- The document describes a statistical computer experiments approach to solving the airline fleet assignment problem using a two-stage stochastic programming framework.
- In the first stage, crew-compatible aircraft are assigned to flights. In the second stage, individual aircraft are assigned within families based on demand scenarios.
- A Latin hypercube design is used to generate initial crew-compatible allocations for the first stage. For each allocation and demand scenario, a deterministic fleet assignment model is solved. The expected revenue is estimated as the average over scenarios.
Automata Theory Models and Algorithms for Controlling an Aircraft Group in Av...IRJET Journal
This document discusses models and algorithms for controlling aircraft during landing when their positions are unknown due to external factors like weather. It proposes using finite state automata and generalized synchronization methods to model the aircraft system and develop dispatcher instructions to guide the aircraft safely. The problem is formalized and modeled as a finite state machine. Algorithms are described to find synchronizing sequences of commands to move the aircraft system to a safe configuration regardless of their initial positions. Linear finite state machine models are considered, which allows solving the problem using linear algebra methods to obtain short synchronizing sequences. This approach aims to help dispatchers efficiently guide aircraft groups during adverse conditions.
The objective of the project is to predict whether a flight will be delayed or not by studying the various features of the flight and calculating the probability for delay using Bayesian Classification.
This document summarizes a computational analysis of flow behavior over a multi-stage launch vehicle with strap-on boosters. A structured grid and commercial CFD software Fluent were used to model the 2D flow field. Finer meshes were generated near vehicle surfaces to capture shocks accurately. Both Euler and Navier-Stokes solvers were tested using various turbulence models. Results aimed to understand complex flow interactions over the multi-component vehicle configuration.
This document discusses aerodynamic modeling and simulation of aircraft at high angles of attack. It outlines some of the challenges, including developing accurate mathematical models from wind tunnel data and flight tests. Nonlinear effects like separated and unsteady flow require dynamic modeling approaches. Qualitative analysis of the nonlinear dynamics can reveal critical flight conditions like departures and spins. Flight simulations are used alongside flight tests to evaluate control laws and train pilots. Accurate modeling of phenomena like aerodynamic asymmetry is important for understanding spin behavior.
International Refereed Journal of Engineering and Science (IRJES)irjes
1. The document presents collision risk models for free flight based on course adjustment and velocity adjustment to resolve conflicts between aircraft.
2. The models establish probability distributions for aircraft distances based on their relative motion and incorporate factors like separation standards and positioning error.
3. Collision risk is calculated as the probability of overlapping protection zones for aircraft in both horizontal and vertical planes, and an example is analyzed to demonstrate the models.
International Refereed Journal of Engineering and Science (IRJES)irjes
International Refereed Journal of Engineering and Science (IRJES) is a leading international journal for publication of new ideas, the state of the art research results and fundamental advances in all aspects of Engineering and Science. IRJES is a open access, peer reviewed international journal with a primary objective to provide the academic community and industry for the submission of half of original research and applications
Automata Theory Models and Algorithms for Controlling an Aircraft Group in Av...IRJET Journal
This document discusses models and algorithms for controlling aircraft during landing when their positions are unknown due to external factors like weather. It proposes using finite state automata and generalized synchronization methods to model the aircraft system and develop dispatcher instructions to guide the aircraft safely. The problem is formalized and modeled as a finite state machine. Algorithms are described to find synchronizing sequences of commands to move the aircraft system to a safe configuration regardless of their initial positions. Linear finite state machine models are considered, which allows solving the problem using linear algebra methods to obtain short synchronizing sequences. This approach aims to help dispatchers efficiently guide aircraft groups during adverse conditions.
The objective of the project is to predict whether a flight will be delayed or not by studying the various features of the flight and calculating the probability for delay using Bayesian Classification.
This document summarizes a computational analysis of flow behavior over a multi-stage launch vehicle with strap-on boosters. A structured grid and commercial CFD software Fluent were used to model the 2D flow field. Finer meshes were generated near vehicle surfaces to capture shocks accurately. Both Euler and Navier-Stokes solvers were tested using various turbulence models. Results aimed to understand complex flow interactions over the multi-component vehicle configuration.
This document discusses aerodynamic modeling and simulation of aircraft at high angles of attack. It outlines some of the challenges, including developing accurate mathematical models from wind tunnel data and flight tests. Nonlinear effects like separated and unsteady flow require dynamic modeling approaches. Qualitative analysis of the nonlinear dynamics can reveal critical flight conditions like departures and spins. Flight simulations are used alongside flight tests to evaluate control laws and train pilots. Accurate modeling of phenomena like aerodynamic asymmetry is important for understanding spin behavior.
International Refereed Journal of Engineering and Science (IRJES)irjes
1. The document presents collision risk models for free flight based on course adjustment and velocity adjustment to resolve conflicts between aircraft.
2. The models establish probability distributions for aircraft distances based on their relative motion and incorporate factors like separation standards and positioning error.
3. Collision risk is calculated as the probability of overlapping protection zones for aircraft in both horizontal and vertical planes, and an example is analyzed to demonstrate the models.
International Refereed Journal of Engineering and Science (IRJES)irjes
International Refereed Journal of Engineering and Science (IRJES) is a leading international journal for publication of new ideas, the state of the art research results and fundamental advances in all aspects of Engineering and Science. IRJES is a open access, peer reviewed international journal with a primary objective to provide the academic community and industry for the submission of half of original research and applications
This 3-sentence summary provides the high-level information from the document:
The document describes a zonal approach for simulating external vehicle aerodynamics using a panel method for the inviscid flow field, a boundary layer method for viscous effects, and a wake model to simulate separated flow. The methods are coupled iteratively, with the potential flow solved using a panel method, boundary layer separation determining the starting point for the wake model, and the wake influencing subsequent boundary layer and potential flow solutions. The approach aims to provide a fast simulation tool for vehicle aerodynamics in early development phases when analyzing multiple variants.
This document discusses autonomous formation flight control for aerial refueling missions. It models the dynamics of a KC-130J tanker and F-16 receiver in close formation flight. A control strategy is designed using a rotating reference frame attached to the wingman aircraft. The strategy aims to control the wingman's position relative to the leader through state error feedback between the aircraft. Specifically, it controls the lateral and forward distance between the aircraft in the level flight plane, as well as the vertical distance. The control system design decomposes the problem into inner loop attitude control and outer loop trajectory control of the formation geometry.
An Interactive Approach For Hierarchical Analysis Of Helicopter Logistics In ...Allison Thompson
This document summarizes a study that develops mathematical models to aid in planning helicopter missions for disaster relief operations. The planning problem is broken down into two hierarchical sub-problems - a top-level problem that determines tactical decisions like fleet composition and crew assignment, and a base-level problem that handles operational routing and loading. An iterative procedure coordinates the two levels to transfer information and improve solutions. Multiple objectives require a multi-criteria analysis and interactive process to evaluate alternative solutions.
An independent flush air data system (FADS) was designed and installed on a Boeing B707 aircraft to provide real-time air data for flight testing without interfering with the aircraft. The FADS used pressures measured by sensors on the nose radome to compute air data parameters like angle of attack and sideslip through a mathematical model. The model was calibrated and computations were performed in real-time at a rate of 2.5 Hz during flight tests to generate air data for simulator development. Post-flight processing recomputed the air data at 50 Hz for further analysis.
This document discusses an algorithm for estimating the safe maneuvering envelope of damaged aircraft using optimal control and reachability analysis. The algorithm performs a robust reachability analysis by formulating the problem as an optimal control problem and solving the resulting Hamilton-Jacobi-Bellman equation using level set methods. The approach is applied to an aircraft example involving structural damage. Validation tests confirmed the approach can estimate the safe maneuvering envelope for damaged aircraft.
This document describes a Kriging component for spatial interpolation of climatological variables in the OMS modeling framework. Kriging is a geostatistical technique that interpolates values based on measured data and the spatial autocorrelation between data points. The component implements ordinary and detrended Kriging algorithms using 10 semivariogram models. It can interpolate both raster and point data and outputs the interpolated climatological variable values. Links are provided for downloading the component code, data, and OMS project files needed to run the interpolation.
FLIGHT TRAJECTORY RECREATION AND PLAYBACK SYSTEM OF AERIAL MISSION BASED ON O...cscpconf
Recreation of flight trajectory is important among research areas. The design of a flight
trajectory recreation and playback system is presented in this paper. Rather than transferring
the flight data to diagram, graph and table, flight data is visualized on the 3D global of
ossimPlanet. ossimPlanet is an open-source 3D global geo-spatial viewer and the system
realization is based on analysis it. Users are allowed to choose their interested flight of aerial
mission. The aerial photographs and corresponding configuration files in which flight data is
included would be read in. And the flight statuses would be stored. The flight trajectory is then
recreated. Users can view the photographs and flight trajectory marks on the correct positions
of 3D global. The scene along flight trajectory is also simulated at the plane’s eye point. This
paper provides a more intuitive way for recreation of flight trajectory. The cost is decreased
remarkably and security is ensured by secondary development on open-source platform.
Airline Fleet Assignment And Schedule PlanningAndrew Molina
The document discusses airline fleet assignment and schedule planning. It describes how optimizing fleet assignment and scheduling is important for airline efficiency and profitability. It then reviews several models for integrated fleet assignment and scheduling, including the basic fleet assignment model, and discusses techniques for solving these optimization problems.
NAME _____________________ (317 points possible) earned ________ .docxhallettfaustina
NAME _____________________ (317 points possible) earned ________
Page 1
Basic Aviation Cross
Country Planning Assignment
Due date May 6, 2021
Objective: Plan a cross-country flight in a Cessna 172 from Will Rogers Airport (OKC)
in Oklahoma City to Cox Airport at Paris, Texas (PRX) to Addison Airport (ADS) at
Dallas, Texas. Use pre-flight planning resources available within to ensure that the flights
can be conducted safely and legally. Departure date 4/7/21.
GIVEN:
Use Dallas-Ft. Worth sectional chart
OKC departure time -- 1900Z
True Airspeed (TAS) -- 110 kts
Distances -- measure in nautical miles
Fuel Consumption -- 11.0 gallons per hour
Pilot -- (name – Ace) 190 lbs. Private pilot 17 years old
Passenger -- (name – Joker) 150 lbs. Student pilot 16 years old
Luggage -- 20 lbs.
Aircraft empty weight -- 1350 lbs. & Moment 51.5
Fuel Tanks Full – (48 gal) (Min. octane – 100) 12/7/19 at 1830Z
Cessna 172 -- single-engine, 150 hp, fixed pitch prop, flaps, non-retracting gear
What is the
youngest legal age
Ace (as PIC) could be on this x-country? ___________
The minimum
medical certificate
required for Ace is a _____________ Class.
What telephone number will Ace use to call for a
Flight Service Station briefing?
___ _____ ______ _________
When Ace received an initial FSS briefing more than six hours prior to departure, he
requested an ________ briefing. A complete briefing an hour before the proposed
departure is a _________briefing. Another call to the FSS just minutes prior
to the flight for a specific segment is referred to as an ___________ briefing.
A
Special VFR Clearance
can be requested by Ace at night. T F
Ace’s proposed departure time from OKC is 1900Z, convert that to local time. _________ central standard time _______ 24-hour clock time
TFR stands for _____________ ____________ ___________
Can Ace knowingly violate a FAR during an emergency? Yes No
Find weather surface observations (
METAR
) to determine if the airports
Page 2
are currently reporting VFR conditions.
What are the
cloud base(AGL), visibility(SM ) and visibility restriction
at:
OKC --
PRX --
ADS --
METAR KOKC 301245Z 28009KT / 17SM / SKC / 27/11 / A2970
METAR KPRX 301245Z 21012KT / 3SM BR/ FEW 004 SCT120 / 27/11 / A2968
METAR KADS 301246Z 18011G25 / 5SM -RA / BKN11 / 28/24 / A2974
Based on these METARs, do
VFR conditions exist
? Y N (circle one)
PILOT REPORTS (PIREP)
(UA) OKC UA /OV OKC/ TM 1917/ FL 045/ TP C172/ TA 11/ WV 18037/
TB Occasional moderate turbulence surface to 50. Smooth above.
1. What
kind of turbulence
might Ace
expect while departing OKC to PRX? __________________
2. Ace may have to slow the airplane to
what V-speed?
_______________
3. Why?
4. If Ace encountered severe turbulence, which
flight
condition
should he attempt to maintain? _______________________
5. Determine the following from the
PIREP
above:
Altitude - ____.
This document discusses using a multiplicative time series model (p,d,q)x(P,D,Q)s to forecast air transport demand.
It analyzes passenger data from Yemenia Airlines in Yemen over many years. The data shows seasonality with a 12-month period. To account for this, the model includes polynomials in the backward shift operator B with period s=12.
The model is fitted to the logarithms of the monthly passenger totals, which minimizes residuals. This confirms the appropriateness of a logarithmic transformation for this time series data. The model provides a means of linking observations within and across years to generate forecasts of air transport demand.
This document summarizes the design of the Jayhawk Economic Turboprop Transport (J.E.T.T.) aircraft for the 2013-2014 AIAA Undergraduate Team Aircraft Design Competition. It provides the mission specifications for a new regional turboprop airliner, including a 400 nautical mile economic mission carrying 75 passengers and a 1,600 nautical mile design mission carrying 67 passengers. It then describes the initial configurations considered and presents analysis on determining design parameters through statistical modeling techniques. Key aspects of the preliminary aircraft design are summarized, including engine selection, wing design, and layout of major systems.
Flight trajectory recreation and playback system of aerial mission based on o...csandit
Recreation of flight trajectory is important among research areas. The design of a flight
trajectory recreation and playback system is presented in this paper. Rather than transferring
the flight data to diagram, graph and table, flight data is visualized on the 3D global of
ossimPlanet. ossimPlanet is an open-source 3D global geo-spatial viewer and the system
realization is based on analysis it. Users are allowed to choose their interested flight of aerial
mission. The aerial photographs and corresponding configuration files in which flight data is
included would be read in. And the flight statuses would be stored. The flight trajectory is then
recreated. Users can view the photographs and flight trajectory marks on the correct positions
of 3D global. The scene along flight trajectory is also simulated at the plane’s eye point. This
paper provides a more intuitive way for recreation of flight trajectory. The cost is decreased
remarkably and security is ensured by secondary development on open-source platform.
This document summarizes the design of an optimal feedback control system for the final landing phase of an aircraft. It describes modeling the aircraft dynamics using a linearized transfer function relating elevator deflection and pitch rate. The design uses the parametric expansion method of optimization theory to synthesize time-varying feedback controls by minimizing an error index formulated from performance requirements. Several candidate control systems are compared by simulating aircraft landing trajectories.
A Novel Technique in Software Engineering for Building Scalable Large Paralle...Eswar Publications
Parallel processing is the only alternative for meeting computational demand of scientific and technological advancement. Yet first few parallelized versions of a large application code- in the present case-a meteorological Global Circulation Model- are not usually optimal or efficient. Large size and complexity of the code cause making changes for efficient parallelization and further validation difficult. The paper presents some novel techniques to enable change of parallelization strategy keeping the correctness of the code under control throughout the modification.
This article considers different approaches for autopilot controller gain values adjustment. The correct autopilot
performance is tested using modeling methods. A variant of land-based autopilot is considered. Examined are
scenarios of UAV airplanes in level flight. The latter are applicable to tasks such as remote sensing, controlled
area surveillance, etc.
Triton UAS Technical Design Paper 2020-2021KennyPham19
The document summarizes the design of an unmanned aerial system called the Swallow created by Triton Unmanned Aerial Systems for the 2021 AUVSI student competition. Key aspects of the design include the fixed-wing aircraft chosen, the autopilot and ground control system used, computer vision systems for object detection and classification from images, and algorithms for autonomous path planning that avoids obstacles while completing mission tasks. The team's goal is to have a system that can autonomously identify and classify ground targets from images taken during flight.
М.Г.Гоман, А.В.Храмцовский (1998) - Использование методов непрерывного продол...Project KRIT
М.Г.Гоман, А.В.Храмцовский "Использование методов непрерывного продолжения решений и бифуркационного анализа для синтеза систем управления", Phil.Trans.R.Soc.Lond. A (1998) 356, 2277-2295
M.G.Goman and A.V.Khramtsovsky "Application of continuation and bifurcation methods to the design of control systems", Phil.Trans.R.Soc.Lond. A (1998) 356, 2277-2295
In this paper the continuation and bifurcation methods are applied to aircraft nonlinear control design problems. The search for the recovery control from spin regimes is based on the minimization of an energy-like scalar function constrained by the aircraft's equilibria conditions. The design of a global stability augmentation system for severe wing-rock motion is performed by using bifurcation diagrams for equilibrium and periodical modes. The nonlinear control law, which totally suppresses wing-rock motion, is derived, taking into account both local stability characteristics of aircraft equilibrium states and domains of attraction, along with the requirement that all other attractors be eliminated.
This paper proposes a method to calculate a flight cost of an unmanned aerial vehicle (UAV) considering its change of heading angle though there are many reasons that cause the energy consumption. The proposed approach demonstrates that when a UAV moves from a starting position/point to a target/goal position/point, if the number of obstacle increases, the number of heading change would also increase. As a result, it raises the energy consumption of the UAV. It also shows that the magnitude of heading change would affect the energy consumption proportionally. The theoretical analysis as well as the simulation outcome proves the usefulness of the proposed technique.
j2 Universal - Modelling and Tuning Braking CharacteristicsJohn Jeffery
1) The document outlines tuning the braking characteristics of an aircraft model using flight test data. Key steps included generating test cases from the data, reconstructing the data for analysis, and performing a re-prediction analysis to identify discrepancies.
2) Regression analysis on acceleration errors was used to derive an improved braking friction coefficient table. Additional tuning of thrust reverser dynamics provided better matching of deceleration profiles.
3) Sanity checks confirmed the friction coefficients derived were feasible. The tuned model matched the flight data to tolerances required for pilot training simulation.
Design and Analysis of Solar Powered RC Aircrafttheijes
The International Journal of Engineering & Science is aimed at providing a platform for researchers, engineers, scientists, or educators to publish their original research results, to exchange new ideas, to disseminate information in innovative designs, engineering experiences and technological skills. It is also the Journal's objective to promote engineering and technology education. All papers submitted to the Journal will be blind peer-reviewed. Only original articles will be published.
How To Write An Empathy Essay By Jones Jessica - IGina Rizzo
The document provides instructions for creating an account and submitting assignment requests to HelpWriting.net, an online writing assistance service. It describes a bidding process where writers can submit proposals, and customers can select a writer and provide instructions for their assignment. The document outlines steps for customers to receive their paper, provide feedback, and request revisions to ensure satisfaction with the writing help received.
Rocket Outer Space Lined Paper Lined Paper, WritinGina Rizzo
The document provides instructions for requesting a paper writing service from the website HelpWriting.net. It outlines a 5-step process: 1) Create an account with a password and email. 2) Complete a form with paper details, sources, and deadline. 3) Review writer bids and qualifications and place a deposit. 4) Review the paper and authorize payment if satisfied. 5) Request revisions until fully satisfied, with a refund option for plagiarism.
Weitere ähnliche Inhalte
Ähnlich wie A Statistical Computer Experiments Approach To Airline Fleet Assignment
This 3-sentence summary provides the high-level information from the document:
The document describes a zonal approach for simulating external vehicle aerodynamics using a panel method for the inviscid flow field, a boundary layer method for viscous effects, and a wake model to simulate separated flow. The methods are coupled iteratively, with the potential flow solved using a panel method, boundary layer separation determining the starting point for the wake model, and the wake influencing subsequent boundary layer and potential flow solutions. The approach aims to provide a fast simulation tool for vehicle aerodynamics in early development phases when analyzing multiple variants.
This document discusses autonomous formation flight control for aerial refueling missions. It models the dynamics of a KC-130J tanker and F-16 receiver in close formation flight. A control strategy is designed using a rotating reference frame attached to the wingman aircraft. The strategy aims to control the wingman's position relative to the leader through state error feedback between the aircraft. Specifically, it controls the lateral and forward distance between the aircraft in the level flight plane, as well as the vertical distance. The control system design decomposes the problem into inner loop attitude control and outer loop trajectory control of the formation geometry.
An Interactive Approach For Hierarchical Analysis Of Helicopter Logistics In ...Allison Thompson
This document summarizes a study that develops mathematical models to aid in planning helicopter missions for disaster relief operations. The planning problem is broken down into two hierarchical sub-problems - a top-level problem that determines tactical decisions like fleet composition and crew assignment, and a base-level problem that handles operational routing and loading. An iterative procedure coordinates the two levels to transfer information and improve solutions. Multiple objectives require a multi-criteria analysis and interactive process to evaluate alternative solutions.
An independent flush air data system (FADS) was designed and installed on a Boeing B707 aircraft to provide real-time air data for flight testing without interfering with the aircraft. The FADS used pressures measured by sensors on the nose radome to compute air data parameters like angle of attack and sideslip through a mathematical model. The model was calibrated and computations were performed in real-time at a rate of 2.5 Hz during flight tests to generate air data for simulator development. Post-flight processing recomputed the air data at 50 Hz for further analysis.
This document discusses an algorithm for estimating the safe maneuvering envelope of damaged aircraft using optimal control and reachability analysis. The algorithm performs a robust reachability analysis by formulating the problem as an optimal control problem and solving the resulting Hamilton-Jacobi-Bellman equation using level set methods. The approach is applied to an aircraft example involving structural damage. Validation tests confirmed the approach can estimate the safe maneuvering envelope for damaged aircraft.
This document describes a Kriging component for spatial interpolation of climatological variables in the OMS modeling framework. Kriging is a geostatistical technique that interpolates values based on measured data and the spatial autocorrelation between data points. The component implements ordinary and detrended Kriging algorithms using 10 semivariogram models. It can interpolate both raster and point data and outputs the interpolated climatological variable values. Links are provided for downloading the component code, data, and OMS project files needed to run the interpolation.
FLIGHT TRAJECTORY RECREATION AND PLAYBACK SYSTEM OF AERIAL MISSION BASED ON O...cscpconf
Recreation of flight trajectory is important among research areas. The design of a flight
trajectory recreation and playback system is presented in this paper. Rather than transferring
the flight data to diagram, graph and table, flight data is visualized on the 3D global of
ossimPlanet. ossimPlanet is an open-source 3D global geo-spatial viewer and the system
realization is based on analysis it. Users are allowed to choose their interested flight of aerial
mission. The aerial photographs and corresponding configuration files in which flight data is
included would be read in. And the flight statuses would be stored. The flight trajectory is then
recreated. Users can view the photographs and flight trajectory marks on the correct positions
of 3D global. The scene along flight trajectory is also simulated at the plane’s eye point. This
paper provides a more intuitive way for recreation of flight trajectory. The cost is decreased
remarkably and security is ensured by secondary development on open-source platform.
Airline Fleet Assignment And Schedule PlanningAndrew Molina
The document discusses airline fleet assignment and schedule planning. It describes how optimizing fleet assignment and scheduling is important for airline efficiency and profitability. It then reviews several models for integrated fleet assignment and scheduling, including the basic fleet assignment model, and discusses techniques for solving these optimization problems.
NAME _____________________ (317 points possible) earned ________ .docxhallettfaustina
NAME _____________________ (317 points possible) earned ________
Page 1
Basic Aviation Cross
Country Planning Assignment
Due date May 6, 2021
Objective: Plan a cross-country flight in a Cessna 172 from Will Rogers Airport (OKC)
in Oklahoma City to Cox Airport at Paris, Texas (PRX) to Addison Airport (ADS) at
Dallas, Texas. Use pre-flight planning resources available within to ensure that the flights
can be conducted safely and legally. Departure date 4/7/21.
GIVEN:
Use Dallas-Ft. Worth sectional chart
OKC departure time -- 1900Z
True Airspeed (TAS) -- 110 kts
Distances -- measure in nautical miles
Fuel Consumption -- 11.0 gallons per hour
Pilot -- (name – Ace) 190 lbs. Private pilot 17 years old
Passenger -- (name – Joker) 150 lbs. Student pilot 16 years old
Luggage -- 20 lbs.
Aircraft empty weight -- 1350 lbs. & Moment 51.5
Fuel Tanks Full – (48 gal) (Min. octane – 100) 12/7/19 at 1830Z
Cessna 172 -- single-engine, 150 hp, fixed pitch prop, flaps, non-retracting gear
What is the
youngest legal age
Ace (as PIC) could be on this x-country? ___________
The minimum
medical certificate
required for Ace is a _____________ Class.
What telephone number will Ace use to call for a
Flight Service Station briefing?
___ _____ ______ _________
When Ace received an initial FSS briefing more than six hours prior to departure, he
requested an ________ briefing. A complete briefing an hour before the proposed
departure is a _________briefing. Another call to the FSS just minutes prior
to the flight for a specific segment is referred to as an ___________ briefing.
A
Special VFR Clearance
can be requested by Ace at night. T F
Ace’s proposed departure time from OKC is 1900Z, convert that to local time. _________ central standard time _______ 24-hour clock time
TFR stands for _____________ ____________ ___________
Can Ace knowingly violate a FAR during an emergency? Yes No
Find weather surface observations (
METAR
) to determine if the airports
Page 2
are currently reporting VFR conditions.
What are the
cloud base(AGL), visibility(SM ) and visibility restriction
at:
OKC --
PRX --
ADS --
METAR KOKC 301245Z 28009KT / 17SM / SKC / 27/11 / A2970
METAR KPRX 301245Z 21012KT / 3SM BR/ FEW 004 SCT120 / 27/11 / A2968
METAR KADS 301246Z 18011G25 / 5SM -RA / BKN11 / 28/24 / A2974
Based on these METARs, do
VFR conditions exist
? Y N (circle one)
PILOT REPORTS (PIREP)
(UA) OKC UA /OV OKC/ TM 1917/ FL 045/ TP C172/ TA 11/ WV 18037/
TB Occasional moderate turbulence surface to 50. Smooth above.
1. What
kind of turbulence
might Ace
expect while departing OKC to PRX? __________________
2. Ace may have to slow the airplane to
what V-speed?
_______________
3. Why?
4. If Ace encountered severe turbulence, which
flight
condition
should he attempt to maintain? _______________________
5. Determine the following from the
PIREP
above:
Altitude - ____.
This document discusses using a multiplicative time series model (p,d,q)x(P,D,Q)s to forecast air transport demand.
It analyzes passenger data from Yemenia Airlines in Yemen over many years. The data shows seasonality with a 12-month period. To account for this, the model includes polynomials in the backward shift operator B with period s=12.
The model is fitted to the logarithms of the monthly passenger totals, which minimizes residuals. This confirms the appropriateness of a logarithmic transformation for this time series data. The model provides a means of linking observations within and across years to generate forecasts of air transport demand.
This document summarizes the design of the Jayhawk Economic Turboprop Transport (J.E.T.T.) aircraft for the 2013-2014 AIAA Undergraduate Team Aircraft Design Competition. It provides the mission specifications for a new regional turboprop airliner, including a 400 nautical mile economic mission carrying 75 passengers and a 1,600 nautical mile design mission carrying 67 passengers. It then describes the initial configurations considered and presents analysis on determining design parameters through statistical modeling techniques. Key aspects of the preliminary aircraft design are summarized, including engine selection, wing design, and layout of major systems.
Flight trajectory recreation and playback system of aerial mission based on o...csandit
Recreation of flight trajectory is important among research areas. The design of a flight
trajectory recreation and playback system is presented in this paper. Rather than transferring
the flight data to diagram, graph and table, flight data is visualized on the 3D global of
ossimPlanet. ossimPlanet is an open-source 3D global geo-spatial viewer and the system
realization is based on analysis it. Users are allowed to choose their interested flight of aerial
mission. The aerial photographs and corresponding configuration files in which flight data is
included would be read in. And the flight statuses would be stored. The flight trajectory is then
recreated. Users can view the photographs and flight trajectory marks on the correct positions
of 3D global. The scene along flight trajectory is also simulated at the plane’s eye point. This
paper provides a more intuitive way for recreation of flight trajectory. The cost is decreased
remarkably and security is ensured by secondary development on open-source platform.
This document summarizes the design of an optimal feedback control system for the final landing phase of an aircraft. It describes modeling the aircraft dynamics using a linearized transfer function relating elevator deflection and pitch rate. The design uses the parametric expansion method of optimization theory to synthesize time-varying feedback controls by minimizing an error index formulated from performance requirements. Several candidate control systems are compared by simulating aircraft landing trajectories.
A Novel Technique in Software Engineering for Building Scalable Large Paralle...Eswar Publications
Parallel processing is the only alternative for meeting computational demand of scientific and technological advancement. Yet first few parallelized versions of a large application code- in the present case-a meteorological Global Circulation Model- are not usually optimal or efficient. Large size and complexity of the code cause making changes for efficient parallelization and further validation difficult. The paper presents some novel techniques to enable change of parallelization strategy keeping the correctness of the code under control throughout the modification.
This article considers different approaches for autopilot controller gain values adjustment. The correct autopilot
performance is tested using modeling methods. A variant of land-based autopilot is considered. Examined are
scenarios of UAV airplanes in level flight. The latter are applicable to tasks such as remote sensing, controlled
area surveillance, etc.
Triton UAS Technical Design Paper 2020-2021KennyPham19
The document summarizes the design of an unmanned aerial system called the Swallow created by Triton Unmanned Aerial Systems for the 2021 AUVSI student competition. Key aspects of the design include the fixed-wing aircraft chosen, the autopilot and ground control system used, computer vision systems for object detection and classification from images, and algorithms for autonomous path planning that avoids obstacles while completing mission tasks. The team's goal is to have a system that can autonomously identify and classify ground targets from images taken during flight.
М.Г.Гоман, А.В.Храмцовский (1998) - Использование методов непрерывного продол...Project KRIT
М.Г.Гоман, А.В.Храмцовский "Использование методов непрерывного продолжения решений и бифуркационного анализа для синтеза систем управления", Phil.Trans.R.Soc.Lond. A (1998) 356, 2277-2295
M.G.Goman and A.V.Khramtsovsky "Application of continuation and bifurcation methods to the design of control systems", Phil.Trans.R.Soc.Lond. A (1998) 356, 2277-2295
In this paper the continuation and bifurcation methods are applied to aircraft nonlinear control design problems. The search for the recovery control from spin regimes is based on the minimization of an energy-like scalar function constrained by the aircraft's equilibria conditions. The design of a global stability augmentation system for severe wing-rock motion is performed by using bifurcation diagrams for equilibrium and periodical modes. The nonlinear control law, which totally suppresses wing-rock motion, is derived, taking into account both local stability characteristics of aircraft equilibrium states and domains of attraction, along with the requirement that all other attractors be eliminated.
This paper proposes a method to calculate a flight cost of an unmanned aerial vehicle (UAV) considering its change of heading angle though there are many reasons that cause the energy consumption. The proposed approach demonstrates that when a UAV moves from a starting position/point to a target/goal position/point, if the number of obstacle increases, the number of heading change would also increase. As a result, it raises the energy consumption of the UAV. It also shows that the magnitude of heading change would affect the energy consumption proportionally. The theoretical analysis as well as the simulation outcome proves the usefulness of the proposed technique.
j2 Universal - Modelling and Tuning Braking CharacteristicsJohn Jeffery
1) The document outlines tuning the braking characteristics of an aircraft model using flight test data. Key steps included generating test cases from the data, reconstructing the data for analysis, and performing a re-prediction analysis to identify discrepancies.
2) Regression analysis on acceleration errors was used to derive an improved braking friction coefficient table. Additional tuning of thrust reverser dynamics provided better matching of deceleration profiles.
3) Sanity checks confirmed the friction coefficients derived were feasible. The tuned model matched the flight data to tolerances required for pilot training simulation.
Design and Analysis of Solar Powered RC Aircrafttheijes
The International Journal of Engineering & Science is aimed at providing a platform for researchers, engineers, scientists, or educators to publish their original research results, to exchange new ideas, to disseminate information in innovative designs, engineering experiences and technological skills. It is also the Journal's objective to promote engineering and technology education. All papers submitted to the Journal will be blind peer-reviewed. Only original articles will be published.
Ähnlich wie A Statistical Computer Experiments Approach To Airline Fleet Assignment (20)
How To Write An Empathy Essay By Jones Jessica - IGina Rizzo
The document provides instructions for creating an account and submitting assignment requests to HelpWriting.net, an online writing assistance service. It describes a bidding process where writers can submit proposals, and customers can select a writer and provide instructions for their assignment. The document outlines steps for customers to receive their paper, provide feedback, and request revisions to ensure satisfaction with the writing help received.
Rocket Outer Space Lined Paper Lined Paper, WritinGina Rizzo
The document provides instructions for requesting a paper writing service from the website HelpWriting.net. It outlines a 5-step process: 1) Create an account with a password and email. 2) Complete a form with paper details, sources, and deadline. 3) Review writer bids and qualifications and place a deposit. 4) Review the paper and authorize payment if satisfied. 5) Request revisions until fully satisfied, with a refund option for plagiarism.
1. The document discusses a 5-step process for getting college research paper writing assistance from HelpWriting.net, including creating an account, submitting a request, reviewing bids from writers, ensuring paper meets expectations, and requesting revisions.
2. Key aspects include completing a 10-minute order form, choosing a writer based on qualifications and reviews, authorizing payment upon satisfaction, and getting free revisions or a refund if needed.
3. HelpWriting.net uses a bidding system where writers submit bids for requests, and clients can choose a writer and start the assignment process by placing a deposit.
Research Paper Executive Summary How To WriteGina Rizzo
The document provides instructions for writing an executive summary for a research paper in 5 steps: 1) Create an account; 2) Complete a form with paper details; 3) Review writer bids and choose one; 4) Review the paper and authorize payment; 5) Request revisions to ensure satisfaction. It emphasizes providing clear instructions to writers and choosing a writer based on qualifications to receive an original, high-quality paper.
The document provides instructions for creating an account and submitting an assignment request on the HelpWriting.net website. It involves 5 steps: 1) Create an account by providing a password and email. 2) Complete a form with assignment details, sources, and deadline. 3) Review bids from writers and choose one. 4) Review the completed paper and authorize payment. 5) Request revisions to ensure satisfaction, with a refund option for plagiarism. The process aims to match clients with qualified writers to complete assignments for a fee.
The document provides instructions for creating an account on HelpWriting.net in order to request writing assistance services such as paper writing or revisions. It explains that writers will bid on the request and the client can choose a writer, make a deposit, and the writer will start on the assignment. It also notes that clients can request revisions to ensure satisfaction and will receive a full refund if the work is plagiarized.
Writing A Personal Letter - MakeMyAssignments BlogGina Rizzo
The document provides instructions for requesting writing assistance from HelpWriting.net. It outlines a 5-step process: 1) Create an account with a password and email. 2) Complete a 10-minute order form providing instructions, sources, and deadline. 3) Review bids from writers and choose one based on qualifications. 4) Review the completed paper and authorize payment if satisfied. 5) Request revisions to ensure satisfaction, and HelpWriting.net offers refunds for plagiarized work.
How To Write Better Essays Pdf - BooksFreeGina Rizzo
The document provides instructions for requesting writing assistance from HelpWriting.net. It outlines a 5-step process: 1) Create an account with a password and email. 2) Complete a 10-minute order form providing instructions, sources, and deadline. 3) Review bids from writers and select one based on qualifications. 4) Review the completed paper and authorize payment. 5) Request revisions to ensure satisfaction, with a full refund option for plagiarized work. The service aims to fully meet customer needs through original, high-quality content.
This document summarizes the life and reign of Ivan the Terrible, ruler of Russia. It describes his complicated upbringing and early life experiences, including the murder of his father. As Tsar, Ivan initially ruled capably and consolidated Russia, but he grew increasingly paranoid and distrustful later in life after the deaths of his beloved wife Anastasia and son Dmitri, which sent him into a spiral of erratic and violent behavior. The summary captures the complexity of Ivan's character and reign as both a capable ruler but also a man whose mental state deteriorated later in life, leading to his notorious and brutal acts.
The document provides instructions for requesting writing assistance from the HelpWriting.net website. It outlines a 5-step process: 1) Create an account with a password and email. 2) Complete a 10-minute order form providing instructions, sources, and deadline. 3) Review bids from writers and choose one. 4) Review the completed paper and authorize payment if satisfied. 5) Request revisions until fully satisfied, with the option of a full refund for plagiarized work.
The document discusses Bandura's theory of self-efficacy as a sub-theory of social cognitive theory. It defines self-efficacy as one's beliefs in their ability to successfully perform tasks. Bandura derived self-efficacy in 1977 and later updated the model in 1997 to include individual factors influencing self-efficacy levels, with mastery experiences and perceptions of prior accomplishments being most important. Imagined experiences have also emerged as an important influence on self-efficacy.
The document provides instructions for requesting paper writing assistance from a website. It outlines a 5-step process: 1) Create an account with a password and email. 2) Complete a 10-minute order form providing instructions, sources, and deadline. 3) Review bids from writers and choose one. 4) Receive the paper and authorize payment if satisfied. 5) Request revisions until fully satisfied, with a refund option for plagiarism.
How To Start Your Introduction For A Research Paper. How To WriteGina Rizzo
The student interviewed a counselor who works at Cuyahoga Community College. While the student's own interests lie in community or private practice counseling, they chose to interview this counselor to gain perspective from someone in the college counseling setting. The interview explored the counselor's career path, typical duties, clientele served, and rewards and challenges of the role. Insights from the counselor provided the student with valuable information as they continue pursuing their own counseling career goals.
Custom Admission Essay Dnp A Writing Service WiGina Rizzo
The document describes the steps to get assignment writing help from the website HelpWriting.net. It involves creating an account, submitting a request with instructions and deadline, reviewing writer bids and choosing one, making a deposit, and then reviewing and authorizing payment for the completed assignment. It notes that free revisions are provided and plagiarized work will result in a full refund.
Blank Torn White Paper Template Premium ImageGina Rizzo
The document discusses a freemium pricing strategy used by Dropbox to attract users and generate revenue. The strategy involved offering basic cloud storage for free (2GB) and premium paid storage plans. This allowed Dropbox to attract millions of free users and generate most of its revenue from the minority of users who purchased premium plans. However, the strategy also carried risks like low conversion rates of free to paid users and inability to sustain profitability against large competitors.
Green, Yellow, Red The Keys To The Perfect PersuaGina Rizzo
1. Research has found Instagram to have the highest negative effect on mental health compared to other social media platforms due to the emphasis on curating an idealized image of one's life.
2. The constant exposure to highly curated images of others' lives on Instagram can lead to negative social comparison and negatively impact self-esteem and body image.
3. Spending excessive time on Instagram browsing images and focusing on likes and comments can exacerbate symptoms of depression and anxiety by disrupting real-world social connections and encouraging attention-seeking behaviors.
FCE Exam Writing Samples - My Hometown Essay WritiGina Rizzo
This document discusses key factors to consider when choosing a child care provider. It summarizes the results of a survey of 67 parents in Spokane County. The top priority is how loved and happy the children seem. Safety is also very important, including caregiver background checks, drug testing, and security. Additional high priorities are the quality of education, staff qualifications, and student-teacher ratios. Price is lower in priority compared to the child's well-being, safety, and quality of care and education. The document recommends prioritizing a care provider's feelings of safety, quality, and price when choosing where to place a child.
The document outlines a 5 step process for requesting and receiving a custom written paper from the website HelpWriting.net, including creating an account, submitting a request form with instructions and sources, choosing a bid from qualified writers, reviewing and authorizing the completed paper, and having the option to request revisions if needed. The purpose is to provide a simple overview of how the writing assistance service works from start to finish.
How To Teach Opinion Writing Tips And Resources ArtofitGina Rizzo
1. The document provides steps for requesting writing assistance from HelpWriting.net, including creating an account, completing an order form, and reviewing writer bids before choosing a writer and placing a deposit.
2. After receiving a completed paper, the customer can ensure it meets expectations, authorize payment if pleased, and request free revisions.
3. HelpWriting.net promises original, high-quality content and offers refunds for plagiarized works, allowing customers to choose them confidently knowing their needs will be met.
Fantasy Space Writing Paper By Miss Cleve TeaGina Rizzo
This document provides instructions for requesting writing assistance from the website HelpWriting.net. It outlines a 5-step process: 1) Create an account with a password and email. 2) Complete a 10-minute order form providing instructions, sources, and deadline. 3) Review bids from writers and choose one based on qualifications. 4) Receive the paper and authorize payment if pleased. 5) Request revisions to ensure satisfaction, with a refund option for plagiarized content.
हिंदी वर्णमाला पीपीटी, hindi alphabet PPT presentation, hindi varnamala PPT, Hindi Varnamala pdf, हिंदी स्वर, हिंदी व्यंजन, sikhiye hindi varnmala, dr. mulla adam ali, hindi language and literature, hindi alphabet with drawing, hindi alphabet pdf, hindi varnamala for childrens, hindi language, hindi varnamala practice for kids, https://www.drmullaadamali.com
How to Fix the Import Error in the Odoo 17Celine George
An import error occurs when a program fails to import a module or library, disrupting its execution. In languages like Python, this issue arises when the specified module cannot be found or accessed, hindering the program's functionality. Resolving import errors is crucial for maintaining smooth software operation and uninterrupted development processes.
Main Java[All of the Base Concepts}.docxadhitya5119
This is part 1 of my Java Learning Journey. This Contains Custom methods, classes, constructors, packages, multithreading , try- catch block, finally block and more.
How to Manage Your Lost Opportunities in Odoo 17 CRMCeline George
Odoo 17 CRM allows us to track why we lose sales opportunities with "Lost Reasons." This helps analyze our sales process and identify areas for improvement. Here's how to configure lost reasons in Odoo 17 CRM
Leveraging Generative AI to Drive Nonprofit InnovationTechSoup
In this webinar, participants learned how to utilize Generative AI to streamline operations and elevate member engagement. Amazon Web Service experts provided a customer specific use cases and dived into low/no-code tools that are quick and easy to deploy through Amazon Web Service (AWS.)
A workshop hosted by the South African Journal of Science aimed at postgraduate students and early career researchers with little or no experience in writing and publishing journal articles.
Strategies for Effective Upskilling is a presentation by Chinwendu Peace in a Your Skill Boost Masterclass organisation by the Excellence Foundation for South Sudan on 08th and 09th June 2024 from 1 PM to 3 PM on each day.
Your Skill Boost Masterclass: Strategies for Effective Upskilling
A Statistical Computer Experiments Approach To Airline Fleet Assignment
1. A Statistical Computer Experiments Approach to Airline Fleet
Assignment
Venkata L. Pilla1
, Jay M. Rosenberger1
, Victoria C. P. Chen1
, Barry C. Smith2
1
Department of Industrial and Manufacturing Systems Engineering
The University of Texas at Arlington
Campus Box 19017
Arlington, TX 76019 USA
2
Sabre Research Group
3150 Sabre Drive, MD 8203
Southlake, TX 76092
December 27, 2005
1
2. Abstract
The fleet assignment model allocates a fleet of aircraft to scheduled flight legs in an airline timetable. The robust
fleet assignment model addressed in this paper utilizes a two-stage stochastic programming framework along with
the Boeing concept of demand driven dispatch to assign crew-compatible aircraft in the first stage, so as to enhance
the demand capturing potential of swapping in the second stage. A design and analysis of computer experiments
approach is utilized to reduce the computation involved in solving the problem. A Latin hypercube-based design is
developed within the feasible region, and regression splines are employed to obtain the approximation for the second-
stage recourse function. The results on the accuracy of the fit for a real airline carrier are presented and future work is
discussed.
1 Introduction
As airline woes continue to persist, mainly due to rising fuel costs, major domestic airlines are battling to reduce
costs, increase load factors, and improve revenues. The International Air Transport Association reports that North
American carriers are expected to lose $8 billion dollars in the year 2005 (IATA 2005b). One way to reduce costs and
increase revenue is to strike a balance between supply (seats) and demand (passengers). Utilizing a smaller capacity
aircraft would result in spill due to seat shortage, and usage of a larger capacity aircraft would result in flying empty
seats. To avoid this, airline Fleet Assignment Models (FAMs) are used to assign aircraft to the scheduled flights
in order to maximize profit (revenue − cost). FAMs have been credited with an annual savings of $100 million at
Delta (Subramanian et al. 1994), $15 million annually at USAirways (Rushmeier and Kontogiorgis 1997), and a 1.4%
improvement in operating margins at American Airlines (Abara 1989).
The quality of a FAM solution depends upon the accuracy of cost and revenue estimates. While cost estimates are
relatively stable and well known, revenue estimates depend on demand forecasts. In addition to seasonal variations,
airline passenger demand has been affected by various events, such as the terrorist attacks on September 11, 2001,
the outbreak of SARS, and recently rising fuel costs forcing the airlines to increase fares and indirectly affect demand
(IATA 2005a). This uncertainty in demand makes it challenging to assign the right type of aircraft to each flight
leg in the schedule, which is published 90 days prior to the departure of the flight. Consequently, modeling demand
stochastically and delaying the fleet assignment decision closer to departure would likely improve profit.
In this regard, stochastic programming and the Boeing concept of Demand Driven Dispatch are very useful. A
brief introduction to the two concepts are given in Sections 1.1 and 1.2.
1.1 Demand Driven Dispatch
Since most demand for flights is realized after the schedule is published, a robust FAM approach would include
reallocation of aircraft much closer to departure. Berge and Hopperstad (1993) introduced the concept of Demand
Driven Dispatch (D3
), in order to match demand to aircraft close to departure of the flight. In this approach, they
2
3. use two stages of decision-making. The first stage occurs 90 days prior to the departure of the flight, when the
flight schedules are published. During this stage, crew-compatible families of aircraft are assigned to flights in the
airline timetable. Two aircraft are said to be crew-compatible, if they have the same cockpit model; hence, the same
crew could operate either aircraft type. The second stage occurs two weeks prior to the departure of flights, when
most of the demand is realized and individual aircraft are assigned within the crew-compatible families based on the
demand. Swapping can take place with the assignment of specific aircraft in the second stage. For example, Boeing
757 and 767 models are crew-compatible, but a 767 can fly more passengers than a 757. Suppose flights A and B are
initially allocated a 757 and 767, respectively. If two weeks before departure, flight A has realized a higher demand
than expected, while demand is lower than expected for flight B, then the airline can swap the 757 and 767 without
affecting the crew schedule. Higher profit is achieved via this swapping since more revenue is captured, and the the
cost of swapping is usually insignificant for the airline.
Using this concept, Berge and Hooperstad have reported 1% - 5% improvement in profits. The difficulty in imple-
menting a D3
-FAM approach is assigning aircraft types to flights such that swapping is feasible within the practical
constraints of the airline network. Most notably, two planes can only be swapped if they depart from the same airport
and have similar departure times.
1.2 Stochastic Programming
The uncertainty in demand, which is realized close to departure of the aircraft, can be modeled using two-stage
Stochastic Programming (SP). SP models use probability distributions for random events, and their goal is to maximize
the expectation of some function of the decisions based on these random events. In two-stage SP, a decision is taken
in the first stage, after which a random event with a known probability distribution is realized, and a second-stage
recourse decision is taken based on the random event and the decision from the first stage. The optimal decision
consists of a single first-stage decision and a set of recourse decisions to be taken based on the random outcome.
As described in Birge and Louveaux (1997), the basic two-stage stochastic linear program with fixed recourse is
given by:
min z = cT
x + Eξ
min q(ω)T
y(ω)
(1)
s.t. Ax = b, (2)
T(ω)x + Wy(ω) = h(ω), (3)
x ≥ 0, y(ω) ≥ 0, (4)
where x ∈ Rn1
is the first-stage decision vector with linear costs c ∈ Rn1
, y(ω) ∈ Rn2
is the second-stage decision
vector with linear costs q(ω) ∈ Rn2
, A is an m1 x n1 first-stage linear constraint matrix with right hand side b ∈ Rm1
,
and T(ω) and W are, respectively, m2 x n1 and m2 x n2 matrices specifying the second-stage linear constraints on x
and y with right hand side h(ω) ∈ Rm2
. For a given realization of the stochastic variables, ω ∈ Ω, the second-stage
problem data, q(ω), h(ω) and T(ω), become known, and the second-stage decision y(ω, x) can be obtained. Let Ξ
denote a set of scenarios, and let the vector ξT
(ω) represent a scenario with different components of the second stage,
3
4. i.e., q(ω)T
, h(ω)T
and T(ω), such that ξ ∈ Ξ. The objective function represented by (1) contains a deterministic term
cT
x and the expectation of the second-stage objective q(ω)T
y(ω) taken over all realizations of the random event ω.
For a given ω, we can write the second-stage value function as
Q(x, ξ(ω)) = min
y
{q(ω)T
y|Wy = h(ω) − T(ω)x, y ≥ 0}. (5)
Suppose, the expected second-stage recourse function is defined as:
ℑ(x) = EξQ(x, ξ(ω)), (6)
then the deterministic equivalent of the two-stage stochastic linear program can be written as:
min z = cT
x + ℑ(x)
s.t. Ax = b, (7)
x ≥ 0.
The primary difficulty is determining the recourse function ℑ(x). As stated above, each evaluation of ℑ(x) at just
one value of x requires calculation of an expected value of an optimized objective. For large, complex problems, like
fleet assignment, the iterative approximation methods described by Birge and Louveaux (1997) can be very slow to
converge because they typically require a large number of scenarios to adequately represent the stochasticity of the
second-stage optimization in (5). Thus, in each optimization iteration of the first stage, there is a high computational
cost for evaluating ℑ(x) in the second stage, making the problem defined in (7) even harder to solve.
In order to control the computational requirement, Chen (2001) proposed discretization of the x-space to a finite
set of points and solving for ℑ(x) only at those points, followed by a function approximation technique to estimate
the entire surface of ℑ(x). This approximation, ℑ̂(x), will be computationally trivial to evaluate in (7). In selecting a
discretization, it is important to select only those x values which result in a feasible solution in the second stage.
1.3 Contribution
The FAM problem is modeled in this paper using a two-stage SP framework. The two stages correspond to the stages
of decision-making in the D3
concept. Similar to Berge and Hopperstad (1993), the FAM model presented here assigns
crew-compatible aircraft to flights in the first stage. Given an initial Crew-Compatible Allocation (CCA), the second
stage obtains many equally likely scenarios using a probability distribution, and for each scenario, a deterministic
FAM, using the current CCA, solves for the Linear Programming (LP) relaxation of assigning aircraft within the
crew-compatible families. This LP is separable by crew-compatible family. The estimate of the expected revenue is
determined as the average over the scenarios. As mentioned in Section 1.2, the basic SP solution approach can be
computationally intensive if an excessive number of scenarios is required to adequately represent the stochasticity of
the second-stage optimization.
A computationally tractable solution method is presented that employs statistical methods from Design and Analysis
of Computer Experiments (DACE, see Chen, Tsui, Barton and Allen 2003; Sacks et al. 1989). An experimental design
4
5. is used to generate the initial CCA and then for each CCA and for each scenario (demand realization) in the second
stage, a computer experiment (in our case, an optimization problem) is solved. The average over the scenarios is used
to estimate the expected revenue value of the recourse function for the two-stage SP model. A statistical model is fit
to these data to provide a continuous approximation of expected revenue over the initial CCA space. In our case, the
experimental design was based on a Latin hypercube, and the statistical model was based on Multivariate Adaptive
Regression Splines (MARS, Friedman 1991). Optimization of the recourse function is not considered in this paper.
A literature review regarding the FAM problem is presented in Section 2. Section 3 discusses more on DACE and
how it is applied in this paper, and in Section 4, the two-stage SP model for fleet assignment is presented. In Sections
5 and 6, we demonstrate the methodology of generating the first-stage experimental design and second-stage recourse
values, respectively. In Section 7, we present results of the application of this methodology to a real airline carrier, and
finally in Section 8, conclusions and future research are discussed.
2 Literature Review
The fleet assignment problem has been a well researched topic for the past fifty years and has been credited for
increasing profits at several airlines. Ferguson and Dantzig formulated a combined FAM and aircraft routing problem
with deterministic demand (1955) and stochastic demand (1956). Abara (1989) solved the FAM problem using an
integer LP model. He defined “turn” as the successive assignment of an aircraft to two consecutive flights and this
decision variable causes practical limitations to the FAM problem.
Hane et al. (1995) presented a detailed description on solving a fleet assignment problem as a multi-commodity
network flow problem, which formed the basis for a large portion of later FAM research. They modeled the FAM
problem as a Mixed Integer Programming (MIP) problem and reported run times twice as fast as the standard LP
based branch-and-bound code. Some of the concepts addressed in the Hane et al. (1995), like node aggregation and
islands at stations with low activity, are employed in this paper to preprocess the FAM model before solving.
Subramanian et al. (1994) implemented the Coldstart model to solve the fleet assignment problem at Delta Airlines.
Theoretical properties of the FAM problem were presented in Gu et al. (1994). Since the solution of FAM affects
subsequent planning decisions like maintenance requirements, aircraft routing and crew scheduling, extensions of
FAM were considered in Clarke et al. (1996), Rushmeier and Kontogiorgis (1997), Barnhart et al. (1998), Rexing
et al. (2000), Ahuja and Orlin (2002), and Klabjan et al. (2002). Disruptions during the operational phase of the
problem were addressed in Rosenberger et al. (2003).
An important component of a successful FAM is modeling the objective function. The objective function can
be maximizing revenue, minimizing passenger spill (lost revenue due to assigning smaller aircraft) or minimizing
the number of aircraft being used. In this paper the objective is modeled as maximizing revenue and, similar to the
research mentioned earlier, we assume demand for each flight leg is independent. In practice, this is not true, since
in a multi-leg itinerary, the demand on one leg affects the others. This is also referred to as Origin Destination (O-D)
FAM. The network affect will be considered in future work. Jacobs et al. (1999) presented an O-D FAM formulation
5
6. and solved it using Benders’ decomposition. Barnhart et al. (2002) developed a Passenger Mix Model (PMM) that,
given a schedule with known flight capacities and a set of passenger demands with known fares, determines optimal
demand and revenue. They proposed a combined PMM and FAM problem. Farkas (1996) demonstrated that revenue
management has a significant impact on passenger volume and mix, and by ignoring these effects FAM produces
sub-optimal solutions.
The stochastic nature of passenger demand was recently addressed in Listes and Dekker (2002) for determining an
optimal airline fleet composition. Given an airline schedule and a set of aircraft types, the fleet composition problem
determines the number of aircraft of each type that the fleet requires to maximize profit. They developed a two-stage
SP model to determine a single fleet composition that maximizes profit in the first stage across all demand scenarios
generated in the second stage. In the second stage, they solved a deterministic FAM model for each demand scenario
allowing swapping to occur (similar to Berge and Hopperstad (1993)) and employed a scenario aggregation approach
(Rockafellar and Wets 1991) to reduce the computational complexity.
3 Design and Analysis of Computer Experiments Approach
A Design and Analysis of Computer Experiments (DACE) approach is useful when a computer experiment is the only
means for representing a complex system (see recent reviews Chen, Tsui, Barton and Allen 2003; Chen et al. 2005;
Kleijnen 2005). DACE-related methods have been successfully employed for solving stochastic dynamic program-
ming and Markov decision problems (Werbos 1998; Chen et al. 1999; Chen 1999; Chen, Günther and Johnson 2003;
Tsai et al. 2003; Si et al. 2004; Cervellera et al. 2005), and their use was first suggested for SP by Chen (2001). This
paper represents the first implementation of these ideas for SP. Typically, the computer experiment is a simulation
model; however, in this paper it is an optimization model.
Specifically, our DACE Phase conducts the following:
• An optimization model (computer experiment) of system performance is constructed based on knowledge of
how the system operates.
• Design of Experiments (DoE) is used to select the set of sample points as input to the optimization model, which
then provides the corresponding responses.
• A statistical model is to fit these data.
For our robust FAM, the DACE Phase is illustrated in Figure 1. In particular, DoE points for the CCA space must be
feasible, which is an issue not typically handled by DACE methods; thus, a new approach is devised and described in
Section 5. In order to estimate the second-stage recourse function, scenarios are generated based on known probability
distributions, and for each CCA, average revenues from the second-stage FAM optimization are collected as responses
y (see Section 6 for details). A MARS statistical model, ŷ(CCA), is fit to these data to generate an approximate
second-stage recourse function, which can then be employed for more efficient future evaluations in the first-stage
optimization. The run time for MARS is dependent upon the number of basis functions that the user specifies with the
6
7. parameter Mmax. A variant of MARS that automatically selects Mmax, developed by Tsai and Chen (2005), is used
to decrease the time required to fit MARS.
Figure 1: DACE Phase of the Two-Stage SP Framework
4 Two-Stage SP FAM Formulation
Given an airline schedule and a set of fleets of different aircraft that can fly each flight leg, the fleet assignment problem
allocates the fleet of aircraft to the scheduled flights subject to the following operational constraints:
• Balance: Aircraft cannot appear or disappear from the network.
• Cover: Each flight in the schedule must be assigned to exactly one aircraft type.
• Plane Count: The total number of aircraft assigned cannot exceed the number of available aircraft in the fleet.
The objective is to find a feasible assignment that maximizes profit. In this paper, the fleet assignment problem
is formulated as an integer multicommodity flow problem on a time line network similar to Berge and Hopperstad
(1993) and Hane et al. (1995). For a given fleet type, a time line is a graph that represents the arrival and departure
events occurring at each station (airport serviced by the carrier) over a specified time period as shown in Figure 2.
Flights below the time line indicate arrivals, flights above the time line indicate departures, and the numbers indi-
cate the corresponding flight legs. Any flight, which arrives at a particular station, will not be available for departure
immediately because of the time required for fueling, loading passengers/baggage etc. As such, a turn time is added to
all the arrivals before they are ready to take off. The turn time is dependent on the particular fleet type and the station.
A node in a time line begins with an arrival and ends before the next arrival, with at least one departure in between. In
Figure 2, BC, DE, and FA represent nodes. The arcs that connect within these nodes are called ground arcs, and the
arc that connects the last arrival on the time line to the first departure is called the overnight arc. These arcs denote at
least one plane being on the ground at a station and are defined as continuous variables. Because all flight variables are
integral the values corresponding to these arcs will be integral as well. The sum of all the times corresponding to the
arcs represent the total ground time of the planes at that particular station. The overnight arc includes a plane count
7
8. Figure 2: Time Line
hour (typically 4 A.M. EST) that is used for developing the plane count constraints. A detailed description of how to
generate the constraints based on the time line is given in Section 5.1.
4.1 Model Assumptions
The following assumptions are taken to model the two-stage SP FAM problem:
• The revenue coefficients are calculated assuming that the demand for each flight is independent.
• All passengers for a flight have equal revenue. This is inconsistent with yield management practices because as
the capacity is added, incremental passengers have lower average revenue.
• Spilled passengers are assumed to be lost by the carrier and are not recaptured.
• The decision variables are relaxed to be continuous.
• Fleets within a crew-compatible family have same turn time. In practice, we can choose the maximum turn time
within a crew-compatible family.
These assumptions tend to bias FAM solutions to the over use of large aircraft. Considering an OD-FAM is more
reasonable and a subject of future research.
4.2 Stochastic Model
Let L be the set of flight legs (indexed by l). Let F denote the set of fleet types (indexed by f), and G be the set
of crew-compatible families (indexed by g), which can be used for each of the legs l ∈ L. Since we assign, crew-
compatible families in the first stage, for each leg l ∈ L and for each crew-compatible family type g ∈ G, let a binary
variable xgl be defined such that
xgl =
1 if crew-compatible family g is assigned to flight leg l,
0 otherwise.
In the second stage, we assign specific aircraft within the crew-compatible family. As such, for each leg l ∈ L, for
each aircraft f ∈ F, and for scenario ξ ∈ Ξ, let a binary variable xξ
fl be defined such that
8
9. xξ
fl =
1 if aircraft f is assigned to the leg l for scenario ξ,
0 otherwise.
Consider the following FAM parameters:
• S = set of stations, indexed by s,
• V = set of nodes in the entire network, indexed by v,
• f(v) = fleet type associated with node v,
• Av = set of flights arriving at node v,
• Dv = set of flights departing at node v,
• Mf = number of aircraft of type f,
• R(xξ
fl) = revenue if flight leg l is assigned to aircraft type f for scenario ξ,
• C(xξ
fl) = cost if flight leg l is assigned to aircraft type f for scenario ξ,
• aξ
v+ = value of ground arc leaving node v for scenario ξ,
• aξ
v− = value of ground arc entering node v for scenario ξ,
• Of = set of arcs that include the plane count hour for fleet type f, indexed by o,
• L0 = set of flight legs in the air at the plane count hour.
The two-stage formulation can be represented as:
max
X
l∈L
X
f∈F
E[ R(xξ
fl) − C(xξ
fl)] (8)
s.t. aξ
v− +
X
l∈Av
xξ
f(v)l −
X
l∈Dv
xξ
f(v)l − aξ
v+ = 0 ∀v ∈ V, ξ ∈ Ξ, (9)
X
g∈G
xgl = 1 ∀l ∈ L, (10)
X
f∈g
xξ
fl = xgl ∀l ∈ L, g ∈ G, ξ ∈ Ξ, (11)
X
o∈Of
aξ
o +
X
l∈L0
xξ
fl ≤ Mf ∀f ∈ F, ξ ∈ Ξ, (12)
xξ
fl ∈ {0, 1} ∀l ∈ L, ξ ∈ Ξ, (13)
xgl ∈ {0, 1} ∀l ∈ L, g ∈ G, (14)
aξ
v+ ≥ 0 ∀v ∈ V, ξ ∈ Ξ. (15)
9
10. The objective is to maximize profit (revenue − cost) in the second stage by assigning aircraft within the crew-
compatible allocation made in the first stage. For each flight leg l ∈ L, for an aircraft f ∈ g and in scenario ξ ∈ Ξ,
revenue is calculated as:
R(xξ
fl) = (min {[Demand]ξ, [Capacity]f }) ∗ (Average fare per passenger)l.
The block time of a flight leg l is defined as the length of time from the moment the plane leaves the origin station until
it arrives at the destination station. Let bl be the scheduled block time for flight leg l. The cost for each flight leg is
calculated as a function of block time and operating cost of a particular fleet type per block hour, and is given by:
C(xξ
fl) = bl ∗ (Operating cost per block hour)f .
Constraints in set (9) represent the balance constraints needed to maintain the circulation of aircraft throughout
the network. Cover constraints (10) represent the first-stage crew-compatible assignment, and set (11) guarantees
that aircraft within the crew-compatible family (assigned in the first stage) are allocated. For formulating the plane
count constraints (12), we need to count the number of aircraft of each fleet being used at any particular point of day
(generally when there are fewer planes in the air). As such, the ground arcs that cross the time line at the plane count
hour and the flights in air during that time are summed to assure that the total number of aircraft of a particular fleet
type do not exceed the number available. Binary constraints (13) and (14) are relaxed to model the decision variables
as continuous in this paper. As in practice, most crew-compatible families include only one or two aircraft types, so
integer solutions result, as mentioned in Berge and Hopperstad (1993). For families with more than two aircraft types,
an upper bound on the objective function is obtained.
Since the variables are relaxed, the resulting fit for the recourse function obtained using MARS provides an upper
bound for the profit generated. A true recourse function can be estimated with MARS using integral values only at
the expense of higher computational time. This is difficult in the case of solving the two-stage SP with traditional
Benders’ decomposition because there are integrality issues during the generation of cuts.
5 Design of Experiments Approach
The main objective in the first stage is to assign the initial crew-compatible allocation (CCA) using DoE. The design
is generated within the experimental region formed by the first-stage FAM constraints. Since at this stage we are
concerned only with the CCA decision variables, the first-stage constraints are similar to the constraints described
earlier except that they are based on crew-compatible family variables as defined below:
(Balance) av− +
X
l∈Av
xgl −
X
l∈Dv
xgl − av+ = 0 ∀v ∈ V, (16)
(Cover)
X
g∈G
xgl = 1 ∀g ∈ G, l ∈ L, (17)
(Plane Count)
X
o∈Og
ao +
X
l∈L0
xgl ≤ Mg ∀g ∈ G, (18)
0 ≤ xgl ≤ 1 ∀g ∈ G, l ∈ L, (19)
10
11. where Og represents the set of arcs that include the plane count hour for crew-compatible family g, and Mg denotes
the number of aircraft of crew group g.
In order to generate an appropriate experimental design, the above constraints must be preprocessed, removing any
redundant variables, so as to specify the feasible polytope, defined by a system of linear inequalities Ax ≤ b, in the
lowest dimensional space. Preprocessing is conducted by exploiting the explicit as well as implicit equalities present
in the constraints as defined in Section 5.1. Savelsbergh (1994) demonstrated the use of preprocessing techniques to
reduce MIP problems.
Let P := {x ∈ R|Ax ≤ b} be a nonempty convex polytope formed by the first-stage constraints. Discretized
points within this polytope can be generated as shown in Section 5.2 to represent the initial CCA decisions. Any
infeasible design points can be projected into the feasible polytope as described in detail in Section 5.2.2.
5.1 Reducing the Decision Space
Before generating the design, it is important to reduce the first-stage FAM to the minimum number of variables
because the precision of the estimates is adversely affected by multicollinearity (or correlations) among the input
variables (Myers and Montgomery 1986). Reducing to the minimum number of variables in the first stage removes the
redundancy as well as dramatically improves the efficiency of the design. In order to generate the feasible polytope P,
we start with the traditional FAM defined by Hane et al. (1995) and pivot out variables using the equality constraints
as demonstrated in Sections 5.1.1 and 5.1.2, and then remove the implicit equalities as discussed in Section 5.1.3. In
particular, Section 5.1.1 discusses the application of the Boeing concept of D3
, and Section 5.1.2 takes advantage of
the first-stage constraints to reduce the number of variables. Once the equalities are removed, we are left with a set of
minimum variables that define the first-stage decision space.
5.1.1 Demand Driven Dispatch
The robust FAM approach presented in this paper incorporates the Boeing concept of D3
as defined in Section 1.1. In
the first stage (90 days prior to departure), before most demand is realized, crew-compatible families of aircraft are
assigned to flights. As such, in the first stage, only one variable needs to be considered when there is more than one
aircraft belonging to the same crew-compatible family that can be allocated to a flight leg l ∈ L. For example, suppose
flight leg l can be assigned to six different fleet types; that is F = {1, 2, 3, 4, 5, 6}. The cover constraint for flight leg l
is given by:
x1l + x2l + x3l + x4l + x5l + x6l = 1. (20)
Suppose aircraft 1 belongs to crew-compatible family 1, aircraft 2 and 3 belong to crew-compatible family 2, and
aircraft 4, 5, and 6 belong to crew-compatible family 3; that is G = {1, 2, 3}. In this case we substitute x2l and x3l
with just x2l, and we replace x4l, x5l and x6l with x3l. Then, the cover constraint for flight leg l becomes:
x1l + x2l + x3l = 1. (21)
Thus, the number of variables has been reduced from six to three.
11
12. 5.1.2 Constraints
The special structure of each of the constraints in the first stage can be exploited to reduce the number of variables over
which the design has to be generated. Section 5.1.2.1 addresses the reduction due to cover constraints, and Section
5.1.2.2 describes the reduction achieved due to balance constraints.
5.1.2.1 Cover Constraints
For every leg l, the cover constraint obtained after reduction through the Boeing concept can be manipulated to reduce
the dimensionality by one. For example, for equation (21) we have:
x1l + x2l + x3l = 1 ⇒ x3l = 1 − x1l − x2l. (22)
We also have the constraints:
xgl ∈ {0, 1} and g ∈ G = {1, 2, 3}. (23)
Since we are modeling the variables as continuous, we have
0 ≤ x3l ≤ 1 ⇒ 0 ≤ 1 − x1l − x2l ≤ 1 ⇒ 0 ≤ x1l + x2l ≤ 1. (24)
5.1.2.2 Balance Constraints
The balance constraints maintain the circulation of aircraft throughout the entire flight network. Based on the ground
time of the planes and the traffic intensity, stations are typically classified as hubs and spokes.
Balance Constraints for Spokes
Stations with low traffic intensity are called spokes. A spoke time line consists of sudden activity during a period of
time where we can see equal numbers of arrivals and departures. Therefore, we can find time periods in a spoke time
line when there are no planes on the ground. The ground arcs corresponding to these periods can be dropped. Islands
can be defined as a set of nodes in which the incoming and outgoing ground arcs are zero, and a complex island has
two or more nodes. Figure 3 represents a spoke time line. In Figure 3, A, B, C, and E represent nodes; B and D
represent islands while A forms an overnight island. Island D is a complex island with nodes C and E in it. The time
line for each crew-compatible family assists us in framing the balance constraints at each station. Consider a spoke
time line for crew-compatible family g as shown in Figure 3. The balance constraint for node B is:
xg3 − xg4 = 0 ⇒ xg3 = xg4. (25)
Figure 3: Spoke Time Line
12
13. Thus, the crew group type of xg3 will determine that of xg4. For node C, the balance constraints are:
xg5 + xg6 − xg7 − ag = 0 (26)
ag ≥ 0 (27)
where ag represents the ground arc from node C to E. For node E, the balance constraint is:
ag + xg8 − xg9 − xg10 = 0. (28)
From (26), (27) and (28), by dropping the ground arc variable ag the balance constraints for the island D are:
xg5 + xg6 − xg7 ≥ 0 (29)
xg5 + xg6 − xg7 + xg8 − xg9 − xg10 = 0 ⇒ xg10 = xg5 + xg6 − xg7 + xg8 − xg9. (30)
Similarly, for node A:
xg11 + xg12 − xg1 − xg2 = 0 ⇒ xg2 = xg11 + xg12 − xg1. (31)
Thus, the total number of decision variables is reduced from twelve to nine.
Balance Constraints for Hubs
A hub consists of continuous activity with periodic flights coming from the spokes. Because of the high traffic intensity
at hubs, we cannot eliminate all of the ground arcs as we did at spokes. Consider a hub time line for crew-compatible
family g with nodes A, B and C as represented in Figure 4. The balance constraints for nodes A and B are given by:
Figure 4: Hub Time Line
ag + xg1 + xg2 + xg3 − xg4 − xg5 ≥ 0 (32)
ag + xg1 + xg2 + xg3 − xg4 − xg5 + xg6 − xg7 ≥ 0 (33)
ag ≥ 0. (34)
Since the time line ends with the last departure at node C, the constraints for node C will determine the overall balance
constraint for the hub. Therefore, we have:
xg1 + xg2 + xg3 − xg4 − xg5 + xg6 − xg7 + xg8 − xg9 − xg10 = 0
⇒ xg10 = xg1 + xg2 + xg3 − xg4 − xg5 + xg6 − xg7 + xg8 − xg9. (35)
13
14. For each of the hubs there will be one equality constraint for the entire time line to maintain the balance of flights
arriving and departing. Hence, the number of decision variables is reduced by one, and there is one ground arc
variable for each crew-compatible family.
5.1.3 Implicit Inequalities
In this section, we generate equations based on implicit opposing inequalities from the set of constraints Ax ≤ b.
This is achieved by formulating a LP problem that finds two implicit opposing inequality constraints using convex
combinations of the explicit inequalities. Any feasible solution to the LP refers to an implied equation that can be
generated.
An inequality αx ≤ β from Ax ≤ b is called an implicit equality (in Ax ≤ b), if αx = β for all x satisfying
Ax ≤ b. To check if there are any implicit equalities among the system of constraints defined by
Pn
j=1 aijxj ≤ bi, ∀i = 1, 2, ..., m, (36)
we formulate a LP. Let z ≥ 0 be a multiplier for constraint (36), so
Pn
j=1 aijxjzi ≤ bizi, ∀i = 1, 2, ..., m.
P is full-dimensional if and only if P
′
:= {z|AT
z = 0, bT
z = 0, z ≥ 0} = {0}. The proof for this statement
is detailed in Gao and Zhang (2002), Sierksma and Tijssen (2003), and Pilla et al. (2005). The formulated LP is a
maximization problem as we are looking for a non-zero solution. Since the constraints in P
′
form a cone, we need to
add a constraint of the form
Pm
i=1 zi ≤ 1, to get a feasible solution. Thus, the LP can be represented as:
max
m
X
i=1
zi (37)
s.t. AT
z = 0, (38)
bT
z = 0, (39)
m
X
i=1
zi ≤ 1, (40)
z ≥ 0. (41)
If a nonzero solution exists for the LP, then an implicit equality can be generated. Figure 5 shows the flow chart used
for reducing the dimensionality of the FAM problem.
5.2 Generating the Experimental Design
An underlying form for the expected revenue function of a FAM model cannot be assumed; thus, we desire an ex-
perimental design that adequately represents the CCA space. This is challenging for several reasons that will become
apparent in the design generation process. After reducing dimensionality, the design is constructed in three steps:
14
16. • A Latin hypercube algorithm is devised to generate DoE points that satisfy the triangular FAM cover constraints.
This will provide a set of well-distributed points in the CCA space that are close to the feasible region.
• The current DoE points are projected to extreme points of the (convex) feasible region.
• Another Latin hypercube is used to generate interior feasible points as convex combinations of the extreme
points.
Given the high dimensionality of the CCA space, the Latin hypercube is a practical choice of design. Because perfect
correlation between two input variables is possible in a Latin hypercube design, our algorithm generates a larger design
and columns are eliminated so that the final design has low input variable correlations.
5.2.1 Latin Hypercube Algorithm for the Reduced Cover Constraints
After dimensionality reduction, suppose m is the number of crew-compatible types and n is the number of legs. Let the
total number of design points be N = λp mn, where p represents design “levels” and λ is a frequency parameter.
• Draw a “grid” of p points in a m-dimensional hyper-triangle representing a reduced cover constraint over crew-
compatible types.
• Order the p points according to distance from origin (randomly break ties) → 1, 2, . . . , p. Each is an m-tuple.
• Generate a Latin hypercube with N = λp in n dimensions.
• Randomly assign the Latin hypercube levels for each dimension:
1, . . . , λ → 1
λ + 1, . . . , 2λ → 2
.
.
.
(i − 1)λ + 1, . . . , iλ → i, for i = 1, 2, . . . , p
Since each of the reduced cover constraints has a different number of crew-compatible types (m), our design generates
N = λp1p2 . . . pm points.
To determine the p points in hyper-triangles of the cover constraints, we maximized the minimum Euclidean
distance between points (Johnson et al. 1990). For a constraint with a single variable, we spread the points evenly
across the dimension, but for a constraint with more than one variable (two variables form a triangle, three or more
form a hyper-triangle), we spread them evenly across the design space. The problem of maximizing the minimum
pairwise distance among n points in a unit square is equivalent to the problem of finding the maximum diameter of
equal non-overlapping circles contained in a unit square (Maranas et al. 1995). For example, a constraint with two
variables can be represented by a triangle, and its six equidistant points can be found as shown in Figure 6.
5.2.2 Projecting to the Feasible Region
It is critical that the final design points lie within the (convex) feasible region. Different options were explored,
including the omission of any infeasible points and the projection of points into the feasible region. The most promising
16
17. Figure 6: Overlapping circles for generating design
option was projecting points to nearby extreme points of the feasible region. Given that the original set of design points
is well-distributed, the resulting set of extreme points should provide good coverage. However, it should be noted that
it is not computationally practical to identify all extreme points, so it is possible that some regions of the feasible
region will not be covered. The infeasible points (x0
) are projected onto the feasible region using the L1-norm as
represented below:
min kx − x0
k (42)
s.t. Ax ≤ b, (43)
x ≥ 0. (44)
Suppose the solution to the above LP is (x̃). The solution (x̃) actually lies on the face of the feasible polytope
P := {x|Ax ≤ b, x ≥ 0}. Finding a nearby extreme point to a point on the face of the polytope is a difficult task.
We searched for an extreme point in the direction of the gradient of the L2-norm given by kx0
−x̃k2
and exploited
the features in ILOG CPLEX 8.0. Since every iteration of primal simplex provides a basic feasible solution that is an
extreme point, we can search in the direction of the gradient of the L2-norm by providing x̃ as the starting information
and stopping the primal simplex method after the first iteration. This will not guarantee the nearest extreme point, but
by providing x̃ as the initial primal values and iterating primal simplex once, it is likely that an extreme point close to
x̃ is obtained. The LP to optimize the gradient of the L2-norm at the point x̃ is represented as:
min (x0
− x̃)T
x (45)
s.t. Ax ≤ b, (46)
x ≥ 0. (47)
Nearby extreme points could also be obtained using neighborhood search methods like breadth first search (Chvatal
1983).
17
18. 5.2.3 Interior Feasible Points
Although a MARS approximation could be fit to the extreme points alone, interior points would enable better repre-
sentation of the shape of the second-stage recourse function. By Minkowski’s finite basis theorem (Minkowski 1896),
interior points can be obtained by using convex combinations of the extreme points. A second Latin hypercube design
is used to specify the coefficients of the convex combinations.
6 Optimization
Once the initial CCA values in the first stage are obtained via the experimental design, random scenarios are generated,
and a deterministic second-stage FAM problem is solved for each CCA design point and scenario. The two-stage
optimization model was presented in Section 4. An estimate of the revenue was calculated as an average over the
scenarios. The objective of the second stage is to maximize profit (Revenue − Cost) of assigning aircraft to each
flight leg in the schedule. For calculating the revenue, average fares are multiplied with the minimum of demand and
capacity for each scenario. Passengers spilled (customers who cannot be accommodated due to insufficient capacity)
are assumed to be lost revenue and are not captured.
6.1 Scenarios
The purpose of scenarios is to provide estimation of an expected value, which is equivalent to numerical integration.
Swan (1983) developed an approach to estimate true demand from historical booking patterns and showed that booking
data from American Airlines and Swissaire followed either a normal or gamma distribution. In this paper, demand
scenarios for each flight are generated based on the normal distribution assuming that the mean and variance are
known. Since an average of the scenarios is estimated in the second stage, one would expect the performance of the
estimation in the second stage to improve as the number of scenarios increases.
To identify the appropriate number of scenarios, a sample size calculation based on confidence level information
is used. Suppose ȳ is used as an estimate of average revenue (µ). Then given an estimate for the error variance σ, we
can be approximately 100(1 − α)% confident that the error (|ȳ − µ|) will not exceed a specified amount E when the
sample size is:
2 zα/2 σ
E
2
'
.
It should be noted that the iterative approximation methods in Birge and Louveaux (1997) require enough scenarios
to adequately estimate the expected value and an expected supergradient of the recourse function, and it is this latter
expectation that ultimately determines the necessary sample size for their methods. By contrast, our DACE approach
only requires accurate estimation of the expected revenue, which is a much smoother function than the supergradient.
18
19. 7 Case Study Results
We applied the methodology described in the previous sections to a real airline carrier with a weekly schedule contain-
ing 50 stations, 2358 legs, seven fleet types and four crew-compatible families. The traditional FAM model addressed
in Hane et al. (1995) included 6537 integer variables. The turn time was taken as 29 minutes and was assumed constant
irrespective of the crew-compatible family. This assumption can be relaxed to make the problem more realistic, but
this will make the problem even more complex. Based on the ground time for each station and the traffic intensity, six
hubs were identified. The results were computed using a Dual 2.8 GHz Intel Xeon workstation and ILOG CPLEX 8.0
software.
In order to use DoE for generating the first-stage CCA values, the reduced CCA decision space was obtained using
the approach presented in Section 5.1. As discussed in Section 5.1.1, the Boeing concept of D3
was used to create
the cover constraints with only crew-compatible families as variables. The obtained cover constraints were used to
pivot out one variable for each flight, as mentioned in Section 5.1.2.1. The balance constraints were generated for
each station using a time line, as discussed in Section 5.1.2.2, and the equalities obtained were used to pivot out more
variables. The plane count constraints were added to the model to maintain feasibility in the first stage. Finally, after
reducing the model to the least number of variables using the equalities in the operational constraints, the model was
checked for implicit equalities. In our first-stage problem, we did not find any implicit equalities. Pilla et al. (2005)
showed that it is possible to generate implicit equalities using the formulation defined in Section 5.1.3. The results of
the first-stage dimensionality reduction are presented in Table 1.
Table 1: Computation Results
Total number of variables 6537
Variable reduction due to
Demand Driven Dispatch (D3
) 2441
Cover Constraint 2358
Balance Constraint 474
Implicit Equalities 0
Remaining variables 1264
The obtained 1264 CCA variables represented 972 legs, with the corresponding cover constraints forming a poly-
tope. Following the methodology discussed in Section 5.2, a Latin hypercube design was generated based on the cover
constraints. Out of the 972 legs, the cover constraints of 680 legs involved one variable, and the remaining 292 legs
involved two variables. Since n = 972, m = 1 or 2 and N should be greater than 1264 (680*1 + 292*2), we considered
N = 1980 design points, with three levels for 680 legs, and six levels for 292 legs as shown below:
• p1 = {0, 0.5, 1} (corresponding to three levels)
• p2 = {(0,0), (0,1), (1,0), (0, 0.5), (0.5, 0), (0.5, 0.5)} (corresponding to six levels).
These six levels correspond to the six centers of the circles as shown in Figure 6.
19
20. The generated Latin hypercube design was mapped to the corresponding levels, which resulted in 1980 first-
stage CCA decision values. However, none of these satisfied the first-stage operational constraints. Following the
methods in Section 5.2.2, the 1980 infeasible points (x0
) were projected onto the feasible polytope formed by the
cover constraints, resulting in 141 points (more than one infeasible point getting projected to the same point, x̃). Since
the optimal solution lies at an extreme point, nearby extreme points were obtained by considering the gradient of the
L2-norm and using one iteration of ILOG CPLEX 8.0. This resulted in 1525 extreme points (x∗
). The spread of these
points with respect to the original infeasible DoE points was measured using the ratio
||x0
− x∗
||
||x∗||
, (48)
and the histogram of these ratios is shown in Figure 7. There are chances of some corners in the polytope being left
behind during this process as there are a total of
3695
1264
combinations of extreme points, and it is impossible to take
into account all of these. Given that there are no gaps in the histogram, the points appear to be well spread across
the polytope. Finally, interior points were obtained using convex combinations of these 1525 extreme points. Thus,
Figure 7: Histogram to check the spread of points
around 3562 total first-stage assignments were obtained, which included 141 feasible projections, 1525 extreme points
and 1896 interior points.
In order to solve the second stage, demand scenarios were generated using a normal distribution with known mean
and standard deviation values for each flight leg. As discussed in Section 6, for a desired error of approximately 0.1%
of the average revenue, we obtained the sample size of scenarios to be 30. For each design point and for each scenario,
the second-stage LP as discussed in Section 4 was solved, and the expected revenue was calculated as the average over
the scenarios.
Using the average revenue response values at the CCA design values, a MARS approximation was fit using an
automatic stopping rule as discussed in Tsai and Chen (2005). The fit resulted in 320 basis functions with a coefficient
of determination (R2
) of 99.013%. A validation data set was generated (utilizing new convex combinations of 1525
extreme points) to test the MARS approximation, and relative errors were calculated using the formula
|y − ŷ|
|y|
. (49)
20
21. The maximum relative error was obtained as 3 x 10−5
and the median relative error was 6 x 10−6
, both of which
are very small. A new set of extreme points was generated to evaluate the MARS approximation in regions possibly
outside of that represented by the experimental design, and the relative error was on the order of 10−3
. Thus, the
MARS approximation is a very good fit for the second-stage recourse function.
8 Future Work
The main contribution of this paper is a method to estimate the recourse function for the two-stage SP model using a
MARS approximation over a discretized first-stage decision space based on a Latin hypercube design. The obtained
fit can be used for future evaluations and optimized so as to generate a single first-stage decision. In this aspect,
derivatives of the MARS function can be utilized to generate revenue cuts. The revenue generated from using our
approach will be compared to the present methods utilized.
21
22. References
Abara, J. (1989). Applying integer linear programming to the fleet assignment problem, Interfaces 19(4): 20 – 28.
Ahuja, R. and Orlin, J. (2002). Very large scale neighborhood search in airline fleet scheduling, SIAM News 35(9).
Barnhart, C., Boland, N., Clarke, L., Johnson, E., Nemhauser, G. and Shenoi, R. (1998). Flight string models for
aircraft fleeting and routing, Transportation Science 32(3): 208 – 220.
Barnhart, C., Kniker, T. and Lohatepanont, M. (2002). Itinerary-based airline fleet assignment, Transportation Science
36(2): 199 – 217.
Berge, M. and Hopperstad, C. (1993). Demand driven dispatch: A method of dynamic aircraft capacity assignment,
models and algorithms, Operations Research 41(1): 153 – 168.
Birge, J. R. and Louveaux, F. (1997). Introduction to Stochastic Programming, Springer, New York, New York.
Cervellera, C., Chen, V. and Wen, A. (2005). Optimization of a large-scale water reservoir network by stochastic dy-
namic programming with efficient state space discretization, European Journal of Operational Research, to appear
.
Chen, V. (1999). Application of mars and orthogonal arrays to inventory forecasting stochastic dynamic programs,
Computational Statistics and Data Analysis 30: 317 – 341.
Chen, V. (2001). Measuring the goodness of orthogonal array discretizations for stochastic programming and stochas-
tic dynamic programming, SIAM Journal of Optimization 12(2): 322 – 344.
Chen, V. C. P., Tsui, K.-L., Barton, R. R. and Meckesheimer, M. (2005). Design, modeling, and applications of
computer experiments, IIE Transactions, to appear .
Chen, V., Günther, D. and Johnson, E. (2003). Solving for an optimal airline yield management policy via statistical
learning, Journal of the Royal Statistical Society, Series C 52(1): 1 – 12.
Chen, V., Ruppert, D. and Shoemaker, C. (1999). Applying experimental design and regression splines to high-
dimensional continuous-state stochastic dynamic programming, Operations Research 47: 38 – 53.
Chen, V., Tsui, K., Barton, R. and Allen, J. (2003). A review of design and modeling in computer experiments,
Handbook of Statistics 22: 231 – 261.
Chvatal, V. (1983). Linear Programming, W. H. Freeman and Company, New York, New York.
Clarke, L., Hane, C., Johnson, E. and Nemhauser, G. (1996). Maintenance and crew considerations in fleet assignment,
Transportation Science 30(3): 249 – 260.
Farkas, A. (1996). The Influence of Network Effects and Yield Management on Airline Assignment Decisions, PhD
thesis, Massachusetts Institute of Technology.
22
23. Friedman, J. H. (1991). Multivariate adaptive regression splines, The Annals of Statistics 19(1): 1 – 141.
Gao, L. and Zhang, Y. (2002). Computational experience with lenstra’s algorithm, Technical Report TR02-12, Rice
University, Department of Computational and Applied Mathematics.
Gu, Z., Johnson, E., Nemhauser, G. and Wang, Y. (1994). Some properties of the fleet assignment problem, Operations
Research Letters 15(2): 59 – 71.
Hane, C., Barnhart, C., Johnson, E., Nemhauser, G. and Sigismondi, G. (1995). The fleet assignment problem: Solving
a large scale integer program, Mathematical Programming 70: 211 – 232.
IATA (2005a). Passenger growth slows in august - fuel prices begin to dampen demand,
http://www.iata.org/pressroom/pr/2005-09-30-02.htm.
IATA (2005b). Us$7.4 billion losses for global aviation in 2005, http://www.iata.org/pressroom/pr/2005-09-12-01.htm.
Jacobs, T., Johnson, E. and Smith, B. (1999). O-d fam: Incorporating passenger flows into the fleeting process,
AGIFORS Symposium Proceedings, 39, New Orleans, LA, USA.
Johnson, M., Moore, L. and Ylvisaker, D. (1990). Minimax and maximin distance designs, Journal of Statistical
Planning and Inference 26: 131 – 148.
Klabjan, D., Schaefer, A., Johnson, E., Kleywegt, A. and Nemhauser, G. (2002). Airline crew scheduling with time
windows and plane-count constraints, Transportation Science 36: 337 – 348.
Kleijnen, J. P. C. (2005). An overview of the design and analysis of simulation experiments for sensitivity analysis,
European Journal of Operational Research 164(2): 287 – 300.
Listes, O. and Dekker, R. (2002). A scenario aggregation based approach for determining a robust airline fleet compo-
sition, Technical Report Econometric Institute Report EI 2002-17, Erasmus University Rotterdam, The Netherlands.
Maranas, C., Floudas, C. and Pardalos, P. (1995). New results in the packing of equal circles in a square, DMATH:
Discrete Mathematics 142.
Minkowski, H. (1896). Geometrie der Zahlen, Teubner Leipzig.
Myers, R. and Montgomery, D. (1986). Response Surface Methodology, John Wiley and Sons Inc, New York, New
York.
Pilla, V., Rosenberger, J. and Chen, V. (2005). An lp formulation for implicit dimensionality reduction, Technical
Report COSMOS-05-02, University of Texas at Arlington, Department of Industrial Engineering.
Rexing, B., Barnhart, C., Kniker, T., Jarrah, A. and Krishnamurthy, N. (2000). Airline fleet assignment with time
windows, Transportation Science 34(1): 1 – 20.
23
24. Rockafellar, R. and Wets, R. (1991). Scenarios and policy aggregation in optimization under uncertainty, Mathematics
of Operations Research 16(1): 119 – 147.
Rosenberger, J., Johnson, E. and Nemhauser, G. (2003). A robust fleet assignment model with hub isolation and short
cycles, Transportation Science 38(3): 357 – 368.
Rushmeier, R. and Kontogiorgis, S. (1997). Advances in the optimization of airline fleet assignment, Transportation
Science 31(2): 159 – 169.
Sacks, J., Schiller, S. and Welch, W. (1989). Designs for computer experiments, Technometrics 31: 41 – 47.
Savelsbergh, M. (1994). Preprocessing and probing techniques for mixed integer programming problems, ORSA
Journal on Computing 6(4): 445 – 454.
Si, J., Barto, A., Powell, W. and Wunsch, D. (2004). Handbook of Learning and Approximate Dynamic Programming,
Wiley, New York: Wiley.
Sierksma, G. and Tijssen, G. (2003). Degeneracy degrees of constraint collections, Mathematical Methods of Opera-
tions Research 57: 437 – 448.
Subramanian, R., Scheff, R., Quillinan, J., Wiper, D. and Marsten, R. (1994). Coldstart: fleet assignment at delta air
lines, Interfaces 24(1): 104 – 120.
Swan, W. (1983). Traffic losses at high load factors, AGIFORS Symposium Proceedings, 23, Olive Branch, MS, USA.
Tsai, J. and Chen, V. (2005). Flexible and robust implementations of multivariate adaptive regression splines within a
wastewater treatment stochastic dynamic program, Quality and Reliability Engineering International 21(7): 689 –
699.
Tsai, J., Chen, V., Chen, J. and Beck, M. (2003). Stochastic dynamic programming formulation for a wastewater treat-
ment decision-making framework, Annals of Operations Research, Special Issue on Applied Optimization Under
Uncertainty 132: 207 – 221.
Werbos, P. (1998). A Brain-Like Design To Learn Optimal Decision Strategies in Complex Environments, Springer,
Springer, London.
24