Professional Documents
Culture Documents
Airline Operations
Stephen Altus, PhD
May 2007
Overview Take Another Look
850
750
550
450
250
150
– Weather 50
55000 75000 95000 115000 135000 155000 175000
The fuel required to climb varies nonlinearly with the weight of the aircraft
950
• Fuel for an
850
arbitrary 2000’
enroute climb –
Fuel required (lb)
750
650
typical altitude
550
change as the
aircraft gets lighter
450
• Aircraft from
350
multiple
250
manufacturers
150
• Significance:
50
55000 75000 95000 115000 135000 155000 175000
Weight of the
aircraft at any point
Aircraft weight at start of climb (lb)
impacts the
Note: this plot generated with proprietary data; the actual aircraft, optimal decision at
altitudes, climb speeds, and temperatures cannot be disclosed that point
Relevant Domain Aspects
Aircraft Performance Take Another Look
Cruise fuel flow rate varies nonlinearly with the weight of the aircraft
6500
• Fuel flow rate at
6000
an arbitrary
altitude and
Fuel flow (lb/hr)
5500 temperature
• Aircraft from
5000
multiple
4500
manufacturers
• Significance:
4000
Weight of the
aircraft at any
3500
50000 70000 90000 110000 130000 150000 170000 point impacts the
Aircraft weight (lb) optimal decision at
that point
Note: this plot generated with proprietary data; the actual aircraft,
altitudes, climb speeds, and temperatures cannot be disclosed
Relevant Domain Aspects
Aircraft Performance Take Another Look
530000
510000
• Medium- and
490000 long-haul flights
• Aircraft from
Weight (lb)
470000
450000 multiple
430000 manufacturers
410000 • Significance:
390000
Cannot assume
370000
constant aircraft
weight in
350000
performance
330000
0 1000 2000 3000 4000 5000 6000 models
Distance (nm)
Note: this plot generated with proprietary data; the actual aircraft, payload, route,
profile, climb speeds, winds, and temperatures cannot be disclosed
Relevant Domain Aspects
Expanding the domain… Take Another Look
So what’s the big deal? We have all these (mostly) smooth curves, why
not just let the aero engineers write some simple trajectory
optimization software?
5500
weight
5000 • Aircraft from
multiple
4500
manufacturers
4000 • Significance:
Temperature at
3500
• Jakarta to Honolulu
Wind-optimal
• NWS Winds
route
• 25 October 2002
• Route is 8% longer but
Great circle
3% less fuel and 2% less
time than minimum-
distance route (11%
longer than great-circle)
• Real (although extreme)
result from a commercial
flight planning system
Relevant Domain Aspects
Weather Take Another Look
Not every method gets the right answer all the time
The same CGK-HNL example
was run with two Optimal
Control algorithms:
• Dynamic Programming
• Neighboring Optimal
Paths
Even today, most flights follow defined static airways and waypoints
(navigational aids or arbitrarily defined latitude/longitude points)
Relevant Domain Aspects
Route and Altitude Structure Take Another Look
Defining the grid is non-trivial, because there are different rules about
where you can fly
Now that we’ve learned about the domain, let’s re-visit the basic problem
Could use Optimal Control methods that are currently used for trajectory
optimization (e.g., spacecraft)
Advantages:
• Correctly captures all coupled effects (best route depending on details
of profile, avoidance of 4D areas, etc.)
• Globally optimal unconstrained solution (subject to uncertainty in input
data)
Disadvantages:
• Must fit the path to discrete allowed route/altitude structure, may lose
optimality
• Restrictions increase the likelihood of sub-optimal solutions once
translated
Optimization of Route and Profile
Solution Strategies Take Another Look
Operational Cost
Data Data
Advantages:
• Optimized solutions are domain-feasible with no post-translation
• Much easier to enforce specific restrictions
Disadvantages:
• Not guaranteed to converge to the global optimum (sequential subspace
optimization)
• Limitations on handling of 4-D (and even 3-D) avoid areas
Optimization of Route and Profile
Domain Decomposition Approach Take Another Look
• Optimal Altitude and Speed should Minimize Total Cost (not fuel)
– Not necessarily max lift-to-drag ratio
– Include compressibility drag and climb fuel • Each aircraft
50000
type is most
45000
efficient at a
Speed constant
De
40000
specific value of
with altitude
ns
W/(ρV2)
ity
Spee
35000
d
to fly higher
of so
25000
Speed increases
with altitude • Fuel/time
und
20000
tradeoff
15000
– Higher for
10000
less fuel
5000
– Near 27,000’
0
for max
Variation of temperature, density, speed of sound with altitude speed
Optimization of Route and Profile
Domain Decomposition Approach Take Another Look
Profile/Speed Optimization
35,000’
700lb 650lb
115,000lb.
5,100lb/hr * 19.4 mins = 1650lb
113,350lb.
33,000’
• Cruising higher (and lighter by 700lb) leads to 2% lower fuel flow rate
• But delaying climb reduces climb fuel by 6%
• Total fuel if climb is delayed is 30lb less
Optimization of Route and Profile
Domain Decomposition Approach Take Another Look
Optimum altitude
Besides the basic route and profile optimization, other decisions are
relevant (and coupled to the route and profile optimization)
Three examples:
• Economic Fuel tankering (coupled to profile and speed optimization)
• Optimal reclear point selection
• Minimum Cost Routing with Overflight Charges
Specific Topics
Fuel Tankering Take Another Look
Accurate operational
Exchange data (e.g. MEL/CDL)
rates
Flight-planning system
Fuel prices
Flight plan with optimal
tanker calculation
Takeoff and
Aircraft schedules landing
for sector linking weight limits
Specific Topics
Fuel Tankering – Expanded Domain Take Another Look
Accurate operational
Exchange data (e.g. MEL/CDL)
rates
Fuel prices
Flight plan with optimal
tanker calculation
Range/Payload Diagram
• The Fuel/Payload
exchange region (limited
Limited by Maximum Payload by max takeoff weight) is
the most interesting
Maximum Payload – Indicates correctly-sized
Payload (weight)
Range (distance)
Specific Topics
Optimal Reclear Point Selection Take Another Look
Origin Planned
destination
• Instead of percentage contingency fuel for the full flight, need the
greater of:
– Percentage from origin to reclear airport
– Percentage from decision point to planned destination
• Optimal solution is both scenarios equal – “equal fuel point”
Specific Topics
Optimal Reclear Point Selection Take Another Look
RTE 3
Avoiding LE
RTE 1
Shortest RAD-
compliant route RTE 2
Avoiding LP
Specific Topics
Optimal Routes with Overflight Charges Take Another Look