital
• Superb spread of data points– Costs at 1000, 1500, 2000, 3000, 4000, 5000km distances– Much larger than spreads of averages for airlines– Comparable overall average distance– Much greater sensitivity to slope
• Objective is to learn the shape of the relationship– Find appropriate algebraic form
• For ratios of costs at different distances
Cost is Linear With Distance737-700 Example
0
5
10
15
20
25
30
0 1000 2000 3000 4000 5000Distance (km)
Trip
Cos
t (in
dex)
Cost DataLinear (Cost Data)
Explanation:Why is Cost Linear With Distance?
• Most costs are per hour or per cycle• Time vs. distance is linear: speed is constant
– (roughly ½ hour plus 500 mph)• Departure/arrival cycle time is about ½ hour• Some costs are allocated
– Allocation is per hour and per cycle– Ownership, for example
• Very small rise in fuel/hour for longer hours• Beyond 8 hours, crew gains 1 or 2 pilots
• All airplanes’ cost vs.. distance was linear• Calibration using 6 “perfect” data points• Least squares• Slopes per seat-km similar• Intercept in equivalent km cost similar• 757s designed for longer hauls• Otherwise comparable capabilities
Why we Care
• Costs Linear with distance means– Average cost is cost at average stage length
• We generally know these data• We can adjust and compare airlines at standard
distance– Cost of an extra stop are separable
• Stop cost independent of where in total distance
• Simplifies Network Costs– Costs are depend on total miles and departures
Costs Are Linear with Airplane Size(Example at 1500 km)
• Costs Linear with Seats means– Average cost is cost at average size
• We generally know these data• We can adjust and compare airlines at a standard size
– Cost of Frequency and Capacity are Separable• Frequency cost is independent of capacity
• Powerful Independence in Network Design– Costs and values of Frequencies– Cost and need for capacity
Calibration for Planar Formula
• NOT Cost = a + bSeats + c*Dist + d*Seats*Dist• Yes:Cost = k * (Seats + a) * (Dist + b) = k*a*b + k*b*seats + k*a*Dist
+ k*Seats*DistNOTE: only 3 degrees of freedom
Why We Care
• Planar function is VERY easy to work with• Decouples frequency, size, distance• Vastly simplifies network design issues• Allows comparison of airline costs after
adjustment for size and stage length• Calibration with broad ranges of size and
distance means slopes are very significant
Calibration Techniques
• Calibrate each airplane vs.. distance– Two variables, k and b
• Calibrate a for least error– Unbiased– Least squared
• Compare to least % error (log form)• Compare to size-first process• Results very similar• Results also similar to 4-variable values
Calibration Formula
Cost = $0.019 * (Seats + 104) * (Dist + 722) Where Cost means total cost 2001US $ per airplane trip,
non-US cost functions.Seats means seat count in standard 2-class regional
density.Dist means airport-pair great circle distance in
kilometers.
One try at “Fair” Relative Seat CountsRegional Configurations