Back to top

Beyond Data: A Trip to Delta Airlines’ Operational Core

WINTER 2013 --The next time your flight is delayed or cancelled, know that the reasons are varied and complex, but sometimes it comes down to something as simple as this: When an airline has a choice of flights to scratch, it rarely cancels one with unaccompanied minors on it—for fear of losing a child.

That was just one bit of insider information that ITS doctoral students learned on a two-day trip to Delta Airlines’ operational headquarters in Atlanta last September in an effort to better understand why airlines make the choices they do as they consider fuel efficiency, on-time performance, optimum utilization of their aircraft, customer relations and a host of other factors. 
The students' visit also resulted in a new six-month project co-sponsored by Delta and the International Council on Clean Transportation (ICCT) aimed at reducing the amount of fuel used by aircraft. 
Transportation engineering professor Mark Hansen and three of his students flew to Atlanta, along with Frank Ketcham, a Delta pilot who has worked with ITS air transportation students since 2008, for an inside look at the complicated workings of the country’s second-largest airline.
Hansen’s students work on a number of projects funded by the Federal Aviation Agency that deal with the behavior of airlines. Because airlines are competitive, the data they collect is mostly kept under wraps and unavailable to researchers. For modeling purposes, researchers must rely primarily on publically available data—which has its limits.
“There’s a certain amount of research you can do with that data, but the opportunity to talk to the people who are making decisions day in and day out is really important,” says Hansen, who is a co-director of the National Center for Excellence in Aviation Operations Research (NEXTOR), a consortium of universities contracted by the Federal Aviation Administration to provide research support for a variety of aviation issues.
Delta pilot Frank Ketcham, ITS student Yi Liu, Professor Mark Hansen, and ITS student Michael Seelhorst
Which is how ITS doctoral student Michael Seelhorst learned that airlines hate to cancel flights with unaccompanied children aboard.
Seelhorst has been modeling flight cancellations using publically-available data for more than a year with the goal of determining how cancelling one flight affects the decision to cancel other flights.
“Let’s say Delta has a flight in the morning going from Atlanta to Knoxville. Typically, the same aircraft is used for both legs of the flight. If the first leg is cancelled, the aircraft is not available to return on the second leg, and it will be cancelled as well. That’s one example, but there are many.
“My modeling deals with where these interdependencies lie and figuring out the scope of flights that are affected by a single cancellation decision.”
In Delta’s vast operations control room, the people making the decisions to cancel a flight told Seelhorst what the data he’s been using could not.
“Some of what they said surprised me, like how they hate to cancel flights carrying unaccompanied children or high-value customers who fly frequently. But one of the most important things that affect which flights to cancel is crew scheduling—because crews can only work a certain number of hours. “
If a flight is canceled and the crew is scheduled to keep working, that crew won’t get to the destination where they’re expected for the next flight.
“For that reason,” explained Seelhorst, “if they have a choice, airlines try to cancel a flight that is the end of the crew’s last shift so as not to effect another flight.”

Keeping customers happy

While passengers may not believe it, the airlines also try to keep down the number of hours that passengers will be inconvenienced by using software that shows them what the average passenger “re-accommodation time” will be if a particular flight is canceled. In other words, if all passengers on a canceled flight were re-routed, how long will it take each of them to get to their destinations? One hour? Eight hours?
Complicating that equation is the fact that there are fewer empty seats on other flights than were available a few years ago, making it more difficult to accommodate passengers from canceled flights.
“If they’re trying to choose between two flights to cancel, they really don’t want to anger their passengers,” added Seelhorst. “That’s probably more important to them than lost revenue.”
Frequently, weather accounts for delays or last-minute flight cancellations—particularly summer thunderstorms, which are harder to forecast and prepare for than winter snowstorms or hurricanes.
“Even when thunderstorms are forecast, you never know exactly when they’re going to happen,” said Seelhorst. “So the number of flights they have in front of them that they’re choosing from to cancel, that is, the flights that are within this window of being dependent on each other, is very small. It may be only an hour or so.”
Forecasters can predict with more accuracy snowstorms and hurricanes like Sandy, which allows airlines to be more strategic and to work out cancelation problems to their best advantage, added Seelhorst.

Burning fuel

Lack of predictability in airline operations is a broader problem than weather and flight cancelations, however, and students Lu Hao and Yi Liu are trying to understand and solve other problems that lead to inefficiencies and increased cost for airlines.
Hao has pursued the trade-offs of a longer, more costly “scheduled block time,” versus a shorter one. Scheduled block time refers to a flight’s allotted time between departure and arrival.
“On the one hand, the airlines want a shorter scheduled block time because
longer ones are big cost drivers for them,” said Hao. “Every minute of block time is expensive. They have to pay the crew, load more fuel, and they’re not using the aircraft as efficiently.”
“On the other hand,” adds Hao, “A longer scheduled block time means the flight is less likely to be delayed—you’re building more error into it—and on-time performance is very important to customer relations. “
Lu Hao is studying scheduled block times in order to find more efficient solutions.
Yi Liu is working on several projects aimed at helping airlines prioritize the relative importance of different performance goals such as fuel efficiency or the ability to use their fleet of aircraft more efficiently.
Fuel—and how much of it to load on an aircraft—is an important consideration because it takes fuel to carry fuel, and the more fuel that is burned, the more expensive and the greater the environmental effects. And the one part of the trip that burns a lot of fuel unnecessarily is on the ground waiting for take off.
One area that Liu has focused on is the period when a pilot pulls out of the gate and taxis to the runway and the actual time the aircraft takes off, which plays a significant role in fuel efficiency.
“Most passenger aircraft have two engines. When you taxi on the ground you only need one engine,” explains Liu. “Before you take off, however, your other engine needs to warm up for five minutes before you can take off. So if I tell you, okay, take off now and you only have one engine on, you cannot.”
Minutes are critical in airport surface operations. Three aircraft can take off during a two-minute time period.
“So if you’re sitting there, blocking everyone while you warm up your airplane—that’s really bad,” she adds.
Liu’s research is about making this part of a journey more predictable and more fuel-efficient.

Real world experience from the cockpit

With these and other research and modeling challenges, pilot Frank Ketcham has been extremely useful. Although two days at Delta provided students with many new insights, their ability to meet with a pilot who regularly flies throughout the National Air Space and who is familiar with various airports and problems unique to them, gives these researchers insight into the real world of aviation operations throughout the year. He visits the campus weekly to work with students and answer their questions.
"Airlines are producing a lot of data, but not necessarily capturing it and assessing it to its full potential. The people at Berkeley are very good at doing just that." 
Ketcham, who’s been a pilot for Delta Airlines for 15 years and flew for TWA prior to that, enthusiastically embraces those questions.
As a student at UC Davis, he studied economics and finance, but his other passion was flying: his father had been a pilot, his mother a flight attendant.
In 2008 he began working with Berkeley students.
“I can give them a real, first-hand perspective from my experience about everything from the high altitude routings, the terminal operations where we’re down low being vectored in, and airport surface issues,” he said. “They’re all very challenging and they are all very, very complex.”
For Ketcham, the students provide “a really wonderful feedback loop where I can go out and fly the system, see a problem and then have a place for that information to go. It’s very therapeutic.”
He also provides a bridge between Delta Airlines and UC Berkeley’s NEXTOR program.
“Airlines are producing a lot of data, but not necessarily capturing it and assessing it to its full potential. The people at Berkeley are very good at doing just that. We’re in a position now where this is more important then ever because air transportation is such a large part of our economy. We need to dig deeper and bring in smart people like those at NEXTOR to try to address some of these challenges.”
“For Michael who is modeling flight cancelations, it was really important for him to get a first-hand account of what types of processes they go through in making their decisions. How much are they looking at revenue and those types of things? Or are they just trying to triage,” he explained. “Once he can understand that decision-making process he can model it.”
“Yi and Lu are looking at ground delay programs and surface management, and that’s where there are very significant delays and very significant variance. Once an airplane is in the air, there is, for the most part a known trajectory. On the ground is where they experience higher uncertainty and delay.”
Ideally, says Ketcham, the airlines will try to take the airborne delay and put it on the ground, take the ground delay and put it in the terminal, and take the terminal delay and put it at home. Delays have different characteristics and understanding those characteristics helps manage and minimize them.
“Wouldn’t you rather be sitting at home and hear several hours before you leave the house that your flight’s been delayed?” he asks. “That’s really the concept of predictability that the airlines need to be working toward. We obviously can’t do that with all types of delay but it’s the direction we need to go.”
Near the end of the two-day visit to Delta’s headquarters, the students got the chance to try Ketcham’s job: landing a 767-400—or at least, the simulator version.
“I don’t think any of us landed the plane successfully,” said Hansen.
“I failed that pretty tremendously,” admitted Seelhorst, bringing to mind a comment Amelia Earhart is said to have made: “Trouble in the air is very rare. It is hitting the ground that causes it.”

Solving a Fuel Dilemma

Several months after the students returned to Berkeley, Delta and ICCT proposed a project that would utilize the students' skills. 
For the next six months, they and students working with ITS alumna Megan Smirti Ryerson, now an assistant professor at the University of Tennessee, will determiine how to improve the decisions dispatchers make about how much fuel to load on aircraft.
"Weight is the enemy because it takes fuel to carry fuel," explained Hansen.
Dispatchers at Delta's central control make the decision about how much to load based on various criteria, including how much extra fuel may be required if problems arise, such as being diverted to another airport due to weather problems.
Over the next six months the students will talk to dispatchers to try to understand what factors influence their decisions. Then they will present possible solutions.
"Our relationship with Delta has been a very fruitful one. We expect it to be very productive," said Hansen.