The tragic disappearance of Malaysia Airlines flight 370 just took an intriguing turn, as the Wall Street Journal reported this morning on U.S. investigators’ suspicion that the plane’s engines remained running for about four hours after last contact with the flight. The source of this revelation? The plane’s engines themselves, which return monitoring data to the manufacturer while in operation. The collection and use of monitoring data on MH370 presents an interesting case study in big data, internet-of-things and industrial internet technologies in action, and also prompts interesting questions about the flight.
The Boeing 777 plane carrying flight 370 was powered by twin Trent 800 engines manufactured by London-based Rolls-Royce. Each engine can be fitted with up to about 25 different sensors, each of these capable of monitoring one or more key engine operating characteristics such as temperatures, pressures, speeds, flows and vibration levels at various places in the engine. Data from the aircraft may also be captured to provide context for monitoring systems–including altitude, speed, air temperature, cabin air quantity, and electrical power.
The data captured by the engine’s sensors is collected by the plane’s Airplane Condition Monitoring System (ACMS) which, in the case of the Boeing 777, is essentially a hardened server provided by a company like Honeywell or Teledyne called the Airline Information Monitoring System (AIMS). The ACMS/AIMS generates snapshots of the airplane’s state on takeoff, during climb, and periodically while the aircraft is cruising.
Another system, the Aircraft Communications Addressing and Reporting System (ACARS) is used to transfer data from the ACMS to the ground via VHF radio or a satellite link while the plane is in-flight. According to reports, the snapshots are transmitted every 30 minutes during cruise. Transmitted data is sent to a Rolls-Royce facility in Derby, U.K. for storage and analysis.
While each snapshot is limited to 3 KB in size, this can add up quickly, with enough engines in the air. In the case of Rolls-Royce, a Trent 800 powered Boeing 777 takes off or lands every 96 seconds. With over 13,000 engines in service across a variety of commercial and military planes, a Rolls-Royce powered aircraft takes off or lands every 2.5 seconds, equating to over 5.5 million flights per year, covering 12 billion flight-miles. As a result, in 2012 the company collected approximately 9 GB per day in half a billion data reports.
All of this data is analyzed as it flows in, with software automatically checking for validity and correcting for errors. Trends are extrapolated from individual snapshots and compared to expected data, identifying anomalies or subtle changes from one flight to the next. The company reports using neural network-based algorithms to perform this analysis, ‘fusing’ information from multiple sensors to provide a highly sensitive detection capability.
Questions a Data Scientist Might Ask About MH370
Enterprises value big data because it allows us to ask, and hopefully answer, new questions. Here are a few of the questions a data scientist might explore using data transmitted from the flight’s engines:
- What is the correlation between the altitude and the speed. Does this look like an aircraft that crashed or one that landed?
- What does air temperature at a given altitude tell us about the path the place was taking?
- A look at the cabin pressure over time would allow us to accept or reject the theory that a loss of cabin pressure took place.
No doubt Rolls-Royce engineers have been heads-down in their analysis of this data since the disappearance of the flight. Hopefully they reveal some of what they find.
UPDATE: 3/13/14 – 5:26 CDT. When I originally posted this this morning, I noted that just as I hit post CNN began reporting that officials are denying that the planes engines kept running. As of the time of this update, they are now reporting that the engines were in fact pinging Rolls-Royce. CNN’s coverage of the story is here.