AINsight: Data-entry Errors Impede Takeoff Safety

 - February 28, 2020, 10:49 AM

We’ve all “fat-fingered” a text message. The outcome from this error at worst is a misspelled word, a chuckle from the recipient, and an embarrassing, but forgettable, moment for the sender. These keyboard input errors seem harmless enough, right? But beyond the smartphone, humans inputting erroneous information into an automated system can have a much more dire consequence.

So, what or who is at fault? Is it the machine or system—or is it the operator? In the case of the smartphone, a convenient excuse may be to blame the engineer for designing a keyboard that is far too small and lacks tactile feedback for the user.

In the financial markets, these fat-fingered errors can have a huge financial impact. This happened in 2014 when an errant mouse-click from a single Japanese trader—inputting an order for 1.9 billion shares instead of the intended 1.9 million shares—nearly cost his company $711 billion. Fortunately, a “smart” algorithm built into a trading platform flagged this trade and canceled the order.

In the medical field, according to one study, a staggering 400,000 patients die each year due to errors. Often these mistakes are related to transcription errors or a lack of interoperability between medical devices. Erroneously administering a 100-cc dose of medicine instead of the prescribed 10-cc could create a serious complication or, worse, become fatal. In a discipline that deals in life or death, it’s unfortunate that there are automated systems that create precise solutions, only to rely on a human—potentially under stress and fatigue—to accurately input that information into a dissimilar system.

In aviation, the story is much the same. According to FAA studies, fat-fingered or other data input errors into a flight management systems (FMS) or takeoff performance systems (TPS) result in a takeoff-performance related incident at a rate of one in 100,000 flights. These events are rarely deadly—only two fatal accidents in the past 50 years—but almost always reduce regulatory safety margins, might result in major aircraft damage, and have the potential to be catastrophic. What’s most alarming is that aviation is a system built on checking and double-checking, verifying and cross-verifying; yet these events continue to occur.

The most recent example of a takeoff performance event occurred just weeks ago. On February 1, a MyCargo Airlines Boeing 747-400 freighter operating on behalf of Saudia Airlines struck its tail on the runway during takeoff from Dammam, Saudi Arabia. The aircraft became airborne, stopped its climb initially at 7,000 feet, later continued to climb to 10,000 feet, and ultimately diverted to Jeddah, Saudi Arabia.

Photos of this aircraft after the event indicated major structural damage to the empennage and rear fuselage. While the investigation is ongoing, a database of similar events suggests erroneous takeoff data based on a calculated takeoff weight that was much lower than the actual takeoff weight. These miscalculations often result in erroneous takeoff V speeds (V1, Vr, and V2) and/or lower-than-required takeoff thrust settings.

Netherland’s National Aerospace Center (NLR), KLM, and Martinair recently completed a takeoff performance incident study. The report, published in September, analyzed 49 takeoff performance incidents and accidents from 1998 to 2018. Only those events investigated by state accident investigation authorities (AIA) were included in the study. The authors of the study concluded that these events are a serious threat to flight safety and are often under-reported; the number of actual events could easily be increased by a factor of two.

The scope of this project focused on performance-related takeoff events that resulted from flight crew data input errors. Often, these errors resulted in an attempted takeoff at speeds lower than required (V1, Vr, and V2), a takeoff thrust setting lower than required, or a takeoff from an intersection with the wrong thrust setting. The study also investigated the feasibility of onboard “real-time” aircraft takeoff performance monitoring systems and various flight-data monitoring algorithms to identify these events, post-flight, after flight data is processed.

Of interest, the most common error identified in this study was an attempted takeoff with takeoff speeds that were too slow for the actual takeoff weight. The greatest risks associated with these events are a tail strike or collision with obstacles. These “wrong calculated takeoff weight” errors—accounting for nearly 60 percent of all events in this study—were primarily the result of an entry error in the FMS or TPS.

Common factors included entering a value that was different from the actual value by a factor of 10 (example an error of 10 tons, 100 tons, or 100,000 pounds) or mistakenly using the zero-fuel weight or landing weight as the takeoff weight. In each of these events, the calculated takeoff weight was 12 to 42 percent lower than the actual takeoff weight.

Nearly one-third of the events in the study involved aircraft attempting to take off from the incorrect position on a runway. Common errors include using full-runway-length data and taking off from a position with less takeoff run available (TORA), such as an intersection.

Other related errors include entering data in the TPS for the wrong runway (Runway 17L versus 17R), the wrong runway intersection (Runway 17/AA versus 17/BB) or a runway transposition error (Runway 01L versus 10L). In each of these cases, if the entry error results in the use of data for a runway with a longer TORA, then there is a risk of a runway excursion/overrun or collision with obstacles.

Entering the wrong thrust setting in the FMS or thrust management computer (TMC) was another error identified in the study. These errors resulted in a thrust setting that was too low for the conditions. The risks associated with these events are a runway excursion or collision with obstacles. Often these errors are caused by a communication error between the pilots when selecting assumed or derated takeoff thrust settings.

It’s easy to see how these events occur in aviation. For each successful takeoff, a tightly orchestrated chorus of people (dispatchers and other planners) and systems (weight-and-balance software and takeoff-performance tools) must funnel information—either electronically or paper-based—to the pilots to further input into aircraft systems, all just minutes from departure. The opportunities to make a mistake are huge. Most systems are not connected, the margin for error is slim, and the responsibility rests on the pilots to “get it right.”

To support pilots, operators must develop sound procedures to prevent and trap errors such as transposing numbers or selecting the wrong runway in the TPS, omitting a zero in the FMS, or selecting the wrong temperature in the TMC. Aircraft manufacturers must also design aircraft systems to eliminate a scenario where a fat-fingered data entry error might result in a catastrophic accident.

Pilot, safety expert, consultant and aviation journalist - Kipp Lau writes about flight safety and airmanship for AIN. He can be reached by email.