HUMANFACTORS Evolving fatigue risk management systems increase two airlines’ confidence about alertness. BY WAYNE ROSENKRANS Scientific Scheduling or about 15 years, Air New Zealand pilots departed from Christchurch levels are reasonable and fatigue predic- periodically assessed pilots and between 2100 and 2200 local time and ar- tions are sufficiently accurate. flight attendants during flight op- rived back at Christchurch at about 0700 The company recently monitored erations and attempted to scientifi- the following day. “It is the sort of duty for three months the benefits of pilots Fcally identify links between measured done around the world,” Powell said. self-reporting their fatigue level about levels of fatigue and safety indicators. “Changing the aircraft [Boeing 737- 30 minutes before the safety-critical top Today, fatigue risk management systems 300/Airbus A320] doesn’t make any of descent phase on every flight. In all, (FRMSs) “mirror the pillars of safety difference, but providing a night stop in 9,000 paper-form responses represented management systems,” says David Pow- Brisbane makes a big difference. Reaction long-haul, regional and domestic opera- ell, aviation medicine specialist for the time [on in-flight psychomotor vigilance tions. One finding for regional trips airline. Nevertheless, airlines are finding tests], compared with that in all of our was that starting duty from morning to that discussing an FRMS is easy while studies, was quite high towards the end. midday kept the peak fatigue level well actually implementing all the elements … [Objective] reaction time data and the within an acceptable range, but starting is “particularly hard to do,” he told [pilots’ self-reported] subjective data tend duty in the evening or the middle of Flight Safety Foundation’s 64th annual to tell exactly the same story.” From such the night could cause fatigue levels to International Air Safety Seminar (IASS) studies, predictive analyses have red- increase quickly toward an unacceptable in Singapore in November. flagged situations requiring changes in level. Ability to isolate risk factors within A few years ago, a company study the timing of departures, crewing level or this “wealth of data” then convinced the focused on two-crew flights for the details of the pattern to mitigate fatigue. company to require top-of-descent alert- Christchurch, New Zealand–Brisbane, Equally valuable, he said, has been confir- ness ratings from each Boeing 777 pilot Chris Photography Sorensen Australia, city pair, on which the same mation by both types of data that fatigue on the flight deck on every flight. © 24 | FLIGHT SAFETY FOUNDATION | AEROSAFETYWORLD | FEBRUARY 2012 HUMANFACTORS Using the Samn-Perelli alertness scale “In terms of making the call on what with Helsinki University to study crew of 1 to 7 (Figure 1), flight crews reported is safe enough, we have got a long way to fatigue levels on long-range flights, and a higher fatigue level on the return sectors go,” Powell said. “There are not enough in 2008, the researchers expanded data of out-and-back, daylight flights from data out there on fatigue linking with collection to narrowbody aircraft. Auckland, New Zealand, to Melbourne, safety, so I’m here to appeal for [research “In 2008, we asked, ‘What are the Australia, for the 0800 local time depar- on safety metrics].” A promising avenue possibilities of introducing a fatigue ture compared with similar-duration of inquiry is how some crews with a model to actually influence the con- flights at different departure times. high fatigue level or restricted sleep can struction of the schedules from the very “There is probably a little bit of trunca- perform tasks in a flight simulator or beginning, rather than just measuring tion of [pilots’] sleep,” Powell explained. line operations safety audit as effectively fatigue after the process is completed? Another finding, from a three-crew as well-rested crews, or can exhibit fewer Could [we] influence those sequences variant of this flight, was that “the benefit — but more serious — recorded exceed- of flights to end up in the best possible of a third pilot for a daylight sector is less ances of normal flight parameters. context?’” Klemets said. This work led obvious [than assumed],” he said. to the design and early 2011 launch of Powell told IASS attendees, “You Finnair Crew Vulnerability an Apple iPhone application (app) for can monitor fatigue across your entire Tomas Klemets, head of scheduling building alertness into crew schedul- operation easily and cheaply [together] safety, Jeppesen Systems, described to ing, developed with design input from with your flight data analysis programs.” IASS attendees Finnair’s early experi- company pilots. Airlines should expect to frequently en- ence with its evolving, incomplete “Finnair pilots actually fly rosters that courage crewmembers to keep up their FRMS in a presentation co-authored have been produced … using a fatigue in-flight ratings over time; find ways to with Gabriela Hiitola, the airline’s head model guiding the overall construction,” gauge FRMS effectiveness in relation of crew scheduling. Finnair operates he said. Each “planning horizon” is con- to measures in international guidance; widebody jets connecting Europe with tinually revisited and refined from the and produce validated, reliable mea- long-range destinations in Asia via Hel- long-term planning stage to the day of sures of safety performance. sinki. In 2007, the airline began to work operation under the FRMS, he added. Some risk factors are inherently Two-Crew Fatigue at Top of Descent tough to mitigate, however. “When an airline decides to operate to a certain 6 Duty time (hours) station with certain equipment at a cer- 10.5 9.0 tain departure time, that will inevitably 7.5 5 6.0 lead to a certain level of fatigue that will 4.5 3.0 be very difficult to avoid,” Klemets said. 4 Pairing construction, roster construc- tion by automated optimizers and FRMS monitoring have far less influence in 3 those situations, he said. The airline also has added scheduler Samn-Perelli Scale (7-point) 2 and pilot training on key performance indicators (KPIs) of safety. “What Finnair 1 does today is to trend what we call the 0 2 4 6 8 10 12 14 16 18 20 22 24 PA5, the average predicted level of alert- Time of day (hour) ness on the 5-percent ‘worst’ flights,” he Notes: The earliest research using pilot self-assessment on paper forms before top of descent, during said. For any dramatic improvement, two-crew regional operations, enabled Air New Zealand to derive trend curves based on duty start time and approximate duty duration. Higher values on the vertical scale mean greater fatigue. however, the airline “would need to relax Source: David Powell, Air New Zealand or remove some [regulatory/contractual] constraints or sacrifice some other KPIs,” Figure 1 Klemets added. FLIGHTSAFETY.ORG | AEROSAFETYWORLD | FEBRUARY 2012 | 25.