November 01, 2012
Finnair had a problem that costs it about €100 ($130) per minute and it needed a solution.
The culprit was the dual engine bleed air system on its Airbus A330s. The airline suffered two incidents because the system failed—each malfunction resulted in loss of aircraft pressurization and resultant emergency descents.
Finnair flies its eight A330s and A340s about 19 hr. per day “so even a one-hour delay can cause a mess for a week,” says Manu Skytta, assistant VP of component services for Finnair Technical Services. Over the past 12 months, the airline has logged about 20 hr. of delays with these aircraft due to the bleed system, so a solution definitely was needed to ensure that these airplanes would be able to produce the high-dispatch reliability rates they are noted for.
Skytta wondered if the airline could use the parameters from the bleed-monitoring computer to forecast bleed system faults and component removals. It had the data but needed to figure out how to detect anomalies in aircraft subsystems, so it could “predict as many critical events as early as possible,” without generating false alarms, ideally with a signal-by-signal approach, he says.
Enter Frankfurt Consulting Engineers (FCE), a company that develops mathematical algorithms to improve industrial production. One of its tools is Hazard Predictor—a data-driven condition monitoring system that uses algorithms to detect deteriorations.
Seeing its problem growing worse, Finnair decided to put its faith in math because something needed to be done. “The longer a critical event lasts, the more critical it is,” says Skytta.
Finnair started sending raw aircraft signal data via a file transfer protocol server and asked FCE to figure out what was wrong with the aircraft’s system, even though the company has no aircraft background. Just to emphasize this point: FCE’s employees are highly intelligent, but they are not aircraft engineers or maintainers.
FCE received the flight data weekly, input it into its condition-based monitoring tool, analyzed the data clusters and sent Finnair weekly reports detailing which aircraft parameters included anomalies that could become critical.